Sleep strengthens successor representations of learned sequences in humans.

Experiences reshape our internal representations of the world. However, the neural and cognitive dynamics of this process are largely unknown. Here, we investigated how sequence learning reorganizes neural representations and how sleep-related consolidation mechanisms contribute to this transformation. Using high-density electroencephalography and multivariate decoding, we found that learning temporal sequences of visual information led to the incorporation of successor representations during a subsequent perceptual task, despite temporal information being task-irrelevant. Importantly, individuals with better sequence memory performance exhibited stronger successor incorporation during the perceptual task. Representational similarity analyses comparing neural patterns with different layers of a deep neural network revealed a learning-induced shift in representational format, from low-level visual features to higher-level abstract properties. Critically, both the strength and transformation of successor representations correlated with the neurophysiological hallmarks of slow-wave sleep during a post-learning nap, particularly the coupling between slow oscillations and spindles. These findings support the idea that sequence learning induces lasting changes in visual representational geometry and that sleep physiology strengthens these changes, providing mechanistic insights into how the brain updates internal models after exposure to environmental regularities.