“I tucked Joel in, but I feel so guilty I didn’t hold him longer,” Julie Rea said, her voice welling with emotion. That is all she can muster about the worst night of her life. As she tries to say more, she breaks down.

The story remains, still, almost unspeakable. In the early morning hours of Oct. 13, 1997, Rea was jolted awake by a scream. She discovered an intruder, but saw no sign of her son, in her Lawrenceville, Illinois, home. She told police that she struggled with the man, who fled. Then ran for help. But it was too late. Her son, 10-year-old Joel Kirkpatrick, had been stabbed to death.

At the time of the murder, Rea was a single mother working toward a doctorate in educational psychology. She had divorced Joel’s father three years earlier and was leading a quiet, uneventful life in the wake of a turbulent marriage. The mild-mannered daughter of missionaries, Rea had devoted herself to her bright, inquisitive son.

But in 2000, after a protracted and deeply flawed investigation, Rea was charged with killing Joel.

“Surviving your child’s murder, only to find out that you’re being accused of murdering your child, is a kind of trauma that I wouldn’t wish on any living being,” said Rea, now 50. “I wouldn’t wish it on a snake.”

Prosecutors used a forensic discipline called bloodstain-pattern analysis to argue that an intruder never entered her home on the night of the crime and that Rea was, in fact, her son’s killer. She was convicted of first-degree murder in 2002 largely on the strength of the testimony of two bloodstain-pattern analysts.

Four years later, Rea was acquitted at a retrial, after a legal team assembled by the Center on Wrongful Convictions at Northwestern University’s Pritzker School of Law in Chicago mounted a vigorous defense that challenged the state’s forensic testimony. They also presented new evidence that a serial killer of children — a lifelong drifter who was on Texas death row for a nearly identical crime — had confessed to killing Joel. Rea was formally exonerated in 2010.

Today, she belongs to a growing community of victims: Americans who were wrongly convicted with the help of forensic disciplines allowed into courtrooms despite little to no proof of their reliability. Of the 362 people who have been exonerated based on DNA tests in the United States, faulty forensics contributed to almost half of the underlying convictions.

Like Rea, these exonerees have had years of their lives stolen, and many have struggled to find their place in the world after surviving the crucible of incarceration.

A year before Rea’s exoneration, the National Academy of Sciences had released a report that called into doubt the reliability of bloodstain-pattern analysis. Practitioners’ conclusions were often “more subjective than scientific” and open to “context bias,” the 2009 report said.

“Some experts extrapolate far beyond what can be supported,” it said. And it cautioned, “The uncertainties associated with bloodstain-pattern analysis are enormous.”

The report criticized a wide range of forensic disciplines, including the analysis of hairs, fibers, bite marks and shoe and tire impressions. Its authors found that many of these disciplines were not grounded in hard data and extensive, peer-reviewed research, but instead relied on practitioners’ personal interpretations. “The law’s greatest dilemma in its heavy reliance on forensic evidence,” it stated, “concerns the question of whether — and to what extent — there is science in any given forensic science discipline.”

The report called for sweeping reform. Yet nearly a decade later, little has changed. In the field of bloodstain-pattern analysis, rigorous research that might determine the accuracy of analysts’ findings is scant. Bite-mark analysis — which, in 2016, the President’s Council of Advisors on Science and Technology declared had no scientific validity — is still admitted in criminal prosecutions. So, too, is microscopic hair comparison, an outmoded and dangerously flawed technique that has, to date, led to the convictions of 75 people who were later exonerated by DNA testing.

“Forensic science should be treated like any other consumer product,” said M. Chris Fabricant, director of strategic litigation at the Innocence Project in New York City. “Before it’s allowed to be used on human beings, it should be scientifically tested and clinically demonstrated to be reliable, just like toothpaste.” But as long as courts continue to admit forensic evidence proffered by prosecutors, and prosecutors continue to win convictions using it, he added, “There is no incentive to change.”

Bad science also affects defendants whose cases never go to trial. About 95 percent of criminal cases are resolved with plea bargains, and in such cases, forensic work often goes unchallenged. “The mere threat of scientific evidence being used in court is often sufficient to induce a plea,” Fabricant said.

