Search

Summary Memories are thought to be retrieved by attractor dynamics if a given input is sufficiently similar to a stored attractor state [ 1–5 ]. The hippocampus, a region crucial for spatial navigation [ 6–12 ] and episodic memory [ 13–18 ], has been associated with attractor-based computations [ 5, 9 ], receiving support from the way rodent place cells “remap” nonlinearly between spatial representations [ 19–22 ]. In humans, nonlinear response patterns have been reported in perceptual categorization tasks [ 23–25 ]; however, it remains elusive whether human memory retrieval is driven by attractor dynamics and what neural mechanisms might underpin them. To test this, we used a virtual reality [ 7, 11, 26–28 ] task where participants learned object-location associations within two distinct virtual reality environments. Participants were subsequently exposed to four novel intermediate environments, generated by linearly morphing the background landscapes of the familiar environments, while tracking fMRI activity. We show that linear changes in environmental context cause linear changes in activity patterns in sensory cortex but cause dynamic, nonlinear changes in both hippocampal activity pattern and remembered locations. Furthermore, the sigmoidal response in the hippocampus scaled with the strength of the sigmoidal pattern in spatial memory. These results indicate that mnemonic decisions in an ambiguous novel context relate to putative attractor dynamics in the hippocampus, which support the dynamic remapping of memories. [ABSTRACT FROM AUTHOR]

Summary Integration of separate memories forms the basis of inferential reasoning—an essential cognitive process that enables complex behavior. Considerable evidence suggests that both hippocampus and medial prefrontal cortex (mPFC) play a crucial role in memory integration. Although previous studies indicate that theta oscillations facilitate memory processes, the electrophysiological mechanisms underlying memory integration remain elusive. To bridge this gap, we recorded magnetoencephalography data while participants performed an inference task and employed novel source reconstruction techniques to estimate oscillatory signals from the hippocampus. We found that hippocampal theta power during encoding predicts subsequent memory integration. Moreover, we observed increased theta coherence between hippocampus and mPFC. Our results suggest that integrated memory representations arise through hippocampal theta oscillations, possibly reflecting dynamic switching between encoding and retrieval states, and facilitating communication with mPFC. These findings have important implications for our understanding of memory-based decision making and knowledge acquisition. [ABSTRACT FROM AUTHOR]