Investigating conference feedback in Cambodian EFL class: students’ writing accuracy improvement and their views of the feedback

Research of conference feedback on EFL students' writing accuracy is a vital issue; however, it has never been conducted in Cambodia. This study was carried out to fill this void by investigating conference feedback to improve students' writing accuracy in four categories: semantic, grammatical, mechanic, and lexical features. It also explored the students' perceptions of the feedback in the context of EFL writing class. One class, with twelve students, from 18 to 25 years of age, with a ratio of 8:4 women: men, was purposively chosen for the study which was designed in two cycles, each with eight steps. Data were collected from the writing tests, writing assignments, students' self-evaluation survey, and focus-group discussions. Findings revealed that conference feedback had improved students' writing skills by allowing them to use correct grammar and expressions. The feedback has also developed students' engagement in the writing process. It also proposed a pedagogical implication that effective feedback through a conference approach depends to a great extent on the knowledge and experience of the teacher, classroom resources, and interactions between the teacher and the student.


Introduction
Research on the effects of conference feedback has been a vital area (Anast-May et al., 2011;Hyland & Hyland, 2006), as it is undertaken with a "one-on-one" dialogue (Al Noursi, 2014), or a discussion between teacher and students to improve a piece of writing (Boggs, 2019;Hyland & Hyland, 2006). Although several researchers, for example Demirel and Enginarlar (2016), Ho et al. (2020), Huisman et al. (2019), Klimova (2015), Angel et al. (2017), Nurie (2020), Sotoudehnama and Pilehvari (2016), and Yamalee and Tangkiengsirisin (2019) argued that any kinds of feedback helps students become better writers in class and beyond, conference feedback has a distinct feature, as a process-oriented technique, for enhancing students' writing abilities to improve language accuracy and meaning in the written texts (Atai & Alipour, 2012;Chandler, 2003;Ekmekci, 2015;Lerner, 2005;Sotoudehnama & Pilehvari, 2016). In addition, this feedback focuses not only on teacher's feedback but also on teacher-student interactions along a writing process (Eckstein, 2013;Lerner, 2005) which helps develop students' writing skills and speaking abilities (Hyland & Hyland, 2006;Nosratinia & Nikpanjeh, 2015). However, the study on the effects of conference feedback on Cambodian students' EFL writing has never been conducted in Cambodia although it has been widely recognized (Eckstein, 2013;Klimova, 2015;Nosratinia & Nikpanjeh, 2015;Sotoudehnama & Pilehvari, 2016) as an effective strategy for the writing classes. Without any preliminary studies, the teacher-researcher conceptualized conference feedback as a conversational-process design with a clear purpose, structure, and role for the teacher and the students to work on improving the accuracy of the written texts.
The accuracy in writing is referred to as the ability of students to produce sentences without errors (Hammerly, 1994;Richards, 1971); however, the measurements of accuracy have been ambiguous, dynamic, and complex (Kaplan, 1966;Polio, 1997;Polio et al., 1998;Polio & Shea, 2014). For instance, Polio et al. (1998) suggested three measurements: holistic scales, number of error-free units, and number of errors, with or without classification, while other scholars proposed other four types: contrastive analysis (CA), error analysis (EA), interlanguage analysis (IA), and contrastive rhetoric (analysis) (CRA) (Richards, 1971). A study of Chandler (2003), for example, adapted a Hum Investigating conference feedback in Cambodian EFL class: students' writing accuracy improvement and their views of the feedback Journal on English as a Foreign Language,11(1), 104-124 p-ISSN 2088-1657; e-ISSN 2502-6615 106 measurement of Azar (2003) that encompasses various variables: singularplural form, word form, word choice, verb tense, add or omit a word, word order, an incomplete sentence, spelling, punctuation, capitalization, article, unclear meaning, and a run-on sentence, while other studies (Lahuerta, 2018;Shirotha, 2016) were conducted to explore the complexity of accuracy while concluding that the accuracy can be improved by using different feedback types: directive feedback, non-directive feedback, peer feedback, group feedback, and/or a combination of various types. Such complexity; however, (Chandler, 2003) suggested that various measurements used in the accuracy research may generate different effects on improving writing quality. To this end, researchers should be mindful when replicating a method from another study; otherwise, the results and interpretations may be discredited. To avoid such issues, researchers should define specific error categories for the analysis.
