Thursday, June 25, 2015

On the Direction of 19th Century Poetic Style, Underwood and Sellers 2015

Another working paper (title above). Download at:

Abstract, contents, and introduction below:

Abstract: Underwood and Sellers have discovered that over the course of roughly a century (1820-1919) Anglo-American poetry has undergone a consistent change in style in a direction favored by editors and reviewers of elite journals. This directional shift aligns with the one Matthew Jockers found in Angophone novels during roughly the same period (from the beginning of the 19th century to its end). I argue that this change is characteristic of a cultural evolutionary process and sketch a way to simulate such a process as an interaction between a population of texts and a population of writers where texts and writers. I suggest that such directionality is a sign of autonomy in the aesthetic system, that it is not completely coupled to and subsumed by surrounding historical events.

C O N T E N T S

0. Introduction: Looking at Cultural Evolution whether You Like It or Not 2
1. Cosmic Background Radiation, an Aesthetic Realm, and the Direction of 19thC Poetic Diction 8
2. Beyond Whig History to Evolutionary Thinking 14
3. Could Heart of Darkness have been published in 1813? – a digression 19
4. Beyond narrative we have simulation 22

0. Introduction: Looking at Cultural Evolution whether You Like It or Not

I was of course thrilled to read How Quickly Do Literary Standards Change? (Underwood and Sellers 2015). Why? Because they provide preliminary evidence that 19th century Anglophone poetic culture has a direction. Just what that direction, and how to characterize it, that’s something else. But there does appear to be a direction. And just why is that exciting? Because Matthew Jockers made the same discovery about the 19th century Anglophone novel. To be sure, that’s not what he claimed – I’ve had to reinterpret his work (see my working paper, On the Direction of Cultural Evolution: Lessons from the 19th Century Anglophone Novel) – but that’s what he has in fact done.

So we’ve got two investigations making the same observation: there is a long-term direction 19th century literary culture. But not the same, as Jockers looked at novels and Underwood and Sellers looked at poetry. Moreover their observational methods are quite different. Jockers uncovered direction by looking for similarity between texts where similarity judgments are based on a variety of stylistic measures and on topic analysis. Underwood and Smalls bumped into directionality by looking for differences between the general run of literary texts and texts selected for review by elite publications. Jockers’ work, almost by design, uncovered continuity between successive cohorts of texts, but simply ignored elite culture. Underwood and Smalls had no explicit interest in local continuity but, by looking at elite choice, uncovered a possible factor in directional cultural change: the “pressure” of elite preference on the system as a whole.

Thinking in High-Dimensional Spaces

All these differences not-withstanding, I note that both investigations employed the common mathematical trope high dimensional space, 3200 dimensions for Underwood and Sellers and roughly 600 dimensions for Jockers. These spaces are of high dimensionality only in comparison to the three dimensions of the ordinary physical world and the two-dimensions the printed page, the television or movie image, and the computer screen. Physicists are used to spaces of such high dimensionality that, for all practical purposes, they are of infinite dimensionality.

More to the point, some neuroscientists use high-dimensional space to think about the nervous system, where each neuron (of c. 100 billion, which is, for practical purposes, infinite) is assigned one dimension. The point of this exercise, of course, is so we can think of the state of the brain at any moment as a point in high-dimensional space. Correlatively, the evolution of a brain’s state is a trajectory in this high-dimensional space.

Jockers represents a novel as a point in 600-dimensional space. Underwood and Sellers represents their poems in a 3200-dimensional space. The relationship between these spaces and trajectories in human neural space, the trajectories evolved as a person reads the words in question, is somewhat obscure. How one moves, conceptually, from the neural space of individual readers to something like the collective neural space of a population of readers, that is all but beyond imagination. (Though I’ve begun imagining something like it in chapters two and tree of Beethoven’s Anvil, 2001, where I conceptualize a basic neural model interaction of people making music together.) But something like that is nevertheless in play in these investigations for the texts they investigate have been written and read by human beings and the historical trajectory of those texts thus reflects the attitudes, dreams, and desires, the minds, of those people.

