A Model for Analyzing Teachers’ Written Feedback on Adult Beginners’ Writing in Swedish as a Second Language

: This study serves to ﬁll a research gap in the written feedback practices of teachers in a second language (L2) writing by focusing on adult beginners, previously uninvestigated in feedback studies. For investigating this, a feedback analysis model was developed. Unlike previous models in L2 contexts, in this study feedback is divided into two main areas, focus ( what the teachers comment on) and manner ( how the comments are given). The study was conducted in a web-based Swedish as a second language course and included three experienced teachers and their L2 students. The analysis of teachers’ written comments on 60 texts from 12 of the students revealed that in the two main areas both new and previously used categories of feedback were identiﬁed. In the area focus ﬁve new subcategories were identiﬁed within language accuracy, and in the area manner politeness, and reinforcement of learning outcomes were identiﬁed as new categories. Within the area, focus, the teachers concentrated on language accuracy, and within the area, manner, all three teachers mostly provided information, but also made a range of suggestions. There was also individual variation among teachers. The analysis of feedback strengthens the importance of applying the main distinction between ‘focus’ and ‘manner’ in L2 contexts.


Introduction
Recent research on teacher feedback on second language (L2) learners' writing has mainly concentrated on written corrective feedback (WCF), that is, the correction of grammatical and lexical mistakes and mechanics (Jakobson 2018). However, teachers' feedback practices are holistic in the sense that feedback involves different categories. Studies on teachers' written feedback in the L2 classroom with more advanced students have shown that teachers comment extensively not only on grammar, vocabulary, and mechanics but also focus on text-related categories (content, ideas, organization), and that these written comments contain a variety of syntactic and linguistic features (Magno and Amarles 2011). Thus, there is a divide between what research has focused on (mainly narrowly designed effect studies on WCF) and the holistic nature of teachers' feedback practices. Many researchers (e.g., Hyland and Hyland 2010) have analyzed feedback from a broader perspective and claimed that different types of feedback should be used and include in these many focus categories such as content, organization and WCF, and that this can contribute to language development and increased awareness, as well as self-regulation by the writer in the long term. Developing their knowledge, the students need to be primarily supported by formative feedback (feed forward)-the most important component of feedback (Hattie and Timperley 2007) because it creates opportunities to do better next time (Black and Wiliam 2018). Moreover, not only focus categories (what the teachers comment on e.g., content), but also another area, namely the manner of feedback (how the comments are given e.g., praise) is an important part of feedback (Sutton et al. 2012;Straub 1996). The categories in both areas are closely related to classroom teaching and learning (Hyland 2001) and, therefore, in order to better understand feedback, investigating feedback from a more holistic perspective can provide a better description of feedback and help make it more relevant. In addition, both the focus and the manner of the feedback can have decisive influences on how it is received and used by the students. According to Wiliam and Thompson (2007), how students use the feedback is a significant factor for further learning.
Furthermore, previous studies almost exclusively concern feedback provided to university students in advanced and intermediate level ESL/EFL in process writing classes in English-speaking countries, and relatively little is known about what teachers focus on and how they give their comments when they respond to beginner-level texts (Jakobson 2018). In addition, it is more important to describe and understand the complex teaching process in specific contexts instead of trying to generalize about what is true for all classrooms (Goldstein 2016). As research on feedback in the context of adult students in Swedish as a second language (SSL) is generally underrepresented (Jakobson 2019), there is a need for studies on feedback that focus specifically on adults' second language writing in the Swedish context.
In online language teaching, a common mode of education today (Coleman and Furnborough 2009), feedback plays an important role for supporting student learning (White 2003). Online teaching technologies may also influence the teachers' feedback practices and their feedback choices (Kessler 2018). Furthermore, online written feedback for adult beginners is largely unexplored (Jakobson 2015). Therefore, there is a need to investigate teacher written feedback in online language learning contexts for adult beginners. This may provide teachers with new knowledge and help provide better understanding of feedback.
To address the issue of the under representation outlined, the present study constitutes a contribution to knowledge by investigating three experienced teachers' written feedback on adult beginners' written texts in SSL as well as providing a basis for further research.
Based on the holistic perspective of a hermeneutic approach (Gadamer 2010), the aim of this study is to investigate teachers' written feedback from a holistic perspective by developing and testing a model for analyzing feedback. The model is divided into two main areas. The first area is the focus of feedback and the second area is the manner of feedback. Both areas include categories that can be identified as parts that make up feedback as a whole. The starting point for the model is to test previously used feedback categories identified in studies on written feedback for writing, and also to identify any potential new categories of feedback.
With the aim of developing the model, the study first addresses the following question: what categories of feedback can be induced from teachers' feedback practices in the main area, focus, and in the main area, manner? With the aim of applying the model, the second question in the study is: to what extent do the teachers use the different categories of feedback in the main area, focus, and in the main area, manner?

