The loss of a Colleague, Rochester Graduate and Dear Friend: Ivan Andrew Sag (1949-2013)
October 14, 2013
Obituatry: Ivan Andrew Sag (1949-2013)
Editor for this issue: Sarah Fox <email@example.com>
Ivan A. Sag, Professor of Linguistics and of Symbolic Systems and Sadie Dernham Patek Professor in Humanities at Stanford University, died on September 10 after a long and courageous struggle against cancer. He was 63. He is survived by his wife of 28 years, the Stanford sociolinguist Professor Penelope Eckert, and by a niece and nephew in Budapest, Annemarie and Zoltan Sag. The LSA has set up an Ivan Sag Linguistic Institute Fund to commemorate his life (see http://www.linguisticsociety.org/news/2013/09/13/new-fund-established-honor-ivan-sag). Donations (http://www.linguisticsociety.org/donate) will support the Linguistic Society of America's biennial Linguistic Institute, including student fellowships and the Sapir Professorship.
Ivan Sag was a force of nature. His work in many subareas of linguistics was spread over more than four decades and was respected around the world by the thousands of linguists he taught or befriended or influenced. He was honored by election as a Fellow of the American Academy of Arts and Sciences in 2007 and was named a Fellow of the Linguistic Society of America in 2008. Earlier in his career he had held a Mellon Fellowship (Stanford, 1978-79), an Ameritech Fellowship (Chicago, 1987-88), a Logica fellowship (at the Dutch Research School in Logic, Utrecht, 1994), and a fellowship at the Center for Advanced Study in the Behavioral Sciences (2002-03). He was named professor honoris causa at the University of Bucharest, Romania, in 2001.
He was born in 1949 in Alliance, Ohio. He loved pointing out that it is a town with the unusual feature that Main Street dead-ends at both ends. His high school years were spent at the highly selective Mercersburg Academy, but he was expelled late in his senior year for being the ringleader of a small drinking expedition during a French Club trip to Washington DC to see a play.
Many years later an institutional change of heart brought Ivan back to Mercersburg to deliver the 2008 Cum Laude address, and characteristically, instead of avoiding the expulsion incident he told the whole story in his inimitable way (the text of his priceless speech is fortunately preserved at http://lingo.stanford.edu/sag/mburg-speech.pdf).
Sag had been offered early admission to the University of Pennsylvania in his senior year, but that deal was taken off the table when Penn learned of his expulsion. So he went to the University of Rochester to earn his undergraduate degree. There he was very happy, and first became interested in linguistics, particularly Sanskrit and Indo-European studies. He also found time to become involved in rock and blues music, even dropping out for a while to work as a musician and as a part-time concert promoter.
His BA from Rochester was awarded in 1971, and after that he finally did go to Penn, earning a Master's degree in linguistics in 1973, and completing all of the coursework for a PhD in historical linguistics. (One class that he took, a sociolinguistics course taught by William Labov, was much more significant than he knew at the time. One of his fellow students in the class was a young woman named Penny Eckert. Neither of them realized then how important they would one day be to each other.)
Eventually Sag's growing interest in syntax led him to MIT to earn the PhD under Noam Chomsky's supervision. He did publish a couple of papers on Grassman's Law in Sanskrit, but did not pursue historical linguistics any further. His PhD dissertation was in syntax and semantics. It dealt with ellipsis, specifically verb phrase ellipsis, which he was responsible for renaming Post-Auxiliary Ellipsis, on the grounds that the elided material is often not a VP under anyone's analysis, but is always immediately preceded by an auxiliary. Using the lambda calculus to represent VP meanings, and exploiting the relation of alphabetical variance between variables, he was able to account for many of the otherwise odd departures of post-auxiliary ellipsis from involving strict syntactic identity.
While at MIT he formed an alliance with Jorge Hankamer (then at Harvard) that led to a truly classic joint paper, 'Deep and surface anaphora', (Linguistic Inquiry, 1976), and a sequel 'Towards a theory of anaphoric processing' (Linguistics and Philosophy, 1984).
After earning his PhD, Sag was hired at Penn as an assistant professor, but within two years he took a year's leave to go to Stanford University as a Mellon Fellow for the year 1978-79. And after sampling life in California he never returned to the East Coast. He was appointed assistant professor of linguistics at Stanford, and earned tenure there, and it remained his academic home for the rest of his life.
