Studies in Second Language Learning and Teaching High school EFL students’ beliefs about oral corrective feedback: The role of gender, motivation and extraversion

This study employed an explanatory sequential mixed-methods research de-sign to examine the beliefs of Vietnamese EFL students concerning oral corrective feedback (CF) and the role of some individual differences in these beliefs. The data consisted of questionnaires completed by 250 Vietnamese high school students and follow-up interviews with 15 of them. Exploratory factor analysis revealed six latent factors underlying students’ beliefs about CF, namely, (1) output-prompting CF and eliciting recasts, (2) desire for CF, (3) non-verbal cues, (4) important errors, (5) input-providing CF, and (6) less important errors. Descriptive statistics and thematic analysis of the interviews showed that students were positive about CF. They liked both input-providing CF and output-prompting CF for all error types. Metalinguistic feedback was the most strongly preferred, while clarification request was the least preferred. Further statistical analyses revealed some interesting relationships between students’ beliefs about CF and their gender, English learning motivation, and self-rated introversion/extraversion. Females were more positive about CF than males, and extraverted females were more positive about input-providing CF than introverted females. Also, students learning English for exams were more positive about CF than those learning English for communication. Pedagogical implications for effective feedback provision in EFL contexts are discussed.


Introduction
Studies of learner beliefs are essential in second language (L2) research. Like other individual differences (e.g., motivation, working memory and language aptitude), learner beliefs are significant learner individual differences which can influence both the process and outcome of language learning (Dörnyei, 2005;Ellis, 2008). Beliefs about oral corrective feedback (CF) 1 (responses to learners' erroneous utterances) merit increased research attention because such research can shed light on belief in/congruence between students and teachers, and can thus help us to understand and to enhance the effectiveness of CF. As noted by Pawlak (2014), lea rn e r be li e f s , e xp ec t atio ns and p r e f e r en c e s r eg a r din g t he p r o vi s io n o f C F "should indubitably be taken into account if foreign language pedagogy aspires to be learner-centered and the guidelines furnished by scholars are to stand the chance of being transformed into actual instructional practices" (p. 69).
Among various aspects of CF research, CF beliefs have received the least research attention (Akiyama, 2017;Ha & Murray, 2020). Most of the research concerning learners' CF beliefs has included only a few questions probing students' opinions about the usefulness and necessity of CF as part of larger studies investigating other topics, predominantly beliefs about grammar instruction (Jean & Simard, 2011;Loewen et al., 2009;Schulz, 2001). Research looking extensively at learners' CF beliefs is limited (Akiyama, 2017;Li, 2017), but understanding the phenomenon is very important (Kim & Mostafa, 2021). Firstly, learners' CF beliefs can impact the noticeability of CF (Kartchava & Ammar, 2014) as well as its effectiveness (Sheen, 2007). Secondly, it has been noted that learners view CF and grammar instruction as two distinct categories (Loewen et al., 2009), suggesting that investigating learners' CF beliefs as a separate construct is necessary. Thirdly, learners' CF beliefs have been found to be contextspecific (Loewen et al., 2009;Pawlak, 2011;Schulz, 2001), dynamic and complex (Akiyama, 2017;Leontjev, 2016), and trainable (Sato, 2013). Moreover, as noted by Ellis (2009Ellis ( , 2017, both second language acquisition (SLA) researchers and language educators have frequently disagreed about whether to correct errors, the choice of errors, CF types, and CF timing. Understanding learners' beliefs about these aspects of CF, and the individual and contextual factors that influence them may help to resolve such disagreements and enable a more nuanced and tailored approach to classroom practice. Therefore, there is a pressing need for more research on learners' CF beliefs in various contexts.
The present study contributes to CF research in several ways. Firstly, while learner CF beliefs have been found to be influenced by contexts (Schulz, 2001), no studies have been conducted in any Asian high school EFL contexts, which involve a vast population of language learners. In order to provide one missing piece to complete this puzzle, this study aims to provide insights into students' beliefs about CF in Vietnamese high school EFL classrooms, a hitherto unexplored context. Secondly, while it is well established that understanding learner individual differences is important, questions remain about the interactions among those variables, particularly in the field of CF research. Learner individual differences such as gender, extraversion, and motivation for learning English may play a role in learner CF beliefs. This study is the first attempt to investigate some possible relationships between students' CF beliefs and these three variables.