Feedback Categories
As research on feedback in SSL is very limited, previous research on feedback for other second languages is presented below. There are three areas of research that currently dominate in international research in written teacher feedback in second language contexts. These are: (1) teacher feedback practices; (2) students' perceptions of feedback, and (3) teachers' perceptions of feedback (Goldstein 2016). The present study is focused on teachers' feedback practices, and this is the basis for the categorization of feedback in two main categories, focus and manner. First, the focus categories (also called global and local textrelated feedback by Connors and Lunsford 1993) identified in previous studies are described in Section 2.1. Thereafter, the manner categories (also called interpersonal feedback by Hyland and Hyland 2010) identified in previous studies are presented in the same section. Finally, linguistic and semantic features are included as subcategories of feedback and they are presented in a separate section.

Focus and Manner
Focus categories such as content, organization and language accuracy are wellestablished basic categories of feedback used to develop comprehensive writing skills, and these categories originate from L1 composition writing (Ferris 1997). Feedback on language accuracy can be divided into subcategories such as vocabulary, grammar, and mechanics (e.g., Montgomery and Baker 2007). While some studies show that teachers focus on content (e.g., Conrad and Goldstein 1999;Ene and Upton 2014), others have shown that teachers pay most attention to the language (e.g., Junqueira and Payant 2014;Montgomery and Baker 2007;Lee 2009;Lee et al. 2018). Additionally, Hyland and Hyland (2010) found that teachers consider students individual needs while Lee (2008) showed that the teachers do not address students individual needs. The variety and the mixed results in teachers' feedback may be due to many different factors. Such factors can be the type of course, individual differences between students and teachers, consideration of the syllabus requirements, and student proficiency levels (Bitchener and Storch 2016;Ferris et al. 1997;Goldstein 2016). These studies suggest that feedback is a complex concept to investigate.
Considering the L1 context, Straub (1996) argued that it is necessary to investigate both focuses and modes (Straub's emphasizing, p. 233) of teacher comments. It was Straub (1996) who first pointed out the function of feedback as a relationship-building link between the writer and the teacher in Straub's ambitious and creative projects (Ferris 2003). The common categories included in manner are Directives, Suggestions, Praise and Criticism (Lunsford and Straub 2006;Sutton et al. 2012). The different manners of feedback show the type of relationship the teacher creates with the writer in the current context (Lunsford and Straub 2006). The teachers decide the level of control in their feedback based on whether their perception is more teacher-centered or more student-centered Lunsford and Straub (2006). For example, teacher-centered criticism limits the development of ideas and shows a high degree of control, while student-centered praise draws attention to the strengths of the text and has a motivating effect on the writers, encouraging them to continue working (Lunsford and Straub 2006). Ferris (2003) pointed out that studies on praise and criticism, and what the balance between these should look like, have not been investigated. Concerning criticism, this gap in the research still remains. For example, praise is included in ten studies while criticism, interestingly, is investigated in only three studies Hyland andHyland (2001, 2010) 1 and Lee (2009). Research on suggestions in feedback are only found in a few studies (Hyland andHyland 2001, 2010). In their studies on the feedback of three teachers given to intermediate and advanced level students in English proficiency course, Hyland andHyland (2001, 2010) investigated the extent of feedback in the categories of criticism, praise and suggestions. Those categories were not divided into specific and general feedback. The researchers found that the feedback given by all three teachers contained mostly praise (44%), followed by criticism (31%) then suggestions slightly less (25%).
There are some studies that investigate manner categories such as praise, criticism, and suggestions along with focus categories such as content, organization, and language accuracy (Hyland 2001;Hyland and Hyland 2010;Lee 2009). Hyland (2001) explored different types of written teacher feedback on a work from students with different levels of language proficiency studying by distance. Hyland (2001) designed a model for the analysis of feedback using text-related categories in the main category, product, and four categories in the main category, process. Hyland's (2001) model was modified and applied in a Swedish learner context earlier by me (Jakobson 2015) and used to analyze teachers' written feedback on writing and pronunciation to beginner level L2 university students. The results showed that the teacher provided most feedback on language accuracy and focused mostly on grammar, vocabulary and spelling compared to content and organization.
Suggestions concerning writing in L1 and L2 contexts point out that feedback is effective if it contains both praise and constructive criticism (Hyland andHyland 2001, 2010;Straub 2000). However, it is important to consider that different people can encounter and perceive praise positively or negatively depending on their characteristics or cultural background, for example (Hattie and Timperley 2007). Nevertheless, several researchers (Hattie 2012;Hyland and Hyland 2001;Straub 2000) are consistent in their opinion that praise and criticism in feedback should be used carefully, sincerely, clearly, and with caution. Regarding the distinction between general and specific comments, general critical comments do not have any positive effect on learning (Douglas and Skipper 2012). For example, praise should be given specifically and directed towards the achievement and not towards the individual (Hattie 2012). Furthermore, unfounded or general comments create interpretation problems and may not contribute to better writing, especially concerning students in the earlier stages of language learning .

