Lorna Fadden—Language Detectives II (EAC-BC meeting)

After speaking at well-attended EAC-BC meeting in 2012, forensic linguist Lorna Fadden returned to the stage last week for a highly anticipated follow-up. “I hope I don’t disappoint you,” she said. “You know when a sequel comes out, and it sucks?”

With an opening like that, Fadden had no cause for concern.

Fadden lectures in the department of linguistics at SFU, where she studies sociolinguistics and discourse analysis, as well as First Nations languages. She also runs a consulting practice in forensic discourse analysis, examining language evidence for investigations or trials in criminal and civil cases. These cases may involve hate speech, defamation, bribery, internet luring, plagiarism, and extortion, among other types of language-related crimes. She analyzes both linguistic form—grammatical structure, word choice, and prosodics—and linguistic function—meaning, social context, and pragmatics.

Fadden presented a historical case in which forensic linguistics played a starring role: in 1989, the Exxon Valdez, captained by Joseph Hazelwood, struck Prince William Sound’s Bligh Reef and spilled its crude oil cargo, resulting in one of the worst environmental disasters in history. There was wide speculation that Hazelwood was intoxicated at the time, and forensic linguists analyzed his recorded exchanges with the Coast Guard to find evidence that he was impaired.

Alcohol depresses the central nervous system, Fadden explained, impairing coordination, reflexes, and nerve transmission—basically everything you use when you talk. It also impedes your ability to recall words, as well as your ability to utter words in the correct sequence. Intoxication leads to misarticulation of certain speech segments: r and l sounds can become blended, and s and ts sounds can be palatalized to become sh. Suprasegmental effects of intoxication include slower speech, lower mean pitch, a wider pitch range, vowel lengthening, and the lengthening of consonants in unstressed syllables.

According to the Coast Guard’s recordings, Hazelwood’s speech had all of these characteristics 1 hour before, immediately after, and 1 hour after the Exxon Valdez ran aground but was normal 33 hours before and 9 hours after the accident.

At the time, forensic linguistics as a social science was relatively new. Hazelwood’s trial was the first time this kind of evidence was used in court, but because no witnesses could remember seeing Hazelwood drink and the jury may have been uncomfortable with this means of demonstrating drunkenness, he was acquitted.

Fadden then told us about some of her cases, one of which involved a series of menacing and highly critical letters being sent to a large company’s board of directors. These letters were sent anonymously, but the writer claimed to be a member of the company’s front-line staff or a mid-level manager. The language in the letters, accusing the directors of having “zero business acumen” and referring the company’s “value proposition,” as well as referring to “our managers”—unlikely for a low-ranking staff member to do—betrayed the writer’s higher rank. With Fadden’s help, the investigation uncovered that the writer was a high-ranking executive who’d been fired, and he was sent a cease-and-desist letter.

In another case, the mother in a custody dispute received a series of letters, supposedly from her kids, telling her they wanted nothing to do with her. Fadden’s role was to determine whether the children genuinely wrote the letters themselves. One letter, in her 7-year-old’s handwriting, mentioned that the kids did not “fully trust” their mother and agreed that they would spend time with her only on supervised visits. “Kids that age don’t use adverbs like ‘fully,’” said Fadden, and she doesn’t believe that kids have the meta-awareness implied by the letter. Occasionally we write something addressed to one person, knowing it will have a larger audience. In this case, the letters were written in a style that suggested the writer realized that others—lawyers, psychologists, and so on—may read them. Fadden’s analysis, along with a social worker’s assessment and psychologists’ assessments, led to a favourable outcome for the mother, who’d been accused of nefarious things that hadn’t been proven. “You have to be careful asking kids questions, because the questions we ask them often already suggest the answers,” said Fadden. “We rarely ask children information-seeking questions.”

Fadden’s third case was a more complex one: a woman had accused a man of drugging and sexually assaulting her, but eyewitness accounts, video surveillance, and toxicology suggested that her allegation was false. She faced a charge of public mischief, but she claimed she didn’t understand what happened during the police interview. Fadden had to assess whether she was legally competent by comparing her linguistic performance with what we’d expect from a native speaker in the same context. In her doctoral dissertation, Fadden had characterized a series of police interviews of first-time suspects, so she had a robust set of measures as benchmarks.

