An Interview with Paul Postal
Chris Collins
April 2020
The following is an interview with Paul Postal on his career in linguistics. The interview took place via e-mail during the coronavirus pandemic, from Tuesday March 24, 2020 to the end of April, 2020. For readability, some of the questions and answers have been edited and the order of a few of the questions has been altered.
For convenience, I have divided the interview into the following segments:
1. Early Years and Background
2. Education
3. MIT (1961-1965) and CUNY (1965-1967)
4. IBM (1967 - 1993)
5. NYU (1993 - 2009)
6. Influences
7. Accomplishments
8. Foundational Issues
9. Conclusion
10. References Mentioned in Interview
A list of additional documents providing background information can be found at this link:
This interview should be cited as follows (adjusted for style):
Collins, Chris. 2020. An Interview with Paul Postal. Ordinary Working Grammarian (blog), (posted April 30, 2020, revised August 6, 2020).
1. Early Years and Background
Chris:
Where were you born? And when?
Paul:
Weehawken, New Jersey. It is a town right across from Manhattan. There used to be a ferry connecting them. It was terminated by construction of Lincoln Tunnel. I was born on November 10, 1936.
Chris:
Do you have any siblings? Male or female? Older or younger?
Do you have any siblings? Male or female? Older or younger?
Paul:
My sister is my only sibling. She is 4 years older than me and the healthiest person on earth. She has really never been sick in her whole adult life.
Chris:
Can I ask what her occupation is? Or was, if she is retired?
Paul:
Before she was married, she worked for a magazine or magazines in NYC. I forget what exactly she was doing. After she got married, she didn’t work for many years…traditional pattern. Then later she worked for the New York State Nurses’ Association; I think she was their public relations director.
As she is 87, of course she is retired.
Chris:
Did you ever talk about linguistics with your mother, father or sister? If so, what did they think about it?
Did you ever talk about linguistics with your mother, father or sister? If so, what did they think about it?
Paul:
Pretty much not. No interest, no understanding of questions involved.
Pretty much not. No interest, no understanding of questions involved.
Chris:
How many children do you have, and how old are they?
Paul:
Two; one female age 55, one male age 52.
Two; one female age 55, one male age 52.
Chris:
Did they ever express an interest in linguistics?
Paul:
None.
None.
Chris:
Where were you brought up? I remember you telling me that you moved at some point.
Paul:
We lived for a year after I was born somewhere in New Jersey close to Weehawken. Then we moved to Washington DC, for about a year I think. And then to Silver Spring Maryland, a close in suburb of DC. We stayed there until the end of the war, and in 1946 moved to Long Island, a place called Lakeview, where we lived through my years in High School.
Chris:
Do you remember the names of your elementary school, junior high school and high school?
Do you remember the names of your elementary school, junior high school and high school?
Paul:
Parkside Elementary School, Silver Spring, Montgomery County, Maryland
Woodfield Road School, Lakeview, New York
Malverne High School, Malverne, New York
There was no junior high school or middle school, the High School covered grades 7-12.
Chris:
What were your father and mother’s occupations?
Paul:
My mother had no occupation outside the home after she got married. I believe she worked for a school board and they had a rule that married women could not be employees. It was the depression and I guess the feeling was one job per family was the maximum. My father was in Jewish public relations.
Chris:
So, I guess this is him:
BERNARD POSTAL, ASSOCIATE EDITOR OF THE JEWISH WEEK AND A WRITER
Your father was a well-known intellectual. This says to me you came from a very highly educated family.
Paul:
Not really. My father never finished college; my mother never attended college. Almost none of their relatives could have been considered intellectuals.
Chris:
Are you Jewish?
Paul:
Yes, by traditional criterion. My mother was Jewish.
Yes, by traditional criterion. My mother was Jewish.
Chris:
If so, did being Jewish play any particular role in your education?
Paul:
Yes/no. My parents of course insisted I go to Jewish school for the minimum number of years. I was not interested and it had no lasting consequences.
Yes/no. My parents of course insisted I go to Jewish school for the minimum number of years. I was not interested and it had no lasting consequences.
Chris:
Did your father and mother also come from New York, or did they come from other countries?
Paul:
My father and mother both were born in the US, father in New York city, mother in New Jersey. Their parents were late 19th century immigrants from parts of eastern Europe which were sometimes in Poland and sometimes in Russia.
Chris:
What was the intellectual environment of your home like growing up?
Paul:
My father was a sort of intellectual, but at least when I knew him his interests were almost exclusively in Jewish affairs; I was not interested.
Chris:
Did you have lots of books around the house?
Paul:
He had a mass of books, thousands, some general, history, politics, little science, many novels, tons of stuff on Jewish matters, Israel, Jewish history.
Chris:
Did you father talk about the books he wrote or related issues?
Paul:
I have no clear memories of this.
Chris:
Did you read any of your father's 10 books growing up?
Paul:
Maybe one…no memory.
Chris:
Did you have political debates?
Paul:
Yes, I think so, once we got to be older. Traditionally eastern European Jewish families are very argumentative.
Chris:
Did you have relatives or family friends who were intellectuals, artists, scientists, politicians?
Paul:
Not really. My maternal grandfather was said to have been part of the Democratic political machine in Union City New Jersey (I think). But they got overconfident one election, not enough people voted and they got thrown out of office. That is when he got into the bar/liquor store business.
Chris:
What were your intellectual interests in high school? Where you more of a math/science type person, or a language/literature type person, or something else? Were there any professors or courses in high school that had an influence on you?
Paul:
I didn’t really have any intellectual interests. Note that I skipped the fifth grade and thus went to high school very young. From most points of view, I didn’t belong there. I was an immature, undisciplined and uninspired student. That leaked over into college, where I started before I was 17. Again, from most points of view I didn’t belong there.
Chris:
Were you into sports as a child? Which sports did you like?
Paul:
Somewhat, I liked baseball and played stickball whenever possible. Later I played some basketball. I was never very good at any of it.
Chris:
Do you speak, read or write Hebrew or Yiddish?
Paul:
No.
Chris:
Did you parents or grandparents speak Yiddish? Did you ever hear Yiddish being spoken by members of the family?
Paul:
My parents did not speak it. They no doubt understood a little. I know my maternal grandparents spoke Yiddish. I heard them and other relatives speaking it. I do not know about my paternal grandparents as we had little contact with them.
Chris:
Other than English, what languages do you speak? What languages have you tried to learn?
Paul:
French. I studied Spanish in high school and German and Spanish a bit in college, but I don’t think I ever really intended to learn them. I worked on Mohawk but never intended to learn to speak it.
Chris:
Did you actually take French classes, or learn it on your own? Have you lived in France?
Paul:
No, I have never taken any French class nor lived in France. I lived in Quebec for 6 months in 1960, but it is questionable whether they speak French there and I was working on Mohawk and utterly uninterested in French at the time.
Chris:
Do you realize that some Canadians might be offended at this statement?
Paul:
I could care less how many Canadians I offend. The similarities between continental French and Quebecois are substantial. But the differences are many. One can be fooled because I think the people in Quebec learn some standard French in school. But particularly in terms of pronunciation, the unschooled pronunciation is essentially incomprehensible, and lexical differences are massive.
Chris:
You had the biggest collection of tawdry paperback French crime novels that I have ever seen (actually, I have never seen another such collection). As I recall there must have been over 100 books, maybe 200. You gave them all to me, and I in turn tried to give them to the NYU students. Where did you get those novels and what did you use them for?
Paul:
That was just a drop in the bucket. Over the last 50 years I have owned and read somewhere between 4,000 and 5,000 such books. I have given away most of them, but still have say 450-500, probably 100 or more unread. Originally, I got some during trips to France, as gifts (Gilles Fauconnier once came and gave me around twenty or so by the same author), later by trips to Montreal. There was a short period where one could get some in Manhattan, but those were new and thus expensive. Once the internet was established, I got and get them on line, mostly from the site Livreenpoche, where they have endless numbers of used books. One can also download French books from various sites and read them on a kindle, which I have.
I had and have two goals: one hobby activity, two learning actual French rather than the stilted stuff taught in traditional ‘learn French’ volumes. It was also useful in getting a sense of French grammar for linguistic research purposes; I wrote a few papers and one book on French grammar.
Chris:
Did you ever use example sentences from the tawdry French novels in your papers on French?
Paul:
Definitely.
Chris:
Other than reading tawdry French crime novels, do you have any other hobbies or interests that you would care to mention?
Paul:
I would go easy on tawdry…lots of them are not. One has to understand that France had a traditionally extremely prescriptive culture with respect to language as it appeared in writing. After a while, the canons of respectability just rendered ordinary colloquial speech beyond the pale. In Tesniere’s famous grammatical work, I recall that in 1936 he characterized the use of ‘on’ as a replacement of ’nous’ as somehow improper. I gave away that book, maybe to you, so I can’t check his exact words. So, the real fact is that the kind of literature I got interested in was one which simply ignored these unrealistic prescriptive ideas.
I remember vaguely the first time I picked up a volume by an astoundingly widely-read author named Frederic Dard, who wrote mostly under the nom de plume San-Antonio, also the name of his heroic police character. Although I knew a good bit of basic French at the time, when I looked at the first page, it was mostly incomprehensible, filled with slang I had never heard of and much of which is not in standard dictionaries. But after a few years I could read such works, gradually learning more and more of the slang.
Of course, there is always a question in literature of how much vulgarity, sex, violence, etc. is acceptable…but these things change. ‘fuck’ used to be shocking. I remember the first time I heard it (in the mouth of a young woman) in a 1960s movie, it was striking. Now, movies have hundreds of such occurrences and nobody blinks.
Chris:
OK, but putting aside those novels, do you have any other hobbies or interests that you would care to mention?
Paul:
I am interested in various philosophical issues related to linguistics, but in a very amateurish way. Most of the discussions of e.g. the foundations of logic, etc. are too hard for me. But some of it is relevant to linguistics, e.g. to model-theoretic semantics. I was convinced by Katz’s arguments about that, and note that there is no real solution to the problems he pointed out (e.g. all necessary truths come out as identical, as do all necessary falsehoods). And there is another problem, which I have never seen discussed by linguists. It shows up in sentences like:
(1) Absolutely every set (except the null set) has members.
If you consider how this is supposed to be analyzed, it turns out that one needs to talk about the denotation of ’set’, which is going to be a set of sets and for (1), the set of all sets. But it is not hard to show that there is no such set. I can send you works showing that. Nonetheless (1) is a perfectly fine sentence with a perfectly clear meaning (no problem in knowing how to translate it into French, etc.).
I am also interested in political and social things, do a lot of surfing. I would never read any English fiction and haven’t in at least 40 years.
On the lighter side, I like to play pool, and in college I spent a great deal of time at it, rather than studying. I also played a lot at Yale. Since then, I have played only occasionally since I don’t have my own table.
Chris:
Did you serve in the military?
Paul:
No. I registered for the draft but we lived in an unincorporated area with lots of really poor people. I had the deferment of being in college and was never even called to take a physical. I did not volunteer.
Chris:
What year was that?
Paul:
I was 18 in 1954. So, the Korean war was over and there was no new one. But the draft continued until sometime when Nixon was president.
Chris:
Would you describe yourself as liberal or conservative politically? What political party do you belong to?
Paul:
I am very conservative. I have no political party. I think the remark of the wonderful Thomas Sowell is accurate: “The only reason to vote for Republicans is Democrats.”
Chris:
How about your parents? What were their political views?
Paul:
When I was young, I remember my father telling me: ’The worst Democrat is better than the best Republican’. He did not support that claim. Whatever its value decades ago, my view is that the opposite is now the case. There is no Democrat I would vote for for any office.
Chris:
Is it safe to say that you do not agree with Chomsky's political views?
Paul:
You are kidding right? Of course, I don’t agree with him.
2. Education
Chris:
Where did you go to college as an undergraduate? What year did you start? What was your major?
Paul:
Columbia, 1953-1957. Major: Philosophy, Anthropology
Chris:
Since you got into Columbia as an undergraduate, you must have done well in high school.
Paul:
No, not at all. I was a mediocre student. I got into Columbia because my sister got into Barnard, so I had a family connection.
Chris:
How did you end up choosing philosophy and anthropology as your majors at Columbia? Do you remember?
Paul:
No.
Chris:
When you were an undergraduate, did you take any courses that were in some way relevant to linguistics, like logic, perhaps, or something in anthropology?
Paul:
I took a logic course; but understood little. I can’t remember any anthropology course relevant to linguistics.
Chris:
Were there any professors or courses that influenced you intellectually as an undergraduate? Anything that you recall that had a deep impact on your thinking?
Were there any professors or courses that influenced you intellectually as an undergraduate? Anything that you recall that had a deep impact on your thinking?
Paul:
No.
Chris:
Where did you go to graduate school? In what field? What were the dates (beginning and end)?
Paul:
Yale, Department of Anthropology, 1957-1962/3. I left Yale in late spring 1961, finished my thesis in 1962 but too late to meet the submission deadline. So, the Ph.D. was not awarded until 1963.
Chris:
When did you first meet Jay Keyser?
[C.C., Jay Keyser was a graduate student in the Department of Linguistics at Yale from 1958-62.]
Paul:
I think fall of 1957, maybe 1958.
Chris:
Did you discuss linguistics with him at Yale?
Paul:
Hardly if at all. We were at different life stages. He was already married, and living off campus I think in a rented house. He was not much around campus. I lived in campus housing, campus dining hall, etc. Also, our interests did not overlap. He was in phonology, history of English.
Chris:
Did you ever take courses together at Yale?
Paul:
Not that I remember.
Chris:
Jay says he was in Bloch's class with you.
Paul:
Possible…I have no memory of it
Chris:
Did you influence each other's thinking about generative syntax while at Yale?
Paul:
No.
Chris:
Did any of your other fellow students at Yale go on to become generative linguists?
Paul:
I don’t think so. Recall I was in anthropology department. No one else there was interested in linguistics as far as I remember. Well, actually the question is vague. I knew students who were in the linguistics department, and some of them became professional linguists. Not sure about the ‘generative’.
Chris:
Jay sent me this list, but it is only for Linguistics, not Anthropology. Lounsbury [C.C., Paul’s dissertation advisor] is on it several times. So is Jay. I thought you might be interested. Do you recognize any of the names on the list from 61-63?
Paul:
Sure. But the astounding thing is that starting from the most recent people, one has to go back to 1974 to find someone I have heard of, Marianne Mithun, although I never knew her. Then from 1971, I of course knew and even did a bit of work with Jorge Hankamer. I have heard of Sarah Thomason from 1968, but again had nothing to do with her. Frederick Damereau from 66 is interesting. I did not even realize he had a Yale Ph. D, or if I did, I forgot. Anyway, he worked in the real part of IBM, the business part, I forget which division, for many years, then came to Research, and became a member of the linguistics group. He never did anything you or I would regard as linguistic, but he was smart, did programming type stuff, some thinking about machine translation and then worked on the group question answering system. He was very helpful to me understanding the computer system there when we were all assigned terminals in 1979. Also, I learned a lot from him about how IBM worked. He kept his habits from the business area, knew what things cost, how the regulations worked, always wore a white shirt and a tie, etc., pretty rare except for managers in Research.
I knew Percival, an English guy from Leeds; we lived in the Hall of Graduate Studies at same time. I was friendly with Alan Stevens during a period at Yale. I remember he came to visit me for a day or two when I was living in Montreal working on Mohawk, in 1960. We socialized a bit after that but I haven’t heard from or about him in decades and have no idea whether he is alive.
It is normal for people my age to be dead, especially males, as Edgar Gregersen illustrates. He was a fellow anthropology student at Yale and died quite recently although he was slightly younger than I am.
The fact that Noam Chomsky is alive at soon to be 92 is beyond amazing. Of course, I knew and know Jay Keyser. Other names from that period I have some memories about but they are not people I had anything much to do with.
In the picture, the only person I recognize is Larry Horn.
[C.C.,
1964 Alan M. Stevens A grammar of Madurese
1964 W. Keith Percival A grammar of Toba-Batak
1966 Frederick J. Damerau Empirical investigation of statistically generated sentences]
[C.C., Edgar Gregersen’s NYTs obituary can be found here:
Chris:
What was the title of your thesis?
Paul:
Some Syntactic Rules in Mohawk
Chris:
Who directed your thesis?
Who directed your thesis?
Paul:
Well technically a gentleman named Floyd Lounsbury. But in fact, I directed it. Did I not tell you that when I handed it in he wrote me a letter saying he was too busy with his ‘work’ to read it for two years. I am sure he never read it. And certainly, he never sent me any reactions. This seems like a joke today but it was not.
[C.C., Wikipedia entry on Floyd Lounsbury:
[C.C., Anthropology Tree entry for Floyd Lounsbury:
[C.C., the acknowledgments from Paul’s thesis:
“ACKNOWLEDGMENTS
I am indebted to Floyd G. Lounsbury for introducing me to the Iroquoian languages and for many hours of discussion of the complexities of their grammars. I have also profited greatly from his published study of Oneida which has served as a foundation of my understanding of Iroquoian verb structure even when I have been forced to disagree with some of its theoretical implications. To Noam Chomsky I owe my conception of linguistic description as well as thanks for long hours of discussion of the theoretical intricacies of the formal grammatical devices used in this work. I have benefited greatly from extended conversations on Mohawk syntax with G. H. Matthews. In particular, the treatment of incorporation and possessive constructions in this dissertation owes much to his suggestions. I am grateful to Morris Halle for my notions of phonological structure which underly the transcription in the present study as well as for many helpful comments on the manuscript. To my wife I owe a debt of gratitude for aid in preparation of the final manuscript as well as for innumerable aids of a more abstract sort without which its completion would have been impossible. Finally, and most importantly I must thank my Mohawk informants, in particular Mrs. Thomas Lahache of Caughnawaga, Quebec, and Mrs. Margaret Lahache of Brooklyn, New York, for their continuously cheerful cooperation which did not stop at occasionally apologizing for some of the amazing complications of their native tongue.”]
Chris:
Why did you choose him to direct your thesis? Or was he assigned to you?
Paul:
He was the linguist in the anthropology department and an Iroquoianist. I can’t imagine any question of choosing arose.
