Every time someone writes a memo, a note or an e-mail, the person leaves behind something akin to a written fingerprint.

It is not so much the words, the spelling errors or the grammatical mistakes that make writing unique, however. It is the way the words are strung together - the syntax - that give writing a signature almost as individual as DNA.

Carole Chaski, a linguist and criminal justice instructor at Delaware Technical & Community College in Georgetown, is using writing samples to help link people to crimes.

Her linguistic work is at the center of new forensic research being developed in a world that is increasingly paperless and devoid of handwriting because of the use of computers.

Chaski has appeared on national television news shows in recent weeks, and an article in The Washington Post quoted the instructor's thoughts on the meaning of the Washington-area sniper's writings. An online Wall Street Journal column scoffed at her, using quotations marks to set off Georgetown, Del., to question her standing as an expert. In addition to being a Delaware Tech instructor, she runs a consulting business and has helped police agencies around the country with investigations.

The flurry of sniper-related publicity, however, has been just part of what the forensic linguist does. For The Washington Post, Chaski said she simply did a quick read of some of the messages authorities said were left behind for police.

But in a real investigation, Chaski takes sentences apart to see how individuals use nouns and verbs, adverbs and prepositions. In a very basic way, what Chaski does is somewhat similar to elementary school students standing at a blackboard and separating the parts of speech in a sentence.

As teachers tell those students, every sentence has to have a subject or a noun, and a predicate or a verb. Even with simple sentences, there are a lot of ways to combine the words and say the same thing. Because most people don't write sentences that consist merely of a noun and a verb, Chaski's work is complex.

She has developed a computer program to do most of the analysis of sentences. The program takes apart the sentence and finds each part of speech in relation to the rest of the words. Then she counts the patterns and uses a statistical analysis to determine whether someone wrote something.

"I was trained as just a regular syntactical linguist," Chaski said. "I was just a regular linguistics professor at North Carolina State University."

But her career changed when she was asked to help police on a perplexing case about North Carolina State student Michael Hunter, who died in April 1992 from an injection of lidocaine, Benadryl and Vistaril. One of his two roommates reported the death. All signs pointed to a suicide, but Raleigh, N.C., police Detective W. Allison Blackman wasn't so certain.

There were suicide notes, written on a computer and printed out by Hunter's roommate, Joseph Mannino, authorities said.

"I said there's got to be some way to figure out who wrote these notes," Blackman recalled.

The notes had been printed out on a university computer from a disk. Blackman began contacting universities and eventually found Chaski. Chaski said she looked at the pattern and placement of words in the sentences and counted the patterns.

"She was about 99.9 percent sure" that Hunter had not written the suicide notes, he said.

Her work helped police identify Mannino as a suspect. Weeks away from becoming a doctor, he was convicted of involuntary manslaughter and sentenced to seven years in prison. He claimed he had given Hunter antihistamines to treat a migraine.

After working with Blackman, Chaski quickly realized that linguistics and word patterns could play a role in police investigations. She wondered whether a syntax review of writing would work every time.

Then Chaski was called in on a case involving solicitation of murder, and the word patterns again pointed to a suspect.

She applied for a fellowship at the National Institute of Justice, where she began a scientific study of written syntax and developed the computer software that she uses.

To be accepted as evidence in a court case, a system such as Chaski's has to be based in science and must produce reliable results consistently.

She started to pull together a database of writing samples. She asked four women for writing samples on 10 topics, then looked at the word patterns and punctuation. She then asked for a blind sample, and was quickly able to determine which of the women wrote it.

Chaski said she thinks writing is an instinctive process, which is why it tends to be so individualized.

"Isn't it amazing that that's how we work?" she said. "It's something about the way we process language, but it is the minor, tiny things that are different because we understand each other. ... Like DNA, we share 98 percent of our DNA patterns. Our differences are tiny and not something that is easy to find."

Language has about seven basic units, and they are combined in predictable combinations, she said.

The differences are what make each person's writing unique.

"This really gave me something I could do with linguistics," she said.

Nonetheless, Chaski said, her method is not perfect. It works only if there is a limited group of suspects.

Since the Hunter case, Chaski has worked on civil patent cases, an attempted homicide in Florida and a rape case in Washington, D.C.

"My method requires about 200 words, and it's much better if you can get more," she said. "I would like to eventually expand it to do voice. There are really no validation studies of individual voices. ... It's easier to disguise and it's less automatic than writing."

Reach Molly Murray at 856-7372 or mmurray@delawareonline.com.

http://www.delawareonline.com/newsjournal/local/2002/11/16forensiclinguis.html