Back

Colourful predictive templates in early visual cortex

Bartsch, M. V.; Spaak, E.; de Lange, F. P.

2026-06-12 neuroscience
10.64898/2026.06.11.731729 bioRxiv
Show abstract

Predictive processing theories suggest that our brain constantly predicts incoming sensory information, though the level of granularity and concrete sensory content of these predictions is debated. In this study, we investigate whether predictions in visual cortical areas carry information about colour before the onset of the anticipated stimulus. Thirty-seven human participants viewed coloured disc stimuli presented in a predictable sequence, enabling the brain to form expectations about the upcoming colour, while the brain response was tracked with high spatiotemporal resolution (combined electro-and magnetoencephalography). A decoding model trained on a separate colour localizer dataset could successfully reconstruct the predicted colour during the pre-stimulus period, where only a grey placeholder was present. Thereby, our findings demonstrate that early visual regions of the brain form anticipatory representations of colour information, in the absence of external colour input, yet in a representational format similar to veridical sensory input. Significance StatementThe human brain continuously anticipates upcoming events, but the exact features and level of detail of these predictions is a topic of debate. Here we show that when a visual stream has a predictable sequence of colours, the brain pre-activates colour information in early visual cortex before the stimulus appears. Using high-temporal-resolution brain recordings and multivariate decoding analyses, we demonstrate that expected colours can be reconstructed from neural activity several hundreds of milliseconds before the anticipated stimulus is shown. These findings provide direct evidence that perceptual predictions contain concrete sensory content in terms of colour information. By clarifying the features that make up predictive representations, our work advances understanding of how expectations shape perception and guide attention in everyday vision.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.