Figure 1.

A simplified hierarchical PP model based on the circuit proposed by Rao and Ballard. ¹⁸ Here, each level is composed of a population of reciprocally connected expectation units (Ex) and error units (Er). Expectation units signal the expected pattern of activity at the preceding level, given the current perceptual hypotheses. Where discrepancies emerge between these sensory expectations and sensory‐driven input, prediction errors are fed forward to the subsequent level to inform the revisions of prediction necessary to minimize prediction error.