Refutation texts and argumentation for conceptual change: A winning or a redundant combination?

100 undergraduates in three conditions

Undergraduates; topic biological evolution (a robust, complex misconception cluster: essentialist/transformational conceptions of species change).

Three-arm randomised experiment - (1) refutation text followed by peer argumentation over erroneous worked-out solutions, (2) refutation text followed by individual problem solving, (3) control expository text followed by individual problem solving. The design is the useful one because it puts the refutation text against an ACTIVE alternative text and then asks whether the field's other favourite conceptual-change technology (argumentation/socio-cognitive conflict) adds anything on top. Dose is a single session. Outcome is a researcher-built evolution conception measure. Read at abstract depth; effect sizes not obtained.

Strong effects for the refutation text over the expository control on conceptual change in evolution. Subsequent PEER ARGUMENTATION DID NOT further improve learning gains beyond the refutation text alone - the two conceptual-change technologies were redundant, not additive, in this design. Dialogue-protocol analysis showed that the dyads that did gain were more likely to be symmetrical and to discuss core conceptual principles, which is a mechanism finding rather than an effectiveness one. For a founder deciding where to spend instructional time this is the useful result: adding structured peer argument to a text that already refutes the misconception bought nothing, so the effective ingredient is the refutation itself, not the social process the constructivist literature tends to sell alongside it. Note the ceiling on the claim - one session, one topic, undergraduates, no delayed post-test reported.

Low - randomised assignment to conditions.

Single study, not replicated. Its null on argumentation-on-top-of-refutation is consistent with the broader pattern in Pacaci et al. 2024, where strategy type (cognitive conflict vs bridging vs ontological shift) did not moderate the effect at all (Q = 0.95, p = 0.62).