Hippocampal-neocortical interactions in prediction and memory
Helen Barron (University of Oxford)
Hosted by the Computational Neuroscience Unit
Memories for past experience can be used to guide future decisions. For example, we can use memory to repeat actions with known consequences, but we can also draw on loosely related events, to predict the outcome of an entirely novel choice. Using memory to support flexible decision making in this manner is thought to engage a number of brain regions. However, the underlying neuronal computations remain unclear. In this talk I use a cross-species approach in humans and mice to reveal the neuronal computations that describe how memory is used to guide upcoming decisions. During inferential choice, I show that the hippocampus engages a prospective code to forecast temporally-structured learned associations. This prospective code correlates with memory reactivation in relevant sensory neocortical areas, where a transient break in the balance between excitation and inhibition (EI) can be observed. In a second set of studies I relate this hippocampal-neocortical interaction to behavioural measures of memory interference. I show that activity in the hippocampus can predict memory interference, but not when neocortical EI balance is disrupted using non-invasive brain stimulation. Together these studies reveal how the hippocampus interacts with neocortex when memory is used to guide upcoming decisions. Within a predictive coding framework, these findings cast memory as a fictive prediction error that may provide a training signal to optimise generative models of the world, in the absence of sensory data.
Join via BlueJeans: https://bluejeans.com/929295991
For further information contact Cian O'Donnell.