Predictive processing

Predictive processing is the view that the brain continually generates predictions about incoming signals and updates them when those signals differ from expectations. It explains perception and action as parts of an ongoing process of prediction and error correction.

Predictive coding: A neural-processing model in which higher levels send predictions downward and lower levels return prediction errors. It specifies a hierarchical mechanism for predictive processing.
Bayesian inference: A method for updating probabilities in light of new evidence and prior beliefs. Predictive processing is often described as Bayesian belief updating.
Hermann von Helmholtz: A German physicist and physiologist who described perception as unconscious inference. His account anticipated the idea that perception relies on inferred causes.
Visual illusion: A perceptual experience that differs systematically from the physical properties of a stimulus. Illusions can reveal how prior expectations shape interpretation of sensory input.
Feedforward processing: Information processing that passes signals from earlier stages to later stages without relying on feedback. Predictive processing emphasizes reciprocal feedback rather than a purely forward flow.
Prediction error: The difference between a predicted signal and the signal that is actually received. Updating depends on discrepancies between predicted and incoming signals.
Prior probability: The probability assigned to a hypothesis before considering current evidence. Predictions depend on expectations formed before each incoming signal.
Richard Gregory: A British psychologist who treated perception as hypothesis testing shaped by prior knowledge. His perceptual theories helped establish top-down prediction as an explanatory idea.
Perceptual learning: A lasting improvement in sensory discrimination or interpretation through experience. Learning can alter the predictions used to interpret recurring signals.
Stimulus-driven processing: Processing guided primarily by current sensory input rather than expectations or goals. It contrasts with accounts that assign a substantial role to prior predictions.

Linked from 37 pages