The direction that Jockers found, but didn’t conceptualize as such, and the one that Underwood and Sellers found, but weren’t looking for, those directions are, most immediately, directions in the abstract high-dimensional spaces they employed in their computer models. Those models themselves are not temporal. But the texts are dated and when you examine the relationship between those dates and the positions those texts have in those spaces we can see there is direction to those dates, a systematic drift in space. I take that drift as an indication that the underlying causal process is an evolutionary one.

And when you start to think about just what that causal process is, how it operates, where do you end up? Our data is about texts, but we know those texts were written by individuals (most of them), published in moneymaking ventures (not always successful), and were read by thousands and tens of thousands of people, some of them at least. Those people led complicated lives. Reading would have been only a part of those lives.

What part? Here’s my touchstone passage from Kenneth Burke’s, “Literature as Equipment for Living” (The Philosophy of Literary Form 1973). Using words and phrases from several definitions of the term “strategy” (in quotes in the following passage), he asserts that (p. 298):
. . . surely, the most highly alembicated and sophisticated work of art, arising in complex civilizations, could be considered as designed to organize and command the army of one’s thoughts and images, and to so organize them that one “imposes upon the enemy the time and place and conditions for fighting preferred by oneself.” One seeks to “direct the larger movements and operations” in one’s campaign of living. One “maneuvers,” and the maneuvering is an “art.”
Millions of people making sense of their lives, that’s what we’re looking at in those 600 or 3200 dimensions. Why not call it what it is, the evolving spirit of the times, Geist? Or, if not the entirety of the spirit, it’s Jet Stream or Atlantic Current, the large movement in which local and regional turbulences and whorls find their way.

Evolutionary Dynamics and Cultural Process

Neither Jockers nor Underwood and Sellers have advanced evolutionary accounts of their discoveries. Not only that, they’ve all explicitly rejected such accounts. Jockers glosses his rejection by quoting the Wikipedia entry for a book (Darwin’s Dangerous Idea) by the philosopher Daniel Dennett (pp. 171-172):
Darwin’s discovery was that the generation of life worked algorithmically, that processes behind it work in such a way that given these processes the results that they tend toward must be so” (Wikipedia 2011a, emphasis added). The generation of life is algorithmic. What if the generation of literature were also so? Given a certain set of environmental factors–gender, nationality, genre–a certain literary result may be reliably predicted; many results may be inevitable. This is another dangerous idea, perhaps a crazy one.
Underwood and Sellers seem to have a similar reservations when they assert (p. 21):
We hope readers will remain wary of metaphors that present historically contingent standards as an impersonal process of adaptation. We don’t see any evidence yet for analogies to either Darwin or Lamarck, and we’ve insisted on the difficulty of tracing causality exactly to forestall those analogies.
I believe these fears are unfounded. As I’ve devoted a section of this working paper to addressing Underwood and Sellers (2. Beyond Whig History to Evolutionary Thinking) I’ll set that reservation aside for the moment and note part of my response to Jockers:
If cultural evolution should happen to work like that, well then Jockers is right to be worried. For that sounds like Edward Said’s late career plea for an autonomous aesthetic realm is in vain (“Globalizing Literary Study”, PMLA, Vol. 116, No. 1, 2001, pp. 64-68). Literary culture is ‘determined’ by external forces (the forces of production?) and that’s that. QED.

But is that really so? Is evolution algorithmic?

That depends on what you mean by algorithm. In the strictest sense an algorithm is an effective computational procedure that produces a result in a finite number of steps. As far as I can tell, biological evolution is not algorithmic in that sense. Just as “deconstruction” has come to have a popular sense that’s only vaguely related to the work of Derrida, “algorithm” has come to have a popular sense that’s not tightly coupled with the sense it has in mathematics and computer science. In this popular sense an algorithm is simply a procedure that proceeds according to rules, or something. […]That biological evolution is lawful doesn’t make it algorithmic in any strict sense, no more so than the regularity of planetary motions means that they are executing “Newtonian algorithms.”

