Reducing international graduate students’ language anxiety through oral pronunciation corrections

Reducing international graduate students’ language anxiety through oral pronunciation corrections

System 56 (2016) 78e95 Contents lists available at ScienceDirect System journal homepage: www.elsevier.com/locate/system Reducing international gra...

4MB Sizes 0 Downloads 35 Views

System 56 (2016) 78e95

Contents lists available at ScienceDirect

System journal homepage: www.elsevier.com/locate/system

Reducing international graduate students’ language anxiety through oral pronunciation corrections Eun Jeong (Esther) Lee Claflin University, 400 Magnolia Street, Orangeburg, SC 29115, United States

a r t i c l e i n f o

a b s t r a c t

Article history: Received 21 April 2015 Received in revised form 23 October 2015 Accepted 23 November 2015 Available online xxx

This mixed methods study examines the relationships between teachers' oral corrective feedback (CF) and changes in international graduate students' language anxiety levels. The participants were sixty advanced-level adult ESL students, all of whom were training to be teaching assistants (TAs) at a large Midwestern US university. The data were collected through classroom observations, survey questionnaires (pre- and post-surveys), and indepth follow-up interviews with 40 of the 60 students. The study revealed that most of the teachers' oral CF had positive impacts on the students' affective variables, specifically lowering their anxiety about speaking English. However, certain clarification requests increased the students’ anxiety instead. These results highlight the potential affective risks of CF, and the study suggests how ESL instructors can use CF to encourage or even create positive emotional states in their ESL students. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Oral Corrective Feedback Advanced-level adult ESL students Second Language (L2) anxiety Clarification request Oral English proficiency

1. Introduction Many studies suggest that future research on corrective feedback (CF) account for individual learner differencesdthose “likely to influence whether and how learners process CF” (Sheen, 2011, p. 159)dto assess CF's mechanisms and wide-ranging effects (Kim, 2004; Mackey, 2006; Rezaei & Mozaffari, 2011; Russell, 2009; Yang & Lyster, 2010). Despite considerable scholarly attention to how these differences influence L2 acquisition and development (Horwitz, 2001; MacIntyre, 1999; Woodrow, 2006; Zheng, 2008), though, only a few studies have considered the associations between ESL learners' affective (i.e. emotional) variables and teachers' oral CF (e.g. DeKeyser, 1993; Sheen, 2007, 2008). Therefore, the present study specifically extends prior research on oral CF and L2 learner anxiety, applying Horwitz's (2001) observation that “language anxiety is fundamental to our understanding of how learners approach language learning, their expectations for success or failure, and ultimately why they continue or discontinue study” (p. 121). Rather than simply confirming that students with low anxiety are likely to learn English faster or have more positive attitudes toward speaking English, this study takes a different approach: it traces how teachers' oral error corrections affect adult ESL students' existing anxiety levels, and how interactions between CF and anxiety help or hinder students' oral English development.

E-mail address: eulee@claflin.edu. http://dx.doi.org/10.1016/j.system.2015.11.006 0346-251X/© 2015 Elsevier Ltd. All rights reserved.

E.J.(E. Lee / System 56 (2016) 78e95

79

2. Literature review 2.1. The effects of CF on L2 oral production Several recent studies of adult ESL students (Author, 2013; Han & Jung, 2007; Panova & Lyster, 2002; Suzuki, 2004) have shown positive associations between oral corrective feedback, learner repair, and uptake in adult ESL classrooms. Most of these studies use Lyster and Ranta's (1997) taxonomy of CF, which identifies six feedback moves: recasts, explicit correction, clarification requests, metalinguistic feedback, elicitation, and repetition. Studies following this approach have assessed the outcomes of certain types of CF on L2 development. For instance, Ellis, Loewen, and Erlam (2006) compared the efficacy of implicit CF (recasts) and explicit CF (metalinguistic feedback), while Ellis (2007) examined the extent to which recasts and metalinguistic feedback influenced the effects of CF on various grammatical structures. Likewise, Yang and Lyster's (2010) study compared the effectiveness of recasts and prompts (including clarification requests, metalinguistic feedback, elicitation, and repetition), and Chu's (2011) study juxtaposed output-promoting CF (clarification requests, metalinguistic feedback, elicitation, and repetition) and input-providing CF (explicit correction and recasts). While these studies offer useful tools for targeting CF to certain tasks or certain students, they ignore how oral CF might influence L2 learners' anxiety, which can influence their oral English development both positively and negatively. Language anxiety often serves as a catalyst for other affective variables among adult ESL learners. MacIntyre (1999) defined language anxiety as “the apprehension experienced when a situation requires the use of a second language with which the individual is not fully proficient … the propensity for an individual to react in a nervous manner when speaking, listening, reading, or writing in the second language” (p. 5). Similarly, Sheen (2011) reminded us that “foreign language anxiety (as a highly specific kind of situational anxiety) constitutes an important individual variable that is related causally to various L2 criterion measures” (p. 135). Sparks, Patton, Ganschow, and Humbach (2009) affirmed that “proponents of the language anxiety hypothesis argue that language anxiety is a stable personality trait triggered when a student has to perform oral and written tasks in the L2” (p. 728). However, they investigated anxiety in combination with several other cognitive and non-cognitive features. Their subsequent longitudinal analysis showed that students with high anxiety exhibited significantly slower progress in L2 development, compared to students with low anxiety. Anxiety was further investigated by Woodrow (2006) in a mixed-methods study that cataloged the major reported causes of second language anxiety, and analyzed its relationship with second language performance. Woodrow's participants were 275 students who were taking intensive EAP (English for Academic Purposes) courses immediately prior to entering Australian universities. Her results indicated that anxiety is the most crucial predictor of L2 oral proficiency development, and that it is most often brought on by interaction with native speakers of English. Students with high anxiety, especially those from Heritage Confucian Cultures, often experienced retrieval interference and skills deficits. The central role of corrective feedback in L2 pedagogy has made it an important variable in previous studies of individual differences, including differences in language anxiety. DeKeyser (1993) explored the effectiveness of CF on individual differences in language proficiency, grammatical sensitivity, extrinsic motivation, and anxiety. The participants were thirtyfive Dutch-speaking Belgian high school seniors, who were learning French as a second language. Focus on form was their primary instruction method, and all classroom communication was in French. The teachers used oral error correction for an entire academic year, and most of the error correction in communicative activities was related to morphosyntax. Even though the results showed that there were no significant interactions between individual factors and CF, students with high previous achievement, high language aptitude, and low anxiety profited most from their teachers' oral error correction on a written grammar test. Given that these students were learning French, a popular and official language in Belgium, it's possible that language anxiety functioned differently for them than it would for adult L2 learners in American ESL university classrooms. For instance, the Belgian students could expect a large community of other Dutch speakers, even in academic and professional contexts. The same cannot be said for most international graduate students in American universities, who are typically expected to study, communicate, and publish in English. Furthermore, cultural differences between the international students and their colleagues and professors may heighten the role of their affective variables in how they respond to CF. Sheen's (2007) study likewise focused on learner attitude, but in a context more relevant to the present study. Sheen (2007) showed how different types of CF influence the acquisition of English articles, as well as the extent to which individual differences mediate the effects of CF. Here, the participants were 80 intermediate-level ESL adult learners in the American Language Program (ALP) of an American community college. Three grammar assessments were administered: a pre-test, post-test (one day after), and a delayed post-test (12 days after), consisting of a timed dictation test, a writing test, and an error correction test, respectively. In addition, Sheen (2007) administered an aptitude test prior to the pre-test, a questionnaire to investigate the learners' attitudes toward error correction, and an exit questionnaire to explore the learners' awareness of the treatments' focus. The results demonstrated that when students had more positive attitudes toward CF, they improved their test performance. However, this relationship only held if the learners noticed what was being corrected during the initial CF interactions. Sheen's (2008) study investigated the interplay between language anxiety, recasts, and L2 learners' grammatical accuracy. More specifically, Sheen (2008) examined “learners' responses to recasts” in terms of their language anxiety (p. 846). She found that high anxiety prevented some learners from benefitting from the interactional feedback moves that

