Corrective feedback, learner uptake, and feedback perception in a Chinese as a foreign language classroom

The role of corrective feedback in second language classrooms has received considerable research attention in the past few decades. However, most of this research has been conducted in English-teaching settings, either ESL or EFL. This study examined teacher feedback, learner uptake as well as learner and teacher perception of feedback in an adult Chinese as a foreign language classroom. Ten hours of classroom interactions were videotaped, transcribed and coded for analysis. Lyster and Ranta’s (1997) coding system involving six types of feedback was initially used to identify feedback frequency and learner uptake. However, the teacher was found to use a number of additional feedback types. Altogether, 12 types of feedback were identified: recasts, delayed recasts, clarification requests, translation, metalinguistic feedback, elicitation, explicit correction, asking a direct question, repetition, directing question to other students, re-asks, and using L1-English. Differences were noted in the frequency of some of the feedback types as well as learner uptake compared to what had been reported in some previous ESL and EFL studies. With respect to the new feedback types, some led to noticeable uptake. As for the students’ and teacher’s perceptions, they did not match and both the teacher and the students were generally not accurate in perceiving the frequency of each feedback type. The findings are discussed in terms of the role of context Tingfeng Fu, Hossein Nassaji 160 in affecting the provision and effectiveness of feedback and its relationship to student and teacher perception of feedback.


Introduction
In recent years, the role of interactional feedback in second language classrooms has received considerable research attention.Interactional feedback refers to feedback that learners receive on their erroneous utterance in the course of communicative interaction (Nassaji, 2009).A number of descriptive and experimental studies have examined both the provision and usefulness of such feedback in classroom settings (e.g., Ellis, Basturkmen, & Loewen, 2001;Loewen & Philp, 2006;Lyster & Ranta, 1997;Panova & Lyster, 2002;Sheen, 2006;Suzuki, 2004;Williams, 2005;Zhao & Bitchener, 2007).This research has provided substantial information about the various types of feedback teachers use and also the extent to which such feedback contributes to language development.However, most of this research has been conducted in English teaching settings.Many researchers have argued that context of interaction may make a difference in how feedback is used and assists L2 learning.The goal of the present study was to extend research in this area by examining teacher feedback and learner uptake in an adult Chinese as a foreign language (CFL) context.The study also examined learner and teacher perception of feedback types and feedback frequency in this context.This question was motivated by a scarce number of studies of interactional feedback on feedback perception and in particular the extent to which teacher and student perception affect feedback effectiveness.

Previous descriptive studies of classroom feedback
When learners make an error in the classroom, the teacher may decide to correct the error, and he or she may have many options.Classroom feedback studies have identified a number of these feedback types and have also examined what kind of effects they have on students' uptake and learning.As for the effect of feedback, they have shown positive effects for feedback in general, but they have also found different results for the distribution of different feedback types and their effects on learner uptake and learning in different instructional contexts.
One of the first studies that provided a detailed account of classroom feedback is Lyster and Ranta (1997), which investigated four communicative French immersion classes of Grade 4 and 5 students.Among the six identified feedback types-recast, elicitation, clarification request, metalinguistic feedback, explicit correction, and repetition-recasts were the most frequent, making up 55% of all six feedback types, followed by elicitation, accounting for 14%, followed by clarification request (11%).The other three types of feedbackmetalinguistic feedback, explicit correction, and repetition-each achieved less than 10% of the frequency.Panova and Lyster (2002) conducted a similar study in an ESL context.Their study found similar results.Recasts accounted for 55% of all feedback instances, which is the same percentage as that found in the 1997 study, and they found the same percentage for clarification request, which accounted for 11%.One difference was that elicitation only accounted for 4% of feedback occurrences compared to 14% of the original study.Metalinguistic feedback, explicit correction, and repetition were similarly low in frequency.Suzuki (2004) examined ESL classes with intermediate level adult learners and three teachers.Recasts were the most frequent feedback type (60%), followed by clarification requests.However, the percentage of clarification requests was 30%, which was much higher than the other two studies (11%).The other feedback types found were metalinguistic feedback, elicitation, explicit correction, and repetition, which occurred rather infrequently, not more than 5% each.What is significant in this study is the uptake rate; students tended to respond to teacher feedback almost all the time (97%), and recasts led to much more repair (66%) than those in Lyster and Ranta's study (18%).The successful repair rate (54%) was much higher than those cases that were still in need of repair (43%).Jimenez's (2006) study examined feedback in two Italian EFL classrooms at two different levels of language proficiency.This study found a high level of peer interaction with recasts being the most frequently used (37.8% and 38.3% in each class).These rates for recasts are relatively low compared to previous studies.Yoshida (2010) examined feedback in a second-year university level Japanese language course.Results showed that recast was the number one feedback move, which occurred 47 times and accounted for 51% of all moves.
Studies have also compared feedback in different contexts and have found an important effect for context.Lyster and Mori (2006), for example, examined teachers' feedback in two different immersion contexts: a Japanese and a French immersion context.The results showed differences in the distribution of learner uptake in the two contexts, showing a higher rate of uptake following recasts in the Japanese context but a higher rate of uptake following prompts in the French immersion context.Sheen (2006) also found differences related to context, comparing four observational studies of classroom feedback (a French immersion class, ESL classrooms in New Zealand, ESL classrooms in Canada, and an EFL classrooms in Korea).Sheen found less effect for recasts to generate uptake in the Canadian ESL and the French immersion contexts than in the Korean EFL contexts.Meta-linguistic studies of feedback have also shown a significant effect for context.Mackey and Goo (2007) for example found a larger effect for recasts in EFL contexts than ESL contexts.These findings suggest that context plays an important mediating role in the effect of feedback.As can be seen, the focus of studies has been more on English language learning, and fewer studies have examined feedback in other language classrooms.Thus, more research on feedback is needed in non-English language teaching contexts.

