silber@sbphy.ucsb.edu (02/19/89)
I have been reading "Connectionism and Symbols" (MIT Press 1988) which contains critical essays re: connectionism and cognitive systems. In their contribution, Fodor and Pylyshyn stress the inability of many connectionist models to accomodate "constituency relations". E.G. In a connectionist model with a NODE labeled "A & B" which is connected to a NODE labeled "A" and a NODE labeled "B", neither A nor B is A PART OF (constituent) of the node "A & B", WHEREAS in the classical approach using logical EXPRESSIONS CONTAINING CONSTITUENT TERMS, A and B ARE "parts" of the EXPRESSION "A & B". The cited book contains many interesting footnotes in the smallest font this side of a legal contract! I am interested in any comments/evaluations by others re:the issues raised by the contributors to that book...