80

E.J.(E. Lee / System 56 (2016) 78e95

recasts provided them. As a result, Sheen (2008) concluded that language anxiety influences both the specific efficacy of individual feedback (e.g. whether a recast prompts the student to self-correct) and the broader efficacy of recasts in promoting student learning. Though Sheen's (2008) study does show a relationship between students' existing anxiety levels and their receptiveness to recasts, it assumes that those levels are fixed for each student. Accordingly, it does not consider an important corollary relationship: how CF itself potentially impacts L2 learners' anxiety levels. As ESL and EFL classrooms become more and more diverse, teachers must better understand the relationship between anxiety and second/foreign language learning in terms of language achievement, so that they can develop strategies and interventions to support students' self-confidence and minimize their language anxiety. In line with this perspective, the present study argues that CF should play a central role in these strategies, both in terms of delivering feedback and of giving students sufficient interaction opportunities to facilitate their oral development, while diminishing their timidity and shyness. 3. Method 3.1. Research questions 1. What patterns of corrective feedback and learner repair occur in advanced-level adult ESL classrooms? 2. How does oral CF specifically affect students' anxiety about speaking English?

3.2. Participants The targeted participants, sixty advanced-level ESL adult students, were all first or second year PhD students at a large Midwestern university, and were training to be teaching assistants (TAs). Specifically, they were all enrolled in a TA training course in the university's intensive oral English program from March through June 2012. They were from a variety of countries: 36 were Chinese, 10 Korean, 4 Indian, 2 Japanese, 2 Taiwanese, 2 Thai, 1 Mexican, 1 Iranian, 1 Venezuelan, and 1 was from Hong Kong. In addition to the student participants, four American ESL instructors participated in the study, though none were available for interviews. All sixty students planned to teach courses in their respective fields to undergraduate students at the university. Before the ESL Spoken English program begin, all the students took the SPEAK test, an institutional version of the TSE (Test of Spoken English) which determines the spoken English proficiency of people whose native language is not English. Students who receive a score of 230 or above on the SPEAK test are immediately certified to teach, while scores below 220 place students into one of three courses. Students who receive scores between 190 and 220 are placed in the Level 3 course, while scores below 190 earn placement into the Level 1 or Level 2 course. At the end of the Level 3 course, students take a mock teaching test (MTT): a 10 min lecture (in the student's academic discipline) to a real audience, to demonstrate proficient and effective oral English communication. Most students pass the MTT, and are thus certified to teach. The few students who fail take an additional course (Level 4), and retry the MTT at the end of the quarter. Since Level 3 students have already been exposed to significant amounts of oral CF, and since they are actively trying to improve their oral English to pass the MTT, Level 3 is the best place to gather sufficient relevant data for this study. All sixty students completed two surveys on the relationship between oral CF and affective variables, one early in the semester and another a month later (see Appendices A and B). After completing both surveys, 45 of these students showed enthusiastic interest in participating in follow-up interviews. Two of them were unable to complete the interviews because of schedule conflicts. To ass the suitability of interviewing the remaining 43 students, the researcher thoroughly examined their survey responses. First of all, in order to identify how the students' anxiety levels had changed, their relevant Part II and Part III responses were analyzed, and each student's anxiety data were classified into one of three categories: ‘no score change,’ ‘decreased score,’ or ‘increased score.’ In addition to this quantitative data, which helped identify preliminary trends relevant to the study, the students' qualitative responses in Part I helped identify potential interview candidates. Most of the students offered specific, detailed reasons for how CF influenced their oral English improvement and their progress in the class. This suggested that they had carefully considered how and why CF works, and that they would be able to apply those ideas to changes in their own affective variables, including anxiety. However, three students showed inconsistencies between their quantitative and qualitative survey responses, especially in terms of the effects of their teachers' oral CF. Specifically, their responses suggested that they would have difficulty contributing specific observations about CF and anxiety. Since the study's research questions require maximizing the richness of the qualitative interview data, these three students were not interviewed. Among the forty interviewed students, 22 were Chinese, 8 Korean, 3 Indian, 2 Japanese, 2 Taiwanese, 1 Thai, 1 Venezuelan, and 1 was from Hong Kong. The students' academic fields were somewhat diverse, though heavily weighted towards STEM (Science, Technology, Engineering, and Math) fields. They were from five different instructors’ classrooms: 11 from Instructor 1, 9 from Instructor 2, 5 from Instructor 3, 10 from Instructor 4, and 5 from Instructor 5. Though two classrooms were relatively underrepresented in this sample, it was not possible to enforce a strict balance since interview participation was voluntary.

