Peer-feedback as a Translation Training Tool in Web-based Communication

Teaching web-based communication involves website analysis, website design and webcopy writing, often in multiple languages for different target audiences. Localization and foreign language webcopy writing require training in the area of translation, in addition to other skills related to web communication. Such training might be didactically supported with peer-feedback assignments. Surprisingly few studies in the area of localization/translation and web-based communication report about classroom practice of this kind. Drawing on theory from Educational Psychology, Writing Research in L2 and Translation Studies, this paper explores common notions in peer-feedback research. It discusses peer-feedback and its implementation as a translation training tool in the context of web-based communication with a focus on localization as a form of text production. Student feedback and the revisions performed on its basis are investigated. The data reported here derive from an undergraduate web-based communication course, where students produced a translation and an academic translation review. The paper discusses possible future peer-feedback practices, including suggestions for feedback scaffolding tailored to the needs of future language professionals in the area of web-based communication.


Introduction
Website localization involves adapting various aspects of a website to meet the linguistic, cultural and locale-specifi c requirements of the target market (Singh 2012: 124).Locale-specifi c requirements include times, dates, currency and number formats, addresses and phone numbers, colours, and local conventions.For professionals working within web-communication, language, cultural and locale-specifi c awareness are necessary to successfully communicate with the target market.Research has shown that web users are up to four times more likely to shop and purchase online, if the website is available in their native language (Parr/McManus 2001: 3).In addition to increased purchase intentions, website localization can improve website navigation, usability, and consumer attitudes towards a website (Baack/Singh 2007).Therefore, as a minimum, company websites must adapt to the languages of the web users (Yunker 2002: 10).
Website localization is a multifaceted process and introducing our students to this topic results in interdisciplinary discussions on the various elements that constitute website localization on the one hand and the overlap between website localization and web-based communication on the other.One aspect that is common to all websites (original and localized) is text, or webcopy.Consequently, as part of our teaching strategy, we place a strong focus on all aspects of writing for the web, including translation, search engine optimization (SEO) and international SEO.SEO is the process of designing, coding and writing content for a website that makes search engines rank it high (Barr 2010: 393).Once a site is optimized, search engines are more likely to fi nd and rank the website, leading to an increase in website visitors, and ultimately, paying customers (Singh 2012: 226).One very important step in the SEO process is the strategic placement of keywords.Keywords are the words and phrases that web users type into a search engine to fi nd the information they seek (Singh 2012: 228, Barr 2010: 397).When considering international SEO, it is often not the case that the keywords will simply be a direct translation of the English keywords.
Language professionals in the area of web-based communication are increasingly required to provide communication solutions in the form of multilingual webcopy.Even if they do not produce such material themselves, they must be able to judge and evaluate texts written for the multilingual web.Such evaluative judgements are considered to be a professional skill (Cowan 2010) which must be explicitly developed since it supports both critical thinking and refl ective capabilities.In essence, intended learning outcomes require the students to develop a skillset which allows them to produce high quality domain content, including multilingual texts, and to refl ect academically on their communicative choices.Findings from a previous study showed that using peer-feedback in tasks could be a "productive platform for the development of essential professional skills" (Nicol et al. 2014: 116).Before conducting our study, we also believed that peerfeedback could play an important role to help our students develop the required professional skills.Therefore, we applied a teaching approach which incorporates student peer-feedback on webcopy translation and an academic translation review.
We expected our students to write English-language webcopy (original and translations), and to refl ect on their work processes, where they describe the problem-solving processes during the task performance.This use of refl ective writing is commonly used as a training tool within Translation Studies (TS), so that the students not only translate a text but also refl ect on the process "in the form of annotations, of how and why decisions have been reached, based on which criteria" (Adab 2000: 221).When students refl ected on a translation task in our context, we termed this accompanying refl ective commentary an academic translation review.The same process within TS has been termed a descriptive commentary (Álvarez 2007) and translation commentary (Shei 2005).However, translation commentary within TS has previously been used to describe a statement made by a translator about parts of a literary translation carried out by himself/herself, which is not what we aimed to do in our study.We wanted to evaluate how well web-based communication students were able to juggle the challenges of web-based genre production, translation and their theoretical refl ection.Therefore, tasks in this course were designed to mirror professional practice.This supports the decision to include peer-feedback in the task, since peer-feedback requires students to make judgements about the quality of the work.Having an opportunity to develop such ability is deemed "fundamental for life beyond university, where graduates will be exposed to complex information and events in personal, professional and global contexts" (Nicol 2011: 1-2).
This paper reports on a small-scale study with 4 th and 6 th semester students of a content-based undergraduate course in International Business Communication.The study explored whether student peer-feedback is an effective teaching tool to improve communication students' translation skills as well as writing skills in an academic context and in a web-based communication context.This peer-feedback was introduced into the classroom practice using a minimalist approach to scaffolding (see Section 2.2).The data presented and discussed here were derived from a translation/localization task that was part of a larger study comprising three other written tasks within web genres.In addition to four writing tasks, the larger study included pre-and post-course online questionnaires, which we requested students to complete.The questionnaires asked for background information and to rate a number of statements about giving, receiving and implementing feedback.
Our web-based communication course requires the students to critically analyse existing websites, to make suggestions to improve the sites based on literature and theories covered during the course, to present their fi ndings, to design their own website and to present the fi nished product.To achieve credits, students must produce an end-of-semester written exam incorporating tasks covered during the semester.Therefore, given the emphasis placed on the written element of the course, we believed that students would benefi t from receiving extensive feedback on their written tasks.Since it is not often possible for the teacher to provide this extensive feedback, one way of increasing this feedback is to implement peer-feedback tasks.
We used this translation/localization task to investigate the following research questions (RQ): RQ 1.What types of feedback do students provide their peers?
RQ 2. What types of revision do students implement?RQ 3. Which fi ndings could contribute to how we implement peer-feedback as a translation training tool in the future?
The paper is organised in the following way: Section 2 explores peer-feedback as a teaching tool, with a focus on the merits of providing and not providing students with structured feedback criteria; Section 3 outlines a description of our study and the methods used to collect data, while Section 4 presents our fi ndings and interpretations, and, in Section 5, we refl ect on the results and possible avenues of future research.

