What Have We Learned...?
language at sprynet.com
Wed Dec 1 21:16:33 UTC 2010
I found one by-product from our recent discussion truly heartening: I
actually received just one or two off-line messages admonishing me that I
came across as an ignorant amateur, totally unaware of the depth and glory
attained by mainstream linguistics. And this of course explained why so few
true professionals had bothered to respond on-line. But even these messages
were rather friendly and apologetic in tone. So let me also be as apologetic
as I can for what follows.
This was in fact a far cry from the response I first met with back in 1991,
about five years before the Web, when using UNIX modules I crept onto the
USENET group sci.lang.translation, then a hot bed of TGG, to dispute the
various reigning dogmas. On that occasion I was roundly denounced as a rank,
untrained novice, a mere translator who had no right to post on the group at
Not to mention the uproar I provoked when I began publishing articles as
early as 1987 questioning these same dogmas in the pages of "Language
Monthly," "Language Technology," "Language International," "Sci-Tech
Translation Journal" (along with its electronic heir "Translation Journal"),
and the "ATA Chronicle." Or when I dared to introduce my complaints on that
holiest of holy TGG USENET sanctuaries, alt.fan.noam-chomsky.
And let's not forget the even more outraged cries that rang out ten years
ago when I first published my "33 Reasons Why the Chomskyans Are Mistaken"
(soon with help from Sergio Navega to become "44 Reasons...") on
language-related newsgroups and my own website ten years ago.
Consistent throughout all these complaints was the assumption by my critics
that they were true professionals, and I was nothing but a total amateur,
someone who could never even hope to realize how totally all my arguments
had long ago been demolished by TGG experts.
I found this remarkably amusing, since as early as 1990 I had begun writing
my invited paper "Limitations of Computers As Translation Tools," published
by Routledge in 1992 in John Newton's "Computers in Translation: A Practical
Appraisal," alongside papers by a number of MT and Linguistics figures.
In the second section of that essay I demonstrated in no uncertain terms
that the only amateurish arguments were in fact the ones presented by TGG
advocates. I accomplished this by applying the sole applicable standards, as
presented by Leonard Bloomfield in his essay "Secondary and Tertiary
Responses to Language."
Using Bloomfield's benchmarks, it quickly becomes clear that every single
notion in the Chomskyan canon is itself an example of rank amateurism. These
standards still apply today, as you can confirm for yourselves by examining
this piece here:
Once you have reviewed this text, I do not believe you will be able to
refute a single one of its points. As appealing as some mainstream arguments
may sound to the ignorant, no one but an amateur linguist advances the
notion that there is a universal structure underlying all languages, that
any one language is most likely to be a model for learning others, that
"deep structure" and "hard wiring" prove, taking all circumstances into
account, that the creations of one language can be held up as better, wiser,
or richer than those of another (or any other of the many mainstream
In other words, as measured by Bloomfield's standards, all those who
embraced generative arguments proved merely by doing so that they were the
true amateurs, the ones clearly ignorant of language and linguistics.
At this point there is no way of avoiding some truly imposing questions:
Precisely how could an entire school of would-be scholars embrace such a
vacuous theory, what made them coalesce around this theory in the late
Fifties, and why did these events take place at MIT?
The answer to all three of these questions can be found in a single name:
Wiener's 1948 blockbuster "Cybernetics: Or the Control and Communication in
the Animal and the Machine" was probably as available on the shelves of
would-be intellectuals--and as little read--as Stephen Hawking's books are
today. Wiener first came to MIT after World War I and remained a
commanding presence there until his death in 1967. A MIT mathematics prize
in his name has been available since that year.
As we all know, Wiener's work set the stage for everything computers have
done--or failed to do--since then, even prefiguring much of artificial
intelligence. So it was neither brave nor creative that a young would-be
linguist in his twenties would have hitched his wagon to the dominant
academic star--if anything it was simply playing safe. Nor is it remotely
surprising that a number of other young linguists would have followed his
But there was an even deeper problem: not only did these young linguists
have no real notion about the workings of language as described by
Bloomfield, but rather that they were almost equally ignorant of computers.
