I first published this in The Valve on December 7, 2009. This piece tells the story of how, in the early 1970s, I came to abandon "traditional" literary criticism and set sail for cognitive science. I'm reposting it here and how because it is also about description, for the core of my work on "Kubla Khan" WAS descriptive, not theoretical as I'd set out to do. I've also published post as part of a working paper, Lévi-Strauss and Myth: Some Informal Notes (PDF).
* * * * *
Previous posts in the series (links to The Valve):The King’s Wayward Eye: For Claude Lévi-StraussLévi-Strauss 2: Subject and ObjectLévi-Strauss 3: What’s the Subject?* * * * *
My recent posts about Claude Lévi-Strauss are about his ideas on myth, but they are also about the influence of those ideas on me. They are stories in my intellectual autobiography. I would like to continue with that autobiography and say more about how, for better or worse, I moved through Lévi-Strauss to cognitive science.
That could not have happened unless I had become interested in the cognitive sciences even as I was investigating Lévi-Strauss. For Lévi-Strauss & Co. was not all that caught my undergraduate interest when I was at Johns Hopkins in the 60s. I studied psycholinguistics under James Deese and thereby discovered Noam Chomsky’s transformational generative grammar (& how it wiped behaviorism off the map). I learned about Jean Piaget under Mary Ainsworth, who also introduced me to John Bowlby and to primate ethology. Other thinkers were important as well, including Nietzsche, Merleau-Ponty, the early Wittgenstein, and, in a way, Philippe Ariès, though that influence is not so germane to this particular story.
What brought it all together, or, in a way, forced me to tear it apart, was a poem, Coleridge’s “Kubla Khan” (text available here). I became interested in “Kubla Khan” during my senior year at Johns Hopkins in a course taught by the late Earl Wasserman. As you know, Coleridge had declared “Kubla Khan” incomplete: It came to him in an opium-induced reverie (opium was the aspirin of the time); he was interrupted by a man from Porlock; the reverie was dissipated and, with it, most of the poem vanished. The text was oh! so incomplete. I sensed that the two sections of the poem—the first, about Kubla and his pleasure-dome, the second, about a poet and his vision of the damsel with a dulcimer—had the same structure. I also believed that “Kubla Khan” became complete by asserting its own incompleteness: “Could I revive within me . . .” But he couldn’t, and so this incomplete poem asserted its own fragmentary nature, from within itself, thereby recursively attaining closure. It was a trick of the times.
I wrote that paper and, while I thought it was a good one, it didn't come close to exhausting my interest in the poem. I decided to continue that work into a Master's Thesis at the Humanities Center, with Dick Macksey as my director. (Of course I addressed him as “Dr. Macksey” at the time.) This was a few years after Macksey and Eugenio Donato had organized the 1966 structuralism conference that is generally given as the starting point of the intellectual shake-up that has since devolved into Theory. Though I was on campus at the that time I didn’t attend any of the sessions. But I took several courses with Macksey and an independent study as well.
And so, in due course, I undertook a binary inquiry into “Kubla Khan.” Getting started was easy, too easy, for the poem bristled with oppositions—dome vs. caves, light vs. sound, sight vs. breath, Kubla vs. the woman wailing, the woman wailing vs. the damsel with a dulcimer, order vs. chaos, time vs. space, impotence vs. creativity, and others. But I could see no way of making sense out of the plethora of such oppositions; they refused to order themselves into a continuum—from maximum opposition through intermediary steps to minimal opposition—such as Levi-Strauss had shown in his famous discussion of Oedipus or his later treatment of the story of Asdiwal. Nor was any other pattern discernible. “Kubla Khan” appeared to be a chaos of binary order (see Figure 1).
Figure 1: A fragment one of the worksheets I used in exploring “Kubla Khan.”
