Habituation as Cognitive Competence: A Formal Distinction from Neural Mechanisms

Habituation—the progressive attenuation of responses to repeated stimulation—is a ubiquitous phenomenon spanning perception, learning, and attention. Yet its cognitive and neural manifestations are often conflated, obscuring distinct mechanisms and limiting both theory and application. Here, we introduce a formal framework that separates cognitive habituation—an abstract competence defined by belief updating and surprise minimization—from neural habituation—the implementational performance realized through synaptic and circuit‐level dynamics. We develop dynamical‐systems models of synaptic depression alongside Bayesian and variational formulations of predictive coding, integrate them within a competence–performance hierarchy, and derive computational cost–benefit trade‐offs between flexibility and metabolic efficiency. A complexity‐theoretic analysis further establishes that, under bounded‐rationality constraints, no purely neural mechanism can simulate the full generality of cognitive habituation without super‐polynomial overhead. Through simulations and logical arguments, we show how this multi‐level account clarifies developmental habituation profiles, suggests biomarkers for disorders such as autism and epilepsy, and guides the design of adaptive artificial agents with competence‐aware filtering. Our work unifies neural and cognitive perspectives on habituation, offering precise predictions and a roadmap for future empirical and technological advances.