When a Texas House committee launched a dramatic effort to stop death row inmate Robert Roberson’s execution this month, it brought long-running disagreements about his guilt out of the courtroom and into the broader public arena.
Roberson was convicted of capital murder in 2003 for the death of his chronically ill 2-year-old daughter, Nikki Curtis. At his trial, prosecutors accused Roberson of shaking Nikki so violently that she died.
After several state and federal courts shut down his attempts to halt his execution, Roberson, who has maintained his innocence over two decades, was left with few options.
The day before Roberson’s Oct. 17 execution was set to take place, the Texas House Committee on Criminal Jurisprudence Committee held a hearing airing his claims of innocence and lack of due process. The panel’s surprise move to subpoena Roberson on Oct. 16 then successfully forced a delay in his execution.
The campaign to win Roberson a new trial has led to fierce pushback from Texas’ top Republicans and some on the state GOP’s rightmost flank, setting off a bitter political battle of competing narratives about Roberson’s case.
The prosecuting district attorney has stood by the state’s case against Roberson, as has Attorney General Ken Paxton, the state’s top legal official. But Roberson’s supporters, including some celebrities, say that there is enough evidence supporting his innocence claim to warrant a new trial. Some lawmakers also argued that the courts aren’t properly implementing the state’s 2013 “junk science” law, which allows for new trials when convictions rely on scientific evidence that is later discredited.
READ MORE: Robert Roberson's attorneys publicly released transcripts from his trial and a post-conviction evidentiary hearing the trial court held after his execution was stayed in 2016.
Shaken baby diagnosis was quickly accepted
On Jan. 31, 2002, Roberson rushed Nikki’s limp, blue body to a hospital emergency room in Palestine. He said he awoke to find her unresponsive after she fell from their bed.
The toddler had a history of breathing problems and head injuries. She was severely ill in the week leading up to her death, prompting Roberson to take her to the emergency room and pediatrician — where her temperature was measured at 104.5 degrees. She was diagnosed days before her death with a “respiratory illness, possibly viral,” and sent home with prescriptions for promethazine, which a toxicology report later showed to be in high quantities in her system.
After her final collapse, a doctor diagnosed Nikki with shaken baby syndrome, which presumes abuse. Brian Wharton, the lead detective in Nikki’s death who testified against her father at his trial, has said he accepted that medical diagnosis before arresting Roberson.
Roberson’s advocates noted that medical examiner Jill Urban classified the girl’s death as a homicide after law enforcement told her Roberson had been arrested for capital murder.
“This approach has been roundly disavowed in forensic science because it creates bias,” lawmakers wrote in an Oct. 23 letter rebutting Paxton’s claims about the case.
Despite Roberson’s insistence that he was innocent, his defense attorney at trial did not contest the shaken baby syndrome diagnosis and argued only that Roberson had not meant to kill Nikki.
Medical providers and investigators at the time also viewed Roberson’s flat affect as corroborating his guilt. Roberson was diagnosed with autism in 2018. His attorneys have pointed to statements medical providers and local law enforcement made about his non-typical expressions as signs of bias against a disabled man.
Consensus on shaken baby syndrome has changed
In 2002, the year Nikki died, the medical consensus said that if a child presented with a “triad” of symptoms — unexplained bleeding on the brain, bleeding behind the retinas and brain swelling — doctors could presume that the child was violently shaken and make a shaken baby syndrome diagnosis.
But studies later began to show that many of the same symptoms could also be caused by short falls, a wide range of naturally occurring medical conditions and accidental traumas. Research also found that shaking a baby violently enough to cause the triad would likely break the child’s neck.
At least 34 people convicted in cases relying on shaken baby diagnoses have been exonerated since 1992, according to the National Registry of Exonerations.
“When you read the transcript in this case, I think it's very clear what the jury heard: outdated, unverified, unreliable science that was presented to the jury as fact,” Don Salzman, a pro-bono lawyer whose firm reviewed Roberson’s case, told the Texas House panel Oct. 21. “And it was essentially, as this committee has expressed, a diagnosis of murder.”
There are dueling portrayals of the prosecution’s case
Prosecutors and the attorney general’s office have sought in recent weeks to downplay how central Nikki’s shaken baby diagnosis was to Roberson’s conviction, claiming instead that the prosecution showed that Nikki died from “blunt force head injuries” caused by beating.
But Roberson’s attorneys, advocates and a juror in his trial said the state’s case hinged on the shaken baby diagnosis. The trial transcript contains several dozen references to shaking.
Urban, who performed Nikki’s autopsy, testified during trial that “shaking also falls into this definition of blunt force.”
“Although it doesn't seem like, you know, shaking is not necessarily striking a child, when you are — when a child is say, shaken hard enough, the brain is actually moving back and forth within, again, within the skull, impacting the skull itself and that motion is enough to actually damage the brain,” Urban said.
Roberson’s supporters also note that there was no distinction made at trial between shaken baby syndrome and “blunt force head injuries.”
“Everything that was presented to us was all about shaken baby syndrome,” Terre Compton, a juror who convicted and sentenced Roberson, said on Oct. 21. “That was what our decision was based on. Nothing else was ever mentioned or presented to us to consider.”
