Teachers’ Corrective Preferences

^{Mackey et al. (2004)} considered 18 experienced and inexperienced teachers’ use of incidental focus-on-form techniques. Each teacher taught for half an hour during which their behavior was examined. Experienced teachers were found to use more preemptive focus on form, recasts, and explicit correction in comparison to the inexperienced teachers.

Using stimulated recall, ^{Polio et al. (2006)} also examined the relationship between teaching experience and corrective behavior. After a two-way information exchange activity, the videotaped interaction was played to each native speaker teacher. Each error and the teacher’s correction were marked as recast, negotiation, or ignore. The results showed that experienced teachers used more recasts (35% vs. 29%) but not more negotiations (9% vs. 11%). The difference was not significant, however. The cause of this small difference, they found, was that learners are likely to produce more language when talking with an experienced teacher because he or she uses more general statements and questions encouraging learners to speak.

In another study, ^{Zyzik and Polio (2008)} examined the occurrence of incidental focus on form using observation, interview, and stimulated recall in the context of advanced-level literature classes. The results of their study were similar to those of ^{Mackey et al. (2004)} in that teachers preferred to use recast for correcting their learners’ errors. Their major reason was that it did not embarrass the learners. Accordingly, explicit correction occurred only once because all the three observed teachers believed that it might cause embarrassment. Authors speculated that other types of feedback that involve negotiation, in contrast to recast, might be more time-consuming.

In a rather different study, ^{Lee (2013)} examined both teachers’ and learners’ preferences for corrective feedback. He observed and analyzed the use of feedback in advanced conversation-based classes and compared students’ and teachers’ preferences. The participants included four native speaker teachers and 60 adult ESL learners from different nationalities in the United States. After the fourth session, teachers and learners filled in a questionnaire followed by interviews with 20 of the students. The results were almost similar to the previous studies; recast was the most commonly used type of feedback. There was a mismatch between teachers’ and learners’ preferences in this regard; teachers had a high preference for implicit feedback (recast) while students disliked it and preferred explicit forms of correction.

^{Mori (2011)} examined the role of teachers’ beliefs in their correction in more detail by observing and interviewing two English language teachers in Japan. One of the teachers used recast in 92% of the cases and the second teacher only in 5.7%. Learners’ emotional reactions, instructional focus, time constraints, and the frequency of errors were the influential factors in the first teacher’s corrections; his major priority was to avoid hurting learners’ self-confidence and to help them take risks and contribute more to the classroom discussions, which made recast a good choice for him. The second teacher’s students, however, were not willing to express their beliefs because they were very afraid of making mistakes. The larger number of students (35) in comparison with the first class (18) caused further inhibition in the students. Therefore, the teacher needed to “induce” or “coax” them to talk by “prompting” them to express themselves, and, via elicitation, to think of what structures, words, and idioms to use (14.3%). However, as the students were unwilling to risk talking and their proficiency level was quite low, the number of corrections involving the use of recast was limited (5.7%).

In a more recent study, ^{Kamiya (2016)} explored, via interviews and a single classroom observation, four native-speaking teachers’ stated beliefs and practices in providing OCF. The author also examined the role of experience in the correspondence between the teachers’ beliefs and actual corrective practices. The results showed that the teachers did not believe in the effectiveness of correction, particularly recast, which they considered as peripheral. They avoided explicit forms of correction in order to create a comfortable learning environment for the students. The teachers preferred recasts, instead, which are, according to the teachers, not humiliating. A comparison between three teachers’ practices and their views pointed to the existence of a correspondence between the two, particularly in the case of experienced teachers.

^{Roothooft (2014)} included more teachers in his study and tried to compare their beliefs with their practice. He investigated 10 Spanish teachers’ beliefs/practices of corrective feedback through classroom observation and a following questionnaire for exploring their views. The findings from the study showed that although the teachers considered feedback to be important, they were not willing to interrupt their learners for correction and were concerned about the negative affective reactions that may develop following feedback. Furthermore, the teachers were not aware of how much and even what types of feedback they wanted to provide. The results also attested to the mismatch between belief and practice.

