Effects of the types of error, proficiency level of the learners and nature of the recasts on the uptake of learners

This current study examined the effects of the type of errors learners make (that is, phonological, lexical and grammar), proficiency level of the learners (that is, A1, A2, B1, B2 and C1) and nature of the recasts (that is, long and short) on the uptake of the learners. The data of this study came from the video-recordings of A1, A2, B1, B2 and C1-level of Turkish as a Second Language (TSL) classes. 60-h data have been transcribed and analyzed by the researchers. Here, qualitative methods of data collection was employed in this multi-case research along with quantitative methods of data analysis, when necessary, to examine the relations between and among the constructs. Analysis of the study revealed that the learners had a higher percentage of uptake rate against phonological errors (with significant difference); C1 level learners had the highest percentage of the uptake rate (with significant difference) and long recasts yielded a higher percentage of the uptake (but with no significant difference between recast types). Thereafter, findings of this research are compared with those of other prominent studies.


INTRODUCTION
The effectiveness of corrective feedback on learners" interlanguage development has been a major issue of investigation in the recent second language acquisition research.Even though many researchers acknowledge the necessity of positive evidence for language acquisition (Krashen, 1981;Schwartz, 1993;Truscott, 2007), it is believed that there is also a need for negative evidence, in the form of corrective feedback, to make learners notice the problematic aspects of their utterances (Long, 1996;White, 1991).When such focuson-form takes place during interaction, learners not only pay attention to linguistic form, but form-meaning connections will also take place (Long, 1991).
Corrective feedback (CF) is defined as "responses to learner utterances containing an error" (Ellis, 2006: 28) or as "responses to a learner"s non-target-like L2 production" (Li, 2010: 309).Schmidt (1990Schmidt ( , 2001) ) opines hypothesis is the theory on which the role of corrective feedback is grounded.He suggests that the acquisition of a second language (L2) is a conscious process and supplying corrective feedback is one way to trigger the notice of language forms.With that theoretical background provided, many different descriptive (Doughty, 1994;Lyster and Ranta, 1997;Sheen, 2004) and experimental studies (Doughty and Varela, 1998;Lyster, 2004) were carried out both in classroom and laboratory settings, and more recent studies also focused on CF as a part of Computer Mediated Communication (CMC) research (Iwasaki and Oliver, 2003;Loewen and Erlam, 2006;Yilmaz and Yuksel, 2011).One of the most commonly studied aspects of CF-related research is on the effectiveness of recasts as a form of oral corrective feedback.A "recast" refers to "a reformulation of the learner"s erroneous utterance that corrects all or part of the learner"s utterance and is embedded in the continuing discourse" (Sheen, 2011: 2).Recasts have been studied extensively (Doughty and Varela, 1998;Ellis et al., 2006;Lyster, 2004;Nabei and Swain, 2002) because: a) they are very frequent in classrooms; b) they are considered to be implicit, and thus may not always successfully induce learner notice; and c) they provide both positive (that is, input) and potentially (if noticed) negative feedback (Sheen, 2011: 57).In the related literature, recasts were mainly compared and contrasted with other forms of corrective feedback such as prompts (Ammar and Spada, 2006;Yang and Lyster, 2010) and metalinguistic feedback (Carroll and Swain, 1993;Ellis et al., 2006;Lyster, 2004;Sheen, 2007) and the findings of the studies suggested that prompts are more effective than recasts and explicit CF is more effective than implicit CF (Ellis et al., 2006).Another aspect of recasts under investigation is whether it leads to learner uptake or not, but the studies led to conflicting results.In a seminal study, Lyster and Ranta (1997) found out that recasts were the most common type of CF in French immersion lessons; however, they produced the least amount of uptake.Panova and Lyster (2002), in their study carried out in an adult classroom in Canada, produced similar results to those of Lyster and Ranta (1997).On the other hand, Ellis et al. (2001), in their study in intensive adult ESL classrooms in New Zealand, found out that recasts were the most common type of corrective feedback and they led to a high level of uptake.

Turkish as a second language (TSL) context
In addition to its strategic geographic location as a bridge between two continents; namely, Asia and Europe, Turkey"s status as a candidate country for the European Yuksel et al. 1009 Union makes this country an attraction for students taking part in different programs such as Erasmus, Socrates, Leonardo, etc. Thanks to this new status, a need to teach Turkish language to foreigners has appeared in recent years.Taking this need into account, many universities and private institutions are working collaboratively to develop programs teaching Turkish to foreigners.Since the number of foreign students attending Turkish universities is on the rise, language teaching practitioners and materials developers have taken part in many new projects, and some important universities in Turkey have started M.A and PhD programs on this recently popular field.
Even though TSL, as a field of study, has gained popularity in recent years, a lot of studies carried out are related to the use of different methods to teach the grammar and vocabulary of Turkish (Apaydin, 2007;Barin, 1992;Yayli, 2004), materials and activities for Turkish in TSL classes (Dilek, 1995;Diliduzgun, 1995;Kucukler, 2010;Ozdemir, 2010;Top et al., 2015;Zengin, 1995) and individual differences (Eker, 2010).However, the number of studies dealing with the classroom discourse-based studies related to TSL classes is very limited (Fidan, 2015;Inan and Fidan, 2012;Yilmaz and Yuksel, 2011).It is believed that this study is going to contribute to the related literature with its large database and focus on different aspects of the corrective feedback patterns taking place in five different levels.