In his appeals, Roberson presented evidence that Nikki died from natural and accidental causes, not from head trauma.
His most recent appeals cite medical and forensic experts who concluded that Nikki likely died from undiagnosed pneumonia, which suppressed her breathing and was exacerbated by Phenergen and codeine — medications no longer prescribed to children. Her illness progressed to the point of sepsis, experts said, triggering her fall from bed, causing bleeding in her brain and leading to her death.
The state’s highest criminal court most recently rejected Roberson’s appeals on procedural grounds. In a concurring opinion, Judge Kevin Yeary wrote on Oct. 10 that the case was “not just a ‘shaken baby’ case,” pointing to the argument that Nikki “suffered multiple traumas.”
Medical examiner’s methods and conclusions have been criticized
The attorney general’s office and the prosecuting district attorney have sought to portray Roberson as a child abuser with a violent history who beat his daughter to death.
Paxton in particular has promoted claims that there were multiple sites of physical trauma and widespread bruising on Nikki’s body — claims that appeared to be based primarily on Nikki’s autopsy report, which he attached to an Oct. 23 press release.
The treating physicians who first saw Nikki — who had a bleeding disorder that made her prone to bruising, including from medical intervention — did not make note of multiple sites of physical trauma upon her arrival at the hospital.
“Since Nikki wasn’t admitted with these injuries, and especially because she had [a blood-clotting disorder], it’s clear that they came from two days of being picked up, moved and manipulated during emergency procedures,” a group of lawmakers wrote in a rebuttal to Paxton’s Oct. 23 statement.
An initial CT scan before Nikki’s death showed only a single site of physical trauma on the back of her head. The state’s principal medical witness, Dr. Janet Squires, testified at trial to just one impact site. Several medical experts, who were asked to review Nikki’s records by Roberson’s current attorneys, reached the same conclusion.
A single impact site is consistent, Roberson’s advocates said, with his explanation that he had woken up to find that Nikki had fallen from the bed before he rushed her to the hospital.
Roberson’s advocates raised concerns about the autopsy, saying that Urban admitted that she did not consider Nikki’s medical history, CT scan or toxicology report in completing her report. Urban also examined Nikki’s body after days of emergency medical treatment.
Squires, who treated Nikki in Dallas after she was transported from the Palestine hospital, also testified that she found “minimal bruising,” a “little chin abrasion,” “no other indication of traumatic injuries,” “no fractures,” “no old fractures,” “no scars, no unusual bruising or anything.”
The lack of external injuries, Squires said, led her to conclude that Nikki’s condition was caused by violent shaking.
Roberson was accused of having a violent history
Paxton, and others opposed to stopping Roberson’s execution, pointed to testimony describing Roberson as violent to validate his death sentence.
Roberson was previously arrested for burglary, writing hot checks and violating his probation, his lawyers said — but he had “absolutely no accusations, arrests or criminal charges for any violence before he was falsely accused of hurting Nikki.”
The primary evidence of violent behavior presented during Roberson’s trial came from witnesses his attorneys have argued are non-credible: his then-girlfriend Teddie Cox, her underage daughter Rachel Cox, her underage niece Courtney Berryhill, and Roberson’s ex-wife Della Grey.
Roberson’s attorneys argued that law enforcement put “enormous pressure” on the Coxes and Berryhill to testify against Roberson, making Teddie Cox believe that she could be prosecuted for Nikki’s death if she did not.
His advocates also said that Teddie had admitted under oath that her assessment of Roberson changed depending on “how [she] feel[s]” at the moment and whether she was “mad” at him at the time.
Teddie Cox’s sister Patricia Conklin testified that Roberson had always been loving and caring with Nikki, and that Teddie Cox had a track record of lying.
Lawmakers supporting Roberson argued in their rebuttal to Paxton’s assessment of the case that the testimony against Roberson was “so dubious” that the state did not even include it in its statement of facts during his appeals process.
During Roberson’s sentencing, Gray accused him of strangling her with a coat hanger, punching her while pregnant and beating her with a fireplace shovel — allegations that Paxton repeated in his Oct. 23 statement.
But Roberson’s advocates noted that there were no contemporaneous medical, police or court records to corroborate her claims. And they said that Gray did not raise the allegations during their divorce proceedings, when she would have likely been motivated to do so.
“Critically,” lawmakers said, “she admitted that she was unhappy with how the divorce had gone and had only come back to testify at the trial involving Nikki to make Robert ‘pay’ by testifying against him.”
Claims that Roberson sexually abused Nikki were never corroborated
People who believe Roberson is guilty have also repeated claims that he sexually abused Nikki.
But just one nurse made that allegation: Andrea Sims, who said at trial that she was a certified sexual assault examiner before later admitting on cross-examination that she was not, in fact, certified.
Roberson’s attorneys and Brian Wharton, the case’s lead investigator, said that Sims examined Nikki for signs of sexual abuse without being asked by law enforcement.
Sims testified that she saw three anal tears that led her to conclude sexual abuse. The prosecutor then asked her to conject, before the jury, about pedophiles and anal sex.
