Arnold Zwicky's Blog

A bit of intellectual and social history, plus a lesson in the fragility of memory. Set off by the British scholar Chris Knight on his Science and Revolution website, in two postings there:

from 2/18/18, “Chomsky’s Students Recall their Time at the MITRE Corporation”

from 3/4/18, “The MITRE Corporation’s project to use Chomsky’s linguistics for their weapons systems”

I come into this because I was one of those students of Chomsky’s who worked at the MITRE Corp. (in Bedford MA), in 1963-65 (53 to 55 years ago). Some of us have been trying to reconstruct those days, for Chris’s sake but also for our own.

About Chris, from Wikipedia:

Chris Knight (born 1942) is a British anthropologist and political activist.

… Knight’s most recent book, Decoding Chomsky[: Science and revolutionary politics] is a sustained critique of Noam Chomsky’s approach to science and its relationship to politics. Its publication in October 2016 sparked instant public controversy. A reviewer [Tom Bartlett] for the US The Chronicle of Higher Education hailed it as perhaps ‘the most in-depth meditation on “the Chomsky problem” ever published’, recommending it as ‘a compelling read’ . In Britain, The New Scientist [in a review by Marek Kohn] described Knight’s controversial account as ‘trenchant and compelling.’ Chomsky responded dismissively to Knight’s book in both The New York Times and The London Review of Books.

Below I’ll look at another review of the book, by Dan Everett in Language and Cognition (2017). (Knight makes reviews of the book available on his Science and Revolution site.)

First from Chris’s 3/4 posting, which originally appeared on 1/28 (and led to the exchanges that formed the backdrop for the 2/17 posting):

The first, 2016, edition of Decoding Chomsky: Science and revolutionary politics did not explore the fine details of military research at the Massachusetts Institute of Technology. MIT is so well-known as a Pentagon-funded university that specifying the details seemed unnecessary. The only quote I cited which connected Pentagon funding to the new linguistics was this one:

Defense of the continental United States against air and missile attack is possible in part because of the use of such computer systems. And of course, such systems support our forces in Vietnam. The data in such systems is processed in response to questions and requests by commanders. Since the computer cannot ‘understand’ English, the commanders’ queries must be translated into a language that the computer can deal with. … Command and control systems would be easier to use, and it would be easier to train people to use them, if this translation were not necessary. We sponsored linguistic research in order to learn how to build command and control systems that could understand English queries directly.

This statement from 1971 seemed sufficient because the whole point of my book was not to suggest that Chomsky worked directly on military research but, rather, that he had to move mountains to avoid doing so. The mere threat of his work being used for military research was enough, I argued, to prompt him to take countermeasures. After some years, the result was that Chomsky’s linguistics had become so abstract and other-wordly that it could not be used for any practical purposes at all, let alone direct military applications.

However, since finishing my book, I have become aware of a February 1964 article from MIT’s official journal, the Technology Review.

In this, Lieutenant Samuel Jay Keyser explains that once a computer has been programmed in accordance with Chomsky’s theories, it will then be ‘endowed with the ability to recognise instructions imparted to it in perfectly ordinary English, thereby eliminating a necessity for highly specialised languages that intervene between a man and a computer.’  He continues:

In fact a great deal of work in doing just this has already been undertaken. Donald E. Walker of the MITRE Corporation, Associate Professor G. Hubert Matthews of MIT and J. Bruce Fraser have placed a significant portion of the grammar of English on a computer.

… here is the first page of Donald Walker’s [actually, a paper by Zwicky, (Joyce) Friedman, Hall (Barbara Partee as she was then), and Walker] 1965 paper, The MITRE Syntactic Analysis Procedure for Transformational Grammars, where he acknowledges Chomsky’s influence

… The specific documents which clarify Chomsky’s relationship to this entire military project are two restricted-access papers from 1963, both of which name Chomsky as a ‘consultant’ to Walker’s department at MITRE. [Zwicky, “Grammars of Number Theory: Some Examples” (1963); Zwicky & Isard, “Some Aspects of Tree Theory” (1963)]

Available on my blog:

“Grammars of number theory: Some examples” (MITRE Corp., 1963).

Stephen Isard & Arnold Zwicky, “Some aspects of tree theory” (MITRE Corp., 1963).

Zwicky, Friedman, Hall, & Walker, “The MITRE syntactic analysis procedure for transformational grammars” (AFIPS Proc., 1965).

Then from Chris’s 2/17 posting, with an emphasis on my part in these events:

Several of the linguists employed at the MITRE Corporation between 1963 and 1965 were kind enough to email their recollections to me in January and February 2018. I was particularly interested in how scientists whose passion was developing linguistic theory, not military technology, felt about working for the Air Force. My impression is that, for understandable moral and political reasons, they tend to play down the Air Force’s stated mission to apply insights into language structure to military problems of weapons command-and-control.

Extended comments from Barbara Partee, then:

Other linguists working at MITRE agreed with much of Barbara’s version of events. Arnold Zwicky added that it was ‘colonels we had to impress, not generals’ and that he also never heard Chomsky talk politics in his lectures. He also said:

I too was astonished to discover [Chomsky’s intense political] attitudes (but then became an anti-Vietnam protestor). But it’s part of Noam’s intense, passionate belief that any position he comes to is the only possible one and that those who believe otherwise are deluded or even evil. (recall the Skinner review.)

