What Affects Middle School Students ’ English Anxiety in the EFL Context ? Evidence from South Korea

This paper examines what affects adolescents’ English anxiety in the English as Foreign Langauage (EFL) context. A total of 414 adolescents in South Korea participated in the study and the AMOS 20.0 was used in structural equation modeling for statistical analysis. The results are as follows. Girls showed a higher level of English anxiety and self-directed learning ability than boys. Second, adolescents’ English self-efficacy had a partial mediating effect on the relationship between parental pressure expectation and English anxiety. Third, parental pressure expectation had a significant effect on English anxiety through self-directed learning and English self-efficacy. Fourth, a gender difference in the paths of the models was significant. This provides supporting evidence to many educators and parents for the implementation of effective support practices for adolescents who learn English in EFL contexts similar to Korea to reduce thier English anxiety.


Introduction
English anxiety is the unstable physiological or cognitive reactions to stressors arising from the English learning process [1,2].Many South Korean students suffer from a high level of English anxiety, which has a debilitating effect on their English proficiency as well as their English learning motivation.This seems to be relatively more serious for students from China, Korea, and Japan, who are considered to be members of the Confucian Heritage Culture (CHC), and were found to be relatively more serious than those from other ethnic groups [3].This tendency can be interpreted as a result of the learning environment of Northeast Asian countries, which pays attention to Confucian values, such as 'face' and 'silence' [3].However, although this interpretation seems valid when compared to Western cultures, it does not fully explain the nature of Korea's English learning environment.
English anxiety experienced by Korean students is closely related to the academic grading system in South Korea, and middle school years are considered to be a transition period in English learning as midterm and final exams are used as an evaluation tool for subjects, including English [4].Although South Korea is not a country that uses English in daily circumstances, English is considered as one of the most important subjects that evaluates students' academic achievement and their college entrance examination.English is one of the subjects of the college entrance exam and the exam requires accuracy in grammar and vocabulary, which pushes students to memorize grammatical rules and the meanings of English words.Eventually, students are under high pressure to perform well, which in turn causes stress and anxiety.Even after college entrance, it is necessary to get a high score on English proficiency tests, such as the TOEIC or TOEFL, since a high score is associated with opportunities for academic achievement, such as participating in a student exchange program as well as obtaining in a good internship placement.After college graduation, students are required to have scores for English proficiency tests to get into the company they prefer.Getting a high score on an English test is an inevitable phenomenon in Korea.As a result, students are put in a situation that forces them to study English extensively to obtain excellent grades in both academic achievement and entrance examinations, leading to an uneasiness in English.In fact, Korean students' English learning motivation consistently decreases until 9th grade, but increases again gradually starting from 10th grade to 12th grade [5], which means that English learning motivation among middle school students is the lowest.Because Korean students are assessed based on accuracy rather than fluency, they are afraid of being incorrect when they speak English even outside the classroom.
Moreover, South Korea is also known for 'education fever', its strong parent involvement in education, and most South Korean parents are committed to their children's academic achievement and they often push their children to study in a coercive manner [6,7].The enthusiasm for education leads to a high level of participation in private education; according to a survey by the National Statistical Office, it was found that 39.8% of kindergarten to 12th grade students were receiving private English education in 2017 [8].The statistics reveal that many students are under pressure to get good grades at school and parents are deeply involved in this process.As higher pressure by parents results in a higher level of children's test anxiety [9,10], it is probable that children of parents with higher education fever might suffer more from English anxiety when taking English tests.In contrast, a study found that the expectations and pressures of Korean parents had a positive influence on children and led to positive predictions of academic achievement [7].Therefore, it is necessary to verify the relevance of Korean parental pressure expectations to English anxiety.
Pressure from parents can affect English anxiety through self-directed learning and English efficacy.Authoritarian parenting [11] and parenting control [12] were negatively related to children's self-regulated learning, possibly resulting in damaged academic efficacy [13].Previous research [14] suggested that strategies on cognitive and behavioral regulation, such as organization and seeking peer assistance, had a significant impact on students' verbal efficacy, leading to the proposal that self-directed learning influences English self-efficacy.Moreover, a higher sense of parenting control could have a harming effect on academic self-efficacy [15] by increasing anxiety arousal [16].From studies conducted in Korea, English self-efficacy was considered as an important factor in explaining English anxiety.The higher the level of English self-efficacy, the lower the English anxiety [17] and this tendency works the same for English writing [18].Referring to the results from previous studies, it can be inferred that when children are under pressure from their parents, it is harder for them to act on their own initiative, which may eventually reduce their English self-efficacy and could increase the level of English anxiety.However, it is difficult to find any studies that deal with both variables so it would be meaningful to examine the relationships of parent pressure expectations, children's self-directed learning abilities, and English self-efficacy, as well as English anxiety, together.
Furthermore, there is a possibility that gender differences could appear at the level of the variables and even at the structural pathways between the variables.For example, a study in India that investigated gender differences in academic anxiety showed mixed results [19,20].In [14], it was shown that boys' academic efficacy was higher than girls' while girls reported more self-regulated learning strategies than boys.In [10], the authors showed that there were gender differences in terms of parental pressure and test anxiety.Although the previous literature did not look directly at the gender differences in the relationship of the variables, we propose that there could be a gender gap at the level of the variables as well as a correlation between the variables.Therefore, if the variables that have gender differences are correlated with each other, it can be inferred that this can also lead to gender differences in inter-parameter relationships.Furthermore, if we find gender differences in inter-variable relationships, we can implement a discriminatory approach to teach English to manage English anxiety and make English learning effective according to the gender of the students.Therefore, this study established a structural relationship that parent pressure expectations would influence English anxiety through student self-directed learning abilities and English self-efficacy.To be specific, in other words, the study attempted to verify its hypothetical paths, which include the direct path that parent pressure expectations have an effect on English anxiety as well as the indirect paths that parent pressure expectations influence English anxiety only through English self-efficacy or through both self-directed learning and English self-efficacy.Also, it aims to find the possible difference in parental pressure expectation, self-directed learning ability, English self-efficacy, and English anxiety at the level of the variables as well as the structural path between variables.As a result, this study will enable meaningful implications on English education in EFL environments, such as Korea, to be drawn to enable efficient learning of English.