It was assumed at the time that sooner or later computers were capable of
solving everything (or failing that, close to everything). Even then these
machines were conceived on what seemed to be a inconceivably vast scale,
which as we know would only become geometrically more enormous.
The hardware was there, which meant to those working in the field that most
problems would sooner or later be solved. Even if their solution turned out
to be work-intensive or vastly time-consuming, once the work and time had
been invested, the solution would appear in retrospect to have been a
But hardware does not work without software, and it was in the creation of
this software that the whole mainstream project started to fall apart, even
from the very beginning. Software cannot be created without an algorithm,
which in turn must be elaborated first as pseudocode, and finally as working
code that will run without bugs on the machine. But even creating an
algorithm for "language" soon became an intractable task. Perhaps this
explains the chaotic lurching of mainstream theory over the years from TGG
to G&B to P&P to something called "Minimalism."
And finally leading to the devious decision, when these would-be linguists
finally caught on that they weren't getting anywhere with language, that
what they had really and truly been doing all along was "philosophy." But
here too amateurism may have crept in, since they remained blissfully
unaware that the true relationship between language and philosophy had been
spelled out two centuries earlier by the German physicist and wit Georg
Christoph Lichtenberg, when he wrote:
"Language originated before philosophy,
and that is what is the matter with philosophy."
In any case these savants had no chance of creating an adequate algorithm
for language, simply because the work of mainstream linguists has not been
anchored in any genuine realities of language since 1957. Wiener himself
recognized the shortcomings of the mainstream approach as early as 1966,
when he wrote about MT in his work "God and Golem, Inc.":
"I do not believe that linguistic science is so far advanced as to make a
set of rules of this sort practicable, nor that there is any prospect of its
being so advanced in the predictable future."
He was also acutely aware of how dangerous the many errors created by MT
"Short of this state of affairs, a translating machine will have a chance of
error. If any important consideration of action or policy is to be
determined by the use of a translation machine, a small error or even a
small chance of error may have disproportionately large and serious
Wiener also insisted that translators should be actively included in every
stage of creating and critiquing MT, and in many ways he foresaw the advent
of CAT and to some extent even TM systems. But he was not correct in his
supposition that back-translation by an MT system from Language B to
Language A (after another system had gone from Language A to Language B)
would necessarily uncover all errors. And as gifted a mathematician as he
undoubtedly was, he did not grasp the many problems set theory holds for all
MT or TM systems, as discussed just a few paragraphs beyond the Bloomfield
section in my paper. (Citations & summary from Wiener, op. cit., pp. 77-80,
Wiener was also deeply concerned by the contradiction between the practice
of science and funding by the military, and as early as 1947 he authored an
article in "The Atlantic," urging his fellow scientists to avoid such
We all know that a family living beyond its means may risk going into
bankruptcy. The same is true of small businesses or even, as we have
recently seen, vast commercial consortiums previously imagined as "too big
to fail." Even governments or entire nations can fail and be subjected to a
humiliating process of disbanding and total reorganization.
But what is the fate of a whole learned faculty when it has visibly and
dramatically failed in its objectives over fifty years, consuming the lives
of three generations of linguists in the process? It may be time to
seriously ask this question and formulate responsible answers. Julien Benda
came close to this topic in his famous "La Trahison des Clercs,"
translatable at various points between "The Betrayal of the Intellectual
Class" and "The Treason of the Academics." In its wholesale desertion of
provable linguistic principles and willing acceptance of DOD funding, it may
well be that an entire learned class has once again become guilty of a
comparable act of treason.
Once again, I wish to apologize if I have gone too far in describing the
negative aspects of much recent linguistics work. I realize that some of
you have had little choice in this matter and have in fact worked well and
creatively despite an unfriendly climate. I promise not to discuss this
matter any further from a negative point of view, and in my next posting
(early in 2011, if I am permitted to post it) I would like to provide a more
positive vision of how the study of language can develop in coming years and
All the best to everyone!
PS--after my first message on this topic here, I noticed a slight spike in
sales of my 60s book. Thanks to those of you who may have been responsible!
More info still available at:
The principal purpose of language is not communication but to persuade
ourselves that we know what we are talking about, when quite often we do
More information about the Funknet