Skip to content
Teaching and Education Research

An international forum for rigorous research on teaching, learning and teacher education

Research articleOpen access

Meta-analysis of the relationship between teachers’ self-efficacy and attitudes toward inclusive education

Show affiliations and correspondence
  1. aUniversity of Vienna; University of Jyväskylä
  2. bUniversity of Jyväskylä
  3. cNorth-West University; University of Jyväskylä
  4. dUniversity of Vienna; North-West University

DOI: https://doi.org/10.1016/j.tate.2021.103521

Outline

Abstract

This meta-study aims to examine the size of the relationship between teachers' self-efficacy and attitudes toward inclusive education of K-12 students with special educational needs and to identify potential moderators (publication, sample, and research procedure characteristics). We synthesized the research conducted from 1994 to 2018, and 41 studies were included. Bare-bones meta-analysis with random effect model revealed a sample size weighted correlation coefficient between teachers’ self-efficacy and attitudes as r¯ = 0.35 (CI = 0.31-0.39). The between-study variations were not associated with hypothesized publication and sample characteristics. However, the self-efficacy measurement method accounted for a statistically significant strength of association.

Keywords

1. Introduction

Since the importance of the implementation of inclusive education has been reinforced by many international documents (United Nations Educational, Scientific and Cultural Organization [UNESCO], 1994, United Nations Educational, Scientific and Cultural Organization, 2005; United Nations, 2006; United Nations General Assembly, 2015), research about this topic has also increased. The publication of UNESCO’s (1994) Salamanca Statement and Framework for Action on Special Needs paved the way for further reform toward inclusive education. A long term development in understanding inclusive education is that it is nowadays regarded as an effort to developing school systems in which children with disabilities or special educational needs (SENs)dalong with those of any gender, ethnic background, or economic statusdare welcomed in the same local educational organizations (Mitchell, 2005; Savolainen et al., 2012; United Nations Educational, Scientific and Cultural Organization, 2020).

All along this path of development, teachers' attitudes toward inclusive education have been a very popular research topic (see, e.g., Avramidis & Norwich, 2002; Avramidis & Toulia, 2020; de Boer et al., 2011; Lüke & Grosche, 2018), perhaps because negative attitudes have long been considered a major obstacle for inclusion. For instance, searching for the keywords “attitude” and “inclusion” and/or “disabilities” within * Correspondence concerning this article should be addressed to Akie Yada, Faculty of Education and Psychology, Department of Education, University of Jyvaskyla, P.O. Box 35 (Ruusupuisto), 40 014, Jyv€askyl€a, Finland. Email: ado.eika@ gmail.com. * Corresponding author. Department of Education, University of Jyv€askyl€a, Finland. j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / t a t e publications in the ERIC online library in this 10-year period (2010e2020), 7535 hits were found (retrieved on June 10, 2020).

Teachers are key stakeholders in the implementation of inclusive education, and their positive attitudes toward it are of crucial importance in building truly inclusive educational systems (Avramidis & Norwich, 2002; Avramidis & Toulia, 2020; de Boer et al., 2011; Moberg et al., 2019; Schwab, 2018). Moreover, in line with attitudes, teachers' self-efficacy is seen as a crucial element for the successful implementation of a good educational system (e.g., Gibson & Dembo, 1984; Schwab, 2019; Sharma et al., 2012). In a recent systematic literature review, the authors concluded that teachers' attitudes toward inclusive education was one of the main themes studied widely in research on inclusive education (Van Mieghem et al., 2020).

It has been suggested that teachers' negative attitudes toward inclusion or disabilities can be a barrier to implementing an inclusive education (Guralnick, 2008) because teachers' attitudes appear to be associated with their behavior (i.e., inclusive practices) (Ajzen, 2012; Sharma & Jacobs, 2016), in turn, that could influence children's educational experience in inclusive classrooms (van Steen & Wilson, 2020). Further, previous studies have shown that the attitudes are relatively stable over time (Savolainen et al., 2020; Schwab, 2018), which may be difficult to change. However, a recent meta-analytic study by van Steen and Wilson (2020) reported that teachers' attitudes toward inclusive education are highly influenced by an interplay of cultural and demographical factors, and it is necessary to take those factors into account when developing interventions aimed at enhancement of the attitudes.

Several lines of evidence indicate that teachers' selfefficacy is crucial and related to good educational practices (e.g., Gibson & Dembo, 1984; Soodak & Podell, 1993), and high teacher self-efficacy is associated with their positive attitudes toward inclusion (Savolainen et al., 2012, 2020; Yada et al., 2018). More recently, in a substantial number of studies, scholars have reported correlations between teachers' self-efficacy in inclusive practices and their attitudes toward inclusive education in various samples, such as those involving teachers from different countries (Malinen et al., 2013; Savolainen et al., 2012; Sharma et al., 2018; Sharma & Jacobs, 2016; Yada et al., 2018), teachers who were instructing at different levels (Lee et al., 2014; Malak et al., 2018; Munthe & Thuen, 2009), and teachers with different types of certificates (Chan, 2016; Hernandez et al., 2016; Kormos & Nijakowska, 2017).

If teachers' self-efficacy is related to their attitudes, that may open a door to gradually changing attitudes in a positive direction and, thus, improve the possibilities for inclusive education. Previous research findings into the relationship between teachers' self-efficacy and attitudes in inclusive education have been somewhat inconsistent; the authors of some of the studies reported a high correlation among them (Morgan, 2013; Omer, 2015), but other researchers found them as weak or nonsignificant (Chan, 2016; Wood, 2017). Moreover, there has been only a few longitudinal studies conducted (Carew et al., 2018; Kormos & Nijakowska, 2017), but recent research tested the causal direction between attitudes and efficacy with a cross-lagged longitudinal design and suggested that teachers' self-efficacy beliefs affect their attitudes toward inclusion rather than vice-versa (Savolainen et al., 2020).

The aim of this systematic review was to draw a comprehensive picture regarding the relationship between teachers’ self-efficacy and their attitudes toward inclusive education of K-12 students with special educational needs.

1.1. Teachers’ self-efficacy in inclusive practices

The concept of self-efficacy was first articulated by Bandura (1977) as part of his social cognitive theory and popularized in his book called “Self-Efficacy: The Exercise of Control” (Bandura, 1997). In it, he defined self-efficacy as one's belief about his or her capability to conduct a particular course of performance successfully. Additionally, he pointed out that the belief of efficacy influences controlling one's exercise over action and regulating one's cognitive and psychological processes. In the literature, teachers' self-efficacy can broadly be defined as their beliefs in their ability to positively influence students' outcomes (Gibson & Dembo, 1984; Klassen et al., 2011; Ross & Bruce, 2007; Tschannen-Moran and Woolfolk Hoy, 2001).

Based on Bandura’s (1997) argument that self-efficacy beliefs have a context-specific nature, the authors of several studies have investigated teachers' self-efficacy in specific teaching domains, such as math, language, science, and technology (Klassen et al., 2011). Although such studies offer useful insights for teachers to make context-specific decisions on their specific teaching behaviors (Klassen et al., 2011), critics have also argued that the more the domain specificity of self-efficacy research increases, the less the generalizability to other contexts, and, thus, it may lose practical usefulness (Pajares, 1996; Tschannen-Moran et al., 1998). In this vein, a number of authors have explored general teacher self-efficacy, which refers to teachers' self-efficacy related to most teaching situations, such as beliefs regarding suitable instruction, managing the classroom, and developing students' motivation and engagement (Klassen et al., 2011; Tschannen-Moran et al., 1998).

Recently, a considerable amount of research has been published on teachers' self-efficacy in implementing inclusive education, echoing the international debate on inclusion (UNESCO, 1994; United Nations, 2006; United Nations General Assembly, 2015). Teachers' self-efficacy in inclusive practices might be considered a domain-specific construct (Sharma et al., 2012). However, inclusive practices could be applied in every kind of school setting despite the school levels and subjects that are taught (Yada, 2020). For instance, when we compare two scales that are widely used in the research area, the Teachers' Sense of Efficacy Scale (TSES; Tschannen-Moran & Woolfolk Hoy, 2001) developed for measuring teacher self-efficacy in most teaching contexts and the Teacher Efficacy for Inclusive Practices scale (TEIP; Sharma et al., 2012) made for assessing teachers' self-efficacy in inclusive education, both measures similar domains such as self-efficacy in instruction and classroom management.

Thus, teachers’ self-efficacy in inclusive practices can be seen as a construct that combines both domain specificity and generalizability. Several lines of evidence suggest that teachers' behavior in inclusive classrooms is influenced by their self-efficacy. For instance, one study by Gibson and Dembo (1984), which compared highefficacy and low-efficacy teachers, demonstrated that teachers with high teaching efficacy rarely provided critical feedback to students' incorrect answers and were more persistent in students’ failures. Furthermore, a number of scholars have indicated that mainstream education teachers with a high sense of self-efficacy were more willing to utilize new teaching instruction methods and strategies (Ghaith & Yaghi, 1997; Guskey, 1988; Milner, 2002) and less likely to request referral of difficult students to special education (Meijer & Foster, 1988; Soodak & Podell, 1993).

1.2. Relationship between teachers’ self-efficacy and attitudes

In a large and growing body of literature, researchers have linked teachers' self-efficacy and their attitudes toward inclusive education. The term “attitude” can be broadly defined as a person's cognitive and affective evaluation of an attitude object or specific behavior, whether it is favorable or unfavorable (Ajzen, 2012; Bohner & Dickel, 2011; Eagly & Chaiken, 1993). In 1980, Ajzen and Fishbein articulated the theory of reasoned action, which provides an understanding of an individual's attitude and behavior. Based on A. Yada, M. Leskinen, it, Ajzen (2012) formulated his theory of planned behavior, centered around three factors that predict one's behavioral intention and subsequent behavior: (a) attitude toward the behavior; (b) subjective norm; and (c) perceived behavioral control.

The concept of perceived behavioral control is regarded as similar to Bandura’s (1997) self-efficacy, and the three factors are intricately intertwined and influence behavioral intention and actual behavior (Ajzen, 2012). Teachers' attitudes toward inclusive education refer to their beliefs and feelings about including children with diverse educational needs in mainstream classrooms (Avramidis & Norwich, 2002; Forlin et al., 2011). In several previous studies, researchers found that teachers' self-efficacy was positively correlated with their attitudes toward inclusive education (e.g., Lifshitz et al., 2004; MacFarlane & Woolfson, 2013; Malinen et al., 2013; Savolainen et al., 2012; Yada & Savolainen, 2017).

