Logo IMG
HOME > PAST ISSUE > July-August 2014 > Article Detail


Belles lettres Meets Big Data

Quantitative analysis of poetry and prose has roots deep in the 19th century.

Brian Hayes

Shakespeare’s Spectrum

Mendenhall published nothing more on word-length studies until 1901. Writing then in Popular Science, he sheepishly admitted he’d had an ulterior motive all along: To show that Francis Bacon wrote the plays of Shakespeare. For work on this famously vexed and vexing question, Mendenhall was able to secure funding. Augustus Hemenway, a Boston philanthropist (and Baconian partisan) agreed to pay the salaries of two Worcester women hired as letter counters, as well as the cost of building a special tabulating machine. The nature of the machine is not explained in detail, but it had a button for each possible number of letters in a word.

One of the counters, with book in hand, called off “five,” “two,” “three,” etc., as rapidly as possible, counting the letters in each word carefully and taking the words in their consecutive order, the other registering, as called, by pressing the proper buttons.

2014-07CompsciFp263bot.jpgClick to Enlarge ImageThe two Worcester counters tallied 400,000 words of Shakespeare, 200,000 words of Bacon, and works of various other Elizabethan authors. They soon made a curious discovery: Whereas the word spectrum of most authors writing in English has its highest peak at words of three letters, Shakespeare exhibits an exceptional fondness for four-letter words. A computer survey of the complete works of Shakespeare confirms this observation. (See illustration at right.)

A second discovery must have been disappointing to Hemenway and Mendenhall: Bacon’s word-length spectrum looks quite different from Shakespeare’s, with a more conventional peak at three letters. Perhaps there was some consolation in seeing the spectral method itself vindicated, in that the two authors are clearly distinguished by their word-length curves.

comments powered by Disqus


Subscribe to American Scientist