Refutations in science texts lead to hypercorrection of misconceptions held with high confidence

114 eighth-grade adolescents, 16 science misconceptions each

Grade 8 students (age ~13-14) - one of the very few refutation-text experiments in this literature run on school-age children rather than undergraduates.

Within-topic experimental comparison of standard expository text against refutation text across 16 common science misconceptions. The design's distinctive feature, and the reason it earns a place here, is that it measures PRIOR CONFIDENCE: students answered pre-reading questions on each misconception and rated how confident they were in each answer, then read the texts, then took a post-test with both true/false and open-ended items. That lets the study ask not "does refutation text work" but "on whom" - which is the applicability question this archive exists to record. Outcome is a researcher-built misconception test; the post-test is immediate and no delayed test is reported, so the study says nothing about durability. Read at abstract depth.

Refutation text corrected HIGH-CONFIDENCE misconceptions more often than standard expository text did, while LOW-CONFIDENCE misconceptions did not benefit from refutation at all. This is the hypercorrection effect applied to instruction: being told you were wrong is informative in proportion to how sure you were, because the surprise recruits attention to the correction. Two practical consequences follow. FIRST, the refutation-text advantage is not uniform - it is concentrated on exactly the beliefs students hold firmly, which are also the ones ordinary teaching leaves untouched, so the technique is complementary to normal instruction rather than a substitute for it. SECOND, and less comfortably, it means the measured effect size of a refutation text depends on the prior-confidence distribution of the sample, which no meta-analysis in this literature codes; Schroeder & Kucera 2022 report prior knowledge as "not reported" for 37 of 44 comparisons.

Low - within-subject comparison of text formats on matched misconception items.

Single study; the underlying hypercorrection effect is well replicated in the memory literature, but its application to refutation text has not been independently repeated on children.