A neglected area: Elementary school EFL education and teacher self-efficacy

Abstract

Most students begin learning foreign languages in elementary school, and yet little research is conducted in this area. This mixed-methods study investigated teacher self-efficacy (TSE) among 138 classroom elementary teachers in Japan tasked with teaching English as a foreign language (EFL). Policymakers promote EFL in their schooling systems, but with a lack of specialist EFL teachers, many classroom teachers must conduct lessons. This study began by developing a measure of TSE specially designed for elementary school teachers who teach EFL, and then investigating factors that predict TSE. Rasch analysis was used for initial validation, and a regression model was constructed to determine predictors of TSE. Results suggested the questionnaire was unidimensional, and the regression showed that English teaching experience was the strongest predictor of TSE, with study of English outside of formal education also a significant predictor. Interview data with eight participants revealed teachers were confident about teaching the content, but expressed apprehensions about their EFL teaching skills, being a role model for learning, and that certain contextual factors affected their engagement with the subject.

Introduction

Elementary school is where most students take their first steps towards learning a foreign language and may form lasting impressions and attitudes about studying English. Despite this, there is very little research in this context. Teacher influence over students is particularly important at this early stage, and classroom practices of teachers are shaped by their beliefs that they possess the necessary competencies and skills to carry out actions in order to produce certain outcomes (Bandura 1997). Investigating language teacher self-efficacy (LTSE) is important because these beliefs inform aspects of teacher reasoning and behavior in the language classroom. In this paper, TSE is defined as a personal assessment of a teacher’s expected level of performing a specific task involving student learning, while evaluating the resources available and the constraints they face within that particular language teaching context. It should be noted that we are deliberately using the term TSE rather than the more established LTSE to highlight the fact that these teachers are not specialized language teachers.

English as a foreign language (EFL) in the elementary school curricula has become common across Asia (Zein and Butler 2022). With the perceived idea that learning English earlier will improve final proficiency and can lead to sustained socioeconomic development, countries are introducing foreign language education to increasingly younger students (Hayes 2022). However, with a shortage of licensed EFL teachers, more elementary school classroom teachers are required to teach English, not only in Japan, but in other countries such as Finland (Hahl and Pietarila 2021) and the Netherlands (Michel et al., 2021). With an established link between TSE and teachers’ belief that they can positively influence student learning (Caprara et al., 2006; Zee and Koomen 2016; Kim and Seo 2018), it is important to measure the perceived self-efficacy of these elementary school teachers. Furthermore, in the field of second language acquisition (SLA), there continues to be a strong research focus on high school and tertiary contexts, while elementary schools may be overlooked because English is a relatively new subject, or because researchers based in universities struggle to gain access to this context. This study aims to address this gap in the research, by first developing a questionnaire to measure TSE in this context, and then examining factors that predict it. Finally, we analyzed interview data to ascertain teachers’ views on issues related to English teaching in elementary school.

Literature review

Language teacher self-efficacy

Over the past ten to fifteen years, the amount of research on TSE in second and foreign language education has flourished. Researchers have moved beyond earlier general TSE measures (e.g. Tschannen-Moran and Hoy 2001, 2007; Klassen and Tze 2014) to develop more specific instruments designed to better measure language TSE (e.g. Thompson 2020). This improved instrumentation has enabled investigations into TSE in various settings, such as Türkiye (Ortaçtepe and Akyel 2015), Vietnam (Phan and Locke 2015), Korea (Choi and Lee 2016), Iran (Moradkhani and Haghi 2017), and Canada (Karas and Faez 2021). Findings from these studies have recognized that teachers’ self-perceived English proficiency can influence general self-efficacy beliefs towards teaching English (e.g. Choi and Lee 2016; Faez, Karas, and Uchihara 2021). Additionally, reviews of research into language teachers’ self-efficacy (e.g. Houng 2018; Wyatt 2020; Faez, Karas, and Uchihara 2021; Wyatt and Faez 2024) have further enhanced our understanding and advanced the field of research investigating language TSE.

In Japan, there have been several studies that investigated teacher beliefs and TSE with English teachers, although they focused on contexts other than elementary school. Nishino (2009) investigated English high school teachers (n = 139) beliefs towards communicative language teaching (CLT). Using a mixed-methods approach, Nishino concluded that there was a gap between the teachers’ beliefs, their perceptions of how they conducted lessons, and what they were actually doing in the classroom. Praver (2014) considered TSE among 440 Japanese university instructors and found that the L1-English-speaking participants reported higher levels of perceived TSE than the Japanese English teachers. Gender was found to be insignificant, but years of teaching experience, the ability of their students, and working with administration and colleagues were factors that affected TSE. A qualitative study with three private high school English teachers (L1 Japanese) (Glasgow 2014) reported low TSE in their ability to conduct lessons in English and difficulties finding opportunities to improve their English-speaking skills. Those conclusions were corroborated by Omote (2017), who examined the instructional speech among 179 junior and senior high school English teachers. Omote concluded teachers struggled to find a balance of using the target language of English while maintaining a high sense of TSE and linguistic identity.

Thompson and Yanagita (2017) explored how two English high school teachers at one school collaborated to address policy changes requiring teachers to use English as the medium of instruction and implementing CLT. They found that TSE is reflected in prior beliefs about English education and that the challenges of conducting communicative lessons influenced LTSE. Thompson and Woodman (2019) used exploratory factor analysis to examine the LTSE survey results from 141 high school English teachers. The authors found the factors of teaching in English, CLT, teamwork, student achievement, and workload to have influenced TSE. Although these studies have contributed to our understanding of TSE, teacher beliefs towards CLT, and language teachers’ work in Japan, none investigated elementary school teachers.

