Tuesday, December 30, 2014

When experts stray out of their field

This end-of-year post has nothing to do with networks, or even phylogenetics, although the general principle involved might apply to both. My point here is simply that experts sometime look foolish when they commentate on fields outside their own area of expertise.

As an introductory example, I remember reading a paper in a physics journal that tried to convince the readers that humans could potentially live forever. Unfortunately, the authors confused the concepts of lifespan and longevity, which is pretty basic stuff in population biology. Lifespan is the length of time for which humans normally live. We have more than doubled this over the past millenium, due to changes in sanitation, medication, surgery and safety. Longevity is the length of time for which humans are capable of living. We have not changed this by even one year, as it seems to be related to phenomena like programmed cell death. Changes in lifespan do not therefore entail changes in longevity; all that has happened is that our expected lifespan is now closer to our observed longevity than it previously has been.

More recently, an electrical engineer drifted into the field of literature while claiming to be a scientist — Mikhail Simkin (2013) Scientific evaluation of Charles Dickens. Journal of Quantitative Linguistics 20: 68-73. Sadly, his article displays neither of the characteristics of science (replication and control), nor does it appear to contribute anything much to literature.

As noted on his web page, the author had trouble publishing this article, and he has subsequently received "a flood of criticism", which he naively seems to believe he has rebutted at the Significance blog.

His intention was a simple one: a comparison of the writing style of Charles Dickens and that of Edward Bulwer (later known as Edward Bulwer-Lytton). His premise was: "Edward Bulwer-Lytton is the worst writer in history of letters ... In contrast, Charles Dickens is one of the best writers ever." He put online a quiz with "a dozen representative literary passages, written either by Bulwer-Lytton or by Dickens." The takers had to nominate the author of each quote. Simkin discovered that on average the votes were "about 50%, which is on the level of random guessing. This suggests that the quality of Dickens's prose is the same as that of Bulwer-Lytton." The results are shown in the graph above.

Simkin's intention seems to have been to demonstrate that currently revered and non-revered authors do not differ much in style, which is a contention that I see no reason to disagree with, but if so he has gone about showing this in a remarkably unscientific manner.

Let us take the premise first, for which the author provides no personal justification nor any reference to a published one. It seems patently true that the current fashion is for Dickens to be widely read but Bulwer not. This on its own means little, however, as even the Shakespearean works have had a century or so of being out of fashion, although not in the past couple of hundred years (to the dismay of anyone who has had an English-language education).

Was Bulwer a bad writer? Well, first, the results of Simkin's poll imply "no", at least in comparison to Dickens. But more importantly, many other sources say "no", as well. Indeed, Wikipedia makes a strong case both for his popularity in his own time, and for considerable influence on literature since then. Indeed, he is so 'obscure' that towns as far apart as Canada and Australia are named after him. His works are so 'poorly known' that we continue to use his expressions "pursuit of the almighty dollar" and "the pen is mightier than the sword". His works have been so 'derided' that several operas are based on his books, including one by Richard Wagner; and authors such as Edgar Allan Poe have paraphrased his words. His books are such 'poor examples' of English that people have felt compelled to translate them into Serbian, German, Russian, Norwegian, Swedish, French, Finnish, Spanish and Japanese, among other languages.

Clearly, the premise that Bulwer represents the nadir of English-language literature holds no water. He is currently obscure, but as John S. Moore has noted, the fact that he is not read does not mean that he is not worth reading.

Indeed, a scientist would immediately note the lack of replication here. Why are "best" and "worst" writers not replicated in the experiment? This would immediately address any possible mis-judgements about potential literary worth. It is repeated patterns that provide convincing evidence in science, not isolated pairwise comparisons. This poll is hardly a "scientific evaluation", as claimed by the author.

Now let us consider the experimental procedure. This consisted of choosing "representative literary passages", without any explanation for how this was done or what were the criteria for choice. Clearly, this choice is the key to the experiment. After all, all the experiment does is show that one can find passages by both Dickens and Bulwer that are hard to distinguish. That could very well be true of almost any pair of writers from the same culture (ie. country and century). The experimental comparison has thus not been controlled, as it would be in science.

What would experimental control look like in this case? Clearly, the issue is one of style, since authors vary their writing style depending on the book, the plot situation, and even the character involved. (One of Bulwer's passages is actually taken from the dialog of one of his characters, which hardly represents the author's own writing style!) The objective, then, must be to find passages that represent the range of styles present in the corpus of each writer. One might try grouping the passages into topics or styles, for example, or whether they describe actions or locations, etc.

Without either replication or control, this literary evaluation cannot be considered to be scientific. Sadly, on his website Simkin has several other so-called scientific comparisons within the arts, designed in exactly the same inadequate way.

As a final note, we can ask why was Bulwer chosen for this comparison in the first place? The choice seems to be almost solely due to various extant parodies of the opening of one of his books, Paul Clifford (1830): "It was a dark and stormy night; the rain fell in torrents ..." For example, this was chosen by Charles Schultz in his Peanuts cartoons, as the opening of one of Snoopy's failed attempts to be a world-reknowned author. The full sentence does not actually seem bad, although it tries to cram a bit much information into the number of words available. Thomas Hardy later tried the same thing, but with more success, in The Return of the Native (1878): "A Saturday afternoon in November was approaching the time of twilight ..."

However, the award for sheer bravado surely goes to D.H. Lawrence, in his short story Tickets, Please! (1919), which starts with a paragraph consisting of a sentence of 118 words, followed by sentences of 15 words, 27 words and finally 113 words.** A plethora of commas, colons, semi-colons and dashes are needed to keep the meaning coherent in this page-long paragraph. You and I could not get away this, which is why Lawrence is considered to be one of the great English literary stylists. Apparently, Bulwer did not get away with it, either.

** My count is based on the original publication in The Strand magazine, which is slightly different to subsequent versions.

No comments:

Post a Comment