A typology of the characteristics of teachers’ written feedback comments on second language writing

Abstract Written feedback commentary (WFC) on L2 student writing is a widespread and time-intensive teacher practice, serving a range of roles and purposes. One of the challenges in providing effective WFC is attending to the many content and delivery options that are possible, some of which have been shown to exert tangible effects on students and their texts. This article presents a typology of characteristics of teachers’ written comments, synthesised from 30 years of empirical research. Ten strategies for providing WFC (focus or target, mode and tone, syntactic structure, text specificity, location, explicitness, length, presence of mitigation strategies, pen-and-paper versus computer-mediated delivery, and temporality) are outlined. Thereafter, the paper presents the available options relating to student response to written feedback commentary. Each characteristic is illustrated and research into its effectiveness reviewed.


ABOUT THE AUTHOR William S. Pearson is a Postdoctoral Research
Fellow with the Graduate School of Education at the University of Exeter. His research activities address candidate preparation for the IELTS (International English Language Testing System) test, teacher written feedback on L2 writing, and pre-sessional English for academic purposes preparation programmes. His recent doctoral study explored how four student writers preparing for IELTS engaged with content-and formfocused written feedback on essays undertaken as simulated practice for the test. The present study constitutes an attempt to generate a novel typology that characterises the complex, negotiated written feedback content and delivery decisions that arose during the project. The typology can be applied to other learning-towrite settings involving the provision of teacher feedback commentary.

PUBLIC INTEREST STATEMENT
Feedback commentary provided on essays written by students for whom English is a second language is a widespread and time-intensive teacher practice. One of the challenges in generating effective feedback comments is addressing the various content and delivery options that are possible, some of which exert a tangible effect on students and their texts. This article presents a typology of 10 strategies for providing written feedback comments, drawing on 30 years of classroom-based research. These include, 1) the selection of textual feature (e.g., grammar) that occupies the focus of the comment, 2) a comment's mode and tone (e.g., advisory, positive), 3) syntactic structure (e.g., declarative), 4) specificity, 5) location within the text, 6) whether an explicit revision strategy is provided, 7) its length (in words), 8) the presence of "mitigation" to soften the blow, 9) pen-and-paper versus computer-mediated delivery, and 10) whether comments are provided while students write or are delayed. The paper then presents the available options relating to how students respond to commentary.