Last year, then-Attorney General Jeff Sessions dashed hopes for reform when he disbanded the National Commission on Forensic Science, an Obama-era advisory panel that sought the input of research scientists in improving the standards and soundness of forensic analysis and testimony.

Meanwhile, defendants convicted on the strength of dubious forensic testimony languish in prison, while those who have won their freedom, like Rea, try to put their lives back together. The $87,057 in compensation she received from the state of Illinois has proven to be cold comfort. Struggling with PTSD and reeling from abuse she endured behind bars, she continues to suffer in ways both unbearably public and excruciatingly private.

Being branded a child killer “separated me from society,” Rea said. “I’ll always have a scarlet letter. Nothing will ever make that go away.”

---

On the morning of Oct. 13, 1997, a neighbor, Lesa Bridgett, awoke around 4:30 to the sound of Rea pounding on her front door. Rea was standing outside, barefoot, clad only in a T-shirt and underwear. She looked frantic, and she was speaking so rapidly that she was hard to understand. Hoping to calm her, Bridgett pulled the 28-year-old into a bear hug. “Joel’s gone,” Rea shrieked.

Court records show she told police she’d run into Joel’s room after hearing a scream and saw only his empty bed. A man in a ski mask lunged at her, and Rea — who had a black belt in taekwondo — struggled with him, chasing him through the house and into the backyard. He slammed her head to the ground before making his escape. Hysterical, she told Bridgett she thought Joel had been kidnapped.

But when Lawrence County Deputy Sheriff Dennis York arrived and searched Rea’s home, he found Joel. The fifth-grader lay on his bedroom floor, between his empty bed and the nearby wall. His pajamas were drenched in blood. He had been stabbed 12 times, twice through his aorta.

When Rea was told that her son was dead, she erupted into screams.

She was taken to the hospital with a black eye, rug burns on her knees, a wound on her right arm that required stitches, and scratches and abrasions on her head, both shoulders and the tops of her feet. In her fragile state of mind, she did not grasp that investigators were already eyeing her as a suspect. They had not been able to find any obvious signs of forced entry into the home; Rea could not remember locking the back door but told police it was safe to assume she had.

Rea hardly fit the profile of a woman who would drive a knife into her child’s heart. She had no history of mental illness and no criminal record, nor did her behavior in the hours leading up to his murder suggest anything was amiss. She and Joel had spent an unremarkable evening at home with her best friend, Trena Woodward, and Woodward’s children. The two women had spent the evening scrapbooking.

A plodding investigation followed, with few leads. The murder weapon — a steak knife pulled from a butcher block in Rea’s kitchen — did not yield any fingerprints. Investigators had little to go on, having failed to do basic police work that might have pinpointed the identity of an intruder. They never dusted Joel’s bedroom, or the butcher block that the knife was pulled from, for fingerprints and did not preserve critical trace evidence on Joel’s bedspread.

Instead, they focused on Rea. Looking for any remnants of blood that she might have tried to wash away, investigators dug up her septic tank, inspected her sink and shower drains, and examined the clothes in her washing machine. They sprayed luminol, a chemical that detects the presence of blood, around her house. They found nothing.

Nor could they find any discernable motive. Rea, who shared custody of Joel with her ex-husband, Leonard Kirkpatrick, had fought for more time with their son. Kirkpatrick had been awarded physical custody of the boy after he remarried — a decision Rea had fiercely protested, filing her latest appeal two weeks before her son’s death.

In the wake of the murder, she could not bear to be alone. Terrified of the dark, she rarely slept. She stayed in an apartment an hour-and-a-half’s drive away, in Bloomington, Indiana, where she had been commuting to Indiana University’s Ph.D. program. There, a rotation of friends stayed with her around the clock. “To fall asleep, I had to have someone on either side of me and the lights on,” she said.

She got a German shepherd that she took with her nearly everywhere. She managed to hold down a teaching position at the university but was plagued by panic attacks and debilitating anxiety. Several times, investigators showed up to her home and workplace unannounced.