Peer-feedback as a teaching tool
Peer-feedback is a popular means to aid students in text production.While it is commonly applied out of the necessity that large content classes make it diffi cult for teachers to provide adequate feedback on writing (Bok 2007), it has been shown to have positive effects on student learning and student writing (Topping/Ehly 1998, Yang et al. 2006, Nicol et al. 2014).Besides cognitive, affective, social and linguistic benefi ts, feedback is claimed to be a motivational instrument for future work (Hyland 2000), essential to the development and execution of self-regulatory skills (Topping 2009), and supports increased learner autonomy (Yang et al. 2006).In addition, it is claimed that feedback from a peer allows students to revise drafts more successfully (Cho/ Schunn 2007), learning from feedback improves the students' ability to evaluate their own writing (Lundstrom/Baker 2009, Nicol et al. 2014) and learner confi dence is increased by feedback tasks (Lin/Chien 2009, Mei/Yuan 2010).An enhanced sense of audience and ownership of text is also attributed to peer-feedback (Mangelsdorf/Schlumberger 1992, Tsui/Ng 2000).Furthermore, students who provide feedback are empowered (Coit 2004) and students providing feedback gain more from the task than students receiving feedback (Lundstrom/Baker 2009, Roscoe/Chi 2007).
The aforementioned studies are from the fi elds of Writing Research and Educational Psychology, but are applicable to several other areas of text production.One such area is Translation Studies, which forms a part of our web-based communication curriculum, namely through website localization.

Peer-feedback in Translation Studies
A recent article by Galán-Mañas/Hurtado Albir (2015) discusses assessment types in translation training.Among others, peer assessment is suggested as a means of formative assessment.Despite this, only a small number of studies on peer-feedback in Translation Studies (TS) exist.Wang/Han (2013: 63) mention studies on learner autonomy in interpreting and translation training.The focus of these studies is on oral interpreting training.Lindgren et al. (2009) present peerdiscussion in computer-based translation exercises.So far, Wang/Han (2013) and Lindgren et al. (2009) appear to be the benchmarks in the fi eld of peer-feedback in translation.Wang/Han (2013) see pedagogical methodologies from interpreting training to hold great promise for written translation training.In addition, they believe CALL technology to be a promising tool for providing and receiving peer-feedback which is unchained from face and politeness issues and allows for maximized feedback numbers in groups.We agree with Wang/Han (2014: 63) that "the adop-tion of peer-feedback in translation training is well motivated".Their reasons are L21 language training, adaptability of L2 learning from interpreting research, and emerging possibilities for L2 training with CALL means.Yet our reasons and motivation for incorporating feedback in our web-based communication course were different.We were unable to use interpreting-based feedback approaches, and instead took inspiration from previous peer-feedback research from Educational Psychology and the neighbouring fi eld of Writing Research.We believed that TS could draw on the current understanding of peer-feedback processes from these two fi elds.In doing so, a decision needed to be made regarding how the feedback task should be introduced in the classroom.A distinction that needed to be made by the teachers was whether students should follow instructed or non-instructed feedback tasks.A common instruction-led practice in the abovementioned fi elds is scaffolding and this is elaborated below.

Scaffolding vs. Non-scaffolding approach
Integrating peer-feedback in classroom practice is not an easy task.Students are often not familiar with either self-refl ective tasks or with giving and receiving feedback from a peer.This is perhaps surprising since they have received feedback from teachers throughout their education.Yet students can be reluctant -if not scared -to criticize an equal status learner.It may be that they do not want to seem overcritical or judgmental, they do not want to lose face, and they may prefer to receive praise and compliments rather than suggestions for improvement (Sampson et al. 1999, Nilson 2003).To help overcome this reluctance, it has been suggested that mitigating language could be used instead of pure praise (Nelson/Schunn 2009), since it includes criticism and praise.

Pro scaffolding
To address student issues with peer-feedback, feedback research since the 1980s has suggested and discussed a best practice approach, which is to introduce peer-feedback tasks with scaffolding.This provides students with pre-defi ned feedback criteria, assessment criteria or other forms of guidance or instruction.Nicol (2014: 205) terms this set of criteria a rubric which is provided by the teacher.It has been claimed that such scaffolding moves learners along in their thinking and their learning (Wood 1975).Writing research has drawn on this fi nding over the last decades (see for example Rijlaarsdam/Couzijn 2000).In order to scaffold, Hillocks (1986) suggests direct instruction in evaluative criteria.In a similar vein to this procedural facilitation, Chanquoy (2008), MacArthur et al. (1991) and Stoddard/MacArthur (1993) propose structured peer interaction and reciprocal peer revision.Mangelsdorf/Schlumberger (1992) generally recommend training students in peer-interaction and response.In support of scaffolding, Stanley (1992) has argued that the lack of organized practice could be unfair to the students.

