Neural heterogeneity enables adaptive encoding of time sequences

The timing mechanisms in biological systems operate across a vast range of scales, from microsecond precision for sound localization to annual cycles. A key open question concerns the mechanisms for encoding intermediate time intervals-hundreds of milliseconds to minutes-that are essential for navigation, communication, memory, and prediction. Here we present a theoretical framework that explains how neurons can represent such intervals using a common biophysical property: neural fatigue, where activity diminishes during sustained stimulation. Our Bayesian framework combines parametrically heterogeneous stochastic dynamical modeling with interval priors to predict available timing information independent of the actual decoding mechanism. We find that a trade-off emerges between accurately representing the most recent interval and retaining information about previous ones. We show that cellular diversity is not just tolerated but required to encode sequences of time intervals. Our work highlights the computational role of biological heterogeneity in shaping memory for time, with implications for understanding temporal processing in neural circuits.