Impact of a content-rich literacy curriculum on kindergarteners' vocabulary, listening comprehension, and content knowledge

Two RCTs combined; 47 schools in two large urban US districts, 1,194 kindergarten students, ~80 teachers

Kindergarteners in large urban US districts, mixed SES, substantial share English language learners

The most direct causal test yet of E. D. Hirsch-lineage content-rich curriculum, and the single most informative source in this cluster because it measured the SAME construct with an aligned and an independent instrument at once. Schools (not classes) randomly assigned to implement the Core Knowledge Language Arts KNOWLEDGE STRAND - systematic interactive read-alouds building science and social studies domain knowledge, 45-60 min daily - or to a waitlist control; teachers taught on average 71.8 of an expected 76-77 lessons over roughly one semester, after two days of PD plus light-touch facilitator support. Study 2 is a conceptual replication of Study 1 in a different district; combined analysis reported, with separate results in appendices. Design meets WWC Standards v4.0 without reservations. Cost reported as USD 316 per student. Evaluators are university reading researchers, not the publisher. Crucially, EVERY domain was measured twice: a proximal, researcher-developed measure aligned to what the curriculum taught, and an independent standardized measure of the same construct (PPVT and Woodcock-Johnson Picture Vocabulary for vocabulary; CELF Sentence Structure and Test of Narrative Language for listening comprehension; Woodcock-Johnson Science and Social Studies subtests for content knowledge).

Content-rich instruction taught the content it taught, and nothing beyond it. Proximal social studies knowledge (curriculum-specific Native Americans test) ES = 0.93 (p < .001) - one of the largest effects in this archive. The INDEPENDENT standardized measure of the same construct, the Woodcock-Johnson Social Studies subtest, moved by ES = 0.00 (p = .999). Exactly zero. Same story elsewhere: taught vocabulary 0.63 vs standardized receptive vocabulary null; taught science knowledge 0.26 vs standardized science 0.11 (p = .054, ns). The one apparently significant standardized effect - WJ Picture Vocabulary, ES = 0.09 - vanished (ES = 0.01, p = .702) when the authors removed three items whose words (canoe, moccasins, loom) were explicitly taught by the curriculum. Standardized listening comprehension was flat or negative: CELF Sentence Structure -0.05, Test of Narrative Language +0.02, both ns. Second finding, equally important: the benefit was LARGER for children who arrived with stronger vocabulary. On the proximal social studies measure the effect ran 0.53 at 1 SD below the mean PPVT, 0.92 at the mean and 1.31 at 1 SD above - a within-trial Matthew effect, and the same asymmetry appeared for English learners (0.58) versus native speakers (0.99). The authors' own conclusion is that the field needs "rethinking the theory of change underlying Tier 1, whole class language-focused interventions." Limitation the authors flag themselves: several treatment teachers reported dropping other content-area instruction, so part of the contrast is substitution rather than addition.

Low. School-level randomisation across two independent trials, with pretest covariates; gene-environment correlation cannot generate the treatment-control difference. The measure-alignment problem, not selection, is what limits interpretation.

Internally replicated: Study 2 was a conceptual replication of Study 1 in a different urban district, and the authors state the Study 1 findings were reproduced. This is the strongest replication status in the cluster and it is a replication of a NULL on independent measures alongside a replication of a large positive on aligned ones. It also converges with Hwang, Cabell & Joyner (2022), whose meta-analysis found standardized content-knowledge effects unestablished on six effect sizes, and it is in obvious tension with the Grissmer Core Knowledge lottery's 0.24 SD on state reading tests - a tension best resolved by dose and duration (one semester of read-alouds versus six years of whole-school schooling) rather than by discarding either result.