Kenneth Alarcon had a bad feeling as he and a friend left the restaurant. He noticed two figures following closely behind, and suddenly there was a gun in his face.

“Give me all your stuff or I’ll shoot you,” a voice commanded.

Alarcon emptied his pockets and turned over a wedding band that belonged to his grandmother.

Don’t look at me, one of the two assailants warned as he snatched the ring. “I will kill you!”

And then the gunmen were gone, leaving Alarcon and his companion face down on the dirty sidewalk just south of downtown Minneapolis. Police arrived minutes later and a description of the suspects given by the shaking Alarcon led to arrests and robbery charges.

But that was not the end of this case. When it went to trial last summer, the defense argued that police ignored basic photo lineup science, broke department policies and coached Alarcon into picking the suspect they wanted — claims police denied in court and in recent interviews.

Debates over the reliability of eyewitness memory rarely make it to trial in Minnesota, but a growing body of scientific research is driving law enforcement agencies across the country to adopt new policies designed to reduce the chances of wrongfully identifying an innocent person. The Department of Justice requires all federal agents to follow procedures based on this science, such as recording the confidence of a witness directly after the lineup, and not months later in court, when the memory is more likely to fail. Many states, from Wisconsin to Texas, have passed laws mandating scientific procedures.

Minnesota has no statewide policy, and most police agencies have no written policy at all, according to a 2016 survey conducted by the Minnesota Chiefs of Police Association.

“It’s all over the board,” said William Ward, Minnesota’s chief statewide public defender.

Minneapolis’ Deputy Chief Bruce Folkens said officers embrace all tools that can help point investigative resources toward criminals. “We want to minimize any potential for false identification,” he said. “If there’s a better way to do things, we try to implement that.”

Yet the Alarcon case isn’t the first to raise scrutiny over how the department conducts lineups.

In one case, a questionable ID led to a wrongful arrest in a violent rape, and the wrong guy spent months in jail. In another, an officer e-mailed a lineup to a witness.

“Dear God. That to me is a problem,” said Julie Jonas, legal director for the Innocence Project of Minnesota, the national organization that works to overturn wrongful convictions. “You have no idea — the guy could be drunk when he’s looking at it.”

Choosing from six

Three months after the robbery, in mid-October 2014, Alarcon sat in an interview with Minneapolis police, studying the faces of six men as Sgt. David Wilson presented them one by one.

At the time of the robbery, Alarcon thought he’d be able to recognize the guys if he saw them again. But now he wasn’t so sure.

“Can I see them all together?” Alarcon asked, according to an audio recording of the interview.

“No, it has to be one at a time,” Wilson replied.

The lead investigator in the case, Sgt. Mark Suchta, had explained the department policy before the lineup began. Officers must show the photos one at a time — called a “sequential lineup” — to avoid a phenomenon in which the brain finds the face closest to the suspect’s and thinks it’s a match.

The policy also instructs officers to use what’s called a “double-blind”: the officer showing the photos doesn’t know the suspect, and couldn’t, even subconsciously, give cues to the witness. Because Suchta put the lineup together, Wilson conducted the test.

But when Alarcon couldn’t pick a suspect for the second time, Wilson left and Suchta entered the room. He laid out all six photos at once.

“It’s hard to know with 100 percent certainty,” Alarcon repeated.

“You just pointed at one,” said Suchta. “You just have to say for the record which one you pointed at. You pointed at number six.”

Alarcon acknowledged pointing to a photo, but said he wasn’t sure — that it was only “possible.”

Suchta ended the interview. He filed a report saying he’d completed a sequential lineup, and that Alarcon identified number six as the person “who could have possibly robbed [him].” The document says Wilson conducted the lineup — even though Alarcon made the statement quoted in the report to Suchta.

The Hennepin County Attorney’s Office later filed felony robbery charges against number six: A 28-year-old Brooklyn Center man named James Moore.

Falling behind

Not long ago, Minnesota was considered a leader in the science of photo lineups.

In 2004, then-Hennepin County Attorney Amy Klobuchar helped implement new countywide policies for law enforcement, including using the double-blind method and showing photos sequentially. Officers are also trained to construct lineups based on similar characteristics, like height, weight and skin color, using at least six photos made up of one suspect and the rest fillers.

The body of research behind identifications has grown since then, leading to more reforms, and the Innocence Project counts bad identifications as a factor in 70 percent of wrongful convictions the group has overturned nationally. In January this year, the Department of Justice released a revised rulebook on how to conduct lineups. Nineteen states have adopted statewide policies, according to the Innocence Project.

Minnesota is not among them. Of the 156 law enforcement agencies that responded to a 2016 survey by the police chiefs association, about 60 percent had no written policies.