Meanwhile, the authors of other studies have identified the self-efficacyeattitudes relationship as weak or non-significant (Carew et al., 2018; Chan, 2016; Lee et al., 2014), which indicates that teachers’ self-efficacy and attitudes are divergent concepts (Saloviita, 2015). The inconsistent findings of previous studies may be influenced by several potential factors. In a meta-analysis of teachers' attitudes toward inclusion, van Steen and Wilson (2020) considered whether their attitudes are moderated by demographical and cultural factors. These include publication year, teacher type, gender differences, type of disability, cultural differences, and the measures which were used to evaluate teachers' attitudes.

It is found that those factors are intricately intertwined and influenced teachers' attitudes toward inclusion (van Steen & Wilson, 2020). Publication type, publication year, and sampling method are the factors that are often included in meta-analysis studies (Card, 2011). It has been indicated that the studies whose authors found negative results or no statistically significant results were less likely to be accepted for publication (i.e., publication bias) and might be reported only via unpublished dissertations, theses, or conference papers (Card, 2011; Cooper et al., 1997). Empirical studies on the relationship between teachers' self-efficacy and attitudes in inclusive education may not be an exception.

Furthermore, an evaluation of publication year as a moderator in a meta-study may reveal historical trends of the effect sizes over time (Card, 2011). For instance, Qi and Ha (2012) reviewed 75 studies on inclusion in physical education and illustrated that the quantity of studies was increasing with time, which might be echoing the international trends toward inclusion (UNESCO, 1994; United Nations Educational, Scientific and Cultural Organization, 2005; United Nations, 2006; United Nations General Assembly, 2015). In the case of teachers' selfefficacy, Klassen et al. (2011) conducted a literature review and concluded that the number of teacher self-efficacy studies has significantly increased year over year, with increasing diversity in the methods used, specific teaching domains, and the use of international samples.

Concerning the sampling method, whether the authors used convenience sampling or random sampling may have affected the accuracy of the study results (such as estimated standard errors) as well as the generalizability of the findings (Fan et al., 2017). It has been suggested that using random sampling is rare in many fields (Card, 2011), which is also the case for the subject area of this study. In a meta-analysis study by Fan et al. (2017), the authors found that the effect sizes of studies whose authors used non-random sampling were larger than those of studies that used random sampling. Overall, differences in publication type, publication year, and sampling method may have a potential influence on the effect sizes of the relationship between teachers’ self-efficacy and attitudes in inclusive education.

The cultural differences among samples are another likely factor that might correlate with the effect sizes of the selfefficacyattitude relationship. Leading scholars have seen studying inclusive education as well as learning and teaching in a crosscultural context and have found each to be extremely relevant and important (Artiles & Dyson, 2005; Mitchell, 2005; Pajares, 2007). Of course, cross-cultural comparisons are complex, and it is important to ensure that the constructs are measured in the same way (e.g., by ensuring measurement equivalence) in different countries as, for example, culturally valued orientations and response styles among countries might exist (see Vieluf et al., 2013).

However, some scholars have provided empirical evidence that self-efficacy scales are measurement-invariant across countries (Avanzi et al., 2013; Teo & Kam, 2014; Yada et al., 2018). Previous researchers have conclusively shown that teachers' selfefficacy in inclusive practices and their attitudes toward inclusive education vary by country (Moberg et al., 2019; Savolainen et al., 2012; Sharma et al., 2018; Yada et al., 2018). One framework used to illustrate cultural differences in the understanding of self is the classification of cultures into two categories, in that they rely mainly on individualism or collectivism (Markus & Kitayama, 1991). It has been demonstrated that those from individualistic cultures view the self as independent, whereas those from collectivistic cultures see the self as interdependent with contexts (Markus & Kitayama, 1991).

Klassen (2004) analyzed the data from 270 South Asian immigrant and Anglo-Canadian nonimmigrant students and concluded that self-efficacy in collectivistic cultures might be more influenced by relationships with others. Moreover, a systematic literature review found that the cultural dimension of individualism had a significantly positive effect on teachers' attitudes toward inclusion when other cultural dimension factors were taken into account (van Steen & Wilson, 2020). Although caution must be applied against generalizing a specific population (Bandura, 2002; Klassen, 2004), this cultural classification framework may provide useful insights into explaining variation among the effect sizes of correlations between teachers’ self-efficacy and attitudes.

In previous research, scholars suggested that the association between self-efficacy and attitudes in inclusive education might be moderated by the grade level teachers are instructing, as well as their gender. For instance, Klassen and Chiu (2010) found that teachers who taught in elementary schools and kindergartens showed higher self-efficacy in student engagement and classroom management. This view is supported by Tschannen-Moran and Woolfolk Hoy(2007), who wrote that, among the experienced teachers, teaching the youngest children was linked with a higher sense of self-efficacy. Research findings into the effect of grade level on teachers' attitudes toward inclusive education have been somewhat inconsistent.

While the findings from some studies indicate that upper secondary school teachers have significantly more positive attitudes toward integration than those of lower secondary and primary school teachers (e.g., Leyser et al., 1994), findings from other studies suggest that teachers of grades kindergarten through five show more positive attitudes toward students with disabilities than those of high school teachers (e.g., Leahy, 2014). In terms of teachers' gender, the authors of various studies have pointed out contradictory findings, in which some authors indicated there were gender differences in teachers' selfefficacy and attitudes, with higher self-efficacy in classroom management associated with male teachers (Klassen & Chiu, 2010) and more positive attitudes among female teachers (Alghazo & Gaad, 2004).

On the other hand, several lines of evidence suggest there is no significant effect of gender on teachers' self-efficacy or their attitudes toward inclusive education (Desombre et al., 2019; Leyser et al., 1994; Yada & Savolainen, 2017). In summary, evidence regarding the influence of the roles of instructional grade level and gender as determinants of teachers’ attitudes or levels of efficacy A. Yada, M. Leskinen, has been inconsistent and contradictory but will be included as a moderator in this meta-analysis. Finally, several measures were used in previous studies of teachers' levels of self-efficacy and their attitudes, with some scales showing high reliability and validity (e.g., Sharma et al., 2012; Tschannen-Moran & Woolfolk Hoy, 2001), whereas others have found moderate reliability (e.g., Benton-Borghi, 2006; Forlin et al., 2011).

In addition, some of the studies measured teachers’ selfefficacy and attitudes with regard to including all kinds of children (e.g., Avery, 2017; Desombre et al., 2019), while other authors assessed them in relation to including children with special educational needs or specific disabilities, such as children with hearing loss (Eriks-Brophy & Whittingham, 2013), dyslexia (Kormos & Nijakowska, 2017), and emotional and behavioral difficulties (MacFarlane & Woolfson, 2013). Since Bandura (2012) mentioned that the assessment of self-efficacy is influenced by the validity of the measures and by contextual factors, the above characteristics of measurements may likely serve as potential moderators in this study.

1.3. The present study

The purpose of this study was to explore the magnitude of the relationship between teachers’ self-efficacy and attitudes in implementing inclusive education of K-12 students with SENs. Specifically, the aim was to answer the two research questions (RQs): RQ 1. How strong is the overall correlation between teachers' self-efficacy and attitudes in inclusive education? RQ 2. How is the relationship moderated by the following potential moderators? RQ 2a. Publication characteristics: publication type (peerreviewed articles vs. academic dissertations and theses) and year of publication. RQ 2b. Sample characteristics: cultural differences (collectivism vs. individualism), grade level taught, and gender differences.

RQ 2c. Research procedure characteristics: sampling method, measurement of teachers' self-efficacy and attitudes, context and reliability of scales, number of items, and type of Likert scale.

2. Method

2.1. Data collection

Studies investigating the relationship between teachers’ selfefficacy and attitudes in implementing inclusive education were identified through a search of several databases and the perusal of relevant journal webpages. The search process followed the Preferred Reporting Items for Systematic Reviews and MetaAnalyses (PRISMA) statement protocol (Moher et al. & The PRISMA Group, 2009). The literature search criteria and inclusion process are presented in Fig. 1. The databases were consulted for the time span between 1994 and 2018; this period was selected to include studies conducted after the Salamanca Statement on Principles (UNESCO, 1994) was published, which is considered as a primary point to implement inclusive education in policies and practices.

To avoid publication bias, in which only those studies with statistically significant results would be included (Greenwald, 1975), we searched both peer-reviewed and non-reviewed studies as well as unpublished dissertations. First, we searched the online databases (ERIC, Web of Science, PsycArticles, PsycINFO, EBSCOhost Academic Search Elite, Science Direct, Scopus, JSTOR, ProQuest Education and Psychology Journals, and Google Scholar) for publications in English with the following query in the title and abstract: [(teach*) AND (efficacy) AND (inclusi*) AND (attitude*)]. The query enabled us to search the studies that included terms such as teacher(s), teaching, efficacy, self-efficacy, inclusion, inclusive education, inclusive settings, and attitude(s).

Second, online dissertations and theses were searched through the databases (ERIC, ProQuest Dissertations & Theses, and Google Scholar) using the same query. Third, we searched the web pages of journals related to inclusive education, educational and psychological research, and comparative education for relevant studies (e.g., American Educational Research Journal, Comparative Education, International Journal of Inclusive Education, Journal of Educational Psychology, Teaching and Teacher Education, and The Journal of Special Education). The search procedure yielded 855 potential studies; these were scanned for duplicates, which resulted in the removal of 326 studies. Finally, 529 were identified and reviewed to determine whether they met the inclusion criteria for further analysis.

2.2. Inclusion and exclusion criteria

Both an initial screening of the studies' titles and abstracts and a full-text screening were conducted using a computer software package called Covidence (Veritas Health Innovation, 2019). Two people (one of the authors and a student pursuing a master's degree in education) separately reviewed the studies in both screening processes, and conflicts between the two raters were carefully discussed to achieve agreement. Inter-rater reliability was 86.5% in the initial screening and 74.0% in the full-text screening. The relatively low inter-rater reliability in the full-text screening was due to uncertainty regarding the definition of inclusive education in the criteria, and, thus, two raters carefully discussed the conflicts and agreed on inclusion/exclusion.

