How effective are recasts in teaching / θ / and / ð / to EFL learners ?

This s tudy explored the effect of recasts as corrective feedback (CF) on teaching the sounds /θ/ and /ð/ to Iranian EFL learners , which are absent in Persian. A group of 30 intermediate level s tudents s tudying English were assigned randomly to the experimental group (EG) and control group (CG). The EG was instructed by giving recasts as CF; while the CG received no feedback when they were taught these sounds . In this experimental study, two teacher-made tests on sounds /θ/ and /ð/ were administered to the participants before and after the treatment. The collected data were analysed using paired samples t tests . Language learne rs in the recast group outperformed those in the CG in the posttest. The results showed that recast was an effective feedback type in teaching both the sounds . Findings of the present study can help language teachers and teacher trainers in teaching these complex sounds to Iranian EFL learners .


Introduction
Pronunciation should be worked on among the students who learn English as a second or foreign language (Levelle & Levis, 2014;McCrocklin & Link, 2016).Unfortunately for such students, pronunciation is usually paid less attention than the other skills since they are assumed to be more prominent (Isaacs, 2009;Kelly, 1969;Lang, Wang, Shen & Wang, 2012).In effect, development of pronunciation is not that much easy, and thus, pronunciation learning is regarded as a complicated task since it needs knowledge of appropriate sounds in particular contexts as well as the use of vocal organs to articulate those sounds, which needs extensive practice and feedback.Skill s and strategies are required to enrich the students in their pronunciation so that they can approach to native -like pronunciation.As a result, the more similar one's pronunciation is to that of a native speaker, the more likely his speech will be recognised as intelligible and understandable (Riaz, 2015).
Among the teaching techniques, many researchers have recently been interested in exploring the contribution of corrective feedback (CF) in second/foreign language learning.Feedback which is given on students' utterances plays a crucial role in classrooms.Brandet (2008) explained feedback as information provided which is relevant to learners' performance on a task.Lyster and Ranta (1997) recognised six different CF types based on a descriptive study, which they conducted on the interaction between teachers and students in French immersion classes.These six categories are as follows: recasts, explicit correction, elicitation, metalinguistic clues, clarification requests and repetition.
Several research studies have investigated how effective are different CF types in EFL context (Fungula, 2013;Haryanto, 2015;Karimi & Asadnia, 2015;Mohammadi Darabad, 2015;Roothooft & Breeze, 2016;Zarei & Rahnama, 2013;Zohrabi & Behboudnia, 2017 to name a few).Pedagogically, CF is an important component of form-focused instruction (FFI), referring to a teacher's response to learner's errors (Zhao, 2009).FFI focuses on catching learners' attention on some specific issues in teaching.Indeed, FFI intends to teach some particular structures and highlights them during teaching process.Relevant to FFI, CF tries to draw learners' attention on mistakes or errors that they make during language production.The present study is an attempt to probe the effectiveness of CF as a part of focus-on-form instruction in teaching the pronunciation of /θ/ and /ð/ to Iranian EFL learners.To the best knowledge of the researchers, no project has been conducted investigating the effective techniques to teach these two sounds, which are also absent in Persian.As such, applying recast as a CF can be considered as a novel contribution to the teaching of these sounds in the Iranian context.

Teaching pronunciation
In the 1970s, pronunciation teaching was considered as a crucial issue in second/foreign language classrooms, by those who centred on native-like pronunciation (especially phonemic contrasts), by applying minimal-pair drills and imitation of proper models (Celce-Murcia, Brinton & Goodwin, 1996).However, the nativeness assumptions were not reinforced by studies done on second language (L2) speech, showing that having a foreign accent is inevitable and has led many researchers as well as practitioners to consider that pronunciation is an unteachable subject, and consequently, as Celce -Murcia et al. (1996) and Levis (2005) pointed out, to completely disregard pronunciation teaching in their L2 instructional syllabi.Yet nowadays, a revived interest in pronunciation teaching is observed due to the idea that for successful L2 communication, the achievement of 'intelligible' pronunciation is required.Rather than removing pronunciation errors and thus developing speech void of a foreign accent, proponents of this view claim that intelligibility should be focused on for successful L2 communication (Derwing & Munro, 2005;Field, 2005).

Form-focused instruction and corrective feedback
The term FFI was defined by Lightbown and Spada (2013), as a type of 'instruction that draws attention to the forms and structures of the language within the context of communicative interaction.This may be done by giving metalinguistic information, simply highlighting the form, or by providing CF ' (p. 218).Several FFI research studies have been conducted on grammar instruction; however, it is not confined to this category alone.Indeed, form has been substantially used in SLA research to represent grammar or grammatical form, although this reductionism has not been approved by some scholars (Ellis, 2001).The term form he adds, is intended to include phonological, lexical, grammatical and pragmalinguistic aspects of language (Doughty & Williams, 1998;Ellis, 2001).
CF is a prominent aspect in FFI, as it focuses on drawing learners' attention to the mistakes and errors that they make during their production.CF has been practiced in many English as a Second Language (ESL) or English as a Foreign Language (EFL) classes.A large number of studies on CF have been conducted in morphosyntactic aspect (Kim & Han, 2007;Lyster, 1998;Mackey, Gass & McDonough, 2000), grammatical aspect (Ammar, 2003;DeKeyser, 1998;Ellis, 2007;Lyster, 2004;Sheen, 2011;Yang & Lyster, 2010), lexical aspect (Dilans, 2010;Elgort, 2011;Mackey & Goo, 2007) and pragmatic aspect (Joan & Kaya, 2006;Nguyen, Pham & Pham, 2012;Nipaspong & Chinokul, 2010;Takimoto, 2008).Research has also been found in the area of phonology (Jensen & Vinther, 2003;Li, 2012;Lyster, Saito & Sato, 2013;Sato & Lyster, 2012;Sheen, 2004Sheen, , 2010;;Zhao, 1997), however, few studies, if any, have been found to investigate application of recasts as a feedback type in teaching pronunciation of /θ/ and /ð/.Recasts engage the teacher's reformulation of all or part of the student's production, except for the error, as shown in the following example: T: When you were in school?L: Yes.I stand in the first row T: You stood in the first row?L: Yes, in the first row, and sit, ah, sat the first row (Ding, 2012) 1.3.Empirical studies Saito and Lyster (2012) were among the pioneers who considered the application of FFI techniques to teach new sounds in conjunction with the production of /ɹ/ to Japanese learners.The results showed that providing CF (i.e., recasts) in situations, where learners have mispronunciations during FFI treatment contributed greatly in improving their L2 pronunciation.Indee d, the dual pedagogical function of CFs were responsible: Pronunciation-focused recasts provided students with pronunciation models while, at the same time, eliciting self-modified output.
In Mohammadi Darabad's (2014) study, the effect of two CF techniques (recasts and prompts) was explored on students' pronunciation performance.Seventy-two students from SAMA High School in Ardabil were assigned as the participants of this study.Over the two-week treatment, the -s and -es ending pronunciations, which was difficult for these learners was instructed.Group mean scores were compared through a one-way and repeated measure ANOVA.The results indicated that CF types had a positive effect on the learners' accuracy in pronunciation.Recasts were even more effecti ve than prompts to increase accuracy in pronunciation of -s and -es ending words.Dekeyser (1994) examined the effect of teacher's recast on students' phonological errors.He found that, recast was effective in students' learning, especially in phonetic le arning.By the same token, Price (2011) examined the effect of different types of teacher's feedback on students' pronunciation and reached the conclusion that recast can be an effective form of CF.Long, Inagaki and Ortega (1998) probed the effectiveness of recasts in adverb placement among learners of Spanish and in required adjective order and a preferred locative construction among learners of Japanese.The findings revealed some short-term benefits for recasts over speech models (i.e., not contingent on learners' utterances).Mackey and Philp (1998) explored interactions with and without recasts.They, indeed, investigated the impact of recasts on the production of question forms.Their results denoted that advanced learners claimed to benefit more from interaction with recasts.Other classroom studies (Lochtman, 2000;Lyster & Ranta, 1997;Panova & Lyster, 2002) conducted on different CF types have also reported that the most frequently used CF was recasts.Sheen (2007) compared two types of feedback given on English articles: Recasts and a type of feedback including correct forms and the explanation (e.g., 'You should use the definite article 'the' because you've already mentioned 'fox').The experimental group (EG) who received the correct forms and explanation, significantly outperformed the recast and control groups (CGs), whereas, the CG even performed significantly better than the recast group.She concluded 'the more informative type of correction resulted in the acquisition of articles whereas simply providing learners with the correct form through recasts did not' (p.318).Ammar (2008) conducted the quasi -experimental study, by using an oral picture-description task, that revealed prompts and recasts are more effective than no feedback and that prompts may be more effective than recasts in leading to SL morphosyntactic development, especially for low-proficient learners.
Haryanto's ( 2015) study aimed at describing teachers' CF strategies used by the teachers, the timing of CFs and their influences on students' psychological performance during speaking activity at the Daffodils English course.The results indicated that recast as one corrective strategy proved to be effective in speaking activity of EFL learners.Moreover, Baleghizadeh and Dadashi ( 2011) examined the effect of recast feedback on students' spelling.They found that recast feedback is effective in rectifying students' spelling errors.Noruzi Azar (2012) investigated whether teachers' recasts are effective in promoting students' oral performance.The statistical results had revealed that recasting led to the mastery of oral performance, and teacher's recasts had a positive effect on learners' oral performance.
However, there are some studies which did not regard recast as an effective CF.Havranek (2002) studied 207-EFL learners across different levels who received direct feedback and recast from their teachers.She found that feedback that highlighted the correction provoked learners to direct their attention to the error and make correction.In this sense, recast was not effective.Likewise, Lyster and Ranta (1997) who collected data from different grades of French students and were primarily concerned with the reaction by the student immediately following a recast.They found that recast did not have an impact on subsequent production.Therefore, the results showed that most teachers liked recast, but it led to the lowest rate of uptake including the lowest rate of repair.Becoming proficient at pronunciation in EFL context, where there is no native speaker available to model and follow, makes it an extremely perplexing activity for students to acquire such languages and correct pronunciations of the words of such languages.As Geylanioglu and Dikilitas (2012) put it 'The difficulty posed by pronunciation is closely related to little exposure to interaction with native speakers, distinctive phonological system of L1 as in Turkish, a shallow orthographic language' (p.38).Therefore, it is a crystal clear fact that students of L2 assume achieving the pronunciation of that language to be problematic to learn and understand, unless they either have highly educated teachers with correct pronunciation and sufficient practice or use effective techniques to learn correct pronunciation.Likewise, another problem with pronunciation is that some sounds are absent in some languages while are present in other languages.It means there is no exact corresponding among sound systems of all languages.Two phonetic sounds /θ/ and /ð/ (th) are absent in Persian; as such, it is problematic for Iranian EFL learners to learn such sounds.Thus, it is worth employing an appropriate method of teaching these sounds to EFL learners and this study was an attempt to teach these two problematic sounds to Iranian EFL learners effectively.
The current study is motivated by the question whether recasts can affect teaching pronunciation of sounds /θ/ and /ð/ to Iranian EFL learners.The following research questions (RQ) guide the study: (1) Do recasts affect pronunciation of /θ/ among Iranian EFL learners?(2) Do recasts affect pronunciation of /ð/ among Iranian EFL learners?

Method
This study followed an experimental pretest-posttest treatment design in which random sampling procedure was used and it was conducted in a language institute in Dezful, Iran.In the current study, attempt was made to probe the effectiveness of recasts in teaching pronunciation of /θ/ and /ð/ to Iranian EFL learners.The participants were randomly assigned to an EG and a CG.In the EG, recast was the CF used and in the CG, the participants received no feedback.Each group followed the same teaching procedure, pretest, treatment and posttest.

Participants
The first participant group of the present study consisted of 30 female intermediate Iranian EFL learners.The participants' age range was between 18 to 26 years old and they were Iranian and their mother tongue was Persian1.They were selected randomly from the EFL learners studying English at Tak English Language Institute in Dezful, Iran.They took a placement test.It was the Quick Placement Test developed by UCLES in 2001.The placement test was administered among 60 students; those whose scores on the test fell one standard deviation above or below the mean score were selected and formed the sample population of the study.The placement test was administered to ascertain that the sample was homogeneous in terms of English proficiency.They were intermediate EFL learners.Then, the participants were randomly assigned to two groups: One EG and one CG.Each group included 15 participants.
The second group of participants were three non-native raters who were Iranian and their age ranged between 30 and 45.They were English language teachers who had IELTS certificates and their overall band scores were 8.0 or above.They were both BA and MA Teaching English as a Foreign Language (TEFL) holders and their English teaching experience varied from 5 to 15 years.They were female and their mother tongue was Persian.They were recruited to rate the speech tokens recorded during the pretest and posttest.
The third group included the instructor who was the first author of the study.She was Iranian and her native language was Persian.She was an MA student of TEFL and has had taught English in language institutes in Iran for 4 years.In order to ensure the quality of instruction, the classes were held on an odd-even schedule.The EG had classes on even days and the CG on odd days.The whole treatment period lasted for around three weeks.

Materials and instruments
In the current study, a series of materials were used during the data collection period.A set of words which included the sounds /θ/ and /ð/ were prepared by the researchers and were presented in isolation or in context.Since there was no material for teaching these sounds, the researchers prepared the required teaching materials by developing individual and contextualised words considering the learners' proficiency level.In addition, certain instructive video and v oice clips were also downloaded from Youtube and Engvid websites (www.youtube.comand www.engvid.com).The clips were specifically developed by English native teachers for non-native students.In the video clips, these two sounds (/θ/ and /ð/) are taught by showing the place of articulation and repeating words with target sounds in isolation and in context.
A quick placement test (UCLES, 2001) was administered to guarantee the participants' homogeneity in terms of their proficiency level.This placement test contained 60 multiple-choice questions on grammar and vocabulary and the participants' responses were scored on a scale of 60 points.Moreover, a pronunciation placement test was employed to ensure that the participants were at the same level in terms of pronunciation.It was downloaded from teacherspayteachers website (www.teacherspayteachers.com) developed by Gunther Breaux.This test uses the fact that ESL/EFL learners have problems differentiating these sounds: Right, white, light, night.Thus, it can quickly and accurately predict English speaking ability and overall English ability.
A pronunciation test was administered to the participants as the pretest and posttest before and after the treatment to measure their phonetic ability before the treatment and to measure their improvement after the treatment.The pretest and the posttest which were in parallel consisted of some individual and contextualised words that included these sounds /θ/ and /ð/.The tests were researcher-made and were composed of four parts.The parts were single words with target sounds, some sentences, texts and two questions for interview.The interview questions aimed at examining the pronunciation of these two sounds in a context which is natural.Indeed, it was intended to see how the participants pronounced these sounds in a naturally occurring speech.The tests enjoyed a Likert scale in which each word/sentence/text with the target sound should be scored in the following way: 1 = very heavy non-native pronunciation 2 = poor pronunciation 3 = reasonable pronunciation 4 = close to native pronunciation 5 = native like pronunciation.The pronunciation test was piloted on a group of 20 learners who shared some commonalities with the sample population of the study.The reliability coefficient was calculated and it was reported as 0.78.The validity of the tests was confirmed by three experts in the field so that where required, the words were modified based on the experts' comments to improve.

Sampling procedures
To begin the data collection procedure, a placement test was administered to 60 Iranian EFL learners.30 learners were selected and formed the sample population of the study based on their scores on the placement test.The selected participants were randomly assigned to two groups of 15, one EG and one CG.Then, the pretest was given to all participants of the study.The test consisted of four parts which the participants read aloud and they were recorded.After the pretest was administered, the three raters scored the test individually.Then, inter-rater reliability of the test was computed and its Cohen's Kappa coefficient was reported as 0.75.Therefore, the test had acceptable reliability and the participants' mean score of the three scorings were determined as their performance in the pretest.
In the recast group, the participants were taught through using recast as CF.The sounds in question were first taught applying different techniques such as video as well as voice clips, the teacher's direct elaborations on these sounds and employing different examples in isolation, in words, or in sentences.Then, the participants read some texts which contained these sounds.If there were any errors in pronunciation of these two sounds, the teacher indirectly reformulated the errors o r provided the correction without a direct indication that the student's utterance was incorrect.The CG was taught through the same materials; moreover, the same teaching procedure was followed except for the CF.They received no feedback when an error was made by the participants.
Treatment period took 8 sessions and each session lasted for 45 minutes.At the end of the eighth session, the posttest was administered to all groups.The data collection procedure from its inception to its termination took approximately three weeks during which the placement test, pretest and posttest were administered.After the administration of the posttest, the same raters scored them.To ensure the rater-reliability, inter-rater reliability test was computed and the Cohen's Kappa coefficient was 0.73.

Data analysis
Normality of the data was checked through Kolmogrov-Smirnov test.To demonstrate whether there was a statistically significant difference between the participants' performance on pretests and posttests, two paired-samples t test -one for /θ/ and one for /ð/ were conducted to see whether each group had a significant change after the treatment.

Normality of data
The Kolmogorov-Smirnov test was used to examine the normality of the data collected from the pretests and posttests are shown in Table 1.The results of this test showed that in both groups, distribution of collected data was normal.The collected data included two pretests of sound /θ/, two pretests of sound /ð/, two posttests for sound /θ/ and two posttests of sound /ð/.For both groups, the significance level was higher than the probability level (0.05).Therefore, paired-samples t tests can be run on the data.

Homogeneity of the participants on placement test
In order to make the sample homogeneous in terms of English proficiency, a placement test was administered among 60 Iranian EFL learners.30 learners were selected as the participants of this study.They were selected based on the mean score of the test.Those whose scores fell one SD above or below the mean were selected.The descriptive statistics of the placement test are presented in Table 2.

Research question (RQ) one
The first RQ was: Do recasts affect pronunciation of /θ/ among Iranian EFL learners?The purpose of this question was to probe whether recasts were effective in teaching sound /θ/.An independentsamples t test as well as a paired-samples t test was run on the data and the results are displayed in Table 3.The CG's mean score equaled 13.22, while the EG's mean score turned out to be 15.31, which is not quite a large difference.Hence, to determine, in solid terms, whether or not the difference between these two mean scores (and thus the two groups) was statistically significant, the p value under the Sig.(2-tailed) column in the t test Table 4, had to be checked.Based on the information presented in the Table 4, it was found that there was no statistically significant difference between the posttests of the EG and the CG, as the p value was greater than 0.05 (p = 0.183).To check whether the EG had any improvement after the treatment, the pairedsamples t test was computed and the results are displayed in Table 5.Table 5 showed that there was a significant difference between the participants' difference in pretest and posttest of sound /θ/ (M = 16.73,SD = 1.40), t(14) = 9.02, P = 0.000 < 0.05 (two-tailed).The P value was less than 0.05 (P = 0.000); thus, it was claimed that recast was effective in teaching sound /θ/.

Research question two
The second question was: Do recasts affect pronunciation of /ð/ among Iranian EFL learners?This question explored the effect of recast on pronunciation of /ð/.To this end, through an independentsamples t test as well as a paired-samples t test run on the data, the participants' performance in pretest and posttest was compared.The results are shown in Table 6.Table 6 shows that the CG's mean score was 12.35 and the EG's mean score was found to be 14.95.To understand whether the difference between these two mean scores (and thus the two groups) was statistically significant or not, the researchers consulted the p value in Table 7.As it could be understood from Table 7, there was not a statistically significant difference in the posttest results of the CG and those of the EG, t(28) = 0.589, p = 0.560 (two-tailed).In other words, the two groups did not differ significantly in terms of the posttest.To check whether the EG had any improvement after the treatment, the paired-samples t test was computed and the results are displayed in Table 8.Table 8 demonstrated that there was a significant different between pretest and posttest in recast groups' performance (M = 17.00,SD = 1.46), t(14) = 3.83, P =0.002 < 0.05 (two-tailed).Since the P value was less than 0.05 (P = 0.002), it was concluded that recast was effective in teaching sound /ð/.

Discussion
The findings emerging from the present study substantiated that there was a significant difference between recast and CG's performance in the posttest.As a statistically significant difference was observed between the mean scores of the recast group and CG, it was contended that the recast group outperformed the CG in posttest and thus, the recast feedback was an effective technique in correcting the participants' error while teaching sounds /θ/ and /ð/ to Iranian EFL learners.
One explanation for this finding may be attributed to the application of recast that can probably have mildly attracted the learners' attention toward learning and stimulated their curiosity.This finding is consistent with the findings of some other studies.For example, Saito and Lyster's (2012) study showed that recasts played an important role in changing their L2 pronunciation performance.Moreover, Mohammadi Darabad (2014) revealed that CF conditions had a positive effect on the learners' pronunciation accuracy.The recasts were even more effective than prompts in increasing accuracy in the pronunciation of -s and -es ending words.Dekeyser's (1994) and Price's (2011) studies were also in compatible with the results of the current study, substantiating that recast was effective in students' pronunciation.
Recasts can be used as a facilitator in learning complicated subject matter, because they provide scaffolded help for the students particularly in contexts where students' level of ability is lower than the teaching point.Contemporaneously, recasts are examples of positive evidence (Braidi, 2002;Leeman, 2003) and as such, they are to simplify the encoding of new target points when they are brought in proper discourse contexts.Moreover, Doughty and Varela (1998) reported that applying formed focused instruction which includes recasts in communicative classroom is effective.Van Patten (2003) also advocated that explicit CF in the form of negotiating for meaning can help learners notice their errors and create form-meaning connections, and this facilitates acquisition.
In line with the results of the current study, a variety of techniques have also been used to assess the effectiveness of recasts.Ammar (2008), Haryanto (2015), Lochtman (2000), Long et al. (1998), Lyster and Ranta (1997), Mackey and Philp (1998), Panova and Lyster (2007) and Sheen (2007) all claimed that recast could contribute positively in learning a variety of language elements such as adverb placement, adjective order, question forms, English articles, and oral production.Few classroom studies have examined the impact of recasts on interlanguage development.Doughty and Varela (1998) reported changes in learners' L2 use following a period of corrective recasts (i.e., recasts that included an attention-focusing element).Moreover, Baleghizadeh and Dadashi (2011) found that recast feedback is effective in rectifying students' spelling errors.The finding also verifies that found by Noruzi Azar (2012), who revealed that recasts are effective in promoting students' oral performance.
However, the findings of the present study were not compatible with the findings of some previous ones.For example, Havranek (2002) found that recast was not effective at different levels.Likewise, the findings run counter to Lyster and Ranta who claimed that recast did not have an impact on production.Therefore, the results showed that most teachers liked recast, but it led to the lowest rate of uptake including the lowest rate of repair.

Conclusion
Concerning the effectiveness of recasts in teaching these two sounds, the conclusion that is drawn is that reformulation of the whole sentence or a part of it the learners identify their errors, indeed, the teacher's reformulation of the student's production gives the hint to the learner how to correct the error that has been made.In addition, FFI is concerned with CF which catches learners' attention on some specific issues in teaching.The findings of the study suggested that recasts can catch the learners' attention successfully, and it is concluded that recasts are very efficient in removing learners' mistakes and errors, and thus, recasts can be helpful in promoting L2 development.
The present study focused on the efficacy of employing recasts as a CF in teaching two absent sounds in Persian.It was of high significance from two perspectives, the way of providing feedback and teaching absent components of a foreign language to EFL learners.Therefore, different groups can benefit from the results of this study.Teachers should know how to teach and what technique to apply according to the context in which they teach, such as EFL context or ESL contest, and the proximity of phonetic systems of the two languages.Thus, the effectiveness of recasts in an EFL context in which these two sounds were absent can be a great help in teaching these sounds in other EFL and ESL contexts.This technique can help teachers enrich the students in their pronunciation; thus, they can come closer and closer to native-like pronunciation, and as such their pronunciation will be recognised as intelligible and understandable (Riaz, 2015).Also, this can influence feelings of belonging to the community of the native language as well as confidence and having interest in communication in this language.Teacher trainers can benefit from the findings of the study, as well.They can teach pre-and in-service teachers to get familiar with different types of feedback and the way of teaching absent components.Material developers can also use the findings of this project in designing materials for language learners in a way that they can put emphasis on the recast as a type of CF in the teacher's book, which accompanies the student's book and workbook.
Since the data in this study have been taken merely from the Iranian context, it is important not to overgeneralise the results of the study and admit that replicational studies can contribute to building a rich body of knowledge.

Note
1. Dezful is a city in Khouzestan Province located in the South-West of Iran.As this province has border-lines with Iraq, Arabs live in certain cities.Although the researchers already knew that no Arabs live in Dezful and all the inhabitants are native speakers of Persian, the participants were checked in terms of being born to Persian mothers and fathers.

Table 1 .
Normality of data

Table 2 .
Descriptive statistics of placement test

Table 3 .
Descriptive statistics for the EG and the CG on sound /θ/

Table 4 .
Independent-samples t test for the EG and the CG on sound /θ/

Table 5 .
Paired-samples t test for recast group on sound /θ/

Table 6 .
Descriptive statistics for the EG and the CG on sound /ð/

Table 7 .
Independent-samples t tes t for the EG and the CG on sound /θ/

Table 8 .
Paired-samples t-test for recast group on sound /ð/