Introduction
Teacher written feedback on L2 student writing has attracted much research attention in recent years (Ellis, 2009;Liu & Brown, 2015;Pearson, 2020). While many empirical studies have examined the learning potential of written corrective feedback (WCF; Bitchener, 2008;Brown, 2012;Han & Hyland, 2015), a growing body of literature addresses the role of written feedback commentary (WFC) as a discrete approach to feedback on L2 writing (Christiansen & Bloch, 2016;Conrad & Goldstein, 1999;Ferris, 1997;Grouling, 2018;Lee & Schallert, 2008;Sugita, 2006) or in conjunction with WCF (Nurmukhamedov & Kim, 2009;Saito, 1994;Tang & Liu, 2018). Teacher commentary is a widespread and important feature of response to L2 writing. It provides a sense of audience, allowing the reader to address the writer through written prose that explicates pre-eminent features of a text. The ability to pass judgment through WFC establishes the all-important right of the teacher to evaluate a student's text (Hyland & Hyland, 2001;Smith, 1997). Concomitantly, the choice of textual features and how points are conveyed contribute to the interpersonal relationship between the teacher and student as well as determining the complicated, dynamic, and social nature of written feedback (WF) processing and response (Hyland & Hyland, 2019b;Lee & Schallert, 2008).
Commentary fulfils a range of (possibly conflicting) purposes (Leki, 1990), varying from analysing, responding to, engaging with, and improving a student's text (Grabe & Kaplan, 1996;Purves, 1984;Smith, 1997). In order to realise feedback's informational, pedagogical, and interpersonal goals, the teacher-as-reader makes a complex range of deliberate and subconscious language and stylistic decisions (referred to as the teacher's stance) in the design of WFC's content and delivery (Hyland & Hyland, 2019a). Among other qualities, these encompass the focus, mode, tone, syntactic structure, specificity, and length of written comments (Conrad & Goldstein, 1999;Ferris, 1997;Grouling, 2018;Nurmukhamedov & Kim, 2009). Research has shown the linguistic form of the teacher's WFC stance influences how students revise (Conrad & Goldstein, 1999;Ferris, 1997;Ferris et al., 1997;Hyland & Hyland, 2001;Nurmukhamedov & Kim, 2009;Sugita, 2006), with repercussions for "good practice" feedback guidelines. How research operationalises student response to WFC is complex and less certain than with WCF, where there is often a clear "correct answer". Consideration of the ways in which student response to written commentary can be evaluated may feed-forward into more informed practitioner approaches, possibly untapping the often-unfulfilled learning potential of written feedback (Zhang & Hyland, 2018).
Current research into the outcomes of written feedback commentary appears inconsistent, stemming from the complexity and variation across student (e.g., proficiency level, beliefs, motivation, feedback literacy), and contextual (language learning settings, written tasks, presence of pair work) variables (Ellis, 2010;Goldstein, 2004Goldstein, , 2006. As a way forward, there may be value in systematically identifying the various options available to teachers for conveying written feedback commentary on L2 student writing, either as a basis for pedagogical decision-making or for the design of future research, which has yet to comprehensively consider the characteristics of WFC (Ferris & Hedgcock, 2005). This paper presents a typology of written feedback commentary characteristics targeted at researchers and practitioners in second language writing settings. While not a systematic review of the literature, the study draws on three decades' worth of descriptive, experimental, and interpretive research to yield 10 discrete comment characteristics. It begins with a tabular overview of the options available, before illustrating and discussing the characteristics. The article is written as a companion to Ellis' (2009) A typology of written corrective feedback types. Like Ellis (2009), the affordances of the various approaches, conceived of as their effects on the extent or success of student revisions and students' attitudinal responses, are briefly discussed. Since practitioners should be mindful that no approach to the content and delivery of WFC should be unilaterally imposed on learners (Leki, 1991) and that studies into the effects of the formal characteristics of WFC do not always take into consideration the type of revision being requested (Conrad & Goldstein, 1999), the article refrains from categorically advocating particular strategies. Table 1 presents the typology of options for providing written feedback commentary on L2 student writing. The framework synthesises the many potential characteristics of teachers' WFC refined over the last thirty years, often through inductive analysis (Ferris, 1997). It is applicable to classroom contexts where teacher commentary is provided on content issues, perhaps preceding WCF (Ashwell, 2000) or accompanying it in the form of brief metalinguistic explanation (Tang & Liu, 2018). Section A contains the 10 delineated considerations for content and delivery strategies, while section B addresses ways of investigating student response. The options available to practitioners, synthesised from the literature, are outlined in the "Commentary characteristics" column, with definitions of the characteristics and research investigating the options outlined in additional columns. Aesthetic considerations such as the colour of the ink and legibility of the teacher's handwriting are not included.

Focus or target
Focus denotes individual or multiple textual features of learner writing addressed through written commentary (perhaps incorporating the order in which they are dealt with). Institutional guidelines may recommend or dictate teachers focus on particular features (Lee, 2008;Montgomery & Baker, 2007), reflecting current understandings of good practice, such as providing balanced coverage in WF (Hyland & Hyland, 2006;Lee, 2008) or addressing content and organisation before or separately to form (Ferris, 1997;Hyland & Hyland, 2006;Sommers, 1982). However, in many contexts, teachers are free to select focal areas, promoting a conception of WFC as a flexible response that fits the student and task (Ferris, 2014). Research has shown teachers' choices of textual features for WFC are motivated by wide-ranging beliefs about language learning and "good" writing (Ferris, 2014;Junqueira & Payant, 2015), although studies reveal inconsistencies in what teachers believe they are focusing on and textual realities (Lee, 2009;Montgomery & Baker, 2007), perhaps due to practical constraints that impinge on response (Junqueira & Payant, 2015) or difficulties implementing principles acquired from teacher education (I. Lee et al., 2016). As such, there is merit in teachers critically reflecting on their beliefs and whether/how they align with practice (Ferris, 2007;J. J. Lee et al., 2018;I. Lee et al., 2016) and undertaking classroombased action research involving WF innovation (Lee, 2008).
Research has shown many L2 learners value comprehensive, detailed written commentary (Elwood & Bode, 2014;Ferris, 1995;Hedgcock & Lefkowitz, 1994). Nevertheless, selecting salient textual features that warrant a response from the myriad of options available poses challenges for WF providers, who should be mindful not to cognitively overload learners (Bitchener & Knoch, 2008;Sheen, 2007). In contrast to the WCF literature, there are fewer categorisations of WFC focal areas (e.g., Christiansen & Bloch, 2016;Ene & Upton, 2014;Grouling, 2018). One important reason is that there is generally less agreement on categories of potentially problematic textual features in comparison to lexicogrammatical error types (which are incorporated into WFC coding schemes). Most widespread is the long-standing distinction between content/meaning-and formfocused textual issues (Ashwell, 2000;Nurmukhamedov & Kim, 2009;Yu et al., 2018), though is of limited value due to the very general categories of analysis . A scheme delineating specific focal areas into global (overall remarks), discourse (content and organisation), and form (lexis, grammar, and mechanics; Ene & Upton, 2014; J. J. Lee et al., 2018), summarised in Table 2, captures in more detail the goals of teacher response, though has yet to be widely adopted in empirical studies. This refers to the range of textual features the teacher focuses on, which can address the quality of the text overall (general), content and organisation (discourse), and/or lexis, grammar, and mechanics (form). As with WCF, teachers may focus intensively on a particular feature (focused) or two to five (mid-focused). Alternatively, their interest may lie in an extensive (or unfocused) array of issues.
2 Mode and tone: a advisory b correction c criticism d description e give information f "need to" g praise h question posing i reader reflection The mode (or semantic/ pragmatic function) and accompanying tone (i.e., positive, negative, neutral) of teacher commentary varies widely, although research has identified several modes that persist across studies, albeit labelled using varying terms (e.g., "descriptive" comments are also referred to as "neutral" and "characterise the text").
3 Syntactic structure: a declarative b imperative c exclamative d interrogative The syntactic structure of comments vary from declarative, imperative (and exclamative), and interrogative, depending on the teacher's intent.
4 Text specificity: a text-specific b generic Text-specific comments are either explicitly or implicitly tied to the particular text being read. Generic comments are applicable to any text (Ferris, 1997). Conrad and Goldstein (1999), Ferris (1997) and Ferris et al. (1997) examined the text specificity of teacher WFC.

