From cold cases to criminal trials: the power of forensic linguistics

In 1967, Swedish linguist Jan Svartvik authored “The Evans Statements: A Case for Forensic Linguistics“. In it, he illustrated the different grammatical styles between four incriminating statements by Timothy Evans, supposedly dictated to police by the illiterate Evans, which were used to convict and execute the accused for murdering his wife and baby daughter. It was later found that Evans’ landlord, John Christie, had committed the double murder. Despite Svartvik’s pioneering of forensic linguistics as a discipline, it remained a curiosity rather than a burgeoning aspect of the policing and legal spheres.

In 1995, the Unabomber case was cracked as a result of authorship analysis in the United States. An unidentified person had been sending letter bombs between 1978 and 1995, resulting in the deaths of three people and injuring at least 24. No fingerprints or DNA traces were evident, and the crimes continued over 17 years. It was one of the longest and most expensive criminal investigations in US history. In 1995, the so-called Unabomber told police he would pause his attacks on the condition that a newspaper publish his 35,000-word manifesto.

It duly appeared in the Washington Post, the New York Times and Penthouse, leading to a number of people coming forward claiming they recognised the writing style and voice – not least, the brother of the perpetrator. Sociolinguist Roger Shuy had been studying the bomber’s letters, finding patterns in the language. Shuy supposed the perpetrator was from Chicago, based on peculiarities in spelling “wilfully” for “willfully”, and “clew” for “clue”.  Theodore Kaczynski’s family provided letters and writings that enabled FBI linguist James Fitzgerald to write a ‘Comparative Authorship Project’ affidavit that was used to obtain a search warrant for Kaczynski’s Montana cabin, where investigators found copies of the manifesto and homemade bombs.

Forensic linguists were employed by the New York Times to investigate the founders of conspiracy theory QAnon. Florian Cafiero and Jean-Baptiste Camps, together with OrphAnalytics used computational linguistics to analyse the social media posts of suspects. Paul Furber, a South African software engineer, and Ron Watkins, an American conspiracy theorist were named in a report published in 2022. While the technology they relied on was emerging in complexity, the rapid advances in AI technology – including large language models (LLMs) – enable faster, more detailed analysis.

The International Association of Forensic Linguists was founded in 1993. It holds a biennial conference, including events in Australia, and since 1994, has published The International Journal of Speech, Language and the Law. Increasingly, universities have courses dedicated to forensic linguistics, though it remains primarily an area dominated by English-speaking experts.

What is forensic linguistics?

There are three central areas of focus for forensic linguistics:

The language of written legal texts: discerning a layperson’s understanding of the sometimes outdated vocabulary, complicated grammar and infrequent punctuation, which typifies many legal texts, in order to make practical meaning of it.

The spoken language of the legal process: studying the conduct of police interviews with suspects and factors such as guiding language, insinuations and tone, the boundaries upon expression based on rules in courts of law, the limitations upon vulnerable witnesses to convey testimony, and the implications of seeking evidence from those who do not speak the language of the court (including concerns with accurate translations)

Provision of expert witness: while criminal law is the most obvious and cited example of forensic linguistics, it is broadly applicable. Experts might provide evidence towards the differentiation between rival trademarks (including in several cases in which McDonald’s has sued other businesses for the use of “Mc”, with mixed results), or the truthful authorship of documents, insight on the meaning of words and expressions, and evidence on how language and grammar can point towards the place of origin of asylum seekers.

Australia’s recent history of forensic linguistics

Professor Helen Fraser is the Director of the Research Hub for Language in Forensic Evidence at the University of Melbourne. She has looked at a broad range of issues at the intersection of language and the law, including the factors affecting the accuracy of police transcripts.