The error categories were conceptualized in various ways. For example, local errors cover minor linguistic elements such as nouns, verbs, prepositions, and auxiliary verbs, which have a trivial effect on messages in communication, while global errors compose of sophisticated components (i.e., word orders, sentence structures) that can disrupt the understanding of messages (Touchie, 1986). Other error categories are referred to as "performance errors and competence errors" (Hammerly, 1994;Richards, 1971). They explained that "performance errors" are superficial and explicitly recognizable and occur when language users were tired or hurried, while "competence errors" are implicit and serious. In this light, Clouse (2012) proposed twelve types of errors ( e.g., word choice, sentence fragment, run-on sentence, verbs, pronouns, modifiers, punctuation, capitalization, spelling, abbreviation, and numbers), whereas Butler (2014) categorized errors into group 1: capitalization errors, plural, spelling mistake, missing word; group 2: wrong word, wrong word order, subject-verb agreement, and group 3: punctuation errors, verb tenses, verb forms, sentence fragments, and run-on sentences. Such complexity of errors has caused ambiguity in conceptualizing a framework for the research.
Being aware of the ambiguity in error conceptuality, researchers, for example, Atai and Alipour (2012), Polio (1997), and Kim (2012), suggest that researcher's awareness of the reality in the context and the targeted population is the key factor for conducting a study of language accuracy. This means that the researcher should have a good understanding of the curriculum, the materials, and the language levels the students are taking. Based on the contextual knowledge and experience as the teacher in the selected center, only nine errors classified into four categories [Grammatical features: Subject-Verb The nine errors were identified as the most common mistakes made by the targeted students while also being noted as the most frequent errors made by non-native students (Chandler, 2003;Kim, 2012). To minimize the ambiguity, global errors or competent errors such as clauses, conditional sentences, fragments, run-on sentences, and parallelism were excluded from this action research study.
Action research (AR) has been a popular method for decades in language research (Burns, 2005(Burns, , 2019Koshy, 2005). AR is a systematic and cyclical inquiry for observing, identifying, analyzing, and constructing knowledge of what is happening in classes and what the teacher should do for better changes (Koshy, 2005). It was initially used by social psychologists to observe social and political phenomena; it was later utilized by liberationists to examine social, political, and economic constraints (Burns, 2005). More recently, AR has become a multidisciplinary approach in language education research (Burns, 2019;Koshy, 2005). Therefore, AR has become an empirical quantitative and qualitative inquiry that researchers have applied to examine students' writing accuracy in various contexts (Alvira, 2016;Cavkaytar & Yasar, 2010;He, 2015;Sapkota, 2013;Wang, 2016). In this essence, however, AR has rarely been carried out by Cambodian EFL researchers.
Very few studies in Cambodia have focused on EFL writing, let alone the use of action research project. Six years ago, one study by Chan (2015) utilized an AR, but it aimed to integrate EFL students' social awareness to promote reading and writing skills. Another study by Chea and Shumow (2015) attempted to investigate the issues of Cambodian students' writing; however, the study purpose was to measure self-efficacy, writing goals, and writing achievement, while the method was in a quantitative survey. Only Sou (2010), who was a decade ago, conducted a more relevant study to improve EFL writing; however, Sou (2010) tried to compare directive and facilitative feedback strategies effectively. Overall, no studies have been done on the focus of improving ELF writing accuracy through an AR project. Thus, this study was carried out to fill this void, and in response to a call from Chan and Sotith (2016) that Cambodian EFL students often achieve a poor performance in writing, both in grammatical structures and language use, despite having received formal writing instructions over many years. They suggested that this problem needs to be further explored. Therefore, this study was designed with an action research project to seek answers to the following questions.
