Report of the HPSG ADL

[Stephen M. Wechsler]

REPORT OF THE HPSG ANTI-DEFAMATION LEAGUE

On Mar. 24 I posted a query asking for cases of ignorant bigotry against
HPSG and other non-transformational frameworks.  I got about a dozen
replies, of which some highlights are briefly summarized below.  Since I
promised confidentiality, all names are removed.

1. Tilting at windmills.
A few of you who have been in the field a while made comments along the
following lines:

>This
>has been going on for over 30 years now. Any theory that deviates from the
>"Standard" theory (unless it's an earlier stage in the history of the
>Standard Theory) is dismissed as (a) missing the boat on explanatory
>adequacy or (b) a notational variant of the Standard Theory. I think this
>method of dismissal must be taught in introductory classes at MIT. In my
>experience most of the people who use this method have never read anything
>on the variant theory, and are just passing on the dismissing statements
>they hear in the corridors.

I might add a comment from a friend whose GB work I respect quite a lot (he
doesn't grab for the latest fad, and does solid empirical research).  He
pointed out to me that the same charges (not explanatory, etc.) are
frequently levelled against him and others like him WITHIN the GB world.

2. Ignorance is strength.
Also noted by several of you (see e.g. the above quote): the dismissive
comments generally come from people who know next to nothing about the
alternatives they dismiss.  I have had several experiences that definitely
corroborate that.

3. Adequacy inadequate?
One respondent rejected Chomsky's distinction between 'descriptive' and
'explanatory' adequacy as meaningless.  Personally I don't agree; Chomsky's
distinction makes sense to me and I find it useful, although the
terminology is certainly abused. Anyway, that may be a topic for a separate
discussion.

4. Flaming the framework.
Several of you commented on the tendency to incorrectly conflate theory
with framework:

>   Another common mistake made by MIT loyalists is to claim that feature
>structure theories are so powerful as to permit the formulation of
>virtually any sort of account. A well known semanticist once raised this
>objection to a talk I gave in which I argued for an HPSG account of
>raising and control verbs. The mistake is obvious. It involves confusing
>a formal framework (a feature logic or a feature geometry) with a
>particular theory formulated within the framework. Only the latter makes
>claims about the world (natural language) and imposes empirically interesting
>constraints upon phenomena. The conflation of the framework with a theory
>is analogous to objecting to relativity theory on the grounds that the
>mathematical system within which it is formulated permits the expression
>of virtually any physical theory. ...

5. Real or Memorex?
I think many of you will recognize the following phenomenon (I do, anyway):

>After I suggested to him that his notion of chains looked like a formally
>awkward (I was being diplomatic) version of a feature structure path
>with structure sharing to handle long distance dependency, he asked me
>to tell him what "all this" feature structure grammar was about. I suggested
>that he look at some representative work by Gazdar, et. al, Bresnan and
>Kaplan, and Pollard and Sag. He responded by saying that he wanted a brief
>(fifteen minute?) summary of the main ideas because he didn't have time (and
>didn't think it was worthwhile) reading this work. I had a similar encounter
>with XXXXXXX several weeks ago here at XXXX. After she presented a
>paper in which XXXXXXX, I pointed out that this relation (completely undefined
>and vague) looked like an attempt to formulate a feature structure path of the
>kind that we have been using in HPSG for some time.
>She replied that she was pleased to hear this, because Chomsky had recently
>moved in the direction of using abstract feature relations rather than
>movement chains to express syntactic relations, and so this was now
>possible.

6. Argumentation in vacuo.

>one speaker gave a
>talk showing different ERP (i.e., brain activity) patterns for raising and
>equi sentences.  He used this to argue against HPSG, which he claimed made
>no syntactic distinction between raising and equi.
...
>In the Q&A session, I pointed out that HPSG does
>make a syntactic distinction between raising and equi.  The speaker
>immediately whipped out an overhead with some quotes from Pollard & Sag
>1994.  The crucial one was part of (what I now see is)  footnote 40 on p.
>140, where they speculate about other ways the distinction could have been
>formalized in HPSG.

We discussed this and concluded that the speaker should have said that in
HPSG there is no PRO/trace occupying a phrase structure position in
equi/raising-- not that there is no syntactic distinction.  Thus the quoted
fn is irrelevant.  It seems clear the speaker did not understand what he
was reading, but this did not stop him from having a strong opinion about
it.  At least he seemed to grant that HPSG has an analysis which differs
from that of GB and which has different empirical consequences!

CONCLUSION
(exercise for reader)

--Steve

Stephen Wechsler, Assoc.Prof. |   http://uts.cc.utexas.edu/~wechsler/
Linguistics Dept.             |   Calhoun Hall 403
University of Texas           |   ph. (512)471-1701
Austin, TX 78712-1196         |   fax (512)471-4340