Connectionism Need Not Be (Strongly) Associationistic
In a previous post, I briefly discussed the relationship between connectionism and associationism. Thanks in part to the helpful feedback I received, I have now revised the relevant section of the paper I am working on. I'd be interested in any additional comments or references that anyone might have. The text of the paper reads as follows:
The relationship between neurocomputational approaches and associationism is more complex than many suppose. We should distinguish between strong and weak associationism. Strong associationism maintains that association is the only legitimate explanatory construct in a theory of cognition [cf. Fodor 1983, p. 27]. Weak associationism maintains that association is a legitimate explanatory construct along with others such as the innate structure of neural systems. (We should also note that there are different associative mechanisms, some of which are more powerful than others; comparing associative mechanisms goes beyond the scope of this paper.)
To be sure, some connectionists profess strong associationism [e.g., Rosenblatt 1958, p. 387]. But that is beside the point, because connectionism per se is consistent with weak associationism or even the complete rejection of associationism. Some connectionist models do not rely on association at all—a prominent example being the work of McCulloch and Pitts [1943]. And weak associationism is consistent with many theories of cognition including classicism. A vivid illustration is Alan Turing’s early proposal to train associative neural networks to acquire the architectural structure of universal Turing machines [Turing 1948, Copeland and Proudfood 1996]. In Turing’s proposal, association may explain how a network acquires the capacity for universal computation (or an approximation thereof), while the capacity for universal computation may explain any number of other cognitive phenomena.
Although many of today’s connectionists and computational neuroscientists emphasize the explanatory role of association, many of them combine association with other explanatory constructs (weak associationism) [cf. Smolensky and Legendre 2006, p. 479; Trehub 1991, pp. 243-5; Marcus 2001, pp. xii, 30]. What remain to be determined are which neural networks, organized in what way, actually explain cognition and which role association and other explanatory constructs should play in a theory of cognition.
Categories: connectionism, Cognition
Tags: connectionism cognition
Perhaps this is relevant to your paper. Two important studies confirming that the representation of space is an innate property of the mammalian brain were just published in Science. These direct neuro-physiological probes of rat brain lend strong empirical support to the basic claim of the retinoid model which, as such, is a connectionist but not an associative mechanism. Here is an excerpt from the commentary on this recent work:
Science 18 June 2010: Vol. 328. no. 5985, pp. 1487 – 1488
PERSPECTIVES
NEUROSCIENCE:
A Kantian View of Space
Linda Palmer and Gary Lynch
Department of Anatomy and Neurobiology, University of California, Irvine, CA 92697, USA.
"How does the brain represent space? Is this representation entirely the result of learning from experience? In his Critique of Pure Reason, Immanuel Kant argued that there must be certain “a priori conditions” of cognition, which could not be derived from experience but must instead be given prior to it. His theory includes two “a priori pure forms” of space and of time, regarded as constraints of thought rather than results of investigation or experience (1, 2). On pages 1576 and 1573 of this issue, Langston et al. (3) and Wills et al. (4) both refer to Kant’s theory and report that critical components of the brain’s spatial representation systems are already in place when an animal first encounters an extended environment. This supports the view that spatial representation indeed includes an innate component prior to experience."
Reply to this
Arnold, thanks a lot for this excellent reference, which indeed is extremely relevant.
Reply to this