Cognitive deficiency is correlated with a slow speech rate, but the suspect had a relatively high speech rate, and it didn’t drop significantly from the beginning of the interview to the end (so not much of a fatigue effect). Fadden also looked at her turn latency (how much time elapses between the end of the interviewer’s question and her answer) and her pause ratio (how much she pauses compared with how much she speaks). All of these temporal elements were within normal ranges; nothing suggested that she was incompetent.

A stronger indicator of the suspect’s competence was in the way she manipulated specificity associated with details. Take, for instance, “this talk,” from most generic to most specific:

  • Fadden’s giving a talk on Wednesday (type identifiable, not specific)
  • There’s this talk on forensic linguistics on Wednesday (referential)
  • The talk on forensic linguistics on forensic linguistics will be on Wednesday (uniquely identifiable)
  • This/that talk on forensic linguistics is on Wednesday (familiar)
  • I’ll be at that on Wednesday (activated)
  • It’s on Wednesday (in focus—you don’t even have to name it)

(Fadden made sure we noted the distinction in specificity between “this talk” and “this talk.”) Through 2 hours of interviews, the suspect was adept at adjusting the level of specificity based on context, using generic language to describe what she claimed to have witnessed when she allegedly found herself in an unfamiliar environment but specific language when talking about details that the police officer had told her. Fadden concluded that she had normal cognitive status. The suspect eventually confessed to fabricating the story because she didn’t want her husband to find out she’d willingly slept with another man.

To end the evening Fadden challenged us to an exercise of authorship analysis. She gave us two writing samples from different blogs with similar topics and writing styles. We had to figure out who’d authored a third sample. From a superficial reading, most people in the room guessed that the first blogger was responsible, but Fadden showed that by comparing features like

  • the number of words per sentence,
  • the length of words,
  • the use of adjectives and adverbs,
  • the use of parentheticals,
  • the use of discourse markers,
  • the use of conjoined phrases, and
  • the use of independent clauses,

her analysis showed that the second blogger was the likely author. Authorship analysis is a contentious field now because its effectiveness and accuracy aren’t completely understood, and there’s no standard method for carrying it out. As a result, it’s not admissible in court. But, like a polygraph, authorship analysis may help steer the direction of an investigation.

Crimes involving words: some recent cases

Dr. Lorna Fadden captivated the audience at Wednesday evening’s November EAC-BC meeting with her fascinating talk about forensic linguistics. Dr. Fadden is an assistant professor of linguistics at SFU and also runs a consultancy as a forensic linguist, analyzing language evidence for law enforcement and legal counsel in criminal or civil cases.

Examples of crimes or offences involving language include extortion, ransom, solicitation, harassment, Internet luring, threats, coercion, perjury, hate speech, defamation, bribery, and plagiarism, and language evidence can turn up from a spectrum of sources, from emails, text messages, and letters to police interviews and emergency phone calls. Forensic linguists may analyze linguistic form (e.g., grammatical structure and word choice) or linguistic function (e.g., social context).

Modern forensic linguistics, Fadden explained, has its origins in the Evans Statements. In 1949 Timothy Evans of London was accused of murdering his wife and daughter. He was tried, convicted, hanged, then posthumously pardoned—an inquiry found that a neighbour had killed Evans’s family—and this miscarriage of justice caused the UK to throw out its death penalty. In 1968, Jan Svartvik, a Swedish professor of English, reviewed Evans’s statements, which were allegedly a verbatim transcription of what he had said. Evans had the vocabulary of a fourteen-year-old, but Svartvik found portions of his statements that showed the grammar and word choice of someone with a much higher level of education. As a result, Svartvik concluded that the police had made up portions of the confession, and the case put forensic linguistics on the map.

Fadden then took us through four of her cases.

Case 1

Fadden was asked to study a transcript of a 911 call, in which the caller gave an elaborate backstory before, five or six sentences in, telling the operator the reason for the call—that he saw a man with a gun. Most calls to 911, she said, tend to follow a script:

  • the operator takes the call,
  • the caller identifies the problem (“My wife is choking!”) or makes a request (“Send an ambulance!”) in the first few words, and
  • the operator solicits details as needed.