We used the following inclusionary and exclusionary criteria for the initial screening: (a) The study must report original empirical data; (b) The study must be written in English; (c) The study must assess the relationship between teachers' self-efficacy and attitudes in inclusive settings, which includes the broad meaning of inclusive education (e.g., attitudes toward disabilities, students with special educational needs, and students with behavioral problems); and (d) Participants of the study must be teachers working in preschools, primary schools, or secondary schools (K-12). (e) Case studies and qualitative studies were excluded. According to the inclusion and exclusion criteria, 212 studies were selected in this first round.

The full text of these studies was retrieved from online databases (or the authors were contacted with a request to provide them). Full-text versions were not available for five studies; thus, 207 were uploaded into Covidence and read by the two raters. The inclusionary and exclusionary criteria for the full-text screening were that: (f) Participants in the study must be in-service teachers working in pre-schools, primary schools, or secondary schools (K-12); (g) The concepts (both teachers' self-efficacy and attitudes) must be measured quantitatively (e.g., using scales); (h) The study must assess the relationship between teachers' self-efficacy and attitudes in inclusive settings using Pearson's or Spearman's correlation coefficient (r or rho); and (i) The study must report a sample size or degree of freedom for Pearson's or Spearman's correlation coefficient (r or rho).

If criteria (f) and (g) were satisfied but the study did not report a A. Yada, M. Leskinen, correlation coefficient, the study was tentatively included in a further coding process to ask the authors to provide the necessary information. This second round of selection resulted in 96 studies for coding.

2.3. Coding procedures

Coding was done by the first author of the article; the process included two phases. In the first phase, pilot coding was conducted with eight studies. The content of the tentative coding table was discussed with all the authors to include as many study features as possible, which was considered relevant to further analysis. After the pilot coding was finished, the research team discussed the issues raised during the pilot coding and made necessary revisions to the coding table. In the second phase, all the included studies were coded using the revised coding table, and the issues encountered in the coding process were again discussed with the research team. The study features included in the final coding table were: (a) publication year (1994e2018); (b) publication type; (c) sampling method; (d) number of participants; (e) country; (f) grade level taught; (g) rate of female participants; (h) the type of measures used for evaluating teachers' self-efficacy and attitudes; and (i) reliability of the used scales.

In some studies, measures were coded so that a negative correlation coefficient indicated a positive relationship between teachers' self-efficacy and attitudes. Thus, in those cases, correlation coefficients were recoded so that a positive correlation always meant that higher self-efficacy was associated with more positive attitudes. Also, when a correlation coefficient was not reported, we contacted the authors and kindly asked them to provide it through an email or social networking service (e.g., ResearchGate or LinkedIn) if an email address was not available. We contacted the authors a second time if they did not respond to the message. Altogether 23 authors were sent requests and 14 responded (response rate ¼ 60.8%).

Of those who responded, three were not able to provide the information for some reasons (e.g., old data set). The studies that the authors’ contact information was not available, authors did not reply, or authors were not able to provide necessary information were excluded from the analysis. It should be noted that some primary studies measured teachers' attitudes and concerns separately and reported each correlation coefficient between teachers' self-efficacy (Chan, 2016; Sokal & Sharma, 2013). In such cases, we included only the correlation coefficient using the attitude scale because the attitudes scales closely reflected our aforementioned definition of attitude. Furthermore, Fig. 1. Flow diagram for the literature search and inclusion process.

A. Yada, M. Leskinen, the authors of several primary studies used the Scale of Teachers’ Attitudes Toward Inclusive Classrooms (STATIC [Cochran, 1997]) or the Opinions Relative to Integration of Students with Disabilities scale (ORI [Antonak & Larrivee, 1995]), which contains a sub-scale that measures concepts similar to self-efficacy (i.e., “professional issues” and “perceived ability to teach students with disabilities,” respectively). Because these sub-scales would have caused a confounding bias, we asked authors to provide information without the sub-scales, and the studies were included only if the authors provided the information.

Studies with a small sample size (less than 30) were also excluded to assure the reliability of the scale and the correlation coefficient (Samuels, 2015). Among the 96 studies selected as relevant to this one, 41 were finally identified as having the necessary information, and thus, they were included in our meta-analysis. The authors of these 41 studies reported findings according to 43 independent samples, which resulted in an accumulated total number of 10 929 participants (M ¼ 254.16). The authors of two studies reported results from two independent samples from different countries. Moreover, three studies adopted longitudinal design. Since this study is not focusing on the intervention effect, correlation coefficients of the first time point were included in the meta-analysis.

The detailed information included in the final coding table for all the studies in this meta-analysis is presented in Table 1.

2.4. Moderators

Three types of moderators were coded in our meta-regression analysis: publication, sample, and procedural characteristics. The publication characteristics included publication type (journal article vs. thesis) and year (1994e2018). The sample characteristics consisted of cultural differences (collectivism vs. individualism), grade level taught (primary vs. secondary vs. mixed), and percentage of female participants. The cultural differences moderator was a dichotomous factor, where studies were coded as individualism or collectivism using Hofstede's country comparison tool on his website (hofstede-insights.com) (Hofstede et al., 2010). Based on his cultural dimensions theory (Hofstede, 2001), the indicator enables us to divide different countries into individualism/collectivism cultures.

Finally, the procedural characteristics contained the sampling method (random vs. non-random), measurement of teachers' self-efficacy (the TEIP scale vs. the TSES vs. the Teacher Efficacy Scale [TES; Gibson & Dembo, 1984] vs. other), the context of teachers' self-efficacy scale (self-efficacy with including all students vs. students with SEN vs. students with specific disability), the context of attitude scale (attitudes toward inclusion of students with SENs vs. students with specific disability), reliability of scales, number of items, and type of Likert scale.

2.5. Calculating effect sizes

The effect size measure adopted in this study was the Pearson's correlation coefficient (r) between teachers' self-efficacy for inclusive practices and their attitudes toward inclusive education. In addition, the authors of three studies reported Spearman's correlation coefficient (rho), which was also included in the analysis. The majority of the effect sizes were correlation coefficients reported in the original studies or provided by the author of the studies. Some authors did not provide an overall correlation coefficient between teachers' self-efficacy and attitudes but reported several correlations among sub-scales. In this situation, the mean of correlation coefficients was calculated, in which each effect size was first transformed into the Fisher Z value and the averaged Z value was transformed back into Pearson's correlation coefficient.

2.6. Meta-analytic integration

The estimation methods were the bare-bones meta-analysis with a random-effects model, meta-regressions with the restricted maximum likelihood (REML), and the Hartung-Knapp (van Aert & Jackson, 2019) method. Using the Comprehensive Meta-Analysis (CMA, Version 3) software package (Borenstein et al., 2013), sample sizes were used as weights (i.e., inverse of variances). The funnel plot asymmetry was elaborated first with a classic fail-safe N method (Rosenthal, 1979) to identify the possibility for publication biases (alpha 0.05, two-tailed p-value). We did not use Orwin's failsafe N, because setting values for trivial correlation and speculating the level of correlation in missing studies would have been products of arbitrary guesses.

Furthermore, Duval and Tweedie's fill and trim method did not suggest trimmed studies to the left or right of the mean for the random-effects model (Shi & Lin, 2019). Therefore, only observed studies were included in the funnel plot. Meta-regressions were done separately for each of the bivariate, dummy-coded moderator variables to explain the heterogeneity of the correlation coefficients (reference category 0 represents other combined categories; 1 represents an interest category, and; intercept was included in the models). We obtained statistically significant estimates (referred to next as b-coefficients) to indicate whether the interest category held higher explanation power over other categories (positive sign) or lower explanation power (negative sign).

We omitted analyzing the significance of moderation factors by the multiple meta-regression method due to the relatively low number of studies. Also, we were not able to take into account the possible interaction effects of the moderators. To illustrate heterogeneity in the models, we chose to report variances among the studies (t2, REML estimate). We also chose to report the percentages of variability not associated with sampling errors (I2) to illustrate heterogeneity in the models.

3. Results

3.1. Analysis of effect sizes

The search for the literature resulted in 855 papers, and we narrowed them down to 529 candidate studies by removing duplicates. Further title and abstract screening yielded 212 potentially relevant studies. After the reading and coding of the full text, we had 41 individual studies consisting of 43 unique samples. The major findings, including meta-analysis based on effect size and study-effect meta-analysis, are presented in Table 2. The overall correlation between teachers' self-efficacy and attitudes in inclusive education was r ¼ 0.351 (k ¼ 43, p < .001), with a 95% confidence interval ranging from 0.307 to 0.394. The results indicated that although there was inconsistency among the individual studies, a moderate correlation existed between teachers’ selfefficacy and attitudes in inclusive education.

The funnel plot graph provided in Fig. 2 shows where all the correlation coefficients were plotted relative to their standardized errors. The symmetrical distribution of plots around the vertical line indicates that there is no publication bias (Card, 2011). Our data showed symmetrical patterns; thus, publication bias might not exist. In addition, the classic fail-safe N analysis was conducted to check for robustness. The results indicated that 3164 studies with null results were needed for the effect to be nullified. Overall, these results indicated that the effect of publication bias on effect sizes would be minimal and negligible. Fig. 3 presents a forest plot graph, which graphically describes the results from the random-effects modeling analysis of the 43 effect sizes.

