Predictive Architectures Cannot Be Modular

Drayson (2017) explores the relationship between predictive and modular architectures of the mind and concludes that predictive architectures must exhibit some kind of modularity. To do so, Drayson discusses two requirements of predictive architectures that seem to conflict with modular architectures: the continuity claim, the idea that cognition and perception rest on a continuum, and the non-isolation claim, the idea that no brain processes are informationally isolated. Although these features seem to repel modular architectures, Drayson finds reasons for reconciliation. In this paper, I explain such reasons and provide difficulties in Drayson’s argumentation. I conclude that there is no place for reconciliations.