Imagine that you are a judge presiding over the trial of a man named Bill, accused of a grisly murder. The physical evidence is overwhelming, and witnesses have yielded damning testimony. There seems to be no reasonable doubt that Bill committed the murder. Suddenly, the defense asks if it can present images of Bill's brain, produced by magnetic resonance imaging (MRI). Bill's attorneys want to introduce the pictures as evidence that their client has a brain abnormality. They will argue that the abnormality justifies either a verdict of not guilty (because Bill lacked the intent to kill or premeditation to commit murder), or a verdict of not guilty by reason of insanity (because Bill lacked control over his actions), or, at least, a conviction on a lesser offense (because Bill is not fully responsible or possibly just because jurors should feel sorry for people with brain disorders). The prosecution argues that you should not admit the scans, because pictures of Bill's brain and testimony by revered scientists might influence the jury much more than such evidence warrants.

Would you, as judge, allow the brain scans to be exhibited? How would you assess such evidence?

This scenario is not merely theoretical. U.S. courts have been allowing positron-emission tomography (PET) scans of brain metabolism and, more recently, MRI scans of abnormal brain structures or function. In these cases, attorneys have used the images as evidence of brain damage that might impair a defendant's behavior and therefore reduce his or her culpability--the degree to which the defendant should be held responsible for the crime. Strong editorials have been written against these actions. Yet much of the public as well as participants in the court system believe that imaging, particularly the newer anatomical and functional scans produced by MRI, could provide an independent assessment of the root cause of a person's aberrant behavior.

As a result, neuroscientists are invading our courts. Companies such as No Lie MRI, Inc., and Cephos Corporation have been started on the expectation that the legal system will increasingly seek brain scans. Proponents say the images can detect lying by witnesses, prejudice in jurors or judges, and mental incapacities among defendants. If these prospects pan out, neuroscience could greatly influence the direction of our legal system; after all, other forms of advanced scientific technology, such as DNA testing, already work well in courts. Opponents object that such technology removes the human element that is essential to law as we know it. They also fear that brain scans will violate precious rights to privacy and due process.

Although the arguments on both sides are forceful, in Bill's case they come down to a single salient question: Can brain scans reveal a lack of responsibility? We think not. They should not be allowed as evidence in trials, at least not in the foreseeable future. Never say "never," but nothing close to current technology is trustworthy enough for legal settings.

What Made Bill Do It?
Naive faith in the latest imaging technology is misguided at this time. To understand why, consider the questions one must navigate to decide whether this evidence could be truly informative in a criminal trial.

First, if a brain scan indicates an abnormality, then the brain really has an abnormality, right? Wrong. This simple inference overlooks a crucial problem: almost every biomedical test, from MRI to the prostate-specific antigen (PSA) test, can suggest that a condition is present when in actuality it is not. Such cases are called false positives. This problem is not too serious for common medical ailments, such as prostate cancer, when doctors can independently confirm the diagnosis using other tests. The kinds of brain abnormalities that might cause grisly murders, however, are very rare and hard to confirm. When a condition is rare, even a low rate of false positives leaves a relatively large number of errors--not a very reliable means for establishing that the person being scanned has a condition that provokes violence. Even if Bill's scan suggests a brain anomaly, it might be very unlikely that he has any deficit at all.

That is not the only problem. Suppose for the sake of argument that we are absolutely certain that Bill has an abnormality. We still do not know whether that condition caused Bill's criminal behavior. Some people with this kind of irregularity might not be violent at all, whereas others could become violent on a regular basis. With this much variability, even if we assume that Bill does have an abnormality of the right size in the right place, we cannot know that his condition had anything to do with the alleged illegal behavior. Furthermore, even if Bill's condition does cause him to be violent in some way, it still might not cause the particular kind of attack in question: ugly premeditated murder.

To be confident that an abnormality such as Bill's plays a causal role in a particular murder, researchers would have to have studied many more murderers than anyone has ever studied. The best an expert witness in a courtroom could do is establish a weak correlation between brain injury and criminal behavior. But without additional information, no scientist could be justified in claiming that Bill's abnormality caused him to become a murderer or prevented him from making a decision to kill on the day in question.

The defense might argue that the brain scan is just one piece of evidence that when combined with psychological or psychiatric assessments, paints a better picture of Bill's mental state at the time of the crime. Yet we do not know what the relation is between the scan and the other assessments. What percentage of people with a certain psychiatric diagnosis will test positive for this abnormality? What percentage of those who test positive for this abnormality will receive that psychiatric diagnosis? Without such information, we cannot say in the least whether the brain scan supports the diagnosis. In this setting, the behavioral findings must stand on their own.