Location: a marginal b interlinear c end comment
Comments may be written in the margins, between lines of text, or as an end comment beneath the text. The former are often targeted, while the latter, global.
6 Explicitness: a explicit WFC b implicit WFC c revisions not required In learning contexts requiring revisions, commentary may explicitly spell out the required revisions through a directive, correction, statement of a textual problem, or evaluation. In other instances, WFC may indicate something is wrong or elicit clarification without providing a revision strategy (implicit). Not all L2 learning contexts require students to revise their texts, while many comments (e.g., praise) do not solicit revisions.
A few studies have considered the role of explicit strategy provision in WFC (Conrad & Goldstein, 1999;Ene & Upton, 2014; J. J. Lee et al., 2018), although the characteristic overlaps with comment mode (i.e., advisory).
Descriptive studies Grouling, 2018) compared the lengths of teacher comments across modes, whereas Ferris (1997) examined the influence of comment length on textual revisions.

Presence of mitigation strategies: a use of mitigation b no mitigation
It is natural for teachers to want to alleviate the terseness of commentary through mitigation strategies, which come in four forms (Hyland & Hyland, 2001): Several studies have examined student perceptions (Hyland, 2019;Treglia, 2008) or the revision effects of mitigation strategies (Conrad & Goldstein, 1999;Ferris, 1997;Nurmukhamedov & Kim, 2009;Sugita, 2006). Adapted from Ene and Upton (2014).
It may be helpful for teachers to consider the range of focal areas of WFC in terms of their focusedness, a concept borrowed from WCF research (Bitchener & Knoch, 2009;Liu & Brown, 2015;Sheen, 2007). A focused approach denotes WFC that selectively addresses one discrete textual area, while unfocused commentary involves highlighting a diverse range of issues. Directing students' attentions towards a single issue suits scenarios where learners need highly specific WF on a given area, perhaps a short task immediately putting into practice teacher input. For practitioners faced with the pressure of multiple classes with large class sizes, such an approach may reduce the hastiness of WFC generation, which can result in errors or illegibility being introduced (Lee, 2019). Most importantly, a focused approach reduces the cognitive load on learners (Bitchener & Knoch, 2008;Sheen, 2007), which, as indicated in Table 3, could be notable during WF processing if a range of textual issues are dealt with (especially if in conjunction with WCF). However, focused WFC may not be feasible in certain writing settings (e.g., test preparation), could contrast with student preferences for comprehensive WF, or might hinder the rapid progress of gifted learners (Bitchener & Knoch, 2009). A mid-focused approach (Liu & Brown, 2015; on two to five target areas) may offer a suitable compromise that draws on the affordances of both. Research comparing the discrete effects of the focusedness of teacher WFC has yet to be undertaken.