Chris:
Did you ever have meetings with him when you were working on your thesis?
Paul:
Maybe I talked to him a couple of times but in an unorganized way. I was thinking in TG terms and he didn’t know anything about that and didn’t like the movement.
Chris:
So, it is safe to say that he did not influence you intellectually. Right?
Paul:
Right.
Chris:
What framework were you using to do your thesis analyses?
Paul:
TG
Chris:
When you say TG, I assume you mean Transformational Grammar. Which books and papers on TG seemed most useful to you at the time in writing your thesis?
Paul:
There wasn’t much. Anything of Chomsky’s or Robert Lees’ I could get my hands on. I remember getting Lees’ nominalization book while I was in Canada gathering data in 1960. I have no specific memories beyond that. I can scan relevant parts of thesis bibliography if you really care.
Chris:
When did you first learn about TG? How did it happen?
Paul:
Not sure, probably from Robert Lees’ review of ‘Syntactic Structures’ in LANGUAGE.
Chris
Did you take any linguistics classes at Yale in graduate school? Either in TG or in another framework?
Paul:
Yes, I took one class from Bernard Bloch, then editor of LANGUAGE. He was a traditional scholar and I guess a structuralist of the time. The class was awful. He spent enormous amount of time on the various allomorphs of English plurals, and gave a penny reward to anyone who discovered a new one. Edgar Gregersen found one. That is my total memory of the class. Maybe I took a class from someone else, I think it was Rulon Wells. No memory of the content.
Chris:
Was this before or after you learned about TG?
Paul:
Probably more or less simultaneous.
Chris:
I am embarrassed to say that I have not read your thesis, even though you sent me the physical book. Can you summarize what it was about? I read somewhere that you proposed something like the unaccusative hypothesis there. Is that true?
Paul:
It is true that I had an unaccusative analysis, it was almost self-evident. Mohawk has rich noun incorporation, with transitive direct objects incorporating. But also, a subset of intransitives incorporate. So, in the rules I proposed the one argument NP of the intransitives which incorporated originated as a direct object. Pullum published one of his NLLT pieces about the origin of the idea, finding many sources. He cites my letter to Perlmutter, explaining the idea. Perlmutter later published (with my ok) a paper expounding it. Since then the field attributes its origin to him, I don’t think I am ever mentioned except for Pullum’s piece.
The thesis was about handling pronominal agreement (rich system for subjects and objects), possessor nominals, incorporation, broad covertness of nominals, etc.
Incidentally, the following talk I gave at NYU in 2004 might be of relevance:
Chris:
Thanks for the NYU talk. That is very useful for this interview.
Do you still have a copy of the letter to Perlmutter that you could send me as a scan?
Postal:
I don’t have an original copy of my letter to Perlmutter, but Pullum reproduced it in full in his 1988 article in NLLT on the history of unaccusative ideas. Incidentally, he is author of term. It is attached.
Chris:
Other than the unaccusative analysis, were there any other discoveries in your thesis that have been important to you, or that have been significant for syntactic theory?
Paul:
Yes, I was permanently struck in my work on Mohawk syntax by the fact that a good bit of syntactic structure is phonologically invisible.
A typical Mohawk sentence consists of one or more complex verbs, with pronominal prefixes, negative prefixes, tense and aspect suffixes, etc. with some particles thrown in. There are lots of sentences of roughly the following form, where brackets surround verbs:
Note that the orthography of my thesis was a morphophonemic one, which I developed. No Iroquoianist ever adopted it. The usual orthographies are much more phonetic and mark stuff I claimed was predictable.
From page 368:
(1) i?i [khnurnuhwe?s] nehneh [kahurhu?syi] ‘I like the black gun’
It consists of four words:
i?i = ‘I’
khnurnuhwe?s = [1stPsing-gun-like-suffix]
nehneh = particle
[kahurhu?syi] = [3Psing-gun-black-suffix]
The second and fourth word are verbs; the language does not distinguish adjective/verb the way English does.
When I first tried to think about sentences like (1), I was baffled. First of all, each verb has an incorporated noun (stem), the same one in this case, the noun stem meaning ‘gun’. The phenomenon of incorporation is characteristic of the language, as with many others in North America. But Mohawk has a feature which many do not, namely, that the incorporated noun stem in many cases can cooccur with an external noun based on the same stem, as in:
From page 291
(2) kanuhsaragv thikv kanuhsa (the ‘v’ symbol here should be upside down, to represent a nasal vowel, but I don’t know how to get that character) ‘that house is white’
This consists of a verb with incorporated noun ‘house’ nuhs meaning ‘house is white’, a form meaning ‘that’ and an external noun based on the same noun stem. The language has some nouns like that which occur verb externally, but really quite few in comparison to European languages. Lots of semantic equivalents of our nouns are verbs, e.g. ’table’ is a verbal form ‘one puts one’s food on it’, etc. Pronouns for first, second, third person are usually null, indicated by verbal agreement, for both subjects and objects. In a case like (2), the last two words could be omitted yielding a sentence just meaning ‘it is white’.
After a long period of bafflement, it dawned on me that cases like the last verb in example (1) could be taken to represent a relative clause (bracketed <…>) with a null ‘head’. So, it was really, I concluded, something like:
I […gun like] gun particle <gun [..gun black]>
The idea was that each clause had its own original noun, hence the possibility of incorporation in both clauses. Then the external noun sources of the incorporated nouns would be deleted. Today I would posit only a single noun shared by the two clauses, but I had no idea like that in 1960. The relative clause view obviously depended on a good bit of syntactic deletion, as did any general treatment of incorporation of cases like the shortened form of (2).
From the point when I began to develop concrete ideas about the treatment of incorporation and especially the cases I took to represent relative clauses, I could never take seriously any view of syntax which sort of viewed it as some variant of a direct analysis of the visible morphemes. This
made the newly available ideas of the transformational grammar of the time seem extremely attractive as they entirely lacked any commitment which blocked the sort of deletions I took to be required to describe Mohawk.
Incidentally, I found another personal talk I gave at NYU; it is attached.
Chris:
How did you decide to write a thesis on Mohawk syntax? What led up to that decision? Both the fact that the thesis was on Mohawk and the fact that the thesis was on syntax.
Paul:
I was in an anthropology department. A thesis had to be based on work with traditional sources, that is, pretty much nonwestern cultures. I did not wish to and did not have resources to go to some third world local. I lived on Long Island, and there were Mohawks in Brooklyn. Floyd Lounsbury, the linguist in the Anthropology department, was an Iroquoianist. He knew the name of someone in Brooklyn who connected me with a Mohawk in Brooklyn.
I don’t remember how I decided to do syntax, probably because I had seen enough of TG to be interested.
But I can add one thing I am very proud of. Lounsbury had another student after me named Nancy Bonvillain. She wrote a thesis on Mohawk, having worked on a different reservation. In it she has a statement that she did not consult my work, a kind of negative citation. I have never seen anything like that before or after. I took it as a kind of underhanded criticism, as saying my work
was not even worth looking at. But of course, her statement gave no evidence for that.
[C.C., The citation Paul refers to by Nancy Bonvillain might be the only negative citation in the history of linguistics. The relevant reference to Bonvillian is from her thesis:
Bonvillain, Nancy. 1973. A Grammar of Akwesasne Mohawk. University of Ottawa Press.
In chapter 1 (pg. 12, footnote 6), she says:
“The work of Paul Postal (see bibliography) was not consulted as the basis for this grammar.”
In the bibliography she lists two references:
Postal, Paul. “Mohawk Vowel Doubling” in IJAL vol. 35, 1969.
Postal, Paul. Some Syntactic Rules in Mohawk. New Haven: unpublished Ph.D. dissertation, Yale University, 1964.]
[C.C., links on Nancy Bonvillain:
Chris:
I put in those links (to Gregersen, Bonvillain, Lounsbury) to show the academic context that you emerged from, and the kinds of things you were expected to do coming from that department at Yale. It is really quite amazing that you made the transition to TG, given that context. Do you have any thoughts about that?
Paul:
Linguistically, as soon as I heard about TG, generative grammar, that was my intellectual context. I had nothing to do with anything linguistic in the department, except since I planned to work on Mohawk, I had a bit of relation to Lounsbury. But our relations were limited and purely dealt with Iroquoian details. Once I started actually writing my thesis, which really only took place after I was at MIT, I had no relation with him at all. He knew nothing about TG and didn’t like it much.
3. MIT (1961-1965) and CUNY (1965-1967)
Chris:
What was your first job after graduate school? What year did it start?
Paul:
I was a research associate at MIT, research laboratory of electronics. It started in July, 1961, I think.
Chris:
How did you get that job?
Paul:
My memories are vague. Chomsky came to Yale while I was there and gave a talk on the work he and Halle were doing on English stress. I did not meet him. But I was working for some reason on a paper about simplicity in phonology and ultimately sent it to Morris Halle. He invited me to come to MIT for a visit, which I did. I don’t remember anything much about the visit, but I guess that is how I was offered the job.
Chris:
What did you do as a “research associate”? Are there any important events from that period that you can recall?
Paul:
I worked on my thesis. Important events…no. I met various people, Ed Klima, Hu Matthews, Bob Lees, Chomsky, Jerry Katz, and became aware of how little I knew of anything relevant along endless parameters. Basically, nothing I had done before really helped me much.
Chris:
It sounds like a wonderful environment. Is there anybody in particular who you discussed syntax or phonology with at this time at MIT?
Paul:
If I recall, at that early point, the person I talked to most was Hu Matthews, who had worked on an Amerindian language, Hidatsa. But I don’t remember any content. I rarely talked to Chomsky much throughout my stay at MIT. I had a lot of contact with Halle, because I was interested in Mohawk phonology. I remember one day he took me to talk to Roman Jakobson, who was visiting. There was some problem in Iroquoian phonology and Halle seemed to want to know what Jacobson thought about it. When the problem was described to him, I forget what it was, Jacobson said ‘it was just like Oneida’ (which he pronounced [onida]), which was true. Oneida is a sister to Mohawk, and Mohawks will say of Oneidas that ’they speak our language’, although the phonologies are very different. Unfortunately, that provided zero solution so the whole visit was a waste of time.
Chris:
Did you take any classes at MIT during this period? Do you have any memories of them?
Paul:
No, I took no classes. I attended one class of Chomsky’s one afternoon. I found it of little interest; he proposed some ad hoc indexing thing for relative clauses, which I found totally unpersuasive and never went again.
Chris:
Where did you go after your research associate position at MIT?
Paul:
To the City University of New York. I was mostly at Queens College but also, at least pretty soon, was associated with the Graduate Center in Manhattan. I stayed there for two years. Then I went to the IBM Research Center, in Yorktown Heights, New York, about 40 miles north of Manhattan.
Chris:
Wait, I think we skipped a few steps. Your CV says that after your research associate position, you became an assistant professor at the MIT Department of Modern languages from 9/63 until 9/65. Presumably you taught during this time. What did you teach?
Paul:
I think I taught one course on the pre-generative linguistics of the time and one on English syntax, about which I knew almost nothing. But my memories of all that are vague.
Chris:
Is it during that period, when you were an assistant professor at MIT, that you met Haj?
Paul:
Too precise. I met him after the Ph.D. program began and students showed up. It was probably 1963 but I am not sure.
Chris:
I don't know what year Haj started his graduate education at MIT, but his degree was awarded in 1967, so I figured you two overlapped at MIT.
Paul:
Yes, this is not in question, but the precise date is unclear.
Chris:
During this period, you must have also met McCawley and Langendoen, who you later wrote a book with. Is that right?
[C.C., here is a list of MIT alumni with years of graduation
Paul:
Right, but I have almost no memory of McCawley at that period. My only memories of him in person are some hallway conversations at meetings in later years.
Chris:
I have a feeling that this time must have been important to you intellectually. You were no longer writing your thesis (which was finished when you were a research associate). You were teaching a set of very smart students. What kinds of topics were you thinking about? Do any of them foreshadow major themes in your career?
Paul:
Well, I wrote my monograph ‘Constituent Structure’, which was a criticism of nongenerative ideas, and raising to object started with a student named Peter Rosenbaum, who later went to IBM and was the driving force in getting me hired there. But I didn’t start thinking much about it until Chomsky got his bizarre idea that there was no such thing.
[C.C., a link to the book ‘Constituent Structure’:
Chris:
Do you mean that the idea for the rule of raising to object is due to Peter Rosenbaum?
Paul:
I believe so but not sure. One would have to look at his thesis, which I don’t have. Maybe I specify in ‘On Raising’. I will look later.
Chris:
Yes, please check. It is interesting to know the origin of ideas.
Paul:
I looked briefly in ‘On Raising’. It is pretty clear from pages 5-10 that the raising to object idea was indeed Rosenbaum’s.
Chris:
As for your monograph ‘Constituent Structure’, I have not read it. Why did you decide to write a whole monograph on that topic?
Paul:
No idea.
Chris:
Could you summarize it for me?
Could you summarize it for me?
Paul:
No, I hardly remember it.
Chris:
Why did you leave MIT in 1965? It seems that being an assistant professor at MIT would have been a pretty sweet deal. You were in the middle of the intellectual excitement with lots of smart students.
Paul:
Various reasons. I never viewed living in New England, with its distinctive culture as permanent. My whole family and previous acquaintances were in New York. Probably in some sense I did not like the developing culture at MIT, where Chomsky played an outsized role and where any idea of his rapidly became taken as something super important. But perhaps I am building future reactions into that early decision.
Chris:
Your CV says you spent 6/64 to 8/64 (three months) as an assistant professor at Indiana University. Did you live in Indiana and teach?
Paul:
It was an LSA summer school. It was nothing like three months, maybe 6 weeks. Of course, I lived there. I taught two courses, one on Mohawk phonology, the other on transformational syntax.
Chris:
Did you get to meet anybody there who you exchanged interesting ideas with, or learned interesting things from?
Paul:
Ken Hale. I spent a lot of time with him. That is why I recommended him as my replacement when I left MIT.
Chris:
He must have been quite interested in your work on Mohawk. What did you guys talk about?
Paul:
Yes, I remember him attending the Mohawk class. But no memory of 56 year old discussions.
Chris:
You mentioned you wrote a letter of recommendation for Ken Hale to MIT. Do you still have a copy of that letter?
Paul:
I didn’t say that. I said I recommended him. I suspect it was purely verbal to Morris Halle personally, before I left MIT.
Chris:
By the way, I thought that bringing Ken Hale to MIT was a very smart move. It opened up a whole new world to them of cross-linguistic research, on all the languages that Ken was interested in, including Native American languages but also Australian Indigenous languages. He was my advisor at MIT for my 1993 Ph.D. thesis.
Do you have any thoughts about this?
Paul:
Yes, I totally agree. That was my motivation for recommending him, besides his obvious great qualifications. The attitude in the Department was very much focused on European languages. When I met Klima he made disparaging remarks about ‘gook’ languages. This despite the fact that Matthews had worked on Hidatsa. When I gave a talk at MIT on Mohawk, there were a certain number of condescending jokes. This was kind of weird. Chomsky had of course worked on Hebrew. Anyway, I guess establishing Ken there got rid of all that kind of thing.
Chris:
When you were at MIT, did you have much interaction with Klima? Was his office next to yours? Was he teaching there at the same time as you? You and I know his negation article quite well. It is brilliant and groundbreaking. One of the most important articles ever written on negation.
Paul:
No, I don’t recall having any interaction with Klima at all. I don’t know where his office was. I never recall seeing him, eating lunch with him, etc. But he was there.
Chris:
How about Noam? During the time you taught at MIT as an assistant professor, did you interact with him? Did you talk about research projects? Syntax? Plan courses? Did you go to his lectures? Did he come to your lectures? Did you visit him at home and shoot the breeze about syntax?
Paul:
Basic answer to all these questions is no. I had little to do with him. He was always busy, there were always other people hanging around in his office; it was hard to have a private conversation with him. I told you I attended one day’s class of his. He never attended any of mine; I am sure the idea never passed through his head. Why would it have? I was never in his home. I once had a party for people in the department in my apartment and he and his wife came.
Chris:
Your book with Katz appeared in 1964 (‘An Integrated Theory of Linguistic Descriptions’) while you were still at MIT. I want to go into that a little. That book, I have read (or at least parts of it). Could you summarize for our readers what the book is about?
Paul:
Not really. I don’t remember much about it and haven’t had any interest in it for decades. All I remember is this. Chomsky’s TG had no semantics of any kind. Katz and Fodor developed a kind of semantics and published it in 1963. Katz and I tried to concoct a view which permitted a Katz/Fodor semantics to combine with a generative TG. A basic idea was that transformations had no effect on meanings. You can see echoes of that view in the APG [C.C., Arc Pair Grammar] book. Chomsky seemed to adopt elements of our view for a while; you can see it in his Aspects. There, in my opinion, he improved the framework greatly by eliminating the notion of generalized transformation, which were mappings which combined separate trees into single ones. Rather a single tree, called a deep structure, was generated by the base rules and that is what could be interpreted by the type of semantic rules Katz and Fodor had introduced.
Not really. I don’t remember much about it and haven’t had any interest in it for decades. All I remember is this. Chomsky’s TG had no semantics of any kind. Katz and Fodor developed a kind of semantics and published it in 1963. Katz and I tried to concoct a view which permitted a Katz/Fodor semantics to combine with a generative TG. A basic idea was that transformations had no effect on meanings. You can see echoes of that view in the APG [C.C., Arc Pair Grammar] book. Chomsky seemed to adopt elements of our view for a while; you can see it in his Aspects. There, in my opinion, he improved the framework greatly by eliminating the notion of generalized transformation, which were mappings which combined separate trees into single ones. Rather a single tree, called a deep structure, was generated by the base rules and that is what could be interpreted by the type of semantic rules Katz and Fodor had introduced.
But ultimately, Chomsky strangely abandoned the deep structure concept and went back to something like his original generalized transformation idea (in his minimalist program). Of course, he has never had any serious idea about semantics of any kind, so the semantic consequences of his ideas have never loomed large for him. Basically, Katz was responsible for all the semantics in our book while I wrote most of the syntactic stuff.
