Recently, I was recalling the experience of having an uncle read a Sgt. Rock comic to me when I was no more than 8 years old. The event was insignificant in terms of comics–I had been reading comics for some time by then and I was particularly well-acquainted with war comics, especially the aforementioned Sgt. Rock. What made the experience memorable was the manner in which he read the comic, particularly the way he read the onomatopoeic “rat-tat-tat” above a machine gun nest under siege by Easy Company. He enunciated and raised his voice, snapping the Ts and grinding the Rs that border the crack of the long-A sound in the middle. It was enlightening–I had never really heard the machine guns in comics until that point.

All this to introduce the complexity of treating comics as a text. Obviously, comics are full of images–sometimes entirely–and I’m not trying to diminish that, but our analytical approach to them and the even the way we read them are rooted in the attitude we take to reading words–silently, alone, like mature readers historically read. Indeed, some of the these characteristics of reading are held up as what attracted individuals to comics in the first place, and still dominate comics over-arching thematic tropes: secrecy, individuality, history, to name a few.

When I’m not talking about comics here, I work with digital tools for textual analysis a bit. Software and strategies that help us understand patterns in text. And, given my interest in comics, I am consistently struck by how comics problematize and reveal bias in our approaches to textual critique, particularly as those problems and biases are embedded in our traditional techniques as they are transmuted into new methodologies. By that I mean, we look for patterns whether human or machine, we look for characteristics we recognize whether informed by human experience or machine code.

One of the tools often used to examine texts is Google’s NGram Viewer{:target=“_blank”}. Briefly and reductively, the NGram viewer takes all the books Google has scanned as part of the Google Books project and looks for instances (or mentions) of the words you ask it to find. It then graphs these out so that you can measure how many times the word appears.

As is evidenced by the example on the NGram intro page, the viewer does not take into account context (hello Derrida!) and doesn’t offer anything distinguishable about the words that appear in books–Einstein is not a literary character, Frankenstein is both title and character. At the same time though, it does show us how language might develop as illustrated below:

We can see how usages or terminologies shift over time and how one way of describing things falls out of favour or becomes more commonplace. Knowing this might tell us about cultural trends or literary topics or, if you’re really daring, shifting hegemonies (hello Stuart Hall and Gramsci!):

Getting back to how comics complicate these tools for textual analysis–NGram examines books after all–comics, and the subjects that surround them, are difficult to put into such software. As an exercise in keeping with this thread, I thought about how the language of comics might be brought to bear on the NGram Viewer. One spot that seemed particularly fertile was comics’ emphasis on onomatopoeia. I decided to avoid words such as “bang” or “boom” that could occur in other textual contexts and set an arbitrary date limit of 1950 for some of what’s to come in order to head off words that were in common usage before comics ever existed. I also inserted an exclamation point on two of the words to try and distinguish them from regular usage.

Here’s the result for “rat-tat-tat, pow!, bam!”:

It’s interesting how “rat-tat-tat” tends to spike during was periods (1918 and 1941), where “pow!” and “bam!” tend to uptick around 1950 when comics–particularly superhero and war comics start to enter the mainstream. At the same time, “rat-tat-tat” holds pretty steady.

When I insert a contextual term such as “comic books,” we can see how the spike that is the context jives with the upticks in the onomatopoeia present in the genre. At the same time, it’s risky making any real judgements about these graphs and patterns. There is simply too much about comics signalled by the visual. Is “pow!” a superhero word? a war word? Just a word that appears in lots of texts. In short, there is no way one can avoid thinking about how genre and the assumptions the NGram Viewer reveals–its based on a textual–book–corpus; not a cultural one. Texts are not culture or context when treated this way. Trying to say something about comics using the NGram Viewer is a dead-end as far as textual critique is concerned.

Moving on, I figure I’ll see what comics’ artists get more mentions than others and when–and here I set 1950 as an arbitrary date and avoid listing women because they’re simply not there–this is telling in and of itself.

It’s interesting that the graph follows the trajectory of popularity common to most comics–male–scholarship. More recent, more known, more mentions. Of course, both “Frank Miller” and “Alan Moore” are very common Anglo-Saxon names. We could be looking at mentions of people who are not comics artists at all. “Stan Lee” is interesting, especially that dip in 1993 (Harvey Kurtzman dies, Deadpool debuts, and Conan the Barbarian is cancelled), given its steady rise and then sudden collapse in 2007. Why did we stop talking about Stan Lee in 1993 and 2007?

While this NGram of authors doesn’t do much but confirm what most of us know, it ably shows how bias creeps into analysis, with currency being a key marker of attention–this is how patriarchies and hegemonies keep on top. As if to confirm this fact, look at how the NGram Viewer bears down on "superman, batman, wonder woman, aquaman, and Watchmen:

Obviously, Watchmen comes off oddly, with a steady decline in usage since the 1800s unless we remember that NGrams don’t deal with context or comics. But, the uptick around 1996 is interesting given the publication timeline–smack between the comic and the movie–we can almost see it gain a little traction. Wonder Woman and Batman have a strange spike in 1976 (Saturday morning cartoons maybe?). Superman follows the trajectory of Nietzsche’s √úbermensch, particularly the 1909 publication of the translated title (by Thomas Common). The NGram shows how genres such as comics that fall outside traditional scope are often manipulated by the reception and existence of earlier analysis. It’s hard then to build an unbiased tool for textual analysis or more broadly the analysis of big data.

What the NGram view does when I try to use it to think about comics is show me how the interpretation of comics is informed by textual assumptions. It jars me in the same way that my uncle’s voice jarred my understanding of “rat-tat-tat”: until he read the comic aloud, I thought I understood how “rat-tat-tat” was supposed to sound. But, he made visible the preconceptions and ignorance I brought to reading comics, innocent though they may have been. I think using comics as data for machines, and watching how machines try to cope with comics’ conflation of images and text, comics’ conflation of individuality and commonality, of past and future, is a wonderful way to remind ourselves that we too conflate our assumptions for text and image with our analysis of cultural currents in comics.


Leave a Reply