Whenever we play chess, plan a vacation, or cook a meal, we draw on past experience to decide what to do next. Our group studies how the brain extracts structure from memory, predicts what comes next, and reuses what it has learned in new situations. We combine behavioural and neuroimaging studies (fMRI, MEG, EEG and sleep recordings) with AI-inspired models from reinforcement learning and neural network research, treating them both as models of brain function and as tools to understand human behaviour. These are our key research areas:

Neural replay

Fast, sequential reactivation of neural patterns, "replay", lets the brain revisit past experience offline and build maps of the task at hand. We develop methods to detect these sub-second sequences non-invasively in humans, and ask what they compute: how they support generalisation, the formation of successor representations, and inference about the latent causes of experience. A special focus is on replay during sleep, and how offline processing reshapes memory and can trigger sudden insight. This work builds on Schuck & Niv, 2019, Science and Wittkuhn & Schuck, 2021, Nature Communications, with more recent advances in Wittkuhn et al., 2025, PNAS, Löwe et al., 2025, PLoS Biology, and our preprint on replay and latent-cause inference, with more to come.

Representations in brains and neural networks

Reinforcement learning needs a representation of the current state of the world. We investigate how the brain builds such abstract, structured representations, in orbitofrontal and medial prefrontal cortex, the hippocampus and entorhinal cortex, and how they support flexible, generalisable behaviour. We also develop and use analytical tools to understand how artificial neural networks themselves learn and represent tasks, for instance how simple networks produce sudden strategy switches and insight. See our work on cortical state representations in Schuck et al., 2016, Neuron and Moneta et al., 2023, Nature Communications, the broader perspective in Moneta, Grossman & Schuck, 2024, Trends in Neurosciences, strategy reuse in Hall-McMaster et al., 2025, PLoS Biology, and network analyses in Löwe et al., 2024, PLoS Computational Biology, with more forthcoming.

Beliefs, uncertainty and decision making in mental disease and aging

To act well in a changing world, we have to track which situation we are in and update our beliefs when it shifts. We study how people infer hidden states and adjust learning under uncertainty, how this is altered in anxiety and other mental disorders, and, as one further angle, how it changes with age. This is reflected in Zika et al., 2023, Nature Communications, Verra et al., 2026, Communications Psychology, and Koch et al., 2024, PLoS Computational Biology.