Police Reporting After Sexual Assault: Victim and Assault Characteristics That Predict Reporting Among Victims Who Seek Medical Forensic Care.

Despite the success of the theorised distinction between episodic (hippocampal) and semantic (neocortical) systems, and long-standing ideas regarding the consolidation of memories from hippocampus to neocortex, how both systems work together to support a wide variety of cognitive functions is unclear. These include recollection and problem solving, which rely on a variable mixture of specific episodic memories from the hippocampus more general knowledge from the neocortex. Here, we lay out a framework for how this co-operation occurs, and make use of recent advances in machine learning to simulate how the resulting system would perform these functions for comparison with human performance. This builds on many years of theoretical insights into the relationship between schemas and memory, and their potential correspondence to large language models. We explore three questions, and implement our hypothesised solutions as a computational model (see Table ): We hypothesise that a deep generative model in the neocortex is trained to predict its own inputs, using memories of episodic experiences replayed from the hippocampus (and during the experiences themselves). In this way, it develops generalisable (semantic) knowledge, enabling it to generate episodes from an initial input. With enough consolidation, the neocortex can approximately reconstruct memories without the hippocampus, allowing the redundant hippocampal trace to fade away.