Time to transfer: Long-term effects of a sustained and spiraled content literacy intervention in the elementary grades

30 elementary schools, 2,870 students, randomised

US elementary students followed from grade 1 to grade 4; high-poverty urban district, ~40% low-income, majority Black and Hispanic

School-randomised longitudinal trial of the Model of Reading Engagement (MORE), the leading test of whether a SUSTAINED, cumulative content-knowledge sequence transfers to general reading. Treatment = thematic content literacy lessons in grades 1, 2 and 3 plus wide summer reading of thematically related informational texts; control = the same grade-3 lessons only (an ACTIVE, partial-spiral comparison, which makes the contrast a test of accumulation rather than of content instruction versus nothing). Content is science-dominant (animal survival in grade 1, scientific investigation of past events in grade 2, living systems in grade 3), with social studies content in the grade-1 units - so this is adjacent to, not squarely inside, a history curriculum question, and is recorded here for the transfer mechanism rather than for its subject matter. Read at abstract/secondary depth: effect sizes are from the published abstract and the authors' own reporting, so attrition, pandemic disruption (the trial was interrupted mid-way) and moderator detail have not been checked against the article.

The strongest existing evidence that knowledge-building transfers to general reading, and it is small. Intent-to-treat effects at grade 3: science reading 0.14, domain-GENERAL reading comprehension 0.11, mathematics achievement 0.12. Sustained at a 14-month follow-up: grade-4 reading comprehension 0.12, mathematics 0.16. Three things deserve weight. First, the comparison is an active control that received the same grade-3 lessons, so 0.11 is the marginal value of two extra years of cumulative content, not the value of content instruction as such. Second, the effect on MATHEMATICS is as large as the effect on reading, which a knowledge-transfer mechanism does not predict and which should temper any story about domain-specific schema; it is more consistent with a general engagement or school-climate channel, or with chance. Third, the magnitudes (0.11-0.16) sit right at the level the archive treats as meaningful-but- modest for a cheap scalable intervention, and are an order of magnitude below the proximal effects that content-rich curricula produce on their own aligned measures.

Low. School-level randomisation with longitudinal follow-up; treatment and control children have the same expected genetic endowment.

Unreplicated by any independent team. It is the longer follow-up of the same trial reported in Kim et al. (2023), so the two are not independent of each other. Converges directionally with the Grissmer Core Knowledge lottery (0.24 on state reading after years of exposure) and with Hwang et al. (2022) meta-analytic standardized comprehension estimate of 0.25.