Contra scaffolding
However, Couzijn (1999) discusses peer-feedback introduced as a task-execution or "learningby-doing" condition, which is in direct opposition to guidance or instruction.Here, instruction is presented without teacher or peer modelling (Braaksma et al. 2002).Caulk (1994) and Rollinson (2005), otherwise a promoter of guided feedback, state that positive feedback results have also been found with students who have not been subjected to prior training.Formalized approaches of guided feedback have a downside.While they are claimed to help provide concrete and useful feedback (Min 2006), such guidance might also result in rubberstamping advice by peers (Sommers 1980), which might not do the peers' text products -unique as they appear -enough justice.
For academic writing and for feedback in general, checklists are available but they tend to focus on mechanics and organizational structure (Oshima/Hogue 2006, Thoreau 2006: 246).Given their generalized nature, they do not focus on domain content or on particular requirements of the genre at hand, and hardly ever on the process of giving, receiving and implementing feedback.It could be argued that students provided with scaffolding might tend to follow guidance or modelling slavishly, more concerned with providing the "expected" results than with peer interaction or incorporation of self-administered feedback.

Combined approaches
Latest peer-feedback research, e.g.Leijen (2014: 179), discusses pre-defi ned feedback criteria vs. self-administered feedback criteria.In a similar vein, Nicol (2014) discusses a mix of teacher formulated criteria for student evaluation of each other's work on the one hand and self-administered criteria to be built up by the students themselves on the other.In addition, Nicol et al. (2014: 118) conjecture a general twofold benefi t of a mixed approach in which they claim that students are able to generate "richer criteria than those provided by the teacher but sounder criteria than those they might be able to formulate themselves."Both approaches draw on the general notion that students are able to make holistic judgements about quality and that they are able to rationalise these through identifi cation and articulation of criteria (Sadler 2013).
The questions to what degree instruction, scaffolding or modelling should be applied and how students are supposed to master the feedback process remain unanswered.When we introduced a peer-feedback task into the classroom, we believed that it was necessary to fi rst gain an overview of the students' abilities and skills before peer-feedback criteria could be established and provided to the students.It is not always clear if existing peer-feedback models and checklists might be suitable for different settings, learners, learning situations and within the framework of institutional restrictions.Therefore, we took a minimalist approach to scaffolding in this study.This means that we explained to students that comments should be constructive and related to linguistic errors, academic writing and domain content.We adopted this approach to ensure that we could retrieve original student performance data which do not show traces of or refl ect elements from prior instructional treatment.

Feedback types and revision types
Studies in peer-feedback tend to focus on areas such as feedback processes, feedback types, levels of expertise and feedback implementation and less so on subject areas, content and issues related to the application of peer-feedback as a teaching tool.We now elaborate on each of these points.
First, attention is given to the classical threefold division of feedback processes (Van Steendam et al. 2014): detection (of errors in the partner's text), diagnosis (preferably with an explanation of the problem) and correction (in the form of a suggested solution to help the peer).To this we should like to add a fourth process element: the action (implementation or non-implementation of the feedback in the revised text).While researchers understand peer-feedback as a learning tool which can bring the benefi ts discussed above to the classroom practice, students might understand feedback tasks as merely a tidying-up activity.
Second, feedback research distinguishes between different types of peer-feedback (Faigley/ Witte 1981) provided by the peers, and, here, on the types of revision implemented based on the feedback.Taxonomies of feedback types differ across studies, but are typically based on the pragmatic content and the function of the feedback comment.Artemeva/Logie (2002), for example, use local context, content, organization, language, format, writing process, advice, and evaluation as their categories, while Cho et al. (2006) use directive, nondirective, praise, criticism, summary, and off-task as categories.To establish types of revision made based on peer-feedback, Cho/ MacArthur (2010) draw on Faigley/Witte (1981) and Sommers (1980).The resulting revision types include surface changes (simple repairs), micro-level meaning changes (complex repair and extended content), macro-level meaning change (new content and organization), and reference (adding and changing supportive materials).
Third, feedback research also explores the question at what level of expertise these feedback types are provided (e.g., student vs. teacher).Furthermore, it investigates the quality level of the comments provided.More experienced peer-feedback providers appear to have a clearer conception of the feedback task, its goals and scope and revise more "elaborately and effectively" (Hayes 2004: 12).Trained and experienced feedback-providers fi nd more higher-order concerns (HOCs) in texts.These refer to the global level of text, such as content, structure and stylistic problems which change the meaning of the text.It is argued that suggestions of the HOC kind are "better" in peer-feedback, because they enhance suffi ciency, relevance and organization of a text.Untrained or less experienced feedback providers, on the contrary, are said to focus mostly on lower-order concerns (LOCs), often also referred to as simple repairs or surface changes (Sommers 1980, Fitzgerald 1992).Surface changes are deemed to not alter the meaning of the text (Faigley/ Witte 1981: 400).It is claimed that students conducting peer-feedback tasks make more surface changes (Faigley/Witte 1981) than macrostructure or HOCs (Min 2006: 134).Given the typical limited length of student tasks in research studies (Ferris 2003), it is not surprising that they often only contain few macrostructural changes and numerous surface changes.This should not automatically be translated as an inability on the students' part to provide quality feedback, as surface changes can improve texts signifi cantly (Paulus 1999, Min 2006).In this connection, it must be noted that L2 learners have been found to focus on surface changes when they give other students feedback on their writing (Fitzgerald 1987, Stevenson et al. 2006).This might be due to lower profi ciency levels in their L2.In addition, an abundance of LOCs in L2 texts might distract from HOCs.This focus on form instead of content with L2 learners might be the product of a long tradition of a focus on form in language classes (Lindgren et al. 2008).
Fourth, feedback research is concerned with the question of why feedback is implemented or not and how it is implemented.Feedback implementation by the feedback receiver is investigated to understand the conditions under which students implement feedback they receive, as this is assumed to be critical to promote improvement (Nelson/Schunn 2009: 376).Feedback implementation is dependent on the three elements mentioned above.It can be understood as the action taken in the feedback process in response to feedback suggestions.The scope of feedback, which according to Nelson/Schunn (2009: 380) ranges on a feedback continuum from local to global, is likely to affect feedback implementation.These authors argue that, as the scope of the feedback increases, the writer (here: feedback receiver, our addition) is expected to be less able to implement the feedback as a result of the complexity of the global issues (Nelson/Schunn 2009: 380).It could be argued that global feedback might have a greater effect on the quality of the text product than local feedback, but it might be more likely that local feedback is implemented.But a study by Matsumura et al. (2002) found that the scope of the feedback did not have an effect on the quality of the text product.Besides scope and quality, the extent of the feedback, the tone of the feedback (praise and mitigating language versus affective language and emotional comments), feedback providers' confi dence, peer composition, ownership issues (e.g. of the assessment criteria) and knowledge of the course material also play key roles in feedback implementation.Our feedback study described below investigated how students provided, received and implemented peer-feedback in a translation/localization task.

