Syntes: Cheng, Wu, Wang & Wang (2026) — seductive details: a small but reliable average harm, transfer included

Källa: Cheng, C., Wu, Y., Wang, R., & Wang, Z. (2026). Seductive details, cognitive load, and learning outcomes: A multi-level meta-analysis and MASEM. Educational Psychology Review, 38, 28. https://doi.org/10.1007/s10648-025-10099-z

Verified numbers (full text read 2026-10-10)

• Scope: 51 studies retained, 9 outlying effect sizes from 5 studies removed, leaving 50 studies, 177 effect sizes, 20,683 participants (published 1989–2025); three-level model to handle several effect sizes per study.
• Overall (Table 3): g = −0.16 (SE 0.03, 95 % CI −0.22 to −0.10, p < .001); I² = 57.8 %.
• By outcome (Table 4): recall −0.17 (34 studies); comprehension −0.19 (11); transfer −0.12 (SE 0.05, 95 % CI −0.22 to −0.03, p < .05; 25 studies, 68 effect sizes).
• Moderators (Table 5, all outcomes combined): learning environment significant (F = 3.74, p = .033): technology-based −0.19 (41 studies), paper-and-pencil 0.00 (CI −0.16 to 0.17; 9 studies). Grade level not significant (F = 2.02, p = .11): secondary −0.05 (CI −0.19 to 0.10; 9 studies), college −0.18 (40). Sample size significant; text language significant (one Chinese study). Duration not significant (long interventions −0.12, ns).
• Mechanism: in 8 studies with cognitive-load data, the harm runs partly through higher extraneous load.
• Publication bias: Egger p = .98; trim-and-fill imputed no studies.

Role for C8

Supports (kill-condition test survived, 2026-10-10): even with dependence corrected and the evidence nearly doubled since 2020, transfer is still lowered on average (−0.12), so the meta-analytic kill-condition is not met. The average harm is about half the size of earlier estimates. The paper-versus-screen and secondary-pupil subgroups are between-study comparisons on combined outcomes, not tests on transfer, and were not applied as a narrowing.