Promoting Conceptual Change in Science: A Comparative Meta-Analysis of Instructional Interventions from Reading Education and Science Education

70 experimental and quasi-experimental studies from reading education and science education (count as reported by later reviews; not verified against the paper)

School and college students taught science concepts on which they held documented misconceptions; US-dominated; grade range spans elementary through undergraduate.

The founding meta-analysis of this literature and the reason refutation text became a recommended practice. It pooled reading-education and science-education intervention studies on misconceptions and reported that the strategies with consistently high average effects were refutational text and augmented activation activities, with the common element that effective procedures produced COGNITIVE CONFLICT. Three problems this archive must record. (1) It aggregates across two literatures with different outcome traditions and different ages, and the pooled figure therefore hides both. (2) Its primaries almost universally used researcher-designed, misconception- targeted instruments - the strategy and the test were built by the same people from the same misconception list - so the measure-inflation discount applies at full strength; this is an inference from the later reviews of the same corpus (Tippett 2010, Pacaci 2024), not something we read in the paper. (3) Its search window closed around 1990, so it says nothing about durability beyond the short delays typical of that era. Recorded as unresolved because the full text sits behind JSTOR and we never obtained the effect-size tables; the ERIC abstract, which is all we read, reports no numbers.

Synthesised the experimental and quasi-experimental research from reading education and science education on instructional interventions targeting misconceptions. Reported that refutational text and augmented activation activities showed consistently high average effects, and that the effective procedures shared "a common element of producing conceptual conflict". This is the source of the received wisdom that refutation text is the best text-based tool for misconception change - and every later review of the field cites it rather than re-deriving it. We could not obtain the pooled effect sizes, so this archive holds the qualitative conclusion and none of the numbers. Later, larger, better-reported syntheses of the same intervention family (Schroeder & Kucera 2022 g = 0.41; Pacaci et al. 2024 g = 1.10 unadjusted, 0.93 bias-adjusted, 0.64 in randomised studies only) are the numbers this archive uses instead.

Not the operative confound. Outcomes are domain-knowledge items; the threats are measure alignment and publication bias.

The qualitative conclusion (refutation text beats plain expository text) has been reproduced by Tippett 2010 and by Schroeder & Kucera 2022 on largely non-overlapping corpora. The magnitude has not - the modern between-subjects estimate is roughly a third of a standard deviation, not the large effect the 1993 framing implied.