Towards a Neural Network Hypothesis for Functional (Dissociative) Amnesia: Catastrophic Forgetting

Abstract

Functional amnesia, also known as dissociative amnesia, psychogenic amnesia, or mnestic block syndrome, is a rare disorder which, although clinically heterogeneous, is most often characterised by dense retrograde amnesia mainly affecting the episodic-autobiographical domain but with relative preservation of anterograde memory function, a pattern dissimilar to that seen in other amnesic disorders. The pathogenesis of functional amnesia remains unknown. Here, appeal is made to the study of artificial neural networks in the hope that, as in other mnestic disorders, this might give insight into the mechanisms underpinning functional amnesia. Specifically, the observation of catastrophic forgetting or catastrophic interference occurring in artificial neural networks, that is the abrupt and complete loss of previously learned information when learning new information, is extended to the human nervous system to develop a novel hypothesis: the Catastrophic Forgetting Hypothesis of functional amnesia.