Remembering the “When”: Hebbian Memory Models for the Time of Past Events

Abstract

Humans and other species can remember how long ago specific events happened. Based on a systematic exploration of neural coding and association schemes, we develop different hypotheses about how neural networks with Hebbian plasticity could enable an automatic reconstruction of the time of past events. We assess these hypotheses from a computational perspective and in light of the available behavioral and neurophysiological data. Existing models of how the time of past events can be remembered are special cases of our theory. We conclude that many models lead to indistinguishable behavior and future experiments with neural recordings are needed to disambiguate different models.