Linguistic and Semantic Features of Feedback
Linguistic and semantic features of feedback such as imperatives, statements, and questions have been analyzed as forms of suggestions in Ferris et al. (1997). Besides those evaluative categories of feedback there is also an additional variant of non-evaluative teacher comments (Elbow 1999) in the form of informative statements.
According to Ferris (2003, p. 24), the first inductive studies that analyze the pragmatic aims and the linguistic structures of feedback implemented on individualized assignments were those by Ferris et al. (1997) and Ferris (1997). Ferris et al. (1997) categorized 1500 comments of an experienced teacher provided to 47 advanced level L2 students. The students wrote four major process-writing assignments. The results showed that the teacher provided both marginal and end comments which were examined according to the teacher's goals (i.e., positive comments) and the linguistic forms. The results showed that the teacher wrote mostly statements and then questions and that imperatives were rarely used. Indeed, the authors argue that there is a need for further development of analysis of teachers' feedback practices because of the tentative character and the difficulties with the models. For example, they maintain that "the teacher's comments did not always break down neatly into single phrases, sentences, or idea units" (Ferris et al. 1997, p. 168), which shows that analysis is a complicated process.
Other studies have mainly referred to and adapted the Ferris et al. (1997) model (e.g., Conrad and Goldstein 1999;Sugita 2006). These studies contain different sets of categories and have further elaborated how linguistic and semantic features of teacher feedback have influenced students' revisions. Lunsford and Straub (2006) label questions as "interactive comments" because they give the writer some freedom to find their own solutions. Statements, according to Lunsford and Straub (2006) are the least controlling manner, as they reflect the teacher's interpretations, while imperatives express a high level of control. Lunsford and Straub (2006) claim that imperatives are more controlling than questions and non-judgmental statements.

Method
The first section of this chapter presents the context of the study, the second section deals with the data and participants. The third section describes the procedure of data analysis and the last section respondent validation.

The Context
The course investigated was a web-based course in communication for immigrant learners of beginner Swedish (SFI). The course was provided in 2016 by a private company specializing in adult education and followed the national curriculum for non-compulsory schooling in SFI, intending to help the students reach a level of language proficiency of A2, according to the CEFR. All immigrants who are at least 16 years old, have a residence permit and are registered in a local municipality in Sweden, are eligible to take the course. The students had college or university education and are non-native speakers of Swedish. All communication between teachers and students took place in an online Learning Management System (LMS) without any physical meetings in classes.
The students followed a study guide that contained a study plan, materials and exercises (reading, grammar, pronunciation) stored in and accessed through the LMS. The students worked through the materials on their own. The course consisted of several modules, and each module represented one topic, such as everyday routines or shopping. The course included nine single-draft writing assignments consisting of various types of texts such as letters, stories, job announcements and advertisements. Task instructions contained information about what content should be included in the text, and in some cases, how the texts should be organized. All students received the same assignments in the same order during the course. They submitted their texts regularly through the LMS according to the study plan and then the teachers provided written feedback on the texts. Teachers commented on each text only once. The main communication between teachers and students was written feedback from the teacher on the students' text. The students were expected to read through the comments and use the feedback to improve their new pieces of writing. Five different teachers provided feedback during the course.