Chris:
Is this the famous Katz-Postal Hypothesis (henceforth, KPH)? Could you state that hypothesis for me?
Paul:
I think it was just the idea that base structures were interpreted (I guess those were the output of phrase structure rules + lexicon) and that transformations had no effect.
I think it was just the idea that base structures were interpreted (I guess those were the output of phrase structure rules + lexicon) and that transformations had no effect.
Chris:
Can you outline a single concrete analysis which at the time was motivated by the KPH? I want to give our readers a feel for the hypothesis.
Paul:
I just can’t think of how that hypothesis could motivate a concrete analysis. It just said that if e.g. something was a topicalized sentence, it had the same meaning as one minus the topicalization, ditto for passives. But it didn’t say anything about how to analyze passives. It was taken later by critics to be falsified by things like (1) having different scope possibilities than (2).
(1) Every student studied no foreign language.
(2) No foreign language was studied by every student.
But of course, one modern view could have different underlying quantifier DPs in different scope positions and thus even in a TG variant, the sentences would not differ just in transformational relations.
Chris:
But isn't it the case that May's theory of quantifier scope would be incompatible with KPH, because the quantifier DP moves to different scope positions resulting in different interpretations?
Paul:
Yes, for sure. I was going to mention that; but of course, I would never do scope via raisings. When I get some energy if ever, I will describe how I think the range of scopes we cover could be handled.
Chris:
I know that APG (Arc Pair Grammar) does not have deep structure and transformations, but do you believe that something like the KPH holds?
Paul:
Well in that framework, the claim was (though no real semantics was spelled out) that the original structures, defined by self-sponsoring arcs, but equivalently by unsponsored arcs, determined meaning. So that is the closest analog of the KP claim.
Chris:
So as of today, you believe that something like the KPH holds. Right?
Paul:
That’s true, but I have never given any real evidence for such a view and have never done any serious semantics to support it. It just has always seemed to me to be a conceptually lovely way to look at things.
Chris:
Personally, I think that the idea that scope positions are underlying positions is interesting, and should be taken seriously. So, the idea is that the DP is underlyingly in its scope position (not moved there), and is simply not spelled out there, but is rather spelled out in its variable position.
Paul:
Yes, spelled out in ONE of its variable positions. For me, that is going to involve lowering of the Q determiner to the appropriate variable position DP, since only NPs originate in variable positions. This eliminates the need for the ad hoc principle which says a quantifier DP in such a position is interpreted as a variable, which to me is entirely bizarre. The lowering will do the work that the claim that scope position DPs are covert does.
Chris:
How and when did you meet Jerrold Katz? Did you call him Jerrold or Jerry?
Paul:
Jerry…everyone called him that. I met him at MIT, but I really have no memory of when, where or what the circumstances were. It must have had something to do with Jay Keyser, who was a contemporary of mine at Yale. We both came to MIT at the same time and had cubby holes next to each other in the infamous Building 20. I believe he knew Katz earlier; they were both from Washington (DC). That must have been the connection.
Chris:
So, the work on Katz and Postal 1964 was done when you were at MIT. As you said, Katz worked out the semantics, and you did the syntax. What was it like working with Katz?
Paul:
Difficult. He had a very strong personality and once he formulated an idea it was very difficult to lead him away from it. There were many disagreements but I have no memory of what they were.
Chris:
Did you work in your office at MIT, in cafes, at home, by phone, by mail?
Paul:
We worked in his office as I recall, because he had the latest high-tech word-processing equipment of the time, an IBM Selectric typewriter with no ribbon but a kind of ball with type on it. It seemed really cool at the time.
Chris:
How did it go?
Paul:
As I said difficult. We would often argue about who would formulate a particular idea or passage, which involved a discussion of who would sit at the typewriter. Since it was his office and typewriter, he usually won. The whole thing seems comical in retrospect and probably did at the time as well.
Chris:
How long did it take to write the book together? Just fill me in on the process a bit.
Paul:
I don’t remember how long. I would guess six months. It would have to have been after he finished the semantics article with Fodor, and done in time to have been published in 1964.
I have written to Jay Keyser to see if he has any memories about my meeting Katz.
Chris:
I cite Katz and Postal because of its use of various empty elements. For example, as I recall, it had null prepositions.
I think that the use of empty elements in Katz and Postal 1964 is very significant for understanding the syntax/semantics interface.
Chris:
Did Katz work at MIT when you were an assistant professor at MIT? Was he in the same department as the Chomsky and Halle? If not, where was he working?
Paul:
Yes, we overlapped the whole time I was there. He was in Philosophy Department. Later I think that and Linguistics became one. Keyser would know about that.
Chris:
OK, so then you went to Queens College and the Graduate Center (both in NYC). Were there any linguists there at the time? I guess this was before Ray Dougherty was at NYU.
Paul:
I don’t think there were any at Queens then. There must have been somebody at the Graduate Center, but I have no memory.
Chris:
What courses did you teach then at Queens College and the Graduate Center?
Paul:
I think at Queens it was English grammar. Not sure at the Graduate Center, probably something on linguistic theory, whatever I took that to be.
Chris:
Also, what kind of research were you doing at Queens College/Graduate Center? What projects were you thinking about?
Paul:
I can’t remember anything. I think the period between the end of the MIT spring semester in 1965 and the beginning of 1968 was linguistics-wise kind of a lost period. During that period, I left MIT, moved to Queens New York and took a job at the City University of New York, my wife had a baby, I changed jobs again to IBM, moved again to Westchester county New York. I can’t think of anything positive I accomplished linguistically during that period. I only began to get what I considered results when I began work on crossover phenomena (probably) in early 1968. Maybe I started work on pronominalization more generally sometime in 1966-67, but I have no memories and it didn’t yield anything much.
With respect to the period 1965-1967 in the interview, about which I said I didn’t accomplish much of anything linguistically, I amazingly forgot to mention one important factor. In 1966, December I think, my then wife had a miscarriage at 6 months of twin girls. She reacted terribly to the loss and was depressed for a long time, with consequences you can imagine. So along with two moves, a new job, a live baby, there was also this in that period.
Chris:
I don't think I have done a good job getting you to talk about what the early days of generative grammar were like. What it was like to be part of a field transitioning from the older system (represented by your professors at Yale)?
Paul:
I do not have strong memories of this period. But for me there was no real transitioning, since I never really internalized the idea systems of the pre-Chomskyan American linguists of the time, and I was ignorant of European ones.
To understand the period of the introduction of generative grammar into American linguistics, one has to understand the existing situation. Ironically, this was itself the result of something of a revolution, whereby supposedly scientific structural/descriptive linguistics for the most part replaced more traditional grammatical ideas in the very few linguistic departments there were at that time. The newly established figures made no secret of their feeling of superiority with respect to previous views, a superiority based on their supposed rigor and scientific validity.
So when a new generation of linguists arrived, specifically people like Chomsky (30 in 1958), Morris Halle (35 that year), Robert Lees (36 that year), me (22 that year), and began claiming that generative grammar was the way to go and that the so-called successful scientific revolution of the descriptivists in fact had little to offer and was full of errors and lack of insight, it is easy to imagine the horrified reaction of the relatively newly dominant figures, then seeing themselves being pushed toward intellectual oblivion. And I think there was a considerable amount of immature behavior on the part of the generative cadre for a number of years, displays of arrogance, put downs, boasting and similar foolishness. Harris captures some of this in his ‘Linguistic Wars', even if he exaggerates.
There was, I believe, much feeling initially of fighting a battle with entrenched forces. But it wasn’t much of a battle for long, at least in syntax, where the in-place linguists had few ideas and rather superficial and limited ones at that. There was probably much more fighting in phonology, with the key issue being the kind of close to phonetics phonological representations then in vogue versus the partially morphophonemic representations advocated by Chomsky and Halle. Based on my work on Mohawk phonology, I wrote two papers and one book on these topics ('Aspects of Phonological Theory’1968). I have not thought seriously about phonology since 1969 so my memories are vague and I have no sense of how the field has developed since then.
It has often been remarked that generative linguistics got a huge boost from the government education financing bonanza which resulted from the shocked reaction to the Soviet launching of Sputnik. A lot of money was made available for language study and linguistics was able to get hold of a lot. New departments and Ph.D. programs were started, including most famously that at MIT, which I believe began in 1963 and which I taught in for two years, although what I knew to teach was laughably limited.
Chris:
I don’t plan to ask you questions about generative semantics and the “Linguistic Wars” since the topic has already been covered by two books. But I did want to get your feedback on a quote describing you from the book ‘Linguistic Wars’ by Randy Allen Harris:
(pg. 72) “Postal was even less loved by the Bloomfieldians. Like Lees, he is warm and genial in personal settings, and quite tolerant of opposing viewpoints. But his reputation for intellectual savagery is well-deserved, rooted firmly in his public demeanor at conferences, especially in the early years. The stories are legion, most of which follow the same scenario. Postal sits through some anonymous, relatively innocuous, descriptive paper cataloguing the phonemic system of a little-known language. He stands up, begins with a blast like ‘this paper has absolutely nothing to do with the study of human languages,’ and proceeds to offer a barrage of arguments detailing its worthlessness - often making upwards of a dozen distinct counter-arguments against both the specific data used and the framework it is couched in. The performances were renowned for both intellectual precision and rhetorical viciousness.”
What do you think of it?
Paul:
First, I appreciated the line ‘Like Lees……’. It seems accurate to me. Second, talk of ‘intellectual savagery’ is way over the top. As is the later ‘rhetorical viciousness’. No doubt I was very critical. But the whole thing is extremely exaggerated because during the period in question I attended at most two or three meetings of the type at issue and, I am pretty sure, spoke publicly during at most two. I did not like at all that the passage is based on opinions from unnamed sources about unnamed conferences at unspecified dates and my supposed criticisms of unnamed talks by unnamed people. This is not the way history should be written. I protested on these points to Harris, telling him he should cite the people, etc. who he claims told him these things or eliminate the passage. But he refused.
5. IBM (1967 – 1993)
Chris:
Why did you decide to leave Queens College/Graduate Center, and how did you get the job at IBM?
Paul:
I basically told you this. Peter Rosenbaum went to IBM about same time I left MIT or maybe a bit later. He wanted me to come there and persuaded whoever to hire me. I decided to go because of financial security and because there was zero interest at Queens College and not much at Graduate Center at that point. But mostly IBM Research Center was a linguistic desert for most of the time I was there. So, I got money and ability to mostly do whatever research I wanted but no stimulation. Rosenbaum left fairly soon after I was hired; I think he had some dream of starting his own company. I lost track of him and have not heard from or about him again.
[C.C., Rosenbaum is now an academic editor:
Chris:
So, you started at IBM in 1967, what was your starting salary?
Paul:
$22 thousand. According to the inflation calculator I looked at, that is around $170 thousand in 2020 dollars. Nice salary for a 31year old, no?
Chris:
Yes, it is great. Really great. It gave you lots of freedom, and allowed you to do what you wanted.
Paul:
Yeah, it was a lot of money. Note though it was a 50 week salary, as starting employees got two weeks vacation, where university people get at least three months.
Chris:
In comparison, do you remember the salaries for either CUNY or for MIT when you worked there?
Paul:
I started at MIT at $6500; when I became a professor, it was $9000, but we got 2k more if we didn’t vanish in the summer. At CUNY I started at 12k or 12.5k; but I soon got promoted, and then if I recall promoted again to full professor; (don’t ask me why, as I have pointed out I accomplished little or nothing while I was there). So, when I left CUNY my salary was 18k, already quite a lot.
Interestingly, the 26 years I worked for IBM was a time of high inflation; my mother was at one point earning, I believe, 18 per cent on investments. Seems unbelievable today. So even though I got numerous raises, I recently calculated that my salary was less in buying power every year I was there, and that in 2020 dollars, it was worth $22,000 LESS in 1993 than in 1967. These calculations ignore the stock benefit employees had. We could buy IBM stock with a certain percentage of our salaries once a year at 15 per cent off the market price on the fixed date. Most of the time the stock went up in value. Ultimately, I sold it old and used the money to pay for my children’s college and some graduate school.
Chris:
So how did that work? Was IBM placing money on generative syntax? What department were you hired into? I just wonder what they thought the benefit to them would be of hiring a generative syntactician.
Paul:
I was hired in the Mathematical Sciences department. It had a primitive linguistics group which was mostly leftovers from an earlier Russian machine translation effort. For a company of the wealth of IBM at the time, the money involved in linguistics was miniscule. No doubt the publicity given to generative grammar played a role in expanding the group with Rosenbaum and me. The leftovers from the Russian effort were fairly soon dehired. No doubt dreams of machine translation played some role. These never died and the company spent lots of money on that and never achieved anything much. But they had a sound view that any applications/products in any area depended on underlying scientific research. And they understood further that most of the work in any field would be done by people outside the company and that they needed in house expertise to keep up with the work done elsewhere. I had a job status which was called a Research Staff Member, which was pretty high. Out of the 1200 people who originally worked in the center where I did, I think only about 300 were Research Staff Members.
They had a nice system such that every year or was it six months one had to file a report on what one had done. And there were different categories of accomplishment at that period which determined one’s salary increase if any. These included publications in scientific journals, patents, specific work leading to actual products, etc. My production usually looked good because of journal articles and books. Few people in the Mathematical Sciences ever published books, but I did, so I was outstanding in the whole department at times for book publications.
Chris:
Your whole time at IBM is a black box to me. If you were at a university, I would know approximately what you did. But I cannot really imagine what it was like at IBM. I want to ask some questions about that.
What percentage of your time was spent doing your own research (vs. IBM assigned tasks)? Did that percentage change over time? Other than your own research, what kind of projects were you involved in? What was your average day like at IBM? Where you able to have study leaves and sabbaticals? Who were the colleagues that you interacted with?
Paul:
OK. When I arrived at IBM there were no computers in offices; writing was done with typewriters, white out, cutting and pasting with scissors and tape, etc. Horrible. What I did was read journal articles, books (there was a library and they would buy some books), talk to people on the phone, communicate by letter. I would say over my whole period at IBM I spent 98-9 percent of my time doing what I wanted. But in the background was the yearly report. So, I had to publish things.
At one point the head of the Math Department began to pressure the group to do something practical…aim for a product. The decision was made to try to construct a question answering system which could answer formatted questions aimed at databases. I worked a small amount on constructing the grammar for the system. I remember nothing about its content. The work on the system went on for years, and when David Johnson finally got a full-time job there, it was to work on that system. Which never really developed into a usable product.
Sometime in the late 70s, computer terminals appeared in offices, those gray screen suckers with the green writing. But they connected to the main frame. There were no office printers (there never were while I worked there). If you wanted to print something, you had to send the file to the main printer room, which I recall was on main floor, we were on third. Later, much later, printers appeared in the halls, at least on the same floor. So, when work switched to terminals, I had to learn to use whatever the word processing system was. I can’t remember its name. It was not until 1986 that we got decent terminals, with color screens, but still just linked to the main frame.
Most of the time I worked there, there was really no one to talk to about linguistics seriously. Susumu Kuno visited, I think for a year. David Johnson had a temporary job in late 70s, which is when we wrote APG. One semester I went on a sort of sabbatical to Rockefeller University; that is when John Grinder and I wrote our paper Missing Antecedents. Johnson came back as a permanent employee sometime in 80s but only to work on product development like the answering system. So, we could never do much of anything together.
Of course, while the internet began around 1980 internal to IBM, that is, one could connect to other IBM locations, real internet was much later and really never much before I got terminated. I remember my first internet message. It just appeared on my screen. It was utterly mysterious to me how it appeared there. But it was sent from another IBM location, if I recall correctly in Australia.
I also remember the first time I really used the system. It must have been in the early 80s. Pullum and I were working on something and he had moved from GB to Santa Cruz. There was an IBM location not too far from there. And somehow, I got in touch with somebody who worked there and arranged to send him the file and have him print it and leave it at the entrance desk where Pullum could pick it up. Complicated and primitive but it seemed REALLY cool at the time to be able to send an article not on paper across country in an instant.
Chris:
Warren Plath appears in the acknowledgements of several of your papers. Who is he? Did you work with him?
Paul:
He was the manager of the linguistics group the entire time of my presence at IBM. No, I never did anything with him which could be considered working together. He is the brother of famed suicide poetess Sylvia Plath, something he NEVER would talk about, not that I ever raised the topic.
Chris:
Was he a linguist? You acknowledged him quite a bit in your papers, which makes me think you discussed linguistics with him often.
Paul:
Yeah, I think he had a degree from Harvard. We discussed things only in context of my reports on what I was doing. He would read my writings and make editorial comments. It seems impossible to find out anything about him via Google because queries just return stuff about his sister, who is famous.
Chris:
So, you are not in contact with him any longer?
Paul:
No, I haven’t seen him or heard from him since I left IBM. After that, I don’t remember when, I wrote him a message/letter thanking him for all of his many kindnesses and considerations during my residence there. But he never responded. I have no idea whether he is still alive. David Johnson probably knows more.
Chris:
See the following link. It says that Warren Plath has a 1964 dissertation from Harvard titled ‘Multiple-Path Syntactic Analysis of Russian’ directed by Anthony Gervin Oettinger.
What was the name of your research group at IBM?
Paul:
Linguistics Group, part of Mathematical Sciences department, with many other groups, of real mathematicians.
Chris:
This book review written by Warren Plath gives more information about him at the end (on page 559):
“Warren J. Plath is a research staff member, emeritus, at the IBM T. J. Watson Research Center, where he formerly served as manager of the Theoretical and Computational Linguistics Group. As a graduate student and later faculty member at Harvard in the late 1950s and early 1960s, he was a member of the machine translation project under Anthony G. Oettinger, in which he worked on automatic parsing of Russian. At IBM in the mid-1960s he managed the transition from applied work on Russian-English MT using special-purpose machines to a broad program of research emphasizing linguistic theory, English grammar, and formal parsing systems implemented on general-purpose computers. His subsequent project and individual research included work on natural language question-answering, machine-aided translation, semantic grammars for spoken-language interfaces, and information extraction from news articles on the Web. Plath’s address is Mathematical Sciences Department, IBM T. J. Watson Research Center, P.O. Box 218, Yorktown Heights, NY 10598-2103”
Paul:
Good find. That has more about Warren than I think I ever knew.