Studies of feedback perception
In recent years, in addition to provision and effectiveness of interactional feedback, studies have also begun to investigate learners' perception of feedback.This issue is important because if learners do not perceive the corrective nature of feedback and if their perception differs from their teachers' intention, they may not benefit from feedback (Amhrein & Nassaji, 2010).Due to the ambiguity of recasts, some researchers have suggested that recasts may not be very effective because learners may not perceive them as feedback on form but on content (e.g., Lyster, 1998;Lyster & Ranta, 1997).Thus, a number of studies have looked in particular at learners' perception of recasts.One of these studies is Egi (2007a), which examined perception of recasts and its relationship with error types and feedback characteristics.Learners noticed 60% of morphosyntactic recasts and 57% of lexical recasts.This indicated that learners were fairly accurate in perceiving the target of recasts.This of course was different from the findings of some other studies that have shown that recasts targeting lexical errors were more easily noticed than morphosyntactic errors (Mackey, Gass, & McDonough, 2000).The author attributed the high rate of noticing of morphosyntactic recasts to the intensity of recasts in this study; in fact, recasts focused on only two morphosyntactic items.The learners were more successful in noticing shorter recasts with fewer changes.The author attributed this finding to learners' limited attention.It was concluded that different degrees of saliency of recast can challenge learners on different cognitive levels, leading to different learner perceptions.
With the same participants and a similar methodology, Egi (2007b) examined learners' L2 development in relationship with their perceptions of recast.The performance in L2 development was measured by tailor-made tests that specifically targeted learners' errors during the treatment.Both immediate and delayed posttests showed that learners obtained most gains when they perceived recast as a combination of positive and negative evidence.This might indicate that the positive evidence of recasts for lexical errors may cause interlanguage changes more effectively than the evidence for morphosyntactic errors.Carpenter, Jeon, MacGregor, and Mackey (2006) examined whether learners could tell recasts from repetitions.Video clips captured task-based dyadic activities when recast or repetition was provided to advanced ESL learners.The results showed that learners who saw the initiating erroneous utterances were more successful in identifying recasts, but they showed no advantage in distinguishing between recast and repetition.This suggested that the utterance-response context might have enhanced the salience of a recast, but recasts remained ambiguous in their corrective nature and therefore were frequently perceived as mere repetition.Kim and Han (2007) examined teachers' intention, learners' perceptions, and learners' accuracy of perceiving recasts as corrective or communicative.A high level of recast awareness was explained by the consistency of providing recasts as the only kind of corrective feedback.However, whether recasts were self-directed or other-directed did not have any effect on learners' perceptions.What we can learn from this study and other studies discussed above is that aspects of feedback such as length, intensity, error types and the context in which feedback is provided can affect learners' perception and subsequently feedback effectiveness.

