The forensic pathologist who had spent the better part of 30 years investigating violent deaths walked into a Minnesota courtroom in 2012, braced to testify at another grueling murder trial.
Jonathan Arden quickly took stock of the case: A 4-month-old boy had collapsed in his father's care and died from lethal head injuries. Damien Marsden, 33, faced decades in prison, accused of shaking the baby to death.
Once, Arden had been a firm believer in Shaken Baby Syndrome, long considered a deadly form of child abuse. But in rural Warren, Minnesota, in April 2012, the former state expert took the stand for the defense, describing how a thin layer of old blood on the surface of the baby's brain was a telltale sign of an injury that had occurred before the baby had been left alone with his father.
Jurors spent less than three hours deliberating before acquitting Marsden of murder.
"A lot of people in this field, especially many of the pediatricians, make statements that are absolute and dogmatic and do not allow for the exceptions that we know exist," Arden told The Washington Post. "Do you want to be involved in somebody's wrongful conviction because you had this dogmatic approach that it must be trauma, it must be shaking?"
Arden is among a number of doctors who once diagnosed Shaken Baby Syndrome but now doubt the science behind it, swayed by more than a decade of research that's documented how diseases, genetic conditions and accidents can, in some cases, produce the conditions long attributed to violent shaking.
The doctors' journeys from supporters to skeptics expose the uncertainty at the heart of a medical diagnosis that has fueled hundreds of abuse and murder cases. In courtrooms across the country, the doubting doctors are now using the same evidence that once supported a shaking conviction -- medical records, autopsy reports and brain scans -- to challenge the diagnosis. The Post chronicled the stories of nine of those doctors through interviews, documents and trial transcripts.
The issue is not whether violent shaking can harm babies: Even doctors who dispute the diagnosis say shaking can damage an infant's fragile neck, torso or spine. But the doctors say that shaking has not been shown to produce the conditions often attributed to Shaken Baby Syndrome -- namely, bleeding on the surface of the brain, bleeding in the back of the eyes and brain swelling.
The challenges have come from doctors and scientists worldwide, including a forensic neuropathologist in Illinois, an ophthalmologist in Colorado, a radiologist in Pennsylvania, a physicist in Idaho, a forensic pathologist in North Carolina, a neurosurgeon in the District, and several doctors in Britain, Sweden, Hong Kong and Argentina.
Although they are outnumbered by the doctors who support the science, those who challenge it are gathering strength. More than a hundred share their ideas on a private email group called "Evidence-Based Medicine and Science."
They have published their concerns in medical journals and teamed up, sometimes as paid witnesses, with private defense attorneys and lawyers affiliated with the Innocence Network. In courtrooms across the country, the doctors have questioned high-profile criminal convictions, drawing attention from journalists at The New York Times, ProPublica and other media outlets. Northwestern University law professor Deborah Tuerkheimer published a book on the subject last year, "Flawed Convictions: 'Shaken Baby Syndrome' and the Inertia of Injustice."
In Fairfax County, Virginia, pediatric neuroradiologist Patrick Barnes, once a well-known state witness in shaking cases, has come to the defense of a mother of two who has so far spent five years in prison. At a widely watched trial, 45-year-old Trudy Mu�noz Rueda was accused of violently shaking a 5-month-old in her home day care in 2009, causing serious brain injuries. She is serving a 10-year prison sentence.
Barnes, a professor at the Stanford University School of Medicine, looked at scans of the baby's head at the request of defense attorneys and came to a different conclusion: The baby had likely suffered from an infection that caused blood clots in the brain, leading to a series of strokes.
"All of the treating physicians simply assumed trauma and stopped looking for alternative explanations," Barnes wrote in a 2012 affidavit. "That is not sound science and cannot be the basis of a reliable prosecution."
A petition challenging the conviction, filed by the Innocence Project at the University of Virginia School of Law, is pending in federal court.
Other doctors have also stepped forward to defend parents and caregivers, including George Nichols, the former state medical examiner of Kentucky, who made a surprising offer at a meeting for public defenders shortly after he retired in 1997.
"I said if they had a case in which I had testified that somebody had died as a result of Shaken Baby Syndrome alone, that they were to contact me and that I would now testify for a reversal," Nichols said. "Shaken Baby Syndrome is a belief system rather than an exercise in modern-day science."
Arden, who spent five years as the District's chief medical examiner before starting his own practice, began questioning the diagnosis a decade ago. Like other doctors, he started asking: If natural causes and accidents could produce the same conditions in babies, how could doctors diagnosis shaking with certainty?
There's no good way to validate the diagnosis. Shaking tests on animals have been inconclusive, and doctors cannot test on babies. Some biomechanical engineers say adults likely cannot generate enough force through shaking to cause the lethal bleeding and swelling, but the injury threshold among infants isn't known.
