Cognitive Penetrability of Perception in Predictive Brains

Abstract

The debate over whether perception is cognitively penetrable, in order to remain philosophically interesting and relevant, should be unpacked in terms of discerning whether propositional states can directly modulate the low-level phenomenal features of one’s perceptual experience. For this, it should also engage with cognitive science, and it is by proposing a scientifically feasible mechanism of how cognitive penetrability could obtain that there is even a reason for considering it. The predictive coding model of perception provides a framework within which such a mechanism can be motivated. By framing perception as a process of probabilistic inference, wherein top-down models of the world are tested against bottom-up error signals, we can discern whether propositional attitudes could affect the inhibitory and excitatory signaling of sensory neurons in a way that either promotes or inhibits the top-down predictive model that constitutes one’s perceptual experience, all the while preserving the necessary kind of relation between perception and cognition. In a three dimensional variation of the Ebbinghaus illusion, this can in fact be shown to happen, establishing cognitive penetrability of perception and a working mechanism by which this could take place.