Data and Participants
The data were collected from the copies of all students' (39 from 23 different countries) texts, with written feedback from teachers on all nine written assignments that were sent by the coordinator at the company to the researcher after the course was finished. Three teachers agreed to participate in this study, and all were qualified teachers of SSL with extensive experience in language teaching (Teachers 1 and 2 had 20 years' experience, Teacher 3 had 10 years' experience). The teachers also had several years of experience teaching Swedish online (between 3 and 6 years). Teacher 1 had provided feedback to 20 students (181 texts), Teacher 2 to four students (33 texts) and Teacher 3 to nine students (85 texts).
The data for analysis was limited to written feedback from teachers on five of the nine written assignments (assignments 1, 2, 3, 6, and 7) provided by these three teachers to different students (N = 12), that is, 60 texts in total. The first assignment averaged 90 words in length and the average length for the fifth composition was 121 words, which indicates some development of L2 writing skills. The four other written assignments were excluded because sometimes another teacher had commented on the text from the same student, so these texts were excluded. Regarding consistency (Révész 2015), only those texts where the same teacher had responded to the same student were chosen for analysis. As Teacher 2 had provided feedback to four students, feedback on all those students was analyzed. Feedback given to four students, even for teachers 1 and 3, was included to create comparability and a manageable amount of data. The four students texts with feedback for Teachers 1 and 3 were chosen with the aim of selecting students with as many different backgrounds as possible. The only information about the background of the students provided by the education company was their name, age, gender, and country of origin. To achieve variety within the group, texts written by students from different countries and proportionally from different age groups (age range between 31-43) were selected. Of the twelve students selected, five were men and seven women.

The Procedure of Data Analysis
Through two initial analyses of the data, end comments were separated from the correction of grammatical or lexical mistakes in all texts. The first analysis involved final comments, all written in Swedish, which appeared only at the end of students' text and not in margins. The comments were in the form of a sentence or a phrase identified as separate feedback comments. The second initial analysis identified and counted error correction units (WCF) with the aim of obtaining an overall picture of the frequency of teacher feedback. Teacher 1 and 2 gave WCF as "hints" ("tips") in the form of single words or phrases after first commenting. Teacher 3 either gave the correct version in red color or underlined errors in the students' text. The central focus of this study concerns the analysis of written comments and not corrections in the text. The written comments were categorized as belonging to either manner and focus and were further organized into the appropriate subcategory. In line with Straub (1996), the focus of feedback and the manner of feedback were separated as two completely different areas.
Three different Word documents, one document per teacher, were created for feedback analysis. Each document contained 20 tables-five texts multiplied by four students' texts. Comments from all texts were read and counted. In previous analysis models, such as Conrad and Goldstein (1999) and Ferris et al. (1997), a feedback comment can consist of several sentences. These researchers have pointed out that it was difficult to determine where one comment or feedback comment ends and the next one begins. A few of these "compound comments" were found in the present study, but every sentence was considered a single comment that allowed for precise coding and classification. Each comment was placed on a separate line in the table and numbered and counted in another column. In total, 292 teacher comments were identified (Teacher 1: 127, Teacher 2: 92, Teacher 3: 73) and these were then further analyzed as follows.
During the coding of written comments, categories were constructed from data (categorization), and data were classified into those categories according to Strauss (1987) recommendation. By following the "unitizing principle" (Lincoln and Guba 1985) and "a feedback point" (Lee 2009, p. 14), if there was more than one focus or manner unit within the same comment, they were identified and counted separately. To create greater precision and consistency, I compared all comments in each category with each other (Corbin and Strauss 1990) after analyzing the feedback of three teachers. I also constantly alternated between inductively identifying new categories in focus and manner and then trying to link to the categories that have appeared in previous studies, which created an interaction between previous studies and the present data. By following Corbin and Strauss (1990) recommendation on systematic comparisons, it was possible to break down several categories into subcategories within the main area focus and manner.
In line with Révész (2015), this coding scheme was tested with an experienced teacher in SSL. Then, several rounds of coding were performed during the continuous coding process to build up the model until no new categories emerged other than those already included in the model. After organizing the comments into the main area, focus, categories such as Content and Organization, Language Accuracy and Reinforcement of Learning Materials were adopted from previous studies (e.g., Hyland 2001; Montgomery and Baker 2007). The subcategory Language accuracy, was further divided into Vocabulary, Grammar, and Mechanics, which are adapted from Montgomery and Baker (2007) and Lee et al. (2018). Drawing on the studies by Ferris et al. (1997) and Ferris (2007), the model concerning the main area, manner, in this study was partly based on the categories Giving Information, Suggestions (in the form of questions, statements and imperatives/commands), Praise and Criticism. All categories were defined so that other researchers can identify them. The coding process lasted for three months, and by repeating the same procedure several times, the process improved (Corbin and Strauss 1990). Consistency meant that specific comments and units could be grouped into the same main category and the same subcategory. Of the total 292 written comments, 481 units were classified as belonging to focus and 422 units as belonging to manner.

