by Megan Taylor Stephens
Dec. 24, 2022
Brian Steven Smith confessed to a 2019 Alaskan murder because of his distinctive South African accent that was recorded in a gruesome video.
Serial killer and pedophile Stanley Everett Rice’s arrest in 1968 was made possible in large part to his severe stutter.
Robert Durst’s block lettering and poor spelling tied him to the 2000 murder of Susan Berman.
A speech recognition expert’s testimony helped convict Nathan Libby, who made a hoax distress call to the US Coast Guard in 2020.
Our communication mannerisms—such as speech, voice, and writing—can give us away. Try as we might, it’s very hard to change something imprinted in us since childhood. You can wear gloves to conceal your fingerprints and you can cut out letters from magazines to write a ransom note, but you can only run so far from your idiosyncratic speech traits and go-to vocabulary words.
With pandemic boredom came my obsession with true crime, mostly podcasts (“Hellooo, fellow Murderinos!”). With the pandemic also came my curiosity about other ways to apply my skills as a speech-language pathologist (SLP) and former English as a second language teacher. Lo and behold, I realized that I am basically qualified to be a forensic linguist. Or, if not qualified, then at least justified to dabble in forensic linguistics.
What is forensic linguistics, you ask? Well, you see, there is a phonetic fingerprint left in audio recordings by criminals. Ransom calls, hoax calls, phone calls to emergency personnel. When run through a spectrogram—which analyzes all the elements of speech including phonemes, speech rate, pitch, and volume—valuable information can be gleaned. Linguists can measure things like the fundamental frequency of a speaker, a measurement of the vibration of the vocal folds that we hear as pitch and tone. Beyond the use of a spectrogram, there are other ways that speech and language can be analyzed and used in court.
Audio evidence, however, is best used as a way to corroborate other forms of evidence or testimony rather than be the sole determiner of guilt or innocence. Experts caution that our voice varies, depending on things like mental state. Two people can sound similar, background noise can degrade the audio, and recordings can be tampered with. For that reason, forensic phonetics techniques and other audio and linguistic evidence should be used with a grain of salt.
That doesn’t dampen my excitement over my new calling as a forensic linguist. This is basically what SLPs like me have been trained in: gathering speech and language samples, mulling over them with a fine-toothed comb, and drawing logical conclusions from the evidence.
What SLPs Know
SLPs, like other applied linguists, have super senses when it comes to speech and language analysis.
Do you pronounce “out and about” closer to “oat and aboat” or “oot and aboot”? That subtle diphthong change can pinpoint Canada as your homeland. (Scottish English can have this pronunciation as well, but there’s nothing subtle about that dialect!)
Do you say “carmel” or “caramel” when you order a caramel frappuccino? If you say “carmel” you’re probably from the western, midwestern, or northern US and if you say “caramel,” you probably grew up in the south, east, or northeast.
Is it hard for you to differentiate between these word pairs when you’re speaking or listening: feet-fit, share-chair, boat-vote, diss-this? You might have English pronunciation that is influenced by a different mother tongue, such as Spanish.
Do you have “glottal fry” when you speak? In other words, is your voice gravelly, creaky sounding, and underpowered, especially at the end of sentences? Well, I’m spying on your frying, and your glottal fry speech profiles you as a youngish white American girl.
In addition to our speech sounds and voice quality, the grammatical patterns we use (aka syntax) are dictated in large part by the subcultures and regions in which we grow up. Case in point, a speaker of African American English and a white Appalachian speaker might say, “Them kids are hers” instead of “Those kids are hers.” The use of “them” as a demonstrative pronoun rather than the object form of a personal pronoun is a perfectly normal use of the dialect.
Some people can code switch effortlessly between dialects or languages, but many people are hard pressed to change how they talk. When someone has a speech or language deficit, it is especially hard to conceal.
Do you have a frontal lisp, where your tongue tip peeks through your front teeth? Do you have a chronically hoarse voice? Is your syntax filled with simple rather than complex constructions? Do you have challenges with word recall and search for the correct vocabulary? Do you have trouble understanding idiomatic expressions and interpreting implicit meanings?
SLPs and other trained linguists can hear it all: speech sound differences, dysfluencies, voice disorders, word-finding problems, social pragmatic issues, developing language proficiency, and more.
This is a critical clarification: a speech and language deficit or difference or any other type of developmental or acquired disability does not equate to pathological or antisocial behavior. There is no causation or correlation between communication characteristics and crime. Communication patterns are simply a unique piece of evidence, like a fingerprint or filament of hair, that can help validate or invalidate someone’s association with a crime.
