Online research in philosophy

This paper explores the difference between Connectionist proposals for cognitive a r c h i t e c t u r e a n d t h e s o r t s o f m o d e l s t hat have traditionally been assum e d i n c o g n i t i v e s c i e n c e . W e c l a i m t h a t t h e m a j o r d i s t i n c t i o n i s t h a t , w h i l e b o t h Connectionist and Classical architectures postulate representational mental states, the latter but not the former are committed to a symbol-level of representation, or to a ‘language of thought’: i.e., to representational states that have combinatorial syntactic and semantic structure. Several arguments for combinatorial structure in mental representations are then reviewed. These include arguments based on the ‘systematicity’ of mental representation: i.e., on the fact that cognitive capacities always exhibit certain symmetries, so that the ability to entertain a given thought implies the ability to entertain thoughts with semantically related contents. We claim that such arguments make a powerful case that mind/brain architecture is not Connectionist at the cognitive level. We then consider the possibility that Connectionism may provide an account of the neural (or ‘abstract neurological’) structures in which Classical cognitive architecture is implemented. We survey a n u m b e r o f t h e s t a n d a r d a r g u m e n t s t h a t h a v e b e e n o f f e r e d i n f a v o r o f Connectionism, and conclude that they are coherent only on this interpretation.

The notion that the mind is a physical symbol system (Newell) with a determinate functional architecture (Pylyshyn) provides a compelling conception of the relation of cognitive inquiry to neuroscience inquiry: cognitive inquiry explores the activity within the symbol system while neuroscience explains how the symbol system is realized in the brain. However, the view the the mind is a physical symbol system is being challenged today by researchers in artificial intelligence who propose that the mind is a connectionist system and not simply a rule processing system. I describe this challenge and offer evidence that indicates the challenge may be well motivated. I then turn to the question of how such changes in the conception of the activity of the mind will affect our understanding of the relation of neuroscience to cognitive inquiry and sketch a framework in which the cognitive system consists of several levels and in which both neuroscience and cognitive science can make contributions at several of these levels.

Ramscar and colleagues (2010, this volume) describe the “feature-label-order” (FLO) effect on category learning and characterize it as a constraint on symbolic learning. I argue that FLO is neither a constraint on symbolic learning in the sense of “learning elements of a symbol system” (instead, it is an effect on nonsymbolic, association learning) nor is it, more than any other constraint on category learning, a constraint on symbolic learning in the sense of “solving the symbol grounding problem.”.

We examine a proposal of Eric Lormand's for dealing with perhaps the chief difficulty facing holistic theories of meaning—meaning instability. The problem is that, given a robust holism, small changes in a representational system are likely to lead to meaning changes throughout the system. Consequently, different individuals are likely never to mean the same thing. Lormand suggests that holists can avoid this problem—and even secure more stability than non-holists—by positing that symbols have multiple meanings. We argue that the proposal doesn't work, however, since multiple meanings are unstable for much the same reason that single meanings are.

We should not admit categorial restrictions on the significance of syntactically well formed strings. Syntactically well formed but semantically absurd strings, such as ‘Life’s but a walking shadow’ and ‘Caesar is a prime number’, can express thoughts; and competent thinkers both are able to grasp these and ought to be able to. Gareth Evans’ generality constraint, though Evans himself restricted it, should be viewed as a fully general constraint on concept possession and propositional thought. For (a) even well formed but semantically cross-categorial strings often do possess substantive inferential roles; (b) hearers exploit these inferential roles in interpreting such strings metaphorically; (c) there is no good reason to deny truth-conditions to strings with inferential roles.

I argue that we should not adopt categorial restrictions on the significance of syntactically well-formed strings. Even syntactically well-formed but semantically absurd strings, such as ‘Life is but a walking shadow’ and ‘Caesar is a prime number’, can express thoughts; and competent thinkers both can and ought to be able to grasp such thoughts. A more specific way of putting this claim is that Gareth Evans’ Generality Constraint should be viewed as a fully general constraint on concept possession and propositional thought, even though Evans himself accepted only a categorially-restricted version of the Constraint. I establish this by arguing, first, that even well-formed but semantically cross-categorial strings often do possess substantive inferential roles; second, that hearers exploit these inferential roles in interpreting such strings metaphorically; and third, that there is no good reason to deny truth-conditions to strings with inferential roles.

Having concepts is a distinctive sort of cognitive capacity. One thing that conceptual thought requires is obeying the Generality Constraint: concepts ought to be freely recombinable with other concepts to form novel thoughts, independent of what they are concepts of. Having concepts, then, constrains cognitive architecture in interesting ways. In recent years, spurred on by the rise of evolutionary psychology, massively modular models of the mind have gained prominence. I argue that these architectures are incapable of satisfying the Generality Constraint, and hence incapable of underpinning conceptual thought. I develop this argument with respect to two well-articulated proposals, due to Dan Sperber and Peter Carruthers. Neither proposal gives us a satisfactory explanation of Generality within the confines of a genuinely modular architecture. Massively modular minds may display considerable behavioral and cognitive flexibility, but not humanlike conceptualized thought.

Abstract In the first part of this essay (Russell, 1988a) I argued that ?cognisance? (roughly: a subject's knowledge of his relation to the physical world as an experiencer of it) cannot be explained in terms of a syntactic theory of mind, due to the ?referential? and ?holistic? nature of this knowledge. The syntactic account of the higher mental functions is immediately intelligible to us due to its derivation from computer technology, so this would not appear to be a happy result for scientific psychology. In this paper I outline a strategy for developing a scientific psychology of cognisance, beginning from the assumption that such a theory should have a similar status to Chomsky's competence theory of human grammar, whilst diverging from Chomsky on the question of how such a theory should be answerable to data. Most of the paper is taken up with illustrating the way in which Jean Piaget's theory of mental development in humans is at least a first approximation to such a theory.

According to the Generality Constraint, mental states with conceptual content must be capable of recombining in certain systematic ways. Drawing on empirical evidence from cognitive science, I argue that so-called analog magnitude states violate this recombinability condition and thus have nonconceptual content. I further argue that this result has two significant consequences: it demonstrates that nonconceptual content seeps beyond perception and infiltrates cognition; and it shows that whether mental states have nonconceptual content is largely an empirical matter determined by the structure of the neural representations underlying them.