Elementary school EFL programs in Japan

Compared to other countries in the region, Japan was late in introducing English in its public elementary school system (Butler 2004; Ferguson 2022). Commencing in 2008, under a heading for the promotion of international understanding, once-a-week English lessons were introduced as a non-academic subject into the Grade 5–6 curriculum with the aim of fostering an interest towards learning foreign languages by concentrating on developing basic listening and speaking skills. This was achieved through simple communicative activities and games that focused on learning vocabulary associated with each topic (e.g. food, school, animals) and simple questions/answers, such as: What’s your favorite food? I like sushi and hamburgers. Students were not assessed, and reading and writing were not taught. In efforts to revise English education and recognize the need to develop stronger communicative competency in foreign languages among its citizens, in April 2020, Japan’s Ministry of Education implemented changes to the English curriculum (MEXT 2018). In terms of elementary school foreign language education, the reforms were two-fold. First, the new policy designated English as an academic subject in Grades 5–6 (students aged ten to eleven) and doubled the number of lesson hours per year from thirty-five to seventy. Secondly, it lowered the starting age by shifting those weekly non-academic English lessons (thirty-five lesson hours per year) into Grades 3–4 (students aged eight to nine). A consequence was that the number of teachers required to teach EFL in Japan’s public elementary schools doubled.

In Japan, local school boards, or the schools themselves, decide who teaches English. With a lack of elementary school EFL specialists trained to teach the subject, English lessons might be delivered by the regular classroom teacher, a junior or senior high school English teacher, an assistant language teacher (ALT), typically an L1-speaker of English, or several of these teachers collaboratively working together under the term “team-teaching.” To further complicate the issue, Japan uses a teacher rotation system. This means that an elementary school teacher might teach Grade 2 one academic year, then teach Grade 6 the following year, and possibly even in a different school. This rotation system forces teachers to change yearly and thus they are not required to teach EFL lessons every academic year, and if they do so, it is likely to be at a different level. However, despite the large increase in the number of teachers being required to teach English in elementary school, little is currently known regarding how efficacious elementary school teachers feel about their ability to conduct communicative EFL lessons, and possible factors that influence and shape their self-efficacy beliefs towards teaching English.

TSE in Japanese elementary schools

Prior to the implementation of English as an academic subject in 2020, researchers examined issues of preparedness and confidence among elementary school teachers. Butler (2004) investigated elementary classroom teachers’ perceived English proficiency to conduct lessons in South Korea, Taiwan, and Japan. Butler reported that teachers in all three countries felt their English proficiency was insufficient to meet the new demands of teaching conversational English to young learners. Following that study, Butler (2005) examined teachers’ perspectives towards CLT in elementary schools in South Korean, Japan, and Taiwan. She found these new EFL teachers had differing understandings and views towards using communicative activities and tasks in English lessons. All the teachers expressed the importance of practice and memorization but had concerns about inflexible pattern-practice drills. The teachers also had concerns about teaching English literacy skills, identifying age-appropriate materials and activities, using the L1 more than necessary, and student motivation towards English.

The results of a teacher survey by Fennelly and Luxton (2011) showed that Japanese elementary school teachers (n = 147) felt burdened by the policy which had mandated once-a-week English lessons for Grades 5–6. Teachers lacked confidence to teach these weekly English lessons, suggesting low TSE. Similarly, Machida and Walsh (2015) presented the results of an anxiety survey among thirty-seven elementary school teachers (twenty male, seventeen female) that indicated the teachers were anxious about their English proficiency levels for teaching English and their preparation time. Classroom observations and interviews revealed those teachers had high levels of anxiety when they taught by themselves but were averse to working with L1-English ALTs because of their lack of experience in Japanese schools. As a result, they preferred team-teaching with other Japanese teachers, but seldom used English in their lessons.

Machida (2016) speculated that teacher anxiety would be different among elementary school teachers in two Tokyo school districts because one area started English lessons five years earlier. Machida found years of teaching experience, amount of in-service EFL training experience, and perceived English proficiency levels to be the three key factors influencing teacher anxiety. The results, however, revealed no significant differences between the two school districts. These findings were corroborated in a similar study conducted with elementary school teachers in rural Japan, which revealed that the teachers’ top concerns were their English proficiency, EFL teaching abilities, limited preparation time, and effective use of the teaching materials (Otani and Tsuido 2011). An in-depth investigation of four elementary school teachers reported that they felt underqualified to conduct English lessons because of a lack of formal EFL training, which forced them to rely heavily on the textbook (Harris 2017).

Although studies on perceived proficiency, confidence, and anxiety are important, they only provide general information of teachers’ attitudes towards teaching English and may not necessarily impact actual teaching practice. TSE directly asks how efficacious teachers feel doing specific activities, providing a more detailed understanding of classroom practices. Despite the important new role that elementary school teachers have in foreign language education, there is limited understanding of their TSE beliefs and little research conducted in this context. An important study was conducted by Matsumura (2022) who investigated the self-efficacy beliefs towards teaching English among 304 public elementary school teachers in Japan. Matsumura explored how factors of self-perceived English proficiency, years of experience teaching English, and attitudes towards working with native English-speaking assistant teachers (ALT) all influenced TSE for non-specialized English teachers. He found there was a significant and large effect with the ALT collaboration variable and a small effect for English proficiency towards TSE, but no significant relationship between TSE and years of teaching English. While this study has a large sample size and uses robust statistical analyses, there are a number of issues that should be considered when interpreting the findings. First, the questionnaire was originally designed to assess self-efficacy among junior high school English specialists. Adopting a survey for a completely different group of participants without proper validation can compromise the accuracy and reliability of the findings. Furthermore, the survey items for the teachers’ self-perceived English proficiency asked these teachers to evaluate their efficacy in performing multiple linguistic skills across numerous tasks. This is a measure of perceived English self-efficacy and should not be conflated with proficiency. We aimed to build on the work of Matsumura (2022), while addressing some of these concerns.

