The effects of programming interventions in early childhood: A meta-analysis

25 studies, 82 effect sizes

Children aged 3-7.

Pools programming knowledge, computational thinking, mathematics knowledge and cognitive skills into a SINGLE d, which mixes outcome classes this archive deliberately keeps apart — so the headline number cannot be read as an estimate of anything in particular. Read at secondary depth via a publication summary, so the moderator structure, control-type coding and publication-bias tests are unchecked, which is exactly what would determine its grade. Shares an author with Scherer et al. (2020), so it is not a wholly independent estimate of the same literature.

Overall d = +0.73 (95% CI 0.51-0.96) across 25 early-childhood studies and 82 effect sizes. Unplugged activities outperformed screen-based ones and robotics outperformed other programme outputs. A pooled d of 0.73 in a corpus of 25 early-childhood studies is implausibly large by the archive's benchmarks, and the finding that UNPLUGGED activities beat actual programming is the most awkward result in the subject for any mechanism that runs through learning to program a computer.

Medium. Mixed designs across a small early-childhood corpus.

Consistent in magnitude with Scherer et al. (2020) and Hu (2024), but shares an author with the former, and the outcome construct differs from both.