Kavli Affiliate: Mounya Elhilali
| Authors: Sijia Zhao, Benjamin Skerritt-Davis, Mounya Elhilali, Frederic Dick and Maria Chait
| Summary:
The brain is increasingly viewed as a statistical learning machine, where our sensations and decisions arise from the intricate interplay between bottom-up sensory signals and constantly changing expectations regarding the surrounding world. Which statistics does the brain track while monitoring the rapid progression of sensory information? Here, by combining EEG (three experiments N≥22 each) and computational modelling, we examined how the brain processes rapid and stochastic sound sequences that simulate key aspects of dynamic sensory environments. Passively listening participants were exposed to structured tone-pip arrangements that contained transitions between a range of stochastic patterns. Predictions were guided by a Bayesian predictive inference model. We demonstrate that listeners automatically track the statistics of unfolding sounds, even when these are irrelevant to behaviour. Transitions between sequence patterns drove an increase of the sustained EEG response. This was observed to a range of distributional statistics, and even in situations where behavioural detection of these transitions was at floor. These observations suggest that the modulation of the EEG sustained response reflects a universal process of belief updating within the brain. By establishing a connection between the outputs of the computational model and the observed brain responses, we demonstrate that the dynamics of these transition-related responses align with the tracking of ‘precision’ – the confidence or reliability assigned to a predicted sensory signal – shedding light on the intricate interplay between the brain’s statistical tracking mechanisms and its response dynamics.