Sag's substantial and widely varied research output increased in quantity and breadth of interest as his career went on. His publications were often co-authored, partly because his gregarious nature combined with his love of organizing things and his prodigious output of interesting ideas made it almost a necessity to work with others. He raised grants, organized workshops and conferences on the Stanford campus, arranged visits and summer salaries, traveled and lectured, and inspired hundreds of students and younger faculty around the world in syntax, semantics, computational linguistics, and other subfields.
One service to the profession for which he became world famous was setting up cooperative living arrangements for communities of graduate students and faculty at the LSA's Linguistic Institutes, which at one time were held every year. As soon as it was known which campus would host the next Linguistic Institute, Sag would start making arrangements to rent an entire sorority house, or often two side by side. (Nearly always sororities; fraternity houses tended to be more trashed by the end of an academic year, but sororities kept their properties in better shape. He knew things like that.) Usually he would hire cooks as well, to provide communal meals for everyone in the evening.
The intensity of the resultant scholarly interactions was extraordinary. Scores of linguists, many of them handpicked invitees from among Sag's huge array of friends and acquaintances, would live together, walk to classes together, play volleyball together, eat dinner together, and discuss linguistics at all hours of the day and night. The countless linguistic collaborations and lifelong friendships started during those summers had lasting effects on the discipline. The history of American linguistics from the 1970s on simply would not have been the same without the organizing genius of Ivan Sag, and of course his efforts at promoting participation in Linguistic Institutes were a major boost to their financial as well as academic success. It did not surprise his hundreds of friends when in 2005 the LSA recognized his contributions to the field with the Victoria Fromkin Lifetime Service Award.
One of the Institutes at which Sag set up living arrangements was at the University of Illinois at Urbana-Champaign in 1978. It was there that he first encountered Gerald Gazdar, who was there primarily to participate in a seminar on pragmatics organized by Georgia Green, but in his spare time was working on phrase structure schemata for coordination and their interaction with other syntactic phenomena. Later Gazdar produced two famous samizdat papers, circulated in January and April 1979, and just a month after the second of these, at a conference at Brown University on alternative conceptions of syntax, organized by Polly Jacobson (who had also been at Urbana-Champaign the summer before), two of the papers presented were almost entirely devoted to developing Gazdar's ideas, then only a few weeks old. Sag's paper was one of them, and it established him as an important founding member of the GPSG community.
Both Sag and Gazdar presented at a conference in Groningen in the summer of 1980, together with Ewan Klein and Geoff Pullum, and soon an intense set of interlocking collaborations was under way. Sag's paper from the Brown conference appeared in "The Nature of Syntactic Representation" (Jacobson and Pullum, eds., D. Reidel, 1982), along with Gazdar's long expository paper 'Phrase structure grammar.' Gazdar, Pullum and Sag published a GPSG analysis of the English auxiliary system (in Language, 1982); Gazdar, Pullum, Sag, and Tom Wasow published a joint response to Edwin Williams' critique of Gazdar (Linguistic Inquiry, 1982); Sag published a metarule analysis of parasitic gaps (Linguistics and Philosophy, 1982); Klein and Sag published a proposal to reduce redundancy of by letting semantic types rather than stipulations in semantic rules drive the translation into logical form (Linguistics and Philosophy, 1983). Other papers appeared at a rapid rate. There were transatlantic visits in both directions.
Sag was an enormously valuable collaborator. He always understood ideas that were brought to him, and often developed and improved them, and would then promote them vigorously. But most of all, Sag was an idea generator. Although he loved working with others, he was never just a passenger or a minor team member. All of the linguists who worked with him or came into contact with him benefited from his brilliant grasp of data and formalism and argumentation, his theoretical imagination and problem-solving abilities, and his enthusiasm and energy.
He delighted in working out the details for the best ideas, but was quite happy to abandon an approach that didn't work or didn't find favor with his collaborators. He was never afraid to ferret out data that would refute his own pet ideas. He never exhibited the defensive moves of the insecure scientist who cannot bear to be wrong and would rather invent any number of sheltering assumptions rather than lose face. If Sag suspected some idea was not exactly right he would drag it out into the sunlight of the full range of data, and pin it down and try to identify what was wrong. He would propose complete reworkings of whole papers or theories or projects without fear or hesitation. His goal was never to defend his pride or preserve his consistency; it was to actually find out how language worked. That combination of infectious glee with intellectual integrity was his trademark, and made him the spectacular innovator, originator, collaborator, and catalyst that he always was.