(1) How does conference feedback help reduce students' grammatical errors and increase the expression of precise meanings? (2) What are the perceptions of the students towards the effects of conference feedback in improving their writing accuracy?
The findings from these questions would significantly contribute to the improvement of EFL writing classes in Cambodia while also proposing pedagogical knowledge for better teaching practice. The study would also be significant to fill the literature gap in Cambodia's EFL context. Especially, the study could promote EFL research culture in Cambodia.

Research design
This study applied action research (Burns, 2016;Koshy, 2005) that has been recognized as an effective method that allows researchers to develop appropriate intervention and bring an improvement to teaching and learning quality. Guided by the theoretical frameworks as discussed in the literature, this study was conducted in two cycles of eight steps.
In the first cycle, the researcher assessed the English curriculum practices and evaluated instructional materials and school environment to identify issues and start with the pretest and self-evaluation survey. Data collected in this initial stage helped the researcher formulate the study assumption that 'the oneon-one feedback between teacher and student helped improve the accuracy of the written works'. To test the assumption, an intervention was introduced to the class.
To systematize the intervention, the conferencing schedule, steps, and questions were designed and presented to the students. The students were also told that the purpose of this feedback was to improve writing accuracy. After an orientation, conferencing sessions that lasted between 3 to 5 minutes were undertaken. The first two conferencing meetings were mainly aimed to help students improve the meaning (overall content), while the later conferencing sessions helped students to attend to the accuracy of grammatical features and mechanics. At the end of an intervention, the students edited their works and handed in their final drafts for the teacher-researchers evaluation. The results from this cycle helped the researcher to adjust and reformulate the study design in the next stage.
The second cycle of the study aimed to follow up on the previous assumption. To do so, the second writing assignment was given to students, and the students went through the same writing process. However, the students were assigned to confer with a classmate: one student was selected from a high-language group; another was invited from a low-language group (based on the pretest and assignment in cycle 1). The numbers of meetings for each pair were the same as cycle 1. However, the conferencing period was extended from 8 to 10 minutes, giving students more time to take turns (for 5 minutes each), ask questions, and give responses. To re-evaluate the tested variables, a second assumption was formulated that 'the conferencing feedback between student and student help them to identify specific grammatical and language errors, and that they could work on more effectively to revise their written works'. At the end of this cycle, second final drafts were collected and evaluated.

Participants
This study was conducted for more than three months in one selected Language Skills Center in Battambang, Cambodia. The Center was chosen for its (30-year) prestigious services in providing English language education to thousands of Cambodian students. Uniquely, the Center has a systematic program that all classes meet under an arranged schedule: one hour per day, five days a week. For the selected class, two textbooks: Quest 3 Reading and Writing  and Quest 3 Listening and Speaking (2rd Ed.) , were used for daily teaching. Additionally, a Writing Project module was introduced as a guideline for helping the students to write better paragraphs and essays. To ensure consistent teaching, the Center also conducts a two-hour meeting every Saturday morning that teachers may help design lesson outlines and extra teaching materials for class activities.
The twelve students in one class were purposively selected based on the following rigors. First, these students had a similar learning background and attended the same language program (the same level, coursebook, learning assessment, and study duration). Second, the class size was suitable: eight of them were female and four were male, while students' ages were similar; one group was high-school students, aged below 20, while the rest were freshmen, aged over 20. The third reason, these students' writing achievement varied; for example, one group got writing scores from a very good (A) to excellent grade (A+), and another one obtained grades of average (C) to below average (C-) (based on the test scores obtained from the Center). With such variations, the results from this research may be plausible to explain the issues of students' writing skills.

Data collection
The data collection was undergone in multiple steps. Before the intervention, a 45-minute writing pretest and a 15 minute-self-evaluation survey were used to diagnose students' writing skills. During the intervention, the data were extracted from students' final drafts (two assignments) in cycle 1 and cycle 2. After the two cycles, a 45-minute posttest and a 15-minute self-evaluation survey were used to measure the improvement in students' writing accuracy in relation to the intervention. In the end, the researcher conducted oral interviews with two groups of students to understand their perceptions towards the issues arising during the conferencing feedback.