Four cases involving speech and language expertise
Theodore “Ted” Kaczynski, also known as the Unabomber, was a math professor at UC Berkeley. He was beyond brilliant as a mathematician, as seen by his 1969 Harvard PhD dissertation, where he apparently answered this thesis question in the affirmative: “If A is a given set in C of type Fσδ, and if φ is a function of honorary Baire class ≤2(A, Riemann sphere), does there always exist a continuous function f mapping D into the Riemann sphere such that A is the set of curvilinear convergence of f and φ is a boundary function for f ?” Kaczynski was also an unstable recluse who was eventually charged with killing three and wounding 23.
Over 17 years, from 1978 to 1995, while sending packages containing homemade bombs to universities and airlines through the postal system, Kaczynski scribbled hundreds of tormented letters to his family from a remote cabin in Montana. He also sent in a long manifesto railing against technology that was published in The New York Times and The Washington Post newspapers. An ability to compare Ted’s language in the letters to his family with the Unabomber’s language in the manifesto was a hail Mary for the investigators.
Kaczynski went to great lengths to eliminate evidence from his bombs: making his own glue from deer hooves, stripping the labels off batteries, and ensuring no DNA was on his materials. He may have had an IQ off the charts, and his mathematical reasoning skills suggest he knew all about complex patterns, but he wasn’t smart enough to realize that his language contained telltale patterns as well. It was James Fitzgerald, FBI investigator, who stumbled upon the field of forensic linguistics while cracking the case.
Kaczynski’s tendency to use certain phrases and grammatical structures can be largely credited for his undoing. “Negros, broads, chicks” all pointed to the lexicon of a man of a certain age, race, and generation. The spelling variation of words like “analyse, wilfully, instalment” were familiar Theodorisms. The archaic maxim “you can’t eat your cake and have it too” appeared in his letters. His erudite and unusual word choices showed up regularly in his writing, including words like “tautology” and phrases like “cool-headed logicians.”
No one can argue that Ted was a scholarly man with unusually sharp language skills. He also had extremely rigid thinking and obsessive themes in his writing that suggested psychiatric or neurodevelopmental deficits. An analysis of writing, and the authorship attribution that followed, was only one data point in the criminal investigation, but it was probably the straw that broke Kaczynski’s back.
Brian David Mitchell
In June 2002, Brian David Mitchell and his accomplice spouse, Wanda Barzee, kidnapped 14-year-old Elizabeth Smart from her bed at knifepoint. Elizabeth’s 9-year-old sister, Mary Katherine, pretended to be asleep but overheard Mitchell say to Elizabeth something like, “You better be quiet, and I won’t hurt you.”
Mary Katherine had a fuzzy image of who took away her big sister. She thought he was younger than he actually was (he was 48), she thought he had on light clothing (he wore black), she thought he was wearing a gold hat (he wasn’t), and she thought he had a gun (it was a knife).
Mary Katherine thought about what the kidnapper said. He was calm and soft spoken. She never saw his face clearly, but his voice sounded familiar. Four months later, while flipping through the Guinness Book of World Records that she had in her room, she paused at a picture of a muscular woman and a ‘Harley Davidson’ looking man. Her memory was suddenly triggered and she placed the assailant’s voice: it was Immanuel, a man the family had hired for one day to do work on their house. This voice recognition led to a new drawing from a sketch artist, which was widely dispersed on TV. Relatives of Mitchell saw a similarity in the sketch and were able to release real photos of him, which led to the eventual capture of Mitchell in Sandy, Utah.
It is astounding that a young girl could pinpoint the voice of her sister’s kidnapper after only being around him for one day and only hearing a few words. But no two voices are identical, and that’s why forensic linguistics is so intriguing.
The Delphi Murders
“Guys, down the hill.” Those are the chilling words of the suspected murderer of Liberty German and Abigail Williams that were recorded on German’s smartphone before they were killed in Delphi, Indiana in 2017. Richard Allen is the main suspect in the case. Four words don’t paint a clear picture of a person’s background. But they provide much more information to someone who knows the intricacies of speech and language.
Was the killer someone from Indiana, a Hoosier? In other words, was the killer a speaker of Midland American English, the region the girls were murdered? Midland American English is basically an in-between zone of the northern and southern dialects. There are some southern elements, especially in southern Indiana, and there are some northern elements, especially in northern Indiana. Delphi is in the northern-northwestern area of the map.
Syntax: All we have is a few words to go on, but he comes across as a man of few words. He didn’t say, “You guys need to walk down the hill” or “Walk down the hill, and no one will get hurt.” He chose an incomplete sentence instead of a full one: “Guys, down the hill.” This is a man who might be known as parsimonious in his speech.
