By Ryan Gabrielson and Topher Sanders, ProPublica

Amy Albritton can’t remember if her boyfriend signaled when he changed lanes late that August afternoon in 2010. But suddenly the lights on the Houston Police patrol car were flashing behind them, and Anthony Wilson was navigating Albritton’s white Chrysler Concorde to a stop in a strip-mall parking lot. It was an especially unwelcome hassle. Wilson was in Houston to see about an oil-rig job; Albritton, volunteering her car, had come along for what she imagined would be a vacation of sorts. She managed an apartment complex back in Monroe, La., and the younger of her two sons — Landon, 16, who had been disabled from birth by cerebral palsy — was with his father for the week. After five hours of driving through the monotony of flat woodland, the couple had checked into a motel, carted their luggage to the room and returned to the car, too hungry to rest but too drained to seek out anything more than fast food. Now two officers stepped out of their patrol car and approached.

Albritton, 43, had dressed up for the trip — black blouse, turquoise necklace, small silver hoop earrings glinting through her shoulder-length blond hair. Wilson, 28, was more casually dressed, in a white T-shirt and jeans, and wore a strained expression that worried Albritton. One officer asked him for his license and registration. Wilson said he didn’t have a license. The car’s registration showed that it belonged to Albritton.

The officer asked Wilson to step out of the car. Wilson complied. The officer leaned in over the driver’s seat, looked around, then called to his partner; in the report Officer Duc Nguyen later filed, he wrote that he saw a needle in the car’s ceiling lining. Albritton didn’t know what he was talking about. Before she could protest, Officer David Helms had come around to her window and was asking for consent to search the car. If Albritton refused, Helms said, he would call for a drug-sniffing dog. Albritton agreed to the full search and waited nervously outside the car.

Helms spotted a white crumb on the floor. In the report, Nguyen wrote that the officers believed the crumb was crack cocaine. They handcuffed Wilson and Albritton and stood them in front of the patrol car, its lights still flashing. They were on display for rush-hour traffic, criminal suspects sweating through their clothes in the 93-degree heat.

As Nguyen and Helms continued the search, tensions grew. Albritton, shouting over the sound of traffic, tried to explain that they had the wrong idea — at least about her. She had been dating Wilson for only a month; she implored him to admit that if there were drugs, they were his alone. Wilson just shook his head, Albritton now recalls. Fear surging, she shouted that there weren’t any drugs in her car even as she insisted that she didn’t know that Wilson had brought drugs. The search turned up only one other item of interest — a box of BC Powder, an over-the-counter pain reliever. Albritton never saw the needle. The crumb from the floor was all that mattered now.

At the police academy four years earlier, Helms was taught that to make a drug arrest on the street, an officer needed to conduct an elementary chemical test, right then and there. It’s what cops routinely do across the country every day while making thousands upon thousands of drug arrests. Helms popped the trunk of his patrol car, pulled out a small plastic pouch that contained a vial of pink liquid and returned to Albritton. He opened the lid on the vial and dropped a tiny piece of the crumb into the liquid. If the liquid remained pink, that would rule out the presence of cocaine. If it turned blue, then Albritton, as the owner of the car, could become a felony defendant.

Helms waved the vial in front of her face and said, “You’re busted.”

Albritton was booked into the Harris County jail at 3:37 a.m., nine hours after she was arrested. Wilson had been detained for driving without a license but would soon be released. Albritton was charged with felony drug possession and faced a much longer ordeal. Already, she was terrified as she thought about her family. Albritton was raised in a speck of a town called Marion at the northern edge of Louisiana. Her father still drove lumber trucks there; her mother had worked as a pharmacy technician until she died of colon cancer. Albritton was 15 then. She went through two unexpected pregnancies, the first at age 16, and two ill-fated marriages. But she had also pieced together a steady livelihood managing apartment complexes, and when her younger son was born disabled, she worked relentlessly to care for him. Now their future was almost certainly shattered.