Here, each square dot indicates the effect size of each sample, and the horizontal line indicates the 95% confidence A. Yada, M. Leskinen, Table 1 Description of Studies Included in the Meta-Analysis. Author (years) Pub type Location n Female % TSE measure Reliability TSE Attitude measure Reliability attitude r Alaverdyan (2018) TD The Republic of Armenia 187 95.6 Teacher Efficacy for Inclusive Practices scale (TEIP) .9 Sentiments, Attitudes, and Concerns about Inclusive Education Revised (SACIE-R) .77 .271 Angiulo (2014) TD USA 154 77 Teachers' Sense of Efficacy Scale (TSES) e Autism Attitude Scale for Teachers-Revised (AAST-R) e .51 Avery (2017) TD USA 44 e TSES .929 Stages of Concerns Questionnaire (SoCQ) e .361 Bender et al. (1995) JA USA 127 92.1 Teacher Efficacy Scale (TES) e Mainstreaming attitudes e .36 and .23 Benton-Borghi (2006) TD USA 227 74 TSES .92 Attitude toward inclusion .78 .273 Bettineski (2015) TD USA 90 e TSES e Attitudes Toward Teaching All Students (ATTAS-mm) e .027 Carew et al. (2018) JA Kenya 123 46.1 and 43.3 Teaching self-efficacy .72 Beliefs and feelings about inclusive education .63 and .63 .07 and .26 Chan (2016) TD Hong Kong 309 94.2 General Self-Efficacy Scale (GSE) .91 SACIE-R .65 -.06 Davis (2015) TD USA 65 85 TSES e Scale of Teachers' Attitude Toward Inclusive Classrooms (STATIC) e .303 Desombre et al. (2019) JA France 554 79.6 TES (only used personal efficacy subscale) .74 Multidimensional Attitudes Toward Inclusive Education Scale (MATIES) .88 .43 Eriks-Brophy and Whittingham (2013) JA Canada 63 73 Teacher confidence in teaching children with hearing loss .78 Teacher attitudes toward inclusion of children with hearing loss .66 .3981 Frizzell (2018) TD USA 180 e Perceptions of teacher self-efficacy in meeting the needs of students with disabilities e Teacher attitudes toward inclusion e .608 Glaubman and Lifshitz (2001) JA Israel 136 50 TES .6 The second part of the regular education initiative questionnaire .83 .29 Hernandez et al. (2016) JA USA 118 95.7 TSES .94 STATIC .8 .33 Kormos and Nijakowska (2017) JA e 1187 92.8 Self-efficacy beliefs relating to the successful application of inclusive second language teaching approaches (4 items adapted from TEIP) .82 Attitudes to inclusive second language education and concerns about inclusion and the successful application of inclusive teaching practices (8 items adapted from SACIE-R) .665 and .711 .208 Kuittinen (2017) TD Chile 108 74.1 TEIP .96 SACIE-R .72 .393 Lee et al. (2014) JA Hong Kong 275 100 Efficacy .67 Resistance to inclusion and Endorsement of SEN .79 .23 Lifshitz et al. (2004) JA Israel and Palestine 66 and 95 and 100 Gemmel-Crosby & Hanzlik (1994) Questionnaire .8 Gemmel-Crosby & Hanzlik (1994) Questionnaire .87 .45 MacFarlane and Woolfson (2013) JA Scotland 111 94.6 TSES modified for children with social, emotional, and behavioral difficulties (SEBD) .93 MATIES modified for children with SEBD .75 .44 and .44 Malak et al. (2018) JA Bangladesh 1090 42.3 Teachers' Efficacy in Classroom Management and Discipline (TECMD) scale .77 Teachers' Attitudes Toward Students' Inappropriate Behavior (TASIB) .92 .422 Malinen et al. (2012) JA China 451 85.4 TEIP .91 SACIE-R .69 .41 Mian (2017) TD Singapore 183 72.7 TEIP .927 SACIE-R .701 .403 Japan 189 48.1 TEIP .915 SACIE-R .707 .307 Montgomery and Mirenda (2014) JA Canada 100 87 TEIP .65e.77 SACIE-R .68e.82 .215e.561 Morgan (2013) TD USA 142 e Self-efficacy scale .92 Attitudes towards including a child with autism spectrum disorder .92 .59 Munsell (2015) TD USA 370 82.4 TSES .9 MATIES .78 and .79 .311 and .446 Munthe and Thuen (2009) JA Norway 74 63.5 Teachers' efficacy about pupil learning and behavior .60 and .78 The agreement of inclusion .78 .100 and .256 Omer (2015) TD Ethiopia 76 e TES e Teachers' attitudes toward inclusive education e .638 Randoll (2008) TD Canada 34 94.1 Teacher Efficacy Scale: Long Form e Attitudes Toward Inclusive Education Scale (ATIES) e .399 Samms (2017) TD Jamaica 191 91.15 TEIP .879 SACIE-R .702 .197e.234 Savolainen et al. (2012) JA Finland 855 78.3 TEIP .88 SACIE-R .74 .42 South Africa 322 82.1 TEIP .91 SACIE-R .66 .194 (continued on next page) A.

Yada, M. Leskinen, interval of the effect size. The correlation coefficients of unique samples varied between 0.06 and 0.64. There was large variation across the correlation coefficients, and the result of Q tests for heterogeneity among the effect sizes indicated substantive heterogeneity (Q [42] ¼ 254.77, p < .001) justifying the random-effect approach.

3.2. Analysis of moderator variables

According to the overall heterogeneity test, the effect sizes were shown to be considerably inconsistent among the unique samples, indicating the existence of potential moderators that may have contributed to these inconsistencies. As mentioned above, three types of moderators were included in the metaregression analysis: publication, sample, and procedural characteristics. First, the roles of publication type (journal article vs. thesis) and year (1994e2018) were examined. The results indicated that publication type and year of publication were not statistically significant moderators of the teachers’ self-efficacyeattitudes relationship. The sample characteristics included cultural differences (collectivism vs. individualism), grade level taught (primary vs. secondary vs. mixed), and percentage of female participants.

The included samples were categorized either as representing a collectivistic or individualistic culture based on Hofstede's individualism/collectivism indicator (Hofstede et al., 2010). The mean effect size of the individualistic culture was larger than that of the collectivistic culture, although the result was not statistically significant. Moreover, no statistically significant differences were found in the regression analyses of grade level taught and gender (rate of female participants). The last type of moderator was that of procedural characteristics. The meta-regression analysis revealed that the sampling method (random vs. non-random) was not a significant moderator. The measurement of teachers' self-efficacy (the TEIP scale vs. the TSES vs. the TES vs. other) was also not a significant moderator associated with the variations among studies.

However, the context of the teachers' self-efficacy scale (self-efficacy in including all students vs. students with SENs vs. students with specific disabilities) predicted the relationship between teachers' self-efficacy and attitudes. More specifically, the correlations were lower for the studies using teachers' self-efficacy measures in the context of including all students (b ¼ .111, p < .05). Furthermore, the correlations were higher when the studies adopted teachers' self-efficacy measures in the context of including students with specific disabilities (b ¼ 0.135, p < .05). The context of attitude scale (attitudes toward inclusion of students with SENs vs. students with specific disabilities) moderator showed a similar tendency, indicating that effect sizes for attitudes toward the inclusion of students with specific disabilities were larger than that of students with SENs (b ¼ 0.124, p ¼ .059), though the significance level was slightly above the alpha level of 0.05.

In terms of the reliability of scales, higher reliability of the attitude scale was associated with larger effect sizes (b ¼ 0.973, p < .001), whereas such a relationship was not found for the reliability of the teachers' self-efficacy scale. Finally, the number of items and the type of Likert scale for both teachers’ self-efficacy and the attitude scales were not significant moderators.

4. Discussion

Over the last decade, the authors of several studies have shown that there is a positive relationship between teachers' self-efficacy and their attitudes toward inclusive education (e.g., Savolainen Table 1 (continued ) Author (years) Pub type Location n Female % TSE measure Reliability TSE Attitude measure Reliability attitude r Skuller (2011) TD USA 267 84.3 Teacher Efficacy Scale .696 and .759 The scale developed based on the Autism Attitude Scale for Teachers (AAST) e .146e.499 Sokal and Sharma (2013) JA Canada 131 87 TEIP .91 School principals' attitudes toward inclusion .86 .441 Stanovic (1996) TD Canada 33 78.8 TES .82 Attitude Toward Mainstreaming Scale (ATMS) .93 .43 Tasnuba and Tsokova (2015) JA Bangladesh 400 e TEIP .89 2 sub-scales from SACIE-R .60 and .64 .25 Tsakiridou and Polyzopoulou (2014) JA Greece 416 68.3 The self-efficacy belief scale .864 Attitudes toward inclusive education scale .85 .156 Urton et al. (2014) JA Germany 276 96 Wirksamkeit Lehrer .86 Attitudes toward inclusion .85 .51 Wang et al. (2012) JA China 323 e TEIP e Concerns About Inclusive Education Scale (CIES) e .32 Weisel and Dror (2006) JA Israel 139 100 TES .77 and .86 Teachers' attitudes toward Inclusion questionnaire .97 .67 and .52 Wilson et al. (2016) JA Scotland 145 85 Theory of Planned Behavior (TPB) questionnaires .89 TPB questionnaires .96 .392 Wood (2017) TD USA 47 87.2 TEIP .874 Attitudes Toward Teaching All Students (ATTAS) .771 .12 Yada and Savolainen (2017) JA Japan 359 53.5 TEIP .93 SACIE-R .75 .396 Note. 1 ¼ Spearman's correlation coefficient was reported; TSE ¼ teachers' self-efficacy; TD ¼ thesis and dissertation; JA ¼ journal article.

