Navigation – Plan du site
Autumn symposium

The Literary Form

Antoine Compagnon
p. 50

Texte intégral

1Wondering if there was anything new that could be said about literary forms, styles and genres, I was pleasantly surprised to notice that automated word processing, with its iterative review and algorithmic method, suggests the possibility of statistically surveying literary styles and genres.

2Digital Humanities, which are poorly developed in France, could contribute to the analysis of literary forms and the renewal of the theory of genres.

3As is often the case, it was while seeking to resolve another question that the implications of this method for the observation of genres became apparent. Initially, statistical stylistics were used to re-examine the old problem of Elizabethan theatre, particularly the collaboration between authors and the attribution of Shakespeare's plays to Shakespeare himself or to another playwright, Fletcher for example. Now, digital humanities have shown that, not only are they able to tackle certain questions relating to attribution, they can also raise new ones concerning genres. The pioneering work on the subject has come from John Burrows and Hugh Craig at the Centre for Literary and Linguistic Studies of the University of Newcastle, Australia. They have applied statistical stylistics to the study of attributions, that is to say, using linguistic forms as indicators for the attribution of texts.

4The novelty of this approach consisted in letting the variables – signs or distinctive features – emerge from the texts through an unsupervised statistical analysis. Thus, the most frequently used words were simply counted and the resulting frequencies used as variables. As the most frequently used words in a text are function words, grammatical words, the relative frequency of the most common words will be used to compare texts and put their attribution to the test. According to a strong – and largely counterintuitive – hypothesis, the hallmark of a text resides not in its lexical, full or rare words, its hapax legomenon­ for example, but in its function words, its determiners (articles, possessives, demonstratives), its pronouns, its coordinators and its subordinators. Syntactic terms, as opposed to semantic ones, would therefore be those that best allow for the attribution of a text, that is to say, for identifying an author, a particular style. Trivial as they are, these function words form the insignificant outline of the text, a network of pegs that we do not pay attention to when we read and focus on ­meaning, yet the machine counts them. Drawing on another image, individual style comes from the binder, from the linguistic glue that linear reading moves through without stopping when it is in search of the singular, the original, perhaps even the unique.

5Hugh Craig (2004), applying this method to the body of plays attributed to Shakespeare in the 1623 First Folio, has been able to find some of the generic differences between the plays categorized by the editors as tragedies, comedies, historical plays or late romances, as well as chronological differences between the early, middle and late plays. He thus used statistical stylistics to categorize a body of 25 plays (out of the 38 in the First Folio) by counting their 12 most common words.

6This method was taken further by Jonathan Hope and Michael Whitmore (2010). Broadening the analysis to include all of Shakespeare's plays, they confirmed that clear and stable lin­guistic differences exist between the comedies and the historical plays as categorized by the publishers of the First Folio. They thus confirmed a strong correlation between the linguistic statistics results and genre divisions suggested by the first publishers.

7These critics were also interested in the outliers, the exceptions and the anomalies that statisticians tend to overlook but that are relevant to literary analysis. In this way, statistics confirm an old remark according to which Othello is built upon a “comic matrix” which heightens emotion: the author introduced verbal schemes that are typical of comedy in tragedy, producing a sense of irony.

Haut de page

Pour citer cet article

Référence électronique

Antoine Compagnon, « The Literary Form », La lettre du Collège de France [En ligne], 7 | 2015, mis en ligne le 02 novembre 2015, consulté le 20 janvier 2019. URL :

Haut de page


Antoine Compagnon

Modern and Contemporary French Literature: History, Criticism, Theory

Articles du même auteur

Haut de page

Droits d’auteur

© Collège de France

Haut de page