Manual-supported critical-thinking activities improve preschool reasoning performance: a multisite cluster-randomized study in Kazakhstan

Introduction Structured opportunities to compare, predict, justify, and revise responses may support preschool reasoning, but evidence from multisite cluster-randomized evaluations using active pedagogical comparison conditions remains limited. Methods This study evaluated a 16-week manual-supported critical-thinking programme involving 604 children aged 60–71 months from 36 classrooms in 10 preschool institutions in Kazakhstan. Eighteen classrooms received 32 educator-led sessions targeting comparison, classification, prediction, evidence-based explanation, alternative generation, and answer revision; 18 classrooms received an active pedagogical comparison programme with the same scheduled duration and frequency. Reasoning performance was assessed at baseline, immediate post-test, and three-month follow-up using three counterbalanced forms of 12 unfamiliar tasks scored from 0 to 24. Linear mixed-effects models accounted for classroom clustering, repeated child-level observations, institution, age, sex, and instructional language. Results Immediate post-test data were available for 559 children and follow-up data for 545. The intervention produced greater baseline-to-post-test improvement than the active comparison condition, with an adjusted difference of 1.15 points (95% CI 0.77–1.53, p < .001) and a standardized effect of 0.38. At three-month follow-up, the adjusted difference was 0.61 points (95% CI 0.20–1.02, p = .004), corresponding to a standardized effect of 0.20. Multiple-imputation analyses yielded comparable estimates. Discussion The programme produced a modest-to-moderate immediate advantage in directly assessed preschool reasoning performance, with a smaller but detectable advantage three months after programme completion. The findings indicate partial short-term maintenance and should be interpreted within the study-specific criterion-referenced outcome and the targeted reasoning domains.