She added that criminals who call in their own crimes often go off script, because to them, there is no emergency and no urgency.

Fadden concluded that the transcript she analyzed didn’t meet our expectations for what generally happens in a 911 call, and although she had her theories as to why, it wasn’t up to her to make that determination.

Case 2

Fadden’s second case involved a young child being interviewed by family services for an investigation into allegations that his father sexually assaulted him. The lawyer who hired Fadden wanted to know if there was any evidence of coaching. In cases of coaching, you might see vocabulary or sentence structures that look out of place; children the age of the alleged victim tend to use common nouns and verbs and fewer adjectives and adverbs; they tend to stick to neutral terms or higher-frequency words:

  • look rather than leer or ogle
  • touch rather than fondle or grope
  • creepy, funny rather than lewd, salacious

Fadden concluded that the child’s vocabulary in this case didn’t support the theory of coaching, but she continued her analysis of the interview by coding the questions by type:

  • assertions (to be accepted or rejected)—e.g., Your dad touched your penis?
  • high-specificity questions—e.g., Did your dad touch your penis?
  • closed alternative questions—e.g., Did your dad touch your penis or your bum?
  • open alternative questions—e.g., Did your dad touch your penis or your bum or something else?
  • low-specificity questions—e.g., What did your dad do?
  • wide open questions—e.g., What can you tell me?

Fadden studied the alternative questions closely, because in those cases, details might not be generated by the witness but by the person asking the questions. She found a consistent pattern in the way the witness answered those kinds of questions (e.g., always selecting the first option in open alternative questions) and that they did not generate any new information. She questioned the credibility of those answers, contrasting them with portions of the interview in which the witness supplied completely unprompted details.

Case 3

Fadden analyzed correspondence from a man, the owner of a small business, who had met a CEO of a large firm at a public event and began sending him a series of letters. The CEO had never solicited or responded to the correspondence, but it kept coming, and it contained unsettling language. Did the letters constitute stalking?

Stalking letters, Fadden explained, have particular hallmarks:

  • expressing frustration with unrequited feelings or being ignored or overlooked
  • berating the target for the target’s transgressions
  • an offer of forgiveness for those transgressions
  • allusions to a relationship that doesn’t exist

Even though the correspondence in this case didn’t involve any kind of romantic angle, it nevertheless had all of the characteristics of stalking and was legitimate grounds for a cease and desist order.

Case 4

This final case has concluded, and so Fadden was at liberty to share details with us. On March 19, 2009, Justine Winter had a fight with her boyfriend. After dropping him off, she started to drive home, all the while exchanging text messages with him in which she threatened to kill herself. She crashed her car, and although she survived, a thirty-year-old woman and her thirteen-year-old son were killed. The prosecutors hired Fadden to determine whether these text messages constituted a suicide note; their assertion was that Justine’s death wish was a criminal act because two others died as she carried it out.

A suicide note, Fadden said, tends to have certain characteristics:

  • saying good-bye
  • claiming that life is too difficult to go on
  • expressing a desire to end suffering
  • imploring others to go on
  • expressing remorse or regret
  • expressing revenge
  • apologizing

Although Winter’s text messages had some of these characteristics, they also showed features—such as bargaining—that you wouldn’t typically find. “People who write suicide notes are way past bargaining,” said Fadden. As a result, Fadden concluded that the messages weren’t a suicide note.

But how credible was Winter’s threat to crash? To be a credible threat, Fadden explained,

  • it must be communicated
  • the subject must be motivated
  • the subject must have the means by which to carry it out

Analyzing the text messages, Fadden asserted that Winter’s threat to hurt herself were credible. Winter was convicted and is now appealing her sentence.

***

Fadden’s talk was thoroughly engaging and entertaining, and she’s clearly passionate about her work. What I found interesting was that she was very clear about where her role as a forensic linguist begins and ends. Her job is to assess the language; she can identify if something doesn’t fit the script, but it’s up to psychologists and other professionals to discover why.

Dr. Fadden will be moderating a Philosophers’ Café session, “Is language changing for the better or worse?” on December 5, 2012, at 7pm, at the McGill Branch of the Burnaby Public Library.