The investigation sputtered along until 2000, when the Illinois State’s Attorneys Appellate Prosecutor’s Office, which routinely handles complex prosecutions for rural counties, took the case to a grand jury. On the eve of the third anniversary of Joel’s murder, Rea was indicted.

Prosecutors had no eyewitnesses, little forensic evidence and no motive. They would rely on a forensic discipline that had helped win circumstantial cases before: bloodstain-pattern analysis.

Bloodstain-pattern analysis is rooted in the belief that the drops, spatters and trails of blood at a crime scene are rich repositories of information that can be used to reverse engineer and even reconstruct crimes. Many practitioners are law enforcement officers, not scientists. But their expert testimony often lends a sense of scientific certainty to circumstantial cases, and it can be a powerful tool in the hands of prosecutors.

“An expert who says, ‘This is what the physical evidence shows,’ is extremely persuasive, especially in a circumstantial case,” said Judy Royal, a staff attorney with the Center on Wrongful Convictions who began working on Rea’s case in 2003. “Jurors don’t understand when an expert is overstating findings or going beyond what can be tested and replicated.”

This fact has been laid bare in cases around the country in which bloodstain-pattern analysis has figured prominently — most recently that of Joe Bryan, a former high school principal who was convicted of the 1985 murder of his wife in Clifton, Texas. Bryan was the subject of a two-part investigation by ProPublica and The New York Times Magazine this spring that questioned the accuracy of the bloodstain-pattern analysis used to convict him. The Texas Forensic Science Commission, a national leader in criminal justice reform, found in July that the conclusions of the prosecution’s bloodstain-pattern expert, Robert Thorman, were “not accurate or scientifically supported.” In September, Thorman conceded in an affidavit that some of his conclusions were wrong. Bryan, whose attorneys have asked that his conviction be overturned and whose case is before Texas’ highest criminal court, has spent 31 years in prison. A judge recently recommended that Bryan’s conviction stand, finding that “Thorman’s testimony was not important to the case.”

Bryan’s and Rea’s cases are not the only ones in which a bloodstain-pattern analyst has delivered testimony that was later called into question. From Oregon to Texas, from North Carolina to New York, convictions that hinged on the testimony of a bloodstain-pattern analyst have been overturned and the defendants acquitted, or the charges dropped. As recently as this February, a judge vacated the conviction of a Missouri man named Brad Jennings for the 2006 murder of his wife, Lisa, after evidence emerged that supported his claim that his wife committed suicide. Jennings was released from prison after eight years behind bars.

Bloodstain-pattern analysis’s shortcomings were most clearly on display in the case of an Indiana state trooper named David Camm, who found his wife and two children shot to death in their home in 2000 and was quickly charged with their murders. Over the course of three trials, prosecutors presented a succession of bloodstain-pattern analysts who testified that eight specks of blood found on the T-shirt Camm wore on the night of the crime were “high-velocity impact spatter” from the shooting. The defense produced its own bloodstain experts, who argued that the specks in question were actually “transfer stains” — blood that blotted Camm’s T-shirt as he tried to render aid.

Camm was acquitted in 2013 after spending 13 years behind bars. Another man, a burglar with a long rap sheet whose DNA was found at the crime scene, was convicted of the murders and remains in prison.

The retired biochemist and forensic scientist Robert Shaler, one of the authors of the 2009 National Academy of Sciences report and the founding director of Pennsylvania State University’s Forensic Science Program, testified for the defense in the Camm case that experts on both sides were incorrect; specks of blood, he argued, was too little information from which to draw any meaningful conclusion. “The experts couldn’t even agree on the type of pattern they were looking at,” he said in an interview.

Camm’s lead attorney at his third trial, Richard Kammen, was even more blunt. “People see what they want to see,” he said of bloodstain-pattern analysis. “It’s as accurate as a Ouija board.”

---

When Rea stood trial for her son’s murder in February 2002, bloodstain-pattern analysis took center stage. Like Camm’s case, Rea’s case hinged on a minuscule amount of blood that stained her T-shirt on the night of the slaying. Nearly all of it was her blood, which had seeped from the wound on her arm. But a tiny smear on her shirt’s right shoulder was found to be Joel’s blood. DNA testing of two other small bloodstains on her shirt established a possible, though not definitive, link to Joel. The defense would argue that Joel’s blood had likely transferred onto Rea’s shirt when she scuffled with the bloodied intruder.

