“Bookish Math: Statistical tests are unraveling knotty literary mysteries” (SN: 12/20&27/03, p. 392: Bookish Math) skipped one of the most significant methods for analyzing text for authorship. On March 11, 1887, Thomas Corwin Mendenhall reported in Science a straightforward method of plotting word length versus frequency. The beauty of this method is that it doesn’t depend upon the exercise of judgment on the part of the investigator.
Temple City, Calif.
The article immediately reminded me of the Unabomber’s manifesto, and how people familiar with Ted Kaczynski’s prior writings may have seen clues in the manifesto pointing to him.
Downers Grove, Ill.
Thanks for demonstrating that literary stylometry has progressed far beyond intuition and type-token ratios. In my work on forensic author identification, I have been able to distinguish authors of documents with as few as 200 words by focusing on syntactic rather than lexical units. It may be natural for the literati to focus on vocabulary, but psychological principles of language processing also support the focus on syntactic units in forensic and information retrieval applications.
Statistical analyses were applied many years ago to biblical texts in attempts to establish the origins of various books. It would be interesting indeed to see if the new computerized methods would support that earlier work. Scholarship such as this would be much more useful than the Bible-code nonsense.
Virgil H. Soule
It would be gratifying to use text analysis to distinguish fiction from nonfiction, subjective writing from objective writing, true from false, science from pseudoscience. Or would it be horrifying?
John W. Mathewson
Mercer Island, Wash.
The article should also have considered the possibility of cumulative sum analysis, a technique admissible as evidence in British courts.
Also, the authorship of Royal Book of Oz has been ascribed to Thompson for decades. It is not really disputed, as your article implies.
James D. Keeline
San Diego, Calif.