The results of a recent study by ^{Kartchava et al. (2018)} were also indicative of a mismatch between teachers’ beliefs about correction and their corrective practices. They asked 10 teachers to complete a questionnaire based on theoretical and empirical findings in the literature related to the importance of providing feedback, students’ anxiety and motivation, interrupting the communicative flow, and the delay and extent of feedback. They also observed the teachers for one hour in an authentic language class with intermediate learners. The results revealed that the teachers made fewer corrections (only 17%) than they said they would (54% of the errors). Recast was the most highly preferred and the most frequently used feedback type (66%) followed by explicit correction and prompts. Furthermore, no interaction was found between type of error and the type of feedback the teachers preferred except in the case of plurals and questions for which the teachers preferred regular recast and isolated recast (recasting the erroneous part only), respectively.

In a different study, ^{Roothooft and Breeze (2016)} also found clear mismatches, but this time between teachers’ and learners’ attitudes and preferences for correction. While the teachers believed errors should not be always corrected and showed a mixed attitude to immediate correction, the majority of their learners always expected to be corrected. The teachers believed correction may negatively influence learners’ self-confidence and also their fluency and were less inclined to use explicit forms of correction to ensure positive emotions and a positive reaction to their corrections. Their learners, in contrast, seemed to feel much more positive about explicit feedback forms.

The Focus of the Study

Teacher cognition research and practice have witnessed fundamental changes over the years. The field has gradually become concerned with the nature of teacher learning and the factors contributing to teachers’ mentalities regarding various dimensions of teaching. Along this line, ^{Borg (2003)} reviewed research on teacher cognition and identified four factors feeding teacher cognition: previous learning, professional development initiatives, contextual factors, and experience. This line of inquiry, according to Borg, has been concerned with the way teachers conceptualize their practice. In the field of error correction, teachers’ conceptualization of their practice has received less attention. This might lead to considerable gaps between what researchers recommend and what practitioners do. As an instance, there are certain contradictions between teachers’ practices and research findings. For example, teachers use recast although it does not produce much learning particularly for less proficient learners (see Table 1). Besides, proficiency level has been found to be related to learners’ preferences for feedback (^{Brown, 2009}; ^{Yang, 2016}), but whether teachers differentiate between less and more proficient learners has not been touched upon.

Table 1 Studies on the Effect of Learners’ Language Proficiency Level and Their Use of Correction 

Few studies (e.g., ^{Kennedy, 2010}) have considered whether the level of proficiency influences teachers’ corrective behavior. Kennedy used classroom data from the CHILDES database and analyzed them in terms of the type of errors, type of feedback, and rate of learner uptake and repair for each proficiency group. The teacher recorded in the transcripts was an ESL teacher whose first language (L1) was English. There were 15 Grade 1 ESL learners. Kennedy found that the teacher used recast mainly for the low-proficiency group and elicitation-like types of feedback, which require self-correction, for the Mid/High proficiency groups. Kennedy’s study suffers certain shortcomings such as the use of a limited sample population: She used data from just one teacher.

Considering the importance of informing teachers of the empirical findings (^{Roothooft, 2014}), there is, first, a need to see if teachers’ practice reflects the research findings in this regard. Therefore, the present study attempted to fill this gap by addressing the following research questions:

Participants

This is an observational and survey-based study evaluating language teachers’ practices and perceptions of oral error correction. This study was carried out in two private language institutes in Tehran, Iran. The participants included five EFL teachers and 84 learners. The teacher participants had at least two years of teaching experience in conversation-based classes in private language institutes. They had either a BA (n = 3) or MA (n = 2) in applied linguistics. The number of students in the classrooms was limited (about 10 in each class) because the courses were all conversation-based. The participants were male (n = 39) and female (n = 45) EFL learners ranging in age from 12 to 31. They were pre-intermediate (n = 43) and advanced learners (n = 41) according to the institutes’ classification, which was based on written exams and oral interviews at the beginning of each semester. All the learners had also received formal foreign language instruction either at schools or universities.