Mode and tone
Perhaps the most widely investigated characteristic of WFC is its mode, which can be defined as the semantic or pragmatic intention(s) of the feedback provider coded in commentary. Coding schemes that delineate the mode of comments vary from Hyland and Hyland's (2001) simple three core functions (praise, criticism, and suggestion) to more complex schemes that incorporate up to ten

Commentary characteristics
Description Studies 9 Medium of delivery: a pen and paper b computer-mediated feedback (CMF) Increasingly, commentary on student writing is provided through typed comments or ones handwritten onto an e-document using a stylus.
A number of studies have explored the impact of CMF in the form of typed comments (Ene & Upton, 2014;Uscinski, 2017) or handwritten e-commentary (Grouling, 2018). Very few studies have compared the effects of handwritten and computer-mediated teacher WFC (Yeh & Lo, 2009 Teachers may wish to consider the extent students act on WF comments (e.g., substantially, minimally, or not at all). They are also likely to consider how successful students were in their revisions (e.g., positive, negative, mixed effects).
Assessing the extent students revise their texts in response to teacher commentary and the outcomes of such revisions have been widely researched (e.g., Christiansen & Bloch, 2016;Ferris, 1997), sometimes conceived of as behavioural engagement (Zhang & Hyland, 2018;Zheng et al., 2020).

No revisions required
In certain situations (e.g., product writing, preparation for high-stakes English assessments), students may have their texts returned with a judgment of written performance and no accompanying expectation of revisions.
Relatively little research has queried students' perspectives towards WFC (Elwood & Bode, 2014) or post-feedback behaviours (Zheng et al., 2020) when there is no expectation of revision.
Research into the effects of specific modes indicate comments suggesting or requesting changes to student writing add the most value (measured in terms of student revisions and attitudinal responses; Conrad & Goldstein, 1999;Ferris, 1997;Treglia, 2008), perhaps because they fulfil feedback's core function of closing performance gaps through feeding forward usable information (Price et al., 2010;Walker, 2009). However, such findings should be interpreted cautiously. Clearly, a comment's mode may be interpreted differently by a learner (Hyland, 2019), while teachers may not be conscious of the modal choices they are making. The prospect of mode switching within or across student texts could cause some difficulties for learners to notice the teacher's underlying intent (Leki, 1990), suggesting the need for a consistent approach. Furthermore, there is a fine line between being guided and told what to do (Treglia, 2008), suggesting teachers carefully consider how they integrate modality with their syntactic choices.
Praise is valued by some learners for enhancing motivation and self-esteem (Connors & Lunsford, 1993;Ferris, 1995;Straub, 1997), but can be considered by others as unnecessary (Hyland & Hyland, 2001) or even damaging if it is perceived as insincere (Smith, 1997;Treglia, 2008). Teachers should also bear in mind that praise does not tend to induce revisions (Ferris, 1997) and that many learners seek comments that are valid and appropriate to their subject, point of view, and purpose (Straub, 1997), rather than praise. A similar picture presents with criticism. Some studies urge teachers to be cautious of criticism (Connors & Lunsford, 1993;Hyland, 1998a), since it can undermine a learner's confidence, motivation, and self-concept (Hedgcock & Lefkowitz, 1994). On the other hand, most L2 learners are aware that there are shortcomings in their language and/or writing skills, so expect some form of negative response (Ferris, 1995(Ferris, , 2003Hedgcock & Lefkowitz, 1994;Saito, 1994). What is certain is that teachers should take care to express criticisms constructively, avoiding an exasperated or terse tone. There is likely value in querying students' expectations and preferences for praise/criticism and to always be sensitive towards the potential negative affective responses critical WF can induce.

Syntactic structure
A WFC characteristic that has shown to make a difference in how learners respond is the syntactic forms teachers consciously or sub-consciously employ. Comments can be written in the declarative ("It's very confusing"), imperative ("Explain it more clearly"), exclamative ("Nice start to your essay!"), or interrogative ("What does 'it' mean?"). When considered discretely as a characteristic of WFC, research has shown students tend to engage more substantively with comments phrased imperatively (Conrad & Goldstein, 1999;Nurmukhamedov & Kim, 2009;Sugita, 2006), probably because they are interpreted as an explicit requirement to do something (Sugita, 2006). Yet teachers should carefully consider their learners before incorporating such forms into a broader WFC strategy, as they could be interpreted by more autonomous students as curt or controlling (Straub, 1997). It is also apparent that comments phrased as questions generally lead to less substantial or successful textual changes (Conrad & Goldstein, 1999;Ferris, 1997;Sugita, 2006), particularly in the case of open-ended WH questions (Conrad & Goldstein, 1999). Nevertheless, question forms may induce a profound effect on student thinking not reflected in their writing at the time of research.

Correction
There's a preposition missing after "knowledge".

Description
You have three separate points here.

