Things you can('t) tell from a criminal's voiceDate Released: Sun, 3 July 2011 12:34 +0200
On TV series like 24 and Law and Order, it's not unusual to see detectives whip a recording of a suspected criminal's voice over to the sound lab, and half an hour later get the message “we have a match”, confirming that the suspect is guilty. In a keynote address on the final day of Interactions and Interfaces, a conference about language held at Rhodes University, forensic linguist Paul Foulkes stressed the fact that it is not always so easy to analyse voice recordings, but evidence from linguistic experts can still be extremely useful in criminal investigations.
Foulkes is from the University of York in the United Kingdom and also works as a consultant for JP French Associates, one of the world's best forensic speech laboratories. His address was entitled “Individual variability in English pronunciation: Applications in the forensic domain”.
Some of the things that make forensic linguists' work tricky, according to Foulkes, include background noises on voice recordings (such as speech over traffic noises), problems with telephone transmissions, multilingual interactions and strong accents. One possible solution to these problems, he pointed out, is to use high-quality listening devices and software. Forensic linguists also use certain specialised linguistic analysis techniques to figure out the syntax that the speaker is using, work out what people are talking about from the context, and analyse the sound waves that make up the recording.
There are a few snippets of information that forensic linguists can guess at using these techniques, such as the regional background of the speaker(s), their social and educational background, whether they are reading aloud or speaking spontaneously. However, some characteristics they can't tell from a voice recording are things such as the speaker's height, weight, build, psychological state and the sincerity of their words.
With criminal cases specifically, the pronunciation of vowels and consonants, intonation, voice quality and rhythm are analysed to try to identify the speakers on recordings. Foulkes stressed throughout his address that even with all this thorough analysis, evidence from a voice recording is never one hundred percent scientifically accurate enough to convict someone, because the evidence could still be wrong. Unlike a fingerprint, there is no such concept as a ‘voiceprint’ that is unique to a particular person (as Hollywood would lead us to believe), because voices are flexible and can be changed at will, or can be affected by such things as having a blocked nose. Foulkes says that forensic voice analyses will always be based on impressionistic judgements, which again are not enough to convict someone in a court of law on their own, but can add to other existing evidence in a case.
Foulkes ended off his talk by looking positively towards future research in the rapidly expanding field of forensic voice analysis. He also urged fellow linguists to publish more research in their own fields which could help forensic linguists to give evidence that helps to crack crimes.
Article by Babongile Zulu
Photo by Ju-ann Hockly