Respondent Validations
After the preliminary classification and coding, respondent validation was used because it was difficult to interpret whether a comment was intended as an imperative, as criticism, or information (Teacher 1-seven comments) and whether a request should be construed as a criticism (Teacher 2-four comments; Teacher 3-five comments). In order to avoid misinterpretation, in these doubtful cases the teachers were asked what they meant. This validation, called member checking, (Lincoln and Guba 1985) took place in a personal interview where all doubtful cases were discussed with the teachers. After the interview, the final classification of the comments was carried out. It is interesting to note that Teacher 1 and Teacher 2 had used the same comment, 'Här börjar du många meningar med Jag' (Here you start many sentences with I) but classified this comment differently. According to Teacher 1, this comment was a criticism, while Teacher 2 thought it was information. In this case, I interpreted the quoted comment as criticism. Teacher 2 noticed that, on the one hand, everything that is missing in a text could be interpreted as criticism but, on the other hand, it can also be perceived as information by the teachers and students.
Respondent validation was also used as a second control of validity. The entire classification of the teachers own comments was sent in the form of a table to each teacher via email. The classification consisted of 20 tables per teacher (five assignments and feedback on four students per teacher). The classification also included a coding explanation for at least each unit in the tables. Teachers were asked to carefully review the classification and comment on whether they thought any comments were incorrectly classified. Final revisions of the classification were carried out after the teachers sent their feedback to the researcher.

Results
The findings presented in Section 4.1 attempt to answer the first research question of the study by presenting the categories that occurred in two areas; first focus and then manner. The findings which aim to answer the second research question are presented in the Section 4.2.

Focus
The categorization of teachers' comments after data analysis revealed the following categories: Content, Organization, Language Accuracy, Reinforcement of Learning Outcomes and Reinforcement of Learning Materials. As mentioned before, I adopted Content, Organization and Language Accuracy from previous studies, and Reinforcement of Learning Materials from Hyland (2001). One new category identified in the present study was Reinforcement of Learning Outcomes. In addition, the results showed that the teachers wrote other comments that could not be classified into any of these categories: general comments with no focus on writing and other comments. Concerning the category Language Accuracy, the following subcategories were created during data analysis: Grammatical Structures, Vocabulary, Spelling, Mechanics and Language in General. Two new subcategories were identified: Spelling and Language in General.
All categories extracted from the data with examples of teachers' comments in italics and then translated into English are listed below. Concerning the area, manner, the categorization of teacher comments revealed that five categories emerged from the data analysis: Information, Criticism, Suggestions, Praise, and Politeness. Information was divided into two subcategories: Giving Information in the Form of Statements and Asking for Information about the Content in Form of Questions. Regarding Criticism, the teachers in the present study did not write any general critical comments, which means that they focused on a particular aspect in their comments. Suggestions contain three subcategories: statements, questions, and imperatives. Concerning Praise, in previous studies this category was not divided. In this study, this category is divided into Specific and General Praise. Politeness is a new inductively developed category. In sum, the main category manner includes the following categories with examples of teacher's comments in Swedish translated into English: 1. Information a. Giving Information contains explanatory or descriptive non-evaluative statements that inform about the current text and do not refer to new texts or show any direction. Läsaren förstår vad du vill säga. 'The reader understands what you mean'. b. Asking for Information about the Content are comments in the form of questions that ask for more information than what the current text provides. These comments only concern content. Vilken storlek beställde du? 'What size did you order?' 2. Specific Criticism expresses a negative value where the teacher describes a problem. Dina meningar är ibland väldigt långa. 'Your sentences are sometimes very long'. 3. Suggestions advise a future change and improvement in new texts and leave the student to decide the possible need for a change. Suggestions were formed as statements (3a), questions (3b) or imperatives (3c): a. Statements are non-evaluative neutral comments. Datum och ort är också bra att fa med. 'It is good to include the date and place as well'. b. Questions differ from Asking for content information (1b) because they relate to the general improvement and do not address content. Hur kan du variera meningarna mer? 'How can you vary your sentences more?' c. Imperatives are comments that require a change in the text or action from the student. Skriv om så att det blir mer varierat! 'Rewrite so that it becomes more varied!' 4. Praise is a positive comment and can be interpreted as the opposite of criticism.
a. Specific Praise is a positive comment that is related to a particular aspect that the teacher focuses on in the comment. Ordföljden är korrekt och varierande, bra jobbat! 'Word order is correct and varied; well done!' b. General Praise is an unspecific positive comment. Bra jobbat! 'Good job!'. 5. Politeness is a positive expression, an aspect of interpersonal communication that does not refer to praise Tack för din text! 'Thank you for your text.' In sum, to answer the first research question, it was found that most of the categories used in previous studies with more advanced learners also occurred in the data for the present study where feedback was given to beginners. Furthermore, several new categories were developed both in focus and manner. In focus, similar to previous studies, there were comments that could be categorized both as text-related categories such as Content and Organization as well as Language Accuracy. In the analysis for the present study, Reinforcement of Learning Materials, previously only used by Hyland (2001), was identified, perhaps because this feedback was given to beginner level learners in the present study. In addition, Language Accuracy was used in previous studies but was not divided into as many subcategories as in this study. Concerning manner, the results showed that five categories occurred in this area whereas one of them (Politeness) was a new inductively developed category.

The Frequency of Use of Specific Categories in Teachers' Feedback
To answer the second research question, the results of the frequencies of the different categories of teachers' feedback in the main area focus are illustrated in Table 1. Feedback analysis revealed that all three teachers concentrated mainly on Language Accuracy with 47% of the comments in that category. Within Language Accuracy general comments about language were the most common at 24%, and on Grammatical Structures were 14%. Other subcategories in language accuracy get much less attention. Concerning WCF, there was a difference between teachers. Teacher 1 used the lowest number of error corrections (N = 76), while Teacher 2 used twice as many error corrections (N = 130), and Teacher 3 used almost three times as many as Teacher 1 (N = 199). Comments in the textrelated categories such as Content and Organization were only half compared to Language Accuracy. Another observation was the difference between teachers giving feedback on content and/or giving general comments. Teachers 1 and 2 gave a comparatively high degree of feedback on content and wrote detailed comments. In contrast, Teacher 3 wrote mostly general comments on writing. Teacher 1 did not give feedback that could be classified as Reinforcement of Learning Materials, Teacher 2 did not focus on Mechanics or Learning Materials and Teacher 3 did not Reinforce the Learning Outcomes. All teachers gave almost the same amount of feedback, which was classified as belonging to the category organization. To conclude, the teachers displayed both similarities and differences in the type of feedback they gave their beginner level students. The results in Table 2 illustrate the types of feedback given by the teachers in the main area manner. All three teachers gave feedback mostly in the category of Information (46%) and these were nearly all in the form of Statements. About 33% of the feedback was classified as Suggestions and two thirds of these were in the form of Imperatives. While Teachers 1 and 3 used four to seven times more Imperatives than Statements, Teacher 2 used both Statements and Imperatives in similar proportions. The proportion of data units for Specific Praise was 8% of the total feedback on praise 12%. There were also individual differences between teachers regarding Praise. Teacher 2 rarely used praise (4%) compared to teachers 1 and 3, and Teacher 3 used specific praise more often (15%) than both Teachers 1 and 2. Even Politeness was used more by Teacher 3 (11%) compared to the other teachers. In the same way as for the focus categories, the manner of feedback for all teachers was similar, such as informing and suggesting, but they also chose different ways to communicate with their students. In sum, to answer the second research question, for all three teachers most feedback was provided on Language Accuracy in the main area, focus, and dealt mostly with general comments on writing. Regarding manner, almost half of the feedback was given in form of Information and Suggestions and the form of Imperatives were used relatively frequently. There were considerable individual differences in the proportions of feedback provided by different teachers in both focus and manner.

Discussion and Suggestions for Further Research
The present study aimed to develop a model for analyzing teacher feedback and was carried out in an SLL beginner level context. In the literature there are very few studies about feedback on adult language learner at the beginner level. This section first discusses the results of the first research question concerning the categories generated in the model. Thereafter, the second research question, regarding the application and the modification of the model, is discussed. More specifically, the degree of feedback in the different categories is considered. This section ends with concluding remarks.

