2 Comments on Moretti’s LitLab 15: Patterns and Interpretation [#DH] [now 3 comments]

And now let me add a third comment based on a 2016 lecture which covers much of the conceptual ground in this pamphlet. I've added the third comment after the first two. Accordingly, I've bumped this post to the top of the queue.

Franco Moretti has produced another pamphlet:

Franco Moretti, Patterns and Interpretation, Stanford Literary Lab, Pamphlet 15, September 2017, 10 pp. [In Canon/Archive (2017) as the conclusion, pp. 295-308.]

I have two comments to offer. In the first I contrast meaning with explicit semantic models while I the second I suggest that Lévi-Strauss was in pursuit of the form of myths.

Meaning and Semantics

Early in the pamphlet Moretti remarks (p. 2):

Now, meaning is not one of the things literary critics study; it is the thing. Here lies the great challenge of computational criticism: thinking about literature, removing meaning to the periphery of the picture. But of course this is also the great challenge for computational criticism: you discard meaning and replace it with – what?

Yes.

Meaning is a peculiar beast. It is inherently subjective, by which I mean primarily that it arises within subjects; that subjectivity is its mode of existence. As a practical matter assertions about the meaning of texts varies widely and seemingly idiosyncratically from one critic to another, though agree is possible. The former is subjectivity in the ontological sense while the later is subjectivity in the epistemic sense – a distinction I have from John Searle [1]. That the meaning of texts is ontologically subjective implies that, as investigators, we cannot grasp as an object of examination and investigation. All we can do is translate, paraphrase, wheedle and cajole, in an attempt to indicate a text’s meaning(s). It’s a slippery Heraclitean business.

Moretti, if I understand him correctly, would have us, as computational critics, set meaning aside in favor of pattern and, I believe, form. I’m OK with that in the context of contemporary digital criticism. However, there is a different, though by no means contradictory, “replacement” for meaning.

Semantics –

By which I mean an explicit account of how words have, well, you know, meaning. Back in the 1960s cognitive scientists began to create such explicit accounts, often in the form of computer simulations. I learned one such model when I studied with the late David Hays in the Linguistics Department at SUNY Buffalo and employed it in an account of Shakespeare’s Sonnet 129, “The Expense of Spirit” [2]. The model took the form of a semantic or cognitive network where the nodes represented ‘words’ (approximately, more or less) and the edges connecting them represented the relations between them.

This diagram depicts a small fragment of such a model:

One could then ‘account for’ the text as a path through the semantic model. This depicts a fragment of such a path, which arrows pointing from the text to the network fragments they realize:

I say account for because such a path doesn’t actually explain the text. It doesn’t tell you why that path was taken in the first place, nor why it takes this form rather than that form. That’s a different matter.

My point is that such a model doesn’t “give” you the meaning of the text. When you examine such a model, when you follow a text through such a model, you are not reading the text in any reasonable sense of the word. You are simply examining the operations of a model, and a very complex one at that.

Such computational models are different from Moretti’s computational criticism. In computational criticism computation is used to find patterns in a corpus of texts, but it is not intended as the simulation of a mind reading a text. But simulation is the point of those (now classical) cognitive network models. Such models were developed to the point where they could simulate the reading of small and rather simple texts. Hays and I imaged that one day they might become sophisticated enough to deal with literary texts, but that day has yet to come – and I’m not holding my breath [3]. Those models, however, laid the foundations for the so-called semantic web and they are the basis of those ontologies that pervade computing these days.

Such semantic models give us clues as to how the mind constructs meaning. But if it’s the meaning you’re after, you have to get it directly by reading the text. The model won’t help.

Form: Lévi-Strauss, Ring-Composition

Near the end of the pamphlet Moretti observes (p. 9):

If this pamphlet began by moving away from the lived experience of literature towards abstraction, form has emerged as simultaneously the apex of the process, and the turning point that allows to reverse the direction, and return from abstraction to literary history. Here, form functions very much like Max Weber’s concept of the ideal-type: “a mental construct” that “cannot be found empirically anywhere in reality”, but that, once constructed, can be used “for comparison with and measurement of reality”. This is exactly how we should think of literary form: a mental construct which we will never find as such in individual works, but which we can use to “measure” their relationships. Form will never explain a single text, and is the only thing that can explain a series of them.

That, it seems to me, is where Lévi-Strauss took his structural study of myth. By the time he got to Mythologies he was looking for the internal relationships that govern individual myths by comparing them with one another in such a way that the underlying logic is revealed.

I explain this in a working paper, Beyond Lévi-Strauss on Myth [4]:

Just what does Lévi-Strauss mean by transformation? While one might think that he’s making a genetic argument, that some myth X is derived from some other myth W by applying some transformation t to W to yield X, that is not at all what he means. It’s not at all clear that these transformations are anything but analytic tools, or fictions if you will, as opposed to being operations within the myth system itself. The idea is that myths are expressions of underlying mental structures, including kinship systems and the associated roles, all the arts of living, lore about local flora and fauna, the local geography, and so on. Beyond an insistence upon binary oppositions, Lévi-Strauss has no explicit account of how that knowledge is organized and so cannot offer a very explicit account of how myths are derived from that knowledge, which is largely unconscious. That is, he was unable to say: 1) Here is a body of knowledge organized in such and such a way. 2) Here is a procedure that operates on that body so that, 3) upon execution, we have a myth.