Chris:
So, can we say that Warren Plath was important to your work?
Paul:
Yes, you could say that but not in the sense you are thinking, I gather. He was not important in the way say that the people I have done joint work with have been, Katz, Haj, Perlmutter, Pullum, Langendoen, you. Warren was important in facilitating my relation to the Research Division and thus to the corporation. He was/is a very decent, conscientious, rigorous, disciplined person, not involved in corporate politics, not having as far as I can tell any real desire to rise in the management hierarchy. As far as our relations were concerned, his job was to take whatever reports I produced about my work and speak to his management about it (along with work by others in the group) in such a way as to justify our existence. He no doubt did a good job of that. But my existence really depended on my publications. I produced I think six books and one edited volume while I was there, with real publishers. This looked especially good because mathematicians don’t produce much in the way of books. I also published a constant stream of articles. That was my work product. Warren was also very unimaginative and a very rigid and uptight personality. For instance, every work day he was there he had a tuna fish sandwich for lunch unless there was some kind of off campus lunch. Thousands of tuna fish sandwiches.
Chris:
Did he give you feedback on your work?
Paul:
Basically, editorial feedback.
Chris:
But did he also supervise your work?
Paul:
Not in any real sense.
Chris:
Did he allow you freedom to do your research? Did he have that responsibility?
Paul:
Not much. By being a research staff member, I was primed to have that freedom. Of course, if he had had his own ideas, being my manager, he could no doubt have pressured me to work on lines he was interested in, but he had none. The only pressure I was ever subjected to was when the Mathematical Sciences director decided that the group should do something practical; this led to a project to develop a database answering system. Warren, David Johnson, Stan Petrick and several programmers worked on this for years. I did a tiny amount, sketching some ideas about how the syntax should work…but it looked good because Warren could always say and did that I was contributing to this project, the kind of thing management could understand.
Chris:
Do you still have an example of the IBM yearly report?
Paul:
No.
’
;
Chris:
Counterfactually, if you had stayed in a department of linguistics (like CUNY or MIT), how would your career have been different? You might have published less, because you would have had less time. But you might also have had students and done teaching. You might have also had linguistics colleagues to collaborate with.
I sometimes think that your position at IBM isolated you a bit from the field. In some ways, that could be positive (giving you more time and intellectual freedom), but in some ways it could be negative (isolating you, giving you less chance to interact with students and linguistics colleagues). Have you thought about these things?
Paul:
Everything you say is quite true. But who knows what the alternative world would have yielded except the obvious. Anyway, one makes choices and takes the consequences.
Chris:
You are basically an outsider in various ways. You worked at IBM instead of at a university.
Paul:
This was quite crucial in many ways. Most academics have, I believe, no idea what a real corporation is like, the constraints they operate under. Full time (not adjunct) academics live a rather cushy life; the money that supports them is largely taken for granted. It comes from the government or from student fees which academics don’t have to worry about. They are in a sense a kind of welfare class. In a corporation,’ even in the rather far from the business side Research Division of IBM, money and profits are never far from consideration. It is always kept in mind that the company is competing with others, that it is subject to government controls, often unreasonable and corrupt. The threat of serious losses and financial disaster is never out of mind. And indeed, in late 1993, even an as huge and previously successful and dominant company as IBM came, due to terrible management decisions, to the brink of bankruptcy. This led to the elimination of essentially half of the work force, including me.
Most information in corporations is subject to privacy and secrecy regulations, costs are closely monitored (everything from travel expenses to personal phone calls). It is a world very different from the university world. Lawyers play a very major role. The corporate world is widely misunderstood and hated by many academics, with extreme cases like Chomsky’s absurd claims that corporations are fascist organizations. Funny, all the time I worked at IBM, they didn’t arrest anyone.
Chris:
In addition to working at IBM, you started out in generative syntax, but then rebelled. You are a conservative, where academics are largely liberal (at least that is my impression). So, these things go together. Do you agree with this outsider characterization?
Paul:
Absolutely. And it had both advantages (no group think) and disadvantages.
Chris:
How did you get “terminated” (your words) at IBM? Were they downsizing? Was it time for retirement?
Paul:
Due to massive mismanagement, in 1993 the company, which had been one of major corporations in the world for decades, was on the verge of bankruptcy. This was due to many factors, but crucially ignoring the coming role of small computers and downplaying the importance of software. This permitted Microsoft to come into existence, which never should have happened…and the whole Microsoft world should have been part of IBM. In the early 90s the company tooled up to have sales of 100 billion dollars per year. This involved scaling up employment role to 400,000. But they never reached 70 billion. Thus disaster, leading to offloading of half the work force worldwide, roughly 200,000 people.
Chris:
OK, you were at IBM 26 years. How can we divide up that time for the purpose of this interview? Do you view the 26 years in different segments, based on the kinds of projects you were involved with?
Paul:
I never thought about such a division but I suppose one could do it in terms of books. There are too many articles to worry about. Keep in mind, as I always did, that there was the annual report to fill out. And since I wasn’t generating any patents, or products, etc. my work product was almost entirely publications.
So, I guess my first years were those when I worked on the crossover book and then on ‘On Raising’. These were conceived of in the TG terms of the period. The first was published in 1970, the second in 1974. But before the latter date, Perlmutter and I got into relational grammar. And this was my major activity until the APG book, published in 1980. That work was supposed to yield a giant volume covering all aspects of relational grammar, but it never got produced. Part of the reason was that Johnson and I started developing APG, which for us was just a way of trying to make relational grammar serious. This caused problems with Perlmutter, who felt I should not work on an APG book until after we did the relational grammar book. But for me, after the ideas got going, working on a non-APG version of relational grammar made no sense. And to the extent that I did that, yielding various papers published in the three volumes of ‘Studies in Relational Grammar’, it was out of a sense of duty.
Independently, Garland decided to publish my thesis in 1979. But that represented no new work. It did look good on my annual report though.
After that, Terry Langendoen and I worked on ‘The Vastness of Natural Languages’, which appeared in 1984. That was a kind of isolated work and did not involve a lot of time. After that I sewed together various things and produced ‘Studies of Passive Clauses’, a cross-linguistic thing. Sometime around then I started working seriously on French. I got IBM to pay for an informant; I had two over time, both women, who came once a week or so over several years. This permitted writing ‘Masked Inversion in French’, published in 1989.
Sometime in late 80s I started working on extraction and anti-pronominal contexts, which yielded ‘Three Investigations of Extraction’, but not until five years after I left IBM.
Chris:
OK, let me begin with ‘On Raising’ (1974). First, I must say I only read it a few years ago, and I loved it. It is one of my favorite syntax books. I loved how you took a single construction so seriously and turned it inside out.
So, one question I have is this: What motivated you to look at this construction so carefully? What is the background leading up to that? Is there some paper you read first? Why did you feel it was such an important topic to write a whole book about?
Paul:
This is quite clear. To me the existence of raising to object seemed obvious. So, I was motivated and annoyed by Chomsky’s proposal that there was no such thing.
Chris:
How long did it take you to write?
Paul:
I don’t remember how long.
Chris:
Do you remember where Chomsky proposed that? Was it ‘Conditions on Transformations’?
Paul:
Look on page 29 of ‘On Raising’, attached.
[C.C., the quote from ‘On Raising’ is given below:
“Chomsky (1972: 86) states, in criticizing an analysis of the word remind (Postal, 1970a), which made some appeal to Raising, the following: ‘... I might mention that the permutation rule that gives (25) as well as subject-raising [into object position-this phrase not present in the original circulated version] seem to me to be at best dubious rules.’ This formulation (without the added clarification) might well have been confusing to some, as it was to the present writer, who, in some circulated but unpublished arguments, took it to be a general rejection of the existence of a rule Raising. However, in an earlier personal communication as well as in the published version, the author has clarified his position, indicating that what he had in mind rather was only to call into question the existence of a raising operation in the case of what I am here calling B-verbs. That is, Chomsky does not question the existence of Raising for at least some A-verb constructions. Rejection of Raising for B-verb constructions is also briefly alluded to in Chomsky (1971) and sketched in greater detail in Chomsky (to appear) [C.C., ‘Conditions on Transformations’].”
Chris:
Do you remember how people reacted to your book once it was published? Did it generate useful debate? Were some people convinced?
Paul:
I think Kuno was visiting IBM during period I started on the work and we discussed it and he was convinced if I recall and found something relevant in Japanese. No one much who accepted Chomsky style TG accepted it and Lightfoot wrote a negative review as I recall.
But as relational grammar developed everyone in that movement accepted it. It was obvious on passive grounds alone. That is, if passive is a relational thing based on same clause object becoming subject, then raising is a necessity, or so it seemed. Of course, then the HPSG and later GPSG types cooked up a way to analyze Frank believes Mary to be pregnant such that Mary was only in main clause (where Chomsky had claimed it was only in complement clause).
The worst thing about the debate about raising to object arose when Chomsky started talking in terms of ‘move alpha’. Because one of his arguments against raising to object is that it required a special ad hoc rule. But once it was permissible to talk about ‘move alpha’, that argument dissolved. But that was never observed.
Chris:
Sorry, I couldn't follow this last bit, about 'move alpha'. Could you elaborate?
Paul:
I don’t see the obscurity. Move alpha, not a rule although often called such, is a schema which permits any movement operation whatever. Bad stuff must be blocked by independent principles. So, raising to object is permitted at no cost, contrary to Chomsky’s argument that it was a costly ad-hoc addition to some otherwise excellent system. I will try to find tomorrow where Chomsky made the original argument which I claim move alpha dissolves.
Chris:
OK, I see.
As you said, you started working on Relational Grammar during your time at IBM. What is the origin of RG? Are the original ideas yours, or Perlumtter's or somebody else's? Do you remember the period where you started thinking in those terms? And what were the initial ideas that got you going?
Paul:
I do not think it makes sense in general when two or more people work jointly, as Perlmutter and I certainly did, to ask whether the original ideas of the joint project were A’s or B’s, especially when memories are vague after 45 years or more. I do have a definite view about a couple, I mentioned, the unaccusative notion.
The period definitely began in 1972; I remember Perlmutter visiting for a couple of days and us writing up some ideas. The initial ideas were definitely heavy on passivization, moved on to double object relations (give the book to Chuck vs. give Chuck the book). Originally, we had a wrong view of this as involving the prepositional object being promoted to direct object status; this is entirely wrong as I showed in ‘Edge-Based’.
The early ideas are well documented if I recall in the volumes Studies in Relational Grammar I and II, and III.
Chris:
I think that one thing RG did was to factor out passivization (and other grammatical relation changing operations) into different kinds of systems. There was the advancement rule of 2 --> 1, and that was invariant, but then other things like the presence of an auxiliary, or linear order, or a designated passive suffix, could vary. But these were independent of 2 -> 1. So, this is very different from the passive transformation in Chomsky 1957 that grouped all the effects into one transformational rule.
Another thing that I think RG did was to group all the relation changing operations together, as opposed to other kinds of operations. This seems to be a kind of precursor to the A/A' distinction in Principles and Parameters syntax.
Yet a third thing that RG did was to make available formal descriptions of a wide variety of languages, which are constantly being cited in later literature.
According to you, what are the interesting scientific accomplishments of RG (we are still talking pre-APG here)?
Paul:
I think your characterizations of RG ideas here are in general justified.
Of course, the unaccusative idea received its popular first formulation in RG terms, but as I told you, I had it in my thesis a decade earlier. At that time, it was no big deal since work had not congealed on the idea of a uniform initial rule S —> NP VP, so my rule S —> VP didn’t trample on any received wisdom. It didn’t require any invisible underlying expletive subjects, etc.
I think the major contribution [C.C., of Relational Grammar] was documenting a very large number of relational facts in a lot of languages of different types, and showing how insightful relational descriptions could be given of them, which was not clearly possible in nonrelational terms.
There were also claimed relational laws, probably none of which were really right and some of which were clearly wrong. Like originally it was claimed passive always operates on direct objects, which is just wrong even for English, as I argued in ‘Edge-Based’. Some generalizations may be true, although facts are complicated, like the claim that in clause union cases a transitive subject is mapped to indirect object, a transitive object to direct object, etc. I have not thought about these matters in a long time but they are well represented in the ‘Studies in Relational Grammar’ volumes and also, I think in my ‘Studies of Passive Clauses’.
I will mention one detail. The participial passive in French is like that in English in many respects, but in generally totally restricted to direct objects. But then there are the verbs (des)obeir. These do not take direct objects. And yet participial passives based on them are fine. In ‘Studies of Passive Clauses’, I show how to describe these facts without any complication of the conditions on passivization. Basically, with some simplification, one says that with these verbs and, I think them only, 3s/5s can advance to 2 but only if the resulting 2 is passivized. Problem solved. There are other cases like this; I showed something similar about the se faire type passive in French, which is strictly restricted to working on direct and indirect objects, except that with a few verbs which only take de objects like (se) moquer one can form se faire passives. I show in a paper that this is analogous, that is, the de phrases, taken to be 6s can advance to 3, but only if then submitted to the se faire passive construction. But I gave this analysis in APG terms.
Chris:
Can you summarize briefly what separates RG, which you developed with Perlmutter and others from APG (Arc Pair Grammar), which you developed with David Johnson.
I was at Cornell starting in 1993, and I would sometimes talk to Carol Rosen about RG. She even gave me a small tutorial one day, which I appreciated. But I don't think she adopted APG. I don't think Perlmutter took up APG either. Do you have any insight into that?
Incidentally, I would say that other than Principles and Parameters and Minimalist Syntax, the framework that I know best is APG, since we drew so many trees in that framework over the years, especially in work on ‘Imposters’. I think APG is a wonderfully rich framework that has lots of important ideas to investigate.
Paul:
Well RG developed out of an amalgam of TG ideas with some crude ideas about primitive grammatical relations. It came to have a crude idea of clausal structures as non-tree graphs with arcs. But it did not have a good clear way of talking generally about relations between arcs. For internal relations it appealed to the bad idea of coordinates on arcs defining levels. But for trans-clausal relations, it could say that the object of one clause was also the subject of a lower one, by drawing upstairs and downstairs arcs with the same head phrase, but it did not have a way to represent the priority of the lower relation.
APG with its notion of the sponsor and erase relations between arcs, provided that as well as the formal basis to talk about deletions, to guarantee that when something advanced, demoted, or raised it in some sense ceased to involve (for surface purposes) the previous relation.
No, Perlmutter never really got into APG or really understood it. And as he developed a relatively large body of quite devoted students, this was problematical from my point of view because they also did not get into it and thus I often felt that the analyses they were trying to develop could be significantly better done if they utilized its tools. But as they had contact with Perlmutter on an intensive level and rarely with me, there was no way to influence the situation. Ditto for David Johnson, who also was of course not at a university and had no students.
There was one exception. Judith Aissen wrote an APG book on Tzotzil and we talked a lot about things on the phone. But I would say too the early version of APG in the book was itself very primitive; the version in ‘Edge-Based’ is far superior but that only appeared long after RG had essentially died as a movement.
Chris:
There are lots of things about APG that I like a lot.
For example, I like the way that movement is represented in terms of multidominance. This is an idea that is now coming into vogue in minimalist circles as well. But less intuitively, I like the way that anaphora is also represented in terms of multidominance. I think that is a very powerful syntactic idea that people (outside of APG) have not really tried to investigate. I also like the conception of passive by-phrases as arguments (initial 1s), which corresponds loosely to how I think about them. Lastly, you mentioned to me that its conception is model theoretic vs. proof theoretic, and in fact it is one of the original model-theoretic frameworks. I am not sure how important that distinction is, but it is intriguing that grammar can be looked at from such different perspectives.
So, my question to you is: What do you think the intellectual achievements of APG (the book not the framework) are? Why would you recommend that people look at it? What are the take-away important ideas? Just discuss a couple of the main ideas in lay terms accessible to a wider audience.
Paul:
I wouldn’t recommend that anyone read it. I would recommend that they not. Rather they should look at ‘Edge-Based’, which incorporates all of the important ideas in a more elegant form, leaves out lots of complex and pointless stuff and applies the ideas to lots of data in English.
The fundamentals contributions are to show how to construct a flexible and precise version of RG ideas, defining unclear notions like ‘advance’, ‘demote’, ‘raise’ in precise terms via the notion of successor/predecessor, in turn made possible by introducing two relations between arcs, SPONSOR and ERASE. The latter also permits a precise notion of deletion. SPONSOR and ERASE also permitted introduction of the REPLACE relation which, inter alia, permitted a rudimentary approach to some anaphora in terms of overlapping arcs.
I suppose the idea of stating principles precisely enough to permit some proofs is of some interest, but the principles mostly weren’t serious enough to justify the space. Too much of the book depended on the notion of coordinates of arcs, which proved, I think, to be both quite complicated and ultimately unnecessary. There are none in ‘Edge-Based’. I suppose one could say too that the APG framework shows that talk of the necessity of movement in a grammatical theory is quite wrong. APG has no movement, despite your tendency to talk in such terms. The necessity stuff was also undermined radically by the GPSG introduction of slash categories. Anyway, no one has ever shown how movement ideas actually apply to natural language constructions, e.g. to passives, considering such sentences as:
(1) The two gorillas were respectively taught French by Mike and taught German by Lucille.
60 plus years of TG and not a clue how to describe (1) consistently with talk of movement. What is supposed to have moved in (1) and to where? Of course, there is no APG treatment either, but the latter does not come accompanied by grand claims about revolutionary insights into the human mind.
Chris:
You have mentioned such sentences to me on several occasions in the last ten years. I think one can show the same thing using coordination too. Could you briefly explain what problem this sentence raises?
Paul:
The problem is palpable in any view of syntax which recognizes in passives generally a one to one relation between object (positions) and the subject positions of passives. There is such a relation in simple cases like:
(2) The gorilla was taught by Mike.
But in (1) there are two object positions, one for the leftmost ’taught’, the other for the rightmost.
But only one subject. Moreover, in some sense, the semantic parts of that subject are understood
as distributed to the two object positions. TG analyses just grind to a halt in the face of such sentences.
Chris:
Has anybody ever worked on this problem? Have you written about it?