The Innocence Project has provided training for officers, but prosecutors and police have rejected taking it any further. “We’ve gotten a lot of pushback,” said Michelle Feldman, counsel for the Innocence Project’s national office. “The problem is, if you don’t have a uniform practice throughout the state, justice depends on where you’re arrested.”

Andy Skoogman, president of the chiefs association, said his organization determined that a statewide policy isn’t necessary after conducting the survey. Even though many agencies in the state don’t have official policies, Skoogman said most believed they were up to snuff in practice.

Prosecutors want to see investigators using best methods, but they shouldn’t be dictating investigative processes to police, said Robert Small, head of the Minnesota County Attorneys Association.

“Nobody wants to see wrongful convictions,” Small said. “Nobody wants to see questionable identifications impacting a conviction.”

Wrong man jailed

Yet sometimes lineups do just that.

In November 2015, a 15-year-old girl in north Minneapolis said a man raped her at gunpoint in a nearby alley. The girl offered a description of her attacker: a black man, 30-40 years old, heavyset, with a goatee.

The investigators put together 12 mugshots of possible suspects, all with histories of sex offenses, and visited her at school for a follow-up interview. Upon seeing one, the girl began sobbing and shaking. She told police she was “100 percent” certain it was him.

The man in the photo, 34-year-old Charles Stevenson, had sexually assaulted a 15-year-old in his early 20s. Prosecutors charged him with the rape.

But it was the wrong man.

DNA exonerated him two months later. Police eventually arrested Harold Davis, who lived a couple miles from the victim, and just down the street from where the assault took place. In 2007, Davis had kidnapped and raped a 12-year-old.

Deputy Chief Folkens said he saw nothing wrong with the lineup. But Nancy Steblay, an Augsburg University psychology professor who worked with Klobuchar to collect data and introduce the county’s lineup rules, said the case illustrates how a poorly constructed lineup can taint an investigation. If the victim had had the same reaction to a filler photo, the officers would know her memory of the incident was hazy. But by using only suspects, they “violated the first rule of lineups,” said Steblay.

“Consider, in the absence of DNA evidence, how this case may have ended up very differently,” she said.

In a separate 2015 case, a deaf man reported a stolen iPad. The victim described the culprit as a Hispanic man in his late 20s. But in the lineup, the officer used widely varying race and age, including two Caucasian men, and two much older men, one bald with a white beard and tattoo on his neck.

“That is one of the worst lineups I have ever seen,” said Jonas.

The officer did not show the photos one at a time per policy. Instead, Minneapolis police officer Brendon Schram e-mailed the lineup to the suspect. Schram included only one instruction, in the subject line of the e-mail: “Please let me know if you recognize the person that took your IPAD by the numbered photo.”

The victim replied, “Number is 3 picture. That guy’s little beards.”

To which Schram replied, “Thank you. I will be issuing an arrest warrant for him.”

The man was charged with the robbery, but charges were later dropped because the witness moved out of state.

In this case, Folkens said the lineup didn’t meet the department’s standards. “I acknowledge, that’s not a good lineup,” he said. “Procedure was not good. I would expect more from our investigators in the police department, and certainly in our violent crimes and special crimes investigations units.”

But Folkens said police conduct thousands of lineups per year, and this case is an aberration. “We do have a lot of great investigators doing a lot of great casework that is being upheld in court,” he said.

Jury rules

The trial for James Moore began in summer 2016 — almost two years after Alarcon picked his photo in the lineup with Suchta.

Other evidence also linked Moore to the crime. Someone had tried opening credit cards in Alarcon’s name, listing an address where Moore got mail.

In interviews with police, Moore said he’d been set up, and that he’d even heard an acquaintance talking about the robbery. His attorney later said several people got mail at the address.

At trial, Moore maintained his innocence, and his public defender argued that Sgt. Suchta’s lineup wasn’t reliable.

On the stand, Suchta called the department’s lineup rules only the “preferred method.” He denied steering Alarcon toward Moore in the interview.

“I’m just a fact gatherer,” Suchta said. “So when he says it was a possible suspect, I’m obligated to accurately document that.”

In an interview, Cmdr. Erick Fors agreed it’s acceptable that Suchta was present and showed the photos simultaneously, noting a lineup is just one piece of an investigation. “It was done at the request of the victim,” said Fors. “There’s not a legal precedent that says that an investigator can’t respond to a victim request to look at something.”

For a defense, Moore’s attorney turned to Steblay. She said the lineup was flawed from the beginning, when Suchta waited three months to show photos to Alarcon, a long time for a witness to retain clear memories of a stranger’s face. And the officer should have stopped the lineup after Alarcon failed to identify anyone after two attempts.

She also scrutinized Suchta’s behavior when he came back into the room. He ignored the double-blind. By showing the photos all at once, and asking a witness to pick one “possible,” an officer is putting pressure on the witness to make an identification, or worse — “taking an educated guess.”

The jury found Moore not guilty.