Research questions
This study aimed to answer the following research questions: (1) What type of errors do the A1, A2, B1, B2 and C1 level TSL learners make and what are the effects of the error type on the uptake of the learners in TSL classrooms?(2) Is there a relationship between the proficiency level and uptake of the learners?(3) Is there a relationship between the nature of the recasts (that is, long and short) and uptake of the learners?

The setting and participants
The current study was carried out at two foreign language centers of two state universities in the Northwestern and Western part of Turkey.The participants were A1, A2, B1, B2 and C1 level of learners (aged from 18 to 27 years) who were learning Turkish as a second language (TSL).Native languages and the ethnic backgrounds of the participants varied (such as French, Arabic, Uighur, Persian, Mongolian, Somalian, Egyptian, Indonesian).The total number of learners who signed the consent forms were 94 in five different classes; however participants of the classes changed from week to week in the recordings.The instructors of the classes (aged from 23 to 34 years) were all native speakers of Turkish who They all had the experience in teaching Turkish for at least five years at the time of the study (ranging between six years to 12 years).

Procedure
We used qualitative methods of data collection in this multi-case study.The qualitative nature of the study was very important because the aim of this research was to document, examine, and analyse naturally occurring data at two foreign language centers of two state universities in a TSL setting.We also used quantitative methods of data analysis, when necessary, to examine the relations between and among the constructs.Before the data collection process, the researcher obtained the Ethical Committee approval from Kocaeli University, and permission to access the two language centers from the university administrations.One week before the data collection, the participants taking part in the study signed consent forms and agreed to take part in the study.During the data collection, camcorders (Sony HDR-PJ260) were placed with caution in order to be able to catch the voices of the teachers and students well.A total of 60 h of video recordings from all levels (12 h from each level) of TSL classes was obtained.The researchers played an active role during the data collection process but did not participate in the development of lessons while making the recordings.The first two hours of the recordings have been excluded from the database because of some possible feelings of discomfort of students and teachers as they were not used to being recorded during their lessons.Table 1 indicates the number of recorded lessons from each level.After collecting the data, the following step has to do with transcription of the recordings.The researchers transcribed all videos verbatim based on the coding scheme prepared before the recordings.The personal information of the participants was removed and the researchers renamed the participants.Afterwards, the transcriptions were controlled and coded on the basis of Yuksel (2007).Thanks to this coding process, it was possible to follow turntakings, feedbacks, and uptakes of the learners.

Interrater reliability
In order to ensure the reliability of the findings, a randomly-selected part of the data collected was coded by 4 different raters (including the researchers carrying out this study).There was an external rater who helped the researchers in the coding process.He was an Associate Professor of Turkish Education who had been teaching Turkish as a second language for more than 10 years at the time of the codings.Ten percent of the data was selected randomly for the external rater.Before the process of reliability measurement started, a guide explaining the classifications and analyses of the data was prepared.The external rater was asked to code the selected data according to the guideline provided.The results of interrater reliability revealed that there was 94% consistency in the coding of the types of errors, 96% consistency in the types of the recasts and 89% consistency in the coding of the uptakes.
Only spoken language is considered throughout this study.Written corrective feedback types and uptakes are beyond the scope of this study.On the other hand, the analysis was limited with the recasts of teachers.Other types of corrective feedbacks were not analyzed due to the scope of the present study.

RESULTS
The current study examined the different elements of classroom talk (for example, types of the errors made, nature of recasts and uptake of learners) in five TSL classes with different proficiency levels, to explore the dynamics of discourse and shed light on what is going on in TSL classes.

The types of learner errors according to different proficiency levels
To find out the relationship between the types of errors made by the learners in different proficiency levels and uptake, the specific types of learner errors were tallied as presented in Table 2.
As illustrated in Table 2, the learners in five classes made a total of 1831 errors during the recordings.Out of these 1831 errors, 883 of them were phonological errors (48%), 513 were grammatical errors (28%) and 362 were lexical errors (20%).Table 2 also presents the errors made by learners in different proficiency levels.According to the findings of this research, A1 level learners made the highest number of errors (n=549), followed by A2 (n=421) and B1 (n=359).As the proficiency level of the learners increased, the number of the errors decreased.

