In the 2006 cortical scheme only prediction error travels up the hierarchy, carried by superficial pyramidal cells; predictions travel down from deep ones.

The claim

Friston, Kilner and Harrison (2006) derive, under Gaussian assumptions, recognition dynamics in which each level is driven only by the prediction error from the level below and its own. They map this onto anatomy: forward connections, which come from superficial pyramidal cells, would carry prediction error, and backward connections from deep cells would carry predictions. Since superficial cells dominate EEG signals, the scheme predicts that much measured cortical response is prediction error. It is a hypothesis about cell types, which the paper states as something "we can posit".