Caught red-handed – the truth about lie detectors

PHOTO/renaissancechambara
PHOTO/renaissancechambara

This September, Gary Smith was charged with involuntary manslaughter of fellow Army Ranger and roommate Michael McQueen in 2006. According to the jury, he was guilty; according to the latest brain scanning technology, he was innocent. Which do you trust more?

The technology under question is fMRI (functional magnetic resonance imaging), which is capable of measuring the extent of activity across different regions of the brain. It works by monitoring blood oxygen levels, which increase as brain cells become more active. This has led companies such as No Lie MRI to claim that fMRI can be used as a lie detector, given that certain areas of the brain appear more active when people lie, in comparison to when they are being honest.

In 2009, Sujeeta Bhaat (a research scientist at the US Defence Intelligence Agency) put this claim to the test by scanning the brains of participants whilst they performed a fake police line-up task. Interestingly, there were specific differences in brain activity when participants correctly claimed to recognise faces they had seen before and when they lied; for example, an area of the brain known to be important in visual imagery – the precuneus of the parietal lobe – was more active when participants lied.

No Lie MRI and similar businesses claim that this technology has many applications beyond testing the accuracy of eyewitness testimony during line-ups. For example, criminals accused of assault are presented with photos of the weapon used in the crime, as well as photos of other, irrelevant weapons, and asked to press a button if the weapon presented is unfamiliar. This is a very similar task to that used in a study conducted by Giorgio Ganis (Associate Professor of cognitive neuroscience at Plymouth University). Rather than weapons, Ganis showed participants their own birthdates intermixed with unfamiliar dates on a screen. When their birthday was shown to them, the participant was required to deny that the date was of any relevance.

Ganis was concerned that cunning criminals could easily fool fMRI technology by using a simple strategy – to subtly raise a different toe or finger each time they told the truth but to keep still when lying. The study showed that by introducing these extra motions into the brain scans of honest responses, the uniqueness of the brain scan depicting the neural response to the lie was reduced. This had the effect of drastically lowering the accuracy of Ganis’ lie detector from 100% to 33%.

Even disregarding the possibility of criminals cheating the technology, there is a huge hurdle in deciding where to place the cut-off point for brain activity indicative of lying; in other words, in assigning an all-or-nothing, categorical label (lying or not) to a continuous variable (neural activity). For example, George Monteleone (a graduate at the University of Chicago) scrutinised the results of a 2005 study which found that 71% of participants exhibited greater activity in an area at the very front of the brain – the medial pre-frontal cortex – when lying. Whilst lowering the cut-off point for “greater activity” increased the proportion of participants who showed heightened brain activity when lying, it also increased the proportion that showed increased activity when telling the truth!

So fMRI lie detectors cannot yet outwit the complexity of the brains they are tasked with analysing. And perhaps they never will – because the brain cannot afford to assign the job of lying to an exclusive neural space. Also, perhaps you will be glad to know that Gary Smith was found guilty. In fact, the judge never allowed the jury access to the fMRI evidence which exonerated Smith, claiming that there was not enough expert consensus regarding its reliability.