Customer Reviews

Contemporary Debates in Cognitive Science (Contemporary Debates in Philosophy)

5 star:		(0)
4 star:		(0)
3 star:		(1)
2 star:		(0)
1 star:		(0)

 

 

I am summarising only the part on modularity (chs. 1-3).

The massive modularity thesis is meant be challenging the notion of the mind as a general-purpose computer. Arguments supposedly in its favour are:

Evolution operates in increments. Therefore the mind must be decomposable into subcomponents. This is supposed to be contrary to the notion of the mind as a general-purpose computer, but obviously it is not since computers do indeed consist of lots of subcomponents. It may be said that evolution operates not only in increments but in increments each of which is adaptive. But again computers fit the bill, their historical development being of precisely this type. So this argument is entirely pointless since it does nothing to separate the massive modularity thesis from the opposing view. More generally, if the modularity thesis is to be meaningful it must speak not only of modules but of the nature of those modules.

Computational intractability. It is claimed that modules need to be informationally encapsulated because it would be computationally intractable to sift through our entire memory, all our beliefs, etc., to find the information relevant to a particular task. But this is plainly nonsense. Computer science is full of extremely fast database search algorithms. Some basic indexing can obviously give all the same benefits as encapsulation. All the more so since the human mind is obviously fallible, which means that it can use innumerable heuristic and approximative shortcuts. And of course the mind need not be computational at all.

Empirical evidence. E.g., perceptual illusions encapsulated from belief; disorders that affect only a specific function, dissociating from everything else; and Innumerable forms of empirical evidence (e.g., neural imaging, etc.) for specific modules. It is all very well to study such particulars but it has little to do with a general thesis of massive modularity. Since the notion of module has so many identifying characteristics, it is not surprising that it has many (partial) instantiations. For their thesis to be interesting, modularists must pick some defining properties of modules and show their prominence. Their best hope seems to be domain-specificity and informational encapsulation, but even here there is trouble, and very close to home at that:

The mind performs many tasks that require integration of information from many domains. Therefore modules are not domain-specific. Many modularists avoid the problem by allowing some domain-general modules. But this risks trivialising the thesis: surely it is not surprising that if one divides some function into small enough components then some of them will turn out to be domain-specific. The question is whether domain-specificity is interestingly prominent. An argument against this: the visual system (the modularists' paradigm example) is used for non-perceptual tasks such as conceptual problem-solving and memorisation; there is also evidence the same areas of the brain are recruited to aid the sense of touch in blind people. Similar interconnections exist for other "modules."

Many supposedly encapsulated modules, including perception, are subject top-down influence; e.g., if I am looking for a Kodak film carton then small yellow objects stand out. The modularists reply that top-down influence is fine. Encapsulation means rejection of wholesale theory-ladenness and the like, not top-down influence on the choice of focus, secondary interpretation of the perceptual data, etc.

Supposedly encapsulated modules can also influence each-other. E.g., when speech and lip-movements do not match, auditory perception is systematically distorted.

Further counterarguments:

Modularity seems contradicted by our general reasoning abilities (conceptual integration, inferential holism, task range, etc.). The modularists claim that it is precisely the interplay of modules that makes this complexity and generality possible; but, as this claim is not further substantiated, it is not so much a solution to the problem as a restatement of it given a commitment to modularity.

Cognitive ability in one domain is a good indicator of it in other domains as well. Some disorders affect cognitive abilities generally (e.g., Down's syndrome).