REFUTATION TEXT IN SCIENCE EDUCATION: A REVIEW OF TWO DECADES OF RESEARCH

Narrative review of refutation-text studies published from the mid-1980s to 2008 (study count not obtained)

Readers of refutation text on science topics; the reviewed corpus spans grade 3 through undergraduate and preservice teachers, with undergraduates dominant.

Narrative (non-quantitative) review, so grade D by construction - it aggregates without pooling and inherits every weakness of its primaries. Its value to this archive is as the field's own account of what refutation text is and what its evidence base actually contains. Tippett's structural definition is the one later work uses: a refutation text must state a commonly held misconception, signal that it is wrong (the "cue"), and explain the correct conception. We read the abstract and secondary summaries, not the full text.

Concludes that reading refutation text rather than plain expository text is more likely to produce conceptual change - the same direction as Guzzetti et al. 1993 and Schroeder & Kucera 2022. The durability picture reported through this review and its downstream summaries is the honest one for this literature and is worth stating precisely: fewer than half the refutation-text studies used a delayed post-test at all, the delays used were short, and where delayed tests exist the refutation advantage is usually still present but with MORE readers reverting to the misconception at delay than at immediate test. Named delayed-effect studies in the corpus: Tippett 2004 (grades 3-4, conceptual change still present at 6 weeks), Maria & Johnson 1989 (gifted grades 5 and 7, effects at 1 month), Mason & Gava 2007 (grade 8, maintained at 2 months). Nothing in the corpus tests a 6-month or longer horizon. The review also makes clear that the outcome instruments are researcher-built misconception measures throughout; no standardized science achievement test appears.

Not applicable - narrative review of instructional contrasts.

Superseded quantitatively by Schroeder & Kucera 2022, which reaches the same conclusion with pooled estimates and moderators. Retained because it is the field's inventory of which studies used delayed post-tests and at what delays, which the meta-analysis compresses into four bins.