Students' beliefs about oral corrective feedback
As mentioned earlier, most previous research on students' CF beliefs has included CF as part of broader research agenda concerning grammar instruction. Thus, these studies usually included several questions pertaining to CF in a more general questionnaire about grammar instruction. They have, however, provided some useful insights. Overall, students were found to be positive about CF (Brown, 2009;Davis, 2003;Jean & Simard, 2011;Li, 2017;Loewen et al., 2009;Schulz, 1996Schulz, , 2001, but the extent to which students desired to receive CF varied in accordance with the learning context and the prior language learning experience (Loewen et al., 2009). For example, in Schulz's (2001) large-scale study with post-secondary students, foreign language (FL) students in Columbia tended to be more positive about CF than FL students in the US. Jean and Simard (2011), in a large-scale survey with Canadian high school students, found that a higher percentage of English as a second language (ESL) students felt that grammar errors should be corrected all the time during speaking lessons than did French as a second language students. They ascribed this discrepancy to the students' first language learning experience in their language arts classes. Agudo (2015) replicated Loewen et al.'s (2009) large-scale study with 173 Spanish high school EFL students, revealing that Spanish high school EFL students were more positive about CF than undergraduate FL students in the US in Loewen et al.'s (2009) study. He suggested the EFL high school students in his study expected and wished to be corrected regularly in the classrooms.
Several studies compared the beliefs held by students and teachers regarding the role of CF, revealing that students were much more positive about CF than teachers (Brown, 2009;Jean & Simard, 2011;Li, 2017;Roothooft & Breeze, 2016;Schulz, 2001). Students were willing to receive immediate and explicit correction, while teachers were concerned about the possible detrimental effects of CF on students' emotional states (Li, 2017;Roothooft & Breeze, 2016). Students' beliefs about the importance of CF have been found to mediate their noticing of the corrective nature of the teachers' feedback (Kartchava & Ammar, 2014), and students' preferences for particular CF types could mediate the rate of immediate uptake following CF (Akiyama, 2017).
As research in SLA has demonstrated the importance of CF provision (Ellis, 2017), the critical pedagogical questions of how, when, whom and what to best correct in L2 classrooms are worthy of further research (Ellis, 2017;Ha, 2017;Ha & Murray, 2020Lyster et al., 2013). As for investigations into students' preferences for CF types, the limited literature reveals some mixed findings. In a study with 60 advanced graduate students in the US who were training to be teaching assistants in a spoken ESL course, Lee (2013) found that the students rated explicit correction as the most preferred CF type and metalinguistic feedback as the least preferred one. In a study with Spanish EFL students (282 secondary school students and 113 adults), Roothooft and Breeze (2016) found that students were positive about explicit correction and metalinguistic feedback. Interestingly, adult learners rated recasts much more positively than secondary school students. However, the study did not offer any reasons for this discrepancy. In a study of the relationship between Iranian undergraduate EFL students' anxiety levels and CF beliefs, Zhang and Rahimi (2014) found that their students (80 high anxiety and 80 low anxiety students) rated explicit correction and metalinguistic feedback as the most effective type, and elicitation, recast, repetition and clarification request as equal second most effective type, regardless of their anxiety levels. This mirrors the findings of an early study with ESL students in Singaporean high school and university contexts by Oladejo (1993), who found that metalinguistic feedback received the highest rating score. In a recent study with undergraduate EFL students in China, Zhu and Wang (2019) found that their students preferred to receive output-prompting CF (e.g., repetition, metalinguistic feedback) more than input-providing CF (explicit correction). It appears from these studies that contexts influence students' preferences for CF types, and EFL students in some contexts like to receive metalinguistic feedback more than US ESL students. However, little is known about students' preferences for CF types in such contexts as Asian EFL high schools, including Vietnam.
As for the timing of CF, Davis (2003) reported that 86% of the 97 undergraduate EFL students agreed that CF should be provided as soon as possible to avoid bad habit formation. Brown (2009), in a large survey of 1,600 FL students in a US university, reported that his students expressed moderate agreement that effective teachers should correct errors immediately. Zhang and Rahimi (2014) found that their undergraduate EFL students, regardless of their anxiety level, preferred immediate CF to delayed CF, while the students in Zhu and Wang's (2019) study held a negative view of delayed CF. These studies suggest a tentative conclusion that students generally prefer immediate CF to delayed CF, but the conflicting views between SLA researchers and L2 educators regarding optimal CF timing (Ellis, 2017;Ha & Murray, 2021) warrant further research for a more nuanced understanding of students' beliefs and more persuasive pedagogical implications in this regard (Quinn & Nakata, 2017).
Concerning CF targets, students in Zhang and Rahimi's (2014) study felt that errors that should be corrected the most were those hindering communication, followed by frequent errors. Advanced ESL students in Lee's (2013) study also indicated that they would like errors occurring the most often in their speaking to be corrected. Zhu and Wang's (2019) students wanted their errors to be corrected even when the errors were not serious. In Oladejo's (1993, p. 78) study, students believed that "comprehensive, not selective" errors should be corrected to improve their language accuracy. The secondary school student group expected CF for grammar errors, while university students expected more CF for issues with organization of ideas. As no consistent pattern can be derived from the limited literature, more research is needed to draw meaningful conclusions about students' preferences for error types.