Participants
The sample used for this study was obtained from questionnaires.As shown in Table 1, a total of 414 adolescents (169 male and 245 female students) aged between 12 and 14 participated from 5 middle schools in four major provinces (Seoul, Gyeonggi, Chungcheong, and Daegu).

English Anxiety
English anxiety was measured by the modified version [21,22] of the Foreign Language Classroom Anxiety Scale [1].This scale consists of the four factors, 'Negative Evaluation Anxiety (NEA)', 'Communication Anxiety (CA)', 'Class Comprehension Anxiety (CLA)', and 'Test Anxiety (TA)', and each factor consists of four questions.Each item was measured using a five-point Likert scale ranging from 'Not at all (1)' to 'Very much (5)', with a high score showing a higher level of English anxiety.Internal consistencies (Cronbach's alpha) of the four factor were good (total: 0.94, NEA: 0.77, CA: 0.90, CLA: 0.88, TA: 0.91).

English Self-Efficacy
English self-efficacy was measured using a modified version of the Self-Efficacy Scale [23].This scale consists of four questions, such as "I can understand what I learn in English class".Each item was measured using a five-point Likert scale ranging from 'Not at all (1)' to 'Very much (5)', with a higher score showing a higher level of English self-efficacy.Internal consistency (Cronbach's alpha) was good (0.93).

Parental Pressure Expectation
Parental pressure expectation was measured using a modified version of the Parental Influence-Family Processes Inventory [24].This scale consists of the four questions, such as, "My parents put a strain on my English study".Each item was measured using a five-point Likert scale ranging from 'Not at all (1)' to 'Very much (5)', with a higher score showing a higher level of parental pressure expectation.Internal consistency (Cronbach's alpha) was good (0.83).

Self-Directed Learning Ability
Self-directed learning ability was measured using a modified version of the Self-Directed Learning Ability Scale [25,26].This scale consists of the three factors: 'Cognitive Factor (CF)', 'Motivational Factor (MF)', and 'Behavioral Factor (BF)'.The CF consists of six questions, the MF consists of three questions, and the BF consists of three questions.Each item was measured using a five-point Likert scale ranging from 'Not at all (1)' to 'Very much (5)', with a higher score showing a higher level of self-directed learning ability.Internal consistencies (Cronbach's alpha) were acceptable (total: 0.91, CF: 0.89, MF: 0.67, BF: 0.76).

