An event-related potentials account of brain predictive coding

Abstract Predictive coding is a theory that tries to account for how the brain processes in an anticipatory manner the expected stimuli, and reorganizes the underlying neural networks as a consequence of the outcome of predictions: Correct or incorrect. EEG has the advantage of making a continuous and almost instantaneous record of brain activity. The present report summarizes work on Event-Related Potentials (ERPs) and reviews the neural validity of Predictive processing as a mechanism to predict future events, assess the validity of predictions, and then update the probabilities associated with future events. Using two experimental models: predictive tone sequences and central cue Posner paradigms and Bayesian modelling, the report suggests that Contingent Negative Variation (CNV) would be related to prior expectation, Mismatch negativity (MMN) and P300 to Bayesian surprise and/or prediction error, and Post Imperative Negative Variation (PINV) to the assessment of trial outcome in uncertainty situations. The review tends to support predictive coding as a theory consistent with brain operations indexed by ERPs.