In 1974, a young woman in Alabama was shown a row of photographs and pointed at one of them. The man in the picture was Ervin Harris, 19 years old, from the Birmingham area. According to the court record, she explained her choice by saying that the eyes in the other photographs were not big enough. Police told her they were going to go and get him. They did, that same day.
That identification was the case. Not part of it. Essentially all of it. There was no fingerprint, no DNA testing anywhere in American courtrooms yet, and nothing physical that pointed at Harris. A jury in Bessemer convicted him in late 1975 and the sentence was 99 years. He served 42 of them.
Then comes the part of this story that almost nobody writes about, and it is the part worth sitting with. Harris was released on parole in 2017. He did not walk out exonerated. He walked out a convicted man, with the conviction fully intact and every consequence of it still attached, and he lived that way for another nine years. He was 71 when a judge finally set the conviction aside, and 71 when the case against him was formally dropped a few weeks after that.
Most wrongful conviction stories are shaped like a door opening. Someone is in a cell, the evidence collapses, and they come out into a car park with their family and a crowd of cameras. Harris did not get that shape. His prison sentence and his wrongful conviction ended nine years apart, and in the gap between them he was a free man in the narrow sense and a convicted one in every sense that touches a life.
Start with what the trial actually was. The offence was in November 1974. The woman who was attacked was 21. Nothing about what happened to her belongs in this piece, and the one thing that should be said plainly is that nothing in the eventual ruling suggests she was anything other than a genuine victim telling the truth as she remembered it. She has since died. She is not the villain of this story and never was.
What the record shows is that her memory and the man she picked did not match. The person she described to police differed from Harris in age, in height and in facial hair. She told police the man who attacked her spoke very clearly. Harris had, and still has, a severe stutter. His sister testified at the original trial that he could not talk for long without stuttering, and another witness said the same. The attacker drove her around in her own car before the assault. Harris could not drive, and a witness said she had never once seen him manage it. He also had a severe and contagious sexually transmitted infection at the time, which she did not contract.
He gave police an alibi immediately: he was with family and friends, playing cards and shooting pool. A string of relatives and friends testified to it. Prosecutors were never able to shake any of them. The jury convicted him anyway on 6 November 1975, and that is not as strange as it sounds, because juries then and now tend to find a confident witness pointing across a courtroom more persuasive than a row of people saying somebody was at home.
Semen and blood were found on the victim’s clothing. Nobody tested the blood type, which was possible in 1975 and would have narrowed the field considerably. Tests on Harris’s own clothing came back negative. The Alabama Court of Criminal Appeals upheld the conviction in 1976. He filed a petition in 1978 arguing his lawyer should have made more of the medical evidence, and the courts turned him down. After that, the legal system had said its piece, and he had 39 more years to serve.
Judge David Carpenter, reviewing the case decades later, noted something about the circumstances that is in the record and should be stated once, without ornament: Harris was a Black man with two prior burglary convictions, accused of raping a young white woman in Alabama in 1974. The judge raised it as context for how a case this thin reached a conviction. It is not the engine of this story and it does not need to be dressed up. It is simply part of why nobody looked harder at the time.
Now the nine years. In 2017, after more than 41 years inside, Harris was released on parole and required to register as a sex offender.
That sentence passes quickly in every account of this case, and it should not. Alabama has one of the most demanding registration regimes in the country, and for an adult convicted of a sex offence it runs for life. The statute requires a registrant to appear in person to verify his information during his birth month and then every three months after that, in every county where he lives, for the rest of his life. Over nine years, that is somewhere around 36 separate appearances in front of law enforcement, each one a re-registration of a thing he said he had not done.
The restrictions are not only administrative. Alabama law forbids a registrant from establishing or keeping a residence within 2,000 feet of certain properties, including schools and childcare facilities. In a built-up city, that rules out a great deal of the available housing, and it does not care whether the person has anywhere else to go. There are employment bars on top of that, including work at schools and childcare facilities and on mobile vending businesses. If a registrant wants to leave his county for three or more days, he has to give notice first. His name and address sit on a public database that anybody can search, including employers, landlords and neighbours.
Put those pieces together and you get the real texture of those nine years. A man in his sixties, out of prison for the first time since the Ford administration, trying to find somewhere to live inside a map full of exclusion zones, trying to find work with a record that answers every question before he opens his mouth, and reporting to a counter four times a year to confirm that he is still who the state says he is. Parole is not an apology. It is a licence with conditions, and the conviction underneath it was still standing.
There is a further wrinkle that is easy to miss. Registration in Alabama is not a sentence a judge hands down and a person completes. It attaches to the conviction itself and runs for life, which means that for nine years Harris was being supervised, restricted and published on the basis of a verdict that the same court system would eventually decide could not be trusted. Every quarterly appearance was an act of compliance with a finding that was wrong. And because the obligations are lifelong, nothing in the arrangement contained any prospect of improvement. There was no date in the future on which it was going to get easier. The only thing that could ever end it was the one thing that had already failed twice in the 1970s, which was a court agreeing to look again.
It is also worth being clear about what the restrictions cost in ordinary terms rather than legal ones. Housing, work, where you can sleep, how far you can travel and who gets to know your business before they meet you are not peripheral things. They are most of a life.