The National Institutes of Health funded a shaking study on anesthetized baby pigs in 2009, but the research was unable to determine whether shaking can cause the severe brain and eye injuries linked to the diagnosis.
"The jury is still out and we need to do much more basic science," said Valerie Maholmes, who leads a newly formed NIH branch on pediatric trauma. "We want good, strong science that has been confirmed and verified."
Nagged by mounting doubts, Arden on his own started looking into the theory behind Shaken Baby Syndrome, which was forged more than 40 years ago with just a handful of cases and an intuitive leap.
In a bustling hospital in England in the early 1970s, pediatric neurosurgeon A. Norman Guthkelch discovered an alarming pattern among some infants and children: They had blood on the surface of the brain but no external signs of violence to the head.
One explanation came from social workers, who told Guthkelch that parents were disciplining their babies by shaking them. The idea made sense to the soft-spoken father of four, who had long suspected that some parents in northern England considered "a good shaking" far more acceptable than raising a hand to a child.
Guthkelch studied 23 assault cases and found that five children had bleeding on the surface of the brain, known as the subdural space, but no external marks on the head. He proposed that shaking could tear the veins inside the cranial cavity, triggering the bleeding.
In May 1971, Guthkelch detailed his hypothesis in a two-page paper in the British Medical Journal: "One must keep in mind the possibility of assault ... even when there are only trivial bruises or indeed no marks of injury at all."
Over the next few years, prominent pediatric radiologist John Caffey at the University of Pittsburgh, who had spent decades studying head and bone injuries in children, came up with the term "whiplash shaken infant syndrome." He cited case studies of shaking incidents and called for a national education campaign with the mantra:
Guard well your baby's precious head,
Shake, jerk and slap it never,
Lest you bruise his brain and twist his mind,
Or whiplash him dead, forever.
Over time, the subdural bleeding, along with swelling of the brain and bleeding in the back of the eyes, became the classic markers of Shaken Baby Syndrome, though doctors also consider other signs of trauma as well as the accounts of caregivers.
In 1993, the American Academy of Pediatrics, noting the findings of Guthkelch and Caffey, declared Shaken Baby Syndrome "a clearly definable medical condition."
"I am very confident that the science is well validated," pediatrician Robert Block, past president of the academy, recently told The Post. "We have a long history of studying the science, of reviewing the work."
But in the late 1980s, some doctors and biomechanical engineers had already begun to raise critical questions about whether Shaken Baby Syndrome was flawed science.
In 1987, a team of doctors and engineers from the University of Pennsylvania reported in the Journal of Neurosurgery that they had reviewed the cases of 57 children in Philadelphia with suspected shaking injuries, 13 of them fatal. In every fatal case, the shaking symptoms were accompanied by signs of blunt trauma to the head.
Though the researchers acknowledged their study was limited by a lack of information on the response of the infant brain, they concluded, "Shaking Baby Syndrome, at least in its most severe acute form, is not usually caused by shaking alone."
In the late 1990s, concerns about Shaken Baby Syndrome were publicly aired during a contentious trial in Massachusetts, where 19-year-old British nanny Louise Woodward had been accused of killing a baby in her care. The case drew worldwide attention, with one journalist for the British Broadcasting Corp. describing the trial as "the clash of the medical men."
Backed by a series of doctors -- including Barnes, the pediatric neuroradiologist who would later challenge the diagnosis -- prosecutors argued that a vicious assault by Woodward, which included violent shaking and a slam against a hard surface, had left 8-month-old Matthew Eappen with lethal head injuries, including a skull fracture.
But doctors who testified for the defense at the 1997 trial pointed out that the baby had no neck trauma and that the injuries appeared to be several weeks old. One of the doctors was Chicago neuropathologist Jan Leestma, who once was so convinced that Shaken Baby Syndrome was a valid diagnosis that he wrote about it in the first edition of a 470-page textbook he authored on the brain.
"I was wrong," Leestma told The Post. "The original papers that espoused Shaken Baby were basically opinion papers with essentially no science applied to them."
Woodward was found guilty of second-degree murder in October 1997. A month later, Superior Court Judge Hiller Zobel reduced the count to involuntary manslaughter and Woodward was freed from prison.
"After intensive, cool, calm reflection, I am morally certain that allowing this defendant on this evidence to remain convicted of second-degree murder would be a miscarriage of justice," Zobel wrote.
After the trial, 50 doctors wrote a letter to the journal of the American Academy of Pediatrics asserting that the argument against shaking made by Woodward's defense was "inaccurate, contrary to vast clinical experience and unsupported by any published literature."
"The Shaken Baby Syndrome (with or without evidence of impact) is now a well-characterized clinical and pathological entity with diagnostic features in severe cases virtually unique to this type of injury," the doctors wrote.