Statistical Analysis
In this study, AMOS 20.0 was used in structural equation modeling (SEM), which was applied to examine the causal relationship between parental pressure expectation, self-directed learning ability, English self-efficacy, and English anxiety.It was assumed that English self-efficacy and self-directed learning ability had a mediation effect between parental pressure expectation and English anxiety.We hypothesized that the path of influence would differ according to gender and conducted multiple group analysis.
We used the χ 2 value, RMSEA (root mean square error of approximation), TLI (Tucker-Lewis index), and CFI (comparative fit index) to determine the goodness of fit.A TLI and CFI higher than 0.9 are generally considered acceptable, and an RMSEA below 0.05 is regarded as good and around 0.08 as acceptable [27].Missing values were estimated within the model using the full information maximum likelihood (FIML) method offered by AMOS.Implementing FIML in AMOS is a superior method for dealing with missing data, which needs normality of data [28], so we checked the skewness and kurtosis of variables.Additionally, SEM techniques attempt to account for all covariance among its measures [29], and we checked the correlations among study variables.
The best-fitting model was selected for invariance testing across groups.Then, configural invariance, measurement invariance, and structural invariance were verified in a consecutive order.To compare the degree of fit of the nested models, we performed the χ 2 differences test.
A descriptive analysis, t-test, and correlation analysis was conducted using SPSS 22.0 to identify the characteristics of the study variables and to determine whether they meet the analytic criteria.

Gender Differences
Table 2 provides the means, standard deviation, and t-score of adolescents on all study variables.The significance difference between the genders was significant on English anxiety (t = −2.55,p < 0.05) and self-directed learning ability (t = −2.08,p < 0.05), where female students had more English anxiety and self-directed learning ability than male students.The degree of English self-efficacy and parental pressure expectation was not different between male and female students.

Intercorrelation and Descriptive Statistics
Table 3 presents the correlations for all study variables.The correlation between English anxiety and parental pressure expectation (r = 0.39, p < 0.001) was positive, but the correlation between English self-efficacy (r = −0.48,p < 0.001) and self-directed learning was negative (r = −0.23,p < 0.01).Because the correlations among variables were significant, the assumption of SEM was satisfied.The skewness of variables was under ±2 and the kurtosis of variables was under ±7 [30], which meets the assumption of a normal distribution required for FIML.

Structural Model Between
Parental Pressure Expectation, Self-Directed Learning Ability, English Self-Efficacy, and English Anxiety In the structural equation model that posited self-directed learning ability and English self-efficacy as mediators of the associations between parental pressure expectation and English anxiety (see Figure 1), results indicated the hypothesized model showed a good fit of the data; scaled χ 2 (84) = 339.12,TLI = 0.911, CFI = 0.937, RMSEA = 0.086.
As seen in Figure 1, the path from self-directed learning to English anxiety was not significant.To determine the most parsimonious model, we set a modified model in which the insignificant path was removed.The chi-square difference test was used to compare the nested models.Results of the chi-square difference test showed that the modified model was not significantly different from the research model (CV: 0.95 χ 21 = 3.84).The modified model (see Figure 2) showed an adequate fit to the data; scaled χ 2 (85) = 341.930,TLI = 0.911, CFI = 0.937, RMSEA = 0.086, 90% CI = 0.076-0.095.As shown in Figure 2, parental pressure expectation significantly predicted English anxiety (β = 0.24, p < 0.001) in the modified model.This indicated that higher levels of parental pressure expectation were positively related to English anxiety.Parental pressure expectation also predicted English self-efficacy (β = −0.15,p < 0.01), which in turn were negatively associated with English anxiety (β = -0.46,p < 0.001).As the parental pressure expectation became higher, the level of English self-efficacy decreased and this decreased English self-efficacy increased the level of English anxiety.Parental pressure expectation also predicted self-directed ability (β = −0.25,p < 0.001), which in turn was positively associated with English self-efficacy (β = 0.60, p < 0.01), and which in turn was negatively associated with English anxiety (β = −0.46,p < 0.001).These results demonstrated that as the parental pressure expectation became higher, the level of self-directed learning ability decreased.Additionally, as self-directed learning became lower, the level of English self-efficacy decreased, and as English self-efficacy became lower, the level of English anxiety increased.