It needs saying honestly that Harris himself has not publicly described those years in any detail. The court filings, the registry record and the news coverage all move straight from his release to his hearing, and none of them stop to ask what the decade in between was like. That gap is itself worth noticing. The years a person spends in prison get counted carefully, because they are easy to count. The years afterwards, carrying a conviction that has not yet been undone, do not get counted at all, and there is no line in any statute that compensates for them.
He did not stop trying. The Innocence Project took up his case in 2018, the year after he came out. There was not much to work with. He wanted DNA testing on clothing collected in 1974 by the Fairfield police, which would have settled the thing outright. The court ordered an exhaustive search for it. The evidence could not be found. Half a century of storage had swallowed the one item that could have given him a clean answer.
So the case was rebuilt on the only thing left, which was the identification itself, and on what the last fifty years of research has established about how identifications go wrong.
The structural problem with a single eyewitness is not that witnesses are careless. It is that there is nothing to check them against. In most criminal cases, a piece of evidence is only as good as the other evidence that agrees with it, and a jury weighing a dozen strands can notice when one of them does not fit. When the entire case is one person’s recollection, there is no second strand, and the only test available to the jury is whether the witness seems sure. That test does not work, which is the part the science has settled.
This is where the story stops being about one man’s bad luck. In 2014 the National Research Council, part of the US National Academies, published a report called Identifying the Culprit, which remains the standard scientific assessment of eyewitness evidence. Its conclusions are not comforting for anyone who assumes memory works like a recording. Memories are forgotten, reconstructed, updated and distorted. A witness’s memory can be contaminated by all sorts of influences, including ordinary interaction with police, and the contamination does not feel like contamination from the inside. It feels like remembering.
The report’s most uncomfortable finding concerns confidence. A witness’s confidence tends to grow between the identification and the trial, as she learns more about the case, goes through trial preparation and sits in the pressure of the witness box. By the time she points at the defendant in front of a jury, she may be completely certain, and her certainty tells the jury nothing reliable about whether she is right. That is why the report recommends recording a witness’s confidence verbatim at the very first identification, before anything can inflate it.
The report also recommends that photo arrays and lineups be administered blind, meaning the officer running the procedure does not know which photograph is the suspect, so that nothing in his face or tone can steer the witness. It recommends that witnesses be told clearly that the perpetrator may or may not be among the pictures. It recommends that the whole procedure be video recorded as standard. None of that existed in 1974. What existed instead was a procedure in which a witness looked at photographs and the officers already knew which one they wanted, and then told her they were going to go and get him.
At Harris’s hearing, an eyewitness identification expert named Nancy Franklin testified that the identifications in his case were highly unreliable, and that they were very likely the product of suggestive influences after the event rather than of memory. She made the point that the person the witness described and the person she identified were two different people. An investigator with the district attorney’s Conviction Integrity Unit, which was only set up in 2021, said Harris’s account of that night was the most consistent statement he had ever taken. Nothing in it had moved in fifty years.
It matters to be exact about what happened next, because the distinctions are real and they are routinely blurred. On 19 June 2026, Judge Carpenter vacated the conviction and the 99-year sentence. The legal basis was that modern eyewitness research amounted to newly discovered evidence, and that if Harris had been able to put that science in front of his jury, the verdict would likely have been different. The judge also ordered him removed from the state sex offender registry. The district attorney’s office did not object to any of it. That is unusual and it is worth stating, flatly, without applause.
What the judge did not do was declare Ervin Harris innocent. No court has. A vacated conviction means the verdict cannot stand, that the trial was not a reliable test of guilt, that the state’s case does not hold. It is not the same as a finding that the man did not do it, and the order does not contain one. Then, a few weeks later, on 12 July 2026, the prosecution asked for the case to be dismissed and the court granted it. That is as close to over as the system gets. Harris is no longer a convicted man and will not be tried again. The National Registry of Exonerations lists him among its exonerees, with mistaken witness identification as the cause.
Which brings the story to arithmetic, because arithmetic is the only part of this that resolves cleanly. Alabama does have a compensation scheme for the wrongly imprisoned, set up in 2001 and run by a legislative committee with the state’s risk management division checking applications. The base figure is a minimum of 50,000 dollars for each year of incarceration. Harris served 42 years. At the statutory base, that is about 2.1 million dollars.
He has not been paid any of it, and there is no public record of a claim. There are two obstacles in the way. The scheme is built around people who were imprisoned for felonies of which they were innocent, and the order in his favour does not make a finding of innocence, which is a problem an applicant has to solve rather than a box he can tick. And any amount the committee recommends beyond the base still has to be appropriated by the state legislature, which means a payment is a political act as well as a legal one.
So the honest close is not a cheque and not a scene of vindication. The National Registry of Exonerations has recorded more than 3,800 exonerations in the United States since 1989, adding up to more than 36,000 years of lost life, and Harris’s 42 are now part of that total. But the number that stays with you is the smaller one. He finished the punishment before anyone undid the verdict. The state had already taken everything a 99-year sentence takes, and then nine more years on top of it from a man who was out of prison and still not free, and the correction arrived when he was 71, with nothing left for it to give back.