Paul:
Yes, there is a bit of discussion in my ‘Three Investigations of Extraction’, 4.2.2.14 Interwoven Dependencies. I believe that those working in variants of categorial and logic grammars have worked on such cases. See for instance Bob Levine’s paper ‘Biolinguistics: Some foundational problems’ and references therein found in Christina Behme and Martin Neef, ‘Essays on Linguistic Realism’, Benjamins, 2018.
Chris:
Could you describe for your readers what “model theoretic syntax” is and what the difference is between that and “proof theoretic syntax”?
[C.C., See also David Johnson and Geoffrey Pullum’s mini-interviews for comments on this distinction.]
Paul:
Any grammar is going to be a set of formal objects, probably strings, defined by some syntax. A generative/proof theoretic grammar will interpret these formal objects as the equivalent of some Turing machine; they will be a set of instructions to form in a mechanical way a collection of other formal objects. The interpretation of the output is that all of the objects which the instructions yield are grammatical; the system is supposed to be such that each grammatical sentence is formed by the system after a finite number of applications. While Chomsky early in his career spoke of grammars as analogous to scientific theories, that was never true since his notion of grammar was always of the proof-theoretic kind. It was probably true that a grammar together with a claim that it determined all and only the objects of study was a theory. Such grammars are defined precisely in any book on computer science or formal language theory.
A model-theoretic grammar interprets its contained formal objects as statements, formulas to which truth values can be assigned. The formulas will no doubt be equivalent to quantified statements in some second order logic. They will define the language at issue via satisfaction. That is, X will be a sentence of the language L defined by model-theoretic G(L) if and only if X satisfies every statement in G(L). In other words, such a grammar will be analogous to an axiom system in logic or mathematics, not analogous to a computer program.
As argued in my ‘Skeptical Linguistic Essays’, one of the virtues of a model theoretic approach is that it is not bound to a fixed lexicon, permitting it to characterize direct discourse, foreign metalinguistic statements, etc., which a generative grammar cannot. I terminologically characterized this by saying natural languages were open, whereas generative grammars can only characterize closed systems.
Generative grammars offer no way to even specify that sentences of the form (1) are fine in English:
(1) The word W is a verb in Y
where W is a variable over words from the languages over whose names Y is a variable.
Ditto for sentences like:
(2) Marlyne screamed ftobskriflup three times.
Chris:
As I noted earlier, one of the innovations of APG was to used multi-dominance to describe various syntactic phenomena. Multi-dominance is also being used more and more in minimalist literature. Do you have any thoughts on that?
Paul:
Sometimes an encounter with a particular piece of grammatical research raises questions both small and large about the way linguistic research deals with past work and thus how the history of the development of the field is taught. Such questions arose in my mind when I encountered the Doctoral Thesis About Sharing, a 2007 work by Martina Gracanin-Yuksek accepted at MIT, often rated as the number one linguistics department in the United States. This work was directed by four well-known scholars, David Pesetsky, Danny Fox, Norvin Richards and Sabine Iatridou.
Gracanin-Yuksek’s dissertation seeks inter alia to exploit a notion of multidomination in linguistic structures. That references situations where a single node has more than one distinct mother node. Such structures are expressly barred by the single mother condition part of standard definitions of phrase structure tree. Thus multidomination played no role in the first forty years or so of transformational grammar, which depended entirely on that notion of tree.
In reading the beginning of Gracanin-Yuksek’s dissertation, one is inevitably led to believe that the author assumes that the idea of multidomination is relatively new...she cites four works from the 1980s, none of which is at all typical of the exploitation of this idea at an earlier period and all of which invoke it only in very limited and construction specific ways. She also cites works from the 1990s and later. Apparently then, this author was entirely unaware of whole traditions of work which systematically exploited the idea starting in the 1970s. Such a notion was central to the relational grammar work of the 1970s and 1980s, which she nowhere mentions; see e.g. the three volumes of Studies in Relational Grammar published by the University of Chicago Press in 1983, 1984 and 1990. These contained inter alia work dating from the 1970s. The idea of structures not observing the single mother condition was embedded in an explicit formal framework in Johnson and Postal ‘s (1980) Arc Pair Grammar (Princeton University Press) and exploited in numerous papers and several books by me thereafter, specifically Edge-Based Clausal Syntax, MIT Press, 2010. Moreover, the latter work cited predecessors going back to the 1960s including Hockett’s (1966) article ‘Language, Mathematics, Linguistics’ in Thomas Sebeok (ed.) Current Trends in Linguistics, Mouton. and Morin and O’Malley’s (1969) article ‘Multi-Rooted Vines in Semantic Representation’, in Papers from the Fifth Regional Meeting, Chicago Linguistic Society. I quote from the latter (page 182), but provide only one of the diagrams from the original article:
“Deleting axioms A2 and A3 generalizes the definition of a tree so as to permit more than one node to dominate a single node. Such a structure is a type of directed graph without cycles and will be called a vine. A vine is a set of nodes N with relations H and L such that L is a total order and H is a partial order.”
Johnson and Postal (1980) also cited Geoffrey Sampson’s (1975) paper ‘The Single Mother Condition’, Journal of Linguistics 11, 1-11. This paper argued for rejection of the single mother condition on the basis of various facts of anaphora. The proposal was criticized and rejected in Robert D. Borsley, ‘In defense of single mothers’, Journal of Linguistics 6, 1980. Johnson and Postal (1980): also cited S. Saito’s 1976 article ‘On Dropping the Single Mother Condition: Notes on Split Controllers and Antecedent-Contained Pro-Forms’, Journal of Linguistics, 12, 1976 and George Lakoff’s 1977 article ‘On Linguistic Gestalts’ in Papers from the Thirteenth Regional Meeting, Chicago Linguistic Society.
Variants of the idea of elements with multiple mothers were adopted in Lexical Functional grammar, see Joan Bresnan’s 2001 Lexical-Functional Grammar, Blackwell, Oxford, England also unmentioned, and in the HPSG framework, see Carl Pollard and Ivan Sag’s (1994) Head-Driven Phrase Structure Grammar, University of Chicago Press, and its later developments; see the explicit acknowledgement in note 2, page 2 of that work.
In short, there was extensive discussion of, and proposals about, the posit of linguistic structures in which elements had multiple mothers decades before Gracanin-Yuksek’s thesis was accepted.
Given all this, it is rather remarkable, for example, to see on page 7 of that thesis a diagram of, and discussion of, an extraction construction in which the author speaks of such as involving a node which has a low mother and a higher one representing the position to which it is extracted as well. Reading this passage clearly gives the impression that the author believes she is stating some novel idea here. But e.g. Johnson and Postal (1980), probably written before the author was born, contains various examples of exactly that sort, e.g. (130) on page 263, (134) on page 267, (81a, b) on pa 325, (96a, b) on pages 332, 333. I could go on, but it is evidently tedious.
That the author was apparently ignorant of all this work is perhaps surprising perhaps not. But as a student, she depended on departmental people of authority, specifically her thesis directors. It was their job to dissolve this kind of ignorance, and to provide her with a reasonable breadth of knowledge of the hardly secret undeniable antecedents of intellectual ideas she was involved with. This is especially true since the relevant concepts seemingly ended up providing the core of her thesis. Instead, the instruction in her department seems to have left her completely in the dark.
I can’t specify the reason. Could all of the thesis directors really have been so insulated from ideas outside of the mainstream MIT tradition as to be ignorant of these matters? Had they not heard of the works I cited? Or, worse, knowing of this work, did they decide for some reasons difficult to imagine that it was not their duty to make her aware of it?
The implications of the facts documented above are varied. Of lesser importance, at the time About Sharing was written at MIT, it was possible to successfully submit there a Ph. D. thesis which gave a radically distorted vision of the history of the ideas which it dealt with. This would be of limited importance if the MIT linguistics department was a minor influence in the field. But obviously this is not the case. Raised is the question of to what extent similarly flawed linguistic instruction was characteristic of other departments.
Still more important is the question of whether and to what degree the type of lacuna revealed above has been characteristic of syntactic instruction in linguistic departments for decades. While I have documented only one particular case, I suggest more generally that for a very long time in very many places, and not only in the United States, the history of syntactic ideas has been badly represented in ways which favor one tradition and disfavor others, with such bias not grounded in argument or evidence.
Chris:
The word multidominance appears six times in Collins and Stabler 2016 (including in the abstract). In effect, the concept of multidominance was a central topic of discussion in the paper. But we did not mention Relational Grammar or Johnson and Postal or the other sources you cited. I attach the paper. So we are guilty of the same mistake that Grancanin-Yuksek was. But for us, the mistake is even worse, since I of course knew about Johnson and Postal, and had read large parts of it before Collins and Stabler appeared. In fact, I have always thought that the use of multidominance in Johnson and Postal was innovative and interesting. I am not sure why we did not include the relevant references. I am not even sure that the possibility occurred to me. I am guessing that we were trying to cite relevant literature from minimalist syntax, and did not think of citing historical antecedents. But that is really no excuse. It seems to me to be a big omission in retrospect. What are your thoughts?
Paul:
I am not sure your non-citation is worse than the MIT dissertation case, since no student was involved. There are various reactions and various moves you could take. One is the traditional Japanese one: you could commit seppeku to erase the shame. Probably, to be discouraged. Second, you could do what I did when Robert May pointed out to me failures of attribution in my 1993 paper, Remarks on Weak Crossover Effects, LI 24. When I had a chance, I corrected; see note 2 on page 365 of the attached. I think this is the only time I was ever accused of this sort of thing and the accusation was entirely justified. I did not commit seppeku. In general, one should not be past oriented here…the past is over and done. The question is how to deal with the issue in the future, and the answer is obvious.
Incidentally, I would like to see the Kracht 2008 paper you cite. In the unlikely event you have an electronic copy, please send.
It would of course be good if the trio of thesis directors at MIT saw my comments. But I am certainly not going to send the passage I wrote to them. But if you could think of some way for it to happen without your making any more enemies, that would be good.
I think incidentally that your minimalist commitment faces much much bigger issues than failure of multidominance attribution. See sentences like:
(1) The two words ‘hund’ and ‘erhar’ are respectively the German and Mohawk words for ‘dog’ and they have been respectively studied by lexicographers and ignored by syntacticians.,
There are two vicious issues. First, the two words cited are not English items and thus a generative, Turing machine grammar has no way to get hold of them. Their choice was obviously arbitrary and they could be replaced by words from any language on Earth on from any other languages. A rather ironic consequence is that generative grammars offer no obvious way to do the descriptive linguistics of foreign languages.
Second, there is the ‘respectively’ problem. There are analogous issues with respect to direct discourse sentences…all this is discussed in Chapter 6 of the work I just sent. Basically, I take the metalinguistic and direct discourse facts to show definitively that the whole idea of generative grammars was misguided from the outset.
Chris:
You collaborated with David Johnson on the book ‘Arc Pair Grammar’. What was his background? Was he a linguist, a computer scientist, a mathematician? How did you meet him? What was it like to work with him? How long did it take you to write APG?
Paul:
Yes. I don’t recall much about his history. He was a hippie for a while in SF; he went to graduate school at University of Illinois, Urbana-Champaign, got interested in generative semantics from McCawley students, and then RG. I don’t recall how he got a temporary job at IBM, which permitted us to work on APG. His father was a physicist and he had a lot of formal background and talent, and he really pursued that later after he got a full time IBM job.
Chris:
What was it like working with David Johnson? How did you divide up the work?
Paul:
It was very easy and very pleasant. He is an unassuming person, reasonable, rational…no particular axes to grind. I have no memory of how we divided up the work. The only difficulty was the physical atmosphere. Recall this was during the Carter administration. They issued an edict that no energy could be used to lower temperature below 78. It was a hot summer. The Research Laboratory of Electronics was a closed building…no openable windows. Of course, it was air conditioned. But it had a strange system. First, if I recall, the system cooled the air down, but it got too humid, so they heated it up, or maybe it was the reverse. Anyway, the output was that it was too warm, very humid. I remember sometimes just sitting on a desk paper got wet. And of course, we did the whole thing on typewriters with white out, cutting and pasting. It seems inconceivable.
Chris:
In McCawley 1968 ('Concerning the Base Component of a Transformational Grammar'), he proposed that "...the base component is a set of node admissibility conditions". Is this proposal a precursor of model theoretic syntax proposals?
[See also David John’s response to this question in his mini-interview.]
Paul:
Yes, there is a grain of the idea there, but it is really developed in the framework of generative conceptions. And note he was only talking about a base component, so that he was thinking
that the output was input to generative transformational rules, an idea he never abandoned…see his 1998 book on English syntax. George Lakoff, as I recall, also had some nongenerative thoughts somewhere, but I forget details. I suspect that the APG book is the first place where a pure conception is advanced. But note we got some ideas from a guy named Epstein (I think) who was hanging around IBM at that point. Maybe few advances in any field are the work of a totally novel conception unique to one person or one associated n-ad of people.
that the output was input to generative transformational rules, an idea he never abandoned…see his 1998 book on English syntax. George Lakoff, as I recall, also had some nongenerative thoughts somewhere, but I forget details. I suspect that the APG book is the first place where a pure conception is advanced. But note we got some ideas from a guy named Epstein (I think) who was hanging around IBM at that point. Maybe few advances in any field are the work of a totally novel conception unique to one person or one associated n-ad of people.
Chris:
Do you know of other model theoretic approaches other than APG? Are GPSG/HPSG model theoretic approaches?
Paul:
Well, Pullum and a fellow named Rogers were writing a whole book on the topic, but the latter got sick and the project crashed. It would have had a lot of formal stuff. Here is a paper which will lead to where they were/are going insofar as they are still going anywhere.
I don’t know enough about GPSG/HPSG to comment seriously. I know they talked about a system being ‘constraint-based’, but one would have to look at their whole mechanisms which changed over time and evolved. I always found their terminologies and diagrams and symbolism offputting and never really tried to internalize any of it, except on a few occasions for minor points.
Chris:
How did you first meet Geoffrey Pullum?
Paul:
In 1974 I spent part of LSA summer institute in Amherst. He was there at least for a while and we met at a party.
Chris:
You met Pullum 1974, then you guys wrote your first article together in 1978 basically on wanna-contraction. I believe you wrote five papers together, so something about him resonated with you. Tell me about your collaboration with him
Paul:
Looking back, it seems quasi miraculous to me that we were able to jointly write five papers before the existence of the internet. That because we were rarely in the same place, usually not even on the same continent. I guess we did it by hard mail, although I have no precise memories.
For a while GKP was interested in relational grammar, and for a short period in APG. Then he fell into GPSG and those interests lapsed but did not become hostility.
GKP is a remarkable scholar, unbelievably learned in a wide range of areas. He is very sophisticated in formal language theory, can grapple successfully with non-European languages, is a superb writer (probably the best in linguistics, certainly the best I have ever known). He is or used to be a workaholic. If you look at his website vita you can’t help but be amazed at the work he has produced: http://www.lel.ed.ac.uk/~gpullum/
Chris:
We have mostly talked about your book, but I wanted to talk a little bit about on your very influential 1969 paper 'On So-Called Pronouns in English'. Your basic idea was that pronouns are definite articles. After more than 50 years, what do you think of your proposal?
Paul:
No, I haven’t thought about it, don’t want to think about it. I suspect it was wrong but don’t care enough to worry about it at this point.
Chris:
OK, that response will disappoint the readers. Can you give a hint of why you suspect it was wrong?
Paul:
Aa I recall it was all based on ‘we/you Americans’…but *U I/you American’. And French has articles, but *Nous Americains”.
Chris:
Other than MIT, CUNY and IBM, did you ever get any other job offers, or other departments expressing interesting in hiring you?
Paul:
Yes, I got job offers from each of the major branches of the University of California, UCLA, UC San Diego, and UC Berkeley. I don’t remember the dates, but the earliest was still in the 60s and the last, UCLA, was in 1979.
Chris:
Were these part of a job search, or were you just offered the positions? Also, why did you decline?
Paul:
The current notion of ‘job search’ is really inapplicable. At that time, when the powers in a department had an opening or could expand their staff, they talked about who they wanted and contacted that person with an offer.
I don’t really remember why I didn’t accept. The whole idea of moving to California seemed foreign to me, none of the departments AT THE TIME was staffed by people I had close relations with. My wife at the time was from Los Angeles, and had basically fled and had no desire to move back.
6. NYU (1993-2009)
Chris:
After you were terminated at IBM in 1993, you started to work at NYU. How did you land the NYU job? What kinds of responsibilities did you have?
Paul:
I think I contacted various people about a job, including Mark Baltin, and he was interested and managed to get me an adjunct position. I think my only responsibility really was to teach a course, maybe one a year or one a semester, I don’t remember. After a few years hardly anybody signed up. And I remember at one-point Anna Szabolsci was kindly arranging things (as chairman) to increase official registrees. I guess I also was supposed to meet with students at their demand, read the occasional qualifying paper. This rarely took much time. I almost never came into NYU more than once a week.
Chris:
What years did you work at NYU? Why did you finally end up leaving?
Paul:
1993-2009. I ended up leaving because they did not renew my contract, which, from the beginning was on a year to year basis.
Chris:
Did they explain to you why your contract was not renewed?
Paul:
No, I think not. I went to see Marantz about it. He just made it clear that he/they had no interest in retaining me.
Chris:
After your contract was not renewed, what was your status at NYU?
Paul:
I had a visiting scientist appointment which let me keep my e-mail and electronic access to libraries. Again, it has to be renewed every year.
Chris:
When did you move out west?
Paul:
At the end of March 2017.
Chris:
Where are you living now?
Paul:
Lynnwood, Washington, twenty plus miles north of Seattle.
Chris:
Do you remember any particular students you worked with? I know you have mentioned Chris Potts in the past.
Paul:
Well, the one student whose thesis I directed was Jayashree Nadahalli, who wrote about the syntax of her native language, Kannada. I of course interacted with her and Potts. Nothing special about any of the others comes to mind.
Chris:
Other than me, which other faculty members did you interact with at NYU?
Paul:
Well, a little bit Mark Baltin. For a while some with Anna. I can’t think of anyone else.