Four years after the Woodward trial, skeptical doctors from Minnesota to North Carolina to Britain began to challenge each of the conditions linked to shaking.
The most obvious symptom was the subdural bleeding, which had long been attributed to tears in the bridging veins that run from the surface of the brain to larger veins underneath the skull.
In 2001, a team of doctors in Britain reported that it had studied the brains of several dozen infants with inflicted head trauma, including those who were allegedly shaken, and found the subdural bleeding "trivial."
In research papers that followed, Pennsylvania State University radiologist Julie Mack and other doctors suggested that the slight bleeding could have come from millions of much smaller blood vessels, similar to capillaries. Violent shaking was an unlikely cause of the bleeding because the tiny vessels were connected to the skull and would not shift from shaking.
"Every time we carefully examine the anatomy, we learn things," said Mack, who has testified for the defense at shaking trials.
Doctors who support the Shaken Baby science criticize the studies, arguing that bridging veins are thin and can break easily. "That's an interesting hypothesis, but there's no proof," said Carole Jenny, a pediatrician at the University of Washington.
Mack countered: "It is an established anatomic fact that the vessels can bleed without trauma and produce a subdural."
Doctors have also investigated the origins of retinal hemorrhages, which Shaken Baby experts say have been observed in a majority of children with shaking injuries. A 2002 guide to investigating child abuse from the Department of Justice's Office of Juvenile Justice and Delinquency Prevention called retinal hemorrhages "conclusive evidence of Shaken Baby Syndrome in the absence of a good explanation."
But in 2001, North Carolina forensic pathologist Patrick Lantz began to take a closer look at the conditions that can produce the bleeding. Lantz, a state medical examiner who once supported the Shaken Baby diagnosis, examined the eyes of more than 180 deceased children, including 116 infants. He found retinal hemorrhages in babies who had died from an obstructed airway, congenital heart disease and infections such as encephalitis and meningitis.
"If doctors see retinal hemorrhages, they say it's abuse," said Lantz, who has published four studies on the issue. "But it's as scientific as a fortuneteller reading tea leaves."
Jenny criticized Lantz's findings, saying the bleeding he documented wasn't as copious as the hemorrhaging seen in Shaken Baby cases. But Lantz said that the more extensive blood patterns are not exclusive to shaking. "If you look at other cases, you see them," he said. "They're there, but if you don't look, you won't find them."
As doctors challenged the conditions linked to shaking, biomechanical engineers started asking whether accidental falls from sofas, beds, changing tables and playground equipment could trigger the same deadly injuries. The results raised even more questions about the science behind the diagnosis.
Engineers experimenting with crash-test dummies found that "short falls" produced far more acceleration to the head than shaking. The Washington Post commissioned its own test late last year in an engineering lab outside Detroit. The study was conducted by biomechanical engineer Chris Van Ee, an accident reconstruction specialist who has been hired as a defense expert in cases alleging child abuse.
In the tests, a 195-pound man vigorously shook a 22-pound crash-test dummy. The dummy also fell from a couch.
The shaking generated a peak between 6 to 8 Gs, or as much as eight times the acceleration due to gravity or twice what would occur during jumping jacks, compared to a peak of 112 Gs when there was direct impact to the head during a fall.
Through accident reconstruction, biomechanical modeling, cadaver studies and scaling from animals and adults, engineers have estimated that the threshold for serious injuries in babies is about 50 to 90 Gs. But since researchers cannot test on infants, the actual threshold remains unknown.
"It may be impossible to replicate the perfect experiment," said Susan Margulies, a bioengineering professor at the University of Pennsylvania.
Doctors who support the diagnosis criticize the tests, saying that dummies are not babies and the results add little to the debate. They point to a 2008 paper by California pediatrician David Chadwick and others, which tracked research and cases on lethal falls and concluded that "such deaths are rare" -- less than 1 in 1 million. The doctors also say the pattern of injuries from falls is often different from that of shaking, with bleeding in different parts of the brain.
Doctors who have challenged Shaken Baby Syndrome point out that a 2001 study using data from the U.S. Consumer Product Safety Commission documented the cases of 18 children, ranging from 1 to 13 years old, who died after reportedly falling from playground equipment. Most had brain swelling and bleeding on the surface of the brain. Of the six children who had their eyes examined, four had extensive retinal hemorrhages.
The research was done by forensic pathologist John Plunkett, a longtime medical examiner in Minnesota who supported the Shaken Baby diagnosis until the late 1990s.
"I believed that shaking was a viable and legitimate diagnosis," he said. "That's the way I was taught, but I wasn't smart enough to understand the mechanics of injury. That's when I started to learn the science."
Supporters of the science have criticized Plunkett's study, saying among other things that it did not include any infants and that several children had pre-existing conditions that may have increased the risk of death.
