Teaching confidence is not a unitary construct but rather a multidimensional phenomenon, reflecting teachers’ self-perceived competence across distinct domains of practice (Charitaki et al., 2025c; Skaalvik and Skaalvik, 2007; Tschannen-Moran and Hoy, 2001; Umugiraneza et al., 2023). This multidimensional perspective is theoretically grounded in Bandura (1997) social cognitive theory, which conceptualizes self-efficacy as task- and context-specific, shaped by mastery experiences, social persuasion, and environmental affordances.

Empirical research increasingly supports this differentiated conceptualization, demonstrating that teachers may feel confident in some instructional domains while experiencing uncertainty or avoidance in others (Skaalvik and Skaalvik, 2007; Zee and Koomen, 2016). However, much of the existing literature continues to rely on broad or aggregated measures of teaching confidence, limiting the capacity to identify domain-specific strengths and vulnerabilities, particularly in inclusive and special education contexts where instructional demands are inherently complex (Savolainen et al., 2020). Within inclusive and special education settings, such differentiation becomes especially salient. Teachers are required to simultaneously address content mastery, behavioral regulation, and pedagogical flexibility while responding to heterogeneous learner needs (Savolainen et al., 2020; Zee and Koomen, 2020). Yet, prior studies often examine teaching confidence as a generalized disposition, obscuring how confidence operates differently across pedagogical domains and how these variations relate to inclusive practice outcomes.

Confidence in mathematics and science instruction is particularly salient in primary education, where these subjects frequently pose challenges for students with special educational needs and/or disabilities (SEND) due to their abstract, cumulative, and symbolic nature (Florian, 2019; Therrien et al., 2013). Empirical studies consistently show that teachers with higher mathematics teaching confidence are more likely to employ conceptual explanations, inquiry-based learning, and problem-solving strategies, as well as to persist when students struggle (Auslander et al., 2006; Gresham, 2008; Papic and Papic, 2025; Pietsch et al., 2003; Pratiwi et al., 2022; Auslander et al., 2006). These findings suggest that confidence in mathematics and science instruction plays a crucial role in shaping instructional quality and supporting learning for students with diverse educational needs.

At the same time, the literature reveals a persistent pattern of mathematics anxiety and content insecurity among both general and special education teachers, which negatively affects instructional quality and limits the use of adaptive pedagogies (Hodgen et al., 2018; Woodcock and Hardy, 2022). These findings underscore a key limitation of existing confidence measures: when subject-specific confidence is not distinguished from broader instructional self-efficacy, the field risks underestimating the unique pedagogical challenges associated with mathematics and science in inclusive classrooms and obscuring which supports are most likely to improve practice.

The second domain, instructional innovation, refers to teachers’ confidence in adopting and integrating novel pedagogical approaches, digital technologies, and differentiated strategies into classroom practice. Innovation is widely recognized as a cornerstone of inclusive education, as it enables teachers to design flexible learning environments responsive to diverse learner profiles (Florian and Black-Hawkins, 2011; Peng et al., 2024; Tomlinson, 2017; Trust and Whalen, 2020). Accordingly, confidence in instructional innovation represents a key mechanism through which inclusive principles are translated into everyday teaching practice.

Empirical evidence indicates that teachers with higher confidence in instructional innovation demonstrate stronger technological pedagogical content knowledge (TPACK) and are more likely to implement blended learning, Universal Design for Learning (UDL), and assistive technologies (CAST, 2024; Ertmer and Ottenbreit-Leftwich, 2010; Howard and Mozejko, 2015). However, studies also show that professional development alone is insufficient to ensure implementation, as confidence mediates the translation of training into practice (Clipa et al., 2023; Kim et al., 2013). Critically, existing teaching confidence instruments rarely operationalize innovation as a distinct domain, instead subsuming it under general instructional efficacy. This conflation limits the ability to examine how confidence in innovation uniquely contributes to inclusive pedagogy and may help explain mixed findings regarding technology integration in special education contexts (Sharma et al., 2008; Skaalvik and Skaalvik, 2007). Taken together, this evidence underscores the need to conceptualize and measure instructional innovation confidence as a distinct dimension in order to better understand its role in the effective enactment of inclusive and technology-enhanced teaching practices.