E.J.(E. Lee / System 56 (2016) 78e95

81

3.3. Setting This study was conducted in an international teaching assistant (ITA) training program which focused on oral English, and which is part of a Midwestern university-affiliated ESL institute. Only ITAs whose native language is not English are eligible to participate. To help ensure that these students can teach effectively in the university, program instructors and staff members place incoming ITAs into one of four levels based on an institutional placement test, provide formal coursework as needed, and give teaching performance tests after each course to determine whether the students are ready to teach. As mentioned above, the program has four sequenced courses, referred to here as Level 1, Level 2, Level 3, and Level 4. This study collected data from students in the Level 3 course, both because Level 3 instructors are expected to use oral CF frequently in the classroom, and because students in this course have typically been exposed to a lot of CF by the time they start. The researcher's own classroom observations, both those for this study and those for several previous studies conducted in this program, confirm these trends. In fact, CF practices are a mandatory part of the curriculum: instructors are trained to stop and correct students whenever a problem in the students' communication occurs. Students are then asked to repeat their utterances, making the suggested corrections. The ultimate goal of the course is to improve students' natural and readily intelligible language, so they can teach university-level courses to a Midwestern American English-speaking population. Fortunately, since the researcher was allowed to observe nearly all the Level 3 classrooms for this study (one teacher declined to be observed), the researcher could measure and assess multiple CF strategies and techniques. 3.4. Instruments The data were collected by means of two questionnaire surveys (the pre- and post-surveys; see Appendices A and B) which were developed by the researcher, follow-up interviews, and classroom observations. With respect to classroom observations, the researcher observed four Level 3 classrooms over the course of four weeks, for a total of 24 h: 6 h (one week) in each classroom. The observed classes were audio-recorded and video-recorded, and then transcribed. As expected, all the observed classes included at least some corrective feedback. Each class had twelve students, and one and a half hours of instruction per period. One class was not observed at the instructor's request, but the students did complete the surveys. During the observations, the researcher did not directly participate in the class, but used observation field notes to record the class description, teachers' CF types, students' repair attempts, and students' verbal and nonverbal responses, as well as to track the teacher's and students' positions and movements in the classrooms. These notes also captured the students' physical responses to their teachers' feedback (including nonlinguistic signals) that the video recording might have missed. To bookend these observations, one survey was given at the end of the first week of classes, and another after an additional month of instruction. The pre-survey gathered data on the students' previous English learning experiences, to assist in interview planning and data analysis. Because this article draws on a larger project on affective variables in ESL study, both the pre- and post-surveys also included students' self-assessments of four affective variables: anxiety, attitude, motivation, and self-confidence. Levels of each variable were measured using a 10-point Likert-type scale ranging from “Low” to “High” (see Appendices A and B), and each affective variable was measured using three items. The post-survey included a total of thirtyeight declarative statements to gather students' views on anxiety (Items 1e10), attitude (Items 11e20), motivation (Items 21e28), and self-confidence (Items 29e38). Obviously, Items 1e10 are most relevant for the present study. A few statements (Items 6, 12, and 20) used to measure anxiety and attitude were excerpted from Sheen's (2011) experimental study on individual differences and CF, and for several other questions (Items 1, 4, 9, 11, 13, 15, 17, and 18) the researcher modified Sheen's phrasing to investigate how teachers' oral CF specifically influenced adult ESL students' anxiety and attitude. All the remaining items were composed by the researcher. To ensure the validity and reliability of the questionnaire, drafts of each survey were anonymously reviewed and revised by three professors in TESOL and SLA, and by one expert on quantitative research, evaluation, and measurement. Afterward, the revised questionnaires were tested by five ITAs who had taken the Level 3 course the previous quarter. This process allowed the researcher to allot enough time for the participants to complete the survey, and identified potential difficulties the participants might have with unclear wording or concepts, especially in the declarative statements. Overall, the goal of these tests was to make the questionnaires appropriate for the study's participants, and to assess how well they measured all four affective variables. Finally, follow-up interviews were conducted to gather qualitative data to buttress the quantitative data derived from the questionnaires. All interviews were conducted in English, in a quiet and empty classroom, and each interview was recorded and immediately transcribed. The interview questions drew on both the students' quantitative and qualitative survey responses. In particular, the qualitative interviews were designed to articulate the surveys’ quantitative results more precisely. For instance, if an interviewee reported changed anxiety scores on the post-survey, the interviews were designed to obtain more qualitative data about what the students meant by the scores, how specifically their anxiety increased or declined, and why those changes happened. The analysis sections quoting interview excerpts include the relevant interview questions. 3.5. Procedures At the end of the first week of classes, the researcher administered the pre-survey questionnaire survey to all sections at the end of class. From the second week through the fifth week, the researcher completed all the classroom observations. In

