STATS ARTICLES 2012
Are Brain Scans Better Than Lie Detector Tests?
Rebecca Goldin, Ph.D., August 31, 2012
Should courtrooms admit fMRI scans as proof of innocence - or are we being outwitted by technology?
The Washington Post reported on the latest effort to convince the justice system that functional magnetic resonance imaging (fMRI) scans can read someone’s mind. Specifically, a man accused of murder claims he is innocent. His defense team wants to include an MRI brain scan (an fMRI) as part of the evidence of his innocence.
How can an fMRI scan suggest innocence? Simple, the defense team and The Washington Post say: the defendant speaks some known lies while in the scanner, then he speaks some known truths while in the scanner, and then he states his innocence in the scanner. The brain patterns exhibited while he states his innocence are close to the brain scans during his lies.
To read the Washington Post story, this is how “science is put to the test.” The article consists of a series of documented scientific studies in which an MRI scan predicted lies and truths with a fair amount of accuracy. The Post certainly reports doubts about the validity of using scans to demonstrate guilt or innocence, but the paper failed to explain why there are good scientific reasons for those doubts
The correlation between a defendant’s brain pattern while he is saying he is innocent could look very similar to his brain pattern while he is telling a known truth, and very different from his brain pattern while he is telling a known lie, and yet still the defendant could be guilty. Why?
Perhaps comfort causes the brain patterns observed. People telling the truth are comfortable with it, and people telling lies are uncomfortable with what they are saying… except that a murderer who has been in prison for four years could feel comfortable with the lie that he is innocent. He could have been saying it for many years, and it may no loner feel foreign or wrong to him. Alternatively, as with lie detector tests, it might be possible to beat the system by distorting the controls – in other words, thinking of lies while being asked to tell the known truths could change what “truth” looks like in the machine.
Lie detector tests – which measure bodily functions such as sweat or blood pressure – are notoriously inadequate to predict who lies and who tells the truth. Nervous people telling the truth look like liars, and comfortable, confident liars look like they tell the truth.
The Washington Post reported, “accuracy rates for picking out the deceptions topped 90 percent in some cases.” Yet the accuracy rates all involved controlled experiments in which subject involved had nothing particularly to gain by successfully lying or “gaming” the system, and the subject had not been practicing the lie or “holding onto” the lie for more than a minute or two.
The Washington Post didn’t make the obvious connection to the long history and discussion over the use of lie detector tests. While lie detector tests are notoriously unreliable, and considered little better than chance by the American Psychological Association, the American Medical Association, and the National Academy of Sciences, people advocating the use of lie detector tests also see an extremely high success rate: 82-99% according to a meta-analysis carried out by the American Polygraph Association. In other words, fMRI claims are similar to lie detector claims. The comparison sheds a little more light on the question of whether fMRI results could be useful in the courts.
Brain patterns are not uniquely matched to specific “realities”. There are some correlations, for sure, but many possible thought patterns could cause a particular brain pattern, and no one knows how or why.
In the case of Gary Smith, the defense team wanted to argue that the brain patterns that occurred when he answered questions like, “Did you kill Michael McQueen?” were dissimilar to brain patterns when he was told to lie in responding to “Were you ever deployed in Iraq?” But could the dissimilarity result from any one of the other myriad differences between the responses, if he really did kill McQueen? Perhaps stress would result from lying about killing McQueen but not from lying when told to lie about his deployment in Iraq.
As STATS noted in its investigation into the reliability of fMRI, one has to be very careful to distinguish between cause and correlation. The cause behind the specific brain patterns cannot be discerned from the patterns themselves, and we have yet to understand why certain facts or emotions correlate with them. The fact that the courts agreed with the testimony of scientists saying the science is too young suggested a more compelling story. By way of contrast with the Post’s coverage, the Maryland based Gazette.net presents a careful story in which many scientific doubts are aired, including both the possibility of tricking the system and the comparison with lie detector tests. The Washington Post seems to have missed the science in the science.
Rebecca Goldin is the Director of Research of STATS at George Mason University.