Trees, pheno, tectogrammar

Roger Levy rog at stanford.edu
Fri Jul 2 18:59:27 UTC 2004


"Tibor Kiss" <tibor at linguistics.ruhr-uni-bochum.de> writes:

> a very specific remark to Carl's recent comment:
>
> > I agree the right way to think of syntactic derivations is as proofs,
> > though I don't think it is essential that the logic be substructural:
> > if you take Curry's advice (as Dowty, Reape, Kathol etc. did) to
> > separate purely combinatorial aspects of syntax
> > ("tectogrammatical" structure)
> > from how syntactic entities "surface" ("phonegrammatical"
> > structure) then
> > you can just use ordinary (intuitionistic) propositional logic for the
> > derivations.
>
> Separating tectogrammar (combination->interpretation) from phenogrammar
> leads to the assumption that linear re-ordering does not have effects on
> meaning. A version of this view also seems to be endorsed by MP people, and
> earlier by GB people who assumed that extraposition and scrambling are to be
> handled as phonological operations. (The last time I came across such an
> idea, but in the context of verb-second, is Reuland's (2001) paper on
> Binding.)
>
> BUT: This assumption is clearly wrong for a wide variety of dislocation
> operations, such as extraposition and scrambling. For extraposition, this
> has already been discussed in Culicover/Rochemont (1990), and my _Semantic
> Constraints on Relative Clause Extraposition_ is almost entirely devoted to
> promote the view that extraposition has semantic effects and is constrained
> by rules of interpretation (earlier views in the same direction are
> Wittenburg 1987 and Stucky 1987, both in Syntax and Semantics 20).
>
> For scrambling, the effects on quantification and variable binding are
> dramatic (and have been completely ignored in e.g. Kathol's thesis).

Hi Tibor (and others),

The arguments you present in your article "Semantic Constraints on
Relative Clause Extraposition" are interesting in terms of the
tectogrammar/phenogrammar divide, but they actually seem to me to
_support_ the divide rather than undermine it.  As I see it, the key
of your argument has to do with constraints in German on when an NP
can bind a pronoun within an extraposed relative clause modifying a
sister NP:

   NP_j NP_i Verb [RelC_i ... Pronoun_j ... ]
  *NP_i NP_j Verb [RelC_i ... Pronoun_j ... ]

but according to the German speakers I consulted, these binding
patterns are exactly the same whether the relative clause is
extraposed or in situ.  (Shameless self-promotion: there's a more
extensive outline of the facts and analysis in my dissertation
proposal, at http://www.stanford.edu/~rog/papers/diss-proposal.ps)

This points to an interesting middle ground in the case of pronoun
binding you examine: the phenogrammatical realization of NP argument
ordering is semantically potent, but the phenogrammatical realization
of relative clause position (in situ vs. extraposed) is not.

But in the larger picture, isn't it clear that word order variation
is, at the very least, always closely tied to the information
structure of the sentence?  So the extreme position that
phenogrammatical realization has _no_ interpretive reflex whatsoever
is something of a straw man.

The practical use of the pheno/tecto distinction in syntactic analysis
has always seemed to me as a clear way of factoring out one level of
tree-structured meaning -- basic predicate-argument structure -- from
word order phenomena that aren't easily captured with context-free
trees.  But there are lots of other semantic and pragmatic phenomena
left over, and they may be associated with predicate-argument
structure, word order, or (most commonly) a combination of both.

Roger



More information about the HPSG-L mailing list