Biological evolution is an open-ended process that is constrained and lawful, but it is not algorithmic in any useful sense. Nor is cultural evolution.
Assuming you accept my assertion that evolution isn’t the mechanical process Dennett implies that it is, so what? What does evolutionary thinking get you?
As far as I know, adopting evolutionary thinking will not give digital critics access to off-the-shelf concepts and models that will help with the problems they want to solve. What’s it good for?

The computational models are obviously important. But they aren’t everything. We need ways of thinking about the phenomena being modeled. How do we think about thousands and tens of thousands of texts being read by millions of readers (even if we have no direct information about those readers we can use in our models)?

Evolutionary thinking is population thinking. Biologists thinking about interacting populations of organisms. Digital critics have begun thinking about texts by the population, not one by one, and readers are always implicit in their investigations.

Still, what’s to be gained? Two things: 1) Agency and, 2) an aesthetic realm that is a force in history. It is not a fully autonomous force, but it is a force, long-term and unrelenting. I discuss it a bit at the end of my first post in this series: 1. Cosmic Background Radiation, an Aesthetic Realm, and the Direction of 19thC Poetic Diction. As for agency, Underwood and Sellers hint at it (p. 21): “If individual writers are already historical agents, then perhaps the system of interaction between writers, readers, and reviewers also tends to establish a resonance between (implicit, collective) evaluative opinions and directions of change.” If those writers are historical agents, then so are the readers and reviewers and the resonating system they embody is an aesthetic agency. To think in evolutionary terms is thus to think of aesthetic force as a historical agent.

The Problem of Directionality in History

What Underwood and Sellers found particularly problematic, however, was the apparent fact of directionality, for it raised the specter of Whig history, of reading history as a narrative of progress. There are a number of things one can say in response to this, and I’ve said many of them in this paper, but one I haven’t said was brought up by José Angel García Landa in response to one of the posts in this paper:
Whig history shouldn't be entirely avoided, in the sense that a Whiggish sensibility does have some use in detecting those elements that (oh yes) eventually lead to a Whiggish disposition, which after all has not emerged from nowhere. Of course there is the attendant danger of hindsight bias etc. which has to be kept in mind. But at bottom this is an issue much resembling the question of the anthropic principle in evolution: wrong if you make higher consciousness a teleological aim in evolution, right if it forces you to recognize that, after all the whole tangled web of life, consciousness is both the most complex evolutionary phenomenon (or perhaps in a weaker formulation, one of the most complex evolutionary phenomena) and the only one where reflexive models on evolution are built. […] That's one sense in which evolution IS, after all, directional, and not a mere matter of wiggling around. It is cumulative in many ways detailed both by Darwin and by population genetics, and it is also cumulative at the level of culture, which has been called Lamarckian in the sense that cultural evolution does preserve favorable acquired traits in a direct way.
I am, of course, in sympathy with this and would argue that evolutionary thinking provides a way of accounting for directionality without positing an ultimate telos to history.

In any event, Stephen Greenblatt has placed directionality on the table in two of the essays in Learning to Curse (Routledge 1990). In “The Cultivation of Anxiety: King Lear and His Heirs” (pp. 80—98) Greenblatt opens with a long passage from an early 19th century magazine article on the how the Reverend Francis Wayland broke the will of his 15-month old child. Greenblatt notes that “Wayland’s struggle is a strategy of intense familial love, and it is the sophisticated product of a long historical process whose roots lie at least partly in early modern England, in the England of Shakespeare’s King Lear.” To be sure, one need not read that as any more than a statement of historical contingency, that Shakespeare’s play just happened to have been written before Wayland’s article. But when one considers the larger institutional changes Greenblatt considers – from the public space of the king’s court (and Elizabethan stage) to the privacy of the bourgeois home – one may suspect that Greenblatt is tracking the directionality of literary time, that one text must necessarily have been earlier in the historical process in which both texts exist.