Oral corrective feedback and extraversion, English learning motivation and gender
As the role of extraversion, motivation, and gender in learner CF beliefs has not been empirically investigated, it makes sense to review studies exploring the role of these differences in the engagement in, occurrence and effectiveness of CF. Extraversion appears to be consistent over time, and it can influence one's behavior (Matthews et al., 2003). Kim and Nassaji (2018) looked at the influence of extraversion on the occurrences and effectiveness of CF techniques in two ESL classrooms. They found that more extraverted learners were more likely to engage in more CF episodes, but more introverted learners demonstrated more successful uptake. However, there has not been adequate empirical evidence to draw firm conclusions regarding the role of extraversion in CF engagement. Kim and Nassaji (2018) call for more research investigating the relationship between learner extraversion and their participation in CF. Motivation has been found to be a critical variable influencing L2 learning. According to Dörnyei (2005), motivation "provides the primary impetus to initiate L2 learning and later the driving force to sustain the long and often tedious learning process; indeed all other factors involved in SLA presuppose motivation to some extent" (p. 65). Numerous studies have investigated learners' reasons for language learning and its role in learning process and outcome (Dörnyei, 2020). Regarding the role of motivation in CF, it is not clear whether learners' motivation influences their noticing of CF, which might in turn influences their learning outcome. As highlighted earlier, learners' CF beliefs were found to influence their noticing of CF (Kartchava & Ammar, 2014) and CF effectiveness (Akiyama, 2017;Sheen, 2007), but it is not known whether motivation has any relationship with learners' CF beliefs.
Regarding the role of gender in the occurrence and effectiveness of CF, in a study looking at the negotiation of meaning in child interactions, Oliver (2002) did not find any impact of gender on the occurrence of negotiation for meaning, but it played a role in some other studies. Ross-Feldman (2007) found that in a picture story task, either female-female or male-female dyads engaged in more language related episodes and resolved them better than male-male dyads. Pawlak (2020) found a mediating role of gender in the occurrence and outcome of CF in communicative tasks. Same gender arrangements produced the majority of instances of negotiated interaction (90%) and output modification (75%). He concluded the study with an urgent call for more research investigating the role of gender, among other variables, on the process and product of interaction, including CF. Nakatsukasa (2017) found that gender did not influence the effectiveness of verbal recasts during two communicative tasks, but females benefited from gesture-enhanced recasts more than males in the long run. Due to the limited research investigating the role of gender in CF and the mixed findings, more studies are needed to investigate the role of gender in the beliefs about, occurrence and effectiveness of CF.
In conclusion, previous research has shown that students are positive about the importance of CF, but further research is necessary for a more nuanced understanding of students' beliefs about other pedagogical aspects of CF such as CF types, timing and targets. Within the available literature, little is known about students in Asian EFL high schools, although students' CF beliefs have been found to be context-dependent (Schulz, 2001). Moreover, learner individual differences such as gender, extraversion, and English learning motivation may influence students' CF beliefs, but there is a lack of empirical evidence. There is, therefore, a need for more studies investigating students' CF beliefs in a broader range of contexts, and investigating the impact of the learner individual differences on their CF beliefs. The present study is thus timely in addressing the following research questions (RQs): 1. What are Vietnamese high school EFL students' beliefs about oral corrective feedback? (RQ1) 2. Are there any relationships between Vietnamese high school EFL students' beliefs about oral corrective feedback and their gender, extraversion, and English learning motivation? (RQ2)

Method
This study adopted an explanatory sequential mixed-methods research design. Ethics approval and consent for both questionnaires and interviews in both the pilot study and the main study were obtained prior to conducting the research.
The following sections present the context, the participants, and the data collection instruments, procedures and analysis.

Context
The main study took part in a typical public high school in a central province of Vietnam. The school consisted of 39 classes in all three grades: 10, 11 and 12. English is taught by Vietnamese teachers with university qualifications in EFL teaching. English is a compulsory foreign language subject, which is taught in three 45-minute lessons per week. In Vietnam there has long existed a conflict between the macro-level curriculum objectives and the approach to testing. The national English language education curriculum aims to help students to obtain a preliminary level of English upon graduation (Level B1, Common European Framework), but the actual teaching and learning remains highly test-driven (Ha & Murray, 2021). All the students are required to take a final national exam as part of their graduation requirements. This exam is in a written multiple-choice question format, with most questions focusing on assessing students' grammar and vocabulary knowledge. There is no oral assessment.

Participants
Based on the convenience sampling method (Dörnyei, 2007), 250 students, aged from 15 to 17, in grades 10 and 11 at a typical public high school in Vietnam, were invited to complete a paper-based questionnaire and express interest in a follow-up interview. In order to ensure the feasibility of the study, the first 15 volunteers, including seven males and eight females, resulting in an appropriate gender balance, were invited to take part in follow-up semi-structured interviews. Students of grade 12 were not invited because they were busy preparing for their exams, and the highly exam-oriented focus may have influenced their beliefs about CF. Most of the students had started learning English in grade 6, some in grade 3. The classroom was the main place for most of the students to interact in English. Some students chose English as one of the main subjects, with the final exam result considered for admission to universities. Generally, these were students who spent more time learning English outside of the classroom, which involved engaging in self-study at home or taking extra classes at school (after mainstream class time), at their teachers' own home, or at private language centers. Regarding students' proficiency, while considered by teachers to be pre-intermediate in reading, grammar and vocabulary (speaking, listening and writing may be at a lower level), the annual national graduation exam reports indicate that the students' level of English proficiency varies considerably among individuals at the same level.