Chris:
In the interview above you said: “Sometime in late 80s I started working on extraction and anti-pronominal contexts, which yielded ‘Three Investigations of Extraction’, but not until five years after I left IBM.” To me this seems like a big shift from what you had been working on in the previous 10 years, which was mostly relational grammar. Can you tell me about what was behind the shift in interest?
Paul:
Because of interest in and work on French, I read Hans-Georg Obenauer’s and Guglielmo Cinque’s stuff which had original steps toward recognizing the role of antipronominal contexts. Once I saw that and began checking around in English, it became fascinating. And after a while I connected it to Haj Ross’ insight about resumptive pronouns and islands and David Perlmutter’s 1972 copy pronoun article. My memories of all this, dates, etc. are too vague to say much more.
Chris:
I think ‘Three Investigations of Extraction’ is a remarkable book for the amount of English data that it sifts through. I have taught it in my graduate Syntax II class several times. I know that several people in the field are now looking at the data and ideas from different perspectives. Could you summarize in three lines what an anti-pronominal context is and why it is important for extraction?
Paul:
An AC is a nominal context not permitting weak definite pronouns, illustrated by:
(1)
a. Jerome thinks that chimps have rights but Tod doesn’t think that/*it.
b. Jerome believes that chimps have rights but Tod doesn’t believe it.
ACs are important for the study of extraction because they correlate with many restrictions on extraction inter alia. Note:
(2)
a. *That chimps have rights, Tod doesn’t think.
b. That chimps have rights, Tod doesn’t believe.
(3)
a. They discussed the weight of the beast but they could not determine/*tell it.
b. The weight of the beast, they could not determine/*tell.
c. They could not determine/tell the weight of the beast.
One can connect these facts to a claim that some extractions leave invisible resumptive pronouns, and to a further claim that invisible resumptive pronouns have to be weak definite pronominals. Other extractions don’t require invisible resumptive pronouns, hence:
(4) What were they able to determine/tell about the weight of the beast?
Most previous work, took extractions to be pretty uniform…study of ACs shows pretty clearly that is not the case. Compare for instance restrictive vs. nonrestrictive relatives:
(5)
a. They painted the vehicle green/*it.
b. the color which they painted the vehicle...
c. *that color, which they painted the vehicle, ...
Chris:
Did you ever talk to Obenauer, Cinque, Ross or Perlumutter about the work? What did they say?
Paul:
I probably talked to Haj about the issue but I have no real memories. I only met Obenauer once in Paris, but I have no memory of what we talked about. I only met Cinque after the work and only once. I never talked to Perlmutter about ACs, etc. Recall that most of the relevant period was before the internet really got going, before Skype. So, any talk had to be by phone, hardly ideal for linguistics. And hardmail letters. I think I did exchange a couple with Obenaueer and Cinque but I can’t remember details, I think I did once apologize to Obenauer because I felt my original work on ACs did not give him enough credit. He didn’t seem to care.
I probably talked to Haj about the issue but I have no real memories. I only met Obenauer once in Paris, but I have no memory of what we talked about. I only met Cinque after the work and only once. I never talked to Perlmutter about ACs, etc. Recall that most of the relevant period was before the internet really got going, before Skype. So, any talk had to be by phone, hardly ideal for linguistics. And hardmail letters. I think I did exchange a couple with Obenaueer and Cinque but I can’t remember details, I think I did once apologize to Obenauer because I felt my original work on ACs did not give him enough credit. He didn’t seem to care.
Chris:
Your book ‘Edge-Based Clausal Syntax’ was published by MIT Press in 2010. This was two years before ‘Imposters’ was published.
Can you tell me a little about the origin of the ideas in ‘Edge-Based’? How did they come to you? When did you start working on the book? How did they relate to previous work of yours?
Paul:
I don’t have good memories of the particular origins, still less of the dates. I suspect that I had some of the ideas in the book for a long time but it took some particular impulse to work on them in such a way as to make a whole book.
Clearly the framework is an outgrowth of the 1980 APG volume and of RG work. It simplifies the original APG version, adds some new twists, relates the ideas to phrase structure in a nice way I thought. That is chapter 1. Chapter 2 is just a working out of the distinction between 2s, 3s and 4s. I began to see those distinctions early in my time at NYU. A key element was the abandonment of a dogma which animated RG from the beginning and was shared by Perlmutter and me. This was the claim that passives uniformly operated on 2s. Clues that this was wrong popped up early, but we sort of tried to ignore them. Any time we found a non-2 which seemed to passivize, we took it to be a case of an X advancing to 2 and then passivizing. The core of this idea in English is visible in cases like:
(1) Jerome sent a magazine to Clara.
(2) Jerome sent Clara a magazine.
(3) A magazine was sent to Clara.
(4) Clara was sent a magazine.
To keep (3) and (4) consistent with the passive claim, (1) had to be analyzed such that ‘a magazine’ was a 2, while (2) had to have ‘Clara’ as a 2. This could be done via a relational parallel to the Fillmore TG treatment in which the PP object became a 2 in (2). It was when I concluded that this was wrong, that the structure of (2) involved ‘Clara’ as a 3 and all that followed from it, that the basis of chapter 3 arrived. It required some noncommon ideas, such that there is a relation named 5 which 3s can demote to. So that where previous work had mostly assumed the advancement to 2 analysis for (2), I adopted a demotion of 3 to 5 analysis. There were precursors by other people for the essence of this analysis, but the demotion to 5 was my idea.
Chris:
You and I have worked together on camouflage (expressions like “your Majesty”) and imposters (expressions like “yours truly”) and negation. Actually. you were already working on negation when we started our work on camouflage and imposters (which was just after I came to NYU in 2005). When and how did you first become interested in negation? Once again, it seems like a big shift from the topics you had been looking at earlier.
Paul:
This is an easy question. In the summer of 1995 I visited Haj at his lake house northeast of Montreal. Somehow, we noticed the form squat and began looking into it, and found multiple weird facts. We continued back and forth for months; I think we were planning
a paper but nothing ever got done. At that point I was pretty much totally ignorant of anything beyond the grossest facts of negation so I started looking into the literature. Anna was interested and later in 2000 she even invited me to take part in her semantics course, where I took some of the time to talk about squat. That ultimately led to my 2005 talk at the conference Anna organized. All that time I was planning to write a book on negation, but I could never develop the ideas in a way which I found intellectually serious and which was simultaneously sellable to a general public. The first time that seemed feasible was with our NEG raising work.
a paper but nothing ever got done. At that point I was pretty much totally ignorant of anything beyond the grossest facts of negation so I started looking into the literature. Anna was interested and later in 2000 she even invited me to take part in her semantics course, where I took some of the time to talk about squat. That ultimately led to my 2005 talk at the conference Anna organized. All that time I was planning to write a book on negation, but I could never develop the ideas in a way which I found intellectually serious and which was simultaneously sellable to a general public. The first time that seemed feasible was with our NEG raising work.
Most of the things I have worked on over time have, no matter how much time and effort I put in, left me frustrated. As it seemed to always end up in dead ends and unanswered questions.
Chris:
Ah, nice story.
You say: “Most of the things I have worked on over time have, no matter how much time and effort I put in, left me frustrated. As it seemed to always end up in dead ends and unanswered questions.”
I think this is really important for syntacticians to hear. What kinds of things do you have in mind? Can you give me an example?
Paul:
There are more than I care to remember. When I was still at MIT I worked on a book on Mohawk Phonology, but abandoned it as I didn’t have enough data, no opportunity to get more and not enough grounding in phonology.
I was planning at one point a book on Mohawk syntax, but the same issues arose. Plus, grammatical ideas kept evolving to the point that I was not sure what framework I could adopt.
The planned relational grammar book with Perlmutter of course never came to fruition.
I worked a lot and wrote a lot on an APG account of the strong crossover phenomenon in the 1990s I think, but was never happy with it and abandoned it.
In work on negation in 2008-9 I tried extensively to write a book on negation and NPIs but never was satisfied with the accounts I had and abandoned that.
Throughout, in work on English syntax, I was always aware of the respectively problem and even in the sixties wrote about it, in an unsatisfactory way no doubt. The paper is attached.
Chris:
As I recall, you and I started working on negation around Spring 2011, because I was in Ghana then (teaching at NYU-in-Ghana), and I remember starting to think about things then. I had never really thought about negation before, so it was all new to me. But by that time, you had already been working on negation since 1995. So, I think that the syntactic framework we adopted, as well as many of the goals of our joint research on negation were due to you. I did contribute important empirical discoveries to the volume, and I certainly helped you work things out, especially with respect to issues having to do with the syntax/semantics interface. But I have always had the feeling that it was really your project, and I had gotten on board in the middle of it.
How do you see our relative contributions to Collins and Postal 2014?
Paul:
I think your timing and the fact that I was working on it earlier is right; I in fact wrote an article on NEG raising, which you criticized, and then after a while on the basis of your criticisms I asked you to take part in joint project to fix and expand it.
As to the final result, the key and really only relevant fact is that without our cooperation, there never would have been a work like we jointly produced. I have demonstrated that again in fact since I have been trying to write a new book responding to all of the many attacks on us but really can’t do it.
Chris:
You seem to have an ambivalent attitude toward formal semantics. In our interview, and in your interview with Huck and Goldsmith you point out the issue of the meaning of necessary truths. From Huck and Goldsmith, pg. 137:
“Note that the Montague type of semantics was already shown clearly to be a false theory in 1977, since it determines that all necessary truths have the same meaning and that all necessary falsehoods have the same meaning [see Katz and Katz 1977]. Supporters of Montague Grammar know about this problem. But their defense tends to involve vague promises about future revisions and irrelevancies like, ‘The advantage of this approach is that it has a precise, highly developed formalized semantics.’ Precision and formal development are undeniable virtues, but not ones capable of overcoming the flaw of false consequences.”
But then in our book Classical NEG Raising, we crucially rely on notions from formal semantics in defining the compositional semantics of negation. In particular, we gave negation a semantic value and showed how it combines with general quantifiers to yield the denotation of a negated quantifier phrase.
Could you comment on this ambivalence? Can a theory like Montague's be false in one way (e.g., with respect to necessary truths) but true and useful in another way (e.g., in studying the semantics of negation)?
Paul:
Yes, it could seem that way. The way to make the two things coherent and the hope for model-theoretic semantics is to take it as (at best) a theory of denotations, not a theory of meaning. And these are different.
Consider:
(1) The expression Jerome has the flu is an English sentence.
The bolded part has a clear meaning, but this cannot be by virtue of the combination of the denotations of its parts as model-theoretic semantics assumes, since e.g. Jerome here has no denotation. And thus, while the bold part is a declarative sentence, it makes no assertion.
Chris:
How does this resolve the issue that Katz brought up?
Paul:
The issue Katz and Katz brought up arises internal to the view that the model-theoretic ideas give an account of meaning. Once that is abandoned, the fact that necessary truths/necessary falsehoods come out with the same value, true in all possible worlds, is harmless, since that fact is not taken to determine they have the same meanings.
The latter determination is absurd, as one sees e.g. from the fact that one cannot translate (1) and (2) into the same French sentence.
(1) The only even prime is two.
(2) No circle is square.
That follows since translation is intended to preserve meaning, but (1) and (2) do not mean the same thing even though each is true in exactly the same possible worlds.
Let me add that there is probably some analogy to the way I handled the model-theoretic account of negation which you built into Classical NEG Raising and the fact that I do not think it gives an account of meaning and the way we had in all our joint works tried to work out a lingua franca mode of description which neither of us finds really accurate but which each finds contains enough truth to be worth working in for particular projects. I think there is some lack of frankness with readers in both cases.
Chris:
Could you explain the concept of lingua franca (as we used it) to the readers of this interview? It was absolutely crucial for our collaboration.
Paul:
This would be REALLY hard. Here is something.
Linguistic work is done at various levels of formality, tending toward the not very formal. Intermediate levels of precision have both advantages and disadvantages. One obvious advantage is that really precise, formalized work demands tremendous knowledge, time and effort and, to be serious, a degree of insight greater than what is often available. The vast complication of natural languages is such that one can arguably understand some things without having a broad enough or deep enough understanding to permit the wide, deep formal account one would ultimately like.
The approach usually adopted is to state things with various degrees of precision, adopting some overall framework of reasonably precise ideas and stating things in the terminology and conceptual system of the chosen framework. But intermediate degrees of precision permit joint work by linguists who do not necessarily favor the same particular framework, as in our case. They can then adopt common features they do agree on, while avoiding or backgrounding ideas they do not jointly accept. This may amount to creating what one can call a theoretical lingua franca containing an amalgam of ideas some of which would fit in one framework, some in another, and conceivably some in neither. Such an approach may be respectable for some studies, but is evidently far from the ultimate ideal of a single correct framework with fully explicit and correct concepts and principles.
Chris:
On the other hand, with ‘Imposters’, I think our contributions were much more evenly balanced, in terms of empirical discoveries and theoretical ideas. Recall that Imposters was an outgrowth of our work with Simanique Moody on AAE camouflage constructions. I think it is the richest work empirically that I have ever produced. Literally every page is covered with new observations. And I remember clearly the excitement that I felt in writing it. I was probably spending around 20 hours a week during that period just thinking of imposters. I feel very lucky that I got to participate in a project like that.
Do you have any thoughts that you would like to share on our collaboration in writing ‘Imposters’?
Paul:
Well, I agree on the balance question, but again it was a joint work and we both contributed. I don’t have clear memories of the overall process. It is true we discovered lots of cool stuff and I think we had excellent insights into some things. Particularly, our account of pronominal agreement with its notions of primary and secondary source is, in my opinion, really a breakthrough. And our account of person assignment in coordination was really superb. The fact that these seem to have raised no general interest is rather depressing.
On the other hand, I remember never being satisfied with the structures for either imposters or camouflage structures, and they still are mysterious in my mind. And negatively, the problems we raised for super-semantic approaches like Polly’s and Tanya Reinhart’s were really devastating. I don’t know if anyone since has taken them into account. For a while you stimulated a good amount of cross-linguistic study of imposters. I have no idea how much interest or attention that continues to lead to.
I also remember extreme frustration with respect to yours truly is proud of myself. Every time I googled I found new examples of this type, but never, never despite real efforts could I ever find a living being who would accept them. I still don’t know what to make of this, especially given the perfection of plural cases like Jane and yours truly are proud of ourselves, which I have not found anyone who rejects.
Chris:
In writing Imposters, we used the internet to a greater degree than I (at least) had ever used the internet before in my work. I think that experience really shifted my perspective on syntactic data. Do you have any comments on the use of the internet in syntactic research? What does this new kind of data mean for the field? What are the kinds of things we need to watch out for in using it?
Paul:
Evidently, internet searches provide a vast new source of data. This has obvious advantages but also disadvantages. Because one has no control over the sources, are they different dialects, foreigners with imperfect language knowledge, jokes, deliberate distortions, etc. And there are complete mysteries, like the fact that internet provided many instances of ‘yours truly..myself’, but no human has been found to accept it in person. One thing I conclude is that the range of dialect and individual syntactic differences is greater than one would have imagined. For instance, I thought usages like ‘didley beer’ were exclusively British, but I was assured by a respectable source that they are found in a data base of American English. To me such usages are entirely foreign. Apparently, we know much less about the range of variation than we imagine we know.
I suspect that the internet data will be a boon for those who like to pick holes in generalizations, and a problem for those who like to formulate them.
Chris:
I will call those writings of yours questioning Noam Chomsky's behavior your anti-Chomsky writings. During our collaboration together, over the last decade, did we ever discuss your anti-Chomsky writings, or did we focus completely on discussions of syntax?
Paul:
Not only did we not discuss them, because you found the topic upsetting, we instituted a policy of not discussing the issue at all.
7. Influences
Chris:
As you look back on your career, from graduate school until today, who were your biggest intellectual influences?
Paul:
Originally Chomsky, Lees a bit, Halle for phonology. Haj became a big influence through his thesis and introduction of island notions. Later Perlmutter had some influence, bringing to bear vast knowledge of European languages, some Japanese…some relational ideas. After a while Jerry Katz became a major influence for foundational issues and semantics ones (he presented objections to model-theoretic semantics which have never been answered adequately to my knowledge). David Johnson contributed a lot when we worked on APG. Nicolas Ruwet was someone I communicated with about French at a stage when I was massively ignorant about it. I think Richie’s thesis was a real influence for a while; I learned a good bit about French grammar from it. But my ideas deviated more and more from his over time.
Chris:
Believe it or not, I think you and Richie have important intellectual characteristics in common. Both of you are firmly committed to syntactic explanations of phenomena at the syntax/semantics interface. Both of you love to dig in and find interesting new data in English. Both of you have informed my opinion about how vast and unexplored English grammar really is. Both of you are committed to the existence of empty elements and their importance in syntactic explanations, to a greater degree than the average syntactician. Of course, there are lots of differences, but these similarities are substantial.
What do you think of that?
Paul:
I agree that on this level of generality these similarities are real.
Chris:
Tell me more about Ruwet. What ideas of his in particular did you find interesting?
Paul:
I don’t remember much. He was a big fan of TG and Chomsky’s ideas at first and, if I recall, wrote the first book introducing generative grammar in France. But he became disillusioned, and felt the application to French at least was too superficial and ignored all sorts of specific restrictions and semantic facts.
Chris:
You said you were influenced by Lees. Did you ever meet him? If so how and when? And what kinds of things did you talk about?
Paul:
Sure, I met Lees. He was around MIT when I arrived. I don’t remember what we talked about, probably TG, superiority of that to the non-TG of the time. He had one amazing trait. He wrote everything in one sitting. I mean if he wanted to write an article about X, he just typed up the article and it was done. I really admired that. Because I am not the greatest writer even now, at that point I was truly terrible, and everything had to be done, redone, fixed, improved, etc.
Chris:
Did you see this announcement?
Antisymmetry and Comparative Syntax
CELEBRATING 25 YEARS OF ANTISYMMETRY AND 75 YEARS OF RICHARD S. KAYNE
Paul:
No.
Chris:
I was wondering if you had any thoughts on Kayne's thesis?
Paul:
As for his thesis; it was a very long time ago, .5 centuries in fact. So, my memories are few and vague. I remember that it was an impressive work for the time, uncovered a mass of interesting facts, and was surely a milestone in the generative study of French Syntax. It was a major factor getting me interested in the topic. Of course, it was a work of its time, but what isn’t?
