Proceedings of the Annual Meeting of the Cognitive Science Society

Stable trait estimates from sequential task performance fail to separate ability from adaptability: the real-time recalibration of cognitive control parameters that improves execution without acquiring new knowledge. This conflation biases ability estimates. We develop a tractable framework measuring adaptability through trial-by-trial performance dynamics in mathematical cognition tasks. Across large datasets (N>10,000), adaptability strongly predicts math achievement and is uncorrelated with traditional ability measures, establishing it as an independent construct critical for understanding individual differences in cognitive performance.