Instruments
The data were collected via a questionnaire and semi-structured interviews. The questionnaire was developed by the researchers based on the synthesis of literature investigating learner CF beliefs (e.g., Kartchava & Ammar, 2014;Loewen et al., 2009;Schulz, 1996), following basic guidelines for questionnaire construction (e.g., Dörnyei & Taguchi, 2009). All the items were original. The questionnaire was initially developed in English by the authors. After the authors polished and revised the items through various rounds of meetings and discussion, the questionnaire was translated into Vietnamese by the first author in consultation with two colleagues who were bilingual (Vietnamese and English). Then, content validation was conducted via group discussions with three teachers and five students from one of the two schools where the pilot study was conducted.
The validation was conducted with the teachers before trialing with the students. The teachers and the students were asked to read the questionnaire items carefully and complete the questionnaire in the presence of the first author. They were asked to exchange their ideas with the researcher regarding the content and the wording of the items. They were encouraged to openly discuss any issues or hesitations they experienced while completing the questionnaire. Amendments to some items were made based on the comments of both the teachers and the students. A pilot study was then undertaken with 100 students in two high schools that were not used for recruiting participants for the main study but were comparable to them in terms of age, proficiency and learning contexts. Analysis of the pilot study helped identify several flawed items, which were then excluded to improve scale reliability. The time range for the completion of the questionnaire was estimated (15-23 minutes), and a satisfactory reliability was achieved (α =.83).
The final version of the questionnaire (see Appendix A for the English translation) 2 comprised two main parts. The first part included ten demographic questions eliciting students' age, gender, self-rating in terms of one key personality trait (extraversion/introversion), English learning motivations, satisfaction with their teachers' English proficiency, and teaching methods. Accordingly, the participants' beliefs about their level of extraversion were elicited via one item (Item 3), asking them whether they described their personality as more introverted or more extraverted. We allowed the participants to rate their extraversion and introversion, so the construct was assessed in terms of the students' beliefs about their personality rather than through a psychological questionnaire. Similarly, their purposes in learning English were elicited via one item (Item 4) asking them to select one of the five options describing their main motivation. In the data analysis, these options were grouped into three broader categories, namely, motivation for (1) communication in English, (2) future career, and (3) other reasons (including exams and compulsory requirements). The first of these could be classified as intrinsic (Ryan & Deci, 2000) or integrative (Gardner, 2010) in that it relates to a desire to use the language for purposes of connection and interaction with others rather than practical gain or obligation. It should be noted that in the context in question, where language study is compulsory, the application of more complex motivational models becomes less appropriate because of the absence of learner choice. As there were no sensitive or intrusive questions, this section was placed at the beginning of the questionnaire. The main part of the questionnaire, the students' beliefs part, consisted of 44 Likert scale items. The questionnaire was to elicit students' beliefs about various aspects of CF: importance, types, targets, timing, providers/sources, and possible affective aspects. The questionnaire was administered in Vietnamese to eliminate any language issues on the part of the participants.
The interview questions were developed based on both the synthesis of the literature and the preliminary analysis of the students' completed questionnaires.
The purpose of the interviews was to elaborate on the quantitative findings. The final interview protocol included questions eliciting additional information about students' biography and students' beliefs about CF. Seventeen guiding questions were used to ask students to elaborate on their beliefs about CF importance, CF source, CF targets, CF timing and possible affective responses to CF (see Appendix B for the English translation of the interview guide).

Procedures
Two hundred and fifty students received the questionnaire to complete at their convenience. After one week, 247 questionnaires were returned. However, 11 were incomplete, leaving 236 for analysis. Regarding the robustness of the sample size for factor analysis, although some researchers recommended ten times as many participants per variable, more recently researchers have considered five to be adequate (Cutillo, 2019;Pallant, 2016). Therefore, this sample size (236 participants for 44 items) can be deemed sufficient for factor analysis. The interviews were conducted individually in Vietnamese by the lead author three weeks after the completion of the questionnaires. Each interview lasted approximately 22 minutes on average. All interviews were audio-recorded for data transcription and analysis.

Data analysis
Firstly, an exploratory factor analysis (EFA) was conducted following rigorous procedures using SPSS version 24. Before performing the EFA, the suitability of the data and sample for EFA was assessed. Cronbach's alpha (α) for the questionnaire was .85, indicating excellent reliability (Pallant, 2016) for the scale. Most communalities were above .50, indicating that the sample size (N = 236) was adequate (Field, 2013). The correlation matrix showed many coefficients of .3 and above. The Kaiser-Meyer-Olkin value was .742, and Bartlett's test of sphericity reached statistical significance (χ 2 = 2,921.257, p < .001), supporting the factorability of the correlation matrix (Field, 2013;Pallant, 2016). An oblique rotation method (direct oblimin) was initially used, showing a low degree of correlation between the extracted factors (no correlations reached .3). Therefore, an orthogonal rotation method (varimax) was employed to help the interpretation of the results. After the EFA, the descriptive statistics for each extracted factor were calculated. To explore the relationship between the students' beliefs and their gender, extraversion, and English learning motivation (RQ2), a multivariate analysis of variance was initially deemed appropriate, but the assumptions for running it could not be met. As a result, separate general linear models were fitted.
The interview recordings, transcribed verbatim by the first author who shares the first language with the participants, were analyzed thematically with the support of the NVivo software. A theory-driven, deductive approach (Gilgun, 2011) was used for this study. The data were analyzed according to predetermined macro themes derived from the literature, including students' views about the importance and the necessity of CF, CF types, timing, targets, and possible affective responses to CF. The qualitative data were used to support and elaborate on the findings of the EFA and descriptive statistics to address RQ1.

