Which cognitive abilities underlie computational thinking? Criterion validity of the Computational Thinking Test

1251

Spanish students in grades 5-10.

A cross-sectional criterion-validity study by the developers of the Computational Thinking Test, correlating it against the Primary Mental Abilities battery and the RP30 problem-solving test. Read at secondary depth. Recorded here not as evidence about an intervention but as evidence about what the field's outcome instrument is actually measuring — which is load-bearing for every effect size in the topic.

The Computational Thinking Test correlates r = .67 with the RP30 problem-solving test, r = .44 with spatial ability and r = .44 with reasoning ability, with alpha = .79. The authors' own framing is that computational thinking should not be entirely independent from other scales used to predict problem-solving ability. The implication for this topic: a substantial share of variance in any computational-thinking test score is ordinary reasoning and spatial ability, so an intervention that raises a CT score has not thereby been shown to have taught anything beyond what the test shares with an ability battery.

HIGH, and that is the point of including it. Reasoning and spatial ability are substantially heritable, so cross-sectional findings that programmers or programming students score higher on computational thinking are confounded by ability selection into programming.

The g-loading finding is independently reproduced by Aranyi et al. (2024) in programming-naive adults, with r = .56 against fluid intelligence.