1. A general
purpose mechanism could not solve the diverse array of problems facing our
ancestors.
C&T:
"General mechanisms turn out to be very weak and cannot unassisted perform at least most and perhaps all of the tasks humans routinely perform and need to perform" (Tooby and Cosmides 1992, 39).
A woman who used the same taste preference mechanisms in choosing a mate that she used to choose nutritious foods would choose a very strange mate indeed, and such a design would rapidly select itself out (Cosmides and Tooby 1994, 90).
Buller:
"there is no reason to thinkÉthat a domain-general problem solver would simply apply a solution that works in one problem domain to every other domain it encounters" (145)
Consider social learningÉ
"Cosmides and Tooby fail to show that domain-general mechanisms can't generate domain-specific solutions because their arguments rely on a misrepresentation of how a domain-general problem solver would function in different problem domains." (146)
Compare a 'domain general' immune responseÉ
2. In
order for a mechanism to have evolved and spread in the population, it must
have been selected for in response to 'recurrent structure' in the
environment. That is, there must
have been a transgenerationally stable problem for it to solve. But the 'general domain' has no such
stable structure. "There is
no such thing as a general problem."
Buller:
a. Stable vs. Rapidly changing environment (Stability vs. plasticity)É
b. A general mechanism might evolve in response to domain-specific problems.
3. General
learning mechanisms would have to solve adaptive problems by inferring the
solution from experience. But
"many relationships necessary to the successful regulation of behavior
cannot be observed by any individual during his or her lifetime"
(92). (This is the poverty of the
stimulus argument).
Buller
a. The poverty of the stimulus argument requires some sort of innate solution, but it does not require innate modules that are massive numbers of specialized modules.
Fodor and Samuels
An innate library vs. an innate module
From Burlap
(forthcoming):
But
other accounts of the faculties underlying the domain-specific inferential
patterns are possible, including Richard Samuels's (1998) suggestion that
domain-specific expertise may be underwritten by domain-general computational
mechanisms together with domain-specific bodies of mentally represented
knowledge - what he calls the "Library Model of Cognition." For
example, we might posit that ordinary people have an innate 'library' of
mentally represented rules such as,
IF
PERSON A SETS A
CONDITION C ON THE
BORROWING OF A'S
PROPERTY P, THEN IF
PERSON B SATISFIES C, THEN A OUGHT NOT TO MIND THAT B BORROW'S P.
The
domain-specificity of the inferential pattern Cosmides and Tooby identify would
be explained by a domain-general inferential mechanism that operates on
domain-general mental representations, but exploits the domain-specific
information encoded in this library.
I find Samuels's arguments that the Library Model of Cognition is a distinct and viable explanatory alternative to the dedicated neurocomputational mechanisms that Cosmides and Tooby propose to be compelling.
Cosmides and Tooby reply:
Referring to Richard SamuelsÕ library model, [Burlap] points out that domain-specific inferential results need not imply domain-specific inference procedures (which is true). He then suggests that the domain-specific pattern of results for social exchange could be produced by domain general computational procedures operating on a library of evolved, domain-specific information about social exchange: ÒFor example, we might posit that ordinary people have an innate 'library' of mentally represented rules such as, IF PERSON A SETS A CONDITION C ON THE BORROWING OF A'S PROPERTY P, THEN IF PERSON B SATISFIES C, THEN A OUGHT NOT TO MIND THAT B BORROW'S P.Ó (ms p. 3, italics added).ÉRather than containing rules of inference, the library entry on social exchange might be a set of inert facts, a packet of declarative knowledge that is Òlooked upÓ by domain-general procedures. (This interpretation is most consistent with the library metaphor.) Unfortunately, this version of the library model has an empirical problem: It does not explain why performance elicited by social contracts and precautions is so much higher than that elicited by other deontic rules or, indeed, by familiar indicative conditionals.
b. Fodor and Cosmides are too concessive by a long shot, since much less innate stuff is needed than what they suggest.
The case of facesÉ
Could an attentional filter do the job?
Elman: face recognition becomes progressively modularizedÉ
Three blob detector -> highly plastic pattern recognizing deviceÉ
"Ranchers use the cortical device to distinguish among the cows in their herds, and dog breeders use this area to discern different dogs. But most of us use this area just to recognize human faces, and as a result most cows' faces look alike to us." (151).
Innnateness vs. acquisitionÉ
c. Problems with the poverty of the stimulus argument