Exploratory factor analysis
The principal components analysis revealed 14 factors with eigenvalues over 1.0, explaining 62.5% of the total variance. A parallel analysis (using Monte Carlo PCA) was used to determine the number of significant factors (Pallant, 2016). It showed seven factors with eigenvalues exceeding the corresponding criterion values for a randomly generated data matrix of the same sample size (44 variables x 236 participants x 100 replications). To support the parallel analysis, an investigation of the scree plot also showed a break after the seventh factor. Therefore, the first seven factors, explaining a total of 43% of the variance of students' CF beliefs, were retained for further analyses. The scale reliability was investigated for the seven factors, and the α values were .684, .694, .755, .659, .608, .729 and .532 respectively. Factor 7 had the lowest α value (.532), so it was decided to omit it. Therefore, the first six factors, including 20 items and explaining a total of 39.5% of the students' CF beliefs variance, were finally retained for interpretation (see Table 1). .643 34. Some errors are not related to the focus of the lesson, but they need to be corrected.
. 866 33. Some errors are not common in the class, but when they occur, they need to be corrected. .792 As shown in Table 1, Factor 1 consists of five items. Three of them are about students' preferences for metalinguistic feedback, clarification requests and repetitions. The two remaining items are about preferences for integrated recasts (i.e., the reformulation of the erroneous utterance is integrated into additional discourse that continues the topic of conversation; e.g., Item 42) and interrogative recasts (i.e., the reformulation of the erroneous utterance is made in an interrogative form; e.g., Item 43), respectively. Although these two subtypes of recasts provide the correct forms, they are formulated in a way that encourages the learners to continue the topic of conversation. Therefore, Factor 1 was labeled "output-prompting CF and eliciting recasts." Factor 2, labeled "desire for CF," consists of four items addressing learners' desire and expectation for CF in response to their oral errors. Factor 3, labeled "non-verbal cues," contains three items addressing learners' noticing of and the role of non-verbal cues as a form of feedback. Factor 4 contains three items addressing the need to correct errors which are perceived to be highly significant, and so this factor was labeled "important errors." Factor 5, which contains three items about full recasts, isolated recasts and explicit corrections, was labeled "input-providing CF." The last factor containing two items addressing the need to correct less significant errors was labeled "less important errors."

Descriptive statistics and qualitative findings regarding students' beliefs about oral corrective feedback
This section reports selected interesting findings revealed in the descriptive statistics which show the extent to which the students agreed with the items related to the six factors extracted above (out of a 5-point Likert scale). These beliefs were then elaborated based on the evidence from the follow-up interviews.
Regarding items probing students' desire for CF (Factor 2), students' responses to the general question about their desire for CF (Item 3) resulted in the highest mean (4.42), indicating that students were highly desirous of feedback. The students also liked to receive feedback when they were presenting in front of the class (Item 5; 4.12). However, students' expectations for CF when doing group-work activities (Item 6) was the lowest (3.89). In the interviews, all the students stated that CF was important and necessary for learning. Ten of the students wished to be corrected as much as possible. They said that CF helped them understand errors, memorize the target language features, and avoid bad habits, and CF could improve accuracy (more important than fluency), which improved exam scores consequently. For example, Student 1 stated: Teachers' provision of corrective feedback is definitely useful because correcting an error means that we can avoid an error in our exam. The more errors we can reduce in exams, the better, which helps us achieve higher marks in exams.
Regarding output-prompting CF and eliciting recasts (Factor 1), metalinguistic feedback (Item 37) received the highest rate of approval (4.12), whereas clarification requests (Item 38) received the lowest rate (3.46). The two question-form recasts (Items 42, 43) also received high rates of agreement. In the interviews, twelve of the students said that they liked metalinguistic feedback and elicitations because they could have a chance to self-correct, and self-correction helped them to think and memorize correct forms. They also said that their teachers used metalinguistic feedback often and it was effective, and teachers sometimes used L1 to explain language rules. Three of the fifteen students said clarification requests were not effective because they would not know what their errors were. They felt confused, uncomfortable, and worried if the teacher used clarification requests.
Regarding input-providing CF (Factor 5), explicit corrections (Item 44) received a higher rate of approval (3.95) than isolated recasts (Item 41) and full recasts (Item 40). These recasts received the same rate of approval, which was relatively low (3.7). In the interviews, eight students said explicit corrections were the most effective CF type because they were easy to understand and suitable for weak students. Interestingly, five students considered metalinguistic feedback similar to explicit corrections, saying that metalinguistic feedback "is a direct strategy as the teacher informs me directly of the error and helps me to correct my error" (Student 2). Eleven students said recasts were sometimes not salient enough and they would easily forget the correction as they did not really think about their errors and did not try to self-correct.
When interviewed about CF timing, an aspect that might influence students' choice of CF types, 13 of the students appreciated immediate correction because it would provide them with opportunities to repeat the correct forms (i.e., uptake), which in turn made them memorize the correct forms better. In response to how immediate errors should be corrected, they said it was best for their teachers to correct their errors as soon as they finished a sentence rather than correcting them while they were speaking (i.e., in the middle of the sentence). It minimized the possibility of interruptions, and it increased the chance to remember the errors and the correct forms. Ten students said that correcting errors after the speaking activity or by the end of the lesson was not effective because they might have forgotten what they said and what mistakes they made. Four students said CF timing should depend on the error types, as illustrated by the following excerpt: When asked about the possible negative effects of CF, which might also influence students' preferences for CF types, all the students felt comfortable receiving teacher CF because CF provision was a very frequent practice and they expected CF to improve their speaking. Four students said CF had some possible negative effects. Interestingly, they said teachers' correction strategies were not the main cause of negative affective responses (except that three students mentioned clarification requests above), but they considered that teachers' negative attitudes to students' errors may have some negative effects on students' feelings.
Three items explored students' noticing and evaluation of the usefulness of non-verbal cues as CF (Factor 3). Students showed slight agreement that non-verbal cues (Items 8, 10) are a useful method of CF (around 3.73). This may be influenced by the fact that their teachers did not frequently use non-verbal cues along with CF. As shown in the mean score of Item 9 (3.48), the students did not frequently notice their teachers use non-verbal cues in their CF provision. Similarly, in the interviews, all the students said they did not have a strong impression about their teachers' use of gestures in identifying their errors and said non-verbal cues were not effective because they did not help them locate or correct their errors.
Regarding beliefs about error types (Factor 4), students were positive about receiving CF for important errors. Interestingly, all three error types (Items 29,31,32) which are important in their different ways (most frequent errors, errors impeding communication, and errors related to the lesson foci) received a similar rate of agreement (around 4.08). In the interviews, all students stated that frequent errors, and errors pertaining to the lesson focus were the most severe and worth correcting, as reflected in the following comment: Frequent errors (committed by many people) are the most important. Correcting these errors is useful for many people at the same time. Also, we study English mostly for exams, therefore, correcting errors related to the foci of the lessons . . . will help us to do better in our exams and get high marks. (Student 4) Five students said pronunciation errors were the most important because they influenced communication. Errors influencing communication were said to be important, but not for everyone because some students learned English for communication while others learned English to pass exams only. They considered correcting grammar errors the most important because grammar errors could influence exam results and correcting grammar errors was part of the teachers' main job.
Surprisingly, students also showed a similarly high level of approval of CF for less important errors (Items 33,34;around 4.02;Factor 6). In the interviews, 12 students said that simple errors, such as subject-verb agreement errors, plural use errors and articles are worth correcting because they helped improve accuracy and exam results.