Instruments and Procedure

Data were collected using 20 observations and audio recordings followed by interviews with the observed teachers. Each of the five teachers was observed over four sessions (two in their lower-intermediate and two in advanced class) making up 20 one-hour observation sessions. The observed classes were also audio-recorded for further analysis. The teachers were not informed about the purpose of the study before the observations so that their teaching behavior would not be affected. Accordingly, after the observations, which were aimed at understanding teachers’ corrective practice across proficiency groups, in the second phase of the study, semistructured interviews were carried out with the observed teachers to also learn about their beliefs and preferences in this regard. The interviews followed the observations to ensure that the purpose of the study would not be revealed and the teachers’ responses and comments in the interviews would not affect their teaching practice. Each interview lasted from 20 to 35 minutes. The questions asked in the interviews were concerned mainly with the necessity, amount, and types of feedback preferred for different proficiency groups and how teachers differentiated between them (see Appendix). The data from the content analysis of the interviews were finally matched against the results of the observation to figure out the possible areas of inconsistency between the teachers’ practice and beliefs.

Amount of Correction

To find the amount of correction used in practice, almost all the utterances made by the students that had at least one error were reported. Since the focus of the present study was on the different ways of error treatment by the teachers, we will report both the number of errors followed by feedback and errors that did not receive any correction (i.e., non-corrections). Table 2 provides a report on the number of errors made by each proficiency group and the frequency of corrections made by the teachers.

As shown in Table 2, the advanced learners made more errors than the lower-intermediate ones. This was because they had a more active contribution to the raised discussions considering their higher ability to express themselves. However, when the number of words was compared with the number of errors for the lower-intermediate and advanced groups (70 errors in every 1,000 words vs. 111 errors in every 1,000 words, respectively), the proportion of words to errors was higher for the lower-intermediate group meaning that they, in fact, made more errors. Naturally, similar to the studies shown in Table 1, the teachers made more corrections for the lower-intermediate group.

The observed teachers’ individual performance (see Table 3) showed that a majority of them, as in the previous studies (e.g., ^{Jean & Simard, 2011}; ^{Kartchava et al., 2018}; ^{Lee, 2013}), left more than half of the errors untreated. This was particularly so in the case of advanced learners; the teachers, except for Teacher 5, provided fewer corrections for the advanced learners in practice. This was not matched with their belief, however. Except for two, the other teachers said they either do not differentiate between the two proficiency groups (Teachers 2 and 5) or prefer more correction for the advanced learners (Teacher 4). Teacher 1, for instance, noted that “I feel more comfortable when correcting my advanced students” because they have enough background knowledge to understand their errors. Teacher 3 believed low-proficiency learners are more likely to be negatively affected by feedback: “I do not correct all their errors as they might lose their self-confidence.” Teacher 2 stated that “the amount and the types of correction I use depend upon learners’ willingness to learn rather than their proficiency level.” However, he made more corrections for his lower-intermediate students, because, according to him, “lower-proficiency learners depend more on the teacher and expect more corrections.” Similar to Teacher 2, Teacher 5 pointed out that she does not differentiate between the two proficiency groups. She preferred limited correction for both groups: “Overcorrection might be discouraging and embarrassing at lower levels and interruptive for advanced learners.” In practice, she made more corrections for the advanced learners, which might be due to her belief that “high-proficiency learners feel emotionally more comfortable.” Teacher 4 mentioned that she provides more corrections at the lower rather than advanced levels, which was consistent with her practice (see Table 3). She believed that “[learners] do not know much of the language and are not discouraged by correction. In fact, they expect me to correct them.”

Overall, although the teachers provided more corrections for the lower proficiency group during the observations, their beliefs were more consistent with the results of empirical studies. The results of these studies (e.g., ^{Ammar & Spada, 2006}; ^{Kartchava & Ammar, 2014}; ^{Lyster, 2001}, 2004; Lyster & Ranta, 1997; ^{Mackey & Philp, 1998}; ^{Martoccio, 2017}; ^{Sheen, 2006}; ^{Trofimovich et al., 2007}) confirm the teachers’ claim that lower-intermediate learners have the lowest rate of implementation of feedback. Practice, however, does not necessarily follow exact beliefs. In fact, the teachers seemed to be mainly concerned about the practical aspects of providing feedback and learners’ emotional reaction to their correction which, according to them, influences their use of feedback. In other words, their differentiation between less and more proficient learners was mainly related to the different emotional reactions the teachers believed they show to correction; the more proficient the learners are, the less sensitive they will be to correction. ^{Vásquez and Harvey (2010)} and Kartchava et al. (2018) also found that the teachers in their teacher education program initially emphasized the affective dimension of error correction. ^{Roothooft and Breeze (2016)} came up with a similar finding. However, in the present study, concern about creating negative emotions, as well as the context, played a determining role; the reason for using fewer corrections for the advanced learners was that it was not always possible to interrupt the advanced learners due to their higher fluency.

