To make reliable but metabolically efficient perceptual inferences in a changing world, neural systems should dynamically adapt based on surprise and uncertainty about the sensory environment.

Psychophysical measurement and computational modeling show that sensory information cannot contribute directly to a cognitive judgment, but must first be integrated into resource-limited working memory.

Analysis of speech acoustics and human brain activity reveals the neural computations by which listeners combine speech input with prior expectations.

Set size effects in visual working memory are explained as a resource-rational trade-off between an error-based behavioral cost and a neural encoding cost.

Multivariate analyses of human electrophysiological recordings revealed that the brain represents unexpected visual stimuli with greater fidelity than expected stimuli which arose independently of simple habituation arising from repetition.

Changes to sensory predictions are encoded by beta oscillations, surprise due to prediction violations by gamma oscillations, and alpha oscillations may have a role in controlling the precision of predictions.

Neural oscillations are a necessary consequence of efficient coding of sensory signals by a spiking neural network, limited by synaptic delays and noise.

A specific neuronal pathway within the auditory system represents information about the statistical prediction and error for incoming sounds, thereby contributing to efficient representation of complex sounds and sound streams in the brain.

Computational and theoretical analyses offer novel and unexpected insight into how complex, naturally occurring odor mixtures are parsed and normalized at the very first stage of olfaction.

Key variables required for pain perception in the context of a predictive coding model are correlated with distinct oscillatory profiles.