4. Multiple Group Analysis
A multiple-group analysis was conducted to examine whether the pattern of associations in the modified model differed significantly by gender.The first step was to establish an equal pattern As shown in Figure 2, parental pressure expectation significantly predicted English anxiety (β = 0.24, p < 0.001) in the modified model.This indicated that higher levels of parental pressure expectation were positively related to English anxiety.Parental pressure expectation also predicted English self-efficacy (β = −0.15,p < 0.01), which in turn were negatively associated with English anxiety (β = -0.46,p < 0.001).As the parental pressure expectation became higher, the level of English self-efficacy decreased and this decreased English self-efficacy increased the level of English anxiety.Parental pressure expectation also predicted self-directed ability (β = −0.25,p < 0.001), which in turn was positively associated with English self-efficacy (β = 0.60, p < 0.01), and which in turn was negatively associated with English anxiety (β = −0.46,p < 0.001).These results demonstrated that as the parental pressure expectation became higher, the level of self-directed learning ability decreased.Additionally, as self-directed learning became lower, the level of English self-efficacy decreased, and as English self-efficacy became lower, the level of English anxiety increased.

4. Multiple Group Analysis
A multiple-group analysis was conducted to examine whether the pattern of associations in the modified model differed significantly by gender.The first step was to establish an equal pattern As shown in Figure 2, parental pressure expectation significantly predicted English anxiety (β = 0.24, p < 0.001) in the modified model.This indicated that higher levels of parental pressure expectation were positively related to English anxiety.Parental pressure expectation also predicted English self-efficacy (β = −0.15,p < 0.01), which in turn were negatively associated with English anxiety (β = -0.46,p < 0.001).As the parental pressure expectation became higher, the level of English self-efficacy decreased and this decreased English self-efficacy increased the level of English anxiety.Parental pressure expectation also predicted self-directed ability (β = −0.25,p < 0.001), which in turn was positively associated with English self-efficacy (β = 0.60, p < 0.01), and which in turn was negatively associated with English anxiety (β = −0.46,p < 0.001).These results demonstrated that as the parental pressure expectation became higher, the level of self-directed learning ability decreased.Additionally, as self-directed learning became lower, the level of English self-efficacy decreased, and as English self-efficacy became lower, the level of English anxiety increased.

Multiple Group Analysis
A multiple-group analysis was conducted to examine whether the pattern of associations in the modified model differed significantly by gender.The first step was to establish an equal pattern baseline model, as seen Model 1 in Table 4.This configural invariant model showed a favorable fit; scaled χ 2 (170) = 424.146,TLI = 0.912, CFI = 0.938, RMSEA = 0.060.These results indicated that the same pattern of parameters could fit the data for male and female students, but not the parameter estimates that took on the same or even similar value for the different groups.In a second step, a measurement invariance was verified by constraining each factor loading to be equal between the two groups (Model 2 in Table 4).The model-data fit indicators were χ 2 (181) = 431.823,TLI = 0.919, CFI = 0.939, and RMSEA = 0.058.Because the measurement invariant model (Model 2) was nested within Model 1, a chi-square difference test was used to compare the models.Results of the chi-square difference test showed that Model 2 was not significantly different from Model 1 (CV: 0.95 χ 11 = 19.68).This finding confirms measurement invariance and allows for group comparison.
In a third step, a structural invariance model was applied, assuming equal factor loadings and equal regression coefficients across both groups (Model 3 in Table 4).The indices indicated a satisfactory fit for this structural invariant model [χ 2 (186) = 449.416,TLI = 0.917, CFI = 0.936, RMSEA = 0.059], though it was worse than the measurement invariant model.Results of the chi-square difference test showed that Model 3 was significantly different from Model 2 (CV: 0.95 χ 5 = 11.07), which means the measurement invariant model reflects the data better than the structural invariant model.Given this loss of fit, the partial structural invariant model (Model 4 in Table 4) was explored by freeing appropriate regression coefficients, as seen in Table 5.A model with these regression coefficients freely estimated in both groups (Model 4) was a good fit [χ 2 (183) = 431.889,TLI = 0.920, CFI = 0.939, RMSEA = 0.057] and was not significantly worse than the measurement invariant model (CV: 0.95 χ 2 = 5.99).In other words, interrelations among the study variables exhibited a partially different pattern for male and female students.This partial structural model included a direct effect and two indirect effects, as shown in Figure 3. Firstly, the direct path between parental pressure expectation and English anxiety (Male: B = 0.29, p < 0.001, Female: B = 0.12, p < 0.01) were found to be significant.This indicated that a higher level of parental pressure expectation was positively related to English anxiety, and the influence of parental pressure expectation was higher for male than female students.
expectation was negatively associated self-directed ability (Male: B = −0.33,p < 0.001, Female: B = −0.14, p < 0.01), which in turn was positively associated with English self-efficacy (B = 0.81, p < 0.001), and which in turn was negatively associated with English anxiety (Male: B = −0.19,p < 0.001, Female: B = −0.43,p < 0.001).These results showed that as the parental pressure expectation became higher, the level of self-directed learning ability decreased.Additionally, as self-directed learning became lower, the level of English self-efficacy decreased, and as English self-efficacy became lower, the level of English anxiety increased.
These results demonstrated that as parental pressure expectation became higher, self-directed learning ability became lower, and this influence of parental pressure expectation on self-directed learning ability was higher for male than female students.Moreover, it was found that self-directed ability improved English self-efficacy, which in turn lowered English anxiety and this influence of English self-efficacy on English anxiety was higher for female than male students.