Chris:
You also knew Ruwet, right? There seemed to a lot of interesting work at Paris VIII in the 70s.
Paul:
Yes, I knew Ruwet. He was a major figure along with Richie in introducing generative ideas into French linguistics. There was a lot of resistance; Maurice Gross was sort of on board, but he had his own somewhat odd views, but he was a superstar a far as getting money, which was and no doubt still is enormously difficult in France.
He once organized some kind of conference in an old abandoned NATO base, which was a huge Quonset hut complex like MIT building 20. Anyway, the French government somehow sponsored this linguistic conference, paid my way there and once I got there, some employee motioned me to follow him, and we went through lots of corridors to an office with some official in it. I had no idea where we were going or why. But the official had me sign a paper, and then he handed me an envelope filled with hundreds of francs. No one ever explained why. But that was Maurice in action.
Chris:
Did you ever go to Paris VIII or give a talk there?
Paul:
I don’t remember the school numbers, but I did give a talk at Vincennes in 1973. That was before they moved to Jussieux. It was unbelievably disgusting there, dirty, posters and graffiti everywhere, height of hippie culture.
Chris:
What did you talk about?
Paul:
Early relational grammar; pretty bad talk as I remember, which is not very much.
Chris:
I am just trying to figure out the role and atmosphere of Vincennes, which seemed so vibrant.
Paul:
Well, Richie is the one to talk to. He was actually living and teaching in Paris. Ruwet, Gross, etc. are dead; Perlmutter was there for a semester or a year and is not dead. Gilles Fauconnier was a student there; I remember him from my trip in 1973, in particular, we were walking somewhere and he was having a rather nasty argument about something with a university higher up whose name I have forgotten. Pollock was around then. Those are the people who would know something.
Chris:
Who were some of the French syntacticians that you communicated with about French syntax when you were working on it? Were there any that were particularly helpful?
Paul:
Basically, I have never really had the opportunity to communicate seriously with French syntacticians. My earliest memories are getting a few insights from David Perlmutter, who was probably nearly native in French, also from Maurice Gross and Nicolas Ruwet. But as we were rarely in the same place, most of the interaction was by mail, which is evidently a very poor medium for discussing syntax. Of course, I read a lot of papers by various people and looked into the traditional literature. Once I got seriously interested in working on French syntax, most of my contact was with the two paid informants I got IBM to subsidize, not that much money was involved. Also, over time, I contacted a number of French linguists as informants, the way we did e.g. J-Y. Pollock for our imposters book. But during this work, my major interest was trying to show that the facts of French could support some of the ideas of APG. And there was never a French linguist who knew anything about those ideas, so there was not much to talk about.
8. Accomplishments
Chris:
What do you take your most important and interesting syntactic contributions to be? I would put the unaccusative hypothesis, as outlined in your early letter to Perlmutter, to be one of them.
Paul:
1. Getting the idea of primitive clausal grammatical relations like subject, direct object, to a prominent position in syntax (with David Perlmutter). While these so-called relational grammar ideas are in some sense traditional, under the influence of Noam Chomsky, the idea took hold that all such notions could be defined in phrase structure configuration terms.
2. Introduced the conception of unaccusative verbs (term due to Geoffrey Pullum).
3. Developing (with David Johnson) a precise conception of grammar in which the partially vague and partially inchoate ideas of relational grammar could be expressed precisely and effectively. This work based on a complex graph-theoretical conception of syntactic structure showed how the facts treated under various ideas of movement in different versions of transformational grammar could be given an insightful account in graph-theoretic terms.
4. Developing (with David Johnson) the idea of grammars as model-theoretic systems as opposed to the proof-theoretic/Turing machine view made dominant in Noam Chomsky’s work. Offering arguments for the superiority of model-theoretic grammars over proof-theoretic ones.
5. Developing (originally with Jerrold Katz) an extension of Katz’s original arguments that the dominant view of natural languages as psychological (biological) objects is mistaken.
6. Discovering (as far as I know) the strong and weak crossover phenomena.
7. Expanding and developing original work by Ross, Perlmutter, Obenauer and Cinque recognizing the role of (invisible) resumptive pronouns in controlling a range of grammatical phenomena, including extractions, parasitic gaps, passives.
8. Developing and justifying a distinction between three types of English objects, labelled 2, 3 and 4 objects. The former two correspond to the traditional direct and indirect objects; not clear that the third has any direct previous correspondent. Showed how this range of concepts is critical to understanding the facts of English passivization.
9. Developing a dominantly syntactic conception of NPIs as syntactic negative phrases, whose NEGs are not found on the NPIs themselves due to NEG raising and NEG deletion.
10. Developing (with Chris Collins) an account of so-called imposters and so-called camouflage structures with an associated conception of pronominal agreement. As a result, the study of imposters and camouflage structures came to be seen as a general topic of grammatical investigation (cross-linguistically). This work had strong consequences for the characterization of pronominals and their relation to antecedents, overt and covert.
11. Developing (with Chris Collins) a series of arguments favoring a syntactic conception of so-called NEG raising in opposition to the dominant view that the phenomenon can be entirely reduced to semantic/pragmatic principles.
Chris:
There is also your book ‘On Raising’, which I think is significant.
Paul:
Maybe but it is a particular claim about one construction in one language…not worth it.
Chris:
I disagree with you about ‘On Raising’. First, the book set the tone for research. It is possible to go very deeply into one single aspect/rule of English grammar. That is important. Also, it turned out to ultimately convince people, even though it took a while. This was when Lasnik and Saito picked up some of your arguments and restated them in minimalist terms. I definitely think it is a contribution.
There is also the accomplishment of Katz-Postal, bringing semantic concerns into generative grammar.
Paul:
No, not worth mentioning. It had a slight effect on Chomsky for ‘Aspects’, and vanished after that, drowned in wave of interest and support for Montague style possible worlds semantics and other versions later.
Chris:
Also, I don't think Cinque or Obenauer really articulated the notion of “anti-pronominal context” as clearly as you did. They did have empty pronouns, but did they really investigate the notion of “anti-pronominal context” independent of those extractions?
Paul:
You may be right, but I don’t know and I don’t want to do the massive amount of work that would be needed to investigate the issues. I pretty much no longer have the relevant literature.
9. Foundational Issues
Chris:
I have been reading your papers on the ontological status of grammar and sentences. I wanted to go into that a bit. From your point of view, a sentence is an abstract object. Can you explain that?
Paul:
An abstract object is usually taken to be an object like a number, a symphony, a set, a proposition. They are things which exist but do not exist in time or space, do not enter into causative relations, have no physical properties, thus cannot be destroyed. So, one cannot weigh the English sentence ‘Most birds have fur’, one cannot find the location where it is (since there is no such location), one cannot say when it will end, since it had no beginning and can have no end. To maintain such a view of course one must inter alia separate out the sentence from pronunciations of it, which involve physical things moving, the resulting sounds, etc. More importantly, one must not confuse sentences with knowledge of sentences, the latter some kind of mental thing, which cannot share more properties with sentences than the physical coding of a program like WORD in a computer can share with the abstract object which is the program WORD.
Chris:
Now, if I hear a sentence like (1):
(1) Daddy and Mommy will enjoy ourselves at the beach.
How do I know that (1) is grammatical?
Paul:
A short answer is that the internalized grammar says it is. In model-theoretic terms, there is no statement represented in that grammar which it fails to satisfy. But that pushes the mystery one level higher. How did that grammar form? I have nothing to say about that. But there are parallel mysteries which few seem to worry about. How does one know that Modus Ponens is right? How does one know that the sum of 87 and 413 is 500, etc.
Chris:
What do you mean “internalized grammar”? I thought that grammars were also abstract objects [C.C., according to Paul Postal’s writings]?
Paul:
Yes, for sure. Like computer programs are abstract objects. But programs are physically represented in some way in computers, by taking advantage of a one to one correspondence between some physical properties, being in state A or not being in state A, and the symbols 0, 1. We say informally that e.g. computers do calculations. But they really don’t. What they do is various physical things, shift from one state to another, etc. We just interpret what they do as calculations, which is mostly safe to do because they have been designed so that their internal goings on code the relevant properties of calculations. Just so grammars, which are abstract objects, must have some representation in minds or brains if one thinks minds are brain based. There is no equivocation. These matters are much clearer for the computer case because so much more is understood (essentially everything) about what goes on in computers than is understood about what goes on in minds or brains.
Chris:
So, you are saying that there is a grammar that is an abstract object, and an internalized grammar that is part of the mind? Is that right?
Paul:
Yes.
Chris:
Therefore, our grammaticality judgments are based on the internalized grammar (part of the mind). And the internalized grammar is based on the grammar as an abstract object. So, there is an indirect relation between our grammaticality judgments and the grammar as an abstract object. Right?
Paul:
Right.
Chris:
Would it be OK with you if I called the two kinds of grammar: internal vs. external grammar?
Paul:
‘internal’ is no problem. But ‘external’ suggests it has a location, when as an abstract object it is not anywhere. How about ’non-internal’.
Chris:
So, what separates you from Chomsky, as I understand it, is that both of you have an internal grammar, but only you have the non-internal grammar. And for Chomsky, we (as syntacticians) are studying the internal grammar. But for you, we are studying the external grammar, via our grammaticality judgments based on the internal grammar. Is this correct? This allows you to avoid the infinity paradox (described in your papers). Is this correct?
Paul:
Essentially all right, except. Neither of us has an internal grammar; what we have is the idea that there must be such. And nobody has a non-internal grammar either, just fragments of ideas of what it should say.
But Chomsky has in addition, and has had since at least 1972, the unargued claim that there is nothing but the internal thing. But then as you refer to, he falls into endless incoherence because he says things which make no sense if taken to characterize the internal, such that it involves set formation under Merge, has infinite output, etc. A sad summary of his ontological view is simply that he mixes up two fundamentally different things and refuses to really admit it even after it has been repeatedly pointed out. And to add insult to injury, he criticizes other people for not adopting his confusions.
Chris:
Let me rephrase that to get it right:
So, what separates you from Chomsky, as I understand it, is that both of you posit an internal grammar, but only you posit the non-internal grammar. And for Chomsky, we (as syntacticians) are studying the internal grammar. But for you, we are studying the non-internal grammar, via our grammaticality judgments based on the internal grammar. Is this now exactly correct? I have more questions, but I want to get each step right.
Paul:
Yes, but to clarify. I am not telling anyone what to study. My point is that there is an internal grammar and a noninternal grammar, as there is a physical representation of WORD and a nonphysical program. All I object to is failing to distinguish these, to insist that there is only the mental thing (which is claimed to be biological) but ascribing to it properties (infinity, set-theoretic nature) which it cannot have. The fact is we really don’t know much of anything about how to characterize an internal grammar, because so little is understood about human mentation (in contrast to computers, where computer science knows just about everything that goes on in computers because computer science made them).
Chris:
Ok, now some follow up questions. Did Katz have the notion of internal grammar?
Paul:
Surely. Although I don’t remember details.
Chris:
Do we also need to talk about internal sentences and non-internal sentences?
Paul:
Well an internal grammar has some kind of outputs and one needs to call them something.
Chris:
Would it make sense to say my internal grammar matches a non-internal grammar? What word would you use “matches”, “corresponds to”, “represents”?
Postal:
The vaguer terms are far superior since we don’t really know what the relations are, keep in mind we are thinking, rightly or wrongly, that the internal grammar is analogous to an installed computer program and thus it is some kind of physical thing, state, set of states, who knows.
Chris:
People often talk about idiolects. Is there a different non-internal grammar for each idiolect?
Paul:
I would think so. How else could speakers know their different idiolects.
Chris:
How is language acquisition characterized in these terms?
Paul:
Well, this is hugely controversial and I have absolutely nothing to contribute to the matter.
Chris:
Does your concept of grammar allow for levels of grammaticality (e.g., ? vs. ?? vs. *)?
Paul:
Ah, here you need to pay attention to what Pullum said about model-theoretic grammars. They offer effortlessly a super approach to this area, since they permit talk of violations of one or n statements, and of statements of different kinds. This requires no additions to the characterization of the grammars. Forget not though that many of the unclear reactions represented by ?, ??, ???, etc. are probably due to other factors like center embeddings, memory problems, confusions about what was heard, ambiguities leading to non-focusing on the intended reading, semantic constraints, pragmatic constraints, etc.
Chris:
You often compare natural language syntax to math and logic (all three are abstract objects). But to me it feels different from math and logic. Putting aside logic for the moment, it seems to me that syntax is much more closely related to human activity than math is. For example, for natural language syntax we rely on grammaticality judgments. But is there something like this for mathematics? Or to put it another way, we might expect aliens to come up with roughly the same mathematics that we have. But we probably would not expect them to come up with the same natural language syntax. I am really just brainstorming, but I am trying to place my finger on my qualms about your way of looking at grammars.
Paul:
The central claim is that all three as well as music, laws, etc. involve abstract objects. It does not follow that there are not serious differences among the abstract objects. And even in logic, there is not uniformity. There are different logics, even some which claim (I think this is crazy but serious logicians work on it) that there are true contradictions.
Now Pullum has written a lot on the point you make about language and human nature. I answer his views in detail in my ‘Seven Discourses’.
My key point, I would say, is that of course instantiated languages, that is, those that have been learned, relate to human nature indirectly. That follows because they have to meet the conditions of human learning to be learned. For example, there is no linguistic reason to impose a length condition on the number of phonemes in a morpheme. But certainly, there is no learned language with a morpheme 900 phonemes long, because such could not be learned.
This line of thought shows how wrong it is to identify languages with learned languages, that is, to say that all languages are learnable. In fact, for reasons like that alluded to, almost none of them are.
I do not know if syntax is more related to human activity than math is. I haven’t heard of any math not related to human activity. Recall that for a while it was assumed that regular geometry was the only geometry. Then non-euclidean geometries were developed.
I think it clarifies matters to distinguish languages per se from the learned subset. The latter is where the relation to humans shows up, simply because learning X has to meet the conditions provided by the organism.
Incidentally, I have a paper, entitled Books, on LINGBUZZ 004733, which argues that books are also abstract objects, hence not created by authors, hence not destructible. What writers do then is create knowledge of some abstract objects. Such knowledge is in principle obtainable without writers but in fact not because there are so many combinations of sentences forming books that there is no way without an author that a person could ever focus on any specific combination.
Chris:
Let's say hypothetically that a sudden advance in brain science has now made neurolinguistic imaging of sentence structures in the brain possible. And (again hypothetically), the images look more or less like the tree or graph structures that syntacticians commonly use. In fact, the technology allows us to ask and answer questions about how to draw certain kinds of sentences. We just speak a sentence to a human subject, and see what kind of structure comes up on the imaging machine. Through such a device, many of the age-old controversies surrounding various analyses are resolved immediately. Would such a discovery change your mind about the need to have an analysis of sentences and grammars as abstract objects?
Paul:
No way. Start with a simpler case, needing no talk of hypothetical brain imaging. Look at some representation of a linguistic tree you have on paper, or on a screen. No doubt you will see a white surface and straight black lines. Does that tell us that linguistic trees have white surfaces and straight black lines. Of course not. Linguistic trees are sets based on primitive things called nodes, two relations between nodes, etc. What you see on the paper or a screen is not a linguistic tree but a picture of a linguistic tree, a kind of visual coding of a tree. No doubt any coding of anything has to share certain minimal abstract properties with the thing coded. But there are endless ways to code stuff, so there is no necessary inference from the coding properties to the properties of the things coded. So whatever one could see in brains with the hypothetical imagine devices would not replace linguistics.
Another argument. Apple engineers could tell you exactly what the physical aspects of the coding of Microsoft WORD are in a given MAC. The corresponding aspects might be different in e.g. a DELL or a LENOVO. Neither group of aspects would necessarily tell one anything much about WORD. Just like the smoke or whatever it is skywriters uses doesn’t tell us anything about the sentences whose orthographic representations they place in the sky.
Chris:
Do the answers to the foundational questions of what grammars and sentences are (internal or non-internal), have any concrete implications for what syntacticans actually do on a day-to-day basis? That is, did these issues ever come up in any way when we were working on “Imposters” or ‘Classical NEG Raising’?
Paul:
It is possible but cases would be rare. Probably, relevant issues arise when those who follow the Chomskyan line propose arguments against an analysis based on putative learnability issues, which are or at least used to be not so uncommon. But probably a syntax person could live happily without worrying about the issues, which is why the incoherence of the Chomskyan position has aroused so little opposition based on the fact that it has mostly not even been noted. Then too, after it is pointed out, no one wants to hear it, for sociological reasons.
Chris:
What is the ultimate goal of the field of natural language syntax? This question can be broken down: When we write a syntax paper, what are we trying to achieve? And what should the field of syntax as a whole be trying to achieve?
Paul:
This is not a simple question. Its response requires positions about what kind of things languages are, which is of course controversial. Chomsky for one has claimed that the very notion of language makes no sense. I have spelled out my position on this issue several times, most recently in my ‘Seven Discourses’. So, I do not propose to repeat what I said there.
But briefly, I think the most graspable reality of language is represented by sentences. We gain indirect knowledge of these via development of knowledge of sentences. This we can assume is provided by some kind of knowledge system. How this knowledge develops is rather mysterious as is how knowledge of mathematics and logic develops. These are all things where sensory contact with objects is not possible. But given such knowledge, what one does in linguistic investigation is tap people’s knowledge of sentences. When we talk about intuitions about sentences we are talking about intuiting what the knowledge system says. This provides indirectly knowledge of sentences because of the nature of knowledge. That is, if I know that 2 plus 2 = 4, then it does. Likewise, if I know that All dogs eat meat entails No dog fails to eat meat, then it does.
So once one gathers some knowledge about sentences, hence some information about sentences, one wants to generalize and state properties of all sentences in some assumed language. The system of principles which provide that generalization is called a grammar. The nature of grammars is controversial, but their job is to characterize the structure of every sentence and while doing that to give a precise characterization of the distinction between sentence and non-sentence for fixed languages. If there was only one language L, then the grammar of L would essentially be linguistic theory. But there are, vastly many languages. We assume that calling things languages is sound, that is, that there really is a notion language, which means the different ones have to have things in common making them languages and not something else. Hence talk of linguistic universals.