A. Yada, M. Leskinen, Table 2 Relationship between Teachers’ Self-Efficacy and Attitudes in Inclusive Education: Overall Results and Meta-Regression Analyses. k Estimate SE t-value df p 95% CI t2 I2 Null model Intercept 43 .3671 .026 13.95 42 <.000 [.3141, .4201] .023 83.51 Publication moderators Publication year Intercept 43 5.359 10.169 0.53 41 .601 [-15.178, 25.896] .023 83.51 Year 43 -.003 .005 0.49 41 .626 [-.013, .008] .023 83.80 Publication type Intercept 43 .368 .042 8.808 41 .000 [.283, .452] .023 83.51 Article 43 -.001 .054 0.010 41 .992 [-.110, .109] .024 83.88 Sample moderators Cultural differences Intercept 38 .315 .043 7.376 36 .000 [.229, .402] .023 82.11 Individual culture 38 .083 .056 1.494 36 .144 [-.030, .196] .021 80.94 Grade level Intercept 38 .366 .032 11.565 36 .000 [.302, .430] .022 83.35 Primary 38 -.031 .064 0.481 36 .634 [-.162, .100] .022 83.76 Rate of females Intercept 36 .280 .106 2.656 34 .012 [.066, .495] .017 81.89 Females 36 .001 .001 0.709 34 .483 [-.002, .004] .017 82.01 Procedural moderators Sampling method Intercept 43 .368 .029 12.555 41 .000 [.309, .427] .023 83.51 Random 43 -.002 .071 0.026 41 .979 [-.145, .141] .024 82.86 TSE measurement 1 Intercept 43 .380 .033 11.587 41 .000 [.313, .446] .023 83.51 TEIP 43 -.036 .056 0.638 41 .527 [-.149, .077] .023 83.90 TSE measurement 2 Intercept 43 .371 .029 12.656 41 .000 [.311, .430] .024 83.90 TSES 43 -.019 .071 0.263 41 .794 [-.161, .124] .021 83.51 TSE measurement 3 Intercept 43 .349 .028 12.681 41 .000 [.294, .405] .023 83.51 TES 43 .140 .077 1.813 41 .077 [-.016, .296] .022 82.88 Context of TSE scale 1 Intercept 40 .451 .046 9.778 38 .000 [.358, .544] .018 78.39 All students 40 -.111 .054 2.045 38 .048 [-.221, .001] .017 78.94 Context of TSE scale 2 Intercept 40 .367 .027 13.700 38 .000 [.313, .422] .018 78.39 Students with SEN 40 .035 .082 0.423 38 .675 [-.131, .200] .018 78.68 Context of TSE scale 3 Intercept 40 .483 .059 8.134 38 .000 [.294, .402] .018 78.39 Students with a specific disability 40 .135 .065 2.066 38 .046 [.003, .266] .017 78.83 Context of TAT scale 1 (students with SENs) Intercept 42 .464 .056 8.239 40 .000 [.350, .578] .023 83.91 Students with SENs 42 -.124 .064 1.946 40 .059 [-.252, .005] .022 84.22 Context of TAT scale 2 Intercept 42 .340 .030 11.533 40 .000 [.281, .400] .023 83.91 Students with a specific disability 42 .124 .064 1.946 40 .059 [-.005, .252] .022 84.22 Number of items Intercept 42 .345 .063 5.513 40 .000 [.218, .471] .024 83.89 TSE 42 .002 .004 0.384 40 .703 [-.006, .009] .024 83.01 Number of items Intercept 42 .291 .067 4.349 40 .000 [.156, .426] .024 83.89 TAT 42 .006 .005 1.260 40 .215 [-.004, .015] .023 83.32 Type of Likert scale Intercept 42 .346 .110 3.142 40 .003 [.123, .568] .023 83.89 TSE 42 .004 .017 0.217 40 .829 [-.030, .037] .024 84.13 Type of Likert scale Intercept 42 .319 .108 2.966 40 .005 [.102, .536] .021 83.27 TAT 42 .008 .020 0.390 40 .699 [-.033, .049] .022 82.86 Reliability of TSE scale Intercept 35 .240 .244 0.982 33 .334 [-.257, .737] .020 83.85 TSE 35 .136 .288 0.472 33 .640 [-.449, .721] .019 83.37 Reliability of TAT scale Intercept 33 -.402 .190 2.120 31 .042 [-.790, .015] .020 84.34 TAT 33 .973 .243 4.011 31 <.001 [.478, 1.467] .012 76.59 Note. 1 ¼ Fisher's Z coefficient; k ¼ number of samples; Estimate: REML meta-regression estimate with Knapp-Hartung method; SE: standard error; df ¼ degree of freedom; CI: confidence interval; TSE ¼ teachers' self-efficacy; TEIP ¼ Teacher Efficacy for Inclusive Practices scale (Sharma et al., 2012); TSES ¼ Teachers' Sense of Efficacy Scale (Tschannen-Moran & Woolfolk Hoy, 2001); TES ¼ Teacher Efficacy Scale (Gibson & Dembo, 1984); TAT ¼ teachers' attitudes.

A. Yada, M. Leskinen, et al., 2012; Sokal & Sharma, 2013; Weisel & Dror, 2006; Yada & Savolainen, 2017). A recent systematic literature review by van Steen and Wilson (2020) concluded that teachers' attitudes toward inclusion were positive and moderated by an interplay of several factors. However, there has been no study in which a research team quantitatively synthesized the results of the relationship between teachers' self-efficacy and their attitudes toward inclusive education. Thus, the primary aim of this study was to explore the overall relationship between teachers' self-efficacy and their attitudes toward inclusive education.

Many scholars have claimed that improving efficacy might be one way to improve teachers’ attitudes, and, while the results from several studies have suggested, those of this study provideddfor the first timedmore conclusive evidence of the relevance of this hypothesis. Our statistical estimation did not suggest a publication bias, and this is in line with recent findings by van Steen and Wilson (2020) in their meta-study. Further, the analysis revealed an average significant positive relationship (r ¼ 0.351) between teachers' selfefficacy and attitudes, which can be considered a moderate effect size (Cohen, 1988). Additionally, the results indicated that the correlations among single studies varied widely, which highlights the importance of using the results of meta-studies when summarizing the body of literature.

The result is consistent with the previous meta-analyses that have positive links with teacher efficacy. For example, teachers’ self-efficacy has been shown to positively influence their evaluated teaching performance (Klassen & Tze, 2014) and their commitment to the teaching profession (Chesnut & Burley, 2015); it has been negatively associated with their burnout (Aloe et al., 2014). Also, with this study, we aimed to examine whether different moderators influenced the effect size of the teachers' self-efficacyeattitudes relationship. We found that neither publication characteristics (i.e., publication type and year) nor sample characteristics (i.e., cultural differences, grade level taught, and rate of female participants) were significant predictors of the relationship between teachers' self-efficacy and attitudes in inclusive education.

These findings suggest that the relationship between teachers' selfefficacy and their attitudes is quite universal, regardless of the time, culture, grade level taught, or gender, even though the magnitude of teachers' attitudes and self-efficacy might differ based on the factors (Klassen et al., 2011; van Steen & Wilson, 2020). This finding is in line with earlier studies indicating that teachers’ self-efficacy and attitudes have a positive relationship in different countries (Avramidis et al., 2019; San Martin et al., 2021; Sharma & Jacobs, 2016; Yada et al., 2018). In light of procedural characteristics, the meta-regression analysis indicated that the sampling method did not influence the increase or decrease of effect sizes.

Though there is a common understanding that studies with random sampling gather representative data, the effect size may not differ in the case of the relationship between teachers' self-efficacy and attitudes. The efforts involved in randomization are complex and might impact on the mean of variables or constructs but not on the associations among different constructs. However, this result should be interpreted with caution because, in this study, only 7 out of 43 samples were collected with random sampling. Since sampling methods were not used as a moderator variable in previous meta-analysis studies in teachers' self-efficacy research, future work is required to ensure whether the teachers’ self-efficacyeattitudes relationship is influenced by sampling methods.

Different measurements of teachers' self-efficacy did not explain a significant amount of variance in the relationship between their self-efficacy and attitudes toward inclusive education. A possible explanation for this might be that although there are some teachers' self-efficacy scales that are often used in research, Fig. 2. A Funnel Plot Graph of the Correlation Coefficients in the Unique Samples (N ¼ 43). Fig. 3. Overall Random-Effects Model Meta-Analysis Average Correlation Coefficient of Teachers’ Self-Efficacy-Attitudes, Correlation Coefficients in the Unique Samples (N ¼ 43), and Coefficients with a 95% Confidence Interval for Each Study. A. Yada, M. Leskinen, these scales measure similar constructs of self-efficacy in inclusive education.

However, contexts of teachers' self-efficacy scales accounted for the variation among the correlations, in which selfefficacy in including all students predicted lower correlations than scales including students with SENs and specific disabilities. In other words, if the context of the scale is more specific to students with particular disabilities, correlations between teachers' selfefficacy and attitudes become stronger. A similar tendency was found for the contexts of the attitudes scale, although the significance level was slightly above .05. These tendencies can be explained in part by the context-specific nature of self-efficacy (Bandura, 1997). If the scale is more context-sensitive toward including children with specific disabilities, teachers' self-efficacy might become higher, which matches the context of attitudes scale better, and, thus, its correlation became higher.

The results are in accord with a recent study that found only a weak correlation between general teacher self-efficacy and attitudes toward including students with SENs (Saloviita, 2020). Furthermore, these results may reflect the fact that inclusive pedagogy (that is, pedagogy to include all learners) remains ambiguous and is less studied compared to special education pedagogy (Florian & Black-Hawkins, 2011; Schwab, 2020). These findings suggest that single studies need to be interpreted with much more caution, especially in the way of the introduction, whether the study is about including all students, students with SENs, or students with specific disabilities. Moreover, the findings of this study further support the idea that although the relationship between teachers’ self-efficacy and attitudes seems to be universal, how other variables such as experience in teaching students with SENs, the amount of inclusive education training, and sources of self-efficacy might affect the outcomes differently in different countries, as pointed out in previous research (Bonneville-Roussy et al., 2019; Malinen et al., 2013; Savolainen et al., 2012; Yada et al., 2018, 2019).

Therefore, further research should be undertaken to investigate how those variables predict the self-efficacyeattitudes relationship. Finally, the number of items and the type of Likert scale for both teachers' self-efficacy and attitude scales failed to explain the variance in the relationship between the two. However, the reliability of the attitude scale had a significant effect on the self-efficacyeattitudes relationship, in which higher reliability predicted a larger correlation coefficient. Nonetheless, such an effect was not found for the teachers' self-efficacy scale. These results are likely to be related to the variation of reliability, in which reliabilities of attitude scales varied from moderate to high (between .62 and .97; M ¼ 0.78), but those of self-efficacy scales were mostly high (between 0.60 and 0.94; M ¼ 0.85).

If we believe that the studies using more reliable instruments represent results from the actual population, these findings appear to suggest that the relationship between teachers’ self-efficacy and attitudes could be slightly higher than what the general literature would suggest. Because this issue was not discussed in previous research, caution is warranted in interpreting the results, and further work is required.