The effects of the types of error learners made and proficiency level on the uptake of learners
As a second issue, we focused on the effects of the types of learner errors and proficiency level of the learner on the uptake.The numbers and percentages are provided in Table 3.
As portrayed in Table 3, the phonological errors yielded  the highest percentage (60%) of the uptake rate.It has also been observed that 53% of the grammatical errors and 49% of the lexical errors yielded uptake sequences.
To observe the effects of the type of learners" error and uptake, Pearson"s chi square test was conducted (Table 4).The analyis revealed that there was a significant relationship when all proficiency levels were examined together, according to chi-squared results (Asymp.Sig. (2-sided)=.000).
When the relationship between the type of learner errors and proficiency level of the learners was specifically examined, it was observed that there was a statistically significant relationship among some proficiency groups.Table 5 illustrates the results of the Pearson"s chi squared tests.
This study was specifically interested in the relationship between proficiency level of our learners and uptake rate of the errors, and the ratios for it were calculated (Table 6).
As Table 6 demonstrates, C1 level learners had the highest ratio of the uptake (71%) followed by A1 (55%) and A2 (54%) level learners.B2 level learners had the lowest ratio (50%) and the overall uptake ratio was 55%.Also, the relationship between proficiency level and uptake in details was examined (Table 7), and observed that C1 level learners were significantly better than any other group of learners in the ratios of uptake.

The effects of nature of the recasts on the uptake of learners
The instructors of the TSL courses provided different forms of corrected feedback to the errors of the learners.Table 8 illustrates the types of corrective feedback given by the instructors.
As seen in Table 8, the recasts had the highest frequency among different feedback types (n=497).Recasts were also the most common type in all proficiency levels.After recasts, the instructors used explicit correction (n=133) and elicitation (n=123) in feedback moves.The least commonly used type of the corrective feedback was metalinguistic feedback (n=49).
Amongst motivations for this study was the examination of the relationship between the type of the recasts (that is, long vs. short) on the uptake of the learners.With this  aim in mind, the frequencies and percentages of the types of recasts have been prepared.As shown in Table 9, there were a total of 497 instances of recasts, and long recasts (n=273) were observed more than shorter ones (n=224).When the ratio of uptake was analyzed, it was observed that long recasts (60%) yielded a higher percentage of uptake compared to short recasts (49%).Long recasts led to higher ratios in all proficiency levels except C1, where the learners had a very high rate of uptake (73%) towards short recasts.
When the relationship between type of recasts and uptake was examined, the resultant chi-squared analysis revealed no significant relationship (Table 10).

DISCUSSION
This classroom-based study examined the discourse of the TSL classes offered in two public universities in Turkey.Specifically, this research examined the effects of the types of error learners make (that is, phonological,  and C1) and the nature of the recasts (that is, long and short) on the uptake of the learners.The participants were students from A1, A2, B1, B2 and C1-level TSL classes.Previous research on TSL classes mostly focused on some pedagogical issues such as the use of different methods to teach the grammar and vocabulary of Turkish (Apaydin, 2007;Barin, 1992;Yayli, 2004), materials and activities for Turkish in TSL classes (Dilek, 1995;Diliduzgun, 1995;Kucukler, 2010;Ozdemir, 2010;Top et al., 2015;Zengin, 1995) and individual differences (Eker, 2010) and only few studies focused on discourse and/or corrective feedback patterns of these TSL classes namely, Fidan (2015), Inan and Fidan (2012) and Yilmaz and Yuksel (2011).The Study conducted by Fidan (2015) focused on corrective feedback patterns and preferences of TSL learners and classes and found that learners want their errors to be corrected and they mostly prefer immediate teacher correction.In another study, Yilmaz and Yuksel examined the relative effects of communication mode, salience and recasts in an experimental study.To the best of our knowledge, no other study examined the learner errors in TSL classes descriptively; moreover, no previous research focused on the relationship between the types of errors learners make, proficiency level of the learners, and uptake.
In terms of the relationship between the types of recasts and uptake, Sheen (2006) found that shorter recasts tend to have a higher rate of uptake.In a different study Philp (2003) found that shorter recasts result in noticing with a more accurate focus.However, in this study, except for learners of C1 proficiency level, the uptake ratio was higher for long recast; but there was no significant difference.

Table 1 .
The number of recorded lessons from each level of TSL courses.

Table 2 .
The types of the learner errors according to different proficiency levels.

Table 3 .
The effects of type of the error learners made and proficiency level on the uptake of learners.

Table 4 .
The relationship between proficiency level and uptake based on Pearson"s chi-square tests.

Table 5 .
Relationship between the learners" errors (specifically) and uptake based on Pearson"s chi-square tests.

Table 6 .
Numbers and percentages of uptake according to the proficiency level of the learners.

Table 7 .
The comparison of proficiency levels with C1 level based on uptake according to Pearson"s chi-square tests.

Table 8 .
Types of the corrected feedback provided by the instructors in different proficiency levels.

Table 9 .
The frequencies and percentages of the types of recasts and their uptake.

Table 10 .
Chi-square test results for the relationship between the type of the recast and uptake.