Relationships between students' beliefs about oral corrective feedback and their gender, English learning motivation, and extraversion.
In order to explore possible relationships between students' CF beliefs and their gender, extraversion, and English learning motivation, statistical tests of significance were applied. Firstly, the assumptions for parametric tests were checked. Descriptive statistics (through histograms and the skewness values) suggested that the distributions of the factor scores were relatively normal. The homogeneity of variance was checked as part of the general linear model fitting, and results of Levene's test of equality of error variances showed non-significant values for all six dependent variables (the six factors underlying students' CF beliefs), indicating the fulfillment of the assumption. A Pearson correlation test showed that the correlation values among the six dependent variables were non-significant, indicating the suitability for fitting general linear models (Pallant, 2016). Separate general linear models were fitted to find out if there were any significant differences among the participants as related to the underlying factors of their CF beliefs and the demographic variables of gender, extraversion and English learning motivation. Categories of students' motivations included motivation for (1) communication in English, (2) future career, and (3) other reasons (including exams and compulsory requirement). The results 3 showed some statistically significant relationships, although the effect sizes were small.
There was a statistically significant main effect for gender (F(1, 224) = 9.037, p = .003) in students' desire to receive CF (partial eta square = .039). The estimated marginal means (grand mean: .016, male: -.227, female: .259) showed that female students desired to receive CF more than males did.
Also related to students' desire for CF, there was a statistically significant main effect for the motivation type (F(2, 224) = 3.484, p = .032; partial eta square = .030). The estimated marginal means (grand mean = .016, mean for communication = -.263, mean for future career = -.03, mean for other reasons = .313) showed that students learning English for exams or other reasons were the most positively disposed to receiving CF. Students who learned English for communication purposes were more negative about receiving CF.
Regarding beliefs about non-verbal cues, there was a statistically significant main effect for the motivation type (F(2, 224), p = .024; partial eta square = .033). The estimated marginal means (communication = .163, future career = -.151, other reasons: = .336) showed that students learning English for exams or other reasons were the most positive about non-verbal cues as a CF strategy, while students learning English for future career were the most negative. Post hoc tests (Tukey HSD) showed a statistically significant difference between the communication-motivation learners and future-career-motivation learners, and between future-career-motivation learners and other-reasons-motivation learners.
Regarding beliefs about input-providing CF, there was a statistically significant interaction effect for gender*extraversion (F(1, 224), p = .026; partial eta square = .022). This indicates that there is a significant difference in the effect of self-rated introversion/extraversion on beliefs about input-providing CF for males and females. To interpret this interactional effect further, an analysis of simple effects was performed using pairwise comparisons with a Bonferroni adjustment (.05/2 = .025), showing that the introverted females were less positive about receiving input providing CF than their extraverted counterparts (mean difference = .471, p = .02).

