Effects of Computer Programming on Cognitive Outcomes: A Meta-Analysis

65 studies, 432 comparisons

School-age students, largely US, in the Logo and BASIC era.

A pre-PRISMA meta-analysis pooling experimental, quasi-experimental AND pre-experimental designs against control groups, with no risk-of-bias appraisal, no publication-bias test, no active-versus-passive control moderator, no distinction between researcher-made and standardized measures, and no near/far transfer split. Effect sizes are study-weighted means of raw comparisons. Read at abstract depth. Recorded because it is the Logo era's headline number and because of what its moderators say.

Grand mean study-weighted effect size 0.41 across 432 comparisons, with 58 of 65 studies (89%) favouring the programming group, and the authors concluding that the outcomes of learning a computer language go beyond the content of that specific language. Four of seven coded moderators were significant, and the important one is PUBLICATION TYPE. This is the earliest documentation that the field's effect size depends on where the study was published — the same moderator Scherer et al. rediscovered 28 years later.

Medium-to-high. Pools pre-experimental designs with no control for selection into treatment.

Updated by Liao (2000); its two significant artefact moderators — publication type and treatment duration — reappear verbatim in Scherer et al. (2019). Twenty-eight years apart, the same two variables explain the effect.