Give information
In the UK, films are rated PG, 12, 15, and 18 depending on their content.

Praise
Great idea!

Question posing Do you think your audience would know what this is?
Reflection I remember reading about this on the news recently.
The impact of syntactic structure should be considered in conjunction with a comment's semantic/pragmatic function since structures fulfil multiple functions and vice-versa (Ferris, 1997). For instance, one study found declaratives that highlight the necessity of revision were more likely to be acted on than those that characterise the text (Conrad & Goldstein, 1999). Further complicating the issue is that teachers respond to students through WFC and not just their texts (Hyland & Hyland, 2019a). Consequently, teachers should consider their relationship with the learner and be responsive to local realities in the selection of the syntactic/functional forms of comments (Junqueira & Payant, 2015), i.e., students' backgrounds, needs, and preferences (Hyland, 1998a).

Text specificity
Text specificity in WFC concerns whether a comment is tied to a section of text (including the text as an entire object), or not (Conrad & Goldstein, 1999;Ferris, 1997). A text-specific comment could only have been written about the text in question (e.g., "You need to explain and illustrate how international tourism promotes intercultural understandings"), whereas a generic comment could apply to any text ("Keep up the good work!"). The distinction is not quite so clear cut, as clues like the location of the comment and the use of reference words attribute specificity, although these signals may be too subtle for some learners. There appears little pedagogical merit in generic comments. They are highly unlikely to spell out what a learner needs to do to enhance the quality of their writing, omitting feedback's key roles of feedforward and usability (Price et al., 2010;Walker, 2009). Similarly, facile, unspecific comments could unwittingly signal a lack of engagement from the teacher . Nevertheless, practitioners should be cautious with the degree of specificity of comments, since learners usually want and need to generalise messages from WFC to other assignments or tasks .

Location
A common distinction in studies of teacher WFC concerns the location of comments, and by implication, whether they are targeted towards specific issues or are more global. Marginal comments are, unsurprisingly, situated in the margins of a physical or annotatable electronic page or appear as text bubbles beyond the page in word processing applications, typically highlighting specific issues in the text body. In contrast, end comments tend to address the text from a global perspective (Ferris, 1997;Ferris et al., 1997). A comment's presence in the margins does not necessarily mean it is highly specific or form focused. Word processing software conveniently enables large sections of text to be highlighted and assigned a marginal comment. Unique to handwritten WFC are interlinear comments, often addressing form and situated (perhaps awkwardly) between the lines of text. Sommers (1982) suggests the existence of a tension between marginal and interlinear comments. Interlinear comments can convey the impression that revision is a matter of surface-level editing, while marginal comments may suggest more work is required to the content and/or organisation, for instance, requests for additional information (Ferris, 1997). Limited research has compared the effects of marginal versus end comments on L2 student revision. Ferris (1997) found marginal comments solicited 5% more substantive revisions that featured positive written outcomes than end comments and that marginal requests for information led to the most substantive and effective textual changes. Interestingly, end comments about grammar (accompanying underlined examples in the text) were also successful, although it was impossible to say whether this was due to the comments, the underlining, or a combination of both (Ferris, 1997).

Explicitness
When written commentary is provided under the expectation that students revise their texts, one possibility is for teachers to spell out how they anticipate the student could or should undertake the revisions (Conrad & Goldstein, 1999;Ene & Upton, 2014). As one might expect, there are a variety of ways in which explicit WFC can be constructed (Table 4). Direct corrections or reformulations constitute the most explicit approach, often being interpreted as commands to incorporate the suggested form even if they are expressed tentatively, owing to the asynchronous power dynamic between teacher and student. Directives spell out what the teacher wants the student to do, but the question of how this is achieved is left to the student. Revision strategies can also be provided through an example, although whether the teacher expects the student to merely incorporate the given suggestion or generate a parallel response of their own may remain unsaid. Implicit WFC often associates with critical or even terse comments, outlining or explaining a textual problem but providing no strategy pointing the way forward. Yet implicit WFC does not solely associate with critical or perfunctory WF from disengaged teachers, and instead may reflect efforts at engaging learners through questioning (Hyland & Hyland, 2001), encouraging them to consider how their textual choices impact on the reader (Hyland, 2011), and to develop their reflective capabilities and autonomy (Hyland & Hyland, 2006).
From a theoretical perspective, the provision of explicit revision strategies, providing they encompass high-quality, actionable feed-forward information (Price et al., 2010), ought to promote greater learner reflection and capacity for self-monitoring. This is borne out by Conrad and Goldstein (1999), who uncovered explicit WFC on paragraphing, purpose, coherence/cohesion, and examples resulted in 70% of revisions being deemed "successful" (in contrast to implicit WFC's 21%). In contrast, Ene and Upton (2014) found no significant difference in the uptake of explicit and implicit electronic written feedback. One reason could be that, as can be seen from Table 4, generating informative revision strategies requires lengthy comments that can cause students considerable cognitive difficulties during processing. Alternatively, the learners may have resisted incorporating the suggested revision strategies because they interpreted them as overbearing textual appropriations (Sommers, 1982). On the other hand, instructors should be mindful that commentary lacking explicitness has been noted as discouraging by some learners, diminishing their capacities to envisage an appropriate response (Treglia, 2008).

