Effects of computer programming on young children's cognition

18 children in total, two arms of nine

Six-year-olds in the US.

A twelve-week comparison of Logo programming against computer-assisted instruction — a genuine ACTIVE control, which is more than most modern studies in this literature manage, and worth saying out loud. But n = 18 total. Outcomes included Torrance divergent-thinking measures, reflectivity, metacognition and the ability to describe directions. Read at abstract depth.

Reported that Logo influenced reflectivity, divergent thinking, metacognitive ability and the ability to describe directions, with significant gains in Torrance fluency and originality (p < .01). This is the canonical pro-transfer citation of the Logo era and the ancestor of every claim that coding makes children creative — and it has nine children per arm. The largest far-transfer subskill effects in the modern meta-analysis are originality (g = 1.28) and spatial memory (g = 1.25), which is precisely where studies of this size and vintage live.

Low for the internal contrast (random assignment), but with n = 9 per arm the estimate is dominated by sampling noise regardless.

Extended by Clements' own later Logo work. No independent replication at adequate power, and contradicted at the design level by Pea & Kurland's two null year-long studies on the closest outcome.