Later in that collection, in a chapter entitled “Psychoanalysis and Renaissance Culture,” Greenblatt contrasts the conception of the self implicit in the story of Martin Guerre in 16th Century France with the conception of the self implicit in Freud’s psychoanalytic theorizing. Those conceptions are very different. He opens the essay by recounting the well-known story of Martin Guerre, a French peasant who lived in the mid-sixteenth century. In brief, ten years after his marriage he quarreled with his father and walked away from his family: father, wife, and son. A few years later he returned and took up his life again and fathered two daughters. Alas, troubles broke out between him and an uncle and those troubles led to trials. The legal proceedings revealed that this Martin Guerre was, in fact, an imposter, a man named Arnuld du Tilh, who was subsequently condemned for his imposture and hanged to death on September 16, 1560.

Greenblatt takes this as an opportunity to contrast the psychoanalytic conception of a body-rooted subjectivity with a rather different conception of self at play in du Tilh’s trials. Of this other conception Greenblatt (136-137) says:
At issue is not Martin Guerre as subject but Martin Guerre as object, the placeholder in a complex system of possessions, kinship bonds, contractual relationships, customary rights, and ethical obligations. Arnauld, the court ruled, had no right to that place, and the state had the obligation to destroy him for trying to seize it. Martin’s subjectivity ... does not any the less exist, but it seems peripheral, or rather, it seems to be the product of the relations, material objects, and judgments exposed in the case rather than the producer of these relations, objects, and judgments.
What is at stake here is not so much the subject as some ultimate privacy and interiority in the Cartesian manner, but the subject as a crossroads of social relations, a node in a Latourian network of actors. This conception, Greenblatt argues, is very different from the Freudian conception, which is not only characterized by interiority, but by a hidden interiority.

In both of these cases, anxiety on the one hand, subjecthood on the other, there is a difference between two historical situations such that not only is one later than the other, but that order is not merely contingent. It seems that somehow one set of events MUST have been before the other set. This difference seems to involve the fundamental architecture of the self and so cannot be accounted for through the mere accumulation of facts or knowledge of some sort over historical time.

Directionality is there, waiting for us. How are we going to account for it? It is of course one thing for me to say that cultural evolution is the way to go – it’s the only game in town – but it’s quite something else to demonstrate that. That’s going to be the work of a generation, at least.

Finally, I note that I do not conceive of cultural evolution as an alternative to the contingencies of history. Rather, I am interested in the role of cultural evolution IN literary history and in how evolutionary process brings order to strings of contingency. Moreover I certainly don’t think that cultural evolutionary processes are always directional. I rather suspect that in many times and places century scale and even millennium scale cultural processes have no direction. They just drift.. Under what conditions, what contingencies, does directionality arise?

About the Working Paper

The body of this working paper consists of four posts I wrote between late May and mid-June of 2015. Here’s some brief comments about each:

1. Cosmic Background Radiation, an Aesthetic Realm, and the Direction of 19thC Poetic Diction: The basic post in which I make the connection between Underwood and Sellers on poetry and Jockers on the novel. I conclude with some remarks the autonomous aesthetic realm and its relationship to human biology on the one hand and history on the other. As for cosmic background radiation, that’s an analogy. Penzias and Wilson bumped into when they were looking for something else. In effect, Jockers, Underwood, and Sellers bumped into the autonomous aesthetic realm when they were looking for something else. The directionality they found is a sign of that autonomy.

2. Beyond Whig History to Evolutionary Thinking: This is a detailed commentary on two paragraphs in which Underwood and Smalls argue that, while the directionality they’ve discovered may look like Whig history, it may be something else. Regardless, we’ve got to deal with it. As you would expect, I argue that it’s evolution and it’s fine.

