In this project we provide an experimental approach to testing the main prediction of the scenario model: that only the gist is retrieved from episodic memory.
During memory retrieval, a scenario of the past is generated by combining information from episodic memory with information from semantic memory. Here, we study the neural correlate of this interplay between the two memory systems, as well as mechanisms that influence their relative contribution during scenario generation. In two experiments with functional magnetic resonance imaging (fMRI), we will virtual environments to induce realistic episodic memory. Crucially, we will put some elements of this learning task at odds with established semantic memory (e.g. a toaster in the bathroom). During retrieval, memory for these incongruent elements can be either driven by episodic memory (“the toaster was in the bathroom”) or by semantic memory (“the toaster was in the kitchen”). In the first experiment we include manipulations which will modulate the strength of episodic memory and thus will affect the accuracy of memory responses. To achieve this, we will administer cortisol to participants, mimicking stress. Not only is stress a highly relevant factor in everyday life, its effects on episodic memory formation and retrieval will allow us to disentangle the relative contributions of episodic and semantic memory during scenario generation. Furthermore, in the analysis of fMRI we will leverage representational similarity analysis (RSA) to study the neural patterns underlying the formation of the acquired, conflicting episodic memory, its modification by cortisol and how these patterns are related to behavior. While in the first experiment we will rely on established semantic memory (“a toaster belongs in the kitchen”), in the second series of experiments we will experimentally induce new semantic memory about novel, never before encountered objects to take more rigorous control of the contents of semantic memory. Using RSA, we can study the neural correlate of the newly established semantic memory. Importantly, we will again expose participants to conflicting episodic information in a virtual environment learning task. With RSA, we can track the fate of the newly formed semantic memory after encountering conflicting episodic information and relate this to behavioral outcome (e.g. remembering the episode correctly, or committing memory errors based on semantic information). With this proposal, we investigate key assumptions of the scenario model with tightly controlled experiments and powerful new analysis approaches and move towards a better understanding of the neural mechanisms underlying episodic and semantic memory.