The present study
The above research has provided substantial information about the use of various types of feedback, learner uptake, and teacher and learner perception of feedback.However, as noted earlier, most of this research has been conducted in English teaching settings.The goal of the present study was to extend research in this area by examining teacher feedback in an adult CFL context.As can be noted from the above studies, research seems to suggest that learners' and teachers' perception of feedback is also a factor contributing to feedback effectiveness.Thus, the study also investigated the perceptions of feedback frequency and feedback types to see whether there is any relationship between learner and teacher perceptions of feedback and the actual occurrence of the feedback.
The study addresses the following research questions: 1. How often does the teacher provide feedback in an adult CFL classroom?2. What types of feedback does the teacher use and what is their relationship with learner uptake?3. How do the students and the teacher perceive the use and nature of each feedback type, and to what extent does their perception correspond to the actual use of feedback?

Teaching context and participants
This study was conducted in an intermediate level CFL class in a university context.The teaching method used by the teacher was a combination of task-based learning and form-focused instruction.The average age of the students was 20.
The students were of varying first language backgrounds: English (8), Japanese (1), Thai (1), Mandarin Chinese (1), both French and English as first languages (1), and both Cantonese and English as first languages (1).Students were largely female and the gender ratio was 10 (female) to 3 (male).On average, eight students were present at each lesson.Classes were held three times a week.They were 50-minute long each, and data were collected through both video-taping and surveys.The teacher was a native-speaker of Mandarin with eight years of Mandarin-teaching experience.She had also taught other courses including Chinese linguistics and SLA teaching methodology.At the beginning of the course, the teacher and the students were not informed of the research focus being teacher feedback and learner uptake.

Procedures
The data regarding feedback were collected through video and audio recordings of classroom interaction.Altogether, thirteen 50-minute sessions (10 hours) of classroom interaction were recorded.A questionnaire contained 10 questions asking both the teacher and the learners about the use and frequency of each feedback was used.The questionnaire was administered immediately after classroom observations.In order not to make the participants aware of the focus of the study, the questionnaire was not administered in the first seven classes.For the students, the question was about their perceptions of feedback they had received; for the teacher, it was about the perceptions of feedback she had given.

Analysis
The video recording was fully transcribed.Both English and Chinese were used in the transcription.For example, when the teacher was explaining sentence structures or making comparisons between L1 and L2, she used English to make the explanation more accessible to the students.When she was providing sample sentences or relating the current content to previous learning, she used Chinese to "push" the students to process the target language before she explained further.When transcribing Chinese, pinyin was used for mispronounced words.An example of a feedback episode is shown here:

Coding feedback types
A feedback move is issued by the teacher and starts immediately after the student(s) made an error in their target language use, and the move ends when the teacher finishes her explanation.A move can be a simple phrase (e.g., recast) or an extended explanation (e.g., metalinguistic feedback).To code the feedback types, the study initially used Panova and Lyster's (2002) model of seven feedback types, which includes recast, clarification request, translation, metalinguistic feedback, elicitation, explicit correction, and repetition.However, an initial analysis of the classroom interaction revealed that the teacher used a number of new corrective feedback moves.Based on the initial analysis, a framework consisting of 12 feedback types was developed and used.The framework is shown in Table 1, followed by a brief description of each of the feedback types and an example (with English translation).

Explicit correction
A feedback move was coded as explicit correction when the teacher used explicit correction to signal to the student that he/she had made an error.The following shows an example: Example 7
T: Ah? Rice.Not this one."Rice" is without this radical.à Explicit correction Here, the student made an error on a Chinese character.The teacher then explicitly pointed out that the student missed a radical in the character.

Repetition
The teacher sometimes repeated the student's erroneous utterance with a raising intonation to highlight the error: Example 8

Original interaction Translation
Elliot: 他爱自己的国家，uh, something.Elliot: He loves his country, uh, something.T: Something?T: Something?à Repetition 4.1.4.8.Re-asks In addition to repeating the learner utterance, the teacher was also found to repeat a question in a heightened tone after the student produced an erroneous utterance following feedback (e.g., Yoshida, 2010).Re-asks was different from repetition or asking a direct question in that re-ask is a repetition of the teacher's question after an initial feedback turn.The purpose of a re-ask is to repeat the original question as sometimes the students might miss the point of a question while they were in fact capable of answering the question: Example 9 Original interaction Translation T: How would you say the, "three times a year"?T: How would you say the, "three times a year"?Ss: 三年… Ss: Three years… T: "Three times a year!" T: "Three times a year!"à re-ask Elliot: Oh, ⼀年三次.
Elliot: Oh, a year three times.

