Fig. 1.

A simplified schematic of the hierarchical predictive coding in the cortex. This schematic is based on Fig. 1 in Edwards et al. (2012), published under the terms of Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0). Each yellow box represents a cortical column as a predictive coding unit. In this scheme, pyramidal cells are divided into two classes of prediction (black triangles) and prediction-error (red triangles). Predictive coding is then implemented according to a hierarchical scheme: ‘top-down’, ‘backward’ or ‘descending’ neuronal connections (black arrows) transfer predictions from higher processing levels to lower ones, whereas ‘bottom-up’, ‘forward’ or ‘ascending’ neuronal connections (red arrows) convey prediction-errors in the opposite direction. The term ‘prediction-error’ here refers to the (precision-weighted) difference between expectations and predictions at each unit. The precision-weights (green arrows) are controlled by postsynaptic neuromodulation (e.g. conferred by D1-dopamine receptors). The internal feedback loop within each unit constitutes ‘intrinsic’ connectivity, whereas between-unit interactions lead to ‘extrinsic’ connectivity. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)