Data analysis
The data analysis was carried out in several steps. To ensure the analysis consistency, error codes and classifications were critically done. To do so, the researcher repeatedly read and systematically evaluated the students' written works over the study period. The analyses were conducted through the research process through the pre-test to the post-test. In total, 48 written texts (12 x 4 = 48 written texts) were evaluated and rated by the teacher-researcher. All of the written texts were retyped in Microsoft Word for word counts to bolster the reliability, and each error category was read and marked by using error codes (see: Table 1). The teacher-researcher then invited a native-English speaking research mentor to give feedback on error identifications.
To control the consistency of error counts and to ensure no overlapped counts, each error frequency was counted separately and arranged into four superordinate categories for recording (see : Table 1). After completing this process, the researcher took a few weeks to double checks of the data accuracy in order to make sure there were no missing errors in the record. Once consistency was ascertained, the researcher computerized error frequencies in Microsoft Excel, and the calculation of errors was conducted accordingly to each error type. The calculation formula was (number of errors for a given type x 100/total number of words in that text). The final process of data analysis was to compare the mean differences of errors between the pre-test and the posttest, and between the assignment 1 and 2. The improvement in students' writing accuracy was then measured, while discussion and conclusion were made whether conferencing feedback improved the students' writing abilities or not. To extend the discussion, the self-evaluation survey data were analyzed to obtain the patterns of students' perceptions and to evaluate their written works after they had gone through the intervention. Finally, the data from the focus group discussions were summarized to support statistical indications, while discussion and conclusion was made to answer the research questions.

Conference feedback reduces grammatical errors and increases expressions of precise meanings
The results in Tables 2 and 3 responded to research question 1 (RQ.1) that conference feedback has positive effects on reducing grammatical errors and increasing expressions of precise meanings. As evident, before the intervention, the students made the most noticeable errors in two categories: Semantic features (unclear/incomplete-meaning?) and Lexical feature (wrong word choice (WWC). The mean error in "word choice" was 4.9, while the mean in the "Unclear-Meaning Sentence (?)" was 4.3 per 100 words. Comparing the means of these errors in the pretest with the means of errors in the posttest revealed a large difference, decreasing from 4.9 to 1.7 in 'WWC' and from 4.3 to 0.6 in '?'. Decreasing the number of errors showed that the student had improved their abilities to write grammatically correct sentences with appropriate word choice to express precise meanings.
In comparison, the mean differences from the pretest with the posttest (as presented in Table 2) indicate that the students improved their language accuracy in all tested items, mainly decreased from 4.9 to 1.7 in the "WWC" and from 4.3 to 0.6 in "?". In addition, the number of errors reduced to smaller, as can be seen in items (Sp.), (P) and (C) items, ranging from 0.1 to 0.3, respectively. This means that the students are able to identify and correct errors through the feedback intervention. To confirm this assumption, the data collected from two writing assignments were reported in Table 3.  Comparing the mean errors made by students in the two-assignment revealed that the number of errors in "Lexical feature" such as in "WWC" slightly declined from 4.5 to 1.7 per 100 words, while the mean in "Semantic errors" as in the item of "Unclear-Meaning Sentence (?)" dropped from 4.3 to 2.4. However, other error categories remained the same, suggesting that the students made little progress from the first round of conferencing feedback. The teacher-researcher, in the second round of conferencing feedback, decided to let the students work with their classmates in pairs and groups. The students' writing abilities in the second round of intervention seemed to be remarkably improved. As indicated by Table 3, the number of errors in most tested variables, except "WWC" largely decreased. In addition, individual students seemed to have improved their writing accuracy, minimizing errors to a smaller mean to below 1.0 per 100 words. To confirm the effectiveness and explore the feedback intervention issues, the teacherresearcher invited the students to give their responses to a self-evaluation survey (see Tables 4 & 5).