Length
In certain contexts, minimum or maximum limitations may be placed on the length of teacher commentary (Goldstein, 2004). However, more often than not, practitioners decide how long they wish comments to be, notwithstanding the commitment required to generate extensive written commentary and practical constraints (Goldstein, 2004;Kang & Dykema, 2017;Lee, 2019). Ferris (1997) investigated the impact of comment length as a categorical variable, dividing comments Explanation without correction This sounds like a memorised expression that could be used in any essay.

Indicates something is wrong but does not provide explanation
This is certainly clearer, but still comes across as very general.
into four categories: short (1-5 words), average (6-15), long (16-25), and very long (26+). The author found comments of all lengths appeared to influence positive changes, although revisions improved as comments became longer. This could be because longer comments explain more fully how the student should respond (Conrad & Goldstein, 1999). Alternatively, the students may have treated the textual issues associated with longer comments more seriously, perhaps because they considered detailed, explanatory WFC important to learning or good value for money (Elwood & Bode, 2014). A notable caveat to these findings is the uncertainty in delineating when one comment ends and another begins, since teacher comments seldom break down neatly into single phrases, sentences, or idea units . Many studies appear to follow Conrad and Goldstein's (1999) operationalisation of a written comment as "a stretch of discourse having a unified intended function" (p. 153), although this is not always articulated.

Presence of mitigation strategies
Written commentary should contribute to creating a supportive learning environment, centred around a positive interpersonal relationship between teacher and student, fostering the student writer's self-confidence, motivation, and autonomy (Hyland & Hyland, 2001;Hyland & Hyland, 2006). As such, teachers should be mindful that written commentary can pack a significant emotional punch (Han & Hyland, 2019), accentuated by the terseness and finality of evaluative information conveyed unidirectionally from teacher to student (Carless et al., 2011). A very common approach to soften the impact of WFC is through the use of mitigation, delineated by Hyland and Hyland (2001;2019b) as encompassing four discrete strategies (summarised in Table 5).
Hedges denote a diverse variety of structures that indicate the feedback provider's uncertainty, politeness, or indirectness. Ferris (1997) differentiates between lexical (e.g., seems, perhaps) and syntactic approaches (can, might), although Hinkel (2004) breaks them down into more precise categories based on form (see , Table 5). Paired act patterns create a more balanced comment through affixing praise to criticism. A third strategy, personal attribution, signals that a comment reflects the feedback provider's personal opinion. Finally, use of an interrogative syntax weakens the force of a statement by making it relative to a writer's state of knowledge (Hyland & Hyland, 2001). From a theoretical perspective, mitigation reduces the power differential in the teacherstudent relationship and contributes to building a positive, sympathetic teacher-student relationship (Ferris, 1997;Hyland & Hyland, 2001;Treglia, 2008).
Since they are culturally bound and not always visible to L2 learners, hedging strategies in written commentary have been posited as problematic for L2 learners Hinkel, 2004;Hyland & Hyland, 2001;Nurmukhamedov & Kim, 2009). However, a few studies suggest that mitigation in WFC has little (positive or negative) effect on textual revisions (Conrad & Goldstein, 1999;Ferris, 1997), perhaps because of the papers' narrow focus on hedging devices. One study that did uncover an impact was Nurmukhamedov and Kim (2009). Their analysis revealed 87.5% of Paired act patterns (positive softeners) You've raised some good points, but . . .

Interrogative syntax
Is there a more natural expression? Negative effects student revisions deemed effective stemmed from comments that were hedged. The authors caution that positive textual outcomes may have arisen due to hedging being associated with specific comments that gave concrete revision suggestions (generating further evidence of the importance of feed forward). It is also likely the learners' CEFR B2 language proficiency enabled them to better detect the teacher's pragmatic intent. Practitioners are cautioned to ensure feedback messages are as clear as possible (Hyland & Hyland, 2001), to recognise that hedging may frustrate some learners because it indicates a lack of commitment to the truth value of a proposition (Hyland, 1998b), and that such devices may elevate the cognitive burden of processing WF (Nurmukhamedov & Kim, 2009). However, they are also reminded that many developing writers are positively disposed towards the face-saving role of hedges (Treglia, 2008), while failure to mitigate criticism can seriously affect some individuals' motivation to write (Hyland & Hyland, 2019b).