Is Bill Responsible?
Even if these diagnostic problems can be solved, a defendant's medical condition still will not prove any legal status. Assume we know for sure that Bill has a certain brain abnormality and that a high percentage of people with that kind of irregularity commit violent crimes, including murder. Bill might still be culpable.

To see why, imagine that most people with a certain brain condition are thrill seekers. They drive race cars, jump from planes, scale ice cliffs, and so on. These activities are unusual (possibly as unusual as violent crimes), but their correlation with a certain brain condition does not indicate that these individuals do not act intentionally and deliberately or that they suffer a compulsion or delusion that makes them unable to control themselves. Such thrill seekers plan their acts carefully and stop themselves when conditions are too dangerous. That makes them responsible for what they choose to do.

Similarly, even if most people with a given abnormality engage in unusual criminal activities, that abnormality by itself does not indicate that these individuals do not commit their crimes intentionally and deliberately. They are still capable of premeditating or planning their acts carefully. This means they have the ability to "form malice aforethought"--the mens rea that is a necessary element of the crime of murder--and therefore should not be exculpated during the guilt phase of a trial. Moreover, if the trial is one based on a plea of not guilty by reason of insanity and the evidence is presented in the sanity phase of the trial, such abnormalities would not justify a verdict of not guilty, because these individuals do not necessarily suffer from a compulsion or delusion. Like thrill seekers, they might well be able to control themselves and make decisions easily.

To show that Bill is not responsible, a brain scan would have to indicate not only that Bill has an urge and is likely to commit the crime but also that Bill is unable to control his urge. Brain scans show only what is, however, not what could be. They cannot show that Bill could not have stopped himself from committing the murder. Because responsibility depends on such abilities, brain scans cannot show that Bill is not responsible for what he did.

These stringent standards might seem unsympathetic. Shouldn't we feel compassion for people with brain disorders and help them get better? Of course, we should. But if we allow the defense to use brain scans to dismiss guilt, then should prosecutors not also be allowed to use brain scans to indicate guilt? If a brain scan of a defendant reveals an abnormality and some people with that aberration become violent, then a prosecutor might use that brain scan to convince a jury that a given defendant is guilty. Yet innocent people who suffer from known brain disorders will be even more likely to test positive and be wrongly convicted. Or they might be involuntarily committed to mental institutions if the brain scan is taken as evidence that they are dangerous to society. Anyone who has sympathy for these folks should find this new form of evidence discomforting.

Who Bears the Burden of Proof?
Even without such technological uncertainty, the outcome of a trial often depends on who is responsible for proving what--which is a matter of law, not science.

The burden of proof can differ depending on the kind of proceeding, the phase of a trial and whether the case is before a federal or state court. Usually the burden does lie with the prosecution, to prove essential elements of a crime during the guilt phase. But a defendant's legal team may attempt to introduce a brain scan as evidence that negates some element of the crime, such as premeditation, or perhaps raises reasonable doubt. Research shows that when a new scientific procedure is first admitted in trials as evidence, there is significant risk that lay juries and judges will overestimate the value of that evidence. This danger is especially high when the procedure involves eye-catching pictures presented by scientists with impressive credentials. To reduce that risk of error, many jurisdictions require defendants who cite new scientific methods to demonstrate that those methods are reliable and accepted by the wider scientific community. That kind of proof will be hard to present for brain scans because of their low predictive value and the lack of consensus among scientists on the validity of these techniques. If the defense cannot carry that burden of proof, then the scientific evidence should not be admitted into the guilt phase of a trial.

Brain scans might instead be cited as evidence of something like insanity during the sanity phase of a trial. Many jurisdictions give the defense the burden of proving that a criminal act resulted from an individual's mental illness or brain condition. It will be difficult for the defense to carry this burden, for the same reasons just given.

The questions of where to place the burden of proof, which evidence to allow and which disabilities are severe enough to preclude punishment are all considerations for society. And these decisions must indeed be made by society, not by neuroscientists. Data about an individual's brain cannot alone settle whether that person should be held responsible. Responsibility is a social construct, determined by a social group, not by a medical or scientific test result. If society chooses to use forms of brain testing as evidence to assess responsibility, then it needs to make these decisions in light of complete and accurate information about the pitfalls of the various methods being proposed.

We cannot predict the future. Better information, techniques and equipment might come along that will someday make brain scans reliable enough to determine the legal implications of a brain abnormality. The problems might be solved with time, but we are nowhere close today. Brain scans of this kind are, after all, only 15 years old. Neuroscientists need much more basic research, experience and thought about imaging before it invades our courts. Until then, brain scans have too little predictive value to be applied in criminal trials.