nl-kr-request@CS.ROCHESTER.EDU (NL-KR Moderator Brad Miller) (11/03/88)
NL-KR Digest (11/03/88 00:23:34) Volume 5 Number 25
Today's Topics:
Re: References on Writing
Re: incorporation, prediction, and explanation
Submissions: NL-KR@CS.ROCHESTER.EDU
Requests, policy: NL-KR-REQUEST@CS.ROCHESTER.EDU
----------------------------------------------------------------------
Date: Mon, 24 Oct 88 19:02 EDT
From: Don Steiny <steiny@hpcupt1.HP.COM>
Subject: Re: References on Writing
Reading Research Quartly has psycho-linguistics studies.
------------------------------
Date: Wed, 26 Oct 88 23:01 EDT
From: Greg Lee <lee@uhccux.uhcc.hawaii.edu>
Subject: Re: incorporation, prediction, and explanation
From article <161@gollum.UUCP>, by rolandi@gollum.UUCP (w.rolandi):
" If I could interest you momentarily, I would like you to distinguish
""predict" from "explain". Is there any causal analysis associated with
" predictive syntactic theories?
Other than to note that I did not use the term 'explain' in this
discussion, I'll take a rain check on the first matter.
There is no causal analysis associated with the theory under
discussion, cfpsg, so far as I know. I sure wish there were.
I don't think there is for any general syntactic theory currently
being pursued.
There is something to be said, however, for a theory not being
pursued -- regular grammar, i.e. Chomsky type 3, equivalent to
finite state automata. For instance, the difficulty in natural
languages with center-embedded constructions, and the fact that
the two sorts of regular grammar -- left-branching and right-
branching -- correspond to the two language types head-final
and head-initial, seem to support the possibility that natural
language is regular. And of course there is no conflict with
the context-free theory, since every regular language is context
free.
Then, if Chomsky was right in his contention in syntactic structures
that the stimulus-response model is equivalent to the finite
state automaton model, we might have a causal analysis available.
At least, if behaviorism provides a causal analysis. I'll leave
that last step to you.
Greg, lee@uhccux.uc.hawaii.edu
------------------------------
Date: Thu, 27 Oct 88 22:44 EDT
From: w.rolandi <rolandi@gollum.UUCP>
Subject: Re: incorporation, prediction, and explanation
In response to Greg's:
>Then, if Chomsky was right in his contention in syntactic structures
>that the stimulus-response model is equivalent to the finite
>state automaton model, we might have a causal analysis available.
>At least, if behaviorism provides a causal analysis. I'll leave
>that last step to you.
But why to me? Behaviorism does indeed provide a causal analysis of
language. One theory is well described in Skinner's Verbal Behavior.
Before I would presume to take up your challenge, I think you should
at least provide me the benefit of your critique of that work. Tell me,
specifically, where in Verbal Behavior did Skinner go wrong? Where
does he fail to describe language in terms of its causes?
I cannot address Chomsky's contention: I don't know what he
knows about finite state automatons. On the other hand, I would be
willing to bet that it is more than he knows about the "stimulus-response
model".
Walter Rolandi
rolandi@ncrcae.Columbia.NCR.COM
NCR Advanced Systems Development, Columbia, SC
------------------------------
Date: Fri, 28 Oct 88 12:52 EDT
From: Greg Lee <lee@uhccux.uhcc.hawaii.edu>
Subject: Re: incorporation, prediction, and explanation
From article <162@gollum.UUCP>, by rolandi@gollum.UUCP (w.rolandi):
" In response to Greg's:
" > ... I'll leave that last step to you.
"
" But why to me?
Because you know about behaviorism, and I don't.
" Behaviorism does indeed provide a causal analysis of
" language. One theory is well described in Skinner's Verbal Behavior.
" Before I would presume to take up your challenge, I think you should
" at least provide me the benefit of your critique of that work. Tell me,
" specifically, where in Verbal Behavior did Skinner go wrong? Where
" does he fail to describe language in terms of its causes?
Let's go into the merits of Verbal Behavior and Chomsky's review
of it some other time. It seems to me you haven't caught my gist:
Behaviorism uses S-R model.
S-R model equivalent to finite state automaton model.
FSA model predicts natural language is regular.
Regular language cannot have center embedding.
Therefore, behaviorism predicts natural languages will
not have center embedding, which is (more or less) true.
Prediction confirmed -- evidence favors behaviorism.
Phrase structure grammars for regular languages must be
of one of two types: left-branching or right-branching.
Head final languages are left-branching, head initial
languages are right-branching, and vice versa.
Therefore, behaviorism predicts natural languages will be
be head final or head initial (not mixed), which is
(more or less) true.
Prediction confirmed -- evidence favors behaviorism.
And I could have added other evidence. For instance, if a language
is right branching one would expect to see right branching
constructions substituted for expected left branching one, thus
one predicts extraposition in right branching languages. (Argument
stolen from Yngve, "A Model and Hypothesis for Lang. Beh.")
