Forms of inquiry-based science instruction and their relations with learning outcomes: evidence from high and low-performing education systems

151,721 students in 5,089 schools across the 20 highest- and lowest-performing science regions of PISA 2015

15-year-olds in the 10 highest- and 10 lowest-scoring science systems of PISA 2015 (including Finland, Singapore, Estonia, Canada, Taipei, Macao, Vietnam, Hong Kong, Beijing, Japan; and Algeria, Kosovo, Lebanon, Tunisia, Brazil, Dominican Republic, Indonesia, Jordan, Peru, Macedonia).

Cross-sectional PISA data — grade D, no causal identification. But it is the single most decision-relevant observational study in the science tranche because it does the thing the archive keeps asking for: it SPLITS THE INQUIRY CONSTRUCT instead of pooling it. Multi-group confirmatory factor analysis found that measurement invariance CANNOT be established for the nine PISA inquiry items across regions, which by itself invalidates the standard cross-country "inquiry is negatively associated with achievement" headline. Exploratory factor analysis then recovered, in most regions, a two-factor structure: INDEPENDENT INQUIRY (students design their own experiments, do investigations to test ideas, debate investigations, choose their own procedures) versus GUIDED INQUIRY (the same inquiry activities combined with teacher explanations of how science ideas apply to the world). Both factors are student self-report of classroom frequency, and the outcome is the PISA science scale. The authors caution that the strength of the guided-inquiry association is larger in regions where the guided factor is measured with fewer student-centred items, so the contrast is partly a measurement gradient.

Guided inquiry was POSITIVELY associated with science achievement in all 16 regions where that factor appeared; independent inquiry was NEGATIVELY associated in 18 of 20 regions (achievement), 17 of 20 (enjoyment) and 17 of 20 (epistemic beliefs). The standardised coefficients are large and consistent: guided inquiry +0.507 Finland, +0.495 Singapore, +0.408 Vietnam, +0.326 Hong Kong, +0.321 Canada, +0.222 Brazil, +0.149 Dominican Republic; independent inquiry -0.627 Finland, -0.546 Singapore, -0.482 Canada, -0.492 Hong Kong, -0.364 Jordan, and as low as -1.111 Tunisia. Japan was the sole exception, showing a form of independent inquiry positively associated with achievement. The authors' conclusion is the archive's guidance-dose finding restated for science and at scale: "it would be misguided to use PISA findings to support arguments to scale back inquiry" — the negative PISA correlation is an artefact of pooling unguided with guided inquiry into one score.

High for causal claims: cross-sectional student self-report with no random assignment, and reverse causation is plausible in both directions (weaker classes get more unguided activity; stronger teachers give more conceptual guidance). The value here is that the SIGN FLIPS WITHIN the same confounded dataset when guidance is separated, which no confound story explains neatly.

Directly reconciles the archive's existing PISA sources (OECD 2015; Jerrim 2019) with the experimental guidance literature (Furtak 2012 teacher-led vs student-led; Lazonder and Harmsen 2016; Alfieri 2011). It drew a published response — Zhang, "Confusions on 'Guidance' in Inquiry-Based Science Teaching" (2020, 10.1007/s42330-020-00116-4) — arguing the guidance construct is being used loosely.