But Plunkett's study and others turned a series of doctors into doubters, including Peter Stephens, a former deputy medical examiner in Iowa and forensic pathologist for more than 40 years.
"At best, it is an overused diagnosis," he said. "I won't say that shaking can't cause serious injury and death, but it's never been shown to my satisfaction, and it's never been proven."
One of the most high-profile doubters is Guthkelch, whose research on shaking 40 years ago helped spark a new diagnosis for child abuse.
Guthkelch got involved in the issue again in 2011 when he was asked to review the conviction of a young Phoenix man who was accused in 2000 of shaking his girlfriend's 5-month-old son to death. Drayton Witt was serving 20 years in prison.
Guthkelch looked at the baby's medical records and discovered that the infant, born with the umbilical cord wrapped around his neck, had struggled with pneumonia, dehydration and seizures. Guthkelch wrote a declaration to the court, saying there was "simply not enough evidence" to justify a Shaken Baby diagnosis.
Several other doctors brought in by Witt's defense attorney concluded that the baby had likely died from natural causes, such as a blood clot in the brain. The medical examiner who originally ruled the death a homicide said he could no longer stand by his conclusion, writing to the court that multiple conditions can "create the very symptoms and injuries once thought to be nearly exclusively attributable" to Shaken Baby Syndrome.
Witt was released from prison after 10 years.
"I am doing what I can so long as I have a breath to correct a grossly unjust situation," said Guthkelch, now living in Ohio and nearing 100. "I think they've gone much too far."
To Arden, the issue boils down to uncertain science.
Coming up as a young medical examiner in New York in the 1980s, he was taught that bleeding and swelling in the brain, in the absence of any other symptoms or injuries, was clear-cut evidence of shaking. He testified for the state in at least two dozen Shaken Baby cases.
Arden continued to support the diagnosis when he moved to the District in 1999 to take the helm of the city's long-troubled medical examiner's office.
He left after five years in 2003 when city officials complained about high staff turnover and management problems and several female employees accused him of harassment. Arden said the complaints were unfounded, leveled by disgruntled employees who had resisted reforms.
Arden opened a private practice in McLean, Virginia, and was elected to the executive committee of the National Association of Medical Examiners. He started reading new studies on Shaken Baby Syndrome. He learned that Lantz in North Carolina had found bleeding in the eyes of children who had died of natural causes and Plunkett in Minnesota had found bleeding and swelling in children who had fallen from playground equipment. Other studies found that some infants developed bleeding on the surface of the brain following full-term births.
Arden started testifying as a retained witness for parents and caregivers accused of abuse, including Damien Marsden, the father in Minnesota who faced three counts of murder.
In 2009, Marsden was home alone with his 4-month-old son, Rylin, while the baby's mother went shopping. When she returned 30 minutes later, she said Marsden was just coming in from outside, where he had been grilling hamburgers. They said the baby vomited in his bouncy chair and turned limp.
Emergency workers rushed Rylin to the hospital, where a CT scan revealed a swollen brain, bleeding on the surface of the brain and a skull fracture. He died six days later.
Marsden said Rylin had fallen out of bed onto a carpeted floor the day before. But doctors concluded that Rylin's injuries could not have come from the fall. A family doctor with training in child abuse said that Rylin had both "contact and "rotational injuries," as well as a bruise on the inner ear, records show.
The medical examiner noted the baby had a retinal hemorrhage and a contusion on the back of his head, as well as the internal injuries, and concluded that Rylin had died of traumatic head injuries from a physical assault.
Marsden was charged with murder.
"It was a horrible experience to lose my child and then to get blamed for it," he told The Post. "You're basically guilty until you're proven innocent."
As the trial neared, Arden studied the autopsy report, medical records, slides of the baby's brain and, most important, the CT scan of Rylin's head. He found both fresh and old blood on the surface of the baby's brain as well as signs of an impact injury.
"Invoking shaking/rotational injuries is both unnecessary and speculative," he wrote in a report to Marsden's attorney.
At the 2012 trial, the baby's grandmother testified that the infant had been accidentally dropped onto a wood floor at her day-care center about a week before the collapse. Arden told jurors that Rylin had likely died from the two falls. Marsden was acquitted.
He's now married to Rylin's mother, working at an oil company and raising three children.
"My thinking has definitely evolved over time," Arden said. "I haven't joined the camp that says there is no such thing as Shaken Baby, but I am much more cautious and circumspect about invoking it.
"What could be right about getting it wrong?"
• Dan Bauman with the Medill Justice Project at Northwestern University, Cathaleen Qiao Chen and Elizabeth Koh with American University's Investigative Reporting Workshop, and Washington Post researchers Alice Crites and Magda Jean-Louis contributed to this report.