82

E.J.(E. Lee / System 56 (2016) 78e95

the sixth week, the researcher administered the post-survey questionnaire to each class, at the end of the class period. At the end of the post-survey questionnaire, each student was asked if he or she was willing to participate in an individual follow-up interview with the researcher. Interview participants were selected from the pool of interested students, using the methodology described in the “Participants” section above. The selected students were interviewed individually for 60e70 min each. The interviews probed, in greater detail, the students’ responses about their affective stances toward CF. Member checks, also known as informant feedback or respondent validation, were used in order to corroborate face validity. 3.6. Data analysis To scrutinize the frequency of corrective feedback types and each type's learner repair rates (Research Question 1), an external rater coded the audio-recorded and video-recorded classes. To classify CF types, the present study modified Lyster and Ranta's (1997) taxonomy of corrective feedback. CF was divided into six main feedback moves (recasts, explicit correction, elicitation, clarification, repetition of error, and meta-linguistic feedback), but two more types of CF were added: an ‘explicit correction with metalinguistic explanation’ item adapted from Sheen (2011), and a ‘nonlinguistic signals’ item, including the use of gestures and a plastic whistle. Definitions of the eight types of corrective feedback follow those in Ellis (2009), Lyster and Ranta (1997), Panova and Lyster (2002), and Sheen (2004, 2011); see Appendix C for definitions and examples. The error types were classified in terms of phonological, grammatical, and lexical features, and uptake types were classified into repair, needs-repair, or no uptake results. In order to establish rater reliability in measuring error correction classification, frequency, and repair, a second raterda native English speaker who has taught ESL courses in China and Koreadparticipated in the study. This allowed the interrater reliability correlation (see below) to be measured. For Research Question 2, descriptive statistics and nonparametric tests of medians were used to examine the changes in the students' anxiety levels as a result of teachers' oral CF. Based on the quantitative results of Research Question 2, the students' interview responses were coded to emphasize explanations of how and why CF increased or decreased the students' anxiety about speaking English. To ensure more accurate and fine-tuned qualitative data analysis, these codes were re-examined and revised three times by two qualitative data analysts with expertise in ESL education. In order to establish credibility, member checks were utilized by both analysts to corroborate face validity and to confirm the transcripts' accuracy. Table 4 in the qualitative result subsection traces the relationships between teachers' oral CF and students' anxiety. This table was created from two datasets: in-depth interviews with forty students, and those same students’ pre- and post-survey responses. 4. Results 4.1. Research question 1: CF patterns and learner repair in advanced-level adult ESL classrooms 4.1.1. Quantitative results Research Question 1 examines the types of oral corrective feedback present in the observed classrooms, and each type's respective learner repair rate. For this purpose, the researcher recorded the number of times the instructors used each type of corrective feedback, and the number of times the students repaired erroneous utterances in response to that feedback. CF frequency and repair rates were then compared across the four classes and the eight types of CF. Interrater reliability for the frequencies of CF types was assessed based on the observation data of two different ratersdthe researcher and the native-speaker rater mentioned earlierdusing the intra-class correlation coefficient (ICC). ICC ranges from 0.0 to 1.0, with 0.0 representing zero agreement and 1.0 representing complete agreement. In other words, the ICC will be high when all raters give the same or similar scores to each category or item. In this case, the ICC is high (0.9987), with a 95% confidence interval of 0.9974e0.9994. This suggests that despite slight differences in some of the scoring, the frequency ratings given by the two raters are highly consistent. Table 1 lists the frequency of CF offered and utterances repaired for each type of CF. The “offered” column tabulates the number of occurrences of each type of CF in the classroom, whereas the “percentage of total CF” column gives the proportion Table 1 Observed frequency and repair rates of different types of corrective feedback.

#1 Recast #2 Explicit correction w/metalinguistic explanation #3 Clarification request #4 Metalinguistic feedback #5 Explicit correction #6 Elicitation #7 Nonlinguistic signal #8 Repetition Sum

Offered

Percentage of total CF

Repairs

Repair rate

216 135 162 44 34 31 18 9 649

33.28% 20.80% 24.96% 6.78% 5.24% 4.78% 2.77% 1.39%

186 125 137 35 33 27 16 5 596

86.11% 92.59% 84.57% 79.55% 97.06% 87.10% 88.89% 55.56%

E.J.(E. Lee / System 56 (2016) 78e95

83

Table 2 Observed frequency and repair rates of corrective feedback, by instructor.

Instructor Instructor Instructor Instructor Sum

1 2 3 4

Offered

Percentage of total CF

Repairs

Repair rate

232 215 83 119 649

35.75% 33.13% 12.79% 18.34%

214 183 77 90 596

92.24% 85.12% 92.77% 75.63%

Table 3 Paired nonparametric test on the change of median anxiety levels: pretest vs. posttest.

Overall Class 1 Class 2 Class 3 Class 4 Class 5

Pre Post Pre Post Pre Post Pre Post Pre Post Pre Post

Median

Median change

Sign test statistics

Exact p-value

Wilcoxon T statistics

Asymp. p-value

95% Confidence interval of the median difference Lower

Upper

5.667 3.333 6.000 3.333 6.333 4.000 5.833 3.000 4.333 3.167 6.000 3.167

1.667

4.000

0.000

107.500

0.000

2.000

1.333

2.333

0.000

0.001

0.000

0.003

3.333

1.333

1.667

0.000

0.001

0.000

0.003

2.333

1.000

1.667

0.000

0.002

0.000

0.005

3.000

0.833

0.333

3.000

0.344

24.000

0.720

1.167

1.167

2.500

1.000

0.006

1.000

0.003

3.167

1.667

Table 4 Effects of teachers' oral CF on students’ anxiety about speaking English.

Instructor 1 (11 students)

Instructor 2 (9 students) Instructor 3 (5 students)

Instructor 4 (10 students)

Instructor 5 (5 students)

How teachers' oral CF lowered anxiety

Number of interview participants (40 total)

Recognizing errors and knowing how to fix them Having more chances to practice oral English in the classroom Improving oral English Getting positive or friendly feedback Recognizing errors and knowing how to fix them Getting encouragement to speak confidently Recognizing errors and knowing how to fix them Improving oral English Getting positive or friendly feedback Recognizing errors and knowing how to fix them Practicing short presentations for teaching Pre-existing low anxiety, unrelated to CF Wording of clarification requests (“what?”) Recognizing errors and knowing how to fix them Practicing in a relaxed classroom environment Pre-existing low anxiety, unrelated to CF Pre-existing high anxiety, unrelated to CF

5 4 1 1 8 1 3 1 1 3 1 1 (no score change) 5 (increased anxiety) 2 1 1 (no score change) 1 (no score change)

of each type of CF. The number of repairs elicited by each type of CF is recorded in the “repairs” column, and the ratio of repairs to total CF given is listed in the “repair rate” column. As shown in Table 2, 649 examples of CF were observed across the four classrooms. The most common types of CF were recasts (about a third of all the CF observed), explicit corrections with metalinguistic explanation (over 20%), and clarification requests (over 24%). These three types of CF all had relatively high repair rates (>80%), with explicit corrections with metalinguistic explanation garnering the highest cumulative repair rate (over 92%). Less commonly observed were metalinguistic feedback, explicit correction, elicitation, paralinguistic signals, and repetition. Among these five less-frequent types of CF, explicit correction elicited the highest repair rate (about 97%), whereas repetition had the lowest repair rate (over 55%). Of the 18 nonlinguistic signals observed, 16 were gestures used by Instructors 2, 3, and 4, and Instructor 4 also used a whistle twice to correct students' word stresses (see Appendix C). The whistle led to successful repair both times, as did 14 of the 16 gestures. Although using a whistle is an unusual approach to CF, none of Instructor 4's students specifically attributed changes in their anxiety levels to the whistle.