5. Limitations and future direction

Although this study contributed to the understanding of the overall picture of empirical research on teachers’ self-efficacy and attitudes in inclusive education, some limitations should be noted. First, our search procedure may not have included some relevant studies because we only included published studies, theses, and dissertations in which the information was available in online databases. In a similar vein, this research is limited to reports written in English, though there might be related studies written in other languages. Likewise, as in all reviews, zero effect sizes are rarely reported in the publications, and, therefore, the correlation might be overestimated. Further, because of a long process of literature search and coding, the present study was not able to include recent studies conducted between 2019 and 2021 (e.g., Avramidis et al., 2019; Kisbu-Sakarya & Doenyas, 2021; Narkun & Smogorzewska, 2019; Opoku et al., 2020; Saloviita, 2020; Schwab & Alnahdi, 2020).

These issues could cause some publication bias (Cooper et al., 1997; Scargle, 2000). Second, this study was limited by the absence of some moderators that may affect the relationship between teachers' selfefficacy and attitudes. For instance, we could not include teachers' working places (i.e., in regular classes, special classes, or special schools) as a moderator because usually those teachers were included in one sample, though previous studies suggested that teachers in special education were more likely to have positive attitudes toward inclusion (Forlin et al., 2015; Norwich, 1994). Similarly, several variables, such as the number of years of each of their teaching careers, the amount of training in inclusive education, and experience in teaching students with SENs were identified as influencing teachers' self-efficacy and attitudes (e.g., de Boer et al., 2011; Savolainen et al., 2012; Yada et al., 2018).

Moreover, it has been suggested that student-related variables such as severity and type of students' disabilities might affect teachers’ attitudes toward inclusive education (Avramidis & Norwich, 2002). It is unfortunate that this study could not include such student-related variables because only nine studies were focusing on specific disabilities such as autism, hearing, and behavior disability. These variables may explain a significant amount of variance in the selfefficacyeattitudes relationship, and more studies are also required to identify potential interaction effects among the various moderators. Third, another note of caution regarding cultural differences is that the present study examined whether individualism/collectivism culture affected the relationship between self-efficacy and attitudes.

Although this dichotomy is useful, there might be a need to examine the relationship by considering the interplay of other cultural dimensions such as power distance and uncertainty avoidance (Hofstede, 2001; van Steen & Wilson, 2020). Since it was impossible to divide studies into more groups because of the small sample size in this study, future studies on the current topic with a large sample size are recommended. Similarly, inclusive education is greatly influenced by each country's policies and educational system (Yada et al., 2018). Further studies regarding the role of not only cultural contexts but also national policies and educational systems would be worthwhile. Fourth, while there is a common understanding that teachers' self-efficacy comes first and attitudes follow as an outcome (Ajzen & Madden, 1986), solid evidence of such causal direction using longitudinal data is only beginning to emerge (see Savolainen et al., 2020).

Further systematic reviews on the self-efficacyeattitudes relationship using longitudinal studies should be carried out to confirm the causal relationship of teachers’ self-efficacy on their attitudes. Finally, our quantitative synthesis provided an overarching picture of the relationship; however, it did not offer a procedural understanding of how teachers’ self-efficacy influences their attitudes toward inclusive education. Thus, it would be highly useful to include studies whose authors used qualitative methods to provide detailed descriptions of what happens regarding this relationship. Methodically evaluated, we did not correct study artifacts potentially biasing the true correlation between teachers' selfefficacy and attitudes toward inclusive education (Hunter & Schmidt, 2015).

This is, we recognize, a major limitation. However, because we saw large measurement instrument variation among studies, we concluded that we were not able to correct measurements to the level that they are invariant enough across studies. A. Yada, M. Leskinen, Consequently, there is a need to develop further self-efficacy and attitude measurement scales, test their latent structures’ properties, and demonstrate that invariant assumption holds for different settings and contexts. Power is a major concern for all meta-analytical studies (Jackson & Turner, 2017). We identified 43 studies, which is a relatively large number for providing a trustworthy point estimate and narrow confidence interval.

The same is not true for meta-regressions. Some subgroups were rather small, which lowers power and may lead to statistically non-significant results. We decided not to use the multiple meta-regression method because it would have lowered the number of studies to 20. As noted earlier, we did not analyze interaction patterns, which, in turn, should be taken into account in future meta-analysis studies. Finally, future studies may benefit from other and more sophisticated methods than this barebones meta-analysis (e.g., Bayesian meta-analysis; R€over, 2017).

6. Conclusions and implications

The results of this study indicated that the link between teachers' self-efficacy beliefs and their attitudes toward inclusive education is positive and moderate (r¼0.351). More specifically, the correlation between the two concepts was found to vary widely, which underpins the meaningfulness of investigating existing studies' results more systematically. Our findings suggest that single studies should not be over-interpreted when summarizing existing results, as they are somewhat heterogeneous. Furthermore, findings from this study revealed that teachers' self-efficacy has a significant relationship with one of teacher attributes (i.e., attitudes in this study), as has been indicated in previous teachers' self-efficacy research (Aloe et al., 2014; Chesnut & Burley, 2015; Klassen & Tze, 2014).

One contribution of this study has been to provide further evidence in support of two often used theories, that is, the self-efficacy theory of Bandura (1977) and the theory of planned behavior (Ajzen, 2012). Bandura (1994) claims that selfbeliefs of efficacy affect human functioning, that is, cognitive, motivational, selection, and affective processes including one's attitudes. Although our study did not reveal a causal relationship of self-efficacy on attitudes, it provided some evidence on the universality of this relationships and the importance of context specificity of self-efficacy as described by Bandura (1977). The present study also partially relates to the theory of planned behavior in terms of the relationship between self-efficacy and attitudes (Ajzen, 2012) and suggests that studies on inclusive practices should perhaps consider both attitudes and efficacy as possible predictors of teacher behavior.

However, what has remained unclear in this study as well as previous literature is whether selfefficacy and attitudes influence actual behavior (i.e., inclusive practices) (Carew et al., 2018). Thus, a major implication is that further work is needed to fully understand the relationship among teachers' self-efficacy, attitudes, and behavior in classroom. Our findings further highlight the importance of teachers' selfefficacy not only in inclusive practices but also in different kinds of teaching settings. A practical implication of this finding is in line with that of an earlier study indicating that greater efforts are needed to develop preand in-service teacher training, in which teachers can increase their self-efficacy (Avramidis et al., 2019; Desombre et al., 2019; Lee et al., 2014).

Some studies reviewed in this paper (N ¼ 3 out of the 41 studies, 1568 in-service teachers) conducted an intervention in teacher training programs, which aimed to improve teachers' self-efficacy and attitudes in inclusive education, and they found that the intervention enhanced their self-efficacy and attitudes (Carew et al., 2018; Kormos & Nijakowska, 2017; Lifshitz et al., 2004). For example, Carew et al. (2018) reported that intervention programs covering various subjects such as key concepts of inclusive education, child-centered learning approaches, and classroom management increased teaching self-efficacy and produced more positive attitudes toward inclusive education. Further, one longitudinal study found that a short online course related to inclusive practices, which took a taskbased approach for teacher professional development, had a positive influence in terms of elevating teachers' self-efficacy and alleviating their concerns on inclusion (Kormos & Nijakowska, 2017).

In the time of the COVID-19 pandemic, organizing such online courses in preand in-service teacher training might be beneficial to enhance teachers’ self-efficacy and diminish their concerns about inclusive education. When attitudes toward inclusion are regarded as one barrier for inclusive education, it is of utmost importance that new teachers are given a chance to experience inclusive teaching in a safe and supporting environment as a part of their initial teacher education and, thus, hopefully, gain positive experiences and support to have a stronger efficacy to enter the demanding realities of inclusive schools of today. In addition, some previous studies suggested that targeting the four different sources of self-efficacy, that is, mastery experience, vicarious experience, verbal persuasion, and affective state, might be one way to improve teachers' self-efficacy (Bandura, 1997; Desombre et al., 2019; Morris et al., 2017).

Since mastery experience is considered as the most powerful source among the four (Bandura, 1997), creating possibilities to trainees to gain positive and successful experiences in inclusive practices in preand inservice training would be worthwhile and could increase teachers' self-efficacy (Desombre et al., 2019; Yada et al., 2019). Moreover, verbal persuasion, which means feedback from colleagues and other professionals, may have a significant influence on teachers’ self-efficacy (Morris et al., 2017; Yada et al., 2019). Therefore, school leaders and policy makers should pay more attention to developing a learning community in schools, in which teachers can have regular and positive discussions with their colleagues to improve their self-efficacy beliefs, and subsequently their attitudes, to work successfully in an inclusive education context.

Furthermore, the current study found that the relationship between teachers’ self-efficacy and attitudes toward inclusive education became stronger when the context of the scales is specific to including children with disabilities. Looking at this from a different angle, this study revealed that a large amount of research on inclusive education still focuses on including children with disabilities or SENs, even though many international documents emphasize inclusive education for all children (UNESCO, 1994; United Nations Educational, Scientific and Cultural Organization, 2005; United Nations, 2006; United Nations General Assembly, 2015). Further work on including all kinds of children will have to be conducted in order to bridge this apparent gap between policy documents and actual practices.

Finally, our findings are relevant not only for teachers but also for other stakeholders such as students and parents. As mentioned, teachers' self-efficacy and attitudes may be related to their behavior, that is, to their inclusive practices in this particular context (Ajzen, 2012; Sharma & Jacobs, 2016), and it could further influence students’ educational experience in school (van Steen & Wilson, 2020). If students and parents experience a more positive inclusive classroom and get familiar with various kinds of children, they could be more confident to communicate with diverse children and their attitudes toward inclusion might be enhanced in a positive direction. This may contribute toward the development of a more inclusive society in the future.

Funding

Jyväskylän Yliopisto

Declaration of competing interest

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Acknowledgements

We would like to thank Ms. Alexandra Gutschik for helping with this research project.