I decided that I would have to create a theory suited to this poem. Levi-Strauss, yes, and Roman Jakobson, but also Merleau-Ponty, Wittgenstein, Piaget, Nietzsche and others. After seven or eight months of work in which I reviewed the extant literature on the poem and slogged away on that theory I began to feel I was embarked on an endless task. Yes, I was making progress, things would fall into place here and there. But I was well over a hundred double-spaced pages into a draft with no sense of closure in sight. It wasn't the sort of experience where you get through your basic argument, crudely, with loose ends sticking out all over, and then go back to tie them off and find out it takes more thought than you had anticipated. Rather, I simply had no sense of an exposition and argument that was going anywhere but in circles. I would find myself making a certain argument, call it Shangri-La, conclude it, and move on to another line of thought and, ten or fifteen pages later, I was back at Shangri-La, working through it, moving on . . . and right back to Shangri-La. It was like getting lost in the forest and circling back to the same point, except this circling was abstract, not concrete.
I concluded that it was time to call a halt and to retool. I dug out Chomsky, and then stratificational linguistics (Sidney Lamb's elaboration of Hjelmslev). And psycholinguistics. And cognitive psychology. Physical anthropology and also the evolution of language—why not? The neurosciences. I read more and ran it all up against “Kubla Khan.” It was all fascinating. But nothing clicked.
Somewhere in there, early in 1972, I destroyed my aborted manuscript and signed up for trumpet lessons in the at the Peabody Conservatory; I studied with Harold Rehrig, who’d spent his career with the Philadelphia Symphony. This wasn’t a complete break from reality. I’d played the trumpet since childhood and was fairly good. I’d been working on Louis Armstrong trumpet solos since my early teens, Armstrong had just died, July 1971, what better time to walk another path? After a month or two of that I came to my senses and decided to finish out my Master’s degree — though I kept up the trumpet lessons. I would just have to say whatever I could say and drop the elaborate theorizing.
At that point I had a hunch that something might turn up if I attended to line-end punctuation. Punctuation does a number of things, one of which is simply to group strings of words into larger units. Thus semicolons are more inclusive than commas, colons more inclusive than semicolons, and periods (and other sentence-end marks) are more inclusive than colons and semicolons. As “Kubla Khan” consists of more than one sentence, I had to find ways of grouping sentences into larger units; but that was easy enough. The various texts of the poem have line indentations and skipped lines indicating higher-level structure. While the published texts differ slightly on these matters, it doesn’t require very subtle reading to notice, for example, that the last 18 lines are not nearly so closely coupled with a specific geography as are the first 36, or that lines 31 through 36 are more closely related to the previous lines than to the succeeding ones, and so forth.
Figure 2 is one of my work sheets, though not likely the first one. This sheet deals with the first part of the poem, lines 1 through 36. If you look down the right edge you’ll see that I’ve put red boxes around the line-end punctuation marks and have inserted a red pound sign (#) to indicate a stanza break. Over to the left I’ve created a tree—I’d seen lots of trees in the linguistics I’d read—indicating how smaller units are grouped into larger units until we have one unit spanning the first 36 lines of the poem. I numbered the tree nodes using a standard convention.
Figure 2: The grouping structure of the first 36 lines of “Kubla Khan.”
When you examine that tree closely you’ll see that it has an interesting structure. These 36 lines break into three main sections: 1.1 “In Xanadu . . . spots of greenery, 1.2 “But oh! . . . voices prophesying war!” and 1.3, “The shadow . . . caves of ice.” The middle of those three sections (1.2) in turn has three sub-sections, and the middle of those (1.22) has three components as well. Notice that this middle of the middle is about the eruption of the fountain into Xanadu. All other section divisions are binary. How elegant, how very elegant. And symmetrical.
What’s most remarkable is that the second part of the poem,—the 18 lines from “A damsel with a dulcimer” to “And drunk the milk of Paradise”—has the same general structure. And that despite the fact that the two parts of the poem have very different settings and actors – Kubla, river and fountain, wailing woman, caves, and dome on the one hand, damsel and dulcimer, unnamed “I” and unnamed auditors on the other. Perhaps the most surprising finding concerns the middle of the middle of the middle in these last 18 lines. That section, 2.222 by number, consists of a single line, line 47: “That sunny dome! Those caves of ice!” That, of course, is almost a exact repetition of the concluding line of the first part, line 36: “A sunny pleasure-dome with caves of ice.”
