Constructions and Formal Semantics

I have been arguing, for instance in my dissertation, that the correctness of Construction Grammar is pretty much uncontroversial. The point, basically, is that no one has ever proposed a semantic theory for even a simple language that doesn't assume the existence of at least one linguistic construction, usually the subject-predicate construction. So in my view those guys over in Berkeley (and their followers) are on pretty solid ground. (The only way I can see to avoid this conclusion is to argue that predication, or function-application, isn't part of the semantics, but something extra.)

Unfortunately, in virtue of not taking explicit account of the role of constructions in their philosophical semantics, philosophers (and linguistic semanticists) philosophers of language have been led to, in my estimation, very implausible linguistic theses. My personal bugbear is the doctrine of logical forms, construed as a linguistic thesis. I want to be clear here that, although I am not convinced of the need for a level LF in syntax, that notion of logical form is far too weak to do the sort of work required by the sorts of robust semantic analyses posited these days. (Think, for instance, of the neoDavidsonian analysis of eventive sentences!) In order to accomodate these robust semantic analyses, the underlying logical forms would have to be vastly more complex than can be reasonably motivated on purely syntactic grounds.

So why have so many philosophers been suckered into accepting them? I'm not sure, but I wonder if it doesn't arise in part from an implicit acceptance of the Fregean view of the language-proposition relation. According to Frege (or, at least, Dummett's Frege), our only cognitive access onto propositions is via the linguistic structure of the sentences that express them. If Frege's Thesis is correct, then the need for a robust semantics will require a correspondingly complex underlying linguistic structure.

[It is also worth considering, in this quasi-historical context, whether or not Russell's notion of contextual definition and the associated doctrine of "incomplete symbols" doesn't mark out an inchoate construction-based theory of language.]

Comments

[Jason Stanley]

Your question should be -- why have so many *linguists* been suckered into accepted logical form, with rich covert syntactic structures. Once the point is put in this more adequate manner, it becomes clear you're being more than a little dogmatic.
Those philosophers who do accept rich logical forms do so, because, in taking syntax classes for many years, we've been introduced to the notion of a rich logical form with lots of covert structure (is Richard Larson in a philosophy department? Is Chomsky in a philosophy department? Pesetsky?). Robert May's book on logical form in the 1980's had a big impact on syntax and semantics, and many of us who started doing linguistics then were doing GB, and read that book. Minimalist syntax makes different assumptions than GB, and seeks to explain different evidence. But, if anything, it postulates much more covert structure.
In my experience, it's *philosophers* who are reluctant to buy linguistic arguments for covert structures.

[Jason Stanley]

Part of the problem has to do with what's meant by "purely syntactic grounds". If what you mean is, on the basis of judgements of grammaticality and ungrammaticality alone, then that is simply an oversimplistic conception of "purely syntactic grounds". For example, we distinguish bound vs. free readings of pronouns not on the grounds of grammaticality, but on the grounds that they give rise to different readings. We appeal to different potentential attachment sites of modifiers as arguments for underlying constituent structures. And so on -- so your post assumes some conception of "purely syntactic grounds" that is overly philosophical in nature.

[Tony Marmo]

Well, I have tried to discuss this issue in depth, but there is a certain degree of censure in the field of linguistics. Let try to be brief (I know that I cannot talk about all the issues involved):

If you allow me, I would like to point out that the main problem is how to derive the LF of a sentence. Here the term derivation simply means 'the formation of a sentence'. Is LF the endpoint of the syntactic derivation, as Robert May and others propose? Or is it formed in parallel with the Syntactic Structure, as Ray Jackendoff has suggested? Whether one gets a rich LF or a simpler one will depend on how linguists implement such proposals. Which is to say, it depends on what theoretic conditions restrict and shape the format of LF. For instance, given Chomsky's minimalist inclusiveness condition a very rich LF is impossible: the endpoint of the syntactic derivation cannot contain more material than the input.

Nevertheless, in trying to discuss this issue, I have found out that some linguists, especially the dogmatic representationalists, take it as a taboo. They claim that representationalism treats syntactic and semantic structures in parallel, but that is not true. It is not because they assign a representation to a sentence that they provide the right answers.

Representationalism cannot answer the simplest question that matters for syntax: how a sentence is formed. They simplify matters by assuming that in order to make a sentence one just glue things together, but in such a case there is a process of gluing to be explained. Derivationalism, on the other hand, tries to explain how things are glued or assembled.

My guess is that Syntactic Structures and LFs are distinct and manufactured parallely. To construct the Syntactic Structure of a sentence and to construe a sentence involve completely different mechanisms.

But the LF notion generative linguists often refer to is supposed to be just one first 'interpretative' step, semantically speaking. It should not contain all pragmatic details, for instance.

PS: I have posted this discussion in my blog.

[marc]

Hi Tony-

Thanks. Your post is (of necessity) pretty dense and I don't have anything to add. You indicate that you've tried to discuss this issue, if you have some references please send them along.

[Tony Marmo]

Thank you very much.

I have the intention to post a draft version of a paper of mine in my blog or in the syntax group, which is a dense piece of work too and which approaches part of the problem from the minimalist perspective. Let me mention that I am not a minimalist. I work with Minimalism in syntax for a matter of choice, and I greatly respect HPSG and Construction Grammar.

Jackendoff's Architecture of Language Faculty is the most important anti-thesis to May's LF from a derivationalist point of view. (Nowadays Jackendoff is flerting with more representational frameworks like HPSG). May in his classic work and Chomsky in the Minimalist Program argue that LF is the invisible output of syntactic derivation. But Chomsky himself admits that the Phonetic Form of the same sentence is built by a parallel phonological derivation. I have reasons to believe that the minimalist model would be more coherent if the LF of a sentence (or Conceptual Structure in Jackendoff's nomenclature) would also be formed by a parallel semantic derivation. There are also some empirical issues that suggest this is the correct direction, such as scope amibuiguities.