Filler-gap mismatches

[sergi balari]

Dear colleagues,

I would like to comment a bit on the issue Bob Borsley raised concerning
filler-gap mismatches.

I believe the phenomenon is quite widespread and, in fact, I suspect
that the cases Bob mentions conceerning clauses 'extracted' out of PPs
should present a similar paradigm in all languages disallowing
that-clauses as complements of prepositions. Catalan is one such
language (at least some of its dialects, including the standard
variety), unlike, for example, Spanish, which is quite happy to have P+S
sequences. Here are some examples from Catalan in (1) and from Spanish
in (2), as representative of this two putative typological classes of
languages. [By the way, I have not the slightest idea of how to deal
with this difference beyond from mere stipulation.]

(1)
     a. No els vaig poder convèncer que la Maria no ho volia
        not them AUX be-able convince that the Maria not it wanted
       'I could not convince them that Maria did not want it'

    b. No els vaig poder convèncer d'això
        not them AUX be-able convince of-this
       'I could not convince them of that'

     c. *Que la Maria no ho volia, no els vaig poder convèncer

     d. Que la Maria no ho volia, no els vaig poder convèncer d'això

(2)
      a. No pude convencerlos de que María no lo quería
          not be-able convince-them of that María not it wanted

      b. No pude convencerlos de ello
          not be-able convince-them of it

      c. De que María no lo quería, no pude convencerlos

      d. ?De que María no lo quería, no pude convencerlos de ello

Catalan is interesting here because, unlike English, it disallows
preposition stranding, so the only way to actually 'extracting' the
clause is to 'leave' a resumptive PP (or a clitic if the appropriate fom
exists). Note that Spanish allows both possibilities, namely,
'extraction' of the full P+S constituent and resumption (I'm not very
happy with (2d), but I would not say it is ungrammatical).

In the volume that Luca Dini and I edited (HPSG in Romance. CSLI, 1998),
Toni Badia has a paper where he discusses some of the intricacies of the
Catalan prepositional system and deals with some other 'exotic'
situations that arise with verbs selecting for PPs that may contain a
clause.

But the story about missmatches does not reduce to those situations
where clauses are involved. There are a number of constructions in
Romance (I focus on Catalan and Spanish only) where the category of the
'extracted' phrase does not match the selectional restrictions of the
head. Roughly, we can identify two different cases:

A.- Head selects PP and filler is NP.

B.- Head selects NP and filler is PP.

In the A cases, we get those constructions that more or less correspond
to what Guglielmo Cinque (Types of A-bar Dependencies, MIT Press, 1990)
called Left Dislocations:

(3)
     a. Juan, María dijo que no lo ayudarías  (cf. María dijo que no
ayudarías a Juan)
         Juan  María said that not him would-help

     b. Las manchas, una solución quiero  (cf. quiero una solución para
las manchas)
         the stains, a solution want

(3b) is from a TV-ad for some washing powder (so I did not make it up).
Note that in (3a) the NP to be interpreted as the direct object of
'ayudarías' is not headed by 'a' as it is the case with all human direct
objects in Spanish; in (3b), the phrase 'las manchas', which is a kind
of benefactive, is not preceded by 'para' as it is the case with
benefactives in Spanish.

As for the B cases, we find those constructions --in languages like
Catalan and Italian that, unlike Spanish, do not mark human direct
objects-- where the object is focus preposed and is preceded by a
preposition 'a':

(4) A LA MARIA, no he vist  (cf. No he vist la Maria / *No he vist a la
Maria)
     to the Maria not have seen

Similar examples could be constructed for Italian. Now, depending on how
we want to analyze clitics, clitic doubling constructions in those
languages like Catalan, Italian or French, with partitive, locative and
comitative clitics could also be analyzed as filler-gap mismatches, as
partitive, locative and comitative phrases in these languages often are
PPs.

In my own contribution the the HPSG in Romance book, I used these (and
other) examples to argue in favor of the disunity of UDCs, since they do
not present the characteristic properties of 'true' extractions. Thus, I
used the same strategy as BMS, but in the reverse: I identified an
extraction-sensitive phenomenon (subject inversion, in this case) and
set to show that it is absent from certain constructions containing
putative extractions. I concluded that these cases (including those with
clauses) should be analyzed along the same lines that Postal treats
parasitic gaps, that is as XP-NP dependencies falling within binding
theory and not as SLASH dependencies. I know this idea is not very
popular within HPSG, but....

Best,

Sergio

--
_______________________________________________
Sergio Balari Ravera
Universitat Autònoma de Barcelona
Àrea de Lingüística General
Departament de Filologia Catalana
Facultat de Lletres, Edifici B
E-09193 Bellaterra (Barcelona)
Spain
Phone: +34 935 812 350 Fax: +34 935 812 782
_______________________________________________