How do we know where to look for languages? I think we take an anthropological approach here, and take a language to underlie the linguistic communication in a community, that is, idealistically a mono-linguistic community. So, when we contact tribe T, we take the system the speakers in T use to be a language. We might be wrong. For instance, Piraha is so different from so many other languages that, political correctness aside, the idea can arise that maybe it is not a language, at least not a full one. Maybe, it is just a subpart of a language. I don’t know how to resolve this issue, so recognizing the inevitable screaming and vituperation raising the question would cause, let’s not.
So, then the job of syntax is to construct grammars for languages and to search for principles general to all languages hence to all grammars. These principles will characterize what a language is. Suppose for example some version of the coordinate structure constraint is correct. Then a language is a thing which inter alia satisfies that principle, and things which don’t aren’t languages.
This is all quite vague of course and has to be at this level of discourse.
When writes a syntax paper, there can be various goals. One is to focus on some body of data and hope to discover and present principles which characterize it. Many papers will seek to do more and attempt to support new or old general principles of language based on the results of a study of a fixed body of data. For example, one might study some facts in language L and conclude that they support the idea of two types of NPIs, one originating with single NEGs, the other with multiple ones. We have done just that in several papers.
Another type of paper might just focus on one or more principles and cite diverse data from multiple language to support or undermine the relevant principles. It is hard to imagine though work really justifying certain principles of one language without providing some support for more general principles. That follows because any descriptive account must appeal to some general grammatical apparatus and success can’t help but provide at least a tiny bit of support for the worth of that apparatus. One must be cautious here though because often the same descriptive results can be achieved with distinct grammatical tools.
Chris:
OK, this is rather complex, I wanted something more straightforward. Suppose that you had to explain the goal of the field of natural language syntax to a smart undergraduate who had absolutely no background in linguistics or syntax. How would you do that?
Paul:
Syntax is the part of natural language inquiry which studies the principles which determine well-formed sentences in languages, that is, which studies the difference for English between
(1) Frank spoke with his mother on his I-phone.
(2) Spoke Frank his mother with I-phone his on.
Chris:
Let’s move on. I would like to turn to you 2009 article in Biolinguistics. In this article, you bring up the following apparent contradiction (pgs. 251-252):
“Chomsky’s ontological position then incoherently asserts the following:
(8)
a. An NL is a physical state of the language faculty conceived as a biological object, an organ, an aspect of a brain.
b. That organ (brain) state yields an infinite set of expressions.
In the article, you go over various facets of this contradiction, including a useful presentation of the type token distinction for sentences. I think your article did a good job at deconstructing various concepts, and also laying out the issues. I think it is important to the field to have this kind of foundational discussion (as long as it does not involve inflammatory vocabulary). But what was missing from your discussion was the issue of mathematical modelling.
If a scientist or engineer is asked to investigate a physical system, they put together some equations. For example, perhaps the physical system can be described by an equation of the following form f(x,y,x, t) = 0, where x, y and x are spatial coordinates, and t is a temporal coordinate. So, this equation describes the position of an object as a function of time. Perhaps, it describes how a stone falls under the influence of gravity. Or perhaps it models the motion of a planet.
There is no worry about confusing mathematical values of x (e.g., 10.6 meters) with actual locations in the world (a particular physical point of space in our environment). Numbers are abstract objects, and actual locations are physical objects (or parts of physical objects). But we know that it is possible to model locations using numbers.
Furthermore, nothing in the equation prevents the variables x,y,z from taking values that correspond to locations outside of the known universe, but that does not prevent the equation from being an adequate model of the physical system. To incorporate the limits of x into the model we might add the equation: x < 100 meters. So, then our model of the physical system would have two equations: f(x,y,x, t) = 0 and x < 100 meters.
Similarly, in the domain of syntax, I propose that there is a mental system that generates mental representations of sentences (as in Chomsky’s writings). And the way to understand the grammar that we discover and formalize is as a model of that mental system. So, suppose that we write a minimalist grammar G, and that G generates a sentence S (in the standard sense of formal language theory). Then there is a mental system G’ which produces (in some fashion not yet understood) a mental representation S’. Note that I carefully avoid using the formal language theory term “generate” when discussing the relation between G’ and S’. In this example, G models G’ and S models S’. That is in writing G (which generates S), we gain insight into G’ and S’, but there is no way we can confuse G with G’ or S with S’.
Now as you point out there is no way that G’ (the mental system) can produce an unlimited number of S’ (the mental representations). The brain is small, and so the size of the individual S’ is necessarily small. But the formal model itself is free of that limitation. G can generate a sentence S of unbounded length and complexity, as is well known from formal language theory. The only way to prevent G from doing that would be to add an additional equation to the model, something like Length(S) < n (for some constant n), just exactly as we did with the physical system presented above: f(x,y,z,t) = 0.
But here is the crucial point. It may be that G and the equation Length(S) < 0 model two completely different components of the mind. G models G’, responsible for the production and comprehension of natural language, while Length(S) < n models short term memory (which may also extend to other kinds of mental activity).
These comments help us to resolve the contradictory (8). Here are comments on the contradiction from the modelling point of view:
(8)
a.
An NL is a physical state of the language faculty conceived as a biological object, an organ, an aspect of a brain.
Response: Yes, this is true. And it is modeled by G.
b.
That organ (brain) state yields an infinite set of expressions.
Response: This statement is false. But (b) can be replaced by the true statement that follows:
G modelling NL generates an unlimited set of sentences, and any generated sentence S that satisfies Length(S) < n can be realized as a mental representation S’.
In reading your paper (pg. 257), there is the following quote which is very similar to what I have said above:
“In using formal structures, the defense runs, Chomsky is no more committed to claims that NL is formal than a physicist’s use of mathematics involves a commitment to the abstract nature of physical reality. The flaw in this defense is that in fact Chomsky’s appeal to formal science in his characterization of NL and a physicist’s appeal to formal science via e.g. invocation of various complex mathematical equations are entirely distinct. For the physicist never identifies his equations with the physical structures characterized or conversely. But as in (11), Chomsky takes the putative biological entity to be constructing abstract objects, sets. This is entirely distinct from using some piece of formal science to describe something nonformal. And this conflation by Chomsky of formal and physical objects is one of determinants of the incoherence of his ontological position.”
But this is an unsatisfactory reply from my point of view. It is basically not criticizing the physicist analogy (and thus implicitly saying that the physicist analogy could be correct). But then your criticisms of Chomsky’s position may not concern me, since I think about sentences and grammars in terms of the physicist analogy.
Paul:
Yes, it [C.C., a discussion of mathematical modelling on Paul’s part] is missing but I don’t think has any real relevance.
Because inter alia the modelling discussion is so vague and unspecified as to really say very little. Compare the situation with an alien computer we get our hands on. We have no idea as to its physical structure inside or its programming. But it does stuff. That is all we would have to work with. Talk of using that stuff to model the inside physical stuff or the programming is just dreaming. Just so in linguistics. What linguists are doing is not modeling anything in brains or even in brain programming because I don’t think anyone knows much about those and surely most linguists don’t. It approaches play acting to say what e.g. syntacticians do is modeling brain activities. And think of model-theoretic semantics. Are sets of abstract possible worlds supposed to have something to do with modelling brains?
The bottom line, I suggest, is that one can’t model brain properties or mind properties until one has a reasonably strong collection of such things as the basis of the modelling. I don’t see how one can say linguistics has such a collection available, still less that it has had such during the last 65 years of the development of massive working in grammatical investigation. What grammarians study in fact is observed texts and speaker intuitions.
9. Conclusion
Chris:
What do you think of the paper “The Achievements of Generative Syntax” by Roberto D’Alessandro?
Paul:
I had seen it before, and skimmed it then and also today. I would say my general reaction is positive. I think the author has tried to be objective, seems like an unbiased person but is clearly someone whose range of knowledge about approaches to syntax is rather limited. In your terms, it is pretty much a principles and parameters collection. But she has amassed a large set of claims and it will be useful to many.
Some lacunae are kind of shocking; for instance, there is nothing on negation. Klima’s paper is cited but only for c-command (which is odd, since that was invented later). Horn, Ladusaw, Giakannidou, Zwarts, Zeijlstsra, etc. etc. unmentioned. There are many claimed discoveries in the vast literature on negation, for instance, Ladusaw’s claim that negative polarity items need to be under the scope of decreasing operators.
The unaccusative idea is credited to Barbara Partee, but I have never seen a mention of her except in that piece by Pullum. The serious discussion of the origin of this idea is in:
Geoffrey K. Pullum, ‘Citation etiquette beyond thunderdome’, Natural Language and Linguistic Theory 6, 579-588, 1988.
There are many claimed discoveries in the relational grammar literature at least as good as some of those listed in the work in question. These include characterizations of passives, of antipassives, about the relational principles assigning nominals to types under predicate raising, about Visser’s generalization, about the existence of so-called inversion constructions where transitive subjects are mapped to indirect objects, on the nature of pseudo-passives (with arguments that the common renanalysis proposals are untenable, etc. One wonders whether the author has even heard of this literature.
Chris:
Is there any paper or book of yours on syntax that you look back on and think that it was really wrong, or that you regretted having written it?
Paul:
Let us consider first books. I don’t think I regret writing any book. The closest would be the introductory text with SJK [c.c., Samuel J. Keyser]. It wasn’t much of a book. As for all the others, no I don’t regret writing them. I regret that I didn’t do them better and I feel that if I attacked the same topics with what is known now and what I know they would be better. WRT to the APG book, if I had it to do over, I would try to make it much much shorter and more focused. Of course, I would revise the ideas along the lines of Edge-Based, and no doubt the latter could be done better too. As for articles, there are too many for me to remember most. No doubt there are lots of problems in many and certainly the above considerations about books are relevant too. I suspect that most of them have at least some virtue or insight somewhere and that at least I learned something from doing them. Anyway, the bottom line is that I wrote the stuff and it can’t be unwritten.
Chris:
What kind of syntactician are you? What properties characterize your research?
What kind of syntactician are you? What properties characterize your research?
Paul:
I am not sure what the kinds of syntacticians are, or even that there are any. It seems to me easier to talk about types of project. And here I would distinguish those which are concept focused vs. those which are data focused. Most work is no doubt some mixture of these. But for example, the volume ‘Arc Pair Grammar’ by me and David Johnson is overwhelmingly concept focused. There is no body of data which it tried to characterize. In contrast, our paper on Ewe negative polarity items, is data focused. My monograph ‘Masked Inversion in French’ was data focused. It started with a body of facts which seem anomalous and tried to find analyses which reduced the anomalies to regularities.
As a personal preference, I would favor data focused work as the further a project strays from masses of facts, the more likely it is to wander into untested and perhaps untestable speculations. See my LINGBUZZ 004735 paper Chomsky’s One Language/Extraterrestrial Scientist Claims for an example. I also do not like works which skip around from language to language, citing a fragment here and a fragment there to support some idea or claim. I find such a procedure in general quite unconvincing.
On type of syntax project which is concept oriented is exemplified by my monograph ‘On Raising’. I started with a sentence type exemplified by Marvin believes Jane to be a nurse and focused on the question of whether the nominal Jane was an exclusively subordinate clause constituent or whether it was both the latter and main clause constituent. This was in reaction to claims that the former was the case. Grammatical cases like (In spite of her three convictions,) Marvin believed Jane with all of his heart to be a decent person illustrate the sort of evidence which can support the latter possibility. The point is that the adverbial with all of his heart is clearly a main clause modifier, while the nominal whose status is at issue is clearly in the same superficial clause as it is.
Chris:
The distinction between concept focused and data focused seems obscure to me. I would have said ‘On Raising’ was data focused, in that it tried to adduce a large amount of data to argue for a particular analysis.
Paul:
I agree it is rather obscure. But in the case of that work, I did not start with any body of data to be accounted for, but with a conceptual question about constituency. Then I gathered data taken to bear on it. Contrast our Ewe paper, where we started with a bunch of Ewe facts and tried to give an account of them in certain conceptual terms.
Chris:
What kind of tools and methods do you like to use in your research?
Paul:
The basic tool is familiarization with as much literature on whatever topic one is concerned with as is possible. Ideally, one should try to become aware of every previous view of the problems which concern one. In practice, obviously, there are limitations on how much one can do. But if one has an idea/view of some phenomenon, one is obliged to consider existing alternative views. Then too one collects as much data from the literature and other sources as is possible.
The possibility of internet searches has evidently opened up a vast new source of data, but is also not free of obvious dangers. How does one know that a WEB example is not a foreign error, a joke, a grammatical error, etc.
Chris:
What kind of threads have run through the research that you have done?
Paul:
I have no idea and I cannot say there are any. One thing I would say is that it does not hurt to approach past ideas on a given topic with a certain degree of skepticism. Just because a lot of past work, even by prestigious figures, assumes X, does not mean X is the right view. At the same time, one must not cavalierly ignore previous ideas.
Chris:
What are your current research interests? What are you working on now?
Paul:
My research interests are negation. Taking NEG to be the representation of syntactic negation, one interest is in particular defending the existence of syntactic NEG raising, the goal of Collins and Postal (2014). There have been numerous subsequent criticisms of that claim, and at some level I would like to respond to them. I am also interested in defending and elaborating a NEG-based conception of NPIs. This assumes that an NPI like jackshit has a structure including NEG, which has either raised or been deleted (possibly both in some cases).
But I am not sure I am working on anything at the moment. I have a variety of projects, all stalled, perhaps forever.
Chris:
What kind of projects?
Paul:
For instance, one such project involves the word hardly of sentences like He drank hardly any beer. In standard terminology this item licenses the NPI any, which requires taking it to be a negative or at least downward entailing expression. I think it is just semantically a variant of almost, but syntactically it is what Collins and Postal (2014) called a lexical NEG deleter. So the NEG in the example actually originates in the any phrase. A mystery then is the existence of examples like He didn’t drink hardly any beer, which many standard English speakers accept. The question is why, if the analysis I proposed is right, can this have the same meaning as the first example, despite the extra NEG. This mystery deepens when one notices that the passive seemingly corresponding to the second sentence, does not have the same meaning: that is, Hardly any beer was drunk by him is not equivalent to Hardly any beer was not drunk by him. I have proposals about these issues but various complications have prevented constructing a finished product.
10. References Mentioned in Interview
Chomsky, Noam. 1957. Syntactic Structures. Mouton de Gruyter, Berlin.
Chomsky, Noam. 1965. Aspects of the Theory of Syntax. MIT Press, Cambridge, Massachusetts.
Chomsky, Noam. 1973. Conditions on Transformations. In Paul Kiparsky and Stanley Peters (eds.), Festschrift for Morris Halle. Mouton, The Hague.
Collins, Chris, Simanique Moody and Paul Postal. 2008. An AAE Camouflage Construction. Language 84, 29-68.
Collins, Chris and Paul Postal. 2012. Imposters. MIT Press, Cambridge, Massachusetts.
Collins, Chris and Paul Postal. 2014. Classical NEG Raising. MIT Press, Cambridge, Massachusetts.
Grinder, John and Paul Postal. 1971. Missing Antecedents. Linguistic Inquiry 2, 269-312.
Harris, Randy Allen. 1993. The Linguistic Wars. Oxford University Press, Oxford.
Huck, Geoffrey J. and John A. Goldsmith. 1995. Ideology and Linguistic Theory. Routledge, New York.
Johnson, David and Paul Postal. 1980. Arc Pair Grammar. Princeton University Press, Princeton, New Jersey.
Katz, Jerrold and Paul Postal. 1964. An Integrated Theory of Linguistic Descriptions. The MIT Press, Cambridge, Massachusetts.
Kayne, Richard. 1969. The Transformational Cycle in French Syntax. Doctoral Dissertation, MIT.
Langendoen, Terence and Paul Postal. 1984. The Vastness of Natural Languages. Basil Blackwell, Oxford, England.
Lees, Robert B. 1957. Review of Syntactic Structures by Noam Chomsky. Language 33, 357-408.
Levine, Robert. 2018. ‘Biolinguistics’: Some Foundational Problems. In Behme, Christina and Martin Neef (eds.), Essays on Linguistic Realism, 21-59. John Benjamins, Philadelphia.
Nadahalli, Jayashree. 1998. Aspects of Kannada Grammar. Doctoral dissertation, NYU.
Postal, Paul. 1964. Constituent Structure, A Study of Contemporary Models of Syntactic Description, Publication 30, Indiana University Publications in Anthropology, Folklore, and Linguistics, Bloomington, Indiana.
Postal, Paul. 1967. Aspects of Phonological Theory. Harper and Row, New York, New York.
Postal, Paul. 1970. Cross-Over Phenomena. Holt, Rinehart and Winston, New York, New York.
Postal, Paul. 1974. On Raising. The MIT Press, Cambridge, Massachusetts.
Postal, Paul. 1979. Some Syntactic Rules in Mohawk. Garland Publishing Inc., New York, New York. [Postal, Paul. 1962. Some Syntactic Rules in Mohawk. Doctoral dissertation, Yale University.]
Postal, Paul. 1986. Studies of Passive Clauses. State University of New York Press, Albany, New York.
Postal, Paul. 1998. Three Investigations of Extraction. MIT Press, Cambridge, Massachusetts.
Postal, Paul. 2004. Skeptical Linguistic Essays. Oxford University Press, New York.
Postal, Paul. 2011. Edge-Based Clausal Syntax. The MIT Press, Cambridge, Massachusetts.
Postal, Paul. 2009. The Incoherence of Chomsky’s Biolinguistic Ontology. Biolinguistics 3.1, 247-266.
Postal, Paul. Books. Linguistic Frontiers 3.1, 1-7.
Postal, Paul. 2018. Seven Discourses on the Ontology of Natural Languages. Ms., NYU.
Pullum, Geoffrey. 1988. Citation Etiquette beyond Thunderdome. Natural Language and Linguistic Theory 6.4, 579-588.
Rosenbaum, Peter S. 1967. The Grammar of English Predicate Complement Constructions. MIT Press, Cambridge, MA.
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.