Discussion
This study elucidated the structural relationship between parental pressure expectation, self-directed learning ability, English self-efficacy, and English anxiety among Korean middle school adolescents, and demonstrated that there was a gender difference in the model.The key findings of the study are as follows.
First, it was found that female students showed a higher level of English anxiety and self-directed learning ability than male students.However, the degree of English self-efficacy and parental pressure expectation was not different between the two groups.The findings were consistent with previous studies that reported a higher level of English anxiety among female Secondly, the indirect path from parental pressure expectation to English anxiety through English self-efficacy was found to be significant.Parental pressure expectation negatively predicted English self-efficacy (B = −0.16,p < 0.01), which in turn were negatively associated with English anxiety (Male: B = −0.19,p < 0.001, Female: B = −0.43,p < 0.001).As the parental pressure expectation became higher, the level of English self-efficacy decreased and this decreased English self-efficacy increased the level of English anxiety.It was found that the influence of self-efficacy on English anxiety was higher for male than female students.
Lastly, the indirect path from parental pressure expectation to English anxiety through self-directed learning and English self-efficacy was found to be significant.Parental pressure expectation was negatively associated self-directed ability (Male: B = −0.33,p < 0.001, Female: B = −0.14, p < 0.01), which in turn was positively associated with English self-efficacy (B = 0.81, p < 0.001), and which in turn was negatively associated with English anxiety (Male: B = −0.19,p < 0.001, Female: B = −0.43,p < 0.001).These results showed that as the parental pressure expectation became higher, the level of self-directed learning ability decreased.Additionally, as self-directed learning became lower, the level of English self-efficacy decreased, and as English self-efficacy became lower, the level of English anxiety increased.
These results demonstrated that as parental pressure expectation became higher, self-directed learning ability became lower, and this influence of parental pressure expectation on self-directed learning ability was higher for male than female students.Moreover, it was found that self-directed ability improved English self-efficacy, which in turn lowered English anxiety and this influence of English self-efficacy on English anxiety was higher for female than male students.

