Thursday, 15 October 2009

The latest verdict on using brain imaging for lie detection

Excitable tabloids, technophile lawyers and gullible entrepreneurs have all spent the last few years salivating over the prospect of functional brain imaging delivering us the first form of truly scientific, objective lie detection. Not so fast.

Most research that's tested the potential of functional brain scanning for lie detection has compared brain activity between lying and honest conditions by averaging signals across whole groups of participants - no use for real life. Now George Monteleone and colleauges have taken a representative paper from this literature and thoroughly examined its potential for spotting individual liars.

The paper they examine was by Lhan Phan and colleagues in 2005 and involved fourteen participants having their brains scanned whilst they either told the truth or lied about playing cards in their possession. Consistent with several other similar papers, Phan's study showed differential activity in a raft of brain areas when people lied versus told the truth, especially frontal regions involved in working memory and deliberate effort.

Monteleone's team took the brain activity of each individual in Phan's study and compared it with the averaged activity of the other 13 participants to see if the "lying areas" identified at the group level were also extra active when that specific participant was lying.

At the group level, 16 brain regions showed differential activity when lying compared with telling the truth. The brain area that most resembled a true "neural signature" for lying was the medial prefrontal cortex (mPFC). Seventy-one per cent of participants showed heightened activity in this region when they were lying compared with telling the truth. This is better than chance, but far from perfect - really no different from the classic polygraph.

Also, just like the polygraph, brain imaging suffers from the problem of balancing specificity with sensitivity. For example, if the threshold for significant mPFC activity is lowered, then the number of participants showing notable lying-related activity in this region increases, but so too do the number of false alarms - that is, participants who show activity in this region when they're telling the truth. In real life legal settings, these "false positives" could mean innocent people going to jail or worse.

What's more, Monteleone's team warn that it's highly unlikely mPFC activity is a true neural signature for lying. Just as there are many reasons why our pulse might race and our palms get sweaty (thus triggering a polygraph), there are many potential excitors of mPFC activity, including self-consciousness and thinking about other people's mental states.

This also raises the problem of cunning criminals devising simple ways to foil the brain scanner. A participant who performed complex mental arithmetic during truth and lying conditions, or who concentrated on the examiner's mental state throughout a scan, would likely spoil any neat comparison of truth and lying conditions.

The problems don't end there. Monteleone's group further showed that for some lying participants, specific brain regions that appeared to be activated by lying were in fact really part of a far larger spread of brain activation that probably had nothing to do with lying at all. There's also the fact that the playing card lying paradigm is so simple and insipid compared with real-life lying. Also, the researchers observed that a minority of participants showed idiosyncratic brain responses to lying, out of keeping with the general group-level patterns. And finally, there are socio-cultural issues. Problems with language and the cultural appropriateness of deception could both massively distort a person's brain response to lying versus truth-telling.

"...[A]lthough fMRI may permit investigation of the neural correlates of lying," the researchers said, "at the moment it does not appear to provide a very accurate marker of lying that can be generalised across individuals or even perhaps across types of lies by the same individuals."

ResearchBlogging.orgMonteleone, G., Phan, K., Nusbaum, H., Fitzgerald, D., Irick, J., Fienberg, S., & Cacioppo, J. (2009). Detection of deception using fMRI: Better than chance, but well below perfection. Social Neuroscience, 4 (6), 528-538 DOI: 10.1080/17470910801903530

Link to related Wired news story: Evidence from fMRI lie-detection was used in a courtroom for the first time earlier this year.

Post written by Christian Jarrett (@psych_writer) for the BPS Research Digest.

Bookmark and Share

1 comment:

Neuroskeptic said...

Good piece. The mPFC is one of the most researched parts of the brain and it usually rears its head whenever anything to do with mood or emotion is being studied. So if lying activates the mPFC this probably means nothing more than that lying can be stressful.

My favourite "deception-detection" technology at the moment is the EEG P300-based one. No idea if it really works, but the theory is pretty solid and there has been quite a bit of research...

Post a Comment

Note: only a member of this blog may post a comment.