Methodology
The overall study aimed to contribute to a hitherto unexplored use of peer-feedback in the web communication classroom, and in the study reported here we focused on the use of peer-feedback as a translation training tool.

Description of our study
The translation/localization task reported on here required each student pair (eight pairs and one group of three) to produce two texts individually, to swap these with their peer for feedback, and to revise the texts in light of the feedback provided.The fi rst text was a translation of webcopy from Danish into English according to a task description.The task stated that the Danish product pålaegschokolade (conceptual interpretation: thin slices of chocolate often eaten on bread) would be on sale in British shops starting the following autumn, and therefore, the producer, Galle & Jessen, would like to update their website to include an English-language section outlining this news.Once the students had produced their translations, they were required to write a second text, which was an academic translation review.Students also received a task description for this text, in which they were to outline and argue for their translation choices.While there was no word limit for the translation task, we limited the academic translation review to 200 words, and students were expected to use an appropriate academic referencing style (e.g.APA).The majority of the work was completed during class time, but if the students did not complete the task in full, they had until the next day to submit the task to the teacher.The idea behind this approach was for the teacher to monitor the peer-feedback provided, and to provide additional feedback (oral or written) to the student pairs.
Prior to task execution, the students had been to a lecture which covered topics including writing for the web, search engine optimization, website localization and translation.Students were also provided with readings on these topics.We took a minimalist approach to scaffolding for several reasons.First, since our students were from 4 th and 6 th semester, they had previously received detailed feedback from us and other colleagues.Therefore, we believed they understood the concept of providing written feedback.Second, the students completed a module in the 3 rd semester which had a strong focus on translation, and therefore, we assumed that they would be familiar with the practice of translation and the terminology used to analyse a translation.Third, these pairs had already completed two peer-feedback tasks, and received feedback on these tasks from their peers and teacher.This means that the students were aware of the level of detail expected from them for such a task.Last, we had provided students with literature on giving and exploiting feedback, which they were encouraged to read before completing the peer-feedback task on website localization.

Methods
In order to answer our research questions, we needed to code the peer-feedback provided and the resulting peer revisions made.This coding phase was conducted in two stages.

Coding peer-feedback: stage 1
The fi rst stage was to code the peer-feedback comments from the translation task and the academic translation review.Students provided feedback to each other using the commenting function in Microsoft Word, with some students including up to three feedback comments in the one comment box.However, students usually only provided one comment per comment box.In total, the authors coded 184 peer-feedback comments.This coding process followed the approach taken by Artemeva/Logie (2002).Similar to their study, we defi ned our unit of analysis (usually the comment box, but, as mentioned, sometimes more comments were included.If this was the case, the student provided a bulleted list) and devised meaningful types of feedback categories based on the tasks from one pair, before continuing with the coding process independently for the remaining pairs.Following the joint coding of tasks completed by one student pair, we devised fi ve codes for the types of feedback provided by the peer, as explained below.

Motivational Feedback
This peer-feedback praises specifi c parts of the student's writing (translation or academic translation review) or the texts as complete units (e.g."good!You sum up the information on Denmark really well").

Suggestions (Directive)
This feedback includes explicit suggestions from the peer that direct the student to make a particular change to a specifi c unit of text or to the overall text in general (translation or academic translation review).The feedback might include the problem and the solution, or just the solution (e.g."delete this and write: …cocoa beans to produce…").

Suggestions (Expressive)
This feedback includes explicit suggestions from the peer concerning specifi c units of text or the entire text as a unit that the student might like to act upon to improve the accuracy of the translation or academic translation review.Expressive suggestions might include the problem and solution, or just one of these (e.g."maybe use the word box instead of packet").

Questions (Confi rmation/Conversation)
This feedback is in the form of a question to highlight specifi c aspects of the text which the peer believes should be edited.Questions can require students to confi rm the question (e.g."isn't it?"),or be used as a means of initiating a conversation with the peer (e.g."do you think…?").Both question types can include a reference to the problem, the solution, or both.

Statements
This feedback is a non-directive, neutral comment about an aspect of the writing which can be either specifi c to the translation task (e.g."'More popularity' doesn't sound like proper English") or could apply to any writing task, and not specifi cally a translation task (e.g."I think you make an important point in this last sentence, but it is not really linked with the rest of the paragraph").Statements commonly identify a problem without presenting a solution.