References

  1. 1van Aert (2019). A new justification of the Hartung-Knapp method for random-effects meta-analysis based on weighted least squares regression. Research Synthesis Methods, 10, 515. https://doi.org/10.1002/jrsm.1356
  2. 2Ajzen (2012). Attitudes and persuasion.
  3. 3Ajzen (1980).
  4. 4Ajzen (1986). Prediction of goal-directed behavior: Attitudes, intentions, and perceived behavioral control. Journal of Experimental Social Psychology, 22, 453. https://doi.org/10.1016/0022-1031(86)90045-4
  5. 5Alaverdyan (2018).
  6. 6Alghazo (2004). General education teachers in the United Arab Emirates and their acceptance of the inclusion of students with disabilities. British Journal of Special Education, 31, 94. https://doi.org/10.1111/j.0952-3383.2004.00335.x
  7. 7Aloe (2014). Classroom management self-efficacy and burnout: A multivariate meta-analysis. Educational Psychology Review, 26, 101. https://doi.org/10.1007/s10648-013-9244-0
  8. 8Angiulo (2014).
  9. 9Antonak (1995). Psychometric analysis and revision of the opinions relative to mainstreaming scale. Exceptional Children, 62, 139. https://doi.org/10.1177/001440299506200204
  10. 10Artiles (2005). Inclusive education in the globalization age: The promise of comparative cultural-historical analysis.
  11. 11Avanzi (2013). Cross-validation of the Norwegian teacher's self-efficacy scale (NTSES). Teaching and Teacher Education, 31, 69. https://doi.org/10.1016/j.tate.2013.01.002
  12. 12Avery (2017).
  13. 13Avramidis (2002). Teachers' attitudes towards integration/inclusion: A review of the literature. European Journal of Special Needs Education, 17, 129. https://doi.org/10.1080/08856250210129056
  14. 14Avramidis (2020). Attitudes and inclusion of students with special educational needs in regular schools.
  15. 15Avramidis (2019). Teachers' attitudes towards inclusion and their self-efficacy for inclusive practices as predictors of willingness to implement peer tutoring. Journal of Research in Special Educational Needs, 19, 49. https://doi.org/10.1111/1471-3802.12477
  16. 16Bandura (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84, 191. https://doi.org/10.1037/0033-295X.84.2.191
  17. 17Bandura (1994). Self-efficacy.
  18. 18Bandura (1997).
  19. 19Bandura (2002). Social cognitive theory in cultural context. Applied Psychology: International Review, 51, 269. https://doi.org/10.1111/1464-0597.00092
  20. 20Bandura (2012). On the functional properties of perceived self-efficacy revisited. Journal of Management, 38, 9. https://doi.org/10.1177/0149206311410606
  21. 21Bender (1995). Teachers' attitudes toward increased mainstreaming: Implementing effective instruction for students with learning disabilities. Journal of Learning Disabilities, 28, 87. https://doi.org/10.1177/002221949502800203
  22. 22Benton-Borghi (2006).
  23. 23Bettineski (2015).
  24. 24de Boer (2011). Regular primary schoolteachers' attitudes towards inclusive education: A review of the literature. International Journal of Inclusive Education, 15, 331. https://doi.org/10.1080/13603110903030089
  25. 25Bohner (2011). Attitudes and attitude change. Annual Review of Psychology, 62, 391. https://doi.org/10.1146/annurev.psych.121208.131609
  26. 26Bonneville-Roussy (2019). The role of cultural values in teacher and student self-efficacy: Evidence from 16 nations. Contemporary Educational Psychology, 59, 101798. https://doi.org/10.1016/j.cedpsych.2019.101798
  27. 27Borenstein (2013).
  28. 28Card (2011).
  29. 29Carew (2018). The impact of an inclusive education intervention on teacher preparedness to educate children with disabilities within the Lakes Region of Kenya. International Journal of Inclusive Education, 23, 229. https://doi.org/10.1080/13603116.2018.1430181
  30. 30Chan (2016).
  31. 31Chesnut (2015). Self-efficacy as a predictor of commitment to the teaching profession: A meta-analysis. Educational Research Review, 15, 1. https://doi.org/10.1016/j.edurev.2015.02.001
  32. 32Cochran (1997). The development and psychometric analysis of the scale of teachers' attitudes toward inclusion (STATIC).
  33. 33Cohen (1988).
  34. 34Cooper (1997). Finding the missing science: The fate of studies submitted for review by a human subjects committee. Psychological Methods, 2, 447. https://doi.org/10.1037/1082-989X.2.4.447
  35. 35Davis (2015).
  36. 36Desombre (2019). French teachers' general attitude toward inclusion: The indirect effect of teacher efficacy. Educational Psychology, 39, 38. https://doi.org/10.1080/01443410.2018.1472219
  37. 37Eagly (1993).
  38. 38Eriks-Brophy (2013). Teachers' perceptions of the inclusion of children with hearing loss in general education settings. American Annals of the Deaf, 158, 63. https://doi.org/10.1353/aad.2013.0009
  39. 39Fan (2017). Homework and students' achievement in math and science: A 30-year meta-analysis, 1986–2015. Educational Research Review, 20, 35. https://doi.org/10.1016/j.edurev.2016.11.003
  40. 40Florian (2011). Exploring inclusive pedagogy. British Educational Research Journal, 37, 813. https://doi.org/10.1080/01411926.2010.501096
  41. 41Forlin (2011). The Sentiments, Attitudes, and Concerns about Inclusive Education Revised (SACIE-R) scale for measuring pre-service teachers' perceptions about inclusion. Exceptionality Education International, 21, 50. https://doi.org/10.5206/eei.v21i3.7682
  42. 42Forlin (2015). Educational reform in Japan towards inclusion: Are we training teachers for success?. International Journal of Inclusive Education, 19, 314. https://doi.org/10.1080/13603116.2014.930519
  43. 43Frizzell (2018).
  44. 44Gemmell-Crosby (1994). Preschool teachers’ perceptions of including pupils with disabilities. Education and Training in Mental Retardation and Developmental Disabilities, 29, 279.
  45. 45Ghaith (1997). Relationships among experience, teacher efficacy, and attitudes toward the implementation of instructional innovation. Teaching and Teacher Education, 13, 451. https://doi.org/10.1016/S0742-051X(96)00045-5
  46. 46Gibson (1984). Teacher efficacy: A construct validation. Journal of Educational Psychology, 76, 569. https://doi.org/10.1037/0022-0663.76.4.569
  47. 47Glaubman (2001). Ultra-orthodox Jewish teachers' self-efficacy and willingness for inclusion of pupils with special needs. European Journal of Special Needs Education, 16, 207. https://doi.org/10.1080/08856250110074373
  48. 48Greenwald (1975). Consequences of prejudice against the null hypothesis. Psychological Bulletin, 82, 1. https://doi.org/10.1037/h0076157
  49. 49Guralnick (2008). International perspectives on early intervention: A search for common ground. Journal of Early Intervention, 30, 90. https://doi.org/10.1177/1053815107313483
  50. 50Guskey (1988). Teacher efficacy, self-concept, and attitudes toward the implementation of instructional innovation. Teaching and Teacher Education, 4, 63. https://doi.org/10.1016/0742-051X(88)90025-X
  51. 51Hernandez (2016). General education and special education teachers' attitudes towards inclusion. Journal of the American Academy of Special Education Professionals. https://doi.org/10.64546/jaasep.318
  52. 52Hofstede (2001).
  53. 53Hofstede (2010).
  54. 54Hunter (2015).
  55. 55Jackson (2017). Power analysis for random-effects meta-analysis. Research Synthesis Methods, 8, 290. https://doi.org/10.1002/jrsm.1240
  56. 56Kisbu-Sakarya (2021). Can school teachers' willingness to teach ASD-inclusion classes be increased via special education training? Uncovering mediating mechanisms. Research in Developmental Disabilities, 113, 103941. https://doi.org/10.1016/j.ridd.2021.103941
  57. 57Klassen (2004). A cross-cultural investigation of the efficacy beliefs of South Asian immigrant and Anglo Canadian nonimmigrant early adolescents. Journal of Educational Psychology, 96, 731. https://doi.org/10.1037/0022-0663.96.4.731
  58. 58Klassen (2010). Effects on teachers' self-efficacy and job satisfaction: Teacher gender, years of experience, and job stress. Journal of Educational Psychology, 102, 741. https://doi.org/10.1037/a0019237
  59. 59Klassen (2014). Teachers' self-efficacy, personality, and teaching effectiveness: A meta-analysis. Educational Research Review, 12, 59. https://doi.org/10.1016/j.edurev.2014.06.001
  60. 60Klassen (2011). Teacher efficacy research 1998-2009: Signs of progress or unfulfilled promise?. Educational Psychology Review, 23, 21. https://doi.org/10.1007/s10648-010-9141-8
  61. 61Kormos (2017). Inclusive practices in teaching students with dyslexia: Second language teachers' concerns, attitudes and self-efficacy beliefs on a massive open online learning course. Teaching and Teacher Education, 68, 30. https://doi.org/10.1016/j.tate.2017.08.005
  62. 62Kuittinen (2017).
  63. 63Leahy (2014).
  64. 64Lee (2014). What predicts teachers' acceptance of students with special educational needs in kindergarten?. Australian Journal of Educational and Developmental Psychology, 14, 60.
  65. 65Leyser (1994). Teacher attitudes toward mainstreaming: A cross-cultural study in six nations. European Journal of Special Needs Education, 9, 1. https://doi.org/10.1080/0885625940090101
  66. 66Lifshitz (2004). Attitudes towards inclusion: The case of Israeli and Palestinian regular and special education teachers. European Journal of Special Needs Education, 19, 171. https://doi.org/10.1080/08856250410001678478
  67. 67Lüke (2018). What do I think about inclusive education? It depends on who is asking. experimental evidence for a social desirability bias in attitudes towards inclusion. International Journal of Inclusive Education, 22, 38. https://doi.org/10.1080/13603116.2017.1348548
  68. 68MacFarlane (2013). Teacher attitudes and behavior toward the inclusion of children with social, emotional and behavioral difficulties in mainstream schools: An application of the theory of planned behavior. Teaching and Teacher Education, 29, 46. https://doi.org/10.1016/j.tate.2012.08.006
  69. 69Malak (2018). Predictors of primary schoolteachers' behavioural intention to teach students demonstrating inappropriate behaviour in regular classrooms. Cambridge Journal of Education, 48, 495. https://doi.org/10.1080/0305764X.2017.1364698
  70. 70Malinen (2013). Exploring teacher self-efficacy for inclusive practices in three diverse countries. Teaching and Teacher Education, 33, 34. https://doi.org/10.1016/j.tate.2013.02.004
