This is a 5-year project funded by the ERC Consolidator Grant. We started the project in early 2026 together with Sam Whitehurst and Denise Kittelmann, soon to be joined by another PhD candidate and a postdoc. We will use a combination of neuroimaging (3T fMRI, 7T fMRI, fMRS), electrophysiology (EEG), and computational modeling (connectivity models, representational similarity analysis, Bayesian models of belief updating) to disentangle the effects of memory and prediction on processing the present.

Abstract: 

Processing the present (perception) is shaped by past experiences (memory) and current expectations of future events (prediction). The Bayesian brain hypothesis formalizes thisidea, positing that living agents maintain a model of the world, built upon memories and entailing predictions of likely causes of sensations. Prediction is mediated by multiple brain regions such as sensory neocortex and the hippocampus, and my research has pinpointed several neural mechanisms of prediction across task contexts. Brain networks mediating prediction are also crucial for memory, suggesting they are closely related. However, emerging evidence indicatesthat predictive processing may disrupt memory encoding, raising the unresolved question: How does the brain differentiate between mnemonic and predictive processing? Understanding this distinction will not only clarify how memories inform predictions but also why they sometimes interfere.

To address this, MemPred will pursue three interconnected objectives: (1) dissecting the neural mechanisms that distinguish memory and prediction across brain networks, circuits, and functions; (2) investigating how task context modulates memory/prediction signaling; and (3) deriving a unified neurobehavioral model of how memory and prediction interact to shape neural activity and behavior. By combining cutting-edge techniques in electrophysiology and neuroimaging with novel developments in computational modeling, MemPred will offer fundamentally new insights into the dissociable effects of memory and prediction on processing the present.

This research will not only advance cognitive neuroscience by elucidating the neural underpinnings of memory and prediction but will also lead to methodological breakthroughs in generative models for behavioral and neural data. MemPred's findings will extend to fields such as artificial learning and computational psychiatry, offering a novel framework for understanding the interplay between memory and prediction.

Relevant past publications:

Barron HC, Auksztulewicz R, Friston K (2020) Prediction and memory: a predictive coding account. Prog Neurobiol, 101821 [LINK]

Cappotto D, Kang HJ, Li KY, Poeppel D, Melloni L, Schnupp JWH*, , Auksztulewicz R* (2022) Simultaneous Mnemonic and Predictive Representations in the Auditory Cortex. Current Biology 32 (11), 2548-2555e5 [LINK]