Coding peer-feedback: stage 2
The second stage of coding in this study related to the revisions made by students based on the feedback they received from their peers, i.e. implemented feedback.We coded each revision using Cho/MacArthur's (2010: 336) revision coding scheme, which consists of four main categories: surface change, micro-level meaning change, macro-level meaning change, and reference (see Table 1 for more details).According to this scheme, surface changes and reference categories do not change the meaning of the text, while micro-and macro-level meaning changes do.
When coding the revisions, we broadened Cho/MacArthur's (2010) surface change to include two subcategories as proposed by Faigley/Witte (1981: 403).In contrast to Cho/MacArthur's study, our students commonly provided peer-feedback which caused students to revise the text in relation to spelling and grammar, and therefore, two subcategories were needed.The fi rst is formal change (e.g.conventional copy-editing operations) including spelling, tense, punctuation, abbreviations, and format, and the second is meaning-preserving change (e.g.changes that paraphrase the concepts in the text but do not alter the meaning of them).These include additions, deletions, substitutions, permutations, distributions, and consolidations.The next section presents the main fi ndings, including the types of feedback provided by students to their peers, the types of revision made by our students based on the peer-feedback, and the peerfeedback provided but not implemented by the peers.

Results and interpretation
We fi rst provide an overview of the complete data set.Table 2 shows the number of participants in the study, and a breakdown of the feedback data for both parts of the task.For the translation task, the students produced 146 feedback comments, 93 of which were implemented, and 53 of which were not implemented.For the academic translation review, the students produced 38 feedback comments, 10 of which were implemented, and 28 of which were not implemented.Before providing a detailed analysis of our results, we present an overview of the types of peerfeedback coded in our data set, and an indication of how often each feedback type was implemented by the students in their translation tasks (see Table 3) and their academic translation reviews (see Table 4).These tables highlight the combination of feedback types and the resulting student action based on our minimalist approach to peer-feedback scaffolding.As opposed to the data presented in Table 3, the feedback types presented in Table 4 are not as varied and noteworthy as we had originally envisioned.Furthermore, only some student pairs completed this part of the feedback task fully, which resulted in a data set that is much smaller than the authors had expected.
We will now present a more detailed analysis of the fi ndings from this study.These fi ndings are presented in three subsections: implemented feedback in the translation task, non-implemented feedback in the translation task, and feedback comments in the academic translation review.

Data analysis 1: implemented feedback in the translation task
First, we analyse the type of peer-feedback implemented by students in the translation task, the resulting revision types recorded, and the impact the implementation has on the accuracy of the text (see coding process in Section 3.2).To judge the accuracy of the text (original translation, feedback comment, and revised translation), the authors used an accuracy scale from 1 (strongly disagree) to 5 (strongly agree) following the process conducted by Nelson/Schunn (2009).This accuracy scale was applied to each unit of analysis (sentence or heading in the original translation, feedback comment, and revised sentence or heading in the fi nal translation).Accuracy was judged based on the degree to which the authors agreed or disagreed with the content of the feedback and the resulting action (implementation vs. non-implementation).There was 90% agreement between the authors, which can be deemed acceptable.It should be noted that the authors only judged the accuracy of segments and not the text as a whole.
Table 5 provides a broad overview of the impact feedback implementation had on the revised translation in terms of text accuracy.From the 93 implemented feedback comments, 78 of these had an explicit positive impact on the accuracy of the translations, as judged by the authors.The majority of revisions made were surface changes (90%), with the remaining revisions split evenly between micro-level changes: complex repair (5%) and extended content (5%).We found no examples of macro-level meaning change or reference.
Previous studies show that non-directive feedback (coded as questions, statements or suggestions (expressive) in this study) is positively associated with micro-meaning complex repair changes.A cursory inspection of the feedback data in this study indicates the same positive association, yet on a much smaller scale (seven examples) than reported in Cho/MacArthur (2010).Cho/MacArthur (2010) also found that directive feedback is positively associated with simple repairs (e.g.mechanical changes in writing which could be compared with surface changes (formal) outlined above).In our study, peers most often provided suggestions (directive) which caused students to implement formal and meaning-preserving surface changes.Table 6 provides some examples of surface (formal and meaning-preserving) changes implemented in the students' translations which were initiated by directive feedback.Example 1 highlights an incorrect translation from Danish (kakaobønne) to English (cocoa seeds).
Using translation-related terminology, we can claim that the student used the oblique translation microstrategy (sense-for-sense procedure) rather than the direct translation microstrategy (wordfor-word procedure) (Schjoldager et al. 2008: 97), which would have worked in this context.The directive feedback is in the form of a problem/solution comment which adds to the credibility of the feedback, and is also more likely to be implemented by the student (see also Ferris 1997, Nelson/Schunn 2009).
The feedback examples in 2 and 3 are two of 21 directive peer-feedback comments that focus on agreement errors found in the translations.It is not uncommon to fi nd this grammatical error in an English text written by Danish students, since verbs in Danish do not decline according to person, only according to tense, unlike their English counterparts, for example, jeg spiser (I eat), han spiser (he eats), de spiser (they eat).However, it is less common for students in 4 th and 6 th semesters to make these kinds of mistakes than students in 1 st and 2 nd semesters.Nonetheless, this error does occur, and students of all levels can usually identify the error.The debate concerning the extent to which students with English as a second language (ESL) benefi t from corrective feedback continues in the literature (Bitchener et al. 2005, Bitchener 2008, Ferris 1997, 1999, Truscott 1996), with no consensus having been reached so far.This study did not aim to address the merits and pitfalls of providing students with corrective feedback, given the size of the study in terms of participants, tasks and time period.However, the data set provides some insight into how peers provided corrective feedback and the resulting revisions from this feedback.Based on this information and on what has already been discussed in the literature, we can guide our students on how to provide corrective peer-feedback that will empower the student who is implementing the feedback, rather than simply directing them to focus on isolated syntactic and grammatical errors.
In the description of the study, we outlined the translation-specifi c training the students had already received in a previous course.Even though we did not explicitly state in the task description that students should comment on macro-and micro-strategies or writing for the web, these topics were assumed to be an integral part of the task.Peer-feedback that included suggestions (expressive) sometimes resulted in students moving beyond surface changes to add new content and elaborate on existing content.Table 7 outlines three examples of higher-order translation peerfeedback comments and the subsequent revisions implemented by the students, which, however, do not necessarily result in a grammatically correct revised translation.In example 1, the feedback comment was a suggestion (expressive) and non-directive, where the peer highlighted what she considered to be the problem ("It sounds like it is not nice enough for others to like it as musch [sic] as Danes"), and provided a possible solution ("maybe just 'they are not as common abroad yet'?").The student implemented the feedback, using slightly different phrasing, resulting in a more accurate revised translation: (before) pålaegschokolade is not popular among people outside Denmark because they do not like it; (after) pålaegschokolade is not popular outside of Denmark because it is not available in Britain.From this student's academic translation review we learned that the student applied a target-text oriented macrostrategy (Schjoldager et al. 2008: 71-87), and a paraphrase microstrategy (Schjoldager et al. 2008: 89-112) to explain the product's popularity in Denmark to the new target audience, and how it could become popular in the new market.
In example 2, the student provided a suggestion (expressive) feedback comment.She highlighted a possible problem in the text ("Maybe you could specify that it is pålaegschokolade that you are referring to").In the original translation, the student removed any mention of the Danish term pålaegschokolade, and the peer noted that this could cause problems in the target text.It is a diffi cult task to fi nd an English equivalent for the Danish product, and therefore, the solution sug-gested by the peer to keep the Danish word and provide an explanation in English was appropriate in this context and would improve the accuracy of the revised translation.This is an example of using the explicitation microstrategy (Schjoldager et al. 2008: 89-112), and relates to the revision type micro-meaning extended content.However, the peer incorporated the solution incorrectly, resulting in an incomplete sentence, and thus this implementation lowered the accuracy of the revised translation.
In example 3, the peer commented on the part of the sentence "sees the light of the day", which was directly translated from the Danish source text "ser […] dagens lys".The peer-feedback comment implied that this was not an idiomatic phrase in English, and the student has been overly infl uenced by the original Danish text.From a translation point of view, the microstrategy direct translation has been applied correctly in this case, as the English phrase "see the light of day" could be used in this context.The problem centres on the incorrect production of this English phrase (e.g."see the light of the day" -a calque translation ), and not on the fact that the student employed a direct translation microstrategy.However, the feedback comment motivated the student to make a surface change (meaning-preserving), and produced a different idiomatic sentence, and hence a more accurate revised translation.
Given our minimalist approach to scaffolding for this task, it was encouraging to note some higher-order peer-feedback translation comments.Yet the peer-feedback comments and the resulting edits to the translations could have been extended to include cultural references that would be expected when localizing a website.