Positive views of the conference feedback effect on the writing improvement
In addition, the student's responses to the self-evaluation survey, as presented in Tables 4 and 5, revealed their positive views of the conference feedback effects on the writing improvement (RQ.2). Before intervention, the students' responses were in a high percentage on a 'neutral' option. For example, 33% of students selected this option for statement 1 and 50% for statements 4, 7, 8, and 9 (Table 4). Their responses to these statements indicated that, of the majority, the students were not able to, or be reluctant to, evaluate their language accuracy. In contrast, the students chose the "agreed" option to a few statements; for instance, 75% on statements 2 and 3, and 67% on statement 4. These responses were indicative of students' perceived responses to their writing abilities developed after the intervention, as some of them agreed that they could do better writing on grammatical structures such as subject-verb agreement, the use of the pronoun, and verb forms".
The student's responses to the self-evaluation survey, however, changed after they went through the intervention and after they took the posttest. As presented in Table 5, 75% of them, after having participated in the two cycles of conferencing feedback, emphasized that "they could write grammatically correct sentences." Likewise, the students shifted their viewpoints from mostly "neutral" options to mostly "agreed or strongly agreed" responses. This indicated that the students felt they had mastered the language items tested in this study. The evidence was confirmed by the students in statements 4, 6, 7, 8, and 9. This indicated students' positive perceptions or evaluation of their writing abilities after they were involved in the feedback intervention. To support this indication, students were invited to focus group discussions, and the results were discussed.  The data obtained from the focus-discussion group indicated that the students had positive and negative perceptions of the feedback intervention on improving their writing accuracy. The positive views perceived by the highlanguage student group were summarized below: When we talked about our written works with the teacher and classmates, we felt relief since we were able to correct grammatical With these responses, it can be inferred that the high-language accuracy students tended to perceive the important role that conferencing feedback played in their written works.  In contrast, the low-language proficient students seemed to feel less appreciative of the practice of conferencing feedback in classes. As they emphasized in their responses: We felt nervous, inconvenient, and embarrassed (shy) to talk about or to share our written works with friends because we don't want others (classmates) to see our mistakes. This response indicated that the students who had poor or moderate level of writing skills tended to be less involved in the practices of conferencing feedback due to their abilities to attend to specific language errors as well as to come up with effective feedback for improving the written texts of their peers or groups.

Discussion
The extant literature argued that conferencing feedback plays a key role in improving EFL writing. The premise of this feedback is to provide a reign to the teacher and the students to work more interactively and independently since they have an adequate amount of time to incorporate feedback and to polish errors in the texts (Al Noursi, 2014;Atai & Alipour, 2012;Hyland & Hyland, 2006;Yamalee & Tangkiengsirisin, 2019). The findings of this study, therefore, reaffirmed this premise by showing that the students' writing accuracy in the tested items: Grammar, Lexical, Mechanic, and Semantic features have been improved from a moderate to a greater level. The decrease in errors, as revealed in this study, can be a result of the effect of feedback intervention based on some factors.
The first effective factor may relate to a clear goal and structure for the conferencing feedback, as it was aimed at helping students negotiate meanings before directing their attention to specific error types in the later stages. In line with this argument, other researchers, for example, Sotoudehnama and Pilehvari (2016), supported that students who were motivated to improve the meanings in the first step tended to produce better revisions in the next stages. Drawing evidence from the present study (see : Tables 2 & 3), Semantic and Lexical errors were initially higher than other items; however, these errors were minimized along with the decline of errors in other categories. Supporting this discussion, Sou (2010) ascertained that Cambodian EFL students who had gone through facilitative-process feedback improved their written works because they were more engaged through the writing process. The improvement in students' writing skills in this case study has been recognized to be a result of a combination of the action research project and the feedback intervention. As one study in Nepal (Sapkota, 2013) argued that after students participated in an action research project, they improved writing accuracy and were able to solve other writing problems.
