Larry’s Reviews > Concepts: Where Cognitive Science Went Wrong > Status Update

And IF Pinker had gotten his data straight, he could only buy claim 2, but he did not even get his data straight!

Fodor distinguishes 1) the claim that what a word means partially determines the well-formed contexts in which it can occur (perhaps equivalent to: lexical items have features), and 2) the claim that there is a single level at which all and only the semantic features of the item are specified (perhaps equivalent to: lexical items are feature bundles), and denies both

Against Pinker (1989): if there is bootstrapping from semantics to syntax, the semantic concepts doing the work like agent, possession, etc, must be detectable by the child. But those concepts are ill-defined. Consider: “red” expresses a property; is that a semantic fact, or a byproduct of the fact that it’s a modifier of nouns? Contrast Gleitman (1990), who runs the bootstrapping argument the other way around

This is a very alienating experience because I really don’t like his views but I’m incapable of forming a coherent or convincing argument against them

Putnam: you have analyticity at the edge of the web of belief, where analyticity dissolves into one-criterionhood. But that presupposes analyticity (or synonymy—which are the same thing per Quine).

The doorknob argument: if doorknob is not innate (as a concept), then it has to be learned; learnability relies on nomic relations, but… seriously? Nomic relations about DOORKNOBS? Ok, then it’s innate. But… seriously?