Measuring TSE

TSE remains a difficult construct to measure because it is not uniform across contexts, subjects, or the differing types of classroom tasks that make up teacher work (Tschannen-Moran, Hoy, and Hoy 1998; Tschannen-Moran and Hoy 2001, 2007; Klassen and Tze 2014; Zee and Koomen 2016). Furthermore, as Bandura stated, TSE is context specific and is more than a general claim of self-confidence (Bandura 1997; Tschannen-Moran, Hoy, and Hoy 1998); therefore, previous measures of TSE from other contexts, such as those designed for high school English teachers, cannot be effectively applied to elementary school teachers, who teach multiple subjects, have received minimal EFL training (Otani and Tsuido 2011; Machida 2016; Harris 2017), and teach young learners who are just beginning to learn EFL. Moreover, in the field of EFL, the English language is both the content and the intended means of instruction, which can compound the challenges of investigating English language TSE (Houng 2018; Thompson 2020; Wyatt 2020; Faez, Karas, and Uchihara 2021; Wyatt and Faez 2024).

This study followed the theoretical approach posited by Tschannen-Moran and Hoy (2001) and utilized in other studies (Zee et al., 2016; Perera, Calkins, and Part 2019) that TSE is multidimensional. Since this study investigates English TSE among elementary school teachers who are adjusting to the challenges of teaching a new subject, we anticipated that many participants might have several years of classroom experience, but possibly little or no experience teaching English. As a consequence, this study focused on the two teaching dimensions of instructional strategies and student engagement (Tschannen-Moran and Hoy 2001, 2007) during item development, which is described below. Instructional strategies are conceptualized as the perceived ability of the teacher to make changes and use alternative methods for teaching and assessment. Self-efficacy for student engagement refers to the perceived capability to maintain control of student attention and interest during the lesson to promote learning (Tschannen-Moran and Hoy 2001, 2007; Perera, Calkins, and Part, 2019).

A common problem in quantitative research in SLA is the assumption that questionnaires that have been developed are unidimensional and are measuring the intended construct (Sudina 2021). To address this point, we used Rasch analysis to validate the new TSE questionnaire. Rasch analysis offers several benefits over more conventional methods of instrument validation, such as factor analysis (for a detailed description, see Apple and Neff 2012; Bond, Yan, and Heene 2021; Leeming and Harris 2022). Through Rasch, it is possible to assess the unidimensionality of the items (are they measuring a single construct), and also to view how the items interact with respondents in terms of difficulty. Item and person reliability statistics are provided, as well as statistics showing how well the items separate the respondents. Rasch provides information regarding the functioning of the Likert scale, and detailed fit values for individual items, showing how well they fit the Rasch model for measurement. Finally, Rasch converts Likert response data to a scale, suitable for statistical analyses.

An individual’s self-efficacy beliefs are influenced by previous experiences of success and failure (Bandura 1997) and although research suggests that the influence of teaching experience remains inconclusive for language teachers (Karas and Faez 2021), some studies (Praver 2014; Machida 2016) have found this factor contributing to TSE. In terms of gender, none of the aforementioned studies in Japan compared gender except Praver (2014), who found it insignificant among teachers at the tertiary level. Some TSE studies (Klassen and Chiu 2010; Lesha 2017) have included gender and it is considered a factor that is often overlooked in TSE research (Van Eycken, Demanet, and Van Houtte 2024). Furthermore, the ratio of female to male teachers in Japan’s public elementary schools is approximately 6:4 (MEXT 2021); therefore, we decided to include gender as a variable in our analysis. Language proficiency is a concern among elementary school teachers in Japan (Butler 2004, 2005; Fennelly and Luxton 2011; Otani and Tsuido 2011; Machida and Walsh 2015; Machida 2016) and although the relationship between language proficiency and TSE remains unclear (Choi and Lee 2016; Faez, Karas, and Uchihara 2021), we wanted to know if teachers’ willingness to learn English outside the workplace would be important in this context.

Research questions

There were three parts to this sequential explanatory mixed-methods study (Riazi and Candlin 2014). First, it set out to construct and provide validity evidence for an effective and reliable measurement of TSE specifically designed for elementary schools where English is taught as a foreign language. Although this was not the primary aim of the study, it is an important step that is often overlooked by SLA researchers (Sudina 2021). Second, this study sought to identify factors that predict elementary school TSE in teaching EFL. Third, to achieve complementarity (Riazi and Candlin 2014), interviews were conducted with a nested group of eight participants to gain a richer understanding of their TSE beliefs and contextual factors that affect their work in the EFL classroom. These aims are expressed in the following research questions:

1. Does the newly developed TSE measure fit the Rasch model of measurement?

2. To what extent do the independent variables of gender, English teaching experience, and study of English predict TSE for elementary school teachers?

