The Adaptable Mind: What Neuroplasticity and Neural Reuse Tell Us about Language and Cognition - John Zerilli 2021
Modules Reconsidered: Whither Modularity?
In Chapter 4, I argued that we ought to regard dissociability as the sine qua non of modularity. As for what in the brain interestingly meets this standard, the only likely candidate will be something resembling a cortical column. But this is not guaranteed. The effects of the neural network context may so compromise a region’s ability to maintain a set of stable input—output relations that it cannot be considered a genuine module.
1 For a similar criticism of Fodor’s module, see Arbib (1989). Note that, by describing Fodor’s modules as marking “gross” functions, I mean only that they can be specified at the level of proprietary domains (e.g., at the level of vision, olfaction, and aspects of language comprehension) rather than simply at the level of edge-detection or depth discrimination. They are not high-level in the sense that they pertain to complex thought, judgment, or memory. See § 7.2.2 for comment on Fodor’s central/peripheral distinction.
2 I hasten to add, however, that Anderson’s (2014) “dispositional vector” account of brain regions is an alternative strategy for coming to grips with the same set of issues. Others are clearly alive to the problem. Proponents of the Leabra architecture, for instance, resist modularist terminology precisely because it “forces a binary distinction on what is fundamentally a continuum” (Petrov et al. 2010, p. 287). See also Frost et al. (2015).
3 I am heavily indebted to Anderson (2014) for the review that follows.