Saturday, May 5, 2018

Morris, im memoriam

Morris Halle, my PhD co-supervisor, died on April 2. On April 26, I began a paper on long-distance allomorphy. It has quickly bloomed from its intended 500 words into something much longer and, as I craft the argument, I find myself in imaginary conversation with Morris about it. (“An appendix? Of course, you’ll need an appendix. It’s complex data. You have to lay it all out.”) With Morris’s memorial service today, and with me on the wrong side of the Atlantic, I want to say farewell to my teacher by summarising five lessons I learned from him. They map out our time together, from our first meeting at MIT (in late 1998 or early 1999) through my graduation (in June 2003) and into my life as a professional linguist.

1. People don’t come with instruction manuals

When I first met Morris one-to-one, I thought he was obstreperous and cantankerous, and he thought I was a fool.

Optimality Theory was all the rage, upending the processual phonology that Morris had pioneered (as in The Sound Pattern of Russian and, with Noam Chomsky, The Sound Pattern of English) and replacing it with output-oriented constraints. For the first time ever, my year, Ling ’98, was introduced to phonology via OT, ignoring completely the work that had gone before. The result was not positive and, amusingly, when asked at the start of our second year whether we wanted to write our first generals (pre-PhD qualifying) paper on phonology, which we had had a year of, or on morphology, which none of us had touched, the majority went for morphology.

My response to phonology was to work with Christian List on a formalisation of the concepts underlying Optimality Theory using techniques from mathematical decision theory. Along the way, I noticed a property of Classical Hebrew monosyllabic nouns, the best analysis of which involved a generalization over stored forms. Contrary to OT, this was not output oriented and I thought it would be of natural interest to Morris. Morris was not impressed.

“First, you do the analysis,” he grumbled. “After, you get to argue about the theory.” In other words, I was a kid who had turned up to an adults’ game and I was being sent home again to do my schoolwork. This made no sense to me, coming from a background in Oxford philosophy.

Our second meeting did not go better. Writing a generals paper on morphology still meant that I fell squarely within Morris’s sphere of activity. The problem I was working on—apparent superfusion in Kiowa-Tanoan agreement—had been given to me by Ken Hale. “I haven’t been able to figure it out in 30 years,” he had said blithely. “Why don’t you give it a go?” A sensible person would not rush in where Ken had feared to tread. But, as Morris had already established, I was not sensible.

My point of departure was Rolf Noyer’s dissertation. I used its well grounded feature theory and its partial application to Kiowa-Tanoan to begin to chip away at the agreement system. It was slow going. The system, it would turn out, is not a purely morphological problem, but is in full cahoots with syntax, semantics, and phonology. Wayne O’Neil, my committee chair, said I should go to see Morris about it. He had, after all, supervised Noyer.

But Morris spent so much time nitpicking the set up of the problem—a quibble about a Haitian Creole paradigm particularly sticks in my mind—that I don’t remember discussing the problem itself at all. I left the meeting thinking it had been wholly pointless. So, I learned from Wayne, did Morris. “When he does his introductory lecture for Ling ’99,” he told me the next day, “he’s going to switch from phonology to morphology.” I resolved to leave Morris to his cantankerous, obstreperous self.

A year or so later, though, I met with him again, and this time, there was one crucial difference. In that gloriously MIT way, I was writing a term paper on Kiowa for a course on Japanese (the tone systems functioned similarly and so Shigeru Miyagawa signed off). Figuring out Kiowa tone fed my battle of attrition with the agreement system. The only problem was, I was a refugee from first year phonology and had no idea where to start an analysis. When I went to see Morris, therefore, I went with data and questions, not theories and answers. The reception was completely different. He suggested a framework, Metrical Grid Theory. I slotted in the data. Progress was made. Our relationship thickened, until one day, Alec Marantz, who was also on my generals committee, pointed out that it was the kind that PhDs are built on. He and Morris became my joint supervisors and, eventually, my Doktorväter—so that, academically speaking, I come from a two-dad family.

If Morris had come with an instruction manual, we might have gotten on sooner. And that would have been to my detriment. Because, nowadays, when I see young linguists, their hands unsullied by data, waxing all philosophy-of-science on analogies from biology and physics, I roll my eyes. And when I see morphologists cherry pick homophony relations to suit an analysis, I want to pick the nits of out them. In short, I’m a regular Morris minor. The cantankerous old professor won that particular battle.

2. Papers don’t come with instruction manuals either

Morris’s great contribution to morphological theory is Distributed Morphology, which he developed Alec Marantz. DM is committed to three principles (syntax builds, insertion tarries, exponents compete). But one of the criticisms of DM that I have most often heard is the amount of machinery it uses. This criticism originates in Halle and Marantz’s paper. Alongside their three principles, they adopt the (preexisting) morphological operations of fission, fusion, impoverishment, and merger. Four years later, Morris wrote a paper specifically about two of them, and others have followed suit. The impression results that DM needs a hefty box of tricks. But the impression is a major misreading of how Morris viewed his responsibilities as an author.

At some point, I read Morris and Bert Vaux’s analysis of Latin and Armenian case. I was impressed by its handling of data, and by the formal system they developed, but the atoms of their analysis, case features, just made no sense to me. They didn’t have meanings out of which “dative” and “ablative” could be built. Instead, they were self-referential things that told you where in syntactic structure they were—which was wholly superfluous, as you could read that information off the syntax itself.

To my surprise, Morris’s view was the same. His favourite part of the paper was a Latin table that had impressed me: it did in under half a page what traditional grammars spread over dozens. The rest of the paper was, to his mind, an argument that just three features underlie the Latin and Armenian case systems. It was left as a challenge to the right readers to figure out what those features really were. Theirs were placeholders, not proposals.

I was genuinely shocked. Nothing in its presentation had led me think that some parts of the paper were meant more seriously than others. Why would Morris fill his papers with stuff he didn’t believe?

The most basic answer is: because ideas are cheap. Anyone can have a good idea. Go to a pub on a Friday night and there’ll be some guy ready to talk your ear off about his world-changing insights. Has he done anything about them? Probably not. Because implementation is the hard part. Working out details is slow and lonely and boring. (Morris slept only four or so hours a night. After that, he would sit in the kitchen with paper and pencil, working things out.) That’s why everyone I know has more notes than drafts and more drafts than publications. Ideas are like relationships: they fizz at the beginning, but the ones that make it take work. Only implementations, not ideas, were publishable, in Morris’s book.

But there’s a deeper answer too, and it reveals Morris’s true character. Nobody wants a publication to go down in flames because of some technical inconsistency or underived datum. The desire to be right can lead to a kind of dishonesty, the temptation to cover up weaknesses in our analyses, not so as to hide them completely, but so as to delay their discovery and not let them distract from the main point. (Again, ideas are like relationships: we don’t flaunt our faults on first dates.) Morris’s insistence on explicit implementation meant that there was nowhere to hide. It takes bravery, both intellectual and personal, to do that. Honesty hurts, but Morris never flinched.

3. Do wrong right

Detail is revealing. What leads you to think of continental drift? Explicitness does: you draw a map of Africa and South America and you see how one fits into the other. If you really believe in your theory, you want to pursue its full deductive closure, to see that it really maps the lie of land.

After Morris and I had been working on Kiowa tone for a few weeks, I turned up at his door with a worry about words that resist high tone. Out came Morris’s pencil, his clipboard of scrap paper, and his delight at showing someone how the mechanics worked. Off he went, chugging through the formalism, until the output popped onto the page. It was wrong. He checked the result. It really was wrong. Metrical Grid Theory had been tested in rough typological terrain, but it hadn’t encountered Kiowa. The parameter settings that delivered the vast bulk of Kiowa tone sequences entailed that the device for delivering strings of low tones anywhere else in the word was completely inert when word initial. His pet theory had failed him. And he was delighted! “You see, when you write things down, that’s when you learn things.”

Delight at insurrection must be one of the defining experiences of being Morris’s student. I managed it twice. (Kiowa tone doesn’t really count, as I had a fix, not a solution.) The first time was when I convinced him, during my PhD research, that inverse number in Kiowa involved different values of the same feature. Though he had argued for bivalence in phonology almost half a century earlier, he had doubted that morphosyntactic features would be similarly susceptible. Because he liked the result so much, I focused on it when I contributed to his (fourth!) festschrift.

The second time—and the more important one, because it proves that DM machinery was not central to Morris’s thinking—occurred after my graduation, when I entered an old stomping ground of Morris’s: split agreement in Semitic verbs. Morris’s own analysis, published in a 1997 paper (very much in line with Noyer), relied on two morphological operations, fission and impoverishment, and needed each morpheme to be stipulated as a prefix or suffix. I countered that the effects of fission and affixal position followed from general properties of the mapping from syntax, and that the need for impoverishment was based on misanalysis of the data. I’d never known him happier than in that moment. I have that image of him, frozen in time, in my mind’s eye as I write this now.

4. The perfect example

The first year of an MIT PhD involves a survey course, called Survey. It introduces students to faculty members who aren’t teaching the introductory courses. Morris, though already retired, was one of the invitees and I still remember his talk in surprising detail, down to individual examples he used. What gripped me was the way the whole presentation unfurled from a single, perfectly chosen example (two verb forms in Russian). Every subsequent fact refined the initial hypothesis and pushed inexorably towards the next. It was like watching a professional pool player. You think you see them prepare to sink the yellow ball, but when you shift focus, you see the same shot has set things up perfectly for the blue ball.

Morris thought readers should sweat over papers. Major papers need multiple readings. Understanding takes effort. But he didn’t think the responsibility lay only with the reader. Authors had to be clear, they had to be explicit. His approach to examples distilled this to its essence. The perfect fact, free from distractors, could encapsulate a problem, set up the solution, and trigger the domino chain of reasoning between them.

5. Curiosity

Morris taught many lessons. Be clear. Be explicit. Be honest. … Build an edifice. Be a genius. (Some of them are easier than others.) For me, the core is to be true to your own curiosity. This lesson took many forms. His job advice to me: “Figure out the job you want and then persuade someone to give it to you”! But I found it most apparent in his analysis of Psalm 23, “The Lord is my shepherd”. It touches me to recall this today, as he read Psalm 23 at the memorial service to Ken Hale, to whom I also owe so much.

Biblical poetics revolves around syllable counts. It’s not overly sophisticated metrically, but at some point, Morris must have slotted the psalm into Metrical Grid notation. (I can still see his thin-pencil x marks on scrap paper.) Maybe it was in his kitchen before dawn and he got up at some point, leaving his notes on the table. Returning, he might have seen them rotated through 90 degrees. And that was the eureka moment. Suddenly, he saw not a grid, but a building. Searching further, he found depictions on coins and in artwork of the Second Temple in Jerusalem, built after the Babylonian exile and of which the Western Wall is the only remnant. They corresponded quite exactly to his grid.

It is hard to overvalue this discovery. The 23rd is amongst the most widely known of the psalms. It is recited weekly in the Jewish liturgy. But graphic poetry is Hellenic. The rabbinic tradition would have been blind to it. The Greeks with their Septuagint would have been deaf to it. It took Morris to cut through millennia and connect at a new and personal level with the artist who wrote it. This was Morris at his most iridescent, pursuing his curiosity and unearthing riches.

The family, friends, colleagues, and students who gather to remember Morris today are testament to the depth and beauty of the connections he forged. Long may they live after him.