Pitch, stress, and intonation: There was slightly higher pitch and rising intonation on “guys” and flat intonation on the rest of his phrase. His speech was sort of clipped. No vowels were elongated, and no words other than “guys” were stressed.
Pronunciation: The speaker’s short-i vowel on “hill” sounds close to the short-e vowel one would find on “hell.” Like the well-known lack of distinction in the southern pronunciation of “pen” versus “pin,” a Midland American English speaker blurs the two vowels. But he did not have a southern drawl on “guys,” which would make it sound more like “gauze.” We don’t have a lot to go on, but we can hear that the speaker could come from the area.
Semantics: “Guys” is commonly used to address more than one boy or girl in the US. That said, it’s used more in the west and midwest than in the south and northeast. We can infer that the speaker might not be from the southern part of Indiana.
My overall impression of the preserved speech is that the speaker could be from the Midwest or thereabouts. Three seconds of staticky audio isn’t enough to draw any conclusions, and my analysis is purely speculative. Comparing the four fateful words to other speech samples by the alleged perpetrator is definitely needed make headway on this case.
Richard Allen’s trial begins March 20, 2023.
Sometimes forensic linguistics can help clear a person’s guilt and establish innocence. In 1999, 18-year-old Antwuan Cubie was found guilty of killing his friend in Chicago. The police had a signed, two-page letter that they said was a word-for-word transcription of Cubie’s confession. The only problem is that it didn’t sound like Cubie’s speech. It sounded an awful lot like the detective’s.
In the letter, there were sentences with the adverb “then” that were unlike Cubie’s past writing samples. “I then told Jeremy to move his jeep to the end of the alley”; “we both then went into the building after ringing Jamie’s bell.” Cubie’s other writing samples had “then” as the first word in a sentence, or as a connector between sentences to establish sequential events. Cubie never used “then” in this stilted way right after stating the subject.
This transcript was supposedly a verbatim account of Cubie’s spoken words, but there were more than a few eyebrow-raising phrases. One such phrase that was not typical of an 18-year-old’s oral recounting of an event: “I met Jeremy at Cass Avenue and 63rd Street in Westmont at an unknown time on Saturday the 1st of June.” Again, “at an unknown time” might be how people write, but it’s not how many people speak, let alone an 18-year-old kid.
Cubie says that after being beaten by the police, he signed a blank sheet of paper and this signature was used by the police to fabricate a false confession. Cubie’s case remains in limbo while he is incarcerated.
There are times when it is quite apparent that the way someone talks during police questioning means they should be given special treatment. For example, an interpreter should be offered without question to a non-native speaker of the mainstream language. Another example is when a suspect or victim has poor receptive or expressive language skills.
Consider the case of Anthony Templet in Louisiana. In 2019 he admitted to killing his father, Burt. It is obvious that seventeen-year-old Anthony had language and literacy deficits, which were not taken into consideration at first by law enforcement. The trouble he had answering questions, retelling events, using age-appropriate vocabulary and syntactic constructions, displaying nonverbal communication, and expressing emotions all should have been telltale signs that he had developmental delays. When asked if he could write, for example, he said something like, “Yes, I can write all the letters of the alphabet.” Um, that’s what kindergartners do. That’s not how 17-year-olds describe their literacy skills.
Rather than being treated like someone with the cognitive skills and developmental age of a younger person, he was considered a stone-blooded adult murderer. It turns out he was kidnapped by his father as a youngster, had no formal schooling, was not given socialization opportunities, and had clear signs of neglect, if not abuse. Anthony eventually pleaded no contest to negligent homicide, which gave him a reduced sentence—five years of supervised probation—compared to the original charge of second degree murder, which could mean life in prison.
If law enforcement and lawyers had on staff a forensic linguist, or even just an experienced SLP, we could share our insights from the start. Instead, I’m over here shaking my head and yelling, “You’ve got to be kidding me!” while watching the story unfold on Netflix.
I guess it’s time for me to get off the couch and offer my language expertise to the legal system. My husband would agree that my skills are under-utilized, as they are only really evident while I binge watch dubious true crime shows and shout out random profanity-laced observations.
SLP Murderinos, who’s with me?
3 thoughts on “A Speech-Language Pathologist’s Crime Investigation Spree”
Murderinos unite! #SSADGM
We should have new letters after our name CCC-SLP-PI
New passion unlocked! Cheering you on.
I don’t know how people watch or listen to true crime, it’s too scary. But this was a very interesting article.