Discussion
The EFA resulted in six factors underlying students' beliefs about CF. These factors represent six important themes of CF beliefs in the literature (Lyster et al., 2013;Nassaji, 2015). This finding partially corroborates the findings of a recent study conducted by Zhu and Wang (2019) looking at beliefs of Chinese undergraduate learners of English, sharing three factors, namely, desire for CF, inputproviding CF, and CF for less important errors. Students' views separating the input-providing CF from output-prompting CF support the necessity of further investigations of the effectiveness of the two CF categories (Lyster & Saito, 2010;Zhu & Wang, 2019). Interestingly, in our study, students viewed the four subtypes of recasts differently while researchers usually view them as one united construct -input-providing CF. Isolated recasts and full recasts were grouped with explicit correction as one category (Factor 5). On the other hand, interrogative recasts and integrated recasts, the two recasts consisting of questions inviting students to contribute further to the interaction, were grouped with repetition, metalinguistic feedback and clarification requests (Factor 1). This distinction suggests that students' perceptions of the corrective functions of different subtypes of recasts need further investigation.
Similarly to CF types, there was a distinction between students' beliefs about more important error types (Factor 4) and beliefs about less important error types (Factor 6). However, the descriptive statistics showed that students desired to receive CF for both error types at a similar level. This might help to partially resolve the disagreements between L2 teacher educators and SLA researchers regarding which errors should be corrected, as discussed by Ellis (2017). Based on their own experience, L2 teacher educators (e.g., Harmer, 2007) suggest correcting global errors (e.g., errors influencing communication) rather than local errors (e.g., morphological errors), and correcting errors (due to lack of knowledge) rather than mistakes (e.g., a slip of the tongue). SLA researchers (e.g., Lyster & Saito, 2010), however, have focused on investigating the effectiveness of CF on errors which are more likely to function as local errors such as articles or past tense verb forms. Students' eagerness to receive CF on all types of errors might suggest that teachers should not be too concerned about how to distinguish errors from mistakes or global from local errors, which might not in any case be a feasible task. Another interesting finding in our study is that three items pertaining to non-verbal cues formulated one separate factor (Factor 3) of students' CF beliefs. This suggests the necessity for further research looking at the role and occurrences of non-verbal cues as a type of CF in various classroom settings, responding to the call of Ellis (2017).
Descriptive statistics and qualitative analysis of the interviews also revealed some interesting findings. Regarding students' desire for CF, students in this study were very positive about receiving CF, which corroborates the findings in the literature (e.g., Zhang & Rahimi, 2014;Zhu & Wang, 2019). Qualitative data revealed that the students desired to receive CF to improve their accuracy, especially for their exams which were generally in written form and tested accuracy in vocabulary and grammar use.
Regarding preferences for CF types, metalinguistic feedback was the most strongly preferred. This is different from findings by Lee (2013), but it is similar to Zhang and Rahimi's (2014) and Zhu and Wang's (2019) findings. This may suggest that students in EFL contexts prefer metalinguistic feedback to those in ESL contexts, where real-world communication in English is a more pressing objective. Another possible explanation for this preference in the current study conducted in Vietnam, and perhaps other Asian EFL contexts, may be the influence of the test-driven nature of English curricula. Another possible interesting explanation for this preference in the present study is that the teachers gave feedback in their L1, which increased the level of comfort for the students. The interviews also showed that five students considered metalinguistic feedback as a direct and explicit feedback type, effective in helping students to locate their errors and self-correct. Students' self-correction may contribute to feelings of pride and satisfaction as well as enhancing their status, which is of considerable importance in the Vietnamese classroom culture. This may also explain why students preferred interrogative recasts and integrated recasts, where they have a chance to self-repair their errors, rather than isolated and regular recasts.
Students also preferred explicit correction, which corroborates some of the earlier findings (Lee, 2013;Zhang & Rahimi, 2014) but is different from the findings of Zhu and Wang (2019). It is surprising that Chinese undergraduate students in Zhu and Wang's study rated explicit corrections as the lowest among various CF types; unfortunately, no explanations for such an observation were offered. The students in the current study liked to receive explicit corrections because, as revealed in the interviews, this CF type was straight-forward, easy to understand, and familiar to them. Interestingly, clarification requests received the lowest rate of approval (3.42/5). The interviews revealed that this CF-type was less preferred because of its unclear corrective intention, which was claimed to be the main cause of students' confusion and discomfort. This is also reported in Lee's (2013) study. Similar findings from both the questionnaire and interview data were also observed for non-verbal cues. Students commented that they did not notice their teachers frequently using non-verbal cues to identify their errors, and that they did not feel that non-verbal cues were an effective way of giving CF. The students' CF preferences and their comments that their teachers used explicit correction and metalinguistic feedback the most frequently but rarely made use of non-verbal cues in giving CF may indicate that students preferred CF types which are more familiar to them.
Further statistical analyses showed some interesting relationships between students' CF beliefs and their gender, self-rated extraversion, and selected types of English learning motivation. Firstly, female students were more positive about CF than their male counterparts were. One speculation based on our subjective observation is that this difference may be because female students in our context are more successful students than males. In a large-scale survey study looking at the gender gap of adult learners of Dutch as an L2, females were found to outperform males in speaking and writing tests (Van Der Slik et al., 2015). There were some comments in our interviews that those who had stronger English proficiency and who were more confident about their English tend to be more positive about CF than their weaker peers. This suggests that English proficiency may be a moderator of learner CF beliefs, and that further research investigating the role of gender in learner beliefs should consider controlling for the variable of language proficiency. Secondly, it is interesting that extraversion-identifying females were more positive about input-providing CF than introverted females. It is not clear why this pattern was found with the female students in the current study. In the only earlier study looking at the role of extraversion in CF, Kim and Nassaji (2018) found that their extraverted students were more likely to engage in CF episodes, which may partially explain our findings. Further research is warranted to investigate whether and why this relationship exists in different contexts.
In regard to the relationship between learning motivations and desire for CF, students learning English for exams are more positive about CF than those learning English for communication or for future career. This is not surprising because students learning English for exams may want to be corrected more often to improve their accuracy. This was also reflected in the interviews, where some students commented that accuracy was more important for them than fluency and that they would like their teachers to correct both global errors and local errors to improve their exam results. This is also a possible explanation for the finding that students learning English for exams are the most positive about non-verbal cues as a CF type. The short-term objective of accuracy in an assessment context may drive their receptiveness to all kinds of CF. Another possible explanation for this preference for non-verbal cues as a CF strategy is that non-verbal cues are less face-threatening, so they may particularly attract these exam-oriented students who might be sensitive to teachers' evaluation and grade.

