Reframing the Reasoning Gap: Cognitive-Domain Achievement Profiles and Their Predictors Among Eighth-Grade Students in Türkiye Across TIMSS 2019 and 2023
INTERNATIONAL ELECTRONIC JOURNAL OF ELEMENTARY EDUCATION, vol.18, no.5, pp.801-813, 2026 (Scopus)
- Publication Type: Article / Article
- Volume: 18 Issue: 5
- Publication Date: 2026
- Doi Number: 10.26822/iejee.2026.467
- Journal Name: INTERNATIONAL ELECTRONIC JOURNAL OF ELEMENTARY EDUCATION
- Journal Indexes: Social Science Premium Collection (ProQuest), Education Collection (ProQuest), Education Source Ultimate (EBSCO), Scopus, Agricultural & Environmental Science Database, IBZ Online, Education Abstracts, ERIC (Education Resources Information Center), EBSCO Education Source
- Page Numbers: pp.801-813
- Anadolu University Affiliated: Yes
Abstract
Eighth-grade mathematics achievement in Türkiye is often framed around a presumed weakness in higher-order reasoning, yet the relative standing of the TIMSS cognitive domains has rarely been tested inferentially or across time. Using the nationally representative TIMSS 2019 (N = 4,077) and 2023 (N = 4,925) Grade 8 samples for Türkiye, we profiled achievement across the three cognitive domains, tested how the profile changed between cycles, and modelled each domain as a function of mathematics self-confidence, liking mathematics, home educational resources, and gender. All estimates used the five plausible values with jackknife (JK2) standard errors; domain contrasts are reported with
both between-student (d_av) and paired (d_z) effect sizes, and student-level results were corroborated with two-level (student-within-school) models. Contrary to the reasoning-
deficit narrative, reasoning was not the weakest domain in either cycle: in 2019 it exceeded applying and knowing, and by 2023 the domains converged, with knowing becoming
the relative low point. The 2019–2023 gain was concentrated in the applying domain (+18.4 points, p = .002). Across all domains and both cycles, self-confidence (β = .42–.45) and
home educational resources (β = .33–.41) were the dominant correlates, whereas liking mathematics contributed little once self-confidence was controlled (a suppression pattern), and gender differences were small. The predictor structure was near-identical across domains and stable across cycles; multilevel models further showed the home-resources association to be largely compositional (between schools). These findings reframe the Turkish “reasoning gap” and point to school composition and mathematics self-confidence as policy priorities.