Perceptual integration, modularity, and cognitive penetration

Imagine a computational system with a proprietary... database. Imagine that this device operates to map its characteristic inputs onto its characteristic outputs... and that, in the course of doing so, its informational resources are restricted to what its proprietary database contains. That is, the system is “encapsulated ” with respect to information that is not in its database.... That’s what I mean by a module. In my view, it’s informational encapsulation, however achieved, that’s at the heart of modularity (Fodor, 2001, 63). On the best evidence, there seems to be significant interaction between information sources, at all levels of perception: for better or worse, the picture of perception as a collection of independent, non-interacting, exclusively bottom-up feature extractors must be replaced by a picture of perception that more centrally incorporates informational interaction. The question we want to raise in this paper is what impact this lesson should have for our thinking about claims about the modularity and the cognitive penetrability of perception. Many authors have taken the news that perception involves substantial informational integration to require seriously weakening or abandoning idea