The officers allowed her to make a collect call on the coinless cellblock pay phone. She had a strained relationship with her father and with her son’s father as well; instead she dialed Doug Franklin, an old friend who once dated her sister. No one answered. Near dawn the next morning, guards walked Albritton through a tunnel to the Harris County criminal-justice tower’s basement, where they deposited her in a closet-size holding room with another woman, who told Albritton that she had murdered someone. Albritton prayed someone would explain what would happen next, tell her son she was alive and help her sort out the mess. She had barely slept and still hadn’t eaten anything. She heard her name called and stepped forward to the reinforced window. A tall man with thinning hair and wire-rim glasses approached and introduced himself as Dan Richardson, her court-appointed defense attorney.

Richardson told Albritton that she was going to be charged with possession of a controlled substance, crack cocaine, at an arraignment that morning. Albritton recalls him explaining that this was a felony, and the maximum penalty was two years in state prison. She doesn’t remember him asking her what actually happened, or if she believed she was innocent. Instead, she recalls, he said that the prosecutor had already offered a deal for much less than two years. If she pleaded guilty, she would receive a 45-day sentence in the county jail, and most likely serve only half that.

Albritton told Richardson that the police were mistaken; she was innocent. But Richardson, she says, was unswayed. The police had found crack in her car. The test proved it. She could spend a few weeks in jail or two years in prison. In despair, Albritton agreed to the deal.

Albritton was escorted to a dark wood-paneled courtroom. A guilty plea requires the defendant to make a series of statements that serve as a confession and to waive multiple constitutional rights. The judge, Vanessa Velasquez, walked her through the recitation, Albritton recalls, but never asked why she couldn’t stop crying long enough to speak in sentences. She had managed to say the one word that mattered: “guilty.”

Police officers arrest more than 1.2 million people a year in the United States on charges of illegal drug possession. Field tests like the one Officer Helms used in front of Amy Albritton help them move quickly from suspicion to conviction. But the kits — which cost about $2 each and have changed little since 1973 — are far from reliable.

The field tests seem simple, but a lot can go wrong. Some tests, including the one the Houston police officers used to analyze the crumb on the floor of Albritton’s car, use a single tube of a chemical called cobalt thiocyanate, which turns blue when it is exposed to cocaine. But cobalt thiocyanate also turns blue when it is exposed to more than 80 other compounds, including methadone, certain acne medications and several common household cleaners. Other tests use three tubes, which the officer can break in a specific order to rule out everything but the drug in question — but if the officer breaks the tubes in the wrong order, that, too, can invalidate the results. The environment can also present problems. Cold weather slows the color development; heat speeds it up, or sometimes prevents a color reaction from taking place at all. Poor lighting on the street — flashing police lights, sun glare, street lamps — often prevents officers from making the fine distinctions that could make the difference between an arrest and a release.

There are no established error rates for the field tests, in part because their accuracy varies so widely depending on who is using them and how. In Las Vegas, authorities re-examined a sampling of cocaine field tests conducted between 2010 and 2013 and found that 33 percent of them were false positives. Data from the Florida Department of Law Enforcement lab system show that 21 percent of evidence that the police listed as methamphetamine after identifying it was not methamphetamine, and half of those false positives were not any kind of illegal drug at all. In one notable Florida episode, Hillsborough County sheriff’s deputies produced 15 false positives for methamphetamine in the first seven months of 2014. When we examined the department’s records, they showed that officers, faced with somewhat ambiguous directions on the pouches, had simply misunderstood which colors indicated a positive result.

No central agency regulates the manufacture or sale of the tests, and no comprehensive records are kept about their use. In the late 1960s, crime labs outfitted investigators with mobile chemistry sets, including small plastic test tubes and bottles of chemical reagents that reacted with certain drugs by changing colors, more or less on the same principle as a home pregnancy test. But the reagents contained strong acids that leaked and burned the investigators. In 1973, the same year that Richard Nixon formally established the Drug Enforcement Administration, declaring “an all-out global war on the drug menace,” a pair of California inventors patented a “disposable comparison detector kit.” It was far simpler, just a glass vial or vials inside a plastic pouch. Open the pouch, add the compound to be tested, seal the pouch, break open the vials and watch the colors change. The field tests, convenient and imbued with an aura of scientific infallibility, were ordered by police departments across the country. In a 1974 study, however, the National Bureau of Standards warned that the kits “should not be used as sