An Adaptation Study of The Measure of Pronunciation Anxiety in The Foreign Language Classroom into Turkish Language and Culture

The purpose of this study is to adapt the Measure of Pronunciation Anxiety in the Foreign Language Classroom developed by Baran-Lucarz (2016) into Turkish language and culture. The sampling of the study consists of two groups, one of whom 113 junior and senior undergraduate students majoring English language for linguistics validation and the other group was 344 undergraduate students from various faculties of the same university for the construct validity. Subsequent to the linguistic validation, confirmatory factor analysis was implemented to test the model fit in Turkish culture. The revised five-factor structure was seen to preserve and the fit indices were found to be satisfactory and valid in Turkish culture too. As for the reliability, the internal consistency values and the split-half reliability test scores of the revised scale were also found adequate. Anahtar Kelimeler İkinci dilde telaffuz kaygısı İngilizce öğretimi öğretmen eğitimi


Introduction
With a wide range of type (e.g.trait vs. state anxiety, achievement anxiety, discipline-specific anxiety, and test anxiety), source (e.g.genetics, school, and work) physical and affective symptoms (e.g.heart palpitations, trembling, restlessness, trouble concentrating and irritability) and outcomes (e.g.psychological disorders, educational/ occupational failure, social avoidance and prejudice), anxiety has been diagnosed.Many disciplines including Second Language Acquisition (SLA) research and pedagogy have comprehensively investigated anxiety due to its ubiquitous influence on learners' language acquisition and performance (Dewaele, Petrides, & Furnham, 2008;MacIntyre & Gardner, 1994).As Szyszka (2011, p. 287) argues "the articulation of phonological features, represented by both segmentals, such as vowels and consonants, and suprasegmentals, such as weak forms, linking, assimilation, stress, rhythm and intonation may be physically affected by the feeling of apprehension".Likewise, Sparks, Ganschow and Javorsky (2000) state that students with poor first language skills are anxious when they read in another language.
That there is no unique form of anxiety and the influence of anxiety on individuals is possibly due to imprecise conceptualization and measurement of early foreign language anxiety (FLA) studies on the nexus between anxiety and achievement.For this reason, the relevant research may have provided inconsistent and confusing results (Scovel, 1978).Due to the pervasiveness of anxiety in diverse fields, it has been conceptualized and investigated in terms of types, causes and outcomes.A plethora of research (e.g., Aydın, 2013;Bekleyen, 2009;Çağatay, 2015;Horwitz, 2000Horwitz, , 2001;;Lee, 2016;Lev& Keysar, 2010;MacIntyre, 2002;Tóth, 2007) has reached a consensus that there is a negative correlation between language anxiety and performance and affective aspects of foreign language learning and use.
Through a very recent perspective of inquiry, pronunciation anxiety (hereafter PA) is considered to be one of the factors, which affects willingness to communicate in a second/foreign language.Baran-Lucarz (2014b, p.38) defines pronunciation anxiety as follows: Pronunciation anxiety is a feeling of apprehension experienced by FL learners either in the FL classroom or natural setting, deriving from negative FL pronunciation self-perceptions, fear of negative evaluation, and beliefs about the importance of pronunciation, difficulty of learning and the sound of the FL pronunciation, evidenced by typical cognitive, physiological/somatic and behavioral symptoms of being anxious.The construct of FLA (Horwitz, Horwitz, & Cope, 1986) have inspired several language-specific studies focusing on speaking anxiety (Gregersen & Horwitz, 2002), writing anxiety (Atay & Kurt, 2006;Cheng, Horwitz, & Schallert, 1999;Jebreil, Azizifar, Gowhary, & Jamalinesari, 2015;Woodrow, 2011), reading anxiety (e.g.Matsuda, Gobel, 2001;Saito, Garza, & Horwitz, 1999), listening anxiety (Elkhafaifi, 2005;MacIntyre & Gardner, 1991;Melanlıoğlu, 2013;Valizadeh& Alavinia, 2013;Zhang, 2013) or grammar anxiety (VanPatten & Glass, 1999).Nevertheless, despite many studies focusing on L2 pronunciation perceptions, needs, strategies and instructional approaches in foreign language classroom, to our best knowledge, there was no theoretical and empirical initiative focusing on pronunciation anxiety until the most recent research series in Polish context (Baran-Łucarz, 2013(Baran-Łucarz, , 2014(Baran-Łucarz, , 2016)).Baran-Łucarz (2016) developed and tested the construct of PA in terms of reliability, validity and factorial design using university level sampling group with quantitative and qualitative data and reached the final version of the model.
The questionnaire of MPA-FLC consists of five sub-dimensions touching upon various perspectives such as classroom anxiety and FL oral performance apprehension, fear of negative evaluation related to pronunciation, pronunciation self-efficacy and self-assessment, pronunciation self-image and beliefs.This questionnaire is designed merely for classroom settings, though PA is often experienced in real-life situations too.The theoretical framework of the questionnaire has been based on language anxiety dynamics such as test anxiety, fear of negative evaluation (Horwitz et al., 1986), potential sources of language anxiety (Young, 1991, p. 427) and self-perceptions (e.g., Gardner & MacIntyre, 1993).

The Measure of Pronunciation in Foreign Language Classroom (MPA-FLC)
MPA-FLC, which seems the first construct to measure the sources and levels of the pronunciation anxiety in L2 classroom, was developed by Baran-Lucarz (2016).The topic has been handled by some previous research quantitatively and qualitatively (Baran-Lucarz, 2014a, 2014b, 2013).Since the construct is new, it needs to be empirically tested in different contexts and languages with diverse samplings.This model consists of 40 items with five subscales.

Subjects
The adaptation phase was conducted with 113 Turkish university students for linguistic validation.Data were gathered by means of convenience sampling method to get the most available participants for sample.The students have majored in English language teacher education (ELT), and they have provided data for the transliteral equivalence test.Four academics at ELT department and three at Turkish language education department contributed to the study as language experts for comparing and contrasting the translated items.

Procedure
To implement the adaptation of the instrument into Turkish context in terms of language and culture, the required permission was taken from Baran-Lucarz by means of e-mailing.
The instrument then was separately translated into Turkish by three faculty members at the ELT department, given the suggestion that minimum two forward translations should be conducted (Beaton, Bombardier, Guillemin, & Ferraz, 2000).The translated items were then compared, contrasted and reached one agreed form in Turkish language.
In the second phase, three academics at Turkish language education at the same university were asked for evaluating the translated Turkish form in terms of accuracy, appropriateness and intelligibility.Necessary revisions through the suggestions were made.The Turkish form was then back translated into English separately.This new English form and the original English instrument was compared and evaluated, finally both were overlapped by three academics including the author who all major in English language education.The final form was also discussed in terms of content by the academics concerned.As the scale was evaluated in terms of content by the academics at the department of English language education, the item (i=18) related to problematic pronunciation of some English sounds such as the sound 'th' was handled, and the sound 'w' was suggested to be included in the item concerned, since the academics at the English teaching department widely agree that the sound 'w' was problematic among Turkish L1 speakers in terms of pronunciation.The scale was then finalized in regard to linguistic equivalence.

Analysis
The original construct was firstly implemented to 113 juniors and seniors majoring in ELT department of a state university, secondly, the translated Turkish version of the construct was administrated to the same group after 10 days and the correlation analysis between the source and target versions of the construct was implemented to measure the factor structure and the reliability.The correlation scores are given in Table 1.Table 1 displays the correlation direction and strength of two language versions of the PA construct.The Pearson correlation coefficients suggest a positive relationship for the sub-dimensions of the scale.Given the statistically significant and positive relationship, the Turkish version of the construct can be accepted to represent the original construct.Subsequent to the linguistic validation of the scale, another subject group was obtained by means of convenience sampling method to get the most available participants for sampling.Thus, the subjects were 344 university students of the same university from various departments (i.e.dentistry, economics, education, engineering, medicine and tourism), who were required to take English course at least two semesters.English language proficiency has also been perceived to be important for these students' prospective careers.The translated scale was implemented to this group to test the validity and reliability.Throughout the adaptation process, at the end of the analyses for linguistic validity and pilot implementations, some of the total items correlation scores were seen to be below the critical value .30.The items were discussed again with the field experts, and these items were seen not to fit the whole scale.Finally, since 15 items were determined to be included in the other items of the scale, these 15 items were excluded from the scale.To avoid any influence for content validity, statisticians' and field experts' opinions were taken.Subsequent to the exclusion of the items concerned, the scale with 25 items was tested in terms of construct validity and reliability analyses.The internal consistency coefficient value of the 25-item scale was found to be .94and the item total correlation values ranged between .33 and .71.

Construct Validity
The model fit of the revised MPA-FLC scale was tested by means of first-order confirmatory factor analysis.The original scale was comprised of 40 items and 5 factors were investigated in terms of cross cultural fit.The findings of the factor structures at the end of the first-order confirmatory factor analysis are shown in Figure 1.In the revised MPA-FLC scale with 5 factors, CFA scores about the factor loading scores seem to be satisfactory.Given the factor loadings, they seem to range between .30and .82.Since the lower bound of the factor loadings should be at least .30(Kline, 2011;Seçer, 2015;Tabachnick & Fidell, 2001) in psychological tests, the factor loadings of current items can be considered to be acceptable.The model fit indices of the scale comprised of 25 items and 5 factors are given in Table 2.As Table 2 shows, the model fit indices of the MPA-FLC with 25 items appear to be acceptable (Marcoulides &Schumacher, 2001;Kline, 2011).The score of .95 is accepted adequate for GFI, CFI, NFI, RFI and IFI indices, the score of .95 is accepted perfect fit.As for the RMSEA score, .08 is considered to have an acceptable value and the score of .05 is considered perfect fit (Şimşek, 2007).RMS score should be less than .08(Browne & Cudeck, 1993).

Reliability
The revised MPA-FLC scale was tested by means of internal reliability and split-half test reliability procedures.The internal consistency was found to be .77for classroom anxiety and oral performance apprehension, .90 for fear of negative evaluation, .79 for pronunciation of self-efficacy/self-assessment, .71for pronunciation of self-image, and .74for beliefs.The total cronbach alpha (α) value was found to be .84.The total split-half test score was also found to be .74.Given that .70 is accepted as the critical score in scale development and adaptation studies, the scale can be said to have adequate values (Fraenkel, Wallend, & Hyun, 2012;Nunnaly & Bernstein, 1994).

Discussion and Conclusion
Speaking and particularly appropriate pronunciation in English appears to be one of the widely acknowledged challenges in foreign language education at Turkish context.Towards this problematic issue, more research studies to address the oral communication in a foreign language in terms of perceptive and affective perspectives will contribute to a better understanding of how learners can minimize their psycholinguistic barriers to appropriate language use.This study aimed to adapt MPA-FLC, a novel measure to quantitatively investigate pronunciation anxiety in foreign language classroom, with acceptable reliability and validity into Turkish language and culture.At the very outset of the study, the linguistic validity of the scale was attempted to strengthen, and for this purpose, expert opinions from English Language Teaching, Turkish Language Teaching and Measurement and Evaluation Departments were taken.Subsequent to the linguistic procedures, pilot implementations were conducted and correlation scores were investigated.The internal reliability value of 25-item scale was found to be .94and the item-total correlation values were found to be between .33 and .71.These scores were considered to be adequate in terms of both internal consistency and correlation coefficient (Robinson, Shaver, & Wrightsman, 1991).Confirmatory factor analysis was conducted to the five-factor measure of MPA-FLC to test the model fit.Some of the items were seen below the critical value (.30) and did not fit the whole scale.Therefore, these items were excluded from the scale and the fit indices were recalculated.The revised scale form consisting of 25 items with five factors displayed appropriate model fit and the five-factor structure of the scale can be said to preserve on Turkish sampling.
The adapted version of MPA-FLC was also tested in terms of consistency by means of internal consistency and splithalf test analyses.The total cronbach alpha (α) value was .84,and the total split-half test score was .74.Since the obtained scores were found to be over the critical value of .70 (Fraenkel et al., 2012;Nunnaly & Bernstein, 1994;Pallant, 2005), the Turkish scale can be considered to have adequate reliability to be implemented.
Given these statistical procedures, the adapted and revised version of MPA-FLC can be used as a reliable instrument with acceptable validity in a different context and language.However, since the study was conducted in a single context, this may cause some limitations in terms of generalizability.To enhance the reliability level of the adapted version of the scale, it should be implemented on diverse sampling groups.

Figure 1 .
Figure 1.Path diagram about CFA scores of the MPA-FLC