What’s going on? Lots, more than I can summarize here, including the relationship between the rhyme scheme and the tree structure. But these details are readily available (see note below*). The point is simply that this most capital “R” ROMANTIC of poems, a poem born of a drug-induced vision that never completed itself, has an elaborate, rigorous, almost mechanical structure, a highly symmetrical intercalation of binary and ternary groupings. If that structure is evidence of unconscious mentation, that unconscious mind is looking more like a clockworks than a romping-stomping maelstrom of affect and desire. But it’s a clockworks with that ejaculatory fountain (“swift half-intermitted burst”) at its core. Which makes no sense, no sense at all.
Whatever it is, it’s not what I was looking for. I was looking for a nice structure of binary oppositions like I’d seen in Lévi-Strauss’ work on myth, with a bit Roman Jakobson thrown in. As I’ve already indicated, it’s not that those oppositions aren’t there. They are, all of them—dome vs. caves, light vs. sound, sight vs. breath, Kubla vs. the woman wailing, the woman wailing vs. the damsel with a dulcimer, order vs. chaos, time vs. space—and they figure in all my subsequent analytic work on “Kubla Khan.” But I couldn’t play them out in this poem like Lévi-Strauss played them out in those myths. Scratch Lévi-Strauss. Well, perhaps not, but bracket him, and structuralism as well.
It’s worse than that. “Kubla Khan” is one of the best-known poems in the English language and has attracted a great deal of scholarly attention. For all that attention, for all our recognition that it is unique in the body of English poetry, the most basic and elementary set of facts about the poem—its structure—had gone un-noticed and thus without critical commentary. For all our argumentation about whether or not the poem is complete, for all our explication of its meaning, the poem itself has been, in some profound sense, invisible to criticism. It is as though a zoologist were to explicate the lifeways of, say, the Siberian tiger without noting how many legs it had or the nature of its dentition. One would know that it is a predator without knowing how it caught its prey, just as one would know that it ate meat without knowing how the tiger chased its prey and consequently was able to rend that meat from the carcass. How could such a zoology possibly understand the lives of animals?
In a parallel fashion, how could a criticism invisible to the structure of “Kubla Khan” possibly understand how this poem works and why it has been so important? The gap between the elementary facts of “Kubla Khan” and the existing commentary was so wide that I had little choice but to doubt the entire critical enterprise: Is it possible that the elementary formal facts of many texts have gone without notice, and thus without comment?
You might think that’s a bit much: doubt the whole critical enterprise. Really? Remember the times, it was in the early 1970s, and the whole critical enterprise was in doubt, at least at Johns Hopkins. Derrida taught us that Western metaphysics was in flames; that surely included literary criticism. Yes, the particular case that had lodged itself in the center of my intellectual life, this mechanically rigorous but also somehow structuralist analysis of an anomalous poem, was a bit off-center. But then, who could object to being off-center when the center wasn’t holding? No, in tossing over the old ways, I wasn’t engaging in an idiosyncratic act of will, I was just going with the flow.
Besides, there was all this linguistics, cognitive, and neuro stuff just jumping off. Who knows what will develop from that. However, first things first.
And so I wrote up my findings in a Master’s Thesis: The Articulated Vision: Coleridge’s “Kubla Khan.” I didn’t completely abandon the old ways. Many of my old favorites showed up in the footnotes, Wittgenstein, Piaget, Merleau-Ponty, Jakobson, Nietzsche, and, of course, Lévi-Strauss. But the thesis was not dominated by their thinking. It was dominated by that mechanistic structure, and those thinkers were receding into the background and another set showed up in my footnotes, Michael Arbib, Ross Quillian, George Miller, Wilder Penfield, Karl Pribram, Lev Vygotsky, Noam Chomsky, and others. Near the end of the thesis I made an argument that derived from Chomsky. And that’s where I seemed to be headed, to cognitive science.
Institutionally, I headed to the English Department at SUNY Buffalo, which was, at the time, the best experimental department in the nation and which had close intellectual ties with Hopkins. It was my intention somehow to pursue this linguistics and cognitive stuff and to figure out how “Kubla Khan” worked. For one thing, I was attracted to the conceptual style of these linguistics and cognitive models, and they seemed to be the same kind of thing as the structure I found in “Kubla Khan.” That structure was precise, and so where those models. In that, that precision, I was swimming against the post-structuralist tide, which was about indeterminacy.
