Though it would be interesting to examine the language of identical twins for possible exceptions, I venture to say that no two individuals write exactly the same way, using the same words in the same combinations, or with the same patterns of spelling and punctuation. No two adults in the same family (or corporation or motorcycle gang) have read the same books. No one writes consistently fluent sentences. It is that pattern of difference in each writer's use of language, and the repetition of distinguishing traits, that make it possible for a text analyst to discover the authorship of anonymous, pseudonymous, or forged documents.

Police detectives and literary scholars study different but analogous kinds of evidence. There is the "external" evidence, including personal testimony. External evidence on a bomb or murder weapon may include blood type, fingerprints, or traces of DNA. The corresponding evidence to be gleaned from anonymous writing may include such indicators as place of publication, or postmarks and mailing address. If there is an abundance of external evidence, text analysis may not be required at all. But most homicides and many anonymous documents are the work of people who do not wish to be recognized and who have, in fact, taken pains to efface or to falsify the external evidence.

"Internal evidence" is more difficult for the unknown felon or poet to expunge or conceal, often because the culprit is unable to perceive the difference between his own and someone else's work product. Even when leaving no fingerprints or an explicit signature, bomb builders leave traces of their identity on every device, not just identifiable tool marks, but the type and arrangement of initiator, trigger, fuel, oxidizer, wiring, shrapnel, or packaging, all of which may distinguish one pipe bomb from another, one builder from another -- just as the evidence of watermarked paper or typewriter font in an anonymous document may be augmented by matters of page-formatting, punctuation, or vocabulary. In a bomb investigation, the study of component materials may establish where the offender acquired his physical materials. Corresponding study may lead the textual investigator to the sources from which an unknown writer has gathered ideas and phraseology. More broadly, there is the matter of style, the distinctive way in which individuals select and assemble materials to construct either bombs or pieces of writing.

The police or media investigations for which my assistance is sought typically involve one or more Questioned Documents (QDs), texts of unknown provenance or authorship. I look at language, not penmanship. Too often I've seen myself mentioned as a "handwriting" expert, which I'm not, or as a "computer expert," which is a joke. It has also been reported that I have a crystal-ball computer program: you just feed an anonymous, pseudonymous, or forged text into Don Foster's amazing computer, along with some known writing samples by identified authors, and the software spits out a name: "William Shakespeare," "Ted Kaczynski," "Joe Klein." I wish it were that easy. I don't miss much, but the work is time-consuming, and the time spent is sometimes wasted, producing no more than a "maybe," or no attribution at all.

I have learned a lot in the past few years about criminological applications for linguistic analysis and how to present the evidence to investigative agencies or to the courts. While forensic experts like Henry Lee or Roy Hazelwood look at weapons and bodies, I consider words and punctuation. They do the bullet wounds and bite marks, I do indentation and split infinitives. They do body parts and DNA, I handle the ABCs and parts of speech. Text analysis, whether performed for professional literary studies or for the FBI, is a labor-intensive and stressful business. Apprehensive, I sometimes feel as if the work I've done bears comparison with that of an agent whose job it is to dismantle explosives: if I should get it right, my reward is a sigh of relief. If I make a mistake, however slight, I'd better duck, fast.

Copyright © 2000 Don Foster