Noise, uncertainty, and interest: Predictive coding and cognitive penetration

This paper concerns how extant theorists of predictive coding conceptualize and explain possible instances of cognitive penetration. Section 1 offers brief clarification of the predictive coding framework and of cognitive penetration. Section 2 develops more precise ways that the predictive coding framework can explain genuine top-down causal effects on perceptual experience. Section 3 develops these insights further with an eye towards tracking one extant criterion for cognitive penetration, namely, that the relevant cognitive effects on perception must be sufficiently direct. In Section 4, we analyze and criticize a claim made by some theorists of predictive coding, namely, that (interesting) instances of cognitive penetration tend to occur in perceptual circumstances involving substantial noise or uncertainty. We argue that, when applied, the claim fails to explain (or perhaps even be consistent with) a large range of important and uncontroversially interesting possible cases of cognitive penetration. We conclude with a general speculation about how the recent work on the predictive mind may influence the current dialectic concerning top-down effects on perception.