This part of your assignment for Thurs. (in addition to completing Churchland and Sejnowski—to guide) is a relatively short selection from a relatively long paper. As was the case with the selection from Block, the editors have captured the heart of the argument. However, you may feel that some of the context is missing, and I’ll provide a little of that. (The page references below are to a reprint of the paper in Macdonald and Macdonald, Connectionism, Blackwell, 1995, pp. 90-163; hereafter “MM.”)
• First, Fodor and Pylyshyn are assuming a commitment to what they call “representationalism.” In their words,
They oppose this to a view according to which “the appropriate vocabulary for psychological theorizing is neurological or … in any event, not a vocabulary that characterizes mental states in terms of what they represent” (ibid.). They call the latter view “eliminativism,” and it would be fair to suppose that they have in mind something like what you’ve seen in Paul Churchland; however, they hold that connectionists are largely representationalists.
• They assume further that these representational states could function as a distinct “level” that has its own “causal structure” constituting an “architecture” that can be studied in relative autonomy from other levels (MM, p. 95). And they say, “It is, in particular, perfectly possible that non-representational neurological states are interconnected in the ways described by connectionist models but that the representational states themselves are not” (MM, p. 96).
• They take what they call the “classical” position (the one they defend against connectionism) to be characterized by a view that representational states are structured not only by causal relations but also by structural relations with a syntactic, language-like, character and that cognitive operations apply in a way that is sensitive to this structure. They say that an approach that does not assume such a syntactic structure treats representational states as “atomic” or “punctuate” (i.e., point-like). The first selection you will read (from §2.2 of the original paper) addresses this issue.
• The remainder of your selection comes from Fodor and Pylyshyn’s arguments against connectionism (§3 of the original; your selections are from MM, pp. 116, 119, 121f, 131f). This argument cites four closely related properties they see cognition as having. You have selections from their discussions of the first two (“productivity” and “systematicity”) followed by the final three paragraphs of their summary of the argument. In the first of these discussions, they speak of expressions being built up “recursively” or forming a “generated set.” The idea here is that expressions can be formed by repeated (i.e., recurring) application of a limited number of operations. For example, arbitrarily complicated algebraic expressions can be formed from numerals and variables by repeated use of the operations of addition and multiplication.