As things worked out, I was indeed able to pursue the cognitive sciences at Buffalo. On the one hand, the English Department was quite willing to give a bright and headstrong graduate student the opportunity to explore widely—that’s what the department was about. On the other hand, I found someone in the linguistics department to guide me in cognitive science, David G. Hays. Hays had gotten a Ph. D. in Social Relations at Harvard in the mid-50s, spent a year at the Stanford Center for Advanced Study in the Behavioral Sciences, and then took a position at the RAND Corporation. There he instituted the program in machine translation: how do you use a computer to translate from Russian into English? He left RAND in the late 60s to become the first chair of the Linguistics Department at Buffalo. That’s where I found him. He was no longer chair of the department, but he had a small research group in computational linguistics; I joined the group and stayed with it throughout my studies.
Hays taught me to draw diagrams like this:
Figure 3: Hypothetical cognitive network underlying some lines of “Kubla Khan.”
That is a cognitive network and it is closely related to the Model Q that Umberto Eco adopted as his semantic model in A Theory of Semiotics (1976), though he didn’t develop the model at all. The “Q” is for Quillian, Ross Quillian, an AI researcher who’d come up with a model for natural language semantics in the late 60s (which is about when such research began). Such models take the form of a structured network where concepts were related to other concepts by various associative relationships. In my diagram all the lowercase words are nodes in the network and they are connected by links, or arcs as they are often called. Those links have labels, only some of which are actually in the diagram. Those labels indicate the kind of relationship that obtains between the nodes at either end of the link (NRT for inert, APL for apply, PW for part-whole, etc.). A cognitive network is thus a model of the mind’s “deep” semantic structure. Chomsky called it mentalese (though he’s never given it much explicit attention), others simply call it thought or cognition.
The network in Figure 3 depicts some of the semantic underpinnings for lines 31 through 34 of “Kubla Khan”:
The shadow of the dome of pleasureFloated midway on the waves;Where was heard the mingled measureFrom the fountain and the caves.Roughly, the top half of the diagram is about the image (shadow) of the dome being reflected (floating) on the surface of the river, while the bottom half is about the mingled sounds from the fountain and the caves. To go beyond that rough characterization would require . . . a lot. It’s a different world, different assumptions, different problems. A different starting point.
But that’s the cognitive unconscious, or at least (a fragment of) one model of it. That’s the locus of those binary oppositions that are so prominent in Lévi-Strauss’ thinking. That’s a way of thinking about that cognitive base I mentioned in my first Lévi-Strauss article. And what of the subject? If you’ll look carefully in Figure 3 you’ll see two ego nodes, which are, in fact, one and the same. That is not the subject, but it might well be the center of the self.
Maybe.
Perhaps.
I would like to be able to say that my work went well at Buffalo and I was able to create a model through which I could explain both the semantic structure underlying “Kubla Khan” and the surface structure we see in the poem. But I can’t say that. The work did go well, it went very well, but it was not adequate to “Kubla Khan.” I suppose I could have forced things a bit, one can always force things, but I couldn’t see any point to it. Instead, I developed a cognitive model for Shakespeare’s Sonnet 129.**
And that was the last “full” cognitive model I developed for a literary text. I haven’t abandoned that sort of cognitive modeling; for example, “First Person: Neuro-Cognitive Notes on the Self in Life and in Fiction” has a fairly elaborate model for the cognitive structure supporting personal pronouns. But, for various reasons, cognitive networks haven’t proved as useful as I had hoped, nor as many people had hoped.
As far as I can tell, what had at one time appeared to be the conclusion of one intellectual journey has turned out to be but the starting point of a new and heretofore unimagined journey. The point of arrival has become a point of departure. We’re now sailing in a new sea toward parts unknown.
Addendum One
It seems to me that the intellectual guts of the story is there in those three figures, which are quite different in character. Figure 1, of course, is the oldest. This is only a part of a worksheet for the first 36 lines of “Kubla Khan.” Here you see the first 11 lines, typed in triple-space, with various notations on it in four colors, red, blue, green, and black. I probably had something in mind with the color-coding, but it’s not obvious to me what it was. In any event, you see various kinds of annotations: grouping (square brackets left and right, braces), physical inclusion or closure (circle around a dot), various words picked out, and descriptive labels and comments all over the place. And, while I certainly had binaries on the mind, they aren’t very prominent in those annotations. It’s a busy and messy display.
Figure 2 is quite different. Here we have the first 36 lines typed out single-spaced. This illustration has one purpose, to show how those lines are grouped in larger and smaller units. That’s indicated in the tree structure to the left. I’ve also picked out the line-end punctuation marks with small red boxes down the right and transcribed those marks to the tree structure in black. As I said in the narrative, this structure came as a surprise. I wasn’t looking for it, and consequently it took me a long time to find it.
Figure 3 is still different. The poem itself isn’t there at all. But semantic elements from the poem are, eg. done, river, waves, floats, fountain, caves, etc. Just as important, we see semantic elements that are implied by the poem, but not actually in it. For example, at the upper left we have “light beam ---- travels-----path” where the path is said to go from “sun → dome → river → ego,” where ego is the speaker. The poem doesn’t say anything about a light beam nor doe it mention the sun, but how are you going to get the image of the dome on the surface of the river unless there is light going from some source, to the dome, to the river? And what can that source be but the sun? Just about everything we say implies networks of such simple inferences. Much of the challenge in working with semantic networks is simply ferreting out all this commonsense inferences.
As I indicated in the main text, these networks are quasi-formal objects carefully constructed according to a grammar that dictates which kinds of nodes can be connected to one another and by what kinds of links. This grammar also dictates how inferences are made in such networks.
Given that some such network is the cognitive base for all texts, the long-term intellectual project is to come up with a principled characterization of a path through a network like that in Figure 3 that is expressed in a language string broken into parts like the tree in Figure 2. What do I mean by long-term? Maybe in my lifetime, but maybe not. I don’t know.
Addendum Two
A couple years ago Mark Liberman made some remarks on linguistic form that a german to Figure 2 on grouping structure (above):
I've noticed over the years that a surprisingly large fraction of smart undergraduate students have a surprising amount of trouble with what seems to me like a spectacularly simple-minded idea: the simple parallelism between form and meaning that linguists generally call "recursive compositionality", and compiler writers call "syntax-directed translation".I'd say that anyone with a serious interest in describing the structure of literary texts, or movies for that matter, has to be thorougly familiar with this notion. And, Liberman is correct, it's not enough simply to see the concept explained and demonstrated. You have to work through examples yourself.
A trivial example of this would be the relation between form and meaning in arithmetic expressions: thus in evaluating (3+2)*5, you first add 3 and 2, and then multiply the result by 5; whereas in evaluating 3+(2*5), you first multiple 2 and 5, and then add 3 to the result. Similarly, in evaluating English complex nominals, the phrase stone traffic barrier normally means a traffic barrier made out of stone, not a barrier for stone traffic, and thus its meaning implies the structure (stone (traffic barrier)). A plausible way to think about this is that you first create the phrase traffic barrier, and the associated concept, and then combine that — structurally and semantically — with stone. In contrast, the phrase steel bar prices would most plausibly refer to the prices of steel bars, and thus implies the structure ((steel bar) prices).
There are many formalisms for representing and relating linguistic form and meaning, but all of them involve some variant of this principle. It seems to me that understanding this simple idea is a necessary pre-condition for being able to do any sort of linguistic analysis above the level of morphemes and words. But when I first started teaching undergraduate linguistics, I learned that just explaining the idea in a lecture is not nearly enough. Without practice and feedback, a third to a half of the class will miss a generously-graded exam question requiring them to use parentheses, brackets, or trees to indicate the structure of a simple phrase like "State Department Public Relations Director".
* I’ve published two articles on “Kubla Khan.” If you’re interested in the details of that work I’d start with Articulate Vision: A Structuralist Reading of “Kubla Khan.” Language and Style 18: 3 - 29, 1985. You can download a PDF here. I’ve recently published a more detailed study: “Kubla Khan” and the Embodied Mind, PSYART: A Hyperlink Journal for the Psychological Study of the Arts, November 29, 2003. While the second article is more detailed and sophisticated, it’s also more of a slog. The slog will be easier if you’ve first been through the older article.
** Cognitive Networks and Literary Semantics. MLN 91: 952-982, 1976. You can download a PDF of that paper here.
