Because everything we really need to know about law and law enforcement comes from TV, a CBS show, Criminal Minds, picked up on a 2008 Alabama homicide case where Earnest Stokes got nailed because of a peculiar quirk in his writing:

Inspiration . . . came from former FBI special agent and linguist James R. Fitzgerald, who became an adviser to “Criminal Minds” in 2008. Blake, he says, is a combination of him and his fiancee, Georgetown associate linguistics professor Natalie Schilling.

The incident, Fitzgerald says, is based on a 2008 homicide case, State of Alabama v. Earnest Stokes. In a linguistic report he prepared for the prosecution, Fitzgerald said he found the term “light bug” in an anonymous letter attempting to lead investigators off the track (“His [sic] had busted the light bug hanging down”) and in a tape-recording of suspect Earnest Ted Stokes. That was one of the lexical clues leading Fitzgerald to “opine” “with a likelihood bordering on certainty” that Stokes was the author of the unsigned letter.

So the takeaway is to teach your clients the difference between bulb and bug?  Not exactly.

As more of our communication is written, the linguistic fingerprints we leave provide enticing clues for investigators, contributing to the small but influential field of forensic linguistics and its controversial subspecialty, author identification.

The “light bug” quirk will strike most people as fairly obvious and, given  the circumstances, unlikely to have swept up the wrong person.  After all, how many people related to the murder would call a bulb a bug. Unless they had similar spell check problems?

That there is a forensic specialty of linguistics, particularly author identification, may come as a surprise, as it has hardly gotten the historic play of such things as fingerprint or voice analysis.  But then, given how much of our lives happens online now, and that it exists in the written word, the potential impact is huge.

The new whodunit is all about “who wrote it.”

Answering that question becomes ever more urgent as we create a virtual trove of data — in e-mail, in texts and in tweets that are often anonymous or written under pseudonyms. Private companies want to find out which disgruntled employee has been posting bad stuff about the boss online. Police and prosecutors seek help figuring out who wrote a threatening e-mail or whether a suicide note was a forgery.

Textual sleuths find clues not in fingerprints or handwriting, but in word choice, spelling, punctuation, character sequences and in subtle (and usually subconscious) patterns of sentence structure.

Consider Radley Balko’s extensive reporting on the failings of bite mark analysis.  It was good enough to convict people of murder, and yet it was a scientific sham.  But sounding good was good enough for the law.

“There are disputes about the acceptability” of linguistic evidence in court, says Lawrence Solan, a professor of law at Brooklyn Law School who served as president of the 175-member International Association of Forensic Linguists. The sometimes acrimonious controversy, he explains, hinges on the state of the science — whether linguists can provide reliable statistics on their accuracy or “error rate.”

Solan describes an “intellectual and cultural divide” between practitioners such as Fitzgerald who use what he calls an “intuitive” approach, examining among other things idiosyncrasies in spelling and word choice to see whether “constellations of features emerge”; and computer scientists, who perform statistical analyses of such features as character sequences or word length, often by running large amounts of text through software programs. Solan hopes a convergence of those methods might provide sounder science.

Pick your poison, the “intuitive” approach (because experts have magical voodoo powers) or the statistical analysis approach by computer of “character sequences or word length”?  The question isn’t whether these things can work; writers tend to be idiosyncratic. I know I am, at least when I want to be.

But are they reliable enough to put a man in prison?  To put a man to death?

Neither technique is any good, according to Carole Chaski, who founded the Institute for Linguistic Evidence in Georgetown, Del., unless it is based on scientific methodology that is replicable from case to case.

Unless linguists produce verifiable data, says David A. Harris, a professor at the University of Pittsburgh School of Law and a former Maryland public defender, it’s just a matter of time “until they get the wrong person and are proved wrong by DNA.”

Good old scientific method. Actual verifiable data that can be tested and repeated. Except that’s not quite how forensics happens.

“The forensic sciences were born in the police crime lab, not the science lab,” explains Harris, the law professor, and “based largely on human interpretation and experience.” Even the venerable art of fingerprinting is “not really science in the true sense of the word,” he says.

Which is not to say fingerprinting — or linguistics — doesn’t produce valuable evidence.

“Fingerprinting is most often right,” Harris says. But to call it science “cloaks the work in a sort of aura of certainty.”

And yet, law enforcement likes it and, sound or not, linguistic analysis is being used, and its proponents, like Fitzgerald, are making a name and a living at it.  And when wrapped up in jargon, hidden in self-serving studies and approved by organizations whose existence depends on the law’s acceptance of linguistic analysis and whose officials give each other awards and titles like “Expert Linguistic Poobah,” it all comes off as remarkably official.

“Jurors have come to expect the presentation of forensic science in every case, and they expect it to be conclusive.”

And if a judge concludes that the witness is an expert, who is a juror to disagree?