About Novel Observer

In my dissertation, I made a claim that needed evidence. Namely, I asserted that John Dos Passos’s work U.S.A. was especially full of geographical data—placenames, to be precise. This feels true when one reads the work, but why not try to prove it?

My solution was to pull down from my shelf a handful of U.S. novels written around the same time period and do a simple random1 sample for each novel. I picked 20 pages (I think) at random and counted up how many placenames were on each page. Next the computer found the sample mean and the standard deviation. Since I was dealing with a small sample, I used the t-distribution to calculate for 95% confidence.

What this app does is let anyone do the same thing to count whatever it is that is interesting to the person, be it placenames, allusions, pronouns, whatever. The process is not automated, because the idea is that by sampling, the amount of data entry is not terribly high, and the results are just about OK.

I know that, personally, as I add to this database, I will be finding more and more “geographically rich” novels that I could subsequently teach or talk about in some way. Furthermore, I'll be finding what amount of toponymic data is “normal.”

Finally, by graphing the uncertainty (if however awkwardly for the time being), I hope to emphasize Johanna Drucker’s point that humanists deal not with data, but, rather, with capta. These counts are initial attempts to make use of “graphical strategies [that] express interpreted knowledge, situated and partial, rather than complete.”2

  1. This is a computer, so “random” never truly is.

  2. J. Drucker, Graphesis: Visual Forms of Knowledge Production (Cambridge, MA: Harvard UP, 2014), 132.