Arnold also said:

Believe it or not, Noam really was a consultant. [He] came to MITRE several times to look at the English grammar Barbara Hall (Partee) and I and our group were working on.

And that:

When Noam visited, he got a visitor’s badge, which committed him to be under the care of a staff member at all times (and stay within the appropriate clearance-color lines on the floor.) I remember this because on at least one of his visits I was the responsible person, and I had to accompany him to the men’s room. ‘At all times’ was meant quite literally.

Further comments from Haj Ross, Tom Bever, Bruce Fraser, and Jay Keyser (as an Air Force lieutenant, Jay was our liaison with the Air Force, a position later assumed by Bruce).

In between Chris’s two postings, an assortment of MITRE veterans conferred amongst ourselves, trying to establish who worked at MITRE, in which years, on what projects, who shared offices with whom, and so on. Many of our first recollections turned out to be flat wrong: some people who were never involved with MITRE were inserted into memories, and others were totally forgotten. We disagreed about who said what and did what. As I point out often on this blog, memory is a very fragile thing.

In any case, the MITRE summer group was fairly large, comprising linguists, psychologists, mathematicians, philosophers, computer scientists, and anthropologists (at least; I might have forgotten some people). Created and led by Don Walker, who was an able adminisrator, but more important, had an extraordinary ability to bring people together and quietly encourage them to do really interesting work.

Not very long ago, Chris wondered (in e-mail)

how this relationship with MITRE came to an end for you and other linguists associated with MIT. Did it all just stop in 1966 or when? Were the researchers at MITRE still contacting you, Noam or others at MIT into the late 1960s or even 1970s, or had they lost interest in TG by then?

My reply:

as Barbara has now suggested, mostly the summer staff just dispersed, in 1965 or not long after, to academic positions in linguistics, philosophy, computer science, or psychology.  Barbara went to UCLA, I went to Illinois, Joyce went to Michigan, etc. And Don Walker, who created the whole wonderful business, soon went to SRI in Menlo Park CA, where he set up a fresh nurturant place for computational linguistics.

On the arc of Don’s career, in Current Issues in Computational Linguistics: In Honour of Don Walker, ed. by Antonio Zampolli, Nicoletta Calzolari, & Martha Palmer (Springer, 1994), pp. xxi – xxv, “Donald Walker: A Remembrance”, with contributions from Robert Amsler, Woody Bledsoe, Barbara Grosz, Joyce Friedman, Eva Hajičová, Jerry R. Hobbs, Susan Hockey, Alisair Holden, Martha Palmer, Fernando Pereira, Jane Robinson, Petr Sgall, Karen Sparck Jones, Wolfgang Wahlster, and Arnold Zwicky.

On to excerpts from Dan Everett’s review of Becoming Chomsky, reproduced on Chris’s website on 10/19/17 (in the interests of fairness, I note that Dan has tangled publicly with Noam on several occasions):

Chris Knight, an anthropologist at University College London, has produced a well-written, thought-provoking, and controversial examination of the interaction of politics and science in the work of Avram Noam Chomsky, the most dominant figure in linguistics, cognitive science, and philosophy from the latter half of the twentieth century. In linguistics Chomsky’s influence is unique. Perhaps only Charles Darwin in biology has been equally influential in a single field of study. This means that, like Darwin, linguistics for a long period at least, has been constrained by Chomsky’s work and ideas. For many years research could be classified as either advancing it or criticizing it (though for a growing number of linguists, psychologists, and others these days, I suspect that Chomsky’s work has simply become irrelevant [AZ: I was myself an early apostate]). Some believe that Chomsky’s influence has been detrimental. Others believe it has transformed linguistics from a pre-scientific exercise in taxonomy into a genuine science, on solid intellectual footing for the first time in its long history.

Knight’s purpose is to explain, describe, and criticize Chomsky’s nativist, non-sociological theory of the nature, origins, and use of human language, and to draw special attention to what he takes to be an enormous disconnect between Chomsky’s politics, in which social engagement is crucial, and his science, in which it is to be avoided at all costs. Knight agrees with Chomsky’s politics, but disagrees with his linguistics. This is in part because the former concerns itself with the socio-cultural existence of Homo sapiens, whereas the latter ignores it.

This is a courageous project for at least two reasons. First, the history of the field of linguistics is something every linguist has a slightly different take on, so anything spoken against received narratives is going to provoke a chorus of disagreements. Second, any criticism of Chomsky’s work is almost always greeted by the refrain that “This person doesn’t understand Chomsky”. And indeed, Decoding Chomsky has faced vociferous criticisms of both types.

