Impact of programming on primary mathematics learning

Euclidean division N = 1,880; additive decomposition N = 1,763; fractions N = 644; 46 schools with classes randomly assigned

French grade 4 and 5 primary pupils.

A class-level cluster randomised trial across 46 schools comparing Scratch-based programming activities against traditional mathematics activities on the SAME three mathematical notions — an active, content-matched comparison, which is the hardest and most policy-relevant test in this literature and the one the transfer claim most needs to pass. Multilevel analysis. This is the largest randomised test of programming against academic achievement located anywhere in the corpus. Read at secondary depth: the publisher blocked retrieval, so the effect sizes come from the published abstract plus a verbatim description in a later trial report, and should be verified against the paper before this source is marked verified.

Programming did not help mathematics; it hurt it. Significant NEGATIVE effects on all three notions: -0.16 for Euclidean division, -0.19 for additive decomposition, -0.21 for fractions. The authors conclude that transfer between programming and mathematics may not occur automatically. Read against the meta-analytic far-transfer-to-mathematics estimate of g = +0.57, this trial is a direct contradiction under a properly active baseline at scale — and it is the coding analogue of the maths deficit the archive already records for working-memory training.

None material. Class-level randomisation across 46 schools.

No direct replication located. Its direction is consistent with the two-year mathematics deficit found in the Cogmed school trial already held in the archive's brain-training file — both are opportunity-cost findings, and both concern hours taken out of mathematics.