Types of Feedback Preferred

The results related to the types of feedback used by the observed teachers, as presented in Table 4, indicate that the frequency of recasts is much higher in both proficiency groups and the teachers were less inclined to use explicit forms of correction. As one of the teachers noted: “Explicit forms of correction targeted at the individual students might make them feel exposed to others’ judgment.” ^{Rahimi and Zhang (2016)} also found that less experienced teachers did not favor explicit correction, elicitation, and metalinguistic feedback. Furthermore, the teachers generally used output-prompting feedback types (i.e., elicitation and clarification request), though much less frequently, for the advanced learners. This was consistent with the findings of studies (e.g., ^{Ammar & Spada, 2006}; ^{Li, 2014}; ^{Mackey & Philp, 1998}; Philp, 2003); elicitation has been found to be more beneficial to more proficient learners and less helpful for the low-proficiency learners; as ^{Long (2007)} puts it, elicitation requires self-correction which requires a level of knowledge that low-proficiency students lack. With regard to input-providing types of feedback (i.e., recast, metalinguistic feedback, and explicit correction), the teachers used recast and metalinguistic feedback almost equally for both groups but used explicit correction more for the lower-intermediates.

Individually, the observed teachers had different practices and beliefs (see Table 5). Out of the corrections Teacher 1 made for the lower-intermediate group, 81% was in the form of recast though she believed “recast is not beneficial to low-proficiency learners and even if they are corrected by recasting, they will not understand what was wrong with their utterance.” She also used metalinguistic feedback more for advanced learners though she preferred metalinguistic feedback (in belief) for the lower-intermediate group.

Teacher 2 used more recasts for the advanced learners; he preferred to use recast particularly for the lower-intermediate learners believing that “corrections should be made indirectly [using recast] and the feedback should be delayed no matter at what level of proficiency the learners are.” This was because he preferred “not to interrupt the learners during classroom discussion,” which made him rarely use other types of feedback.

Table 5 Distribution of Feedback Types Across Each Proficiency Group 

The feedback type highly preferred by Teacher 3 was also recast, which she believed “should be used mainly for higher-proficiency learners because they are better able to benefit from it due to their higher level of knowledge.” She also asserted that explicit correction is also more appropriate for advanced learners and might hurt lower-intermediate learners’ self-confidence. During the observations, however, as shown in Table 5, she used more recasts for the lower-intermediate learners. She used more explicit corrections for her advanced learners (14.29%) though. Teacher 3’s concern about her students’ emotional reaction was also reflected in her avoidance of “challenging” feedback types, that is, elicitation and clarification request (see Table 5).

Teacher 4 used recast for both proficiency groups. She used elicitation and clarification request only for advanced learners, though quite rarely, despite the fact that she mentioned in the interview that she does not differentiate between the two groups. She also used explicit correction very rarely and only for the lower group, because she believed it might make her students less willing to talk as “direct forms of correction might sound like reproaching to my students.”

Teacher 5’s reluctance to interrupt the learners was reflected in her use of recast due to its unobtrusive nature. This was more so, according to her, in the case of her less proficient students. She pointed out:

Less proficient students normally show a negative reaction to feedback. This might be partly because of their little learning experience; they have been less exposed to corrective feedback as compared to advanced learners…I prefer to use less correction and indirect forms of correction.

This was observed in her practice; she used fewer corrections and mainly recast for correcting the lower-intermediate group. She used elicitation and more explicit correction for the advanced group but more metalinguistic feedback for the lower-intermediate learners.