84

E.J.(E. Lee / System 56 (2016) 78e95

Table 2 provides the frequency of CF offered (regardless of type) and the number of times the students repaired their utterances, for each of the four instructors. As shown in Table 2, the frequencies and repair rates vary substantially across different classrooms. In terms of frequency, 69% of all the CF observed occurred in Instructor 1's and Instructor 2's classrooms. Students showed very high repair rates in classes taught by Instructors 1 (repair rate ¼ 92%), 2 (repair rate ¼ 85%), and 3 (repair rate ¼ 92%), whereas Instructor 4's CF was modestly less effective (repair rate ¼ 75%). Fig. 1 illustrates CF frequency, by type, in each of the four classes. Each bar in the graph represents the CF observed in one class. The upper half of the figure shows the number of times the CF type was used by each instructor, whereas the lower half shows the number of times the students repaired their utterances after the CF. Different types of CF are shown in different colors, and the length of the bars represents the amount of CF/repairs observed. As shown in Fig. 1, instructors seem to have different preferences for specific types of CF. For example, recasts were the most frequently observed CF in Instructor 1's class, Instructor 2 preferred explicit correction with metalinguistic explanation, and Instructors 3 and 4 preferred to use clarification requests. Unlike the other instructors, Instructor 3 never used explicit corrections in class. 4.2. Research question 2: students’ anxiety about speaking English 4.2.1. Quantitative results Research Question 2 investigates whether advanced-level adult students' anxiety about speaking changed after they received their teachers’ oral CF. For this purpose, the medians of the self-reported anxiety levels on the pretest and on the posttest were compared, and the ratings on the perceived relation between corrective feedback and anxiety were examined. Fig. 2 compares the median levels of anxiety reported at the start and end of the quarter. Each bar represents the median rating for one item at a certain point of time (i.e., pretest or posttest). The items are rated on a 10-point scale, where 1 ¼ low and 10 ¼ high. The error bars represent the 95% nonparametric confidence interval. This indicates that we are 95% confident that the true median is somewhere between the two whiskers. As shown in Fig. 2, at the end of the quarter, students reported

Fig. 1. Types and frequencies of corrective feedback offered/responded in each class.

E.J.(E. Lee / System 56 (2016) 78e95 10 9 8 7 6 5 4 3 2 1 0

85

6

5.5

Pre

5

Post 3

3

3

Anxiety about speaking Fear to speak in front of Being embarrassed to English classmates communicate in English Fig. 2. Median anxiety levels with 95% nonparametric confidence interval: pretest and posttest.

lower levels of anxiety compared to those they reported at the beginning of the quarter. In general, anxiety about speaking English, fear of speaking in front of classmates, and embarrassment about communicating in English all decreased substantially from pretest to posttest. Table 3 summarizes the change in the median rating of average anxiety levels, by class. A nonparametric test, Wilcoxon's T test, was conducted on the pretest ratings and posttest ratings for each class, and on the overall sample. As shown in Table 3, averaged across the three anxiety measures (anxiety about speaking English, fear of speaking in front of classmates, and embarrassment about communicating in English), students' self-rated levels of anxiety decreased significantly from pretest to posttest (p < 0.001). Whereas the median self-rating of anxiety was 5.667 (on a scale of 1e10) when the class started, the median rating decreased to 3.333 by Week 6. Specifically, in all but Instructor 4's classes, the students' average anxiety level decreased significantly (p-values < 0.01). The decrease in anxiety level was most drastic in Instructor 5's class (2.5). In sum, the self-reported levels of anxiety (anxiety about speaking English, fear to speak in front of classmates, and embarrassment about communicating in English) decreased substantially from pretest to posttest. On a 10-point scale, the average decrease is expected to be somewhere between 1.33 and 2 with 95% confidence. Specifically, the distribution of the ratings obtained from the ten items measuring the relationship between CF and anxiety (Cronbach's Alpha ¼ 0.731; see Appendix B) demonstrated that most of the respondents did not relate teachers' CF to strong negative emotions such as fear, embarrassment, or avoidance. In fact, they demonstrated a strong willingness to get CF from the teachers, and related getting teachers' CF to their decreased anxiety levels. 4.2.2. Qualitative results While the quantitative results show that most of the students had a strong willingness to get their teachers' oral CF, and that getting it lowered their anxiety about speaking English, Instructor 4's students reported on the posttest that their anxiety levels increased significantly during the course. For this reason, the qualitative analysis examines both how CF can lower students' anxiety about speaking English, and also why the process did not work in Instructor 4's class. Table 4 lists the coding categories relevant to student anxiety, and summarizes the reasons participants gave for their changes in anxiety level. Among the forty interview participants, there were five cases of increased anxiety due to the instructor's wording of clarification requests (“what?”) and three cases of pre-existing low or high anxiety which was unaffected by CF. The remaining thirty-two students reported that their teachers' oral CF helped them lower their anxiety about speaking English. More specifically, twenty-two students argued that recognizing their errors and knowing how to fix them lowered their anxiety levels. Four students stated that having more chances to practice their oral English in the classroom had similar effects. The thirty-two students who reported lower anxiety were given the same set of interview questions. While several of these questions naturally focused on the effects of CF, students were also given ample opportunity to offer other explanations for their lower anxiety levels. To allow for alternate explanations, participants were first asked to “explain more about how and in what ways [their] anxiety level was lowered,” before CF explicitly entered the conversation. At that point, participants were asked, “How have you noticed a direct relationship between your teacher's oral CF and your decreased anxiety level? Could you tell me more of your story about your lowered anxiety?” These questions were designed to draw out the participants' own accounts of the potential CF-anxiety relationship, while two follow-up questions encouraged richer and more complex accounts of that relationship: “Do you think teachers' oral corrective feedback is really helpful to lower ESL learners' anxiety about speaking English? If so, why do you think so?” As Table 4 demonstrates, the majority of participants did connect CF to changes in their anxiety levels. For instance, Participant 18 had this to say: After getting my teacher's oral corrective feedback, I correct my errors. Then, I kind of feel like, “Yeah, this is what the problem with my English is.” So I will no longer have these problems; I feel confident. Because yeah, I can speak much better English, so I don't have to worry about other people judging me or whatever. My anxiety about speaking English is decreased. I know it's impossible that I do 100% correct, but at least I know this problem, I'm aware of this problem, so I kind of pay more attention when I say this word. So that's why I said my anxiety actually lowered (Participant 18's personal interview, May 30, 2012).