The third domain, classroom management confidence, underpins teachers’ capacity to establish structured, emotionally supportive, and responsive learning environments. Effective classroom management is particularly crucial in inclusive settings, where behavioral, attentional, and emotional diversity is pronounced (Emmer and Sabornie, 2014; Friedman, 2003). In this context, classroom management confidence functions as a foundational condition for maintaining instructional continuity and supporting positive classroom interactions in inclusive classrooms.

Research consistently shows that teachers with strong classroom management confidence employ proactive strategies such as clear routines, positive reinforcement, and collaborative problem-solving that enhance student engagement and emotional safety (Kunter et al., 2013; Reupert and Woodcock, 2010; Slater and Main, 2020). Conversely, low confidence in this domain is associated with reactive discipline practices, increased stress, and reduced instructional quality (Friedman, 2003; Tsouloupas et al., 2010). Together, these findings highlight classroom management confidence as a critical determinant of both classroom climate and teachers’ professional well-being in inclusive educational settings.

Although classroom management is included in many efficacy scales, it is often treated superficially, without attention to the relational and inclusive dimensions of management required in special education. This limitation further supports the need for instruments that explicitly capture confidence in managing diverse and inclusive classroom environments.

The relevance of these domains is particularly pronounced within the Greek inclusive education system, which has undergone significant transformation over the past two decades. Inclusive practices in Greece are primarily implemented through parallel support, resource rooms, and differentiated instruction within mainstream classrooms. Primary special education teachers frequently work across multiple classrooms and student populations, often with limited institutional support and high role ambiguity.

Empirical and policy-oriented research highlights that Greek teachers face considerable challenges related to curriculum adaptation, collaboration with general educators, classroom management, and the integration of digital and innovative practices (UNESCO, 2020). These systemic characteristics intensify the demands placed on teachers’ confidence across subject-specific, pedagogical, and management domains, reinforcing the need for context-sensitive measurement tools tailored to the Greek inclusive setting.

A robust body of empirical literature documents the links between teaching confidence, instructional practices, and student outcomes (Klassen and Chiu, 2010; Zee and Koomen, 2016). High-confidence teachers are more likely to adopt evidence-based strategies, differentiate instruction, and persist in the face of instructional challenges (Caprara et al., 2006; Tschannen-Moran and Hoy, 2001), whereas low confidence is associated with instructional rigidity and avoidance of innovation (Holzberger et al., 2013; Wyatt, 2018). In mathematics and science education, numerous studies demonstrate that teaching confidence predicts instructional quality, feedback practices, and students’ achievement gains (Basister et al., 2025; Lazarides et al., 2020; Pietsch et al., 2003; Auslander et al., 2006). Teachers with higher mathematics teaching confidence use conceptual explanations, problem-solving activities, and formative assessment techniques, promoting deeper understanding and persistence among students with learning difficulties (Goddard et al., 2007; Hodgen et al., 2018). Similarly, science teaching confidence has been linked to inquiry-based approaches and greater student engagement, particularly when teachers perceive autonomy and support from leadership (Appleton, 2008; Franklin and Chang, 2025; Lee, 2014; Park et al., 2025). Confidence in classroom management also exerts a significant influence on both teacher behavior and student well-being: teachers with strong management confidence demonstrate proactive discipline, emotional regulation, and relational sensitivity, leading to improved student behavior, academic focus, and socio-emotional adjustment (Kunter et al., 2013; Poulou et al., 2018; Reupert and Woodcock, 2010; Schillinger et al., 2021). Meta-analytic evidence confirms that high classroom management self-efficacy correlates with lower stress and burnout, greater job satisfaction, and enhanced classroom climate (Aloe et al., 2014; Charitaki et al., 2025d; Friedman, 2003). Collectively, this evidence illustrates that teaching confidence operates across instructional and management domains to shape both the quality of teaching practice and a wide range of student academic and socio-emotional outcomes.