Early in the 1980s, at the instigation of Anne Paulson, who had heard some lectures on GPSG at Stanford, a project began at Hewlett-Packard Laboratories in Palo Alto aimed at developing a natural language question-answering system using an implementation of GPSG syntax. Sag and several other GPSG linguists became paid consultants on the project. It was the beginning of Sag's long involvement with computational linguistics.
Around 1983, Sag became one of the core members of the group that founded the Center for the Study of Language and Information (CSLI) at Stanford. This brought an influx of new money (a $20 million gift from the System Development Foundation) and new interdisciplinary ideas, as philosophers like Jon Barwise and John Perry began to interact seriously with linguists and psychologists to build a world-class center for cognitive science.
In 1984 Sag's PhD student Carl Pollard completed a dissertation that began with GPSG ideas but developed from there in a way that was informed by categorial grammar. Pollard generalized the GPSG idea of context-free grammars in a particularly interesting way, leading to a novel class of grammars that were later proved to be weakly equivalent to the tree adjoining grammars and the linear indexed grammars and the combinatory categorial grammars. Typically, Sag did not resist the new thinking that Pollard's work opened up, but began an intensive collaboration with Pollard which ultimately led to HPSG.
The book "Generalized Phrase Structure Grammar" by Gazdar, Klein, Pullum, and Sag appeared in 1985, and that was a watershed. But something else happened around that time that was a more important watershed in the long run: Penny Eckert arrived to teach sociolinguistics at Stanford. She and Sag embarked on a relationship that became the central one in both their lives. Within a year they had married.
The project at Hewlett-Packard was continuing, and needed a name, so the members had constructed one based on a new parser design that involved identifying heads of phrases first and then parsing their dependents: in a certain sense it was head-driven phrase structure grammar, hence HPSG. The fact that the first two letters were H and P pleased the Hewlett-Packard managers. The name stuck, and began to be attached to not just the project but also the framework Pollard and Sag were developing. They did not use the name in the title of their first book ("Information-based Syntax and Semantics, Volume 1: Fundamentals", CSLI, 1987), but there never was a Volume 2; the sequel that arrived seven years later was very different, and bore the title "Head-Driven Phrase Structure Grammar" (CSLI, 1994). It greatly extended the coverage of the earlier volume, including treatments of long-distance dependencies and binding, while also incorporating many refinements to the theory.
HPSG's evolution out of GPSG was motivated by both linguistic and computational concerns. Localizing a great deal of linguistic information in the lexicon and organizing the lexicon as an inheritance hierarchy facilitated both capturing linguistic generalizations and efficient computational implementation. The framework became enormously important and influential. European computational linguists were particularly taken with it, because it was readily implementable yet flexible and powerful. There have been two decades of conferences on HPSG around the world, most frequently in Europe but also in the USA and in Asia.
Work on implementing HPSG grammars for a variety of languages (including English, German, Spanish, and Japanese) has proceeded for many years, and led to actual practical applications. For example, the LinGO system developed at CSLI by a team including Sag and Dan Flickinger from 1993 on is being used as a tool for teaching writing in Memphis public schools.
During the HPSG period Sag worked on many other syntactic and semantic issues. In collaboration with Janet Fodor, he explored the evidence for traces and found it wanting. He worked with Geoff Nunberg and Tom Wasow to re-examine the syntactic arguments based on idioms, and the three of them published a paper in Language arguing that, contrary to standard assumptions in the generative literature, the meanings of idioms could be decomposed to varying degrees. He collaborated with Anne Abeillé, Danièle Godard, and Philip Miller on papers about the syntax of French.
Much of Sag's later work within the HPSG framework was concerned with the syntax and semantics of long-distance dependencies in English. He published a detailed analysis of relative clause constructions in 1997, a book on interrogative constructions with Jonathan Ginzburg in 2000, and a detailed taxonomy of filler-gap constructions in 2010. In the last few years, in conjunction with Philip Hofmeister and other former students, he explored the hypothesis that many so-called island phenomena could be explained on the basis of extra-grammatical factors, particularly independently motivated processing considerations. In the course of this work, he started doing experimental studies, which led to his involvement in a vigorous methodological debate in the pages of Language.
An important innovation coming out of the implementations of HPSG was Minimal Recursion Semantics, described in a 2005 paper that Sag co-authored with Ann Copestake, Dan Flickinger, and Carl Pollard. And this is just a sampling: his output was as prodigious as his breadth of interests.