3. What are teachers’ views regarding conducting EFL lessons, and what challenges, if any, have they encountered?

Methodology

Participants

The data presented in this paper were part of a larger study investigating TSE relating to the core subjects taught by elementary school teachers. The first author contacted fifteen public elementary schools in the Kansai area of Japan that he had previous connections with to seek teacher participation. The purpose of the study was explained to the school principal and a package of ten to fifteen questionnaires was provided to each school. A letter explaining the purpose of the research and ensuring all participant information would remain strictly confidential was given with each questionnaire. The principal explained the request to the teachers in a morning staff meeting and asked for teachers to voluntarily complete the questionnaire.

A total of 138 participants (67 female, 54 male, 17 did not report) completed the questionnaire. All participants gave written consent to use their data. Table 1 shows the participants’ years of overall teaching experience and English teaching experience. Out of the 138 participants, 32 reported taking extra English lessons at some point, but only 5 stated they use English outside of work. We classified these extra English lessons as study of English and used this as a dichotomous predictor variable in our subsequent analysis. Fourteen teachers reported having experience living or studying in an English-speaking country, with five teachers reporting experience living or studying in a non-English-speaking country. One teacher reported having experience living in both. An undergraduate degree and teaching license are required to teach in elementary school in Japan, but in addition to this, thirteen participants reported having a Masters’ degree.

Development of questionnaire items

Accurate measurement is central to quantitative research, and yet instrument validation is often neglected (Sudina 2021). A key challenge for researchers is designing items that accurately measure the construct of teachers’ English language self-efficacy (Karas, Uchihara, and Faez 2024) and ensuring that the validity of the construct is maintained across linguistic and cultural differences (Thompson 2020). In order to ensure this aspect of construct validity, two panels of Japanese elementary school teachers assisted with the development of the questionnaire items. In the first cycle of item development, the authors, who have experience developing measures of self-efficacy (Leeming 2017), compiled a list of candidate items from several studies that focused on language TSE (Nishino 2009; Praver 2014; Thompson 2020). From this pool, we selected twenty potential items that we felt reflected the tasks that comprise the job of teaching English in elementary school and translated them into Japanese. This included items relating to lesson preparation and planning, delivering lessons including the teaching of vocabulary and grammar, and also student assessment, which is now required in the final two grades of elementary school. In the second cycle, the items were reviewed by two Japanese panelists, who each had over twenty years of experience teaching at public elementary schools and approximately ten years of experience conducting English lessons. Both of these panelists individually examined the original English and the Japanese translation for each item. Items that were considered redundant or culturally or contextually ambiguous to the elementary school context were revised or deleted. For example, items covering the TSE dimension of classroom management (Tschannen-Moran and Hoy 2001) were viewed to be inconsistent with the questionnaire and were removed. This dimension of classroom management has been found problematic in other Asian contexts (e.g. Hoang and Wyatt 2021). After revisions, eleven Likert scale items were retained to measure efficacy beliefs teaching English, along with an additional ten general demographic questions.

In the third cycle, the items were further reviewed by a different panel comprised of three elementary school teachers and one elementary school principal. This panel individually completed the questionnaire to ensure there was no redundancy or ambiguity, and that each item was asking about only one aspect of teaching. All items were judged to be culturally and linguistically clear and understandable. This panel’s responses were not included in the analysis. We then reviewed the items before finalizing the questionnaire (see Supplementary Appendix A for the English and Supplementary Appendix B for the Japanese questionnaire items). We named the final version the Elementary School English Teacher Self-Efficacy (ESETSE) Questionnaire.

Quantitative data analysis

The questionnaire data were analyzed using the Rasch model for measurement. Winsteps was used for analysis (Linacre 2020a). Several benchmarks were used to judge the results. For Principal Components Analysis of Residuals (PCAR), the measures should account for more than 50 per cent of the variance, and the unexplained variance in the first contrast should be less than 10 per cent or have an eigenvalue lower than 2 (Wright 1996; Raîche 2005; Linacre 2020b). Benchmarks for reliability are similar to those adopted for Cronbach’s alpha (values above .70 indicate acceptable reliability in measurement), and for separation, item separation should be greater than 2, and people separation should be greater than 3 (Linacre 2020b). Wolfe and Smith (2007) stated that for the Likert data, there should be ten observations for each category, that category measures for people should be higher than the level below, that outfit mean square values should be lower than 2.0, and gaps between levels should be greater than .59 but less than 5 logits. Finally, regarding the individual functioning of each item, high infit values indicate problems with measurement and the items themselves, and are therefore of the most interest to researchers. Generally, values of mean square fit between .6 and 1.4, and standardized z-scores between −2 and +2 are considered to be acceptable (Smith 2000; Wolfe and Smith 2007). Fit statistics are sensitive to n-size, with larger n-sizes benefiting mean square statistics and worsening standardized values (Bond, Yan, and Heene 2021). With a relatively small n-size, we decided to use Z-standardized scores for this study.

To answer the second research question concerning potential predictors of TSE, a regression model was constructed with logit scores of TSE derived from the Rasch analysis used as the dependent variable. Based on previous research (Praver 2014; Machida 2016), it was hypothesized that English teaching experience would be a positive predictor of TSE, and therefore this was entered into the model as an independent variable. Though teachers might develop feelings of self-doubt, it is through classroom experiences that teachers are able to derive feelings of teaching efficacy. Furthermore, we hypothesized that teachers who actively sought out opportunities outside of work to study English, such as at conversation schools, would be more efficacious in English class. Studying English at a conversation school is relatively expensive and time consuming, but shows an interest and motivation to improve one’s English proficiency. Higher proficiency should, in turn, lead to higher self-efficacy regarding the teaching of English (Choi and Lee 2016; An, Li, and Wei 2021; Hoang and Wyatt 2021). Therefore, we added experience of study outside of formal education as a dichotomous predictor variable (0= no; 1= yes). Finally, although results about TSE and gender have been mixed (Klassen and Chiu 2010; Praver 2014; Lesha 2017), considering the ratio of female to male teachers in Japan’s elementary schools is 6:4 (MEXT 2021), we were interested in gender in this context and added it to the model as a predictor variable.