Discussion
This study elucidated the structural relationship between parental pressure expectation, self-directed learning ability, English self-efficacy, and English anxiety among Korean middle school adolescents, and demonstrated that there was a gender difference in the model.The key findings of the study are as follows.
First, it was found that female students showed a higher level of English anxiety and self-directed learning ability than male students.However, the degree of English self-efficacy and parental pressure expectation was not different between the two groups.The findings were consistent with previous studies that reported a higher level of English anxiety among female students [8] [18] and higher self-directed learning abilities among girls than boys [14].Gender gaps in English anxiety can be interpreted as the girls were socialized to express their anxiety while the boys were expected to suppress their anxiety [31].
Second, the results of the structural equation modeling analysis indicated that adolescents' English self-efficacy had a partial mediating effect on the relationship between parental pressure expectation and English anxiety.The findings concurred with previous studies that found that maternal pressure reduced adolescents' sense of self-efficacy [10] as well as English anxiety [13].Also, the paper confirmed the study that academic self-efficacy had a partial mediating role on parental psychological control and test anxiety [15].The results are in a similar context to the results of the previous studies that reported a negative relation between English self-efficacy and English anxiety in elementary school students [17], as well as a negative association between English writing self-efficacy and English writing anxiety among middle school students [18].So, self-efficacy could play a central role in situations where anxiety occurs [16].It confirmed that self-efficacy could play a key mediator role not only between learning motivation and academic achievement [32], but also between parental pressure expectation and English anxiety.Additionally, English self-efficacy could serve as a buffering effect against parental pressure expectation that affects English anxiety.Therefore, enhancing students' English self-efficacy should be considered as essential to lower students' English anxiety.
Third, the indirect effect of parental pressure expectation to English anxiety through self-directed learning and English self-efficacy was found to be significant.That is, parental pressure expectation had a significant effect on English anxiety, first by having an impact on self-directed learning, then on English self-efficacy, and finally on English anxiety, and this path of relationships was found to be significant.In this process, self-directed learning-the mediator variable-had a significant impact on English self-efficacy, another mediator variable in this study.The finding was consistent with a previous study that found that parental involvement negatively affected the self-directed learning ability of middle school students [33].These results can be seen as being similar to the results of the previous studies [13,14], which showed that the higher the level of self-directed learning, the higher the level of academic self-efficacy.
Fourth, the results of the multiple group analysis indicated that gender differences in the paths of the models were significant.In specific, male students' English anxiety and self-directed learning ability were more susceptible to parental pressure expectation than female students, whereas female student's English anxiety was more sensitive to English self-efficacy.These findings can be explained by previous studies [9,10,34], which have shown that male students generally rely on parental pressure more than female students.Gender differences at the level of self-efficacy and English anxiety have been reported, but gender differences in terms of the impact of self-efficacy on English anxiety require further study.
This result supports the implementation of effective support practices for adolescents who learn English in EFL environments similar to Korea by teachers and educators and implies that improving English self-efficacy and self-directed learning skills among teens would help to reduce English anxiety.In addition, specific efforts should be made to develop effective pedagogical strategies and teaching methods to increase students' English self-efficacy and self-directed learning, which will surely help students lower their English anxiety.For example, it may be very helpful for students to experience a sense of achievement in English and improve their English self-efficacy by providing an interesting yet challenging task at the right level.Also, to strengthen students' self-directed learning, educators in practices should seek ways to help students enhance their learning motivation as well as manage their learning behavior.Moreover, there should be a discriminatory approach based on the genders of adolescents.Considering that the negative influence of parental pressure expectation on self-directed learning and English anxiety was higher for boys than girls, parents should be aware of the power of their own words and actions on their children, especially if they have sons.Also, as the positive influence of English self-efficacy on English anxiety was higher for female students than males, improving English self-efficacy will be particularly more effective for girls.
This study examined the structural relationship of variables that could affect adolescents' English anxiety by scrutinizing teenagers, who were feeling insecure about their English, in the context of Korean society where English is not an everyday language, but is important in academic examination and assessment.This study significantly contributes to the research focusing on English education because it revealed that strengthening self-efficacy and self-directed learning of young people lowers English anxiety.However, it was imitated because it did not consider differences in social economic status (SES) and there could be a possibility that the psychological and social attitudes of parents and students towards English may differ depending on their SES.Therefore, it is suggested that a follow-up study is conducted that labels families with different SESs and then looks at parental pressure expectation, students' English self-efficacy, and their level of English anxiety as well as the differences in the pathways between these variables.

Figure 3 .
Figure 3. Final model: Multiple group analysis

Table 1 .
Grade and gender of the participants.

Table 2 .
Differences in study variables by gender.

Table 3 .
Correlations and descriptive statistics for the study variables.

Table 4 .
Fit statistics for tests of invariance.

Table 5 .
Tests of structural invariance.