The role of forensic linguistics in Australia traces back to the first expert linguistic evidence in Adelaide, presented in 1959. Arrernte man and eventually the Central Land Council chairman, Rupert Max Stuart was tried for murder in the Supreme Court, and convicted by a jury’s verdict. Stuart was illiterate and spoke pidgin Arrernte-English, and yet the six policemen who had interrogated him insisted that a typed confession was Stuart’s “literal and exact confession, word for word”. The Lecturer in Linguistics at the University of Adelaide, T.G.H Strehlow gave evidence on linguistic variations in Aboriginal languages in the Northern Territory at a time when Aboriginal rights were still an emerging focus of the law. Strehlow’s evidence was that the confession could not be genuine.

More recently, linguistics evidence has indicated that an Aboriginal man accused in the criminal court could not have used the style of speech attributed to him in interrogation transcripts. That early case was followed by greater interest and evolution in the 1970s and 1980s, and then a more formal recognition and definition in the 1990s. As Fraser and co-authors Georgina Heydon and Diana Eades identified in their 2023 publication “Forensic Linguistics in Australia”, justice is benefited where there is a collaborative, cooperative engagement between forensic linguistic professionals, the judiciary, and legal professionals.

Fraser says, “Forensic linguistics covers a large multitude of different topics, and there are two different branches. We call one of these ‘language and the law’, or legal linguistics, which is studying how language is used in the courtroom itself. The other is for forensic linguistics in the sense of forensic science, where we’re analysing language captured in some kind of recording to be provided as an exhibit for the jury to look at.”

Fraser’s work has investigated the problems for justice created by use of inaccurate transcripts of indistinct audio evidence, and the fact that evaluation of transcript accuracy is left as a matter for non-experts.

The science of analysing captured language involves identifying speakers, transcribing indistinct words, and therefore enabling the court to understand the content of a recording.

She says, “The big problem is that the courts often allow police officers or other non-experts to give that kind of evidence, and sometimes linguists are only brought in as a secondary opinion.”

Fraser explains one of the major reasons for her research into this area is because of concerns about injustice caused by allowing police officers to be responsible for the transcripts of police interviews or covert recordings.

“The court actually allows police officers to provide transcripts on the understanding that it’s easy enough for lawyers, judges and juries to determine if a transcript is assisting them or misleading them, and my research has shown that in fact, that is very difficult evaluating a transcript,” she says.

“Evaluating a transcript just by checking it against the audio is extremely prone to error. And in fact, even if the listener rejects a transcript, and says, ‘No, I don’t think that’s right’, you can demonstrate that they’re still influenced by it. They still hear the recording differently after exposure to an inaccurate transcript.”

The problem is the assumption a conscientious listener will notice errors when checking a transcript against the audio and then be able to correct them.

“Funnily enough, that’s not true,” says Fraser. “Most people would assume that if a careful, responsible listener checks a transcript against the audio, they’ll notice and be able to correct errors. That seems like common knowledge, but that’s not true when you have indistinct audio. You are primed to hear in line with the transcript, even when it’s wrong.”

To that end, Fraser’s team at the University of Melbourne are working to bring about law reform that would require the use of forensic transcription of recorded audio as standard in court.

“The point is that if the audio is indistinct, [deciphering it is] presently considered to be a matter of common knowledge, where [forensic] expertise is not needed. Either side can bring in an expert to offer an opinion, but it’s left up to the jury and to the judge to determine,” she says.

“What we’re working towards is having some law reform that gets forensic transcription seen as scientific evidence, so that it’s up to the discipline to provide evidence-based, accountable methods to ensure that the jury is only, and always, assisted by reliable transcript.”

Like blood samples, or other biological evidence sent to an agency for scientific evaluation, Fraser envisions that indistinct audio – whether covert recordings, interviews, or otherwise – would be sent to an accredited agency for reliable transcription as a matter of course.

“Lawyers typically have very high-level skills in using and understanding language. They typically know a lot about word meanings, etymologies, grammar, and all those kinds of topics,” she says.

“So, this gives a sense of confidence, but [forensic] linguistics as a science is very unusual in that most of its findings contradict common knowledge.”

Examples of this are abundant on Professor Fraser’s website, which is an excellent resource to more deeply understand the value of forensic linguistics within the legal system: