Voiceprinting is used to identify someone solely on the acoustic properties of their speech and, like the name suggests, is said to be as indelible as fingerprinting. But are the techniques behind it as sound as many proponents claim?

Have you ever watched CSI or one of its numerous clones? Remember that scene where X’s speech was recorded, analysed by a computer, and within 5 minutes identified as being, oh I don’t know, a Columbian drug lord, who is promptly arrested. Woo hoo, the heroes win again and another crime is solved in record time.

First, let me say I love CSI, I really do. As I’m sure a modern, flabbier Julius Caesar would say: “I came, I saw, and I watched week after week”. From a writer’s viewpoint, I give them credit for coming up with good storylines that consistently deliver high viewing figures and consistently cause me to grip my sofa’s armrest until my hand goes numb.

That being said, I think it’s necessary when nearly every aspect of our lives is being recorded on CCTV cameras or over the phone that we trust that, when it comes down to it, fact will overall fiction. Our voices could never match that of a wanted criminal, right? Well…

Voiceprinting, the “technique” (and I use that term loosely) displayed on CSI, has been widely discredited by linguists and the scientific community as whole. Simply, it involves comparing spectrograms (visual representations of an acoustic signal) and saying that, for example, the recording you know to be Osama Bin Laden matches the new, unknown recording you think could be him. Pretty easy job for Homeland Security, huh?

The truth of the matter is that such a strong conclusion isn’t possible from looking at two spectrograms side by side. Anyone could do this and it is a lazy and dangerous approach. It’s like trying to guess the murderer in a Murder, She Wrote episode – don’t bother, there’s no logic involved! Spectrograms from the same speaker vary considerably; they contain speech material on top of any other kind of noise that falls within the relevant frequency range. On top of that, the voice is not like DNA or fingerprints. Joe Blow with fifty beers in him will sound very different without them. Our voices vary according to our emotions, stress, physiology, the time of day and even, like ol’ Joe, whether we were drunk the night before.

So, due to the ‘CSI effect’ and voiceprinting, forensic phoneticians and forensic scientists face some tough obstacles in court rooms, where they are often faced with unfair expectations as well as probably some very sleepy and easily distracted jurors.

To date, voiceprint evidence is still admissible in certain U.S. states (I’m looking at you, Alaska) and the FBI maintains a voiceprint unit. But why is this practice allowed to continue when it has been widely discredited by linguists and the scientific community at large? Whatever the case, now you know, if you commit a crime in Alaska and your voice is under scrutiny, don’t worry, you have a 50/50 chance of getting off scot free! Oh, America.