86

E.J.(E. Lee / System 56 (2016) 78e95

Conversely, five students in Instructor 4's class reported increased anxiety levels as a result of his oral CF. The students claimed that the instructor's use of clarification requests such as “What?” and “Sorry?” made them feel more anxious, since they did not know what their errors were. Their interview responses suggest that Instructor 4's CF made them feel intimidated and frustrated about speaking English. The following representative interview excerpts are from two of Instructor 4's students: Participant 28 is from India, and Participant 32 is from Korea. Both excerpts demonstrate why the students felt embarrassed and more anxious after these clarification requests. Both students were asked similar questions to those listed above, with the obvious change of “lowered” to “increased.” Additionally, they were asked several questions specifically about clarification requests: “What do you think of using clarification requests like ‘What?’ ‘Sorry?’ or ‘Pardon me?’ in the classroom? How do you feel when you get clarification requests from your teacher? When you received clarification requests, did you notice what kinds of errors and mistakes you made? After getting your teacher's clarification requests, did you know how to correct these mistakes?” When he asked me, “What? Sorry? Pardon?” I think in most of the cases I felt annoyed. Because I think, according to me, non-native speakers, when they speak English they have the pattern in their minds, so they always translate or they always create a sentence and then they speak it. So in that case, if someone breaks your link, then after returning back from explaining that stuff, you have to go back again to what you were saying, and then there is always the chance of grammatical errors or pronunciation errors, just after coming from that or that break (Participant 28's personal interview, May 29, 2012). “What?” It's really problematic. We are non-native. We are very nervous. So, when I listen to “What?” I just didn't like it. I just want to finish that as soon as possible, regardless of conversation presentation, and whatever. His oral correction was intimidating and threatening a lot, and he had a judgmental attitude. I can give a specific situation, as an example. One of our classmates was trying to say “primitive,” and he was Indian, so his pronunciation was like, ‘padimitive.’ And [Instructor 4] could have done it like this, “Oh, it's not ‘padimitive,’ it's ‘primitive'.“ But he was like, “IT'S NOT 'PADIMITIVE'. WHAT IS 'PADIMITIVE'?' Oh my God. He never did that again. He should avoid that attitude. He was like, “IT'S NOT 'PADIMITIVE'. WHAT IS 'PADIMITIVE'?' Oh my God. What kind of teacher is like that? That is intimidating enough! (Participant 32's personal interview, June 11, 2012). These students' frustrations with their instructor's clarification requests correspond to the researcher's classroom observations of Instructor 4's methods. Specifically, during class on April 20, 2012 the researcher observed that the six presenters felt uncomfortable, not because they were too nervous to present, but because Instructor 4 kept using “What?” during clarification requests. Based on the classroom observation transcript, Instructor 4 used clarification questions 25 times out of 41 error corrections, making clarification requests the most frequently used type of corrective feedback on that day. However, these clarification requests did not change any of the students' utterance errors, but rather made the students repeat what they said, which suggested they thought their mispronunciations prevented their teacher from understanding what they were talking about. This frustrated and embarrassed the students, and even though Instructor 4 subsequently used recasts to correct their pronunciations, those recasts generated very low repair rates. By contrast, in other classrooms the students immediately repaired their mistakes and errors after recasts. Overall, the striking difference between Instructor 4 and his colleagues emphasizes how anxiety-inducing CF can actually hinder students from speaking English freely and comfortably, and thus negatively impact their learning of oral English. 5. Discussion 5.1. Affective risks of clarification requests Any discussion of CF techniques should take into account the potential negative aspects of using clarification requests as output-prompting feedback. While many studies of CF have considered which types of CF are effective and successful in eliminating students' erroneous utterances, there is little research on which types might have a negative emotional impact on ESL students' L2 learning and development. The present study's findings support Sheen's (2011) statements that anxiety might influence L2 learners' ability to pay attention to and process teachers' oral CF, and that the mode in which CF is provided affects student anxiety. Sparks et al. (2009) insist more urgently that “anxiety has a debilitating effect on L2 performance,” such that “as anxiety increases, language skills decrease” (p. 745). Finally, Young (1994) argues that teachers' error correcting methods and the ways teachers interact with their students might provoke students' anxiety. Accordingly, effective ESL teachers should develop strategies for instructional interventions, including corrective feedback, which diminish L2 learners’ speaking anxiety. Specifically, ESL teachers should be attuned to ways in which their oral CF may inadvertently make their students feel anxious or nervous, and be prepared to handle anxiety-provoking situations. Addressing language anxiety effectively and professionally will not only give students more useful input on their speaking errors and mistakes, but also encourage them to reduce their anxiety about speaking English (Chan & Wu, 2004).