Data analysis 2: non-implemented feedback in the translation task
This section focuses on the peer-feedback comments that were not implemented.Following an analysis of the fi nal translated texts and the academic translation reviews, we did not identify any instances where the motivational feedback had been explicitly implemented.Based on the data we are unable to specify the kind of impact the motivational comments had on the receivers.One possibility could be that the students who received motivational feedback were more motivated to take on board suggestions from their peer in general than if they had not received motivational comments.There was only one pair in which one of the students never provided her peer with motivational feedback.However, this did not deter the peer from considering the other types of feedback, and she implemented ten of the thirteen peer-feedback comments provided.Interestingly, the three comments she did not implement all asked her to rephrase a sentence and they provided a part-fi nished solution.Yet she still decided not to implement the feedback.Of the ten feedback comments she did implement, four of these were incorrect.In all ten cases the type of feedback was suggestion (directive), for example, 'misspelling', 'remove comma', 'article 'the''.These fi ndings could indicate that the student considered her peer to have stronger English language skills than she did, which meant that she implemented the directive feedback, but shied away from feedback which related to higher-order concerns (HOCs).
Our fi ndings are in contrast to those from Cho/MacArthur (2010), where their feedback category 'praise', similar to our 'motivational', resulted in revisions being made by students.While the pre-and post-course online questionnaires of the larger study are not our main focus in this paper, it is interesting to note some of the responses we received relating to motivational feedback.In the pre-questionnaire, fi fteen students who completed this translation/localization task said that they agreed or strongly agreed with the statement "I like that feedback includes praise".The remaining two students neither agreed nor disagreed, and one student did not complete the pre-questionnaire.Seven of the eighteen students completed both the pre-and post-course questionnaires.Within this group, six students agreed or strongly agreed with the statement "I was pleased with the feedback I received from my peer", while one student neither agreed nor disagreed.Regarding the statement "I received praise from my peer", all seven students agreed or strongly agreed with this statement.Given the low number of responses in the pre-and post-course questionnaires, it is diffi cult to make any claims regarding motivational type peer-feedback.But, the student respons-es in this small data set could indicate that praise motivates the students to complete the feedback and fi nal revision task, or that it helps their learning process.This could be an argument for including praise in peer-feedback, even though it did not have an explicit effect on improving the accuracy of the revised translation (see for example Ferris 1997, Nelson/Schunn 2009).
In Table 2 we highlighted 53 feedback comments that were not implemented.This number included the 31 motivational comments received by the students, as we were unable to claim whether the students implemented these comments or not.In this section, however, we only discuss the remaining 22 feedback comments categorised under the other four types.Table 8 outlines the types of feedback received, and the types of revision that the students intended when they provided the feedback.
Table 8.Types of revisions based on the type of feedback (not implemented) In seven of the cases mentioned in Table 8, the students were correct to not implement the feedback, as their translations were accurate and the peer-feedback was inaccurate.In all seven examples, students were expected to make surface changes to their translations.Three examples of this incorrect peer-feedback are shown in Table 9.In the table, the original translation, the feedback comment and the revision type expected are shown.We defi ne the revision type expected as the action that would have been taken by the student based on the peer-feedback, if the feedback had been implemented.In addition, there are fi ve examples where the translation was inaccurate, the peer identifi ed an issue and provided a suggestion (expressive or directive), but this feedback was also inaccurate.Therefore, the student was correct to not implement the suggested feedback, but the translation should have been edited differently to improve the accuracy of the revised translation.The examples shown in Table 10 in particular demonstrate that the students were unable to identify how to fi x the error highlighted by their peer, and that the peers were unable to provide a correct solu-tion.In the table, the original translation, the feedback comment, the revision type expected and possible edits, as suggested by the authors, are shown.The remaining ten peer-feedback comments should have been implemented to improve the accuracy of the revised translation.Example 1 highlights how the peer identifi ed a problem and (part) solution, but was unsure of how to word the feedback.Taking into account the concepts of web localization covered during the course, the student should have considered this feedback and edited the relevant segment to clarify the meaning for the target audience.However, the non-committal nature of the feedback which included hedging might have potentially confused the feedback receiver, resulting in nonimplementation of the feedback.
Example 2 illustrates how the peer sensed that there was a problem with the sentence, perhaps relating to word order, but she was unable to identify a specifi c problem and provide a specifi c solution.This resulted in the student leaving the text as it is.The example mirrors previous research (Min 2006) where feedback comments which lacked the problem-solution format tended to be disregarded by students more often than feedback comments that included this format.
Example 3 relates specifi cally to the context of this writing task.This peer comment should have been implemented to meet the criteria for producing a target-text oriented translation.Here, the street name was not relevant for the new target audience, and this was the kind of edit the student should have implemented, based both on the literature for the course and the peer-feedback.

