Texas Might Soon Become the First State To Execute Someone Based on Disputed 'Shaken Baby Syndrome' Evidence

Texas has set an October 17 execution date for Robert Roberson, convicted in 2003 of murdering his 2-year-old daughter.

If Texas follows through with its plans to impose the death penalty on Robert Roberson this October, it will become the first state in the country to execute someone whose conviction was based on what's commonly known as "shaken baby syndrome."

The former detective who arrested Roberson for murdering his 2-year-old daughter is no longer convinced that he did it. Roberson's attorneys say they have new evidence based on previously hidden medical records that his child really died of severe pneumonia, and they argue in a recently filed motion for a stay of execution that, since Roberson's trial, the theory behind shaken baby syndrome—now called abusive head trauma (AHT)—has "been entirely exposed as devoid of any scientific underpinnings."

Nevertheless, prosecutors insist that Roberson, who has been on Texas' death row since 2003, was convicted on sound evidence and should face justice. Earlier this summer, a judge agreed and set an October 17 execution date.

Roberson's case is only the highest-stakes example of a fight playing out in courtrooms around the country.

Last month, the Michigan Supreme Court overturned a 2006 murder conviction in the death of an infant and ruled that the defendant deserves a new trial. The court found that the defendant's expert witnesses, including a doctor who had testified for the prosecution at her original trial, had presented enough new evidence during the appeal process for a jury to have reasonable doubt about her guilt.

The Columbus Dispatch reported on two similar cases in Ohio this February. In one, a court dismissed charges in the retrial of a man convicted in 2003 of killing a 2-year-old child, ruling that a "shift in understanding by the medical community" meant there was "a strong probability of a different result on retrial." In the other case, a court ruled that a woman was eligible for compensation for spending 18 years in prison for the alleged shaking death of a child she was babysitting in 2003. Her conviction was overturned after the pathologist who conducted the autopsy recanted.

In New Jersey, the state supreme court is considering whether AHT evidence should be allowed in courtrooms at all, after a state trial court judge barred testimony from AHT experts in a child abuse trial in 2022. The trial court judge wrote that AHT is "an assumption packaged as a medical diagnosis, unsupported by any medical or scientific testing." A state appeals court upheld the ruling, writing that "the very basis of the theory has never been proven."

Even the pediatric neurosurgeon who first popularized the diagnosis in 1971 had grave doubts about how AHT ended up being used in courtrooms.

Some prosecutors and child abuse specialists have dismissed these objections as "fringe courtroom science" that undercuts very real incidents of abuse. But the number of exonerations points to a larger problem, or at least a live controversy. According to the National Registry of Exonerations, at least 33 parents and caregivers in 18 states convicted based on AHT evidence have been exonerated.

Roberson will join either that list or the much longer list of people who have taken the long walk to the execution chamber in Huntsville.

"I'm just adamantly refusing to acknowledge that's a possibility," his attorney, Gretchen Sween, says of the latter. "But we have not been able to find anybody that's come this close" to being executed.

A Definitive Diagnosis

One undisputed fact in this case is that Roberson's 2-year-old daughter, Nikki Curtis, was ill for much of her short, tragic life. She had chronic, antibiotic-resistant infections.

Roberson and Curtis' mother were estranged, and Curtis' maternal grandparents were the child's primary caregivers until Roberson was awarded custody of her shortly after her second birthday.

Roughly two months later, on January 28, 2002, Roberson took Curtis to the emergency room for vomiting, coughing, and diarrhea. He took her back to a pediatrician the next day, when her fever spiked to 104 degrees. Curtis was prescribed medication and sent home both times.

Curtis was supposed to stay with her grandparents for the next two days. But when one of them fell ill on the night of January 30, Roberson had to pick Curtis up. He was alone with her for the next 12 hours, the last person to see her conscious and breathing.

According to Robersons, he woke up in the middle of the night after hearing a "strange cry" to find that Curtis had fallen out of their bed—a box spring and mattress propped up on cinder blocks. He comforted her until they both fell back asleep. Hours later, Robert woke up again and found that Curtis had stopped breathing and turned blue.

Roberson drove Curtis to the hospital, where she was declared brain-dead. CAT scans revealed a set of internal head conditions: bleeding in the brain, brain swelling, and retinal hemorrhages. Child abuse specialists consider these the definitive "triad" of symptoms for AHT.

Pediatric neurosurgeon Norman Guthkelch released a groundbreaking paper in 1971 theorizing that the whiplash from violent shaking could cause fatal brain injuries in infants, even absent external evidence of abuse such as bruising and broken bones. Shaken baby syndrome not only led to a major shift in public awareness of child abuse; it also provided powerful evidence for criminal charges against abusive parents.

Curtis was taken off life support after two days. Police arrested Roberson for capital murder before an autopsy was performed.

"Instead of exploring all possible causes of the injury sustained by a chronically ill child who had been at the doctor's office with 104.5-degree temperature only two days before," Roberson's lawyers wrote in an August motion for a stay of execution, "a tragedy was deemed a crime and a father, doing the best he could to care for his daughter despite severe cognitive impairments, was falsely branded a murderer."