E.J.(E. Lee / System 56 (2016) 78e95

87

In Author's (2013) study, advanced-level adult ESL students most preferred to get explicit correction, whereas clarification requests were their least preferred type of CFdeven though clarification requests generated over 90% learner repair. Respondents identified several reasons for this preference. First, clarification requests seemed vague and unclear, since the students could not easily recognize their errors. Similarly, they had trouble understanding their teachers' intentions, which further embarrassed and frustrated them. Third, some students thought their teachers had not listened carefully to what they were saying. Fourth, students felt uncomfortable when clarification requests exposed their lack of oral proficiency in front of their classmates. Finally, clarification requests tended to shut down further conversation or presentations. To correct these negative effects, the students suggested that their teachers provide more specific follow-up questions, to indicate more clearly the targeted errors and the goals of the feedback. The students' proposal could lead to more successful teacher instruction in correcting students' errors and mistakes. The present study's participants raised similar objections to Instructor 4's clarification requests, particularly in terms of embarrassment and frustration. The resulting classroom dynamic specifically damaged the students' anxiety and selfconfidence, and it kept them from improving their oral English proficiency. As noted earlier, five of Instructor 4's students increased their anxiety levels as a result of his oral CF because of his use of “What?” and “Sorry?” in clarification requests. They reported feeling anxious, intimidated, and frustrated. As a result, two of these five students, along with two of their classmates, likewise lost self-confidence during the course, citing similar reasons. In addition, Instructor 4's students did exhibit significantly lower repair rates after his oral CF, which may lead to more errors and slower improvement in oral English proficiency. According to Lyster and Ranta's (1997) definition of CF types, clarification requests indicate that the student's utterance was not understood and ask that the student reformulate it; however, Instructor 4's approach to clarification requests discouraged the students to speak in English and offended them. This regrettable result confirms Horwitz, Horwitz, and Cope's (1986) recommendation that “as students appear to be acutely sensitive to target language corrections, the selection of error correction techniques should be based on instructional philosophy and on reducing defensive reactions in students” (p. 131). Instructor 3's classroom offers an instructive example of more emotionally effective clarification requests. There, clarification requests were used often to indicate the students' oral English errors, and led to the highest learner repair rate (96%) among clarification requests across all four classrooms. They also caused no observable negative affective results. This could be explained by Instructor 3's choices in wording and attitude: unlike his colleague, his clarification requests used the phrases “repeat again?” and “say again?”. This clarity removed the students' confusion about how to respond, and they just repeated what they said to their teacher. When asked about this wording, Instructor 3 said that using clarification requests such as “what?” or “sorry?” seemed rude, so instead he asked his students to repeat what they'd said. He attributed these comprehension problems to the students' relative lack of confidence about speaking English. Because these rephrased clarification requests included a specific request, Instructor 3's students did not exhibit any frustration, embarrassment, or discouragement. Instead, as soon as they got Instructor 3's clarification requests, they repeated what they said to him or to their classmates (April 16, 2012, Instructor 3's classroom observation notes). Though clarification requests, by definition, do not offer direct corrections of students' errors (see Appendix C), this repetition is intended to prompt self-repair and improvements in vocal clarity. Instructor 3 reinforced that he was listening carefully to his students through his physical gestures and movements. For example, when he did not understand what the students said to him or how they responded to his questions, he always moved closer to his students so he could hear what they were saying. Also, when appropriate he added gestures to show he did not understand what his student was saying, and thus signaled them to speak again loudly. After seeing these gestures, his students repeated what they said to him more loudly and more clearly. 5.2. The affective filter in the ESL classroom According to the affective filter hypothesis (Krashen, 1982), affective variables such as fear, nervousness, boredom, and resistance to change can have a negative effect on the acquisition of a second language. When the affective filter, an impediment to acquisition caused by negative emotional responses to one's environment, blocks comprehensible input, acquisition fails or occurs to a lesser extent than normal. Although subsequent scholars have critiqued Krashen's (1982) findings (Cook, 1993; McLaughlin, 1987), on the grounds that there is little empirical support for them, I find Krashen's framework useful in assessing why CF sometimes has negative affective results. By using friendly and sensitive clarification requests and gestures, Instructor 3 lowered his students' affective filters, whereas Instructor 4's abrasive and potentially offensive clarification requests arguably increased his students' affective filters. This contrast is likely responsible for the students' different anxiety levels, and suggests that effectively targeted CF can help reduce students' affective filters to create a relaxed and comfortable classroom atmosphere. Simply put, if an ESL teacher's practices or attitudes increase his or her students' affective filters, especially in an environment where anxiety is common, the students may well be less receptive to learning English from that teacher. It is likely, of course, that cultural differences interact with individual sets of affective filters to determine the efficacy of particular pedagogical approaches. As noted earlier, for instance, in Woodrow's (2006) study, Asian students from

88

E.J.(E. Lee / System 56 (2016) 78e95

China, Korea, and Japan with Confucian Heritage Cultures showed more anxiety about learning English than did students of other nationalities. A similar dynamic might affect the present study's results, since most of the participants were from Confucian Heritage Cultures. However, regardless of their students' home countries or cultures, ESL teachers should facilitate students' oral English proficiency while diminishing their timidity and nervousness, in order to produce improved output. Tsiplakides and Keramida (2009) indicate that it is necessary to create friendly and supportive classroom atmospheres, so that students can regard their errors as a natural learning process in mastering a language. The authors significantly remind teachers that there is a need to know who their students are, what their attitudes toward oral production are, and why they exhibit low participation and unwillingness to engage in speaking activities. Such attitudes among teachers help their students to overcome foreign language speaking anxiety without tension and discomfort. 5.3. Study limitations The present research has some limitations. First, as mentioned in the instruments section, this study presents partial data from a larger study on the relationship between teachers' oral CF and a range of affective variables: anxiety, attitude, motivation, and self-confidence. Space limits precluded including the data on all four variables in a single article. Accordingly, future work should consider how teachers' oral CF interacts with adult ESL students’ attitude, motivation, and self-confidence to improve their existing oral English proficiency. Along similar lines, this study focused only on adult ESL students' orientations to their teachers' oral CF, as it occurred in selected advanced-level classrooms, due to limited opportunities to collect comparable data from the participants' teachers. The teachers' tight schedules and the program's strict policies on data collection made it impractical to perform in-depth follow-up interviews with the teachers, which could potentially have revealed more meticulous explanations for their CF choices as well as their broader perspectives on CF and error correction. Therefore, future work should include balanced data from both sides, including more detailed interviews, in order to analyze potential discrepancies between students' and teachers' orientations toward oral CF and error correction. 5.4. Conclusions and pedagogical implications The present study indicates that effective L2 pedagogy must use CF to encourage or even create positive emotional states in ESL students. In the present study, most of the participants were from Asian countries where oral English was deemphasized in their EFL classes. Therefore, they were eager to improve their oral English and to communicate with Americans in English. By providing them with opportunities to receive oral CF, the Level 3 course helped them meet these goals. More importantly, teachers' oral CF had positive impacts on the students’ affective variables, specifically lowering their anxiety about speaking English. As dedicated and enthusiastic L2 learners, these adult ESL students wanted to know what their errors were, and how they should correct them to improve their oral English proficiency. However, this study also showed that the way in which corrective feedback is conveyed can have negative pedagogical consequences. In this case, clarification requests such as “What?”, “Sorry?”, or “Pardon me?” increased the students’ anxiety and discouraged them to speak English, so that their self-confidence was decreased. Using alternate wording, or adding gestures to clarification requests, can alleviate some of these problems. For instance, teachers in speaking classes might use other phrases such as “Could you repeat that?”, Could you say again?”, “Do you mean … ?”, or “Are you saying ….?” In this sense, the teachers might help their students feel comfortable and empowered when they receive clarification requests. Above all, though, ESL teachers should explain how corrective feedback works, ideally at the very beginning of the class. For example, clarification requests are not simply meant to point out bad English or pronunciation, but to ask the student to repeat what he or she said because the teacher could not understand it clearly. In such classroom environments, the students have ample opportunities to practice their English, negotiations for meaning become available, their self-confidence is boosted, and their levels of anxiety are lowered. When teachers deliver oral CF in an emotionally effective way, they encourage their students to speak confidently and freely in the ESL classroomdand beyond.

