Proceedings of the Annual Meeting of the Cognitive Science Society

The Cognitive Reflection Test (CRT) is widely used in reasoning and decision-making research, yet it lacks a formal cognitive foundation. As a result, debates persist over why CRT scores correlate with mathematical ability, why individuals differ in performance, and which mental processes drive observed response patterns. We address these questions using an ecological perspective combined with computational cognitive modeling. We characterize the learning environment by assembling a large dataset of grade-school verbal math problems and specify a learning mechanism that maps linguistic features of problems to arithmetic operations based on prior experience. Our model reproduces the characteristic intuitive errors elicited by CRT problems and explains them as byproducts of adaptive cognition. It further generates novel predictions about how environmental structure and problem wording influence strategy selection and performance, which we test in two preregistered experiments. Overall, our work provides a principled, quantitative account of CRT performance grounded in adaptive generalization.