Another factor contributing to the improvement of students' writing accuracy may result from the atmosphere of a conference conducive to learning and doing the writing. As asserted by the students, "The conference gave us enough opportunities to exchange ideas and experience in writing, and thus we felt relief to discuss our writing problems with teacher and classmates." This statement was supported by Al Noursi (2014), Eckstein (2013), Hyland and Hyland (2006), and Lerner (2005) who claimed that conferencing feedback is not just a platform for improving writing, but it also provides greater opportunities for students to plan, reflect, discuss, clarify, negotiate, and correct mistakes in the text. It is true that this action research study, combined with conferencing feedback, allows students to more easily examine their written works with peers and teachers so that they can attend to language errors and meanings in the texts.
The final factor of conferencing feedback on improving writing accuracy is that the students seemed to benefit from peer engagement. Although this study did not attempt to compare the effectiveness of teacher-student versus student-student conferences, the students' perceptions through the focusdiscussion groups suggested that they were more interested in sharing problems with their classmates than discussing problems with the teacher. Supporting this argument, researchers such as Huisman et al. (2019) and Lerner (2005) ascertained that conferencing feedback was constructive in promoting students' engagement in their writing. It is also a venue for students to think about their works (Eckstein, 2013) as well as to thrash problems out for better revisions (Boggs, 2019;Ho et al., 2020). The combination of conferencing feedback and action research in this case helps reinforce the teacher's understanding of writing lessons, the teaching process, and desired objectives set for a program. Through this vein, the teacher and the students are more engaged in the writing process, which is the most important strategy in classes.
The findings from this study, however, suggested some issues with conferencing feedback. Through the interview responses, the low-achieving language students felt less appreciative about their involvement in the conference, as they sometimes felt it was stressful and inconvenient to ask questions, to give their responses, and to talk about their mistakes. Nosratinia and Nikpanjeh (2015) emphasized in this light that the low-proficiency writers work more effectively with written feedback than other feedback types since the students need to identify specific errors for improvement. Conference feedback thus should be combined with other feedback techniques, for instance, a portfolio (Atai & Alipour, 2012), a reflective inquiry (Bitchener et al., 2005;Ferreira et al., 2007;Ferris, 2007), and/or collaborative feedback (Nosratinia & Nikpanjeh, 2015) to enhance the students' writing accuracy. A combination of other feedback approaches help mixed-level ability students improve content and language accuracy in the texts (Bitchener et al., 2005).
Recognizing this pitfall, conferencing feedback may be challenging to the practice in a mixed-level ability class such as in the case study because the lessproficient language students tended to give less value to it (Eckstein, 2013); therefore, the students in this group were less interactive than those students of greater language proficiency. In line with this, (Touchie, 1986) advised that cultural and social issues tend to be the most influential factors for the effectiveness of the conferencing process. This argument ties well with the study of Chan (2018) that English education in Cambodia has been constrained by the settings of English learning that can influence students' writing selfefficacy and writing mastery goals (Chea & Shumow, 2015).

Conclusion
The finding of the present study lent itself to a theoretical premise argued that conference feedback, whether the teacher-student or the student-student pairs, is an effective teaching strategy for helping students to attend to specific areas of language accuracy and express precise meanings in the texts. This finding thus supports the study assumption. On the other hand, conference feedback seems to be a challenging approach for multi-level language learners in writing classes since the low-proficiency language students may be less engaged in a peer-conferencing process/feedback. This highlights the issues during the pedagogical practice of feedback, which this study aimed to discover from the students' perceptions and interview responses. From this evidence, the effectiveness of conference feedback depends partly on the teacher's ability and flexibility to motivate students, moderate the questions, and facilitate roles and activities during peer and group conferencing. This suggests an implication that the teacher plays a vital role in ensuring that the students gain equal benefits from this feedback approach. Being more caution, the teacher's pedagogical knowledge should be considered to fit into the goal, scope, and structure to ensure the consistent process of feedback and responses.