The prosecution called two bloodstain-pattern analysts who asserted that there had been no intruder. The first was Rod Englert, a retired police detective and past president of the International Association of Bloodstain Pattern Analysts. He began by performing a lengthy demonstration using theatrical blood to show the jury how different kinds of blood spatter are created. Holding forth without interruption, he introduced the forensic discipline’s basic principles and lexicon. Bloodstain-pattern analysis, he explained, allowed investigators “to put the pieces of the puzzle together.”

It was with this imprimatur of legitimacy that Englert delivered his findings. He concluded that the crime scene had been “staged and manipulated” and was “not consistent with her story of a struggle.” His expert opinion was based on both his examination of the evidence, he said, and his “experience of having gone to many, many scenes, many like this, hundreds of scenes.” He saw “no indication that there was a third party in this residence,” he added.

The task of interpreting the bloodstains on Rea’s T-shirt was left up to Dexter Bartlett, a crime scene investigator with the Illinois State Police. Bartlett told the jury that Joel’s blood could not have transferred onto her shirt and was more consistent with her wielding a weapon. Bartlett offered no proof of this — no experiments he had conducted, no data and no explanation of the methodology he used to come to his conclusions.

The defense called its own bloodstain-pattern analyst, a forensic consultant named Paul Kish, who testified that his microscopic examination of the T-shirt led him to believe Joel’s blood had transferred onto it. The blood rested on top of the fibers, he told jurors; it had not penetrated the weave of the fabric, as blood cast off of a weapon would be expected to do.

Who was right and who was wrong was left to jurors to sort out. Rea’s public defender, Brad Vaughn, implored them to remember that none of the claims made by the state’s bloodstain-pattern analysts could be verified; they had not conducted any experiments to test their conclusions. “That’s not scientific,” Vaughn told the jury. “Science is about establishing: ‘This is scientific and true because I can recreate it. I can duplicate it. I know within a range that this can be done this way.’ There was no attempt to do that.”

The other prosecution strategy was to cast Rea – who friends and family described as a caring, attentive parent – as a bad mother. Her ex-husband testified that she had considered having an abortion when she discovered, at the age of 17, that she was pregnant with Joel. Neighbors described her behaving erratically, even coldly, in the immediate aftermath of the crime – with one neighbor telling the jury that Rea, though initially hysterical, never cried. These claims went largely unchallenged because Rea’s attorney didn’t put her on the stand.

In the end, the jury found Rea guilty. She was sentenced to 65 years in prison.

Then in 2004, there was a sensational development in the case: Serial killer Tommy Lynn Sells told Illinois investigators that he had broken into a house in 1997, stabbed a boy to death using a knife he pulled from a butcher block in the kitchen and sparred with a woman both inside and outside the house. Sells was facing execution in Texas for fatally stabbing a 13-year-old girl after breaking into her family’s home, and he had been linked to several other murders of children around the country. That same year, an Illinois appellate court overturned Rea’s conviction because of a legal error and ordered a new trial.

At her 2006 retrial, prosecutors doubled down on bloodstain-pattern analysis when they once again put Englert on the stand. But this time, Englert — who had delivered some of the state’s most damning testimony in the Camm case — focused on a theory he had never mentioned at Rea’s first trial. In examining her T-shirt, he said, he had seen evidence that “a bloody hand pushed up” against it. The source of the blood, he said, could only have been “Joel’s bloodied hands.” His testimony suggesting a horrifying scenario: that Joel had tried to push his mother away as she plunged a knife into him. Englert admitted under cross-examination that he had never requested measurements of Joel’s hands for comparison.

Englert, who stands by his testimony, said recently that any differences in his testimony between trials was likely the result of variations in the prosecutor’s questions.

Testifying for the defense, forensic consultant Kenneth Moses challenged whether Englert had used the scientific method to arrive at his conclusion. “I have seen nothing to indicate that this area was studied for that purpose, and data collected, comparisons made, and some sort of a conclusion or hypothesis drawn,” Moses said, referring to the T-shirt. “There is no scientific basis for making such a claim.”

