Simulated Humanist Mind | Zipf’s Law, Style, and the Literary – Conclusions


Zipf’s Law, Style, and the Literary – Conclusions

Posted by Craig on May 15th, 2015

(This is Part 6, the final part of my series on Zipf’s Law. For Part 5 see here.  I had not planned this conclusion to be so heavily delayed — exams intervened).

My previous posts have argued for the usefulness of a certain parameter — a — gleaned from power-law fitting a text in determining the stylistic concerns of a text.  The question I want to address here is simple: where does this leave the phenomenal experience of reading?  To what degree do certain stylistic experiences (for example, recognizing something as a sample of Puritan plain style, or as “Stein-esque”) supervene on a word distribution deviant from a norm? Though my answer to this question is ambivalent, I believe this exercise has been useful in diagnosing the normal way critics approach texts.

Generally, critics are taught to work backwards from the phenomenal experience of reading a text, tracing the effects the texts produce back to their sources in internal contradictions or the historical grounds of their production.  In a sense, this acts as a recapitulation of the Freudian dream form (perhaps delivered through the vector of Frederic Jameson’s use of Lacan to bolster his historical-symptomatic). By treating the ‘straightforward’ experience of a text as an epiphenomenon of some “lower level” causal factor, for example, the smoothing over of an ideological contradiction, we commit ourselves to a particular conception of the reading mind.  On this account, the phenomenological aspect of reading is not only produced from below — crucially we also tend to believe that given the right method (again, think of Marx) we can trace the provenance of this phenomenological experience back to its source.

This idea of mind has come into question in recent practice. Surface reading, for example, argues that critics should focus mostly on the latent properties of texts rather than the pathology that produces them.  More relevant to the results of this paper, new positions in the philosophy of mind offer potential for the symptomatic reading of texts while de-emphasizing the role of phenomenology. The most extreme of these positions comes in the form of Paul and Patricia Churchland’s “eliminative materialism.”  As Ray Brassier succinctly notes, the Churchlands argue that “we are not as we experience ourselves to be.” 1  While a philosopher like Thomas Nagel might search of an objective phenomenology that connects the neuronal existence of the self (the bat’s brain) with the first person experience of being that neuronal cluster (the bat’s experience) eliminative materialism argues that no such transfer can exist. Phenomenal experience is indexical at best — it can indicate that something is happening on a lower level, but analysis of that experience won’t tell us what.

In effect, accepting this account would mean that a measure like a could give insight into the lower level production of a stylistic experience more reliably than phenomenal experience.  It would be an arbiter — “Oh you think that piece is Stein-esque? Well have I got a value to show you…”

Before anyone assumes I accept this account wholesale, let me go on.

The problem with a measure like a is it does not, or only poorly, captures the semantics of the works it analyzes. 2  Lest we forget that everything old is new again, let  us consult Jonathan Culler on the subject of Roman Jakobson’s poetics.  Culler notes that in Jakobson’s reading of a Shakespeare sonnet, his emphasis on the position of certain words overburdens his analysis.  Culler points out that in the poem “Position does play a role, but not in the way that Jakobson implies; it is subordinated to thematic considerations.” 3 Culler’s point, generalized, is that concerns of a higher level (semantic, thematic) also help comprise the experience of a certain reading. For a simple example of this point, “Heavenly father” might denote the same thing in a Puritan text and a Dickinson poem, but their connotations likely vary wildly.  These variations are not easily captured by something like a, they are contingent not only on the shared semantic notions of the period but also the transformations the author wishes to effect on those notions. 4


The question is how much of each of these factors accounts for style.  The answer is unknown.  But I can conclude this — perhaps approaching such questions from below might open questions the top-down style takes for granted.



  1. Ray Brassier, Nihil Unbound: Enlightenment and Extinction. New York: Palgrave-Macmillan, 2007. 49.
  2. It might actually capture some — co-occurrences of unusual words across a whole text might show that this text is working its hardest to pervert normal semantic concerns, but it won’t get the whole story
  3. Jonathan Culler, Structuralist Poetics: Structuralism, Linguistics and the Study of Literature. Ithaca: Cornell University Press, 1975. 73.
  4. Perhaps this is why I find some of Franco Moretti et. al’s work extremely interesting and some not so much.  An essay like this one that uses a collection of 7000 titles to hypothesize that the change in long summary titles to short ones is the consequence of marketing (“we’ll name this one “The True Coquette” — they think they know what a coquette is but they only really will when they read our book”) is quite satisfying.  Something like his work on clues in detective fiction seems (at least to me) less so. The reason is embedded in the methodology of both works — the titles essay “only” has to reconstruct one period of historical semantics.  The work on clues must also do that, but if we wish to respect the idea that even genre literature tends to reshape the semantics it inherits into its own image, we would have to restore the context of each of those works.  A more difficult task.  I believe Moretti  implies that he thinks most books stick to the semantic conventions of the time — this I doubt. A better approach in my opinion would be to take formal structures that must not vary from a certain structure to be considered valid (perhaps the syllogism) an go from there. Perhaps distant reading is not one thing.

Write a Comment

Your email address will not be published. Required fields are marked *