LINGUIST List 2.98

Wednesday, 27 Mar 1991

Disc: Cognitive Linguistics: Last Posting on this Topic

Editor for this issue: <>

Directory

1. , Response to Pesetsky, Fromkin, and Poser
2. John Goldsmith, MT
3. Vicki Fromkin, Re: Cognitive Linguistics

(1) Is it not remarkable that, instead of arguing about the propriety
of the term 'cognitive linguistics', we are debating cognitive linguistics,
functional and non-modular approaches, the study of language by non-linguists,
and now (with Professor Fromkin's latest remarks) the nitty-gritty issue
of languages vs. language (the latter of which can supposedly only be
studied in terms of something called grammar)? Isn't it grand?

(2) Professors Pesetsky, Fromkin, and (earlier) Poser wish to bew
understood as welcoming the contributions of non-linguists. I am
sorry if I questioned the scope of this welcome, but now I am convinced,
and if any linguist ever raises the issue, I will refer to one of you
heavy guns.

(3) Professors Pesetsky and (earlier) Poser wish to be understood
as granting the viability of cognitive and functional and non-
modular approaches to linguistic THEORY. I emphasize THEORY because
it has often struck me that people will allow the legitimacy of ANYTHING
in linguistics so long as it leaves theory in safe hands. I myself,
remembering the days when linguistic theory was the responsibility
(in this country) of the leading Athabaskanist, the leading Algonquianist,
the leading Uto-Aztecanist, and (somewhat later) the leading Semiticist,
have always found this the only real problem with the state of the
field in our times. So I am gratified again.

(4) I stand corrected specifically on the issue of locality. From
now, no one need to labor under the mistaken assumption (as I did
for so long) that people may work on SYNTACTIC THEORY and BE
TAKEN SERIOUSLY even if they do not account for locality phenomena
as their first order of business. I am so sorry about the misunderstanding.
But you don't know the relief I feel now that it has been corrected.

(5) I am puzzled by Professor Fromkin's remarks about the early days
of machine translation. I take it, on a second reading, that her
point is that machine translation is impossible either without
a well worked-out linguistic theory or without a theory that is
based on autonomous syntax, modularity, etc. And that there have
been many engineers and computer scientists who have had trouble
seeing that. I agree that some kind of theory is required, but
it is not at all clear what kind. There certainly have been attempts
to base computer analysis and translation of NLs on transformational
grammar which (despite the overwhelming linguistic and psychological
evidence for the correctness of this model!) yielded essentially
nothing (although at least one such project continues). On the
other hand, the most (apparently) successful work in this area
has been based on theories and models which are heavily indebted
to non-linguists and the work itself has often been done by
non-linguists (let me recall the names of Shieber, Kaplan, Tomita,
etc., as well as the fact that all such work is very firmly grounded
in earlier work on parsing, logic, and formal grammars, almost
none of which was done by linguists). Without denying the contributions
to this area of such great names in linguistics as Pollard and Sag,
they remain the exceptions rather than the rule. And, it is also
the case that, certainly in practice and increasingly in theory
(as in the work of Karen Jensen and Nelson Correa, for example),
people who work on NLs with computers find themselves compelled to
give up many of the ideas about the organization of language,
including certainly modularity and even the old standby, the competence/
performance distinction.

(6) I was shocked to see Professor Fromkin revive the idea that
there is a contrast between the study of languages and the study
of language (the latter being according to her necessarily something
you do in terms of developing theories of grammar). As I pointed out,
the heyday of American linguistics was the time when such individious
distinctions were NOT made. And I am NOT referring to poor Joos
who incidentally is the ONLY person ever quoted on the subject of
languages varying without limit and not in a book or an article but
in a brief editorial note (if I am not mistaken). I am referring
to, need I say it, Sapir, Whorf, and Bloomfield. I read some years
ago a thing by Jonas Salk deploring the lack of communication between
theory and experimental people in his field, and I laughed, for what
he considered a problem to be remedied is apparently viewed by many
linguists as a major advance to be gloried in and perpetuated. No,
I do not agree that the study of language is different from the study
of languages, except to the extent that the former seems to involve
ONE language and the latter more than one. Nor I do agree that grammars
are the only, the preferred, or even a reasonable approach to theorizing
about language. It would be a priori surprising if they were, since
grammars were designed originally as descriptive devices, without
any trace of the kind of mechanisms that people usually postulate
when they do theory. For example, grammars do not account for real-time
generation and understanding of utterances, they do not account for those
aspects of language whose only explanation lies in history, those which
depend on anatomy, etc. Grammars depend on the performance/competence
distinction, yet certain linguistic phenomena cross this divide. For
example, the very existence of a lexical-grammatical construction
for correcting oneself (which differs in its lexical AND syntactic
form from language to language and dialect to dialect) makes no
sense if you assume that grammars are a self-contained mental box.
(The data are obvious. Some English speakers say ERROR, CORRECTION
rather; others say ERROR, at least CORRECTION. How's that for 
parameter setting?) On a strict interpretation of what is meant
by postulating grammars as mental components, there should be no
such constructions therein, since the (ideal) grammar should have
no cognizance of the fact that (real) people sometimes make mistakes
or change their minds. 

