Computational thinking training and its effects on working memory, flexibility, and inhibition: Randomized controlled trial in fifth-grade children
Robledo-Castro, C., Castillo-Ossa, L. F., & Hederich-Martínez, C. · 2025
grade Crctdeveloper-ledunreplicated
Sample
111 children (57 experimental, 54 control) from 4 fifth-grade classes in 2 Colombian cities
Population
Colombian public-school fifth graders, mean age 10.75, 76% from low socioeconomic strata.
Design
Prospectively registered with a published protocol, CONSORT-reported, sealed-envelope allocation by an external investigator, blinded outcome assessors, and intention-to-treat with imputation. Those are real strengths and rare here. But randomisation was of four INTACT CLASSES — two clusters per arm — analysed with pupil-level repeated-measures ANOVA, which is a cluster-randomisation error: treatment and class effects are not separable. The paper calls the comparison an active control group while describing it as treatment-as-usual following the standard curriculum. Twelve weeks of unplugged activities plus game programming plus educational robotics. No academic outcome and no follow-up.
Key findings
Group-by-time partial eta-squared: auditory-verbal working memory total 0.167; visuospatial working memory (reverse Corsi) 0.081; visual working memory 0.055; Stroop word/colour inhibition 0.038. Cognitive flexibility (Wisconsin Card Sorting errors, perseverations, categories) showed NO change in either group on any indicator. Within-group post-hoc effects are large but the control arm also improved on several tasks. The pattern is diagnostic: the memory-span tasks that a twelve-week programming and robotics course most resembles move, and the ability-like construct — set-shifting — does not.
Genetic confound
Low. Randomised allocation, though at class rather than pupil level.
Replication notes
The first trial of this specific programme. Its cognitive-flexibility null replicates the non-significant pooled flexibility estimate in the computational-thinking executive-function meta-analysis.
DOI / URL
10.1016/j.appdev.2025.101850
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Auditory-verbal working memory (digit span total) | partial eta-squared (group by time) | 0.167, p < .001 | standardized | post-test at 12 weeks | business-as-usual | end-of-treatment | far-transfer |
| Visuospatial working memory (reverse Corsi) | partial eta-squared (group by time) | 0.081, p < .001 | standardized | post-test at 12 weeks | business-as-usual | end-of-treatment | far-transfer |
| Inhibitory control (Stroop word/colour) | partial eta-squared (group by time) | 0.038, p = .042 | standardized | post-test at 12 weeks | business-as-usual | end-of-treatment | far-transfer |
| Cognitive flexibility (Wisconsin Card Sorting errors, perseverations, categories) | partial eta-squared (group by time) | 0.009 / 0.005 / 0.001, all non-significant | standardized | post-test at 12 weeks | business-as-usual | end-of-treatment | far-transfer |
Cited by
- Does learning to code improve general thinking?no effectconf: mediumgc: low