A diagnosis the science has largely abandoned is still deciding who gets to be a parent and who goes to prison, and the country has just watched it die on the record in open court.

Danica Rue, executive director of investigations at the New Jersey public defender's office, said of the state supreme court's ruling: "I would certainly hope that people are able to now seek justice where it was not meted out earlier."

In November 2025, the New Jersey Supreme Court did something no state high court had ever done: it ruled that expert testimony about shaken baby syndrome, the half-century-old diagnosis that a violent shake is sufficient to bleed an infant's brain, is too unreliable to be heard at trial. Six justices to one, the court held that the "triad of symptoms" so long treated as a fingerprint of shaking, bleeding around the brain, bleeding in the eyes, and brain swelling, does not have the scientific standing the law requires. The case, State v. Nieves, was about two fathers who were never going to be tried, but the consequence reaches every parent in the country who ever called 911 for a baby that stopped breathing.

The ruling is a victory. It is also, on its own terms, a correction. New Jersey has been the last state to admit what the rest of the evidence has known for years: the triad is not proof of shaking. Strokes, seizures, short-distance falls, birth injuries, bleeding disorders, and congenital malformations can produce the same findings. A 2024 peer-reviewed review of the literature put it plainly: the triad "does not meet the standards of evidence-based medicine," and the shaking hypothesis has never been validated by the biomechanical science it claims to rest on. Even the neurosurgeon whose 1971 paper gave the diagnosis its name has, in retirement, called triad-based prosecutions a mistake and said the field "needs to go back to the drawing board."

And yet the diagnosis keeps working. Last month, an Illinois court finally dismissed the case against Nathaniel Onsrud, who had spent thirteen years in prison for the murder of his infant son Dax, a 10-week premature baby with a lung monitor and a feeding tube. A new autopsy review by the former chief medical examiner of New York City concluded that the child had died of natural causes tied to lung disease and prematurity, and that the clots in his brain had formed days before his death, before anything Onsrud allegedly did could have happened. "The autopsy evidence shows that Mr. Onsrud confessed to a homicide that had not occurred," the pathologist wrote. The state's attorney's office admitted it had lost documents in the case. The judge who sentenced Onsrud in 2008 told him, to his face, "I've no remorse for you."

That is the shape of this story, and it repeats. Forty-four people have now been exonerated in shaken baby and abusive head trauma cases, the National Registry of Exonerations counts. In Texas, an autistic father on death row, Robert Roberson, was one state legislative hearing away from execution in 2024 when lawmakers, after years of pressure, intervened. In Las Vegas, a mother was cleared of murder in 2022, after her baby's death was traced to sickle cell trait. In Georgia, a new trial was granted this July in part because a judge said the diagnosis is no longer a "presumptive diagnosis." The pace of the corrections has more than doubled: from about one overturned conviction a year in 2008 to roughly four and a half a year in the most recent period tracked.

The diagnosis is now so central to American child-prosecution practice that a name change in 2009, from "shaken baby syndrome" to "abusive head trauma," was quietly designed to make the legal challenges stickier. The American Academy of Pediatrics explained the switch as one of precision, a broader term that covers both impact and shaking. Critics, including the trial attorney who has written the leading book on the diagnosis's forensic unreliability, say the rebrand accomplished something else: it let a contested theory keep operating under a new label while the old name was being shredded in court. "Shaken baby syndrome is alive and well but mostly operates under an alias," Randy Papetti wrote.

At the same time, the professional community that polices the diagnosis has organized its hostility toward anyone who questions it. Last year, three prominent child abuse pediatricians published an opinion in JAMA Pediatrics describing physicians who raise doubts about AHT as "denialists" whose "misinformation may harm children." At the field's own conference, a deputy district attorney from Milwaukee addressed a sympathetic crowd of child abuse specialists, social workers, police, and prosecutors, and asked whether defense medical experts who rebut the diagnosis are "practicing medicine" or "providing reasonable doubt for sale."

Translation: in this ecosystem, the person doing the doubting is the one being treated as the problem, and the institution that is wronging people is the one being defended.

Meanwhile, the cases keep getting built. ProPublica recently profiled a young couple in Blue Ash, Ohio, whose two-month-old son was rushed to the hospital with subdural bleeds after a difficult delivery that left a visible hollow above his forehead. Three independent doctors the parents retained, a pediatrician, a neurologist, and a radiologist, concluded the bleeding was chronic and likely from birth. A child abuse pediatrician at Cincinnati Children's wrote a report that pointed the other way, and the machinery followed: custody petition, search warrant, arrest, a court order barring the father from being alone with his own children. The baby, now healthy, is a toddler at home. The father still faces twelve years in prison if convicted, and his family is still living under court-ordered supervision, cameras in the house, a caseworker who appears without warning. "We have cameras in every room," his wife said. "The nightmare is that I'm found alone with my children."

There is a version of this story in every state, in a different county, usually with a different race and a different income level, and almost always with the same final shape: a child collapses, a doctor names a diagnosis, a detective arrives, and the last person who held the baby becomes the suspect of last resort. When the child is sick in ways nobody caught earlier, when the family is poor, when the mother is a single parent working nights, the probability that the accusation sticks goes up. That is not conspiracy. That is how a system that has never been stress-tested behaves under load.

The New Jersey ruling is real, and it is narrow. It applies to cases where the only basis for a shaking diagnosis is the triad, where there is no evidence of impact, no bruising, no fracture, no other corroborating evidence of abuse. The majority opinion left the door open: if new, reliable science validates the mechanism, the testimony can come back. The dissent, forty-four pages by a single justice, argued that every other state in the country still admits such testimony, and that judges should not be doing the work of scientists.

Both are right about the mechanics. The question the ruling leaves on the table, and the one no court has yet been willing to ask, is a simpler one: how many people are sitting in prison right now, in which states, on the basis of testimony that the science has already walked back? The New Jersey public defender's office says there are "hundreds of people, perhaps more" in the state who were prosecuted under the theory and are now potentially eligible for relief. Other states have no such trigger. They have only the slow, private work of innocence projects finding one file at a time, in a county where the records were kept and the judge is still sitting.

The question is not whether the diagnosis was wrong. The question is why a criminal justice system could rely on it for fifty years, and still is.