So, you see, I'm defending behaviorism, for the moment. You're
the one who expressed interest in finding causal analyses, and you're the
expert in behaviorism. If the above is a causal analysis, I leave
that to you to decide or to demonstrate. That's all I was saying.
" I cannot address Chomsky's contention: I don't know what he
" knows about finite state automatons. On the other hand, I would be
" willing to bet that it is more than he knows about the "stimulus-response
" model".
Time to change sides. You want Chomsky to be right about this.
Get it? Try to think of the matter as a collective effort to get
at the truth of matters, instead of warring personalities.
Greg, lee@uhccux.uhcc.hawaii.edu
------------------------------
Date: Sat, 29 Oct 88 19:27 EDT
From: w.rolandi <rolandi@gollum.UUCP>
Subject: Re: incorporation, prediction, and explanation
In response to Greg's:
>Let's go into the merits of Verbal Behavior and Chomsky's review
>of it some other time. It seems to me you haven't caught my gist:
> Behaviorism uses S-R model.
> S-R model equivalent to finite state automaton model.
I confess that I am less interested in your gist than in ridiculing
what passes for predictive evidence in the rigorous science of
linguistics. I was hoping to engage a discussion regarding the role
prediction plays in causal explanations. I entered this foray only
because I don't like to see a term like "prediction" used in a
scientifically meaningless way.
>Time to change sides. You want Chomsky to be right about this.
>Get it? Try to think of the matter as a collective effort to get
>at the truth of matters, instead of warring personalities.
In the interest of a collective effort to get at the truth, I suggest
that it's more than personalities that are at odds. It's is ultimately
a matter of how one defines knowledge. As far as I am concerned, one
cannot demonstrate knowledge of a given phenomena until they can accurately
predict at least some of the parameters of that phenomena. What
predictions about verbal behavior can be made on the basis of any
syntactic theory? Can any syntactic theory tell us why people say
the things that they say?
Walter Rolandi
rolandi@ncrcae.Columbia.NCR.COM
NCR Advanced Systems Development, Columbia, SC
------------------------------
Date: Mon, 31 Oct 88 00:13 EST
From: Greg Lee <lee@uhccux.uhcc.hawaii.edu>
Subject: Re: incorporation, prediction, and explanation
From article <163@gollum.UUCP>, by rolandi@gollum.UUCP (w.rolandi):
" ... I entered this foray only
" because I don't like to see a term like "prediction" used in a
" scientifically meaningless way.
Hmmm. Clay Bond says my favorite theory is wrong because it
predicts there is no cross-nesting in natural languages, but
there is -- in Dutch. And you say I haven't used "prediction" in
a meaningful way. It's confusing. Can what I said be both
wrong and meaningless?
" ... As far as I am concerned, one
" cannot demonstrate knowledge of a given phenomena until they can accurately
" predict at least some of the parameters of that phenomena. What
" predictions about verbal behavior can be made on the basis of any
" syntactic theory?
I've given several. Weren't you listening?
" Can any syntactic theory tell us why people say
" the things that they say?
No. At least, the theories mentioned in this discussion do not make
that kind of prediction. That doesn't mean they don't make predictions.
Greg, lee@uhccux.uhcc.hawaii.edu
------------------------------
Date: Mon, 31 Oct 88 05:40 EST
From: Celso Alvarez <sp299-ad@violet.berkeley.edu>
Subject: Re: incorporation, prediction, and explanation
In article <163@gollum.UUCP> rolandi@gollum.UUCP (Walter G. Rolandi) writes:
>What predictions about verbal behavior can be made on the basis of any
>syntactic theory? Can any syntactic theory tell us why people say
>the things that they say? ^^^
I don't think so. But neither can a pragmatic theory or a
socio-interactional theory tell us 'Why?'. They can predict, though,
what can be said and how it can be said. That's behavior.
Celso Alvarez
sp299-ad@violet.berkeley.edu
------------------------------
Date: Mon, 31 Oct 88 06:38 EST
From: Clay M Bond <bondc@iuvax.cs.indiana.edu>
Subject: Re: incorporation, prediction, and explanation
Greg Lee:
>Hmmm. Clay Bond says my favorite theory is wrong because it
>predicts there is no cross-nesting in natural languages, but
>there is -- in Dutch.
Hmmm. There may be cross-nesting in Dutch, but there definitely
is in Swiss German. See Stuart Shieber's article in Linguistics
and Philosophy (right now I don't have a more complete reference),
_Evidence against the non-context-freeness of natural language_;
or GKPS, fn.7, p 16. And I don't say it on my own authority, but
on that of Gazdar, Pullum and Sag. (And I might add that GPSG
was also "my favorite theory" until the Institute at Stanford.)
>" Can any syntactic theory tell us why people say
>" the things that they say?