Qualitative data analysis

To address our third research question, we spoke to eight teachers who had completed the questionnaire and volunteered to be interviewed. The interviews, which were conducted in Japanese by the first author, were recorded and then transcribed by a paid research assistant. The English translation of the interview questions on teaching English is given in Supplementary Appendix C. The transcriptions were verbatim but did not go to the level of conversation analysis as this was not deemed necessary for our analysis. The transcriptions were verified by the first author. We then began the analysis by individually listening to the recordings and reading the transcripts, identifying and coding excerpts, and taking notes. In this first cycle of coding, we used descriptive codes to examine the social environment of the schools; evaluation codes to identify judgments of worth, merit, and impact; and process coding to understand how teachers were conducting EFL lessons (Miles, Huberman, and Saldana 2014). We then compared our findings and used second cycle coding to identify patterns within the data and condense that into smaller constructs. Three themes emerged from the interview data, which we labeled: (a) apprehensions of being a role model for language learning; (b) concerns with their perceived abilities to teach English; and (c) engagement with the subject. We address these themes in greater detail in the results section.

Results

Quantitative data

The first research question was concerned with the Rasch validation of the ESETSE Questionnaire. Following Apple and Neff (2012), the first analysis conducted was the PCAR to assess dimensionality. The PCAR for the TSE items showed that 75.3 per cent of the variance was explained by measures, and the unexplained variance in the first contrast was 4.7 per cent (eigenvalue 2.1). Although this eigenvalue is a little high (Smith and Miao 1994), these results generally support the claim that the 11 items are forming a unidimensional measure (Linacre 2020b). The next analysis considered the person and item reliability along with separation. Person reliability was .96 with separation at 4.62. Item reliability was .97 with separation of 5.56. These results suggest strong reliability for the items, and the separation indicates that they are effective in showing different levels of TSE in this group of respondents.

The Wright map for the items was then examined (see Figure 1). The first point of note is that the mean for people is lower than the mean for items by more than one logit, showing that the items were generally difficult for these teachers to endorse, suggesting low levels of TSE overall among respondents. The items showed reasonable spread, although some overlap in the middle of the map suggested redundancy of certain items. Individual items and their difficulty will be discussed in the next section. Generally, the Wright map supports the item separation findings that these items were effective in measuring the range of TSE in this group of teachers. It also shows that was a large variance in feelings of TSE among respondents, with a range of 14 logits from the lowest to highest.

The next analysis was concerned with the functioning of the six-point Likert category scale. Supplementary Appendix D shows that the category structure performed reasonably well in satisfying the majority of the criteria outlined above. The only notable weakness in the scale was the lack of use of the “strongly agree” category which failed to meet the suggested minimum of ten responses (Wolfe and Smith 2007). This means these TSE items were generally quite difficult to endorse for this group of teachers.

The final analysis considered the item fit statistics. As stated previously, while outfit values are of some concern, researchers usually focus on high infit values as being indicative of badly performing items (Wright and Linacre 1994). Supplementary Appendix E shows the results for the eleven items that comprised the self-efficacy measure. Based on the standardized fit statistics (ZSTD) scores, a total of five items are misfitting the Rasch model (Items 1, 3, 4, 9, 11). Items 1 and 11 are underfitting, while Items 3, 4, and 9 are overfitting the model. According to Wright and Linacre (1994), items that overfit the model are not necessarily problematic for accuracy of measurement, and therefore we only consider Items 1 and 11 in the discussion section. Overall, the items are showing reasonable fit to the Rasch model.

The second research question asked to what extent the independent variables of gender, English teaching experience, and study of English predict the dependent variable of TSE in a regression model. The information for English teaching experience, experience of outside study, and gender variables can be found in the description of participants above. Descriptive statistics for the self-efficacy variable is provided in Supplementary Appendix F. As there was no a priori hypothesis regarding the predictor variables, the entry method was used (Field 2009). The regression analysis was conducted following the guidelines outlined by Jeon (2015). The background questionnaire measuring years of English teaching experience had a category of “more than 10 years,” and as this could not be scaled, teachers selecting this option were removed from the analysis. Also, a number of participants declined to report their gender, and they were removed, leaving a final sample size of 115 teachers, which was sufficient for the three predictor variables in the model (Field 2009). Inspection of scatterplots showed that the three independent variables were normally distributed with a linear relationship to the dependent variable. Univariate and multivariate data screening revealed no outliers, and Pearson correlation showed that multicollinearity was not a problem with this data set (Field 2009). Table 2 shows the observed correlations between variables.

For the multiple regression model [F(3,111) = 8.38, P < .001], English teaching experience, gender, and outside study significantly accounted for 19 per cent (R²; adjusted R² = .16) of the variance in TSE for the teachers in this study. An R² value of .16 is equivalent to a correlation of approximately .40 and would be classified as a medium effect size (Plonsky and Oswald 2014). Gender (β = .01) was not a significant predictor and did not add to the model. From Table 3, the β values show that English teaching experience (β = .33) was a stronger predictor in the model than study of English outside of formal education (β = .21).