(7) Going back to the first point, it seems to that it is not
an accident that a whole set of issues got raised (and none of them
terminological, notice!), for, as I said before, the real issue
is who, if anyONE, is to set the agenda for theoretical linguistics.
I am gratified, as I said, that far abler scholars than I have
given such ringing endorsements of the pluralism which I advocated,
and that I was utterly wrong in thinking even for a moment that anyone
intended to assert the primacy of linguists over non-linguists or
of one kind of linguistics over another. 

I'm not going to get into the "cognitive linguistics" argument -- the matter
is far too complex to be treated in a few paragraphs -- but I would like to
point out to Vicki Fromkin and our group that work and workers in
MT (machine translation, aka automatic translation) have been taking it on
the chin for decades from generative linguists for reasons that escape me --
I have some speculation, but maybe Vicki can help on this score. I think
us linguists' criticism of work on MT is off the mark, way off the mark, for
three reasons: MT linguists, in quite goodly numbers, ARE linguists, just
trying to do the best job they can, using the same linguistic tools and
analysis that the rest of us have and work on. I have worked closely
with a number of MT linguists who also write papers on parasitic gaps,
Romance clitics, rule ordering, and so on. Second, and allusions to
amusing MT errors of the 1950s notwithstanding, MT in the 1990s works, and
it pays. There are well over a half a dozen working companies or
institutions that have developed systems that work every day. My general
impression is that most LSA members (i.e., us) are not aware of this.
As Chomsky might say: it's a fact, look it up. Third, while there may
have been some overblown hype thiry, thirty-five years ago in the MT
business, the same is true throughout the umbrella group of cognitive
science, and we can go on to all of the social sciences for more of the
hype. Best to evaluate the work, not the hype.
 I have a review coming out in Journal of Linguistics of Jonathan
Kaye's new textbook in phonology (a good book, incidentally) in which
I point out much the same thing, in response to some gratuitous side
remarks of Jonathan's about MT.
 So, why is there so much hostility, and so little familiarity,
among linguists concerning MT? One speculative hypothesis, and Vicki
may have something to add on this: we might recall that generative
grammar was first developed at MIT in Vic Yngve's MT lab in the early
1950s, where Chomsky, Matthews, and Lees, among a number of others,
were working. There was, for what seem in retrospect to be quite
complex reasons, a parting of the ways early on. I am under the
impression that the parting of the ways between generative grammar
and MT spring directly from this early conflict. (Interested
linguists should note that Fritz Newmeyer's brief account of this
early period does not give a complete picture of how that era is
recalled by a number of the participants.)

I apologize for starting a discussion on 'terminology'. A rose is a rose
is a etc.... I say this because not only do I recoil at what I consider
an arrogant use of the term 'cognitive linguistics' (arrogant because
it defines cognitive linguistics to exclude other views than their own)
but at the pejorative connotation attached to the term 'formalisms' and
'formal linguistics' by various colleagues. The view held by 'formalists'
(like myself although I would never call myself that) is one accepted
in science generally -- e.g. that expressed by Albert E himself in
ESSAYS IN SCIENCE p 54, 69:
 "When we say that we have succeded in understanding a gorup of
natural processes, we invariably mean that a constructive theory has been
found which covers the processes in question. ... The theoretical
scientists is compelled in an increasing degree to be guided by purely
formal consioderations in his search for a theory, because the physical
experience of the experimenter cannot lift him into the regions of
highest abstraction. The predominantly inductive methods appropriate to the
youth of science are giving place to tentative deduction."

 I quote AE because I figure he at least will be listened to, although
I am sure that there are those who feel that quoting scripture or authority
is a poor sort of argument. Also it may be inappropriate to quote from
the physical scientists if Chomsky is right in saying linguistics is in a
pre-Galilean period.(see Chomsky, LANGUAGE AND POLITICS, PP 407-19)

 ANyway, I still think that what is often referred to as 'formal
linguistics' is really 'theoretical linguistics' in the sense of the attempt
to construct a theory which is explicit and explanatory. V Fromkin