I'm wrapping up with a lot of my thoughts on Chomsky's visit here and there was one thing that he briefly mentioned in both his talk and the private Q&A that really struck me; largely because this is something that I have long been considering.
Chomsky noted that all other animal communication systems are purely referential. Human language is at most minimally referential....
Yet, our standard theories of Semantics (and the ones that I believe have made the most progress) are based on an assumption of reference. Model-theoretic/set-theoretic (and most version of truth-theoretic) semantics is built out of the idea that predicates at their core refer back to sets. This seems completely unlike the way that humans actually compose meaning.
A lot of this stems from the fact that as a field, semantics really predates generative syntax. Semantics is born out of studies of logic is philosophy largely in work from the 19th and early 20th century. The marriage of logic and natural language is a messy one. For instance, even simple connectives in logical systems behave massively different in natural language. To take a very simple example:
- p ∧ q ⇔ q ∧ p
- John came in and (John) sat down ≠ John sat down and (John) came in
English and has a temporal ordering that logical and does not have. This is not a fatal blow for such systems, but it something that should at least give us pause. Logic was not built for natural language, we should never forget that.
Similarly, the problem of reference is a real one for philosophers. Yet, it is largely one that does not trouble humans who have not taken Philosophy 101. Sure transporter accidents and other worlds where Hitler gets accepted to art school are interesting thought problems, but they are problems (in my mind) largely divorced from natural language.
Even proper names seem largely divorced from reference (though there are some fairly obvious seeming referential uses: introductions for instance). Rather, most instances of proper name usage instead seems to be referring to a collection of mental attributes about said individual. If this weren't the case then the sentence "I am going to be Heidi Harley in class today" would be exceptional strange barring a world where body-snatching/shape-shifting is possible. When I utter a sentence like that I obviously mean that I well behave as if I were Heidi Harley (teach about DM, be too nice, eat durian, or whatever). Provided a shared set of attributes between the speakers such sentences are easily interpreted...
So how do humans actually compose meaning? I really don't know. Obviously, if reference can't work something mental would be more on track. But that leads us down a path of Fodorian and anti-Fodorian thoughts on concepts-- there is no simple solution. But I believe that somewhere in this thorny, thorny path the answer will be found.
I wish I had a more positive note to end on. Though, I actually do think there is a good deal to be excited about there. I do believe that the next great frontier in linguistics in word-meaning (or in my humble opinion--root meaning) and its contribution to compositionality. And I mean word-meaning in a very mind-internal way, not a dictionary way.
I have two more posts in the work to wrap up my main thoughts on Chomsky's visit (plus hopefully a description on Chomsky's Q&A). I'd like to share my thoughts on Chomsky's public talk on education. There was a lot there. Mostly I will focus on his discussion of the way instruction should be built, which I absolutely loved and have tried to do (with success and failures) my whole career. But as you may now, teaching is one of my great joys in life and there was a lot of great points Chomsky made.
I'd also like to get another post about the rarely discussed 2nd factor out. Obviously, it will be somewhat tied to this one....