Why Feeling Capable May Matter More Than Feeling Good About Math: Landmark Meta-Analysis Maps the Hidden Pathway to Achievement

For decades, educators have debated whether students’ attitudes toward mathematics—their likes, dislikes, and emotional associations with the subject—directly drive their grades, or whether something more subtle is at work. Now, a sweeping meta-analysis published in Educational Psychology Review suggests the answer is both: attitude matters, but much of its power flows through a psychological middleman—self-efficacy, the belief in one’s own capability to succeed. The findings, drawn from 76 primary studies spanning schoolchildren and university students across multiple countries, offer one of the most statistically rigorous portraits yet of how these motivational forces intertwine to shape mathematical performance.

The study, conducted by Ryan Angga Pratama and Sedat Kanadlı of Mersin University in Türkiye, employed Meta-Analytic Structural Equation Modeling, or MASEM, a sophisticated technique that allows researchers to fit a full structural equation model to the combined correlation matrices of many independent studies. Rather than simply pooling effect sizes as a traditional meta-analysis would, MASEM reconstructs the web of interrelationships among variables—attitude, self-efficacy, and achievement—as if thousands of participants had been studied within a single, unified model. The researchers searched Google Scholar, ERIC, ProQuest, and Semantic Scholar, and notably used AI-assisted research tools such as SciSpace and ScholarGPT to streamline the literature identification process, ultimately retaining 76 studies that met their inclusion criteria.

The headline results are striking in their symmetry. Students’ attitudes toward mathematics showed a moderate effect on their mathematics self-efficacy, with a standardized path coefficient of β = 0.517 and a 95% confidence interval of [0.446, 0.588]. In turn, self-efficacy exerted a moderate effect on mathematics achievement (β = 0.240, 95% CI [0.128, 0.352]). Crucially, attitude also maintained a direct, moderate effect on achievement (β = 0.235, 95% CI [0.121, 0.349]), meaning that even after accounting for the self-belief pathway, how students feel about mathematics retains its own influence on how well they perform. The indirect effect—the portion of attitude’s influence transmitted through self-efficacy—was statistically significant at β = 0.124, with a confidence interval of [0.065, 0.188] that excludes zero.

Together, the model explained 27% of the variance in students’ self-efficacy and 17% of the variance in mathematics achievement. Those figures, while respectable for research in educational psychology, also underscore how much remains unexplained. As the authors emphasize, substantial residual variance remains, pointing to the many other variables—prior achievement, instructional quality, anxiety, family background, and cultural context—that operate across different educational environments. The researchers are careful to frame the results as connections rather than proof of cause and effect: MASEM reveals patterns of association across correlational studies, not experimental causation.

The moderation analyses added important nuance. The links in the model were significantly stronger at the university level, with a statistically significant moderator test (χ²(3) = 7.909, p = .048). This suggests that the attitude–self-efficacy–achievement chain may operate differently across educational stages, with adult learners perhaps more attuned to their own competence beliefs when tackling demanding coursework such as calculus or statistics. The measurement of self-efficacy itself also mattered: studies using the well-validated Usher and Pajares (2009) scale of mathematics self-efficacy sources produced significantly different results (χ²(3) = 9.134, p = .028) than studies employing other instruments. Other moderators tested did not reach significance, indicating a degree of robustness across the remaining study characteristics.

The theoretical scaffolding behind the study traces back to Albert Bandura’s foundational work on self-efficacy, first articulated in 1977 and elaborated in his 1997 treatise on the exercise of control. Bandura argued that beliefs about one’s capabilities govern how much effort people invest, how persistent they are in the face of obstacles, and how resilient they remain after setbacks. In mathematics—where failure experiences accumulate quickly and confidence can erode rapidly—self-efficacy has long been considered a linchpin of achievement. Meanwhile, research on mathematical attitudes stretching back to the Fennema-Sherman scales of the 1970s has documented persistent associations between affect and performance, though the causal ordering has always been contested. Prior meta-analytic work, notably Ma and Kishor’s 1997 review in the same journal, found comparatively weak attitude-achievement links, making the present findings a meaningful update to the field’s conventional wisdom.

The methodological machinery deserves attention for readers wondering how such claims can be trusted. The authors followed PRISMA reporting standards, assessed publication bias using established techniques such as funnel-plot-based methods and rank correlation tests, and conducted their analyses in the R statistical environment using the meta package, alongside a dedicated MASEM web application for one-stage modeling. One-stage MASEM, championed in tutorials by Suzanne Jak and colleagues, fits the structural model directly to the raw correlation data while weighting studies by precision, avoiding some of the two-stage estimation pitfalls that plagued earlier approaches. The inclusion of grey literature and the AI-assisted search strategy also helped mitigate the file-drawer problem, in which unpublished null results distort the meta-analytic record.

What, then, should educators take away? The findings suggest that interventions aimed solely at improving students’ feelings about mathematics—rapport-building, positive messaging, enthusiasm campaigns—may yield only partial returns if they fail to build genuine competence beliefs. Conversely, mastery experiences that produce authentic success, which Bandura identified as the most powerful source of self-efficacy, may simultaneously improve both confidence and attitude, creating a mutually reinforcing loop. The moderate direct effect of attitude on achievement also implies that affective engagement has value in its own right, perhaps through increased engagement, persistence, and willingness to take on challenging problems.

The authors caution against overinterpreting the magnitudes. With 17% of achievement variance explained, the model describes real but far from dominant forces. Mathematics achievement is multiply determined, and psychological factors such as attitude and self-efficacy operate within a dense ecosystem of instructional, socioeconomic, and cognitive influences. The study’s cross-sectional foundation—nearly all primary studies measured variables at a single time point—means the direction of the arrows, however theoretically motivated, cannot be definitively confirmed. Reciprocal relationships, in which success builds confidence and confidence fuels further success, remain plausible and have been documented in longitudinal work.

Still, in synthesizing 76 studies into a single coherent structural model, the research delivers a clearer map than the field has previously possessed. For teachers, parents, and policymakers wrestling with widespread mathematics anxiety and underachievement, the message is refreshingly concrete: helping students believe they can do mathematics is not merely a feel-good accessory to instruction—it is a measurable, statistically reliable conduit through which positive attitudes translate into better grades. The psychological architecture of mathematical success, it turns out, runs through self-belief.