'There Had To Have Been Something More Than Just Impact'

At Roberson's trial, prosecutors for the state of Texas painted a very different picture.

Roberson's ex-girlfriend, as well as her daughter and one of her nieces—ages 10 and 11, respectively—testified that Roberson would shake and spank Curtis when she wouldn't stop crying. Roberson's ex-wife from a past marriage also testified that he physically beat and abused her. (Roberson's lawyers argue in his habeas corpus petition that the witnesses were biased, were not credible, and contradicted each other, and that the accusations only came out well after Roberson's arrest, following prodding and leading questions from police.)

Hospital staff testified that Curtis had bruises on her chin and jaw. An emergency room nurse testified at length, without corroborating evidence or certification as a sexual assault examiner, that Roberson had sexually abused his daughter. 

Nurses, police, and doctors also testified for the state about Roberson's odd, emotionless, and lackadaisical behavior. He took the time to dress Curtis and find a parking spot at the hospital, even though she had stopped breathing.

Roberson, who struggled with speech delays growing up and dropped out of high school in ninth grade, was diagnosed with autism in 2018, which his lawyers say led to police and doctors misinterpreting his behavior.

But the most damning evidence came from two medical experts who testified that Curtis' injuries appeared to have been caused by shaking and blunt force impacts. AHT played a large part in both doctors' testimonies.

"It is my opinion…that there was some component of shaking that happened to explain all the deep brain injury out of proportion, I would say, to the injury to the skull and the back of the head," testified Janet Squires, a licensed expert in Child Abuse Pediatrics who has testified in hundreds of child-abuse trials. "There had to have been something more than just impact."

By the late 1990s, the medical and law-enforcement consensus was not just that shaking an infant could produce the triad of symptoms associated with AHT. If the triad of symptoms was present, the default conclusion was abuse by a parent or caretaker.

A 2002 guide to investigating child abuse from the Department of Justice's Office of Juvenile Justice and Delinquency Prevention said retinal hemorrhages were, "for all practical purposes, conclusive evidence of shaken baby syndrome in the absence of a good explanation."

In a criminal justice system where convictions must be based on evidence beyond a reasonable doubt, an expert's testimony that a child's injuries could only have been caused by abuse is devastating to a defendant's case, especially a defendant already facing a highly prejudicial charge.

Faith in the diagnosis was so entrenched at the time that Roberson's trial attorney did not even challenge it, describing the incident in his opening arguments as a "classic shaken baby case." Roberson's current attorneys wrote in his recent petition for a writ of habeas corpus that the statement was an "enormously prejudicial and erroneous concession."

In 2003, a Texas jury convicted Roberson of capital murder and sentenced him to death by lethal injection.

Shifting Science and Polarized Debate

Even at the time Roberson was convicted, cracks were starting to form in the orthodoxy around AHT. 

A growing body of research over the 2000s and 2010s challenged the assumption that the triad of AHT symptoms was conclusive evidence of child abuse. Studies showed that other conditions, such as blood disorders, could mimic AHT symptoms. A 2009 Washington University Law Review article summarized the findings: "Depending upon the clinical picture presented, the differential diagnosis for symptoms previously associated exclusively with [shaken baby syndrome] now contemplates a wide range of nontraumatic possibilities: medical or surgical interventions; prenatal, perinatal and pregnancy-related conditions; birth effects; infections; diseases; disorders; malformations; post-vaccinal conditions; re-bleeds; and hypoxia (lack of oxygen to the brain)."

Guthkelch, the doctor who first popularized shaken baby syndrome, became a critic of how it was being used in courtrooms after he reviewed the case of an Arizona man convicted of murder based on AHT evidence and concluded the man was innocent. In a 2012 paper published in the Houston Journal of Health Law & Policy, Guthkelch urged pediatric abuse specialists to question their findings.

"Often, there seems to have been inadequate inquiry into the possibility that the picture resulted from natural causes," Guthkelch wrote. "In reviewing cases where the alleged assailant has continued to proclaim his/her innocence, I have been struck by the high proportion of those in which there was a significant history of previous illness or of abnormalities of structure and function of the nervous system, suggesting that the problem was natural or congenital, rather than abusive. Yet these matters were hardly, if at all, considered in the medical reports."

Guthkelch, who died in 2016, spent his last years trying to amicably resolve the unusually fierce debates between the two camps—pediatric organizations and abuse specialists, who still represented a majority view, and the skeptics who often worked with defense attorneys and innocence groups.

In 2018, Pediatric Radiology published a consensus statement from nine pediatric and radiology organizations. It declared that there "is no controversy concerning the medical validity of the existence of AHT," but it went on to say that a doctor's workup "must exclude medical diseases that can mimic AHT."

In 2020, the American Academy of Pediatrics released a revised policy statement on AHT stating that the organization "continues to embrace the 'shaken baby syndrome' diagnosis as a valid subset of the AHT diagnosis." It acknowledged the debate over the diagnosis but framed it as "a philosophical one, not a scientific one"—an argument over the degree of certainty physicians attribute to their AHT findings on the stand.