Limitations and directions for future research
Despite the contributions discussed above, there are several limitations that should be noted. Firstly, the students' extraversion and English learning motivation were self-rated with the help of single items. However, the information we requested about these variables was quite straightforward, so we have reason to believe that our participants were unlikely to have had difficulty in understanding what was required. Future research may be able to investigate these relationships further, using a more comprehensive tool to investigate a wider range of students' personality traits, and comparing the results of a comprehensive psychological test battery with a self-rating score of beliefs. Secondly, the 6component solution from the EFA explained only 39.5% of the variance, which may not be an ideal amount. Given the complex nature of the explored phenomenon, learner beliefs about CF, this can be deemed acceptable. Yet, further research to validate the questionnaire in other contexts would be helpful. Thirdly, there is a possibility that the inclusion of the first 15 volunteers in the follow-up interview may not provide a representative sample. Another possible limitation of the study is related to the fact that learner beliefs are dynamic, complex and situated (Ellis, 2008;Leontjev, 2016). This dynamic complexity is unlikely to be revealed in a cross-sectional, one-shot questionnaire-and-interview study. Therefore, the readers should take this caveat in interpreting the findings of the current study. Future research investigating learner beliefs can take an ethnographic and longitudinal approach to uncover the dynamic and complex nature of learner beliefs. For example, there could be a series of followup interviews at different points in time within one or two semesters following the questionnaire. Alternatively, asking the learners to keep a diary to track their beliefs over time may be valuable.

Conclusion and pedagogical implications
The present study adds to the limited literature on students' beliefs about various aspects of CF. The EFA identified six factors underlying high school EFL students' CF beliefs, which were corroborated and elaborated by the qualitative data. The descriptive statistics and thematic analysis of the interview data indicate that high school EFL students in Vietnam were positive about CF and they desired to receive CF in both input-providing and output-prompting forms. The exam-oriented teaching context may explain the students' desire to receive explicit correction and metalinguistic feedback. The study has made some attempts to investigate some relationships between students' CF beliefs and their gender, self-rated extraversion, and English learning motivation, and the statistical analyses revealed some interesting findings.
This study offers some pedagogical implications. Firstly, it opens up the possibility of a dialog within the profession about how teachers can respond to students' beliefs and preferences. For example, students desired to receive CF regardless of error types, which suggests that Vietnamese EFL teachers who chose to accommodate student wishes might select errors to correct without having to be concerned about distinguishing global errors from local errors, or errors from mistakes. Secondly, there is some evidence in the literature that teachers tend to use recasts due to their concern about students' affective responses (Roothooft, 2014). However, in this study, students' desire for both explicit correction and metalinguistic feedback (which five of the 15 students interviewed considered to be direct and explicit), and their comments that CF strategies were not a major source of anxiety and embarrassment, suggest that Vietnamese EFL teachers do not have to be greatly concerned about students' affective responses to some explicit CF types. Vietnamese EFL teachers, therefore, can use a variety of CF types in a flexible way to address different error types in particular situations. Armed with the knowledge to more effectively predict the preferences of different groups of students, teachers can make more informed decisions with better outcomes for all concerned.

STUDENTS' BELIEFS ABOUT ORAL CORRECTIVE FEEDBACK
Thank you very much for taking your time to answer our questionnaire. You will be helping us to understand more about your views regarding oral corrective feedback. Oral corrective feedback means any teacher or peer responses to your erroneous utterances. It can be an explicit correction; for example: Student: I go to the train station yesterday. Teacher: Not go, you should say went. It can just be a signal that there is an error in your utterance, and you need to rephrase it. For example: Student: I go to the train station yesterday. Teacher: I go? Please answer the following questions as honestly as possible. There are no right or wrong answers to these questions. Notes: This questionnaire is only about oral errors and oral corrective feedback, not about written errors or written corrective feedback.

Ranking
Please rank the following four error types (a-d) in the order of necessity for teachers' corrective feedback, with 1 being the most necessary and 4 being the least necessary: (a) grammar, (b) vocabulary, (c) pronunciation, and (d)