Syntes: Kim & Cao (2025) — knowledge goes with comprehension (r = .41); teaching it lifts reading comprehension, least on standardised tests and for knowledge building alone

Källa: Kim, Y.-S. G. & Cao, Y. (2025). Content knowledge and comprehension: A meta-analytic review of correlational and causal associations. Psychological Bulletin 151(10), 1219–1244. https://doi.org/10.1037/bul0000502

Verified content

• [full text] Correlational data (108 studies, N = 68,301): content knowledge and comprehension r = .41, the same for reading and listening comprehension; stronger with norm-referenced knowledge measures (r = .50) than researcher-made ones (r = .39).
• [full text] Causal data (55 studies, N = 18,540): content knowledge instruction raised content knowledge (g = 1.36, 95% CI 0.43–2.29; 18 studies) and reading comprehension (g = 0.44, 95% CI 0.18–0.70; 51 studies; I² = 97.75%); listening comprehension g = 0.13 (95% CI −0.41 to 0.68; 5 studies), not significant.
• [full text] Moderators for reading comprehension: norm-referenced tests g = 0.21 vs researcher-developed g = 0.51 (difference p < .01); knowledge activation g = 0.66 (p < .01) vs knowledge building g = 0.19 (p > .05, not significant; difference p = .045); studies with two or more groups per condition g = 0.18 (p < .01) vs single-class (N of 1) designs g = 0.67. Treatment length, comparison condition, grade, reading skill, language-learner status and SES did not moderate.
• [full text] About 52% of causal studies had an N of 1 design; mean instruction time 671 minutes (5–3,105). Egger's test showed funnel asymmetry for the reading-comprehension treatment effects, but trim-and-fill did not support publication bias.

Limits

• [assessment] Moderator estimates come from meta-regression with very high heterogeneity; "not significant" for knowledge building is not evidence of no effect. Tables are images and were not read; numbers are from the running text.
• [assessment] Tested C17's kill line on 2026-11-04: the pooled causal evidence does not show that knowledge-building teaching fails to improve comprehension, but its average for knowledge building alone is not statistically secure (C17 weakened).

Why it is in the library

The largest meta-analysis of both correlational and causal links between topic knowledge and comprehension; it separates building new knowledge from activating knowledge pupils already have.

Added 2026-11-04 by the project (nightly attacks; read: full text). [abstract] = from the published abstract; [full text] = read in the full text; [assessment] = the project's own assessment.