  71. 71Malinen (2012). Beijing in-service teachers' self-efficacy and attitudes towards inclusive education. Teaching and Teacher Education, 28, 526. https://doi.org/10.1016/j.tate.2011.12.004
  72. 72Markus (1991). Culture and the self: Implications for cognition, emotion, and motivation. Psychological Review, 98, 224. https://doi.org/10.1037/0033-295X.98.2.224
  73. 73Meijer (1988). The effect of teacher self-efficacy on referral chance. The Journal of Special Education, 22, 378. https://doi.org/10.1177/002246698802200309
  74. 74Mian (2017).
  75. 75Milner (2002). A case study of an experienced English teacher's self-efficacy and persistence through “crisis” situations: Theoretical and practical considerations. High School Journal, 86, 28. https://doi.org/10.1353/hsj.2002.0020
  76. 76Mitchell (2005). Introduction: Sixteen propositions on the contexts of inclusive education.
  77. 77Moberg (2019). Struggling for inclusive education in Japan and Finland: Teachers' attitudes towards inclusive education. European Journal of Special Needs Education.
  78. 78Moher (2009). Preferred reporting items for systematic reviews and meta-analyses: The PRISMA statement. Journal of Clinical Epidemiology, 62, 1006. https://doi.org/10.1016/j.jclinepi.2009.06.005
  79. 79Montgomery (2014). Teachers' self-efficacy, sentiments, attitudes, and concerns about the inclusion of students with developmental disabilities. Exceptionality Education International, 24, 18. https://doi.org/10.5206/eei.v24i1.7708
  80. 80Morgan (2013).
  81. 81Morris (2017). Reconceptualizing the sources of teaching self-efficacy: A critical review of emerging literature. Educational Psychology Review, 29, 795. https://doi.org/10.1007/s10648-016-9378-y
  82. 82Munsell (2015).
  83. 83Munthe (2009). Lower secondary school teachers' judgements of pupils' problems. Teachers and Teaching: Theory and Practice, 15, 563. https://doi.org/10.1080/13540600903139571
  84. 84Narkun (2019). Studying self-Efficacy among teachers in Poland is important: Polish adaptation of the Teacher Efficacy for Inclusive Practice (TEIP) scale. Exceptionality Education International, 29, 110. https://doi.org/10.5206/eei.v29i2.9405
  85. 85Norwich (1994). The relationship between attitudes to the integration of children with special educational needs and wider socio-political views: A US-English comparison. European Journal of Special Needs Education, 9, 91. https://doi.org/10.1080/0885625940090108
  86. 86Omer (2015).
  87. 87Opoku (2020). Attitudes and self-efficacy as significant predictors of intention of secondary school teachers towards the implementation of inclusive education in Ghana. European Journal of Psychology of Education, 1.
  88. 88Pajares (1996). Self-efficacy beliefs in academic settings. Review of Educational Research, 66, 543. https://doi.org/10.3102/00346543066004543
  89. 89Pajares (2007). Culturalizing educational psychology.
  90. 90Qi (2012). Inclusion in physical education: A review of literature. International Journal of Disability, Development and Education, 59, 257. https://doi.org/10.1080/1034912X.2012.697737
  91. 91Randoll (2008).
  92. 92Rosenthal (1979). The file drawer problem and tolerance for null results. Psychological Bulletin, 86, 638. https://doi.org/10.1037/0033-2909.86.3.638
  93. 93Ross (2007). Professional development effects on teacher efficacy: Results of randomized field trial. Journal of Educational Research, 101, 50. https://doi.org/10.3200/JOER.101.1.50-60
  94. 94Röver (2017). Bayesian random-effects meta-analysis using the bayesmeta R package. Journal of Statistical Software, 93, 1.
  95. 95Saloviita (2015). Measuring pre-service teachers' attitudes towards inclusive education: Psychometric properties of the TAIS scale. Teaching and Teacher Education, 52, 66. https://doi.org/10.1016/j.tate.2015.09.003
  96. 96Saloviita (2020). Attitudes of teachers towards inclusive education in Finland. Scandinavian Journal of Educational Research, 64, 270. https://doi.org/10.1080/00313831.2018.1541819
  97. 97Samms (2017).
  98. 98Samuels (2015).
  99. 99San Martin (2021). Chilean teachers' attitudes towards inclusive education, intention, and self-efficacy to implement inclusive practices. Sustainability, 13, 2300. https://doi.org/10.3390/su13042300
  100. 100Savolainen (2012). Understanding teachers' attitudes and self-efficacy in inclusive education: Implications for pre-service and in-service teacher education. European Journal of Special Needs Education, 27, 51. https://doi.org/10.1080/08856257.2011.613603
  101. 101Savolainen (2020). Teacher efficacy predicts teachers' attitudes towards inclusion: A longitudinal cross-lagged analysis. International Journal of Inclusive Education, 1.
  102. 102Scargle (2000). Publication bias: The “file-drawer” problem in scientific inference. Journal of Scientific Exploration, 14, 91.
  103. 103Schwab (2018).
  104. 104Schwab (2019). Teachers' student-specific self-efficacy in relation to teacher and student variables. Educational Psychology, 39, 4. https://doi.org/10.1080/01443410.2018.1516861
  105. 105Schwab (2020). Inclusive and special education in Europe.
  106. 106Schwab (2020). Do they practise what they preach? Factors associated with teachers' use of inclusive teaching practices among in-service teachers. Journal of Research in Special Educational Needs, 20, 321. https://doi.org/10.1111/1471-3802.12492
  107. 107Sharma (2018). In-service teachers' attitudes, concerns, efficacy and intentions to teach in inclusive classrooms: An international comparison of Australian and Italian teachers. European Journal of Special Needs Education, 33, 437. https://doi.org/10.1080/08856257.2017.1361139
  108. 108Sharma (2016). Predicting in-service educators' intentions to teach in inclusive classrooms in India and Australia. Teaching and Teacher Education, 55, 13. https://doi.org/10.1016/j.tate.2015.12.004
  109. 109Sharma (2012). Measuring teacher efficacy to implement inclusive practices. Journal of Research in Special Educational Needs, 12, 12. https://doi.org/10.1111/j.1471-3802.2011.01200.x
  110. 110Shi (2019). The trim-and-fill method for publication bias: Practical guidelines and recommendations based on a large database of meta-analyses. Medicine, 98. https://doi.org/10.1097/MD.0000000000015987
  111. 111Skuller (2011).
  112. 112Sokal (2013). Canadian in-service teachers' concerns, efficacy, and attitudes about inclusive teaching. Exceptionality Education International, 23, 59.
  113. 113Soodak (1993). Teacher efficacy and student problem as factors in special education referral. The Journal of Special Education, 27, 66. https://doi.org/10.1177/002246699302700105
  114. 114Stanovic (1996).
  115. 115van Steen (2020). Individual and cultural factors in teachers' attitudes towards inclusion: A meta-analysis. Teaching and Teacher Education, 95, 103127. https://doi.org/10.1016/j.tate.2020.103127
  116. 116Tasnuba (2015). BRAC primary school teachers' teaching-efficacy, attitude, sentiment and concern towards inclusion of children with disabilities in regular classrooms in Bangladesh. Asian Journal of Inclusive Education, 3, 53. https://doi.org/10.59595/ajie.03.1.4
  117. 117Teo (2014). A measurement invariance analysis of the general self-efficacy scale on two different cultures. Journal of Psychoeducational Assessment, 32, 762. https://doi.org/10.1177/0734282914531707
  118. 118Tsakiridou (2014). Greek teachers' attitudes toward the inclusion of students with special educational needs. American Journal of Educational Research, 2, 208. https://doi.org/10.12691/education-2-4-6
  119. 119Tschannen-Moran (2001). Teacher efficacy: Capturing an elusive construct. Teaching and Teacher Education, 17, 783. https://doi.org/10.1016/S0742-051X(01)00036-1
  120. 120Tschannen-Moran (2007). The differential antecedents of self-efficacy beliefs of novice and experienced teachers. Teaching and Teacher Education, 23, 944. https://doi.org/10.1016/j.tate.2006.05.003
  121. 121Tschannen-Moran (1998). Teacher efficacy: Its meaning and measure. Review of Educational Research, 68, 202. https://doi.org/10.3102/00346543068002202
  122. 122United Nations (2006).
  123. 123(1994).
  124. 124(2005).
  125. 125(2020).
  126. 126(2015).
  127. 127Urton (2014). Attitudes towards inclusion and self-efficacy of principals and teachers. Learning Disabilities: A Contemporary Journal, 12, 151.
  128. 128Van Mieghem (2020). An analysis of research on inclusive education: A systematic search and meta review. International Journal of Inclusive Education, 24, 675. https://doi.org/10.1080/13603116.2018.1482012
  129. 129Veritas Health Innovation (2019).
  130. 130Vieluf (2013). Teacher self-efficacy in cross-national perspective. Teaching and Teacher Education, 35, 92. https://doi.org/10.1016/j.tate.2013.05.006
  131. 131Wang (2012). A survey study of Chinese in-service teachers' self-efficacy about inclusive education. Journal of International Special Needs Education, 15, 107. https://doi.org/10.9782/2159-4341-15.2.107
  132. 132Weisel (2006). School climate, sense of efficacy and Israeli teachers' attitudes toward inclusion of students with special needs. Education, Citizenship and Social Justice, 1, 157. https://doi.org/10.1177/1746197906064677
  133. 133Wilson (2016). The impact of social cognitive and personality factors on teachers' reported inclusive behaviour. British Journal of Educational Psychology, 86, 461. https://doi.org/10.1111/bjep.12118
  134. 134Wood (2017).
  135. 135Yada (2020).
  136. 136Yada (2017). Japanese in-service teachers' attitudes towards inclusive education and self-efficacy for inclusive practices. Teaching and Teacher Education, 64, 222. https://doi.org/10.1016/j.tate.2017.02.005
  137. 137Yada (2019). Teachers' self-efficacy and the sources of efficacy: A cross-cultural investigation in Japan and Finland. Teaching and Teacher Education, 81, 13. https://doi.org/10.1016/j.tate.2019.01.014
  138. 138Yada (2018). Teachers' attitudes and self-efficacy on implementing inclusive education in Japan and Finland: A comparative study using multi-group structural equation modelling. Teaching and Teacher Education, 75, 343. https://doi.org/10.1016/j.tate.2018.07.011