Blog Post

Discussion: Stanford Lit Lab, Pamphlet 5

To kick off our conversations in the Literature and Digital Humanities group, I thought it might be nice for us to discuss some important critical work in the field. I'd propose the Stanford Literary Lab's fifth Pamphlet--"Style at the Scale of Sentence"--as something we might start of with. 

It's a brilliant, persuasive argument about the relationship between authorial style and the types of countable elements of language that are susceptible to computerized analysis. In the interests of collocating my money and mouth, here are a few discussion questions that occured to me this time through the pamphlet. 


(1) What do we make of the genre of the article itself? It's written as a narrative of ongoing research that seems to turn the methodology of the sciences into a mystery novel (whether we formulate that as hypothesis-->experiment or idea-->information gathering). Sentence-opening phrases like "The more one looked, the clearer" and closing phrases like "...a comparable surprising was awaiting us..." build suspence by delaying the revelation while emphasizing the ongoing process of discovery. The temporal organization contributes to the same effect. 

It's a brilliant stylistic gambit. The pamphlet is a delight to read. But will this rhetorical gesture stand up over time? How does it relate to the emphasis on the anecdote in (ie) New Historicist writing?

In part, this narrative of discovery is imitating the "feedback loop" of interpretation described on page 28. How different is that loop from Gadamer's hermeneutic circle?


(2) The most interesting insights of the piece, it seems to me, come from insightful theories by particular critics of the general tendences of particular sentence structures. The observations about crescendo and diminuendo, sequencing and stasis, feel sharp and perceptive regardless of the data underlying them. (And, indeed, these generalizations tend not to be defended with recourse to quantity. We rather get phrases like "in case after case.") 

This builds to a head in section 6, with the case of /Middlemarch./ Does the account, on page 23-24, of the difference made by quantiative analysis for the study of that text convince? I find it persuasively put when restated in the abstract on page 28: that in formulating questions that are amenable to digital study, we generate "entirely new type[s] of problem" that lead to new archives and new insight. 


What else leaps out about this piece? What other pieces might we discuss?






Thanks for getting this conversation started!  It might just be because it's my wheelhouse, but I found this idea absolutely fascinating: "Conan Doyle used the unconscious expectations arising from this grammatical fact to perfection when he placed clues in dependent clauses, thus making them visible while suggesting to readers that nothing important was being said" (8). 

But just to be difficult--perhaps the "mystery fiction" genre of this piece that you notice, while undoubtedly fun and convincing, is standing in here for a true conclusion.  In other words, I wonder whether the detective genre of the piece--a genre of hermeneutic struggle--celebrates the process of making meaning to the extent that it conceals the fact that not a lot of meaning is being made.

Obviously Doyle is trying to cover up clues.  Obviously bildungs are different from the Gothic.  While I found this article fascinating, I did get a little bit of the sense of the same trouble I see in Moretti's Atlas of the European Novel.  It's different.  It feels fresh.  But when we get right down to it,  a lot of this work seems to me to get to a certain point just on the edge of meaning-making, and then throw up our hands and say "look, isn't this interesting?" rather than fnishing the thought and telling us why it matters and whodunit. 

So I find myself asking, is this an example of the goal of textual analysis?  Or is it a demonstration of the method which, the suggestion is, might lead to the real discoveries later down the road? 

Along with this, I also was fascinated by the footnote on p. 21 which briefly brings up the distinction between doing a "close reading" and reading textual analysis: "we were not reading Middlemarch, but a series which--as such--did not exist in the text, but was entirely an artifact of our methodology...." I know they do address this idea a bit elsewhere, but I really find myself wishing that this wasn't a footnote.  This seems to me to be at the crux of the question of whether and where this work fits in with other types of meaning-making. 


So I find myself asking, is this an example of the goal of textual analysis?  Or is it a demonstration of the method which, the suggestion is, might lead to the real discoveries later down the road?

It's definitely a demonstration of methodology. You start with something to which you already know the answer, apply new methods to that thing, and if the methods give you a bunch of data that confirm what you already knew, then you know the methods are working. Then you can apply the methods to a thing you don't know about at all and still trust your results. 

Now, there's more to it than just methods-proving, in my opinion. Hopefully the data provide information regarding the mechanisms behind or granular details about what it is you already knew. In other words, hopefully the data explain in more detail how the thing you already knew about actually works. (Everyone already knew that apples fell to the ground if they came off a tree, too.)

Of course, sometimes the methods deliver results that call into question what you thought you knew or give you a completely new perspective on it. (I think the first chapter in Graphs, Maps, Trees does this pretty well, e. g., when it places into a more distant context the various 'rises' or 'falls' of genres, female authors, etc.) But when this happens, you have to be doubly careful about what the methods are or are not showing you, before you abandon your priors. (I think Moretti's first chapter is quite careful, and more persuasive because of it.) 


I totally agree on that footnote, on page 21. The recurring complaint I hear in discussions of this piece (and much quant DH work in general) is that what is being done is traditional close-reading, with quantitative analysis used as a discovery mechanism. I'm not convinced that the insights of the article would change if all the graphs were inadvertently printed with different numbers.  

That footnote suggests one way out: quantitative analysis allows close reading to be performed on a diffferent object. But is that sufficient? When we close-read, aren't we also close-reading a vesion of the text that is entirely an artifact of our methodology?


I haven't read Moretti's /Atlas/. I'll have to check it out!