Within inclusive contexts, teaching confidence further predicts teachers’ attitudes toward inclusion, use of accommodations, and commitment to diversity (Asp et al., 2025; Avramidis and Norwich, 2002; Savolainen et al., 2020; Sharma et al., 2012). Confident teachers are more likely to adopt UDL principles, employ assistive technologies, and collaborate with specialists to support students with special educational needs and/or disabilities (CAST, 2024; Florian and Black-Hawkins, 2011; Kim et al., 2013). They also show greater emotional resilience and lower depersonalization, reducing the risk of burnout in complex classroom environments (Karakose et al., 2023; Klassen and Tze, 2014; Sokal and Sharma, 2013). Consequently, a validated, multidimensional measure of teaching confidence encompassing Mathematics/Science Instruction, Instructional Innovation, and Classroom Management can provide valuable insights for both research and practice. Psychometrically robust assessment tools enable policymakers and school leaders to identify professional learning needs, design targeted training programs and evaluate the impact of interventions on teacher growth and student inclusion outcomes (Organisation for Economic Co-operation and Development, 2019; UNESCO, 2020). Overall, this body of evidence underscores the importance of assessing teaching confidence multidimensionally in order to inform inclusive policy, professional development, and evidence-based educational decision-making.

Despite this robust body of evidence, much of the existing research relies on global or weakly differentiated measures of teaching confidence, limiting the ability to disentangle domain-specific effects across mathematics/science instruction, instructional innovation, and classroom management. This measurement gap is particularly salient in inclusive settings, where pedagogical demands vary substantially across domains, underscoring the need for a validated multidimensional instrument.

The reviewed literature also reveals methodological limitations in the measurement of teaching confidence, with heavy reliance on Classical Test Theory (CTT) approaches that provide limited insight into item-level functioning and multidimensional structure (DeVellis, 2017; DeMars, 2010; Sijtsma and Junker, 2006). Item Response Theory (IRT) offers a powerful alternative by enabling the examination of item discrimination, difficulty, and measurement precision across confidence levels (Boone et al., 2014; Embretson and Reise, 2000; Reise and Revicki, 2015). Thus, the adoption of IRT-based approaches represents a critical methodological advancement for capturing the complexity and domain-specific nature of teaching confidence.

By supporting multidimensional modeling and differential item functioning analyses, IRT aligns closely with the theoretical and empirical need to capture domain-specific teaching confidence in inclusive education (Chen et al., 2006; Putnick and Bornstein, 2016; Tennant and Conaghan, 2007). In this way, IRT provides a methodologically coherent framework for examining whether teaching confidence measures function equivalently across domains and teacher groups in inclusive contexts.

Collectively, the reviewed literature highlights both the central role of teaching confidence in inclusive education and the conceptual and methodological shortcomings of existing instruments. These gaps provide a clear rationale for the development and validation of a multidimensional Teaching Confidence Scale encompassing Mathematics/Science Instruction, Instructional Innovation, and Classroom Management, particularly within the Greek inclusive education context.

The purpose of this study is to examine the psychometric properties of the Greek version of the Teaching Confidence instrument within the context of mathematics education, with a particular focus on inclusive practices for students with Special Educational Needs and/or Disabilities (SEND). Specifically, the study seeks to validate the underlying factor structure of the instrument, assess its reliability, and test its measurement invariance across different groups of teachers. By doing so, the research aims to provide a reliable and context-sensitive tool for capturing educators’ teaching confidence. Based on prior empirical evidence, the following hypotheses were formulated:

The Greek version of the Teaching Confidence Scale is expected to demonstrate an adequate fit to a three-factor structure of teaching confidence.

The items of each measure can be found in the supplementary material.

The Teaching Confidence Scale was adapted into Greek following established guidelines for cross-cultural instrument adaptation. The original English version was independently translated into Greek by two bilingual researchers with expertise in special and inclusive settings. The two translations were compared and synthesized into a preliminary Greek version. This preliminary version was reviewed by experts in special education and inclusive pedagogy to assess item clarity, cultural relevance, and alignment with inclusive settings principles. Based on expert feedback, minor linguistic refinements were made to improve clarity and contextual appropriateness. Prior to the main data collection, the adapted scale was piloted with a small group of special education teachers to assess item comprehensibility and response format. No major issues were identified during piloting, and the final Greek version was deemed suitable for use in inclusive primary education settings.