>No. At least, the theories mentioned in this discussion do not make
>that kind of prediction. That doesn't mean they don't make predictions.
Indeed. My question is: can any theory of any kind tell us why people
behave the way they do? I seriously doubt it, and don't see it as much
of a criticism of PSGs.
Greg, if you are fond of GPSG, have you looked into the Unification-based
grammars, particularly HPSG? See Pollard and Sag, 1987. Information-based
syntax and semantics. CSLI lecture note series, number ???14??? ...
--
<< Clay Bond, IU Department of Leath-er, er, uh, Linguistics >>
<< bondc@iuvax.cs.indiana.edu AKA: Le Nouveau Marquis de Sade >>
<< {pur-ee,rutgers,pyramid,ames}!iuvax!bondc
------------------------------
Date: Tue, 1 Nov 88 08:24 EST
From: w.rolandi <rolandi@gollum.UUCP>
Subject: Re: incorporation, prediction, and explanation
In response to Celso's:
>>What predictions about verbal behavior can be made on the basis of any
>>syntactic theory? Can any syntactic theory tell us why people say
>>the things that they say? ^^^
>I don't think so. But neither can a pragmatic theory or a
>socio-interactional theory tell us 'Why?'. They can predict, though,
>what can be said and how it can be said. That's behavior.
I agree. But if linguistics were to address the question of the
circumstances under which things are said, the predictions of linguistics
might be of interest to people other than linguists.
In response to Clay's:
>Indeed. My question is: can any theory of any kind tell us why people
>behave the way they do? I seriously doubt it, and don't see it as much
>of a criticism of PSGs.
I welcome your scepticism but suggest that you look into the literature
of applied behavior analysis. If you do, you'll find a technology for
measuring and modifying behavior. I encourage you to doubt what you read,
to design and conduct your own experiments. But if you do, you will see
that behavior can be engineered by manipulating environmental events.
When you are satisfied that some behaviors are determined by environmental
manipulation, you might start to wonder about the causes for all types
of behavior, including complex social behaviors like language. By this
time, you will be ripe for Verbal Behavior.
Who else but,
Walter Rolandi
rolandi@ncrcae.Columbia.NCR.COM
NCR Advanced Systems Development, Columbia, SC
------------------------------
Date: Wed, 2 Nov 88 08:11 EST
From: Greg Lee <lee@uhccux.uhcc.hawaii.edu>
Subject: Re: incorporation, prediction, and explanation
From article <14519@iuvax.cs.indiana.edu>, by bondc@iuvax.cs.indiana.edu (Clay M Bond):
" Hmmm. There may be cross-nesting in Dutch, but there definitely
" is in Swiss German. See Stuart Shieber's article in Linguistics
" and Philosophy (right now I don't have a more complete reference),
I got my references confused. There is an article on cross-nesting
in Dutch by Stan Peters and two other authors ...
" _Evidence against the non-context-freeness of natural language_;
" or GKPS, fn.7, p 16. And I don't say it on my own authority, but
" on that of Gazdar, Pullum and Sag. (And I might add that GPSG
" was also "my favorite theory" until the Institute at Stanford.)
Well, the case against context freeness can be and has been made on the
basis of cross-nesting. It's not really clear to me that the
constructions in question are fully general -- that is, not limited to a
small number of "nests" -- and are limited to exactly the cross-nested
type. If not the latter, the arguments are subject to the same
criticism that was leveled against the original "respectively" argument
for non-context-freeness due to Bar-Hillel and Shamir.
Just between us, there is a much worse problem with the context
free theory than the failure to describe cross-nesting, which is
a rather marginal phenomenon if it exists, and that is grammatical
agreement. To describe agreement in a context free grammar, one
must create separate categories for each combination of agreement
features. In English, e.g., to describe subject-verb agreement,
one needs categories: singular-noun-phrase, plural-noun-phrase,
singular-agreeing-verb, plural-agreeing-verb. Thus one predicts,
by the same principle that serves to describe the coordinate
structure constraint, that singular and plural noun phrases cannot
be conjoined in English. But this is obviously false.
In GPSG the book, the authors create a descriptive mechanism that
does allow such conjunctions, but it is ad hoc. One cannot dispose
of counterexamples by inventing a notation for them. (The treatment
of agreement in GPSG is fouled up for other reasons, too, as
Pauline Jacobson noted in her review.)
I still like the theory. There are empirical problems with all good
theories. The thing to do is to look for real solutions, rather than
retreat from empirical theory into mere notational systems.
" Greg, if you are fond of GPSG, have you looked into the Unification-based
" grammars, particularly HPSG? See Pollard and Sag, 1987. Information-based
" syntax and semantics. CSLI lecture note series, number ???14??? ...
I confess I haven't. I will, real soon now.
Greg, lee@uhccux.uhcc.hawaii.edu
------------------------------
End of NL-KR Digest
*******************