12/14/2012 10:58 am ET | Updated Feb 13, 2013

Can Neuroscience Predict Human Behavior?

Written with Justin T. Baker, M.D., Ph.D., and Bruce H. Price, M.D.

Turn to the crime section of any major news outlet and you can't miss the case of Army Staff Sgt. Robert Bales, accused of committing the violent massacre of 16 Afghan civilians.

While the case is on hold, awaiting the outcome of his Article 32 hearing (to determine whether the prosecution has enough evidence to proceed to court-martial), statements from his defense team have alluded to possible trial strategies. Perhaps Sgt. Bales sustained a significant traumatic brain injury (TBI) on a previous tour of duty in Iraq? Perhaps he suffers from post-traumatic stress disorder (PTSD)? Perhaps there will be evidence, neuroscientific evidence, at his trial if his defense is one of "diminished capacity"?

The argument might go something like this. Sgt. Bales suffers from a brain injury (TBI) and mental disorder (PTSD); the combination of these conditions rendered him more likely to respond violently and irrationally to the extreme stressors of war. The defense could argue that these conditions produced a predisposition to behavioral dyscontrol that renders him less culpable (blameworthy) than someone whose brain would allow for better self-regulation and self control.

But how can the law distinguish someone who could not restrain himself from someone who did not restrain himself? Neuroprediction -- or the belief that one can predict individual behavior from neuroscientific data -- is a theory on the rise in courtrooms across the United States. Lawyers have introduced brain scans and genetic tests in hundreds of cases in order to assert that a brain abnormality makes an offender less culpable than someone with a "normal" brain. These arguments seem to have fallen on eager ears. Americans have become accustomed to the scientific method, with its reliance on numbers and metrics. Whether it's the sophisticated weather predictions surrounding Superstorm Sandy, or the precise political calculus of Nate Silver, we've learned to put our faith and trust in mathematical models to understand and predict the outcomes of complex natural and social systems. In fact, we may well put our faith in rigorously-tested models that predict the behavior of groups of individuals or events, for example when tracking and managing epidemics. However, the use of neuroprediction for an individual may be dangerously premature.

Let's consider how the defense team might present a TBI diagnosis in the Bales case. Defense lawyers could try to demonstrate a brain injury with Diffusion Tensor Imaging (DTI), a scan that highlights pathologic changes in the large fibre bundles that connect different parts of the brain. Patients with TBI often show imaging evidence of damage to these information superhighways; this finding has been admitted to strengthen the TBI diagnosis in civil proceedings occuring in several states, including New York, Florida, Louisiana, Indiana, Minnesota and New Mexico.

But these brain scans have substantial limitations: Many patients have normal-looking brain scans and still suffer serious symptoms of TBI, while some severely-traumatized brains look unharmed in the scanner and some purportedly milder-appearing injuries cause debilitating symptoms. The presence or absence of visible damage simply does not always translate into a symptom or behavior, in its presence or severity. In the Bales case, there are also press reports of suspected alcohol and steroid use, which, if substantiated, would complicate any predictions of behavior based on the actual behavioral effects of head trauma.

The same limits apply when relying on neuroscience to predict past or future behavior related to PTSD. The dichotomy depicting TBI as a physical injury and PTSD as an emotional one is becoming a historical footnote as the diagnosis of PTSD is going high-tech. Magnetoencephalography (MEG), a technology similar to electroencephalography (EEG), has demonstrated abnormal bursts of neuronal activity in veterans with symptoms of post-traumatic stress. Using sensors that read magnetic activity from the surface of the brain, researchers have detected signature patterns of neuronal miscommunication in PTSD. Such MEG data, showing areas of abnormal cellular activity superimposed on an image of the brain, might convince a skeptical juror (or judge) that PTSD indeed has a biological reality.

So what's the problem? It is that these studies are early research findings, using homogeneous subjects who did not abuse substances. This makes it harder to replicate the findings in a general population of diverse people or use this data (sometimes called a biomarker) in individual patients with complicated histories, including substance abuse. To add to the complexity, PTSD symptoms vary depending on context or circumstances a person happens to be in, and the vagaries of human memory. In short, this data is hardly the equivalent of a statistical predictor of individual behavior.

Will DTI and MEG figure prominently if Sgt. Bales goes to trial? They might legitimately be part of a defense effort to reconstruct his mental state at the time of the alleged offenses. Combined with other historical and medical data, they may help shed light on how Sgt. Bales functions overall. They should not, however, stand alone to explain his past or to predict his future behavior.

In the years to come, neuroscience may evolve to yield solid predictions about how genetics and brain conditions, in all of their complex aggregates and interactions, can influence a specific individual's particular choices at particular times. But for now, the tools of neuroscience should not be accorded the deference of mathematical certainty.

Tragedies such as the Bales case remain personal. They are not mathematical formulas. They are part brain, part trauma, part motive and part circumstance. Not everything is predictable: That's not the human condition.


Jorge RE, Acion L, White T et al, White Matter Abnormalities in Veterans with Mild Traumatic Brain Injury, Am J Psychiatry 2012; 169: 1284-1291.

Silver, JM. Diffusion Tensor Imaging and Mild Traumatic Brain Injury in Soldiers: Abnormal Findings, Uncertain Implications, Am J Psychiatry 2012; 169: 1230-1232.

A. P. Georgopoulos, H-R M. Tan, S. M. Lewis et al, The Synchronous Neural Interactions Test as a Functional Neuromarker for Post-Traumatic Stress Disorder (PTSD): A Robust Classification Method Based on the Bootstrap, J Neural Eng. 2010 February; 7(1): 16011.