Data analysis 3: feedback comments on the academic translation review
The data collected from this part of the task provide only a small insight into the students' thought processes when translating.Even though all students should have written an academic translation review and provided feedback on their peer's work, of the nine pairs, only fi ve actually wrote an academic translation review, and one of these wrote a joint text without the peer providing any feedback.Therefore, this peer-feedback data set comprises 38 comments from four pairs (nine students).The peer-feedback comments from two pairs focused solely on surface changes (formal): for example, the student should correct verb agreement errors and spelling errors.Despite the lack of content-specifi c comments from these pairs, the students all made reference to web localization considerations relating to the translated text (e.g.cultural and linguistic considerations) and design issues (e.g.country-specifi c URL and the marketing strategy for a new target audience).The two remaining pairs provided a mix of peer-feedback comments.Comments from one pair included motivational comments that agreed with the translation/localization choices made.One problem we identifi ed with the review texts from all four pairs was that there was a mismatch between the content of the translated text and the argumentation for choices made in the academic translation review.In each case, the students produced a source-text oriented translation, staying very close to the structure and content of the source text in Danish, even though they argued for using a target-text oriented translation strategy.Furthermore, little consideration was given to how to translate the name of the Danish product (pålaegschokolade), which was one of the main reasons we selected this website for the task as we knew this would challenge the students from a localization perspective.
One pair, consisting of three students, completed this task with the most success in terms of providing peer-feedback.Although they did not refer to any literature on two pertinent topics for this task (writing for the web and search engine optimization), they did reference two central readings we had used on the course for translation and localization, and all three students provided a mix of feedback comments, some of which would require revisions relating to HOCs (see Table 12, examples 1 and 2).Moreover, we identifi ed a peer-feedback comment, in which a student outlined how her peer's work inspired her to make relevant changes to her own translation (see Table 12, example 3).These results are examples of the kinds of goals we hope to achieve by introducing peer-feedback as a translation training tool into the classroom.

