Perhaps, but as yet we do not have the technology to read it. This is the holy grail of law enforcement, counter terrorism, and intelligence – the ability to detect with absolute accuracy whether or not a subject is lying or telling the truth. It stands to reason that the process (on a neurological level) of fabricating a lie would be different than remembering an actual event. So why shouldn’t we be able to detect whatever differences there are?

This question has now gone from theoretical to very real with the conviction in India of a woman of murder based upon a controversial new technology called Brain Electrical Oscillations Signature test (BEOS) developed by Indian neuroscientist Champadi Raman Mukundan.

Theoretical Background

It is not impossible based upon our current knowledge of neuroscience that there would be detectable difference in brain activity between lying and telling the truth, or recalling a true memory from a fabricated one. There is already a great deal of research looking at this very question. Much of the recent research uses fMRI, which measure blood flow to the brain and from that can infer the relative activity of the different parts of the brain. Using fMRI for this kind of research is doable – but very tricky. The awake brain is constantly in a state of chaotic mental activity and it can be challenging to tease out the relevant activity from the tangential or incidental.

There are also many sub-questions in this research. Do false memories look different from true memories, and if so, how? Can a false memory over time look more and more like a true memory? Can people actually convince themselves of their own lies, and how would that be reflected in brain function? What is the actual neurological process of lying?

We are moving in on many of these questions, but the research is not mature enough, in my opinion, for applications that are reliable enough to be used in court. It is not yet even clear what the limits of this approach will be. At this point we have a technological limitation in our ability to image brain function and our knowledge of how that brain function relates to lying. The question is, once our technology and knowledge advances in this area, what are the inherent limits of the specificity of brain function? What is the theoretical maximum sensitivity and specificity of any brain test for lying? Even if we figure out how to detect lying in brain activity, perhaps we can never can better than 90% accuracy because of inherent variability in brain function and other confounding factors. Also, maybe people will learn how to beat the test, the way good lying can beat a current lie detector test (which measures anxiety levels as a marker for lying).

I think we will develop in the future a lie detector based upon brain scanning (fMRI or something similar) but it remains to be see how reliable such a device can be.

BEOS

This technology, which uses EEG to measure the activity of various brain regions, is much less sensitive than fMRI scanning. It is measuring the electrical activity generated by the brain, rather than blood flow. The primary limitation on sensitivity is the skull – which attenuates the electrical activity measured from the brain. Only large groups of neurons firing together can be detected. Computers can analyze the electrical activity and make pretty color maps of the relative activity of various brain regions, but this technique is not sensitive enough to look as specific pieces of the brain. It looks more at regions of the brain. Also, it is not as good as fMRI at looking at deep structures – the electrical activity of deep structures are not measured well with scalp EEG.

So I doubt that scalp EEG will ever be sensitive enough to be used as a reliable lie detector. I could be wrong – someone might find a way to extract more detailed information, cleverly using some physical property of the electrical fields to infer greater information. But the state of the art now is simply too crude for precise applications like lie detecting, in my opinion.

Even with that concern aside – the real question is, does it work? In order to determine this we would need independent analysis of the technology in a blinded situation. We would need to compare the predictions of this technology to some gold standard, or see if technicians can detect scans of those known to be lying from those known to be telling the truth. There is no such analysis published in any major peer-reviewed journal.

To say the least, it is premature to use this technology to determine the fate of a human being.

Lie Detection in Court

The Indian courts are increasingly relying on this type of technology as a “magic bullet” to determine guilt or innocence.   This recent case is the first time a judge specifically referenced this type of evidence as the reason for a guilty verdict. This is breaking new legal ground.

At this point this is the equivalent of using pseudoscience in the courtroom. This is as irresponsible as basing a verdict on the ramblings of a psychic – except that it comes with the trappings of science and legitimacy.

I hope the Indian courts reconsider their use of this technology, and put it on hold until proper and independent studies can be done. If it really works, then great – the rest of the world can benefit as well. My suspicion is that at present we will get something like a lie detector – it may be 60% or so accurate, but not reliable.  It is doing something, which may convince some people or even court systems that it is useful, but in the end is more susceptible to bias and interpretation than objective information.