Medium of delivery
As language learning and teaching continues to digitalise, more and more teachers, either at the behest of their language teaching organisation or through choice, are adopting computermediated feedback approaches to written commentary. Traditional pen-and-paper annotations and comments are increasingly being applied digitally, often through track changes and comments in Microsoft Word or competitor applications (Ene & Upton, 2014;Uscinski, 2017), synchronous document editing via an app such as Google Docs (Ene & Upton, 2018), handwritten comments on a tablet using a stylus (Grouling, 2018), and bespoke applications (e.g., learning management systems; Yeh & Lo, 2009). The limited comparative research into teacher CMF commentary versus pen-and-paper provision (Yeh & Lo, 2009) indicates it offers affordances for developing writers to better identify written errors. This may be because of CMF's greater legibility (Chong, 2019), the possibility of multimodality (Elola & Oskoz, 2017), convenience in storage and sharing (Yeh & Lo, 2009), and greater ability to monitor L2 or writing skills development (e.g., through submission to Automated Writing Evaluation applications; Zhang & Hyland, 2018). However, the extent students respond and the outcomes of their revisions depend on the way comments are written (Goldstein, 2004), how the teacher fosters response expectations (Uscinski, 2017), and on developing writers' digital literacy (Elola & Oskoz, 2017).

Temporality
In the vast majority of learning contexts, WFC is provided asynchronously, that is, generated temporally (and usually spatially) separate from when learners write. Thus, asynchronous WF is often retrospective, with student queries or response being delayed (or not required). In certain situations, the teacher may be able to pre-empt issues through anticipatory WFC, although if on a wider scale, such issues may be more amenable to whole group feedback. With the emergence of new technologies, synchronous WF provision is increasingly an option for teachers. This involves the teacher generating commentary in real-time, as the learner writes (Shintani, 2015), thereby providing an immediate reader presence. Synchronous WFC provision may take the form of a student-teacher tutorial or less common learning situations when the teacher is present in a computer laboratory with her/his learners (Chong, 2019), providing instantaneous feedback as students compose (Shintani & Aubrey, 2016). Research in WCF settings has shown synchronous WF delivery promotes a more collaborative writing process between teacher and learner (Aubrey, 2014), enables the expeditious resolution of feedback misunderstandings (depending on the teacher's acquiescence to requests; Chong, 2019), and facilitates a focus on higher order concerns (when used in conjunction with asynchronous WF; Ene & Upton, 2018). Such affordances may not translate into WFC settings, since the time needed to reflect on complex or substantial textual issues and write up detailed comments may be unfeasible synchronously, especially with large classes. Furthermore, one might imagine synchronous WCF's tendency to overload learners (Shintani, 2015) or incorporate inefficiency into the revision of texts (Ene & Upton, 2018) would be magnified in WFC provision.

Revisions required
In many L2 learning settings, there is an expectation students respond to teacher commentary, often in the form of revisions to an initial draft. Typically, researchers are interested in classifying students' responses to individual actionable WF points through deductive analysis of their revision operations (e.g., Christiansen & Bloch, 2016;Sugita, 2006). Revision operations in response to WFC differ to WCF in that evaluating student response is more subjective and complex, reflected in the diversity of coding schemes (a sample of approaches is provided in Table 6). The more parsimonious schemes encompass the sole judgment of whether the learner incorporated the suggested change(s) or not (Tian & Zhou, 2020;Yu et al., 2018), suitable if the researcher or teacher's interest lies in the student's ability or willingness to engage with the WF (see, Han & Hyland, 2015;Zhang & Hyland, 2018). More sophisticated approaches (Ferris, 1997;Sugita, 2006) distinguish between substantive and minimal textual changes, with the implication that substantial responses indicate heightened student engagement. Such a delineation may not be appropriate for comments that address form issues, i.e., metalinguistic explanation (Tang & Liu, 2018). Instead, outcomes-orientated measures like target-like/non-target-like revision (Nurmukhamedov & Kim, 2009) may generate more nuanced insights. There is also value in operationalising revision operations as conforming to the teacher's instructions (or not; Christiansen & Bloch, 2016;Nurmukhamedov & Kim, 2009), although such an approach incorporates an implicit judgment of the success of revisions, given following the teacher's instruction is presumably desired.
Most studies that classify the success of student revisions use the broad categories, positive, negative, or mixed effects (Ferris, 1997;Nurmukhamedov & Kim, 2009;Sugita, 2006). How such coding takes place is not always explained or illustrated in research (see, Ferris, 1997), an important consideration for future studies into the impact of teacher written feedback commentary. Through triangulating revision operations with outcomes, analysts are able to able to construct illustrative, quantitative descriptions of student response (Ferris & Hedgcock, 2005). Regardless of the scheme employed, WFC revision operations should be interpreted cautiously. Substantive revisions, while indicative of active learner engagement, might reflect weaknesses in an initial draft. Similarly, disregarding comments can be an indication of disengagement (Han & Hyland, 2015) or growing writer autonomy and agency (Kang & Dykema, 2017). Additionally, deleting content could meet teacher expectations, but also signal a lack of effort or engagement.