Sag taught syntax with legendary enthusiasm to graduate and undergraduate students throughout his time at Stanford. He won a teaching award early in his career, and built a reputation as an inspiring teacher and mentor. With Tom Wasow (and later, for the second edition, Emily Bender) he produced an introductory syntax textbook: "Syntactic Theory: A Formal Introduction."
Sag also served on the committee at Stanford that created an entirely new undergraduate major: Symbolic Systems. It was a course to prepare students for the 21st-century cognitive, linguistic, and computational sciences, combining linguistics, logic, computer science, and psychology, and it has been a remarkable success. He served as its director during 2000-01 and 2005-09.
Forever evolving intellectually, in the last decade of his life Sag found that the degree of detail necessary in an empirically adequate formal treatment of virtually any area of grammar, for example so-called "wh-movement" phenomena, required incorporating into an explicit grammar something like the traditional notion of grammatical construction, along with the mechanism of multiple inheritance. This led him to begin drawing closer to work on the Berkeley variety of construction grammar as developed by Charles Fillmore, Paul Kay, and others. The resulting theory, known as Sign-Based Construction Grammar (SBCG), extended HPSG's idea of using an inheritance hierarchy for the lexicon to constructions (this idea can actually be seen in Sag's work as early as his paper on relative clauses in 1997).
SBCG replaces a lexicon plus phrase structure rules with two kinds of constructions: lexical constructions, which are partial descriptions of single signs, and combinatorial constructions, which are partial descriptions of feature structures containing a mother sign and a list of daughter signs, which, like lexical signs, are associations between sound and meaning. SBCG's constraint-based architecture and surface-oriented and localist grammatical analysis make it suitable for modeling both production and comprehension, and for use in a variety of types of computational implementations. A collection of SBCG papers that Sag co-edited with Hans Boas, including Sag's 133-page "informal synopsis" of the theory, appeared in 2012.
It might seem implausible that a man whose activities spanned copious research, enthusiastic teaching, industrial R&D, program directorship, grant administration, LSA service, and Linguistic Institute accommodation management could find the time for yet another activity, but Sag was also the founder and leader of a rock 'n' roll band, Dead Tongues. Its personnel, shifting continuously over the years, were nearly always linguistics faculty or students, sometimes augmented by other friends from nearby universities, research institutes, or industry labs. For a time it was a large band with a brass section; more recently it was a smaller combo with several excellent vocalists; always it had Sag on keyboards at (and serving as) the heart of it. The repertoire was an eclectic mix of rock and blues classics.
At the end of April 2013, Dead Tongues played a reunion concert. Three ex-Stanford linguistics students and a Berkeley postdoc supplied excellent vocals; Dan Jurafsky was on drums; Peter Sells (now at the University of York) was back in his role as bassist and technical manager; and three returning former guitarists took the stage together: Steve Wechsler, John Beavers, and Geoff Pullum.
The concert took place at the end of the first day of a workshop on Structure and Evidence in Linguistics, honoring Sag's lifetime of work in linguistics and colloquially dubbed the IvanFest. A book of studies inspired by his work ("The Core and the Periphery", CSLI, 2013) was presented to him. Though severely weakened and disfigured by the operations and treatment for the cancer he had endured for some years, Sag mustered all of his legendary energy: he rehearsed with the band's new lineup in the morning, listened closely to every paper presented in the afternoon, asked several penetrating questions, and then played piano with the band right through the evening. The event website (http://www.stanford.edu/dept/linguistics/structures_evidence/website/index.html) includes links to the Dead Tongues performance as well as videos of all the talks.
At the end of the three days of the IvanFest, Sag made a speech of thanks. It was read aloud for him because of the severe articulatory difficulties he was experiencing as a result of radical jaw surgery. His extraordinarily generous and self-effacing remarks contained nothing about him or his work or his achievements or his suffering; it was all about us, the many linguists and other colleagues who had come to Stanford to attend the event, and about his gratitude and appreciation for those who had presented and those who had organized the event.
He died at home with Penny less than six months later. He had told her that he wanted no memorial service or other platform for the delivery of embarrassingly maudlin speeches about him. Many people wanted there to be such a service, but he was very satisfied with the IvanFest as his farewell event. It had brought together just about everything he loved most: Penny, and linguistics, and rock 'n' roll, and socializing, and a room full of scores of his friends and colleagues dancing to the music of his own band right there in the CSLI conference room. It's hard to imagine anything that could have delighted him more. Rock on, Ivan.