The human episodic memory system supports the ability to remember one's past experience, while the semantic memory system supports memory for one's general knowledge about the world and the people, places, and things within it. These systems are interrelated: The semantic memory system allows one to construct representations of the features and details making up an experience, and the episodic memory system binds these semantic representations to a unique spatiotemporal context. Despite the critical importance of these systems, there are major gaps in our knowledge regarding the nature of the neural and cognitive mechanisms that allow us to flexibly access, manipulate, and retrieve episodic and semantic memories. With funding from the National Science Foundation, Dr. Sean Polyn of Vanderbilt University is developing the Control of Semantic and Episodic Memory (CSEM) model, a novel theory designed to explain the how the brain mediates episodic and semantic memory search and retrieval. This project will involve undergraduate students, graduate students, and research staff at every level: data analysis, theory development, and presentation of results to the broader scientific community. The project's objectives will make key contributions to infrastructure for research and education, as the simulation software and the database of anonymized behavioral and neural data will be distributed to the broader scientific and academic community via the internet, along with a repository of behavioral and neural analysis tools.
The CSEM model is a computational model of human memory designed to test hypotheses regarding the specific mechanisms carried out by particular brain regions in the storage, retrieval, and flexible use of semantic and episodic memories. The research team will use functional magnetic resonance imaging (fMRI) to record blood-oxygen-level dependent (BOLD) activity as participants perform semantic and episodic memory tasks. Multivariate pattern analysis, representational similarity analysis, and advanced statistical techniques will be used to characterize the representational structure and dynamics of neural activity, supporting the development of functional, mechanistic computational models of the cognitive system of participants as they perform the memory tasks. The first study examines neural signals for evidence of an integrative contextual representation that allows information about recent experiences to persist in neural activity patterns. The second study examines how the neural system constructs a retrieval cue that can be used to target particular sets of memories. The third study examines how memories interfere with one another during retrieval, and the mechanisms used by the neural system to resolve this interference. Data from these experiments will support the development of a unified theoretical framework to understand how general-purpose executive processes control the semantic and episodic memory systems to give rise to flexible, goal-directed behavior.