Categories Generated in the Model
In the main area, focus, the findings showed that both text-related (content, organization) comments, as well as comments on language aspects were observed as in studies on more advanced language learners (e.g., Lee 2009). Following Hyland and Hyland (2010), meaning and form are closely linked together as writing development not only deals with text-related aspects but also other aspects. Based on the fact that teachers mostly provided text-related feedback, the results of the present study suggest that beginners may need most help with text-related aspects of writing. However, content and organization can be seen as natural categories not only in L1 but also in L2 and the teachers may expect a relatively high level of writing skills from highly educated beginner language learners.
The identification of subcategories within Language Accuracy is the first contribution of this study. Unlike previous research (e.g., Junqueira and Payant 2014), this category was divided into five subcategories in the present study. These subcategories create a picture of the specific aspects of Language Accuracy needed for beginner level adult students. As feedback for beginners in second language research is an unexplored area (Jakobson 2018), it would be necessary to examine the significance of the feedback with regard to language accuracy concerning beginners written texts in Swedish as well as in other languages.
The findings also suggest that Reinforcement of Learning Materials may help students to improve their learning strategies (Hyland 2001). As this category has occurred in few studies (Jakobson 2015;Hyland 2001), more empirical support is needed in the context of adult beginners. Furthermore, it is an interesting observation that Reinforcement of Learning Outcomes was not a separate category in previous studies. However, this focus category can be seen as an institutional factor, as the course outcomes are developed by institutions. Which institutional norms, for example, regarding school policies (Lee 2008), affect teachers' feedback in SSL as well as in other languages might be an area to investigate in the future.
In contrast to focus, the subcategories in the second main area, manner, was not well-researched previously (Ferris 2003) and the development of manner is a further contribution of the current study. The results in the main area manner showed that five categories were identified in this area and one of them, Politeness, was a new category. In contrast to previous studies, Praise was divided into two subcategories, Specific Praise and General Praise. A further two subcategories: Ask for Information about the Content and Criticism also emerged from the data. Those subcategories were not inductively developed in previous feedback research in a second language context and would, therefore, be important for further investigation. Criticism and Politeness are discussed further below.
Concerning Criticism, an interesting question for further research would be how negative comments can be interpreted by different teachers, for example, whether they see them as Criticism or Information, because a negative comment seemed to have different aims for different teachers in this study. Teachers may have different intentions with critical comments and those comments can be understood differently by the recipients. In the present study, the teachers did not use general criticism which could be interpreted as criticism aimed at a learner. As criticism is a very complex category of interpersonal feedback along with praise and suggestions (Sutton et al. 2012), general criticism and specific criticism in writing contexts is an area for future studies.
As Criticism was not divided into Specific and General Criticism in previous research (e.g., Hyland andHyland 2001, 2010), the concept of criticism and what the teachers' intention is regarding negative comments might be an issue for further investigation. The results indicate that criticism is one of the central components of feedback manner, and further research should be undertaken to investigate critical comments from teachers and the students' perspectives on these comments in different contexts such as Swedish as well as other second or foreign languages.
Politeness as a contextual and cultural concept needs further research. As with Criticism and Praise, teachers can signal with the comments not only their personal attitude towards the students, but also the importance of students both learning and knowing the sociolinguistic norms in order to adapt their language use and understand the norms of politeness when communicating with the recipient. Politeness theories and norms could be an interesting area for continued research in feedback contexts.
As stated previously, feedback as interpersonal communication can be given in a variety of ways by using different categories and types of feedback, which create a specific teacher-student relationship (Elbow 1999;Hyland and Hyland 2010;Lunsford and Straub 2006;Sutton et al. 2012). Further studies are needed to examine whether and how individual, contextual and pedagogical factors influence teachers' choice of comment types and manner. Further work should also involve the investigation of manner in different contexts in order to better understand the complexity of feedback.