As an anthropologist, Knight enjoys a vantage point that many historians of science lack, an understanding of the role of culture and society in shaping individual and group thought, not only in science but in life more generally. At the same time, not being a linguist brings at least one disadvantage: Knight fails to provide convincing examples to the effect that formal linguistics cannot account for its own primary empirical focus, grammar, without a theory in which socio-cultural constraints are causally implicated.  After all, Chomsky must be evaluated based on the goals he himself sets for his theory, not on what others think his goals should be. Therefore, Decoding Chomsky would have been more effective had it included more discussion of the empirical shortcomings of Chomsky’s views that devolve from the failure to engage culture and society. Merely juxtapositioning Chomsky’s different approaches to the separate issues of US foreign policy and the nature of grammar does not demonstrate that Chomsky is mistaken to include social considerations in the former and ignore them in the latter. This is the principal flaw in an otherwise wonderful book. But, by any standards, the book affords an interesting, rich, and well-argued perspective on Chomsky’s development and subsequent intellectual influence.

… The first chapter of the book, ‘The revolutionary’, explores the idea that Chomsky’s work is a sterling example of “disruptive innovation” in the 1950s. This chapter is important for a number of reasons. But its principal significance is that it clarifies some of the myths surrounding Chomsky’s ascent, a common one being that he sprung Minerva-like from the forehead of science, a fully formed intellect, presenting a new perspective to the world that had not been available before he appeared. In fact, the truth is more mundane. Chomsky was a brilliant reflection of the zeitgeist – the beginning of a computational and cognitive subculture – in which he was raised and worked. His math came from others, as Knight points out (and is otherwise well known). Some of his leading linguistics ideas came from others (X-bar theory and Transformations came – in slightly different form – from his thesis advisor, Zellig Harris). Whenever we idealize “geniuses”, we forget the social nature of knowledge and progress. As I say elsewhere (Everett, 2017b):

But what is an invention? It is a creation of culture. Edison did not invent the light bulb. He needed Franklin’s work in electricity nearly two-hundred years before him. No one person invents anything. Everyone is part of a culture and each other’s creativity, ideas, earlier attempts, and the general world of knowledge in which they live. Every invention is built up over time, bit by bit.

… From the outset, it is worth mentioning that Knight’s narrative matches a first-hand account relayed to me by someone who was there.

The late Hu Matthews, a former professor of linguistics at MIT cum missionary linguist, recounted to me, about thirty years ago, his version of how Chomsky came to MIT in the late 1950s [this is Dan’s recollection of Hu’s recollection; remember that memory is fragile]. As he told it, Victor Yngve had taken over the direction of the research on Machine Translation from Yeshua Bar-Hillel.

In Matthews’ narrative, Yngve was already working with Morris Halle when he began searching for additional linguistic talent for the project. Zellig Harris strongly recommended Noam Chomsky and Hu Matthews. Hu told me that it was obvious from the first that Chomsky was of high intelligence, even for the MIT crowd. He worked on the project, focusing on his theory of syntax, but, according to Hu’s recounting, would say to others that machine translation was not going to work. Another person working there was Robert Lees. He started using less discretion in repeating Chomsky’s assessment of the possibilities of machine translation to others and, according to Matthews, was consequently terminated by Yngve. But he was then taken on as a PhD student by Chomsky once the linguistics program began, shortly thereafter, in 1961. [This sentence is inaccurate in several ways; see correction below.] From this point, Halle and Chomsky together founded the most important linguistics department ever established, judged by influence in linguistics hirings around the world and citation indices.

Chomsky’s methodology has always been based on the idea that native speakers’ intuition is the best source of data on their grammars. Such intuitions, ex hypothesis, provide better data, for example, than is possible for a nonnative- speaker field researcher from another culture. Clearly native speakers enjoy advantages. But when any methodology relies on intuition, in particular when Generative Grammar ignores standard social science research methods, as it has by and large, it runs the risk of becoming less closely driven by the facts.

From Wikipedia on Lees:

Lees went to the Massachusetts Institute of Technology in 1956 to work on its machine translation project. He first came to notice with an influential review of Noam Chomsky’s Syntactic Structures (1957), and his 1960 book The Grammar of English Nominalizations [the published version of his 1959 Ph.D. dissertation]. Lees was later dismissed by Victor Yngve from his research position, as he had wanted to continue working on straight linguistics rather than on machine translation. He then enrolled in the electrical engineering department at MIT, where he obtained his PhD in linguistics under Chomsky. [I believe that, though Chomsky was Lees’s de facto adviser, Chomsky was not yet accredited by MIT to direct PhD dissertations, so Murray Eden in electrical engineering served as his de jure thesis director. An entertaining fact is that Roman Jakobson, who held a professorship at MIT as well as Harvard, was also on Lees’s oral exam committee.]

Lees then went to the University of Illinois, Urbana-Champaign, where he became the first head of the new linguistics department there. And, in 1965, hired me.

Note on the research community connected to MIT in those days: at MIT itself, Vic Yngve’s group; the umbrella organization overseeing sponsored research at MIT (the Research Laboratory of Electronics); MIT’s Lincoln Laboratory on Hanscomb Air Force Base in Bedford; and BBN Technologies (originally Bolt, Beranek, and Newman) in Cambridge.

 

This entry was posted on March 26, 2018 at 4:01 pm and is filed under Computational linguistics, Language and politics, Linguists, Memory, My life. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.