Translation
The teacher occasionally translated students' L1 utterance into the target language, highlighting the comparison between the two languages and encouraging the students to use the target language for the expression's future use: Example 10 Original interaction Translation Emma: No I did half of it.
Emma: No I did half of it.

Asking a direct question
The teacher can ask a direct question about how to form a specific expression in the target language (e.g., "How do you say that in French?").Similar to elicitation, the aim is to elicit the correct form and ask the students to re-try, adopting an explicit approach: Example 11 Original interaction Translation Teacher: Oh, you want to travel all over the world.
Teacher: Oh, you want to travel all over the world.
Eveline: And help other people.Eveline: And help other people.
Teacher: 或者帮助别人，或者说，how to say "travel all over the world"?
Teacher: Or help other people, or to say, how to say "travel all over the world?" à Asking a direction question Eveline: 去… Eveline: Go …

Directing question to other students
When a student made an error, the teacher sometimes sought a correct response from other students: Example 12 Original interaction Translation Emma: Actually I want to say, old man would be "⽼伙⼦"?Emma: Actually I want to say, old man would be "old fellow"?T: How do you say "old man"?(looking toward Elliot) T: How do you say "old man"?(looking toward Elliot) à Direction question to other student Elliot: (Instantly) ⽼⼈。 Elliot: (Instantly) Old person.
4.1.4.12.Using L1-English "Using L1-English" was used when the teacher felt that it would be easier for the students to understand a certain structure if it was explained in English: Example 13 Original interaction Translation Pinky: uh, jīn chán?
This type of feedback was different from "translation," where the teacher translates the student's English into the target language.When the students were tired or overwhelmed, the teacher used English to lighten up the cognitive load.
In Example 13 Pinky mispronounced the name of a temple.The teacher corrected her pronunciation and at the same time provided the translation of the word "temple" in English to facilitate understanding.

Coding uptake
To code uptake, we used the definition by Lyster and Ranta (1997) and Ellis et al. (2001).We categorized uptake into successful and unsuccessful uptake.An uptake was thus defined as "a student's utterance that immediately follows the teacher's feedback and that constitutes a reaction in some way to the teacher's intention to draw attention to some aspect of the student's initial utterance" (Lyster & Ranta, 1997, p. 49).Successful uptake referred to a student's successful correction of the error after teacher feedback.Unsuccessful uptake referred to a student's partial or off-target correction of an error after receiving teacher feedback.The following episode is an example of successful uptake: coded by another coder and there was 97.5% agreement on feedback types, 100% on uptake, and 100% on uptake types.

Feedback and uptake frequency
The first research question investigated the frequency of teacher feedback.A total of 192 feedback episodes were observed during the 10 hours of classroom interaction.A total of 245 feedback moves were identified to have occurred in these episodes.Students made a total of 285 errors of which 194 received teacher feedback.Thus, the teacher provided corrective feedback to 68.1% of the students' errors.As noted earlier, 12 feedback types were identified.Table 2 shows these feedback types and their percentages.Recast was the most frequent feedback type, accounting for 56.7% of all feedback moves.The second most commonly used feedback type was metalinguistic feedback (10.6%) followed by translation and explicit correction, which together accounted for 7.3% of all feedback moves.The remaining seven types-delayed recast, clarification request, asking a direct question, repetition, directing question to other students, re-asks, and using L1-English-ranged from 0.4% to 3.3%, accounting for 11.1% of all feedback moves.
Table 3 shows learners' uptake following each feedback type.In general, 59% of all feedback moves led to student uptake, and 45.3% of teacher feedback led to successful repair.The uptake rate and the repair rate were much higher than that of Panova and Lyster's study, where only 47% of feedback resulted in student uptake, and only 16% of feedback resulted in learner repair.Due to the great number of recasts, it is not surprising that recasts resulted in the greatest number of learner uptake and repair.However, if we look at the percentage of learner uptake and repair resulting from recasts, it was not the highest.49.6% of all recasts led to learner uptake, whereas 45.3% of all recasts led to successful uptake.Elicitation achieved the highest percentage of learner uptake (94.1%).Of all 17 elicitation moves, 16 resulted in student uptake, and 64.7% (11 out of 17) led to successful uptake.Following elicitation, explicit correction ranked second in facilitating student uptake: 88.9% (16 out of 18) resulted in student uptake, and 61.1% (11 out of 18) led to successful uptake.Metalinguistic feedback was the next best technique.53.8% (14 out of 26) of metalinguistic feedback moves resulted in student uptake, but only 19.2% (5 out of 26) led to successful uptake.Translation had a 50% (9 out of 18) uptake rate, and 33.3% (6 out of 18) resulted in successful uptake.Due to the small numbers of occurrence (1 time to 8 times) of the remaining feedback types, their uptake and repair rates were not as informative.