Acknowledgments I would like to acknowledge and thank to Dr. Alan R. Hirvela and Dr. Leslie C. Moore, my co-advisors. Their continued comments and feedback on my corrective feedback research allowed me to present and publish my work at important conferences and in important journals. I would like to express my deep thanks to them as my advisors, mentors, and respected scholarly models. Their words of advice and support allowed me to conduct this wonderful dissertation study, and their comments and feedback clarified the study's goals and contributions.

E.J.(E. Lee / System 56 (2016) 78e95

Appendix A

89

90

E.J.(E. Lee / System 56 (2016) 78e95

E.J.(E. Lee / System 56 (2016) 78e95

Appendix B

91

92

E.J.(E. Lee / System 56 (2016) 78e95

E.J.(E. Lee / System 56 (2016) 78e95

93

94

Appendix C

E.J.(E. Lee / System 56 (2016) 78e95

E.J.(E. Lee / System 56 (2016) 78e95

95

References Author (2013). Chan, D. Y., & Wu, G. (2004). A study of foreign language anxiety of EFL elementary school students in Taipei County. Journal of National Taipei Teachers College, 17, 287e320. Chu, R. (2011). Effects of teacher's corrective feedback on accuracy in the oral English of English-Majors college students. Theory and Practice in Language Studies, 1(5), 454e459. Cook, V. (1993). Linguistics and second language acquisition. London: Macmillan Press. DeKeyser, R. M. (1993). The effect of error correction on L2 grammar knowledge and oral proficiency. The Modern Language Journal, 77, 501e514. Ellis, R. (2007). The differential effects of corrective feedback on two grammatical structures. In A. Mackay (Ed.), Conversational interaction in second language acquisition: A collection of empirical studies (pp. 407e452). Oxford: Oxford University Press. Ellis, R. (2009). Corrective feedback and teacher development. L2 Journal, 1, 3e18. Ellis, R., Loewen, S., & Erlam, R. (2006). Implicit and explicit corrective feedback and the acquisition of L2 grammar. Studies in Second Language Acquisition, 28, 339e368. Han, J., & Jung, J. (2007). Patterns and preferences of corrective feedback and learner repair. Korean Journal of Applied Linguistics, 23, 243e260. Horwitz, E. K. (2001). Language anxiety and achievement. Annual Review of Applied Linguistics, 21, 112e126. Horwitz, E. K., Horwitz, M. B., & Cope, J. (1986). Foreign language classroom anxiety. The Modern Language Journal, 70, 125e132. Kim, J. H. (2004). Issues of corrective feedback in second language acquisition. Teachers College, Columbia University Working Papers in TESOL & Applied Linguistics, 4(2), 1e24. Krashen, S. (1982). Principles and practice in second language acquisition. Oxford: Pergamon. Lyster, R., & Ranta, L. (1997). Corrective feedback and learner uptake: negotiation of form in communicative classrooms. Studies in Second Language Acquisition, 19, 37e66. MacIntyre, P. D. (1999). Language anxiety: a review of literature for language teachers. In D. J. Young (Ed.), Affect in foreign language and second language learning (pp. 24e43). New York: McGraw Hill Companies. Mackey, A. (2006). Feedback, noticing, and second language development: an empirical study of L2 classroom interaction. Applied Linguistics, 27, 405e430. McLaughlin, B. (1987). Theories of second-language learning. London: Edward Arnold. Panova, I., & Lyster, R. (2002). Patterns of corrective feedback and uptake in an adult ESL classroom. TESOL Quarterly, 36, 573e593. Rezaei, S., & Mozaffari, F. (2011). Corrective feedback in SLA: classroom practice and future directions. International Journal of English Linguistics, 1(1), 21e29. Russell, V. (2009). Corrective feedback, over a decade of research since Lyster and Ranta (1997): where do we stand today? Electronic Journal of Foreign Language Teaching, 6(1), 21e31. Sheen, Y. (2004). Corrective feedback and learner uptake in communicative classrooms across instructional settings. Language Teaching Research, 8, 263e300. Sheen, Y. (2007). The effects of corrective feedback, language aptitude and learner attitudes on the acquisition of English articles. In A. Mackey (Ed.), Conversational interaction in second language acquisition: A collection of empirical studies (pp. 301e322). Oxford: Oxford University Press. Sheen, Y. (2008). Recasts, language anxiety, modified output and L2 learning. Language Learning, 58, 835e874. Sheen, Y. (2011). Corrective feedback, individual differences and second language learning. New York: Springer. Sparks, R., Patton, J., Ganschow, L., & Humbach, N. (2009). Long-term relationships among early first language skills, second language aptitude, second language affect, and later second language proficiency. Applied Psycholinguistics, 30, 725e755. Suzuki, M. (2004). Corrective feedback and learner uptake in adult ESL classrooms. Columbia University Working papers in TESOL & Applied Linguistics, 4, 1e21. Tsiplakides, I., & Keramida, A. (2009). Helping students overcome foreign language speaking anxiety in the English classroom: theoretical issues and practical recommendations. International Education Studies, 2, 39e44. Woodrow, L. (2006). Anxiety and speaking English as a second language. RELC Journal, 37(3), 308e328. Yang, Y., & Lyster, R. (2010). Effects of form-focused practice and feedback on Chinese EFL learners' acquisition of regular and irregular past tense forms. Studies in Second Language Acquisition, 32, 235e263. Young, D. J. (1994). New directions in language anxiety research. In C. A. Klee (Ed.), Faces in a crowd: The individual learner in multi-section courses (pp. 3e45). Boston: Heinle & Heinle Publishers. Zheng, Y. (2008). Anxiety and second/foreign language learning revisited. Canadian Journal for New Scholars in Education, 1, 1e12.