The Amount of Feedback Given
Concerning focus, all three teachers concentrated on language accuracy in their feedback, which was also the case in studies with intermediate and advanced level language learners (e.g., Junqueira and Payant 2014;Lee 2009;Lee et al. 2018). Regarding language accuracy, the teachers focused particularly on grammar and language in general. Regarding spelling, the teachers in the present study gave less attention to spelling compared to a previous study on feedback in which a teacher of adult beginners with a similar background to this study, gave a substantial amount of feedback on spelling (Jakobson 2015). The focus on language accuracy in the present study can be explained by the students' levels of proficiency, that is, adult beginners. Feedback on language accuracy is one of the central aspects of the beginner classroom and continuous feedback may help the students to gain a thorough grounding on which to base their further studies. Beginner level students may need more support with feedback regarding language, as they simultaneously learn grammatical structures, vocabulary and other language and writing-related skills over a short intensive learning period. Additional research is required to determine possible explanations for teachers' focus on language related categories and students attitudes to teachers' feedback within the different text-related categories. The teachers also vary the way in which they write comments with regard to their statements, questions and imperatives in a comparable way to the results in studies by Ferris et al. (1997), Ferris (1997), and Sugita (2006). Further research is needed to examine how linguistic and semantic features can be directed to the students, which may help the students to understand teachers' comments and make use of the feedback to improve their writing.
Even though there were similarities among the teachers' practices, as mentioned before, there was a variation between teachers concerning the proportion of feedback they gave on categories such as content, grammatical structures, and general comments in the main category focus, as well as in the use of statements, specific praise, and politeness. The teachers gave much less praise and criticism compared to information and suggestions. These findings are interesting as they contradict the findings in Hyland and Hyland (2001). The contextual factors, for example teachers' individual attitudes, may partly explain dissimilarities in the feedback between different studies and the individual teachers (Goldstein 2016). However, as suggested by Hyland (2001) and Straub (1996), a carefully balanced combination of praise, criticism and suggestions could enable better relationships to be established between teachers and students, whilst at the same time giving support to students in developing their independent writing. Therefore, it is important to pay attention to how teachers analyze and reflect on the relationship between praise and criticism. One possible issue might concern how much control the teachers wanted to have with their comments in the form of criticism and imperatives, and how much they wanted to encourage and motivate the students to make their own decisions, while supporting them with suggestions in the form of questions and statements (Lunsford and Straub 2006). Variations between teachers' attitudes can explain differences in feedback practices (Ferris 2010). Furthermore, providing feedback is influenced by a "myriad of factors" (Conrad and Goldstein 1999, p. 175) and this variety of contextual factors and teachers', as well as students' characteristics, which interact with each other, affect feedback practice.
Further research is needed to determine and compare the proportion of teacher feedback on language accuracy compared to other text-related categories with the aim of finding possible explanations for differences, for example, by investigating teachers' beliefs. Different manners of feedback affect the students' reactions (Ferris 1997;Sugita 2006) and, thereby, the acceptance of the comments or the refusal to take them into account, which will affect learning (Wiliam and Thompson 2007). Therefore, based on their awareness of the importance of feedback, teachers' choice of feedback regarding what and how the comment and quantity of feedback can be adapted for the needs of different students. Future feedback studies should explore teachers' choices of feedback specifically for beginner learners because the feedback decisions made by the teachers can help to assist the language development of students on their way to becoming independent language users -the ultimate aim of language learning.

Concluding Remarks
This study aimed to develop a model of feedback from a holistic perspective by identifying categories and subcategories in the main area, focus, and in the main area, manner. In line with previous studies where the majority of categories were developed using feedback on intermediate and advanced level language learners, these same categories also appear in the present study within the context of beginner level learners, and additional categories have been identified. A wide range of categories, both in manner and focus, provides additional evidence that the concept of feedback is a combination of different interacting aspects in different dimensions (Bitchener and Storch 2016;Goldstein 2016), which contributes to the complexity of the feedback. Intending to apply the model, the second question in the study concerned the extent of the different feedback categories in focus and manner. The teachers in this study varied their types and the amount of feedback. This variation indicates that the teachers dealt with several aspects simultaneously while commenting on a specific text.
The division of feedback into two different main areas (manner and focus) in the L2 context, specifically in SSL, is the main contribution of this study. However, the results of this study and previous studies (except Lee 2009) are based on data from very few informants and, therefore, provide only an initial picture of teachers feedback practices in the L2 context. Therefore, there is a need to broaden and deepen future feedback research with a more considerable number of participants.
The evolving model for analyzing feedback from a holistic perspective in a Swedish as second language (and other L2) context requires further investigation since only a few studies have investigated feedback from the holistic point of view (e.g., Hyland and Hyland 2010). The model in the present study is not definitive and can be developed and adapted in other areas such as teachers and students perceptions of feedback in SSL, as well as in other teaching contexts such as feedback provided to students with limited educational backgrounds. One model for the analysis of feedback can never be created in research because feedback is context-dependent and extremely complex. An overarching question for further research is how the model can be further developed in beginner contexts and for other language proficiency levels in second languages both in asynchronous and synchronous online contexts and in physical classrooms.