Students' and teacher's perceptions of feedback frequency
Another research question concerned the students' and the teacher's perception of the use and frequency of feedback.To answer this question, the survey asked the students to indicate their level of agreement with answers to questions such as "How often do you think the teacher provided feedback to your/peers' errors?"Each option was assigned a number to indicate the frequency that the students perceived (e.g., 5 means "always used this feedback type," 4 means "used this type 75% of the time," etc.).The options with their corresponding numbers are listed in Table 4.The results of the students' mean, the teacher's mean, as well as the actual feedback frequency for lesson 10 to lesson 13 are presented in Table 5.From the table, we can see that feedback was provided to 64.9% of all student errors.The average of the teacher's perception rating of feedback frequency was 4. In other words, the teacher believed that she had given feedback to 75% of all student errors.The students' mean was 4.7, which corresponded to roughly 92.5%.The students thought that the teacher had given them feedback on a very frequent basis.In comparison, the teacher's perception of feedback frequency was much closer to the actual frequency.
The results were obtained from question 3 to question 9 of the survey that asked how often the teacher provided each type of feedback (e.g., "Did the teacher say the correct form after you/your peer made an error?").Table 6 gives an overview of the actual frequency of seven feedback types, and the students' rating and the teacher's ratings of feedback frequency.The actual frequency rank of each feedback type, together with the students' and the teacher's perception ranking, are presented in Table 7.  From the rankings, we can see that explicit correction was perceived by the teacher to be the most frequently used feedback type.In actual teaching, explicit correction and metalinguistic feedback ranked fourth.Following explicit correction, the teacher perceived recast to be the second most frequently used type.The teacher was quite accurate in perceiving recast, elicitation, metalinguistic feedback, and clarification request, but not so accurate in perceiving explicit correction, repetition, and translation.
The students perceived recast to be the most common type of feedback.Following recast, metalinguistic feedback was perceived to be the second most frequent type.However, in actual teaching, metalinguistic feedback ranked fourth.Students were quite accurate in perceiving recast, explicit correction, and repetition, but not so accurate in perceiving metalinguistic feedback, clarification request, translation, and elicitation.