Qualitative data

The purpose of the interviews was to provide an in-depth understanding of the aspects of elementary school teachers’ self-efficacy beliefs towards teaching English and expose any contextual issues in the school environment that might influence those beliefs. The three themes that emerged from the analysis of the interview data were: (a) apprehensions of being a role model for language learning; (b) concerns with their perceived abilities to teach English; and (c) teacher engagement with the subject. The order of presentation does not indicate a level of importance or suggest any ranking of the themes. Table 4 contains information on the interview participants.

All eight teachers questioned their ability to be an effective role model for their students because being a source of input for their students when speaking and demonstrating proper English usage in communicative activities was difficult for them. As one young teacher said about the realities of now having to teach English:

I wasn’t good at English and didn’t understand it at all when I was in high school or university. I was able to become an elementary school teacher even though I couldn’t speak English. And of course, we don’t have an English license because we’re elementary school teachers, so I don’t have any confidence. (Participant 2)

Five teachers directly expressed doubts that their English-speaking ability was sufficient to effectively demonstrate for students. As two teachers revealed:

I’m not confident in my English skills, so I really felt that I couldn’t do well leading the lessons. (Participant 1)

I can understand the English I hear, but I don’t know how to say it myself in English. It’s difficult. It’s hard to speak English in front of the students. I don’t know what to say. (Participate 5)

Four teachers stated that the language being taught was not beyond their understanding; however, three teachers commented that the number of target vocabulary items and phrases was too large, especially in the upper grades. As one Grade 5 teacher said:

The hardest part was towards the end of the year and the students had to use many of the expressions we had learned during the year, such as “I like”, “I’d like to”, and “I want” and so on. There are many ways to express things and it was very difficult to teach them. The students would ask what the differences were, and they’d get confused because there were so many different expressions. (Participant 3)

One teacher succinctly captured all the participants’ feelings towards the required skills of teaching English when she commented:

Really, elementary school English isn’t that difficult, grammatically speaking, but how to teach. How to do interesting activities, how to get the children to learn the language while enjoying themselves. We need this sort of training. (Participant 4)

Another veteran teacher also stated the importance of the skill to plan effective lessons when he stated:

I think that the hurdle will be lowered a little if teachers understand that while it may be better to have good pronunciation and English abilities, a good foreign language class is not only based on the instructor’s language skills, but also on the teacher’s skill to structure the lesson, the purpose of the lesson, and the steps taken to achieve that. (Participant 8)

These excerpts demonstrate that as these teachers gain experience teaching English, they realize they need the knowledge and skills to plan and deliver effective lessons that keep their students interested to learning English. This is an important issue at this foundational stage because elementary school students are as just beginning to learn English and if they become demotivated, it might be difficult for these students to maintain interest and effort in later grades.

Contextual factors of the schools affected teacher engagement with the subject of teaching English. Teachers conducting EFL lessons all stated that preparation time was an issue. Four teachers (Participants 1, 2, 3, and 6) commented that preparing for English class was difficult because they had to find and organize flashcards, prepare worksheets, get the computer set up to show the digital materials, plus get mentally prepared to use English in the classroom. One teacher expressed her frustrations and said:

I don’t teach one subject like a junior high school teacher. I have to teach math, science, Japanese, social studies, PE, and others. Each lesson needs to be prepared. (Participant 6)

The interviews exposed the issue of time management and how this was affecting their engagement with the subject of English. To provide some relief for classroom teachers, specialized English teachers (senka) have been introduced to some schools. These senka teachers only teach English and might be experienced elementary school teachers, or junior high school or high school English teachers (Ferguson 2022).

The limited use of specialized (senka) elementary school EFL teachers combined with the rotation system has unintentionally created an ambiguous situation for schools and teachers regarding who is required to teach English. Participants 2, 6, and 8 stated that teachers enjoyed having a year break from teaching English. However, some teachers may disengage from teaching English, which does not enable them to gain valuable experience and possibly develop self-efficacy beliefs. As one classroom teacher stated:

Out of 160 schools in the city, there are only about 20 senka English teachers. So, when I get transferred to my next school, they might not have that teacher. There’s a possibility that I’ll have to teach English. But now, I don’t even look at the textbook because the English senka teacher is doing it. I just take my sixth-grade students to the English room. I’m there watching while I mark other stuff and I honestly don’t even know what the English curriculum is like. (Participant 7)

At the school where this teacher worked, the English senka teacher was only teaching Grades 5–6. Therefore, even if this teacher did not transfer, because of the rotation system, there was a possibility the following year he could be teaching another grade (1–4) and be required to teach English. Despite this possibility, the presence of the specialist English teacher enabled him to completely disengage with the subject, which is neither ideal for the teacher nor for his future students.

To conclude, the themes of apprehensions of being a role model for language learning English, concerns about overall English teaching abilities, and teacher engagement with the subject were factors that were not measured in the questionnaire. This demonstrates the complexities of measuring TSE and that a context-sensitive instrument within a mixed-methods approach is necessary to investigate English teaching self-efficacy beliefs.

Discussion

In this section, we explain and interpret the findings by integrating the quantitative data with the interview data presented above. The first research question was concerned with the Rasch analysis of the TSE scale developed for the current study. Results showed that the scale was unidimensional, with good reliability for both people and items (Bond, Yan, and Heene 2021). The separation statistics also showed that it was effective in distinguishing between different levels of TSE in this group of respondents. The Wright map revealed that the items were generally difficult to endorse for these teachers. The easiest item was Item 3, that addressed efficacy to teach reading and writing, which at the elementary school level in Japan is grouped together and basically consists of coping the uppercase and lowercase letters of the alphabet, along with copying simple words and phrases from their textbook (MEXT 2018). This result was to be expected since writing and reading only involves forming the letters of the alphabet and sight-reading simple words as students copy from the textbook, which teachers have commented is easy to assess (Ferguson 2022).

