Forensic Linguists Use Spelling Mistakes to Help Convict Criminals
We talked to one of them about how a text message can give away a murderer.
This article originally appeared on VICE UK.
Whatever happens, your terrible grammar and eccentric spelling will always come back to screw you over. At school, in birthday cards, in the Guardian or Daily Mail comment boxes: Nowhere is safe when the linguistic nitpickers swoop in, like vultures hungry for their latest, misspelled carcass of a word. But the linguistic nitpickers can also be a force for social good. Because it turns out the way you trip over your words is especially crucial if you're a criminal, with language playing an increasingly vital part in identifying suspects and bolstering law court convictions.
Since the mid to late 1990s, a growing community of "forensic linguists" have been employed to look at textual evidence such as blackmail letters, potentially falsified suicide notes, text messages, and online correspondences in order to spot consistencies and lock down suspects in cases ranging from plagiarism to murder and even terrorism.
Hired by defense solicitors, local police, and national police forces like the Serious Organized Crime Agency and the Independent Police Complaints Commission, academics from the Centre for Forensic Linguistics at Aston University, for example, have provided expert testimony in myriad court cases.
Take Dr. Nicci Macleod, who has specialized in the linguistics behind the power dynamics of police interviews with women reporting rape, as well as how we model our identities linguistically on social networks. Her casework includes the analysis of slang in the diaries of the individuals convicted of the attempted murder of Joss Stone in 2013. I gave Nicci a call last week to ask her about how she uses language to nail convictions.
VICE: What exactly is forensic linguistics?
Nicci Macleod: Forensic linguistics is a term that can be used to apply to any interface between language and the law. It tends to be talked about as having three separate strands. The first one is studying the peculiarities of legal language; part of forensic linguistics is looking at the reasons behind the oddness of legal language and sometimes making efforts to make it more accessible to ordinary people.
The second area is to do with the language of legal processes, so things like what happens in a police interview, for example, or what happens in a courtroom trials, and what's special about those contexts and the way the language works within them.
And the last one, which is probably the most exciting, is the use of linguistic evidence. So, if there's something pertinent to an investigation—like a ransom letter, blackmail letter, or suicide note—whoever is investigating it (it tends to be the police) might want to know something about what kind of person wrote this text: their age, their educational background, their geographical origins, maybe their gender. Or, it might be the case that there's a suspect in mind, so we're looking at texts where we know who the author is and comparing it to a text where we don't necessarily know, commenting on whether there are any similarities there that might indicate that the texts have the same author.
What kind of features of language might you look at when trying to identify the author of a text?
I might look at the way that this author is structuring their sentences or the way they're using particular vocabulary. The register of vocabulary they're using might tell us about that person's background or the area that they work in and so on. Things like punctuation, even the layout, and anything that really stands out about the way this person uses language.
Could you give an example of a case you've worked on?
There was a blackmail case. Two or three elderly neighbors had received blackmail letters saying, "We know what you did and we're going to tell everybody unless you put this money in this particular place, behind a shed." In this case, the police were fairly sure that they had the right person; it was a woman who lived locally and was known to them previously.
So what they did was to collect a lot of writings from her home. They had a search warrant and collected various diaries and letters that she didn't dispute she had written. They were in her handwriting and so on. I was provided with them and the blackmail letters and produced a report based on the consistencies between the two.
There were particular spelling errors that kept popping up. We're not always fortunate enough for the same words or features to actually occur in both, but fortunately in this case, several words did come up in both sets. I was able to say that, based on the misspellings and various other features, the letters were consistent with her style.
How reliable are all methods? Are they usually used to back up some other piece of evidence, or can they be used alone?
We would generally say that they should be taken into consideration in the context of other evidence. In the blackmail case, the only thing I could comment on was the letters that had been received by the victims, comparing them to the writings of the police's chief suspect.
What you need to bear in mind is that, although I said there were a number of consistencies here, they'd also found the typewriter that had been used to type the letters in the suspect's shed. It was wrapped up in a bag that only had her fingerprints on it. Obviously linguists can contribute to convictions, but I would shy away from saying this kind of evidence should be used in isolation in these cases.
I noticed one of your personal areas of research is "modeling online identities." How has forensic linguistics changed in the online world?
We had to work very hard to refine our methods, because traditionally when this idea of authorship analysis first emerged we were working with quite lengthy texts. Obviously with internet developments, texting, instant messaging, and so on, we tend to work with very short texts.
This is why we've moved towards these last couple of projects; there was "modeling online identities," and there was also a project on Twitter. We asked, are we able to actually comment on the authorship of 140 characters? We've actually done quite well—it turns out there is potential for commenting on patterns of consistency with those sort of texts.
Professor Tim Grant
What kind of cases and crimes might this research be useful for? Or has it been useful already?
My colleague Professor Tim Grant worked on a case in Staffordshire. A house had burnt down and there was a husband and wife and two children. The husband managed to rescue his two children, but his wife had apparently perished in the house fire. The police had their suspicions, and there were various things that made them think it was the husband who had murdered his wife and set the fire to cover his tracks.
What they'd established is that text messages had continued to be sent from the wife's phone that entire day, long after they thought she'd actually died. And so what Professor Tim Grant did was to compare the historical texting style of the wife—her name was Amanda Birks—and of her husband, Christopher Birks, and then to look at these messages that were sent on the day of the fire to say, are these texts actually consistent with her usual style of texting or are they actually more consistent with his?
He established that they were far more established with the husband's texting style. And, sure enough, I believe he pleaded guilty to actually having murdered his wife, and he had sent those messages from his wife's phone.
Follow Huw on Twitter.