To address the inclusive focus of the study, a systematic comparison between the original and adapted items of the Teaching Confidence Scale was undertaken during the adaptation process. In the original version, items primarily assessed teachers’ confidence in implementing subject-specific instruction and classroom practices within a general education framework (e.g., confidence in preparing mathematics lessons, teaching algebra, using journals, or managing classrooms).

In the adapted version, these items were conceptually refined to reflect inclusive classroom realities. For example, original items assessing confidence in teaching mathematics and science (e.g., preparing mathematics lessons, teaching fractions or algebra, explaining standardized test scores) were reframed to emphasize confidence in teaching these subjects to heterogeneous classrooms that include students with special educational needs and/or disabilities, as well as in supporting conceptual understanding through differentiated and inquiry-based approaches.

Similarly, original items related to instructional practices and innovation (e.g., constructing web-based activities, using cooperative learning approaches, evaluating software for teaching and learning) were adapted to explicitly highlight inclusive instructional strategies, such as flexible instructional design, differentiated teaching, and the use of digital tools to support diverse learners.

Finally, items associated with classroom management and the social context of teaching (e.g., managing classrooms, establishing a sense of community, addressing cultural diversity) were adapted to reflect proactive, supportive, and relational management practices aligned with inclusive settings principles, rather than a sole focus on control or discipline. This comparison demonstrates that the adaptation process extended beyond linguistic translation and involved conceptual alignment of the items with inclusive settings principles, while preserving the theoretical intent of the original scale.

Prior to the main analyses, data were screened for completeness and suitability for factor and item response analyses. Descriptive statistics were examined, and the ordinal nature of the response scale was considered in all subsequent analyses. Data analyses were conducted using R version 4.4.3 (R Development Core Team, 2019). Initially, construct validity of the Teaching Confidence Scale was examined using Confirmatory Factor Analysis (CFA) implemented with the lavaan package (Rosseel, 2012). Given the ordinal nature of the six-point Likert-type response scale, Confirmatory Factor Analysis was conducted using an estimator appropriate for ordered categorical data. Specifically, the WLSMV estimator was employed, as it provides robust parameter estimates and standard errors for ordinal indicators and is recommended for Likert-type data with fewer than seven response categories. Model fit was evaluated using the Root Mean Square Error of Approximation (RMSEA), the Comparative Fit Index (CFI), and the Tucker–Lewis Index (TLI). Following the assessment of factorial validity, internal consistency and model-based reliability were examined using Cronbach’s alpha, Bentler’s Omega, Bollen’s Omega, McDonald’s Omega, and Average Variance Extracted (AVE), calculated with the psych (Revelle, 2019) and semTools packages (Jorgensen, 2019). Subsequently, measurement invariance across teachers’ current role in education (substitute vs. certified special education teachers) was tested through multigroup CFA, evaluating configural, metric, scalar, and strict invariance models.

Finally, Item Response Theory (IRT) analyses were conducted to examine item fit, discrimination, difficulty parameters, and measurement precision. Given the polytomous six-point Likert-type response format of the Teaching Confidence Scale, the Generalized Partial Credit Model (GPCM) was selected a priori for item-level analyses. The GPCM was deemed appropriate due to its flexibility in allowing item discrimination parameters to vary across items, which was theoretically justified by the multidimensional nature of teaching confidence and the expectation that items would differ in their discriminative power. Although alternative IRT models for ordinal data, such as the Partial Credit Model (PCM), Graded Response Model (GRM), or Rating Scale Model (RSM), could also be applied, these models were not formally compared, as the study focused on item-level functioning rather than empirical model selection. Item and test information functions were estimated to evaluate the precision of measurement across different levels of the latent traits. Additional R packages used in the analyses included haven (Wickham, 2019a), psych (Revelle, 2019), MVN (Korkmaz, 2019), semPlot (Epskamp, 2019), caret (Kuhn, 2019), dplyr (Wickham, 2019a), tidyr (Wickham, 2019b), and knitr (Xie, 2019). Together, these diagnostics provided comprehensive evidence regarding the adequacy and robustness of the proposed measurement models.

Firstly, we employed the 3-factor solution as it was suggested by the literature. This solution included the first factor F1, namely, “Confidence to teach math and science” (TC.11, TC.18, TC.10, TC.8, TC.16, TC.27, TC.24, TC.15, TC.12, TC.1, TC.20, TC.9, and TC.6), the second factor F2, namely, “Confidence to use instructional innovations” (TC.13, TC.25, TC.26, TC.19, TC.17, TC.2, TC.22, TC.28, TC.21, and TC.5), and the third factor F3, namely, “Confidence to manage classrooms” (TC.3, TC.7, TC.4, TC.30, TC.14, and TC.29). Figure 1 presents the results from CFA for the aforementioned solution. Fit indices validated the 3-factor solution (RMSEA = 0.027, CFI = 0.994, TLI = 0.956).

Additionally, to overall model fit, factor loadings (λ) were examined to further evaluate construct validity. All items loaded significantly on their respective latent factors, with standardized factor loadings exceeding commonly accepted thresholds (λ > 0.50), indicating strong associations between the observed indicators and their underlying constructs. Specifically, items assessing confidence in mathematics and science instruction demonstrated consistently high loadings on the first factor, reflecting coherent measurement of subject-specific teaching confidence. Items related to instructional innovation loaded strongly on the second factor, supporting the conceptual distinctiveness of teachers’ confidence in adopting innovative, flexible, and differentiated pedagogical practices. Items associated with classroom management also exhibited substantial loadings on the third factor, indicating that they adequately represent teachers’ confidence in managing inclusive classroom environments. The magnitude, consistency, and statistical significance of the factor loadings provide additional evidence that the items effectively capture their intended latent constructs, thereby supporting the construct validity of the Teaching Confidence Scale beyond global model fit indices.

To further support the validity of the Teaching Confidence Scale, content validity was evaluated by a panel of experts in special and inclusive settings. The panel consisted of 5 academics and 10 practitioners with expertise in inclusive pedagogy, special education, mathematics education, and educational measurement. Experts reviewed the adapted items for clarity, relevance, and alignment with inclusive settings principles. Overall, the experts agreed that the items adequately represented the intended domains of teaching confidence and were appropriate for inclusive and special education contexts. Minor wording refinements were suggested and incorporated prior to data collection.

We also estimated four different coefficients to assess the TCS reliability. More specifically, we estimated Cronbach’s Alpha, Bentler’s Omega, McDonald’s Omega, and Average Variance Extracted (AVE) coefficients. Results suggested excellent reliability (Table 2).

Additionally, to overall reliability, factor-specific reliability indices were examined to evaluate the internal consistency of each dimension of the Teaching Confidence Scale. The Mathematics/Science Confidence (F1) and Instructional Innovation (F2) factors demonstrated excellent reliability across all indices (α and ω > 0.95; AVE > 0.75), indicating highly consistent measurement of domain-specific teaching confidence. The Classroom Management factor (F3) exhibited slightly lower, yet still acceptable, reliability (α = 0.83; ω = 0.92; AVE = 0.50). This pattern suggests adequate internal consistency while reflecting the broader and more heterogeneous nature of classroom management confidence, particularly within inclusive settings.

Item-level analyses were conducted using the Generalized Partial Credit Model (GPCM). The results indicated satisfactory item fit across all three factors, supporting the adequacy of the model for analyzing item difficulty and discrimination parameters within the Teaching Confidence scale. Results for S_x², degree of freedom, and RMSEA are presented in Table 3.

All TC items demonstrated excellent fit (Table 3), since the RMSEA values were formed bellow the cut-off value (0.05). The item discrimination parameter presented with a and the item difficulty parameters b_i, with 1 ≤ i ≤ 5 and the standard errors of the items matched with the model fit as per GPCM are presented in Table 4.

Item Characteristic Curves (ICCs) illustrate the probability of endorsing each response category as a function of the latent trait (θ), thereby providing insight into how well an item discriminates across different levels of teaching confidence. Steeper curves indicate higher discrimination, whereas flatter curves suggest weaker differentiation among respondents. Figure 2 presents the item characteristic curves for each item and factor. Results suggested that all items were discriminative.

For illustrative purposes, selected items from the Mathematics/Science and Instructional Innovation factors are briefly examined. The ICCs for these items demonstrate clearly ordered response categories and steep slopes, indicating strong discrimination across low-to-moderate levels of teaching confidence (Figure 3).

Correspondingly, their IIFs display pronounced information peaks within these regions, suggesting that these items are especially effective in differentiating teachers with developing confidence. In contrast, items within the Classroom Management factor tend to show broader and lower information curves, reflecting more diffuse measurement precision across the latent trait continuum. This pattern aligns with the test-level information functions and suggests that classroom management confidence is distributed more evenly across confidence levels, albeit with lower overall precision compared to the other factors (Figure 4).

We assessed measurement invariance across teachers’ current role in education with the use of CFA. The CFA fit indices demonstrated excellent fit for the two groups of participants (Table 5; Figure 5).

We also evaluated goodness-of-fit measures for the nested TC models to validate full measurement invariance across the teacher’s current role in education. All models (Configural, Metric, and Scalar) apart from strict demonstrated good fit. Although, the decrease in CFI at the scalar invariance level exceeded commonly suggested cutoffs, partial scalar invariance was considered acceptable given the stability of factor loadings and the consistency of item functioning across groups. Therefore, the results support configural and metric invariance, with partial support for scalar invariance (Table 6).

Overall, the IRT analyses provided complementary evidence at both the item and test levels, including item fit, discrimination, difficulty parameters, and measurement precision across the latent trait continuum.

Consistent with theoretical expectations, confirmatory factor analysis (CFA) supported a three-factor model representing teachers’ confidence in (a) mathematics and science instruction, (b) instructional innovation, and (c) classroom management. The model achieved excellent fit indices (RMSEA = 0.027, CFI = 0.994, TLI = 0.956), validating the multidimensional conceptualization proposed by Skaalvik and Skaalvik (2007) and further supported by subsequent research on domain-specific teacher efficacy (Klassen and Tze, 2014; Savolainen et al., 2020). These results align with earlier validation studies indicating that teaching confidence is best conceptualized as a set of related but distinct domains rather than a unidimensional construct (Pietsch et al., 2003; Auslander et al., 2006).

The strong loadings observed for the Mathematics/Science Instruction factor highlight the centrality of subject-specific confidence in predicting instructional quality, persistence, and adaptive teaching strategies in inclusive classrooms (Hodgen et al., 2018; Lazarides et al., 2020). Similarly, the robustness of the Instructional Innovation factor is consistent with research identifying teachers’ confidence in adopting new technologies and differentiated pedagogies as a key driver of inclusive practice (Ertmer and Ottenbreit-Leftwich, 2010; Florian and Black-Hawkins, 2011).

By contrast, the Classroom Management dimension demonstrated slightly lower precision in the IRT analyses, although its internal consistency remained within acceptable to good levels (α = 0.83). This finding warrants careful contextual interpretation rather than being viewed as a psychometric limitation per se. In Greek inclusive classrooms, classroom management is often shaped by structural constraints such as large class sizes, limited availability of specialized behavioral support personnel, and centralized disciplinary policies that restrict teachers’ autonomy in implementing individualized behavioral interventions. Moreover, special education teachers frequently operate in co-teaching or parallel support arrangements, which may dilute perceived control over classroom routines and behavioral decision-making. These contextual characteristics may contribute to greater variability in responses and reduced measurement precision for management-related items.

This interpretation is consistent with prior research highlighting the influence of institutional culture, policy constraints, and role ambiguity on teachers’ perceived efficacy in classroom management (Finch et al., 2023; Koutrouba, 2013; Skiba and Peterson, 2000; Vidić, 2022). Within this context, lower precision may reflect genuine heterogeneity in teachers’ experiences rather than inadequacies of the construct itself.

