Effects of Inquiry-Based Approaches on Students' Higher-Order Thinking Skills in Science: A Meta-Analysis

26 effect sizes from 20 studies; 1,349 students

School and (some) tertiary science classrooms, mostly from Turkey, Indonesia, the Philippines and similar settings; PRISMA-screened studies comparing inquiry-based with "conventional" instruction.

Included because it is what a founder will find first if they search for "does inquiry teach thinking", and because it is a textbook case of the red flag METHODOLOGY names: a very large effect from a handful of very small studies with treatment-aligned researcher-designed instruments. Twenty studies and 1,349 students means a median study of ~67 pupils. Effect sizes range from g = -0.37 to g = 3.20 — a single study at three standard deviations should end an analysis, not feature in it. I² = 85.4%. The funnel plot is visibly asymmetric and the authors dismiss publication bias on a non-significant Begg-Mazumdar rank correlation (Kendall's tau 0.24) and a fail-safe N, neither of which is adequate with k = 26 and heterogeneity this high. Outcomes are post-test means on "higher-order thinking skills" instruments supplied by the primary authors; no independent standardised measure appears anywhere. No moderator distinguishes guided from open inquiry in a way that would let a reader separate the treatments, and no outcome leaves the taught unit.

Overall weighted g = 0.893 (95% CI 0.622 to 1.164) for inquiry-based approaches on higher-order thinking skills, described as significant regardless of educational level, discipline or level of inquiry. Taken at face value this would be the largest far-transfer effect in education; taken at design value it is the expected output of 20 small studies with aligned, self-designed outcome measures and an asymmetric funnel plot. Set against the one well-powered randomised trial of a reasoning-focused science programme with independent standardised outcomes (EEF 2016: g = -0.01 to -0.15), the contrast is the whole argument of this topic.

Medium: mixed experimental and quasi-experimental primaries with self-selected classes, but measure alignment and small-study bias dominate.

The pooled estimate has not been reproduced by any randomised trial using an independent outcome measure. The nearest well-powered independent tests (EEF Let's Think Secondary Science 2016; EEF Philosophy for Children 2021) find nothing.