No revisions required
In other settings, texts may be returned to students with the teacher's written feedback commentary without the requirement or expectation of revision (Elwood & Bode, 2014;Zheng et al., 2020). Frequently, such comments serve summative or assessment purposes, centred on the teacher's judgment of how well the writer accomplished the given task. The learner's role could merely involve reviewing or paying close attention to the information, which may or may not be of relevance or feed-forward to future pieces of writing. Instead, there may be interest in other aspects of student response, particularly students' attitudes towards the commentary (Cunningham, 2019;Treglia, 2008), emotions when processing the information (Mahfoodh, 2017;Tian & Zhou, 2020), understandings of (potentially hidden) messages (Hyland, 2013;Yu et al., 2018), and possible learning behaviours upon or after receipt of feedback (Cunningham, 2019;Zheng et al., 2020).

Conclusions
The present study reviewed the available options relating to the content and delivery of teacher written feedback commentary as well as approaches to measuring the effects of student responses through the deductive coding of revision operations. The options synthesised could be used to design new or refine existing written commentary guidelines for use in the evaluation of student writing. Additionally, the characteristics could be incorporated into training sessions to raise practitioner awareness of the options available (including their rationale and insights into their impact from research). As with Ellis (2009), the typology provides teachers with a basis for reflecting on current practices (Junqueira & Payant, 2015; I. Lee et al., 2016) and experimenting with the options in their own teaching (Lee, 2008). The study does not claim to be a systematic account of the literature on teacher commentary in L2 writing contexts nor an exhaustive scheme for conceptualising the characteristics of feedback comments. Length constraints prevented a detailed presentation of the findings of WFC studies, along with excluding research situated in L1 writing settings, where a number of insights into WFC techniques may transfer.
The typology provides a scheme for future experimental research to systematically examine the effects of discrete and combined WFC characteristics, recognised as exerting partial (and not always consistent) effects on language learning or writing outcomes across research (Ferris, 1997;Goldstein, 2006;Nurmukhamedov & Kim, 2009;Sugita, 2006)-in conjunction with contextual (Ellis, 2010) and individual learner factors (Conrad & Goldstein, 1999;Ellis, 2009). More research is needed into the underexplored impacts of commentary characteristics, particularly the focusedness of WFC, CMF versus traditional pen and paper delivery, and whether/how explicit revision strategy provision makes a difference. Existing revision operation coding schemes, while parsimonious, often helpfully separate the extent of a revision from its effect on textual quality, contributing interesting and illustrative quantitative insights into student response (Ferris & Hedgcock, 2005). Yet revision operations provide incomplete and potentially inaccurate knowledge, not to mention the limitations to claims that specific operations equate to writing development. Coding schemes need to be illustrated in research reports in order to improve the transparency of coding and to inform future studies.
No study, no matter how sound its design, can generate a recipe for best practice WFC provision. Clearly, practitioners and researchers need to take into account individual learners' beliefs, expectations, and preferences towards written feedback (Leki, 1991), which may be diverse, idiosyncratic, and potentially uninformed (Diab, 2005). As such, the typology could be further incorporated into interpretive research that conceives of feedback effectiveness from the perspective of the L2 learner. This could take the form of an initial survey, where students indicate their WFC preferences from a menu of options given (Leki, 1991;Saito, 1994), based on or modified from the typology, with follow-up interviews (perhaps post-WF provision) to probe for greater detail. Additionally, descriptive research incorporating all 10 characteristics of written feedback commentary is required to determine the extent to which the given categories reflect established practice (Ellis, 2009;Ene & Upton, 2014), and how commentary behaviours differ across practitioners, institutions, and L2 writing contexts. Descriptive research could also serve to refine the categories and/or create new ones.