Connectionist Psycholinguistics Book

The demands this book makes for a switch from "intuitive" data to objective, observational data in confirming or disconfirming linguistic theories is absolutely to the point. There are descriptions of several good examples of hypotheses supported with intuitive evidence that are disconfirmed by empirical evidence from corpora, including the (im)possibility of multiple center-embedding in actual performance, and limits on the "depth" of left-branching tree structures.

In addition to those welcome demonstrations of the value of evidence and the scientific method in linguistics and the unreliability of linguistic intuitions, there are polemical attacks on what the author sees as the undeserved prestige of Chomskyan generative grammar and a philosophical attack on generative semantics (which amounts to Fodor-Katz semantic markerese).

The criticisms of generative semantics don't address anything like the form of semantics that philosophers of language are most interested in: truth-conditional semantics (in either Davidsonian or Montagovian or any other form). And the attacks on the decompositional forms of semantics don't address any of the more sophisticated contemporary descendants, where there is less emphasis on decomposition and more on the need for more than worlds and individuals in the model (scales, for example, seem necessary to handle the meaning of gradable adjectives), or the idea of "conceptual spaces" as meanings, or the role of categories like CAUSE or AGENT in understanding the meaning of verbs.

The final chapter that argues that semantics is not within the scope of scientific explanation is interesting, but its arguments remain at the impressionistic, analogical level, rather than engaging in detail with current theories. That's a shame, because the author's devotion to gathering and deploying evidence could have a significant impact if directed toward contemporary debates in semantics. Even raising the question of what counts as evidence beyond the intuitions of native speakers in semantics is a relevant question to pose (though I believe there is an answer).