Q/Clf in Thai

[Joan Bresnan]

Yes, that's right.  We probably don't want to use a PRED attribute for
the quantifiers, but simply a semantic feature which is uniquely
instantiated like the PRED attribute (thinking again in classic terms
rather than Glue terms).

J.

> 'Glue logic' could also keep too many floated quantifiers from being
> produced, although you'd have to get into details of a formal semantics
> analysis of plurality to spell it out; if your thesis is due soon it might
> be best not to try to learn about glue logic right now!
>
>  - Avery Andrews
>
> On Thu, 20 Nov 2003, Joan Bresnan wrote:
>
> > Peter, thanks very much for your query.  I've made a couple of
> > observations below.  I hope they'll be helpful.  --Joan
> >
> > > Hello! I'm writing an honors thesis on Quantifier Float in Thai within LFG,
> > > and have run into an interesting problem and would truly appreciate any
> > > input you might have.
> > [...]
> >
> > > With that in mind, it wouldn't be hard to map the floated quantifier +
> > > classifier (FQ) in the f-structure. The main problem is that Thai has
> > > virtually no inflection and is thus extremely endocentric, but I don't think
> > > it would be possible to annotate both the DP and the later ClfP as having
> > > Subj f-structures, since that would violate uniqueness.
> >
> > I think this may be a misconception, if I understand you correctly.
> > It certainly is possible to annotate discontinuous constituents in
> > c-structure with the same function (GF); it is functional uniqueness
> > **in f-structure** that causes them to be merged into a single f-structure.
> >
> > >
> > > Right now, I'm leaning towards making the FQ the right branching spec of VP,
> > > but a distinct problem is that the ClfP, assuming it's a Spec, seems to be
> > > projected in both the Spec of IP and the Spec of VP, as the following
> > > sentences show:
> > >
> > > (2)  a.  dek      pai    rong rian  maj  daj   thuk khon
> > >          child    go    school      NEG  can   all  CLF
> > >          "The children all can't go to school"
> > >
> > >        b. dek      pai     rong rian thuk khon  maj daj
> > >                                      all  CLF   NEG can
> > >           "The children can't all go to school"
> > >
> > > The big problem here is that the DP is already in the Spec of IP, so it
> > > seems impossible to also have a right branching Spec from the same IP. This
> > > problem, I'm sure, will be easier to solve than the other once I think
> > > of/get more data.
> >
> > If you have multiple specifier positions, you could certainly allow
> > the ClfP to be optionally generated in any of them.  The problem would
> > be to explain why it can't be generated in all of them at the same
> > time, if it is carrying agreement information only, rather than
> > semantic features (e.g. a PREd).  If it has a PRED, then by the co-head
> > principle, the various Specifiers will get unified into a single
> > SPEC f-structure, and functional uniqueness will be violated, ensuring
> > that there is only one classifier.
> >
> > >
> > > So my current options are having the FQ be:
> > > A) another subject projected somewhere to the right of the VP in the spec of
> > > IP or VP (unlikely)
> > > B) a non-argument discourse function, kind of like TOP or FOC, again in the
> > > Spec of IP or VP (more likely, but it ignores the fact that it should be
> > > contributing directly to the f-structure of the subject)  or
> > > C) making the floated quantifier a right-branching adjunct to I' or V.' This
> > > possibility has the interesting option of making the floated quantifier
> > > endocentrically unannotated and saying that the classifier is lexocentric
> > > and mapped back to the NP (this is pretty much an ad hoc solution.)
> >
> > Optional (A) seems simplest.  Optional (C) would be unlikely, given
> > the absence of morphology in Thai.
> >
> > You might think of the crawling classifier as analogous to head
> > mobility, except that it is annotated as a specifier ( e.g. by (^SPEC)=|)
> > rather than a head (^=|).  It can be inserted in any or all of the
> > categorially available c-structure Spec positions, but the principles
> > of the theory, functional uniqueness and the unique instantiation of
> > PREds,  will limit the c-structure overgeneration, so that the
> > classifier appears in just one of the available positions.  If it is
> > obligatory, more must be said....
> >
> > HTH, TTFN--
> >
> > Joan
> >
> >
> >
> > >
> > > If you've made it this far, I'd just like to thank you again for taking some
> > > interest in my problem, and also thank you ahead of time for any insights
> > > any of you might have.
> > >
> > > -Peter
> >