A major reason for the perception that AHT is a flawed theory is that exonerations keep happening. In 2021, the Georgia Supreme Court overturned the life sentences of a man and woman convicted of murdering their newborn daughter after finding they had received ineffective counsel. The couple presented evidence that their daughter had a serious birth defect and that county prosecutors had illegally withheld CAT scans that would have supported their defense.

Critics of how AHT is used in courtrooms say that's the rub—not that it's impossible to hurt a baby by shaking it, but that cognitive bias causes doctors to inadequately consider other possibilities and dismiss contrary evidence.

"I've had cases in the last several years where physicians think that a case is abuse," says Kate Judson, executive director of the Center for Integrity in Forensic Sciences, a group that works against what it considers unreliable forensic techniques. "They order tests, and those tests come back outside of the normal limits. So a kid's blood is taking too long to clot or it's clotting too quickly, which can both lead to significant problems, including internal bleeding and stroke and things like that. And then they say, 'Well, those tests are not relevant because we already knew it was trauma.'"

Another problem, Judson says, is that the adversarial nature of the court system doesn't give doctors the kind of feedback that makes them better diagnosticians.

"A physician I know told me you only miss the same terrible cancer once if you're a good doctor," Judson says. "You miss it, you get caught, you feel horrible, and then that's always in your mind. You don't make that mistake again, one hopes, right? But in these cases, there's no meaningful opportunity for feedback, because once the physician decides that it's child abuse, much of the other investigation ceases."

And because the U.S. justice system heavily values finality in verdicts, courts lag behind the times when it comes to science. It can take years to get courts to question the validity of established forensic methods or even pure quackery once they're introduced. For example, Texas only recently banned prosecutors from using statements obtained by police through "investigative hypnosis."

Texas is something of a paradox, though, because in 2013 it also became the first state in the country to pass a law allowing inmates to challenge convictions based on outdated or debunked forensic methods. 

The so-called "junk science writ" already helped Roberson narrowly avoid a trip to the execution chamber once. In 2016, the Texas Court of Criminal Appeals (CCA) granted Roberson a stay of execution to allow him a hearing to challenge his conviction on both actual innocence grounds and under the junk science writ.

Yet a trial court and the CCA both denied Roberson relief and agreed with prosecutors that because the experts at Roberson's trial had not exclusively relied on an AHT diagnosis but a combination of shaking and blunt impacts, the state's cause-of-death theory had not been disproven.

"I am a native Texan," says Sween, "and one of the maddening things about this state is there can be these very bold moves in the right direction to reform criminal justice and then just absolute recalcitrance and unfairness baked into the same system."

A July report by the Texas Defender Service, the first comprehensive review of the junk science writ, concluded that the "law systematically fails to provide relief to innocent people convicted based on false forensic evidence."

"The legislative solution is useless if the courts don't honor it," Sween continues. "Not a single person on Texas's death row who has tried to get relief under the statute has gotten it, not one. And that is statistically alarming."

New Evidence and Old Acquaintances

What's different with Roberson's new round of last-ditch appeals is that his lawyers now say they have evidence, not available in 2016, that Curtis died from undiagnosed viral and bacterial pneumonia so severe that it progressed to sepsis—not from shaking or blunt force head trauma.

CAT scans, X-rays, and a complete autopsy file were rediscovered in a courthouse basement in 2018, and Roberson's lawyers recruited an expert in pediatric radiology, a lung pathologist, and an expert in medical toxicology to review the records.

After inspecting the images and autopsy evidence, the three experts concluded that the symptoms attributed to abuse could all be explained by Curtis' severe pneumonia, complications from her oxygen-starved condition, and side effects of medications she was improperly prescribed during the two doctor visits prior to her death. One of those medications was codeine cough syrup, which depresses respiratory function, and the other, Phenergan, now has a "black box" warning by the Food and Drug Administration against prescribing it to children younger than 2 because it can cause severe breathing problems.

"The condition of Nikki's lung tissue cannot be reconciled with the conclusion that her death was caused by blunt force head injuries, inflicted or otherwise," the lung pathologist, Francis Green, wrote in the habeas petition.

The radiologist also concluded that there were not "multiple impacts" to Curtis' skull, as a doctor testified at Roberson's trial—just a single, minor impact that was insufficient to cause her death.

Among the most unexpected believers in Roberson's innocence is the man who arrested him: Bryan Wharton, the lead detective in Palestine, Texas, at the time of Curtis' death.

In a video interview with The New York Times this July, Wharton, who is now a minister, said that based on what he now knows of the controversy over AHT and Roberson's autism diagnosis, there is "unassailable doubt" that Roberson committed the crime.

"No other possibilities for [Curtis'] injuries were considered," Wharton told the Times. "I regret deeply that we followed the easiest path."

As Roberson's October 17 execution date approaches, the window of opportunity for the state of Texas to choose another path grows smaller and smaller.