What he does, instead, is offer an indirect argument. Let us say that the cognitive base (CB) for a myth is that “region” of cognitive structure, and no more, that supports the myth. When Lévi-Strauss talks of a transformation obtaining between M1 and M2 he’s saying that CB1 and CB2 are the same except in some limited respect. That limited respect is what he incorporates into the transformation. If you transform CB1 in that limited respect, then you have CB2. Notice you could easily think of the difference as going in the other direction as well, from CB2 to CB1. The point is that there is a limited and specific difference. If one then generates myths M1 and M2 from CB1 and CB2, respectively, those myths will be different in a way that can be attributed to the difference between the two cognitive bases. (p. 5)

As I assert later in the paper, Lévi-Strauss is in effect asserting that “there is an economy to myth such that if you change one or three things here, you get correlative changes there” (p. 12). He was trying to theorize the form of myths without quite having the concepts with which to do it. So he adopted terms from the algebra of groups and uses them loosely, metaphorically (which he acknowledges in the introduction to The Raw and the Cooked).

Ring composition presents us with a different problem [5]. We know what the form is:

A, B, C...X...C’, B’, A’

Some element, A, at the beginning of the text must be matched by some element, A’, at the end, and so forth, with a central element, X, marking the turn.

Here I am asserting, contra Moretti, that we can find such form in individual works. But these ring form texts don’t declare themselves as such and there is little to no evidence that they are consciously contrived in the way that, say, sonnets and sestinas are [6]. If, in the end, we are to understand this form it will be by examining and comparing many putative examples so that we can note which features seem general to the form and which are particular to some individual text or group of texts.

I do not, however, see as how computational criticism can identify such texts. That must be done by human investigators.

Only now, with computational criticism, is form visible to us

Moretti says that in a 2016 lecture,  Hidden in Plain Sight: Unspoken Premises of DH Visualization, which he delivered at EPFL (École polytechnique fédérale de Lausanne). He notes that form is an obscure concept, which it is, only now does computational criticism allow us to see it. How so?

If there is a pattern in the data, it’s because behind it there is a form which repeats itself over and over again. Form is the repeatable element of literature, as I wrote elsewhere, and the pattern is nothing but this repetition. [p. 6]

Computation picks up the repetition and reveals the repeated elements to us. Now we can see form.

Very interesting, but I'm not convinced. Yes, computation reveals repetitions what had been otherwise invisible, and that is good. And, OK, form involves repetition – sometimes. Hence Moretti believes that pattern cannot be discerned in individual texts – the the passage I quoted in the ring-composition section, especially the part that I've highlighted.

But now I can see, as well, why he believes there to be an intimate relationship between form and history, which comes up in the recent essay he wrote with Oleg Sobchuk, Hidden in Plain Sight: Data Visualization in the Humanities (New Left Review 118, July-August 2019, pp. 86-115). I'm currently working on a piece in which I discuss that essay.

References

[1] John R. Searle, The Construction of Social Reality, Penguin Books, 1995, pp. 5 ff.

[2] William Benzon, Cognitive Networks and Literary Semantics, MLN 91: 1976, 952-982. URL: https://www.academia.edu/235111/Cognitive_Networks_and_Literary_Semantics

Lust in Action: An Abstraction, Language and Style 14, 1981, 251-270. URL: https://www.academia.edu/7931834/Lust_in_Action_An_Abstraction

[3] Some years ago David Hays and I proposed a hypothetical system that would ‘read’ texts. We called it Prospero: William Benzon and David Hays, Computational Linguistics and the Humanist, Computers and the Humanities 10: 265 - 274, 1976. URL: https://www.academia.edu/1334653/Computational_Linguistics_and_the_Humanist

I have recently recast the old idea in contemporary dress: William Benzon, Virtual Reading: The Prospero Project Redux, Working Paper, September 2017, 32 pp. URL: https://www.academia.edu/34551243/Virtual_Reading_The_Prospero_Project_Redux

[4] William Benzon, Beyond Lévi-Strauss on Myth: Objectification, Computation, and Cognition, Working Paper, February 2015, 30 pp. URL: https://www.academia.edu/235111/Cognitive_Networks_and_Literary_Semantics

[5] I have a number of papers on ring composition, most of them practical criticism, collected here: https://independent.academia.edu/BillBenzon/Ring-Composition

For systematic discussion, see William Benzon, Ring Composition: Some Notes on a Particular Literary Morphology, Working Paper, September 11, 2017, 71 pp. URL: https://www.academia.edu/8529105/Ring_Composition_Some_Notes_on_a_Particular_Literary_Morphology

[6] Mary Douglas believed that such composition was or conscious (at least in some cases), but presents no evidence. See her Thinking in Circles: An Essay on Ring Composition, Yale University Press, 2007, p. 32.