The next easiest item related to being able to follow the textbook and teach using it. In Japan, elementary school teachers are required to use government-approved textbooks that their local board of education selects for all subjects (Ferguson 2022), and therefore most teachers were experienced with following a textbook, even though the English texts were new. However, this point of following the textbook also raises concerns about teachers’ classroom pedagogical choices to make adjustments for effective CLT (Butler 2005; Nishino 2009; Ortaçtepe and Akyel 2015; Omote 2017; Machida 2019). This finding highlights earlier research (Butler 2004, 2005; Otani and Tsuido 2011; Harris 2017) that elementary school teachers who did not feel confident teaching English relied heavily on the content of the textbooks for language activities and how much English they are willing to use in the classroom (Machida and Walsh 2015; Ferguson 2022).

At the top of the Wright map, the most difficult item, Item 11, relates to teachers’ belief that their English proficiency is sufficient to effectively teach English. The difficulty in endorsing this item was supported by the interview data where teachers claimed that they could not be effective role models for language learning due to their limited English proficiency. This point corroborates previous research findings that Japanese teachers had low perceptions of the own English language ability (Butler 2004; Fennelly and Luxton 2011; Otani and Tsuido 2011; Glasgow 2014; Machida and Walsh 2015; Machida 2016, 2019; Omote 2017). Research has shown that teachers in East Asia on average report a lower sense of self-efficacy (Vieluf, Kunter, and van de Vijver 2013; An, Li, and Wei 2021) and Japanese teachers and learners of English generally tend to have very low perceptions of their own English ability (Praver 2014; Leeming 2017). Therefore, it is unsurprising that these elementary school teachers, who are untrained and new to teaching EFL, would report low self-efficacy beliefs in their English abilities. In reality, the beginner-level English covered in the elementary school curriculum means that teachers are more than able to teach the required language, as the teachers themselves acknowledged in the interviews. However, as the interviews also revealed, the real issue might be in recognizing the difference between general language proficiency (English in this case) and classroom proficiency (Butler 2004; Nishino 2009; Choi and Lee 2016) with the need to develop EFL teaching skills for CLT (Butler 2005; Nishino 2009; Wyatt 2020; Glasgow 2014; Thompson and Yanagita 2017).

The second most difficult item, Item 5, asks about teachers’ belief that they can correct students’ errors during English class. Correction is challenging for elementary school teachers new to teaching English (Ha and Murray 2023), and the teachers in this study were not confident that they can effectively help students when they stumble. When students are producing English, teachers must be able to first identify the error, and then offer corrective feedback, often almost instantly. For elementary school teachers, doing this in English might be challenging when they are trying to simultaneously manage a classroom of students. Again, this is related more to the skill of teaching English (Butler 2005; Otani and Tsuido 2011; Machida and Walsh 2015; Harris 2017), which was identified as problematic in the interviews. Finally, the Wright map shows the remainder of the items are grouped around the mean for items, which indicates some possible redundancy in measurement indicated by the overlap of items (Items 4, 6, and 10 all have similar difficulty). Future iterations of the scale could remove one or more of these items.

Finally, the item fit statistics were considered. In total, five items showed poor fit to the Rasch model, although only two of these are potentially problematic for measurement (Wright and Linacre 1994). Item 11, the most difficult item to endorse, references the teachers’ English proficiency, and it is possible that this reference to proficiency causes teachers to focus on their perceived English abilities rather than their efficacy in teaching English (Fennelly and Luxton 2011; Machida 2016; Omote 2017). This supports earlier research (Wyatt 2013, 2020; Choi and Lee 2016; Faez, Karas, and Uchihara 2021) that found that because English is both the object of content and the means of instruction, teacher concern remains high. Item 1 has two parts to the question (understanding the Course of Study and being an effective teacher), which may be responsible for the large infit values. Overall, Rasch analysis showed that the TSE questionnaire was a reliable and effective method of measurement for these teachers, although some refinement with the wording of individual items may be beneficial.

Our second research question concerned potential predictors of TSE. Following the Rasch analysis, teachers’ logit scores for TSE produced by Winsteps software were entered into a regression model as the dependent variable, with years of experience teaching English, gender, and study of English outside formal education as the three predictor variables. Overall, the regression model was significant showing that these three predictors accounted for approximately 16 per cent of the variance in outcomes for TSE. First, gender did not have a significant relationship with TSE, which corroborates Praver’s (2014) findings among university instructors in Japan. Studies that specifically measured classroom management (Klassen and Chiu 2010; Lesha 2017) found gender to be significant, but those findings need to be interpreted with caution. The results of this study showed no advantage or disadvantage towards teaching English for either male or female teachers.