Discussion
This study investigated the provision of teacher feedback and learner uptake in an adult CFL context.The study also examined learner and teacher perception of feedback.Initially, Lyster and Ranta's coding system for feedback moves which involves six types of feedback was used in indentifying the type of teacher's feedback and learner uptake, However, the study also identified additional feedback types used by the teacher.Altogether, it identified 12 types of feedback used by the teacher in this instructional setting: recast, delayed recast, clarification request, translation, metalinguistic feedback, elicitation, explicit correction, asking a direct question, repetition, directing question to other students, re-asks, and using L1-English.Many instances of multiple feedback episodes were also observed, including recast + metalinguistic feedback and metalinguistic feedback + explicit correction.The taxonomy of feedback types identified in this study adds to those identified in previous research on classroom corrective feedback (e.g., Lyster & Ranta, 1997;Panova & Lyster, 2002) and thus can be used as an extended framework to examine feedback in L2 classrooms.It also suggests that teachers have more options to correct learners' errors than what has been frequently discussed in previous studies.The frequency of the feedback types and learner uptake will now be discussed and compared with previous studies.
The results of this study showed that throughout the 10 hours of recorded classroom interaction the teacher made a feedback move every two and a half minutes.Comparing this rate with Panova and Lyster's (2002) result of 48%, this percentage of correction was 20% higher.One reason for this difference could be learners' level of language proficiency.As Panova and Lyster mentioned, their students were at the beginner level.Therefore, the lower level of proficiency might have caused less frequent interaction between the teacher and the students.In the present study, the students were more advanced and hence had already possessed adequate speaking and reading skills.This difference then points to the role of language proficiency in the provision of feedback in L2 classrooms.Another reason could be the teacher's teaching style and the fact that the classroom was a foreign language classroom.When observing the classes, it was found that the teacher frequently called on certain students to either do a quick comprehension check or to ask them to read a paragraph out loud.This could have generated more opportunities for the students to produce errors and therefore more chances of correcting them.
As for the provision of feedback types, out of 245 feedback moves that occurred in 10 hours of recorded interaction, recasts occurred 139 times, accounting for 56.7% of all moves.Following recast, metalinguistic feedback accounted for 10.6% of all feedback moves.Translation and explicit correction each occurred 18 times, ranking third with the frequency of 7.3%.Explicit correction was much more favored in the current study.When looking at the multiple feedback episodes, the most common combination for multiple feedback was recast + metalinguistic feedback.The second most common was metalinguistic feedback + explicit correction, which occurred three times.Single explicit correction occurred 18 times.These results show that the teacher tended to use a more explicit feedback style in her teaching in this Chinese L2 classroom.This might be attributed to the nature and the goal of the course and also the textbook used.The lessons were more grammarfocused and the teacher closely followed the textbook chapters.The students were told that they would be tested on the grammar points in the chapters.Therefore, the teacher could have felt obligated to devote class time to analyze morphosyntactic structures as well as to overtly correct students' nontargetlike utterances.
The least used feedback type in the current study was clarification requests, which occurred only 3 times.Clarification requests were often used when the teacher intended to give the student a second try or when the teacher did not understand the meaning that the student tried to convey (Lyster & Mori, 2006;Lyster & Ranta, 1997;Panova & Lyster, 2002).Therefore, clarification request seemed to be less needed in this intermediate level course where the chances for the teacher not understanding the students' meaning were slimmer.Moreover, there was not as much student-initiated meaning-focused discussion, which would have probably generated more communication breakdowns.
With respect to uptake, as presented in Table 3, two of the three new feedback types unique to this study had promising successful uptake rates even though their occurrences were few: directing question to other students (100% successful uptake in 3 instances) and using L1-English (100% successful uptake in 1 instance).This result is not surprising given the explicit nature of instruction and the classroom focus on grammar.Of all other feedback types, the feedback type that led to the highest successful uptake rate was clarification request (66.7%).Elicitation was the next most successful move, with a successful uptake rate of 64.7%.Following elicitation, explicit correction led to 61.1% of successful uptake.In Lyster and Ranta's (1997) study, the top three feedback types that generated the most amount of successful uptake were elicitation (46%), metalinguistic feedback (45%) and explicit correction (36%).Panova and Lyster's (2002) top three were recast (29%), elicitation (11%) and clarification requests (10%).The current study differed from Panova and Lyster's study the most in that the successful uptake rates were much higher and occurred with the more explicit feedback types.One reason could be that the teacher laid overt emphasis on forms in question, especially when the student engagement level was low.
However, the relatively high uptake rate following explicit types of feedback agreed with previous studies (Ellis, 2011;Lyster & Saito, 2010).This suggests that teachers could consider more explicit techniques when correcting students' errors, especially when the purpose is to raise the level of participation and learner attention.Also, the mismatch between the students' expectation and the actual focus of a language course could be discouraging for the students.
One of the research questions was about learners' and teacher's perception of feedback.The results showed that the biggest difference in frequency was in elicitation and asking a direct question.Elicitation occurred 4 times and 13 times before and after the survey, increasing from 23.5% to 76.5%.Asking a direct question had a bigger increase from 0% to 100% (0 to 8 times); all of its instances occurred after the survey began.This might be due to the possibility that the teacher started to pay more attention to her error correction strategies.On the contrary, delayed recast, clarification request, metalinguistic feedback, explicit correction, repetition, directing question to other students, and using L1-English, all creased in frequency after the survey began.We do need to keep in mind that the occurrences were very few.Metalinguistic feedback and explicit correction were the two types that both decreased by 30%.Both types were explicit and, as mentioned before, both types might have created opportunities for the teacher to elaborate on certain grammar points.It could be true that the teacher had begun to cut back on feedback that would lead to teacher-centered grammar explanations; instead, she called on students to elicit more student-centered discussions.
Also, the teacher had a much more accurate perception of feedback frequency than the students.The teacher provided corrective feedback to the students' errors 64.9% of the time.While the teacher perceived the frequency to be 75%, the students thought it was much more frequent (92.5%).This might be due to the salient nature of the more explicit feedback types.Also, the teacher often called on students, promoting noticing when a mistake was made.Similarly, other students might have been watching and engaging in a similar way while their peers were struggling with a certain question.
Even though the teacher's perception of feedback frequency was more accurate than that of the students, her perceived frequency was still higher than the actual frequency.This could have also been due of the focus of the lesson, which was for the most part on grammar.This could have affected the teacher's perception of feedback and its frequency.
Among the different feedback types, both the teacher and the students were quite accurate in perceiving recast, which accounted for more than half of the teacher's feedback moves.This may be attributed to the fairly explicit nature of recasts.This result is important as it shows that the students' and the teachers' perception of recasts depends on the nature and the context of recasts.