The defense went on to argue that Rea’s injuries on the night of the murder could not have been self-inflicted, and that investigators had missed clues that pointed to an intruder. They also played the audio of Sells’ confession for the jury. The lifelong drifter had confessed to numerous other crimes, and some of his accounts strained credulity. But the defense bolstered its case by introducing evidence suggesting that Sells had been near Lawrenceville, Illinois, around the time of Joel’s murder. Most powerfully, Rea took the stand to tell her story, insisting that she had “absolutely not” killed her son.

When the jury returned a not guilty verdict on July 26, 2006, Rea let out a primal cry. Her knees buckled under her as she fell to the floor.

---

“I thought everything would be better,” Rea said. “I thought I might go to law school, or finish my doctoral degree. I thought I’d be strong and able to help other people and give back to my family and friends who had given so much to me. I just had no idea what it would be like.”

Instead, at the time of her acquittal, Rea was still reeling — not only from her son’s violent death and two criminal prosecutions, but from the trauma of incarceration. She is reluctant to discuss the physical and emotional abuse she endured during the nearly four years she spent behind bars except to say that she was reviled by both inmates and guards. “What I had supposedly done was considered to be the one unforgivable sin,” she said.

Despite her acquittal, she soon discovered that she had to keep living under the weight of suspicion. Prosecutors in the case still spoke of her like a criminal. Edwin Parkinson, the lead prosecutor in her case, told reporters, “The jury found her not guilty; they did not find her innocent.” Parkinson did not respond to a request for comment.

Despite intensive therapy, her faith and a service dog to help ease the effects of her PTSD, Rea struggled. Marriage to a man she wed shortly before her first trial, who believed so fervently in her innocence that he went to law school to aid in her defense, unraveled in the years that followed her acquittal.

She also had trouble finding a job. Knowing her previous conviction would surface one way or another, she was up front with prospective employers. “After the person interviewing me picked their jaw up off the floor,” she said, “the job would be offered to someone else.”

Though she was formally exonerated in 2010, when an Illinois circuit court granted her a certificate of innocence, she had difficulty finding work.

In 2011, Rea moved to Tennessee, where her parents live, hoping for a fresh start. She tried a number of ventures, including opening a bed-and-breakfast, but none came to fruition. She is currently unemployed and apprehensive about the future. “I have no income right now,” she said. “I don’t know what to do.”

She is painfully aware that what happened to her — and to Camm, Jennings and others wrongly convicted on the strength of bloodstain-pattern analysis — changed nothing. Prosecutors continue to use bloodstain-pattern analysis in even the most circumstantial of cases, and judges keep allowing it into the courtroom, largely unexamined.

A working group of forensic scientists and bloodstain-pattern practitioners, which is overseen by the National Institute of Standards and Technology, has recently called for stronger standards and drafted new guidelines for bloodstain-pattern analysts.

But the most significant reform efforts that are currently underway have been brought about by the Texas Forensic Science Commission, whose decisions often prompt other states to follow suit. The commission has determined that bloodstain-pattern analysis that is used to reconstruct a crime must be performed by an organization or individual who is accredited — not anyone who is a self-styled expert — if it is to be allowed in a Texas court. The group also has plans to educate the state’s criminal court judges by creating a resource they can consult that will describe each forensic discipline’s standards and limitations, as well as potential red flags.

“When judges let everything in because they don’t have the tools to do better, we are in a situation where the only thing stopping overstated testimony is the analyst’s own ability to exercise self-restraint,” said Lynn Garcia, the commission’s general counsel. “When life, liberty and justice for the victim is at stake, that is nothing short of terrifying.”

For Rea, any reform will come too late. “Joel still needs justice,” she said. “He’s never seen a day of it. My acquittal just got us back to level ground.” She deplores the fact that Sells, who she believes killed Joel, will never be charged with his murder. Sells was executed in 2014.

Over and over again, Rea has wished that she had awoken sooner on the night Joel was murdered. Then she could have saved her son, she believes, or died trying.