Discussion and conclusions
This study provided us with insights into using peer-feedback as a translation training tool in the web communication classroom, but it has some limitations.The fi ndings were based on a restricted number of peer-feedback comments which do not allow for generalizability across different teaching contexts.We cannot make claims based on the fi ndings as to whether the students who provided feedback were empowered or have gained more from the task than students receiving feedback.Our students were equal-status learners, but without testing them on their language skills in a pre-test post-test design we could not distinguish between higher and lower achievers or measure their language skills directly.Our minimalist approach to scaffolding might have caused students to provide little disciplinary and interdisciplinary declaratives.Furthermore, some of the incorrect and non-implemented feedback comments could be interpreted as indicators of learner autonomy, but without retrospective follow-ups, this entity was diffi cult to measure and it was impossible to evaluate.Another construct which was diffi cult to measure was individual vs. mutual responsibility.
To answer RQ 1, our students provided fi ve types of peer-feedback: motivational, questions, statements, suggestions (expressive) and suggestions (directive).The majority of the comments provided were suggestions (directive).Over half of the peer-feedback comments were implemented in the translation task.The feedback comments that improved accuracy ranged from minor improvement in some texts to major in others.This can be considered a positive fi nding from which to build upon, since none of our students received structured guidance on how to give and implement feedback.
To answer RQ 2, our data showed that, of the feedback that was implemented, suggestions (both directive and expressive) most often triggered formal and meaning-preserving surface change revisions.Suggestions as a feedback type might be easier to understand and implement than other feedback types.A possible downside of providing directive feedback to students of lower ability could be that they implement this feedback without refl ection.Depending on the wording and content of the expressive feedback, this might allow the peer to demonstrate their skills and knowledge, which the student receiving the feedback could benefi t from.It is questionable whether students' learning outcomes were improved by implementing a high number of surface changes.Ideally, we want our students to provide and implement peer-feedback on local and global writing errors which would trigger a range of revision types.However, research has shown that the errors made by students differ, depending on whether they write in their L1 or L2 (Ferris 2011: 10).Furthermore, it is unrealistic to expect L2 writers to produce error-free English texts, or for their texts to read like they were produced by a native English speaker (Valdés 1992).We would argue that surface changes play an important role in peer-feedback.The linguistic errors highlighted by the students refl ect their language ability, and this should be considered when developing peer-feedback tasks.It would be unreasonable to expect the students to focus solely on feedback that would produce a change in the meaning of the text, and ignore surface changes that preserve meaning, but improve accuracy.
Students implemented few micro-level meaning changes and no macro-level meaning changes.The macro-level types of revision would have required them to add new content (e.g., new points or entire new paragraphs) or change the organisation of the existing text (e.g.moving a paragraph, introducing/deleting transitional phrases).The lack of these revision types found in our data could be attributed to a low self-investment on the students' part.Other factors could also play a role here: students' lack of writing skills in the subject area, a low level of academic writing skills, a general lack of competence on global writing and translation issues, a lack of ability to generalize, and few opportunities to practice writing tasks during their studies.
The fi ndings showed that not all peer-feedback was implemented.This could be interpreted in two ways.First, students did not simply implement all peer-feedback suggestions, and based on an informed decision regarding language and content, they decided not to implement some of the feedback.Second, it is likely that students did not know how to interpret the peer-feedback and, therefore, ignored the comment or implemented it regardless whether it was correct or not.Retrospective interviews with the students would help us to better interpret these fi ndings.
To answer RQ 3, we identifi ed translation-related issues from our fi ndings for both the translation and academic translation review tasks.One of our assumptions before the task was that students would demonstrate their understanding of TS by incorporating TS terminology when providing peer-feedback.We perceive incorporating relevant course terminology as a learning step.Some students achieved this learning step: they provided feedback comments which showed a strong TS domain knowledge.Other students failed to reach this step and showed weak TS domain knowledge.Some students displayed an ability to identify translation issues specifi c to website localization.Students in this study have produced academic translation reviews in earlier translation-related courses.Therefore, we cannot ascertain why all pairs did not provide an academic translation review.Our guess is that completing the task in full was perceived as "extra work", since the translation task was considered to be the fi nal product.Students performed relatively poorly in terms of providing peer-feedback specifi c to translation and localization.In addition, they were for the most part unable to provide peer-feedback on global writing issues.Finally, students did not highlight a common mismatch between argumentation in the academic translation review and the translation produced.

Scaffolding in peer-feedback as a translation training tool in web-based communication
The literature review, the minimalist approach taken here and the fi ndings from RQ 1, RQ 2, and RQ 3 provide a basis for scaffolding in peer-feedback as a translation training tool in web-based communication.Scaffolding is a technique that our students could use to improve how they com-municate domain content, refl ect on text production processes, and demonstrate academic writing skills.As teachers, we can use the technique to facilitate the students' learning.We suggest the following elements to be included in future scaffolding approaches: • Clear task objectives for peer-feedback • Clear assessment criteria for peers which specify domain content, process refl ection and academic writing • Discussion on the benefi ts of giving, receiving and implementing peer-feedback • Conceptualisation of feedback as a dialogue3 • Understanding peer-feedback as enhancing professional skills • Discussion of types of peer-feedback expected on the continuum from surface to macro-meaning changes • Instructions on how to phrase peer-feedback, including hedging vs. non-hedging language, mitigating language and problem-solution patterns • Instructions on how to best implement peer-feedback • Guidelines on text genres and how to shift between the different genre • Practical guidelines on how to use word processing tools in the peer-feedback process We understand our suggestions for scaffolding to be applicable to other areas where translation is taught together with other domain content.Besides purely learning-based instructions, students must be made aware of types of feedback, their wording and their function and malfunction in self-refl ection, other-refl ection and self-regulating learning contexts in general (Macfarlane-Dick/ Nicol 2006).It must be highlighted that the problem-solution format (problem identifi cation, explanation, suggestion of improvement) is much more likely to be implemented than comments which only highlight a problem (pointing technique, see Stanley 1992).Likewise, mitigating language is more likely to evoke implementation than criticism or praise alone.But in essence, it is the teachers' responsibility to help the students to develop the skills necessary in providing helpful feedback (Blair et al. 2013).Scaffolding equips the teacher with the means to accomplish this.
Having provided the list above, a word of caution is in order.Merely transposing existing scaffolding concepts and models might not bring about the desired results.Teachers need to know the task and their students so that feedback will be "appropriate for this tutee in this task at this point in task mastery" (Wood et al. 1976: 97).The present study and its fi ndings open up further avenues of exploration.Our next research steps will include, but are not limited to, the following: testing different types of scaffolding, conducting focus group interviews before and retrospective interviews after peer-feedback assignments, and discussing experiences of the peer-feedback process with the students to evaluate their learning outcomes.

Table 1 .
Coding scheme used for revision types in response to peer-feedback comments

Table 2 .
Peer-feedback provided, implemented and not implemented

Table 3 .
Types of peer-feedback in the translation task

Table 4 .
Type of peer-feedback comment in the academic translation review

Table 5 .
Impact of implementing the feedback and the resulting translation accuracy

Table 6 .
Surface (formal and meaning-preserving) changes implemented

Table 7 .
Higher-order peer-feedback comments and resulting revisions

Table 10 .
Peer-feedback not implemented, but revised translation remains inaccurate

Table 11 .
Table 11 highlights some of these examples.Examples of peer-feedback that should have been implemented

Table 12 .
Peer-feedback comments on the academic translation review