“Kubla Khan” and the Quest for Poetic Grammar: A Glimpse of the Pacific?

Earlier this month I asserted that I’d, if not solved, dissolved a problem I’d been working on my entire career: The Fluid Mind and a Problem Dissolved. What does that mean? Do I really believe it?

The problem I had in mind is that of Coleridge’s “Kubla Khan”, which I’ve used as a touchstone for my intellectual work. I’ve been trying to figure out just how that poem works. That, in turn, requires the creation of a model or a theory, whatever, in which one can show the process. It’s the search for that model or theory that’s been driving me.

To solve that problem I would, of course, have to have that model in hand. Whether I’ve created it or simply taken it over from some other investigators is a secondary matter. Given the model I could solve the problem by showing how “Kubla Khan” is generated by the model. QED.

But I don’t have such a model, nor do I know when one will created. What then is the difference between failing to solve the problem and dissolving the problem? Put most simply, the difference lies between how I thought about such things back then and how I think about them know.

The Problem, in Words

I became interested in “Kubla Khan” during my senior year at Johns Hopkins. I filed a Master’s Thesis on the poem in 1972 and took that with me to SUNY Buffalo in 1973, where I entered the English Department in pursuit of a Ph.D. Somewhere in there the problem became enlarged from “Kubla Khan” to include “This-Lime Tree Bower My Prison”. I don’t know just when and how I first formulated that problem, but I published a short note on in in 1981: Metaphoric and Metonymic Invariance: Two Examples from Coleridge. MLN 96: 1097-1105. Here’s how I stated the problem then (pp. 1103-05):

“Kubla Khan” and “This Lime-Tree Bower My Prison” are two very different poems by the same poet. But they have the same two-part structure and they share imagery as well. The roaring dell of “Lime-Tree” corresponds to the savage chasm of “Kubla Khan.” The concern with sight and sound manifest in “Kubla Khan” shows up in “Lime-Tree Bower” in the image of the creeking rook flying across the sun. And the way in which both Charles and the poet have access to that sight gives it a role similar to the sunny dome and caves of ice in “Kubla Khan,” where both the poet and his audience are linked through the image. These two poems share the same world.

But they take radically different paths through it. One path is regulated by metonymy and unfolds through two consciousnesses moving through different parts of the same landscape [“Lime-Tree”]. The other path is regulated by metaphor and so unfolds itself in two different worlds linked by a common image; the path it takes through these worlds is, however the same [“Kubla Khan”].

The burden of my note to that point had been to demonstrate the points I summarized in those two paragraphs. I continued:

If we are to begin serious efforts to construct poetic grammars we need techniques which will allow us to separate poetic worlds from paths through those worlds. The grammar is then the procedure for constructing metaphoric and metonymic paths through those worlds.

That word “grammar” is crucial, as is the notion of a path, which I’ll take up shortly.

What is a grammar? If you look the word up in a dictionary it will talk about sentence structure and rules of language. That “a complete sentence must have a subject and a verb” is a rule of grammar, a rule formulated in ordinary language using terms that refer to language: sentence, subject, verb. During the previous two or so decades, however, linguists had come to formulate grammars in some formal or quasi-formal notion derived from formal logic or mathematics. And there had been efforts to extend such formalization beyond the sentence level to whole texts. George Lakoff, for example, attempted to use Chomskian methods to re-write Vladimir Propp’s work on the fairy tale and story models (PDF) were a hot item in AI for awhile (e.g. in Roger Schank’s laboratory at at Yale in the 1970s). Meanwhile Lévi-Strauss was using imagery from algebraic group theory to talk about myth (Mythologies) and narratology was developing informal and quasi-formal models.

The Problem, with Pictures

There are two things going on here. First, we need a systematic approach to language above the sentence level. We need to account for whole texts. Vladimir Propp had attempted that in his well-known Morphology of the Folktale (1928). Second, we need to formulate that account in some formal language.

I sought to do both. My core texts weren’t stories, but poems, but that was a secondary matter. Here’s how I stated the matter in the final paragraph of that note:

The question is: Can we represent thought without becoming hopelessly enmeshed in the process of representation? The contortions of post-structuralist prose suggest that we are so deeply enmeshed in that process that we have all but given up the possibility of ever inhabiting a common intellectual world. The cognitive sciences have their contortions as well, but at least they have modes of thought in which it is possible to distinguish between the model – which is notated in some formal language, such as a directed graph or a computer program – and commentary on the model, which is constructed in ordinary language. In such models it is possibly clearly to distinguish between a represented semantic world and the paths taken through that world to generate texts. These models are still quite primitive, but there is room for growth. There may even be room enough to grow into a model full enough to show us how “Lime-Tree Bower” and “Kubla Khan” are but different visions of the same world.

Once again that word “path” appears, along with the notion of a directed graph.

Here’s a directed graph:

The most obvious sort of commentary on such a model is that needed to link it to the world. The model itself consists of nodes and arcs (or links) between the nodes. There’s a rich mathematical discourse about such things. But, if we’re going to use such a model to understand texts, then we need to connect the formal objects in the model, the nodes and arcs (and other objects defined over them) and our texts. The nodes might then represent words and the arcs represent relationships between words. That’s the general idea, but it’s a bit more abstract than that in any real model.

Given such a model, here’s one path through the graph:

And here’s a different path:

Those paths are clearly different, hence they should characterize different texts.

The problem I faced was to create a model, a directed graph, such that one kind of path (which I’d identified with metaphor) through it would yield the text of “Kubla Khan” while another, and distinctly different, kind of path (which I’d identified with metonymy) would yield the text of “This Lime-Tree “Bower My Prison.” Further, as “Lime-Tree Bower” is one of a group of poems known as the Conversation Poems, I would expect those to be generated by paths of the second kind as well, but not by a path of the first kind.

What has Changed?

So, what is it that has failed? That’s one question. And why don’t I regard that failure as a failure of my intellectual project? That is, why do I think of that project as having dissolved rather than having failed? That’s the second question.

What has failed? I haven’t been able to create a directed graph that has the desired characteristics. Nor do I think such a graph exists, even in principle. I took another run on “Kubla Khan” in 2003 and published on “Lime-Tree Bower” in 2004. In looking over that work I have concluded the problem isn’t going to be solved by some grand graph of that kind. We need a richer set of tools.

And a richer set of tools has in fact grown up since 1981. We need to figure out how to use them in conjunction with those older tools – there’s nothing wrong with them, but they can’t do the job by themselves. While I certainly don’t know how to use them to cobble together a richer model, one more adequate to the task, I have no reason to think that others won’t be able to make progress.

I further note that my failure on this particular problem is not a merely personal failure. As I explained in my article on computational linguistics and literary study, that whole approach (symbolic processing) to language and mind collapsed in the mid-1980s. I don’t think the problem is that I’ve missed some little trick or technique that’s lying in wait for me out there in that vast literature. Such a thing doesn’t exist.

At the same time the literary critical schools that emerged in the 1970s and 1980s are bottoming out. There’s little or nothing new coming out of post-structuralist discourse. Those ships have run aground as well.

So there’s no point in trying to pursue that old problem in the old ways. That’s over. We’re in a new intellectual world or, if not exactly in it, then we’re getting ready to embark on new investigations.

How can THAT be considered to be a failure? No, the old problems are gone and new ones are waiting to be defined and investigated. We’ve made it across the continent and are looking at the Pacific Ocean. Do we construct a boat and set sail, do we turn and head down the coast (or up), or do we head back inland?