The cognitive effects of computational thinking: A systematic review and meta-analytic study

Review of 19 studies with 1,527 participants; meta-analysis of 11 studies with 862 participants (433 experimental, 429 control)

Preschool to grade 10; studies published 2006-2022.

A multivariate FIXED-effect model with three imputed correlations and a multiverse sensitivity analysis. Grey literature was explicitly EXCLUDED, and the authors concede they were not able to robustly assess publication bias; Egger's test was run for one outcome only. Moderator analyses by age and intervention type were impossible for lack of studies. Only 4 of the 19 reviewed studies had an ACTIVE control group, and of those the two reporting the primary outcome were null. Independence is recorded as developer-involved because the three largest effects in the pool come from the meta-analysts' own laboratory.

The near/far gradient reappears cleanly. Problem solving d = 0.890 (95% CI 0.764-1.016) — but the review itself labels the largest problem-solving effects NEAR transfer, meaning coding problems. Planning 0.364 (0.222-0.505). Working memory 0.199 (0.045-0.353). Inhibition 0.168 (0.057-0.279). Cognitive flexibility 0.118 (-0.07 to 0.306), NOT significant. The authors themselves note that transfer appears to be typically narrow and more pronounced for abilities similar to those trained, and cite the same working-memory-training sources the archive uses to reach a no-effect verdict on brain training.

Low for the experimental core; the pooled studies are mostly randomised or quasi-randomised. The threat here is measure alignment and author-lab concentration, not heredity.

The three largest effects in the pool come from the meta-analysts' own lab and have no independent replication. The core executive-function findings sit alongside null executive-function results the same review tabulates.