The results were generally indicative of the dominance of recast irrespective of the learners’ proficiency level. This result is consistent with the findings of previous studies (^{Kamiya, 2016}; ^{Lee, 2013}; ^{Lyster & Ranta, 1997}; ^{Roothooft, 2014}; Roothooft & Breeze, 2016; ^{Zyzik & Polio, 2008}), which also found that recast is the most frequently used feedback type among teachers. According to the results of the present study, there was no convergence between the teachers in this respect; some used it more for the advanced (e.g., Teacher 2) and some for the lower-intermediate group (e.g., Teachers 1, 3, and 5).

The reasons behind the choice of recast were similar to those of the teachers in the previous studies (e.g., ^{Kamiya, 2016}; ^{Mori, 2011}; ^{Roothooft, 2014}; ^{Zyzik & Polio, 2008}) but different for the two proficiency groups.

Based on the findings of the present study, in the case of lower-intermediate learners, the teachers used recast considering its emotional advantages; it does not interrupt, challenge, or embarrass the lower-intermediate learners. For advanced learners, however, the teachers used this feedback type mainly because of its practical advantages and unobtrusive nature; because advanced learners talked more fluently, the teachers preferred to use recast in order not to interrupt them in the middle of conversation. Most of the teachers were also aware of the disadvantages of recast and mentioned, though on the periphery, its ambiguous nature-particularly for the lower-intermediate learners-and believed it does not benefit them because they do not notice it. However, its practical advantages and the teachers’ great concern about the learners’ emotional reaction, as in the previous studies (^{Kamiya, 2016}; ^{Roothooft, 2014}; Roothooft & Breeze, 2016), made them ignore its disadvantages. This concern was higher in the case of lower-intermediate students as the teachers used recast even more frequently with them.

Another reason for the overuse of recast for the lower-intermediate learners might be the teachers’ misconception about how learners might negatively react to explicit forms of correction. This, as pointed out by ^{Borg (2003)}, might have something to do with the teachers’ previous experience as L2 learners. They might avoid using the techniques they would not have favored themselves as language learners. Explicit correction, nevertheless, does not always result in negative emotions. ^{Tavakoli and Zarrinabadi (2016)} found that explicit feedback not only does not produce a negative response but promotes learners’ willingness to talk and even decreases their anxiety and increases their communication competence. ^{Yang (2016)}, as another instance, discovered that learners generally favored explicit correction and metalinguistic feedback on all types of errors.

In the case of the advanced learners, however, as it was already mentioned, the main reason for the overuse of recast and less frequent use of other more explicit forms of correction was the fact that advanced learners were more fluent in speaking and it was not possible to interrupt them as much in the middle of conversation. In fact, the teachers did not favor interrupting the learners using explicit forms of correction when they were talking, which they believed, would be discouraging and make them lose track of what they were trying to say. Another reason, however, might be the high frequency of errors. According to ^{Mori (2011)}, the more frequent the errors are, the more implicit the corrections become. Overall, it seemed that the teachers preferred to use this feedback type due to its emotional and practical advantages for lower and higher proficiency groups, not because it produces better learning; it does not challenge or embarrass the learners and it does not interrupt them in the middle of conversation.

The findings were also indicative of a lack of correspondence between the teachers’ views and their corrective practice with the teachers preferring more feedback for advanced learners but using more corrections for the lower-intermediate group and pointing to the ineffectiveness of recasting less proficient learners’ errors but still using it. This finding is in line with the findings of other studies (e.g., ^{Basturkmen, 2012}; ^{Roothooft, 2014}). In her review on the congruity/incongruity of teachers’ beliefs and practices, Basturkmen (2012) reported the lack of a direct consonance between beliefs and practice, finding that contextual barriers and curricular constraints impeded the connection between beliefs and practices. This finding seems to support the claim made by ^{Borg (2003}, 2010) in the sense that context makes a cardinal contribution to shaping teachers’ thought processes. In the case of the present study, the conversation-based activities and also the indirect nature of recast, which makes it less embarrassing, made it practically an ideal choice for correcting the errors.