Conclusions
In this study, the focus of grammar teaching could have potentially contributed to the high amount of explicit feedback.Students' proficiency level could have been another factor contributing to a high feedback rate.More proficient students were more capable of producing target language forms and reacting to teacher feedback, which in turn facilitated more teacher feedback.In addition, more explicit feedback types led to the highest amount of learner uptake, and the three newly identified feedback types-asking a direct question, directing question to other students, and using L1 English-had promising uptake and successful uptake rates.Elicitation, metalinguistic feedback, and explicit correction were often used along with recasts to draw students' attention to an error, especially when the level of classroom participation was low.Overall, the students produced uptake following 59.1% of teacher correction.We can conclude that the corrective nature of teacher feedback was frequently noticed by the students.
In perceiving the frequency of providing feedback, the teacher was more accurate than the students.However, the actual frequency was lower than what the teacher and the students perceived.This could have been due to a large number of errors that passed without being noticed or were noticed but were not corrected.In perceiving the frequency of each of the seven feedback types, neither the teacher nor the students were accurate.This could be partially explained by the cognitive demand that the survey had imposed on the teacher and the students.It might have been difficult for them to recall how the correction was made while they were concentrating on correcting the errors.
Finally, the three new types of corrective feedback identified in this study: asking a direct question, directing question to other students, and using L1-English, are all explicit types of feedback.Although the new types occurred only a few times, the uptake rate was promising.This indicates that the explicitness of feedback types might be related to certain classroom dynamics: Quieter classrooms might need more overt corrective techniques.Teachers should consider these techniques or develop new explicit forms of feedback in their classrooms.
There are a few limitations of this study that should be considered when interpreting the results.The first limitation is that the data came from one class.Future research can examine the same questions in other CFL classes.Another limitation is that although one of the questions in this study was about the learners' and the teacher's perception of feedback, this was more about the frequency of the different feedback types and not about how participants interpreted a particular feedback type.To have evidence for the interpretation of feedback, studies with a stimulated recall design are needed (Carpenter et al., 2006;Egi, 2007aEgi, , 2007b)).
In this study, the students' uptake was coded to be either successful uptake, unsuccessful uptake, or no uptake.However, as observed in the data, there were also some instances when the students were not given enough time after teacher feedback.Very often, the teacher corrected several errors in one turn, making it difficult for the students to identify and repair each error.The opportunities were not captured using this coding scheme and thus a more fine-tuned coding scheme for uptake instances is needed.

Table 1
The framework used to code the feedback types One of the feedback types that was identified was recast, which was an immediate reformulation of the learner's erroneous utterance: T: Yeah! 很好。这一次好不容易，好不容易 is a fixed structure to indicate it's very difficult.Same as 不容易.才找到一家挺 (tǐng) 不错的公司。 T: Yeah! Very good.This time, quite not easy, quite not easy is a fixed structure to indicate it's very difficult.Same as not easy.(Untilquitelatehe) found a quite (tǐng) good company.àDelayedrecast4.1.4.3.Clarification requestsThe teacher also used clarification requests, which was defined as a feedback move that requested clarification when the teacher sought meaning-or formrelated clarification after a student made an error:Ron: 对不起 (duì bù chǐ) Ron: Sorry (duì bù chǐ) T: 对不---T: Duì bù ---à Elicitation Ron: 起(qǐ) Ron: (qǐ)

Table 2
Frequency of each feedback type

Table 3
General student uptake following each feedback type

Table 4
Options for rating feedback frequency

Table 5
The students' and the teacher's perceptions of feedback frequency (means) and the actual feedback frequency (percentage)

Table 6
The students' and the teacher's perceptions, and the actual frequency of seven feedback types