The face of an attacker should be seared into memory forever. How could you forget the person who raped, stabbed, or shot you?
Or so crime victims — and, often, jurors — believe.
But as scientists and criminologists have documented, memory is often an unreliable resource, especially when the attacker and the victim are of different races. And advances in DNA analysis in the last 30 years that have brought a harvest of exonerations have underscored the pitfalls of overreliance on eyewitness testimony.
Misidentifications can be reduced, and New Jersey led the way when it reformed police procedures 16 years ago. That was spurred by cases such as that of David Shephard, who spent more than a decade in prison after being convicted of rape and robbery.
But most states, including Pennsylvania, have been slow to follow New Jersey's lead. And accusations can cast a long shadow.
Just ask Shawn Yarbray, who was acquitted by a jury last week after being charged with the May 2017 attack on Philadelphia City Councilman David Oh. Even after the acquittal, Oh remains convinced that Yarbray was the man who stabbed him in the ribs and slashed one of his arms. Oh continues to publicly accuse Yarbray of the attack.
Or ask Robert Williams, who spent three years in jail waiting to be tried for a Germantown corner-store robbery and assault before being acquitted in February by a jury that deliberated for just 73 minutes. Paranoia stays with Williams, who has since found work with a food-delivery company and is making plans to go to divinity school.
"It's scary, because I never know when the police may be at my door," he said. "They took me from my family, it was traumatic." At the time of his arrest, Williams had a 2-year-old son and a pregnant wife. The family had bought a home in Strawberry Mansion. "There's always a chance the police are upset about the verdict, and they may come and try to pin something else on me."
Shephard expressed similar anxieties. DNA evidence eventually cleared him, but the time locked up disrupted his life, and the psychological and emotional toll persists.
"You never move past it," he said.
Besides the personality changes — Shephard is quicker to anger, less trusting — he fears being falsely accused again. He dislikes going out alone, wanting a witness with him at all times. It took him more than a year to stop collecting bus receipts every time he went out, adding them to the box under his bed in case he might have to verify an alibi.
In the cases of Shephard, Yarbray, and Williams, the victims identified them as the alleged attackers, and the victims and the accused were of different races. Police and prosecutors had little other evidence.
"There's a fundamental conflict in how law is made in the courts and what scientific research shows," said Karen A. Newirth, a senior staff attorney at the Innocence Project who focuses on legal reform efforts around eyewitness identification and confessions.
The science is well established. Gary Wells, a psychology professor at Iowa State University who has studied the issue for more than four decades, described misidentification as Psychology 101:
"A fearful stimulus, you don't stand there looking at it and trying to form a good memory of it. Those of our ancestors who upon seeing a bear or something stood there going, 'Look at that nose, look at those eyes, look at those paws. …' Our genes now have programmed us: Fight or flight."
But by not incorporating that science into the criminal justice system, innocent people sometimes are charged and convicted, families and lives are torn apart, real criminals go free.
And although no system is perfect, experts said, the reforms such as those instituted in New Jersey are valuable weapons against wrongful arrests and convictions. In New Jersey, which is considered a national model, landmark court decisions, most recently in 2011, moved the state to require changes such as showing photos of suspects one image at a time and recording a witness' exact words at the moment of identification. In instructing juries, judges address the potential pitfalls of memory. The Philadelphia District Attorney's Office also has adopted those changes, and they are contained in a policy directive.
"What New Jersey is doing right now is leading the country," said J.C. Lore, a law professor at Rutgers-Camden, even if the rest of the nation has been slow to follow.
The result trickles down, Lore and others said: Prosecutors know they will face stiff tests in the courtroom, so they press police for more rigor in the process.
"What we can hope is that we'll have an impact on the front end, because ultimately it's the front end that matters," said Cathy Rosen, chair of the criminal justice department at Temple University. For "the vast majority of the people caught in the criminal justice system, the front end is all that exists."
Consider Robert Williams, identified by a Mexican-born grandfather as the man who shot him and robbed him in the corner store where he worked. Police had no other evidence linking Williams to the crime, and video evidence placed another suspect at the store. The week before Williams went to trial — taking his chances with the jury that ultimately acquitted him — he was offered a plea deal, he said. It would have been for three to 10 years, but he rejected it because he believed he was innocent and didn't want that conviction on his record.
The store employee had identified Williams from a series of police photos of black men from the neighborhood. He also said the gunman sounded like a regular customer whose name he did not know.
"You have to understand that if you are [Hispanic] in a predominately black neighborhood, everybody may look alike to you, everybody definitely sounds alike to you," said Williams, 27, who is married with two children.
"I don't know if he truly believed it was me," Williams said. "I just hope he understands that he was wrong, and that he comes to the realization that he had the wrong person."
Cross-racial identifications have become increasingly recognized as problematic in courtrooms nationwide.
In some cases, police and prosecutors don't have much more to go on than those IDs. When Oh, the Philadelphia city councilman, picked out Yarbray from a stack of more than 300 police photos, investigators had no additional witnesses. Nor did they have any DNA evidence, the knife used in the attack, bloody clothing, video footage, or a known connection between Yarbray, who is black, and Oh, who is Asian.
Still, Yarbray was charged and brought to trial, then acquitted.
Shephard, who is black, was convicted in 1984 of the rape and robbery of a white woman. It took more than a decade for him to be freed when multiple rounds of DNA testing exonerated him.
When John J. Farmer began revising identification procedures as New Jersey's attorney general from 1999 to 2002, cases like Shephard's were on his mind.
At the time, Farmer said, DNA-based exonerations were calling the criminal justice system into question while scientific research was becoming increasingly conclusive about the fallibility of identifications.
"It became clear that the reliability of our criminal justice system was at stake, and steps needed to be taken to improve the reliability of the convictions," Farmer said.
In 1999, the New Jersey Supreme Court ruled that juries had to be informed of the science behind potential errors in cross-racial eyewitness identifications. It was an opportunity, Farmer said, to take a panoramic look at the issue.
He contacted Wells at Iowa State and requested a review of scientific research. Because the New Jersey attorney general wields more power than counterparts in other states, Farmer was able to mandate changes in every police department and prosecutor's office.
The best practices that Farmer ultimately embraced were aimed at ensuring that no suspect unduly stands out, minimizing the risk of misidentification. Changes included having lineups conducted by an officer who does not know the suspect's identity, lest the witness receive unconscious signals.
Police and prosecutors at first were wary of the new procedures, some saying the changes would make it impossible for them to do their jobs. No witness was perfect.
Ultimately, Wells and Farmer recalled, those concerns were put to rest after detectives understood how the rules help them, too, by making sure identifications are as strong as possible.
"It was not an assault on the way things had been done — it was a way of doing things better," Farmer said. "It isn't that police and prosecutors have any interest at all in seeing innocent people convicted. They don't."
Farmer and other experts said they believed the front-end changes have helped ensure that better identifications are made and that fewer innocent people are arrested and charged.