Years of experience teaching English had a positive significant relationship with TSE as other studies have found (Praver 2014; Machida 2016), but contradicting the findings of Matsumura (2022). Matsumura questioned the validity of his own findings due to the lack of variance in English teaching experience among his participants. It should be noted that with elementary school teachers in this context, years of experience teaching does not equate to years of experience teaching English. This suggests that directly engaging with the subject by teaching English in classroom settings, elementary school teachers gain the necessary experience to develop a sense of self-efficacy in their ability to teach English effectively. However, Japan’s public school system uses a teacher rotation system, and therefore, a teacher might teach English one year but not the next. Many of the survey participants reported having only a few years of experience teaching English (seventy-three teachers or about 53 per cent had between one and four years of English teaching experience with twenty-seven teachers or 20 per cent reporting none). It was clear from the interviews, that when teachers are not teaching English, or had a specialist teaching in their place, there was a lack of engagement which might have a negative impact on TSE. This suggests that the use of specialized senka teachers for Grades 5–6 might not be effective in the long term, because it will deny many elementary school teachers the opportunity to gain much needed experience. This could create a gap between teachers, leaving many with little or no experience teaching English, resulting in low TSE, and meaning that in future when they are called upon to teach English, they will be ill-equipped to do so. Instead, encouraging elementary school teachers to regularly be involved in teaching English will increase their TSE, leading to positive classroom experiences for students, and in the long term, strengthening English education at the primary school level in Japan. For future research, a longitudinal study could reveal important factors that help teachers develop positive English teaching self-efficacy beliefs over time because TSE is not a fixed predisposition, but a dynamic set of beliefs that can change positively or negatively with experience (Bandura 1997; Tschannen-Moran and Hoy 2001, 2007).

The other variable found to significantly predict teacher TSE was experience studying English outside of the classroom by attending language schools or English conversation classes. Although English is studied in formal education from elementary school through to university, students often have very limited opportunities to use the language outside of class. As a result, students and young adults who are more motivated to learn often attend English conversation classes. These classes have small numbers of students and are typically taught by L1-English speakers, with the focus on students productively using the language that they already know receptively. Attendance at these schools shows the teachers’ interest in improving or maintaining their English proficiency, which can be problematic for language teachers at all levels and contexts (e.g. Glasgow 2014; Ortaçtepe and Akyel 2015; Phan and Locke 2015; Choi and Lee 2016). Therefore, it is not surprising that this variable positively predicts TSE. The implication is that schools, local boards of education, and possibly even the central government would be advised to provide practicing elementary school teachers with opportunities to use English productively outside of their own teaching. This concern of EFL teachers may mean providing access to language schools, or having conversational language teachers come to elementary schools for in-service training. This in-service training could have a positive impact on TSE, if teachers develop confidence in their English abilities. Teachers, especially at the elementary school level, remain an import source of language input for students in EFL classrooms (Zein and Butler 2022), therefore improving speaking proficiency along with TSE would certainly benefit elementary school students, who are just beginning to learn English.

Conclusions and limitations

The main purpose of this mixed-methods study was to investigate TSE in EFL lessons among elementary school teachers in Japan. Our results show that the questionnaire effectively measures this construct; however, Rasch analysis showed that several items need revision to improve their performance. Results of the regression analysis show that teaching experience is key to developing greater TSE. Though more than ten years has passed since the initial research (Butler 2004; Fennelly and Luxton 2011), our findings show that elementary school teachers’ perceptions of their English proficiency remain an issue. Future research needs to examine in greater detail the aspects of proficiency that may cause teachers to lose TSE, and also which units and particular language forms in the national curriculum are difficult for teachers and why. This would provide stakeholders with valuable information for in-service and preservice teacher education programs to better prepare novice elementary school teachers for the challenges of the classroom. This study raises questions regarding the policy of specialist English instructors in elementary schools, as results suggest that teaching English leads to an increase in TSE, and that when specialists take over, regular teachers may tend to disengage from English classes.

Although we believe this study is an important step in understanding teach self-efficacy in this context, it is not without limitations. First, the convenience sampling with schools by the first author means that we cannot assume to have sampled a cross-section of the entire population of elementary school teachers. The willingness by the schools to ask teachers to complete the questionnaire suggests an interest in English language education that may not be representative of all schools in Japan. Second, the scales used to measure overall teaching experience and English teaching experience did not match. This meant that we were unable to use overall teaching experience as a predictor variable in our regression model. A further limitation was the absence of an outcome variable that is predicted by TSE (Al-Hoorie et al., 2021). Teachers self-reported their TSE, but we have no measure of how this impacts their performance as teachers or the potential impact on the proficiency gains made by their students. Unfortunately, due to the sensitive nature of observation and assessment of teacher performance, this was not possible in this context. Also, students are given formative assessment in elementary schools in Japan, and their performance data on standardized tests is not available, so we were unable to examine the relationship between TSE and student learning outcomes. Future studies should attempt to include some level of classroom observation to attempt to relate reports of TSE with teacher classroom behavior, such as the type of language activities used and the amount of English spoken.

In the future, we hope that the ESETSE Questionnaire developed here can be adapted and used by other researchers to study similar contexts. By deepening our knowledge of TSE and factors that predict it, we can help elementary school teachers feel more efficacious when teaching English, potentially leading to more positive outcomes for students and ensuring they receive high-quality language education. Although most research in our field is conducted in universities, elementary schools provide students with their first experience of learning a language, and this is where long-term attitudes to English are formed. We need more research in this context.

NOTES ON CONTRIBUTORS

Peter Ferguson is an associate professor at Kindai University. He has experience teaching in Japan’s elementary, junior, and senior high schools. He has been involved in elementary school EFL education for twenty years and has contributed to textbook development and been an advisor to the Ministry of Education on primary school English education. His research interests include educational linguistics, education language policy, primary school EFL education, and teacher education.

Paul Leeming is a professor at Kindai University, and an adjunct professor at Temple University, Japan. He has taught at various levels in Japan for more than twenty years. His research interests include task-based language teaching, motivation, and group dynamics in the language classroom.

Acknowledgments

The authors would like to express their sincere gratitude to all the participants for their time and cooperation. They also extend their thanks to the four anonymous reviewers for their valuable feedback and insightful comments, which greatly contributed to the improvement of this paper.

References