Cognitive Architecture and Massive Modularity

The organization of the processes that underwrite our perceptual and cognitive architecture - “the cognitive architecture” - is an important topic of debate among philosophers of psychology.

In contrast to Fodor, many psychologists have argued that the processes underlying some higher cognitive capacities are modular. For instance, Elisabeth Spelke has argued that our capacity to orient ourselves is under­written by a geometric module (Hermer and Spelke 1996), while Nancy Kanwisher has proposed that a module underwrites our capacity to identify individual faces (Kanwisher, McDermott, and Chun 1997). Going further, some psychologists (e.g., John Tooby, Randy Gallistel) and some philosophers (e.g., Peter Carruthers, Dan Sperber) propose that all our psychological processes are modular, a thesis known as “the massive modularity hypothesis.”

Various arguments have been proposed in support of the massive modularity hypothesis (Sperber 1994, 2001; for a systematic overview, see Carruthers 2006, ch.

In reply, Richard Samuels (1998) notes that there are two types of modules - computational modules and Chomskyan modules. Computational modules are mechanisms; they are defined by the nature of their processes. The modules hypothesized by Tooby and Cosmides are of this first kind. Chomskyan modules are bodies of knowledge about specific tasks - they are representations, not processes.

The massive modularity hypothesis has been criticized on various grounds (Fodor 2000, Buller 2005). I discuss here only two problems, the input problem (Fodor 2000) and the brain evolution problem (Quartz 2002). Noting that a specific type of input is required to trigger a given module, Fodor (2000) contends that a psychological process (a routing system) is needed to pair each module with the stimuli that trigger it. Because this routing system would have to be activated by all types of stimuli, it could not be modular. Thus, the massive modularity hypothesis is false. Clark Barrett (2005) has convincingly rebutted this argument by drawing an analogy between enzymes and modules. Enzymes come into contact with a large range of substrates. However, because they have specific binding sites, only some of these substrates are bound with enzymes.

Barbara Finlay and colleagues’ work on brain evolution has also inspired an important objection against the massive modularity hypothesis. They found that across mammals, the volume of the main brain structures is correlated to the volume of the whole brain (Finlay and Darlington 1995). Steve Quartz has argued that these findings show that natural selection did not act on individual brain structures independently of the other brain struc­tures: “These results suggest that neural systems co vary highly with one another as a consequence of the restricted range of permissible alterations that evolutionary psychology can act upon. This makes the massive modu­larity hypothesis of narrow evolutionary psychology untenable” (2002, 189).

Quartz’s argument should be resisted (Machery 2007c). A closer look at Finlay and colleagues’ data shows that across mammals, the volume of the whole brain does not covary perfectly with the volume of the main brain structures, suggesting that natural selection may have acted upon their volume. Furthermore, there is more to brain evolution than the volume of the brain structures considered by Finlay and colleagues. Natural selection probably acted upon the nature of brain cells, their organization, or the connectivity between brain areas.

Finally, it is noteworthy that a careless use of the term “module” has muddled the debate about the massive modularity hypothesis. “Module” means different things for different people (Barrett and Kurzban 2006). Particularly, as noted above, evolutionary psychologists define modules as those processes that have a dedicated function. They need not have any of the properties that characterize Fodorian modules. (Similarly, “module” has a distinctive use in neuroscience.) For instance, modules need not be innate, nor need they be automatic or cognitively impenetrable (Machery and Barrett 2006). Rejecting evolutionary psychologists’ massive modularity hypothesis on the grounds that our central processes do not possess the properties that characterize Fodorian modules (e.g., they are not automatic, etc.) is thus unsound.

3.