Last July, the Texas Forensic Science Commission found that arson investigators had used flawed science in the trial of Cameron Todd Willingham. Willingham was convicted of setting the 1999 fire that killed his three children. He was executed in 2004, despite serious questions about his guilt. The case garnered national attention last year after a David Grann investigation in The New Yorker argued that there was no evidence the fire was intentional.
Now another questionable arson-related death penalty case is percolating in Pennsylvania, where Daniel Dougherty sits on death row for setting the fire that killed his two sons in 1985. Arson experts John Lentini and Angelo Pisani have found that, in a situation similar to the Willingham case, fire investigators relied on outdated and disproved science. One of them, for example, testified that the extreme heat put out by the fire could only have come from the use of a liquid accelerant, an assertion that arson specialists today say simply isn't true. Accelerants have little to do with how hot a fire burns. Subsequent investigators have found no forensic evidence that the blaze that killed Dougherty's children was deliberate.
In Mississippi two men—Devin Bennett and Jeffrey Havard—today sit on death row for the murder of children in their care. Both men insist the deaths were accidents. Both were convicted mostly based on the testimony of the controversial medical examiner Steven Hayne, whose dubious record was the subject of an investigative feature I wrote for the October 2007 issue of reason. In both cases, more reputable medical examiners who later reviewed Hayne's work reported that the signs of abuse he claimed to have found simply aren't there.
In recent years, DNA testing has exposed some serious flaws in several fields of forensic science. A 2009 National Academy of Sciences report found that forensic specialists often make exaggerated claims in the courtroom, expressing levels of certainty that aren't always supported by scientific research. In cases where it's certain that a crime was committed and for which DNA is available for testing, the challenge is getting prosecutors to agree to the tests. But in cases like that of Dougherty, Willingham, Bennett, Havard, and others, the question isn't who committed the crime, but whether a crime was committed in the first place.
DNA isn't of much use in those cases. But during a panel discussion at the Georgetown Law Center last year (disclosure: I moderated the panel), arson investigator Lentini suggested one reform that may give defendants in such cases a fairer crack at justice: bifurcated trials. Courts would hold an initial trial to determine if a crime was committed, then a second to determine who committed it.
The problem, as Lentini explained, is that by asking the jury to answer both questions at once, jurors are nudged toward answering both in the affirmative. When jurors first sit down in the jury box and see someone sitting in the defendant's chair, there's a tendency to think that some crime was committed. And once they've determined there was a crime, the person on trial is often seen as the only—or at least most likely—person to have committed it.
With infant death cases, for example, the only reasonable suspect is the person with the child at the time he or she died. Willingham and Dougherty were the only adults at home at the time the fire broke out. Their mere presence in the courtroom strongly implies from the outset that a crime was committed.
These trials then too easily become little more than judgments of a defendant's character; if the scientific evidence is a wash, the verdict may hinge on whether jurors believe the defendant is capable of violence. In Willingham's case, for example, prosecution witnesses testified about Willingham's emotional reaction after the fire, and whether it fit that of a grieving father or a man who had just committed murder. The prosecution also brought in a jailhouse informant—a notoriously unreliable source of testimony—who claimed Willingham had confessed to him. (Willingham's sentencing trial was even more farcical, with now-discredited psychiatrists testifying, among other nonsensical claims, that Willingham's Led Zeppelin and Iron Maiden posters were indicative of a death-obsessed psychopath who was likely to kill again.)
Such tactics aren't uncommon in these cases, and they can have the effect of getting the jury to look past the scientific evidence related to the important question of whether there has been a crime to be prosecuting in the first place. Once a witness says a guy was acting strangely, once his cellmate says he confessed, it isn't difficult to see how some jurors might then dismiss doubts about the quality of the evidence that a crime had been committed at all.
We already have bifurcated trials in other areas of the justice system. Their purpose is to isolate portions of court proceedings that require a judgment that isn't influenced by evidence from other portions of the case. In civil cases, for example, the awarding of damages is often separated from determining liability. In divorce cases the division of property may be a separate proceeding from the determination of who gets custody of the couple's children. In death penalty cases we separate the guilt portion of the trial from the sentencing.
And so in homicide cases where the defense can plausibly argue that a death wasn't a murder, it makes sense to resolve that question first. After a suspect is charged, defense attorneys would have a pre-trial deadline by which to make a motion for bifurcation. Such a motion would likely come into play not only in arson cases, but also in child molestation cases, infant death cases, and any other situation where there's a legitimate question whether a death really was a homicide.
The state would likely then file a brief in opposition to bifurcation. The judge would rule whether the trial is a candidate for bifurcation, using the same "beyond a reasonable doubt" standard used in criminal cases. If the judge rules for bifurcation, each side would then bring in experts to make their case before the jury. But there would be no defendant in the courtroom, and the state's case against a specific suspect would wait until after a jury first determines that there's a crime to charge.
In the 20 years since DNA testing began exonerating the wrongly convicted, there's been some momentum toward reforming the criminal justice system. Much of it has been understandably focused on improving the quality of the science used in the courtroom. But we should also consider structural changes, ways we can help juries deliberate verdicts as objectively as possible. In cases where the primary issue is not who committed the crime but whether a crime was even committed at all, bifurcated trials might be a way to prevent jurors from confusing those questions. Admittedly, it's a fairly radical idea. But we now have compelling evidence that the government has executed at least one innocent man, and we know for certain that it nearly executed several others. That ought to shock the conscience enough to start considering more radical reforms.
Radley Balko is a senior editor at reason.