Squires, the Dallas pediatrician, refuted Sims’ conclusion, saying she observed one tear and took it as typical for a toddler.
Then, even after dropping the allegation as grounds for capital punishment, the prosecutor discussed sexual abuse in his closing statement to the jury.
Roberson’s attorneys condemned the sexual abuse allegation as unsubstantiated and prejudicial to Roberson’s trial, noting that the state offered no evidence beyond Sims’ testimony. They added that Nikki struggled with diarrhea in the week leading up to her death.
“This is being thrown around, in my opinion, recklessly and dangerously all over the internet right now that Roberson sexually abused and did other things to Nikki,” state Rep. Brian Harrison, R-Midlothian, said Oct. 21.
He asked Salzman, who supported Roberson’s appeals, whether there has “ever been a scintilla of substantiated evidence to support that claim.”
“Absolutely not,” Salzman said.
In his Oct. 23 release, Paxton also detailed a jailhouse informant’s claim that Roberson had admitted to molesting Nikki — a report that even the prosecution did not raise during the trial.
“By including this information, the [attorney general’s office] has repeated a lie with, at best, a complete indifference to the truth,” the lawmakers wrote in their rebuttal. “The ‘jailhouse snitch’ here wove a tale so outrageously contrary to the evidence that prosecutors didn’t use it at trial.”
Relatives of Roberson and Nikki also have conflicting opinions about the death row inmate
In an Oct. 28 letter to the Texas House Committee on Criminal Jurisprudence, members of Nikki’s family — her brother, Matthew Bowman; her aunt, Jessica Rachelle Carriere; and her grandfather, Larry Gene Bowman — said that they were convinced of Roberson’s guilt and had “witnessed the repeated abuse” he allegedly inflicted.
“We do not believe Mr. Roberson should be put to death simply because he is a bad man,” they wrote. “We believe his death sentence should be carried out based on the facts of this case, which remain true today, as well the overwhelming evidence that was presented at the trial that led to the jury’s verdict.”
Roberson’s family members disputed that depiction.
“No one who knew Robert well believed he was capable of harming any child,” Thomas Roberson wrote in a statement about his brother. “He was someone who stood up for kids who were being picked on by others. I never saw him hurt or say a mean word to any child.”
Roberson’s attorneys also noted in a statement that Nikki’s maternal grandparents agreed that Roberson should have custody of Nikki and asked him to care for her the night she collapsed.
“To the best of our knowledge, the record does not show that these family members expressed concern about Robert Roberson caring for Nikki before her tragic death,” his attorneys said. “Respectfully, these family members’ beliefs have been shown by modern science to be misguided and mistaken.”
New evidence has gained little traction with courts
In his appeals, Roberson presented reams of evidence discrediting Nikki’s shaken baby diagnosis and indicating that her death was natural and accidental.
But the state responded by reiterating its case from trial and stating simply that the science had not changed as much as Roberson’s attorneys claimed and that the evidence against him remained convincing.
The court largely adopted the state’s case without integrating the new evidence Roberson had presented, and it declined in 2023 to grant him a new trial, saying there was insufficient doubt over the cause of Nikki’s death to do so.
Roberson has tried again in the past year, filing appeals with new testimony from a broader range of medical and forensic experts who examined Nikki’s records and concluded that her death was natural.
The Court of Criminal Appeals has rejected those appeals on procedural grounds, without reviewing the new evidence or the merits of his claims.
“I do not believe any court has meaningfully acknowledged the evidence,” Gretchen Sween, Roberson’s attorney, told lawmakers on Oct. 16. “Here, we had a show of process that was robust. But if you look at the obstacles — they were immense, and the goal was never really to find out what had happened to Nikki. And it’s a dishonor to Nikki, even more than Robert.”
Some legislators say junk science law isn’t being followed
The course of Roberson’s appeals has raised concerns among lawmakers, defense advocates and former Court of Criminal Appeals Judge Elsa Alcala that the courts were not properly applying Texas’ 2013 junk science law.
That law was created to allow courts to overturn a conviction when the scientific evidence at the center of the prosecution has since changed or been discredited.
In a July report, the Texas Defender Service, a nonprofit working to end the death penalty, found that the Court of Criminal Appeals had applied the law inconsistently, placing a higher burden on convicted people than required by the law, largely ruling against cases with types of flawed evidence other than DNA, and denying relief in all capital cases.
“The court has not used a common sense approach in applying these statutes,” Alcala, who served on the bench both before and after the junk science law was codified, argued to lawmakers on Oct. 21. “It’s been over-technical, and it leans too far to uphold convictions and sentences that I think many other Supreme Courts in other states would reverse.”
Members of the Committee on Criminal Jurisprudence have indicated interest in considering amendments to the junk science law to ensure that it is applied as intended.
“This law, in this way, isn't about one person — it is about the system as a whole,” state Rep. Joe Moody, D-El Paso and the committee’s chair, said on Oct. 16. “The issue in front of us is whether our law, which was a first of its kind in the nation and heralded as a landmark legislation, hasn’t been thrown in the garbage in the courts.”