Visual Mismatch and Predictive Coding: A Computational Single-Trial ERP Study | JoVE Visualize

Area of Science:

Cognitive Neuroscience
Computational Psychiatry
Perceptual Psychology

Background:

Predictive coding (PC) theory suggests the brain infers environmental causes using a generative model and updates it via precision-weighted prediction errors (pwPEs).
While evidence exists for PC in auditory processing (mismatch negativity), experimental validation in the visual domain is limited.
Understanding visual mismatch responses is crucial for elucidating automatic perceptual processes and brain model updating.

Purpose of the Study:

To test whether visual mismatch responses are linked to trial-by-trial precision-weighted prediction errors (pwPEs) within the predictive coding framework.
To investigate the role of pwPEs in updating the brain's generative model during visual perception of unexpected changes.
To compare the explanatory power of a Bayesian pwPE model against a classical change detection model for visual mismatch responses.

Main Methods:

Employed a novel visual "roving standard" paradigm with unexpected changes in face color or emotional expression to elicit mismatch responses.
Utilized a hierarchical Bayesian model to simulate and analyze trial-by-trial pwPE trajectories of a Bayes-optimal observer.
Conducted a comprehensive time × sensor space analysis of electroencephalography (EEG) data, comparing Bayesian and categorical change models.

Main Results:

Found significant modulation of brain activity by both color and emotion pwPEs, demonstrating their influence on visual processing.
Observed that the scalp distribution and timing of single-trial pwPE responses aligned with traditional visual mismatch responses.
Demonstrated that trial-wise pwPEs provided a superior explanation for observed mismatch responses compared to a categorical change detection model.

Conclusions:

Visual mismatch responses reflect trial-wise precision-weighted prediction errors (pwPEs), supporting the predictive coding (PC) interpretation in the visual domain.
The findings extend beyond classical electroencephalography (EEG) analyses, highlighting the utility of computational modeling for studying automatic perception.
This research provides strong evidence for the predictive coding framework in visual perception, akin to findings in auditory processing.