3. Could Heart of Darkness have been published in 1813? – a digression: Another look on directionality in which I argue that, not, Heart of Darkness couldn’t have been published in 1813 because the literary system of the time couldn’t support it. I also address this issue in the concluding section of this introduction.

4. Beyond narrative we have simulation: Here we take a close look at the system of writers, readers and reviewers though the exercise of sketching a computer simulation of that system. The system takes the form of an iteration of cycles in which texts are read by readers. The population of texts changes from one cycle to the next where the population of texts available at cycle N+1 depends on what texts have been read at cycle N. The causal ‘force’ of the simulation is local, from one cycle to the next. Any long-term direction will thus be an ‘emergent’ phenomenon of local interactions.

An Exercise for the Reader

While preparing this introduction, I took another look at the methods section of Underwood and Sellers. I found this very interesting statement: “...the model always (for some unknown reason) sees early works as less likely to be reviewed” (p. 32). They then go on to tell us how they’ve compensated for that bias.

I’m wondering if that bias merits looking into. Their model treats the overall semantic space of this poetry as an atemporal abstraction. The major, and unexpected, finding is that existing poetry drifts “upward” in that space over the course of a century.

One might argue, however, that what is going on is the evolution of the collective cultural space in which poetry is read and interpreted. Thus, the space is not the same in 1820 as it is in 1867 nor as it is in 1919. What this implies is that at least some poems that were written in 1919 couldn’t, in principle, have been written in 1867, nor could some poems written in 1867 have been written in 1820 (and don’t most of us believe something more or less like that?) because the ‘space’ those latter poems inhabit simply didn’t exist at earlier dates.

The converse propositions, about the possible writing of earlier poems at later dates, are tricky. There is a sense in which earlier poetic ‘spaces’ continue to exist at later dates. Early poems are always available to later readers (the converse obviously is not true). But how often are they in fact read, and by what readers in what contexts? But what about writing? There is deliberate anachronism, but how many poets engage in it? Barring that, how often do poets move from the contemporary ‘space’ write in an older one?

What doe these questions mean? Can we see the model’s bias against early works as an effect of these kinds of issues? Why or why not? The major work, it seems to me, is clarification.

4 comments:

  1. I'd be very interested to see how GM Hopkins would plot out, using the actual dates for the MS of his specific poems, against the overall trend described, Bill. To my human eye and ear, he was such an outlier -- hence inter alia the almost complete absence of publication of his major works in his lifetime (1844-89), the posthumous first publication of the majority of his poems in 1918 & their final critical acceptance on publikcation of the second edition 1930. The Windhover, for instance, was apparently written in 1877.

    Of note & perhaps use:

    George Landow, Gerard Manley Hopkins and the Difficulties of Victorian Poetry
    http://www.victorianweb.org/authors/hopkins/difficulty.html

    It'll be interesting when what I'll call the 'wavefront' approach you have been exploring meets the tracking of individual poems in a particular poet's corpus in this way.

    ReplyDelete
  2. Given his publication date it's possible that Hopkins wasn't included in the study at all. You might want to read the original paper, which says a bit about the characteristics of the poetry selected by elite journals. They tend to be dark in tone and to use concrete language, a rather broad specification.

    ReplyDelete
  3. This comment has been removed by the author.

    ReplyDelete
  4. [ edited to correct typos, improve style, etc]

    He may not have been published (much) between 1820-1919, but he was certainly "doiing poetry" well within those boundaries. But I'm not too worried whether he was included or not -- even if he wasn't it would be interesting to make a comparison.

    Adding:

    I've now downloaded Underwood & Sellers, followed their trail to GitHub, and found that Hopkins' first volume was included:

    https://github.com/tedunderwood/paceofchange/blob/master/poems/gerardmhopkins1918.poe.tsv

    --- though 1918 is entirely the wrong date at which to place a poem such as The Windhover!

    ReplyDelete