The TCS demonstrated excellent internal consistency and model-based reliability, with Cronbach’s alpha and Bentler’s and McDonald’s Omega values ranging from 0.83 to 0.98, exceeding conventional benchmarks for research instruments (DeVellis, 2017). The convergence between classical reliability indices and IRT-based parameters provides strong evidence for the internal coherence and stability of the scale. In addition, Average Variance Extracted (AVE) values (0.50–0.77) supported satisfactory convergent validity across all domains.

IRT analyses using the Generalized Partial Credit Model (Muraki, 1992) further strengthened the validation evidence. All items exhibited excellent fit (RMSEA < 0.05), indicating appropriate functioning across the continuum of teaching confidence. Discrimination parameters were particularly strong for items assessing mathematics/science instruction and instructional innovation, suggesting that these items are especially sensitive to differences in teachers’ confidence levels, likely reflecting the influence of targeted training and pedagogical support in these areas (Boone et al., 2014; Kim et al., 2013; Reise et al., 2025).

Information function analyses indicated that the TCS provides the greatest measurement precision at low-to-moderate levels of teaching confidence. From a practical standpoint, this pattern is advantageous for professional development contexts, as it enhances the instrument’s capacity to identify teachers who may benefit most from targeted interventions. At the same time, the relatively lower information provided by certain classroom management items at higher confidence levels suggests opportunities for future item refinement. Specifically, adding items that capture advanced or context-sensitive management strategies such as collaborative behavior planning, co-teaching coordination, or inclusive restorative practices may enhance precision at the upper end of the confidence continuum (Linacre, 2002; Thomas, 2011).

Measurement invariance analyses supported equivalence of the TCS across professional status groups (substitute versus certified special education teachers) up to the scalar level, indicating that the scale measures teaching confidence comparably across these subgroups (Putnick and Bornstein, 2016). The observed decline in fit at the strict invariance level is consistent with findings from previous validation studies and is commonly attributed to differences in response consistency rather than construct non-equivalence (Kim et al., 2013; Savolainen et al., 2020).

Overall, the findings corroborate international research conceptualizing teaching confidence as a multifaceted construct shaped by instructional, emotional, and contextual dimensions (Bandura, 1997; Florian, 2019; Schillinger et al., 2021). The three-factor structure aligns with evidence that confidence in STEM-related instruction, instructional innovation, and classroom management represents distinct yet interrelated domains influencing inclusive practice (Florian and Black-Hawkins, 2011; Savolainen et al., 2020).

Notably, the comparatively lower precision observed for classroom management confidence is consistent with international findings identifying behavior management as one of the most demanding and stress-inducing aspects of inclusive teaching (Aloe et al., 2014; Lehikoinen et al., 2025; Park et al., 2025; Poulou et al., 2018; Savolainen et al., 2022). In contrast, the strong performance of instructional innovation items may reflect increased exposure of Greek teachers to digital tools and inclusive pedagogical frameworks, such as Universal Design for Learning, through recent professional development initiatives (CAST, 2024; Organisation for Economic Co-operation and Development, 2019). These patterns highlight both shared international trends and context-specific challenges, reinforcing the importance of culturally and institutionally sensitive validation studies.

These findings hold several theoretical, methodological, and practical implications for teacher education and inclusive schooling. From a methodological perspective, the study highlights the potential of IRT as a powerful framework for educational scale validation. By modeling latent constructs and item-level precision, IRT enables researchers to move beyond global reliability indices and examine how specific items function across varying levels of teaching confidence (DeMars, 2010; Embretson and Reise, 2000). The use of the GPCM (Muraki, 1992) provided robust evidence that the TCS yields consistent, interpretable, and fair measurement across subgroups of educators. This approach enhances the psychometric rigor of teaching confidence research, supporting cross-cultural comparisons, longitudinal tracking, and program evaluation (Kim et al., 2013; Reise and Revicki, 2015). The integration of IRT methods into teacher education research thus promotes a deeper, data-driven understanding of professional growth and efficacy.

From a practical and applied standpoint, the validated TCS serves as a diagnostic and developmental tool that can inform both individual and systemic improvement. For teacher educators and policymakers, the identification of domain-specific confidence profiles in mathematics/science instruction, instructional innovation, and classroom management offers targeted entry points for professional learning design. For instance, teachers with lower classroom management confidence might benefit from evidence-based behavior support frameworks or peer coaching models (Kunter et al., 2013; Poulou et al., 2018), whereas those with weaker innovation confidence could engage in collaborative digital pedagogy workshops that build competence in UDL and technology-enhanced teaching (Appleton, 2008; CAST, 2024; Florian and Black-Hawkins, 2011; Peng et al., 2024).

At a school and policy level, the TCS can support data-informed leadership decisions, guiding resource allocation, mentoring structures, and evaluation of training effectiveness (Organisation for Economic Co-operation and Development, 2019; Shields and Hesbol, 2019). By integrating confidence assessment into professional development cycles, schools can track progress, identify collective needs, and promote reflective, evidence-based practice. Furthermore, longitudinal use of the scale could facilitate monitoring of how teaching confidence evolves in response to curriculum reform, inclusion policies, or technological change (Florian, 2019; Savolainen et al., 2020).

Finally, for research, the availability of a psychometrically validated, multidimensional scale opens avenues for comparative and predictive studies examining the interplay between teaching confidence, pedagogical quality, and student outcomes. Future investigations could explore how variations in teaching confidence influence inclusive classroom dynamics, student engagement, and learning achievement (Lazarides et al., 2020; Zee and Koomen, 2016). The scale thus provides both a measurement innovation and a practical instrument that bridges the gap between theory, teacher development, and educational equity.

While the findings of this study provide robust evidence for the multidimensionality and psychometric soundness of the TCS, several limitations must be acknowledged to guide interpretation and future research.

First, the composition of the sample poses certain constraints on the generalizability of the findings. Participants were predominantly female and employed in primary special education, reflecting the demographic reality of the Greek teaching workforce but limiting extrapolation to male teachers, general education contexts, or other educational levels. The reliance on a convenience sample of 305 primary special education teachers may also restrict the representativeness of the national teaching population. Future research should therefore validate the scale with more gender-balanced and diverse samples, including secondary and mainstream educators, to examine its applicability across educational settings. Expanding the research to more heterogeneous and cross-cultural contexts would further allow exploration of whether the structure and functioning of teaching confidence differ across educational systems with distinct inclusion policies (Klassen and Chiu, 2010; Putnick and Bornstein, 2016; Florian, 2019; Savolainen et al., 2020). Nevertheless, the present study provides an important first step toward understanding the multidimensional nature of teaching confidence in inclusive settings.

Second, the study relied exclusively on self-reported measures, which, although common in teacher efficacy research, may introduce response bias, including social desirability and overestimation of competence (Wyatt, 2018; Zee and Koomen, 2016). Combining self-report data with external or performance-based measures such as peer observations, classroom behavior assessments, or student feedback would allow for triangulation and stronger construct validation (Klassen and Tze, 2014; Tschannen-Moran and Hoy, 2001).

Third, the cross-sectional design limits the ability to infer developmental changes or causal relationships. Longitudinal research could explore how teaching confidence evolves over time in response to professional experience, mentoring, or policy reforms (Day et al., 2016; Skaalvik and Skaalvik, 2007). Moreover, examining the scale’s predictive validity in relation to objective indicators of teaching quality, classroom climate, and student learning outcomes would clarify its practical utility and ecological validity (Lazarides et al., 2020; Zee and Koomen, 2020).

Finally, qualitative and mixed-methods approaches including teacher interviews, focus groups, or classroom ethnographies could enrich understanding of how confidence is enacted in everyday teaching and how it interacts with contextual variables such as leadership support, collaboration, or school culture (Florian and Spratt, 2013; Tsouloupas et al., 2010). Such insights could deepen theoretical understanding of the mechanisms linking confidence with professional practice and emotional well-being.

In summary, future research should pursue a multi-method, multi-level agenda, integrating longitudinal, experimental, and qualitative designs to capture the dynamic, contextual, and relational nature of teaching confidence. This approach would extend the present study’s contribution, moving from psychometric validation toward a more comprehensive framework linking teacher beliefs, inclusive pedagogy, and educational outcomes.