LINGUIST List 3.261

Wed 18 Mar 1992

Disc: Reality of Rules

Editor for this issue: <>

Directory

1. John S. Coleman, 3.250 Rules
2. Penni Sibun, Re: 3.250 Rules
3. Stavros Macrakis, 3.250 Psychological reality of rules
4. AVERY D ANDREWS, Rules
5. David Powers, ITK Visiting Fellow, Re: 3.250 Rules

Ellen Prince asks

> uh, where exactly *are* these grammars, if not in people's brains?

In the Platonic World of Ideals, of course, along with pi, the square root of 2
 etc.

--- John Coleman

Message 2: Re: 3.250 Rules

 >From: FASOLDguvax.georgetown.edu
 >Subject: Re: 3.235 The Psychological Reality of Rules
 >
 >
 >Is psychological reality the only reality to be contemplated? I think it is
 >entirely possible that linguistic grammars might be much more elegant than
 >whatever goes on in human brains when people talk to each other.
 >
 >Ralph Fasold

 uh, where exactly *are* these grammars, if not in people's brains?

well, grammars are indubitably in *some* people's
heads/brains/minds---the heads/brains/minds of linguists. it's far
less clear whether they are in those of people in general.

				--penni sibun

Herb Simon had some interesting remarks about inferring mechanism from
regularity in an essay in his collection "The Sciences of the
Artificial". His point was that we often learn more about the
mechanism at its limits, where the regularity breaks down, than in the
regular part of its behavior. For instance, if we want to find out
how our computer does arithmetic, it is more illuminating to notice
that 8000000+4000000 gives -4777216 than that 3+4 gives 7. In the
former case, we can infer that the computer is performing 24-bit two's
complement arithmetic, whereas in the latter, it could be representing
numbers internally as decimal, one's complement binary, or as pounds,
shillings, and pence for all we know.

This is in some ways the core of the method of linguistic inquiry
looking for counter-intuitive constraints on syntax. Perhaps
phonology is just too regular to inform us much about mechanism?

	-s

David Eddington writes (Linguist 3.250, Rules)

>A wealth of literature exists which
>demonstrates that the majority of what is done in the name of
>'empirical linguistics' is not empirical at all but rather should
>be categorized along with the non-empirical sciences such as formal
>logic (see Derwing, Botha, Linell, Itkonen, Skousen).

I think this literature fails on the whole to impress people like me
because most of it demonstrates insufficient appreciation of how
remarkable it is that one can find any kind of neat patterns in
linguistic data at all. E.g., if you tried to write elegant rules
to describe the pebbles that your kid brought home from the creek,
you wouldn't get very far. And some of it is just off the wall,
such as Esa Itkonen, who seems to think that linguists have
explicit intuitive knowledge of grammatical generalizations.

On the other hand, I don't see Skousen and Linell as being in this
category, and would tend to be sympathetic to their general outlook,
on the grounds that that generativists often seem
to me not to give sufficient consideration to be possibility of
non-psychological causes for the patterns they find (especially historical
ones, for complicated systems of morphophonemic alternations). I think
we should also take seriously the possibility that the standard methods
of generative grammar won't be in themselves to answer the basic questions
of the field, just as the standard methods of chemistry are insufficient
to answer many basic questions of that field, such as the nature
of valence, & what's going on in benzene.

I believe (hope?) that most generative grammarians would believe that
external evidence is in principle relevant and to some degree necessary
to determine the extent and nature of the psychological reality of
analyses, but this agreement in principle is quite a different thing
from being convinced by particular cases.

Avery.Andrewsanu.edu.au

Message 5: Re: 3.250 Rules

[In the message entitled 3.250 Rules,
Bill Bennett (WAB2phx.cam.ac.uk) writes

> ...
> or those in "Fleet Street"who write whom it is believed' as outside the scope
> of our model and devoid of regularity?

What, under ANY model is wrong with something like "The police
identified a man whom it was believed the assailant raped"? It
may be a bit awkward without some parenthetical commas around "it
was believed", but it is a proper usage as far as I am concerned.
Or are you thinking of something like "whom it is believed was
responsible" (active) or "whom it is believed was raped"
(passive) or "whom it is believed it was" (copular). I don't read
"Fleet Street"! Does that sort of thing actually occur in the Times?
The latter three examples do, of course, jar horribly.

dP

--
Dr David M. W. Powers	 Email: powerskub.nl	 SHOE E xtraction
ITK, Tilburg University, Tel: +31-13 663116	 O f
P.O. Box 90153		 Fax: +31-13 663019	 H ierarchical
5000 LE TILBURG		 Sec: +31-13 663060 S tructure

SHOE is an international project in Machine Learning of Natural Language

Alternate email: powersdfki.uni-kl.de, davidpmacadam.mqcs.mq.oz.au