Filler-gap mismatches

[Robert Levine]

Re Yehuda's point---

>The filler-gap mismatches which the subject line of this thread refers
>to are a case in point. They are old news for LFG, and not considered
>a problem.  However, they are apparently (at the very least) not as
>well known in HPSG, and I would imagine even less known in
>derivational circles.


There is *no* formal problem that such mismatches present in HPSG, as
long as it is understood that lexical entries of the form X[SLASH
{Y}]] do not have to be derived by either lexical rule or some kind of
relational contraint on VALENCE/ARG-ST/SLASH specification, a la BMS,
that links such entries to other entries of the form X[COMPS <...,
Z[LOC X], ...>, ARG-ST <...Z[LOC X, SLASH {}],..>, SLASH{}]. We have
examples of corresponding failures of correspondence between actives
and passive, and no one has the slightest difficulty with the idea that
some V[pass] entries are just one-offs in the lexicon, without
matching V[fin] forms. Cases like

       Robin is someone who I can assure you __ to be absolutely
       honest and diligent.

       *I can assure you Robin to be absolutely honest and diligent

(noted, I believe, in some relatively recent work of Paul Postal's)
are yet another kind of instance of the same phenomenon; the idea that
there is an entry for `assure' of the form (assuming the BMS story
about SLASH termination for the moment)

   assure[SLASH {[1]}, COMPS <[3], VP[inf, SUBJ <[2][LOC [1]]>]

but no form

   assure[SLASH {}, COMPS <[3], [2][LOC [1]], VP[inf, SUBJ <[2]>]>]

doesn't seem especially odd, does it? Is it any different from what
you have to deal with in the case of `Robin is said to be a villain by
everyone who knows her/*Everyone who knows her says Robin to be a
villain'? Items like `assure' exist, are relative few and far between,
and seem to be exactly what you'd expect given any approach to
extraction which took SLASH to be a property of lexical heads
(including, but not restricted to, the BMS story about
extraction). Examples of filler/'gap mismatches are mentioned in
P&S94, and most of the people I know who work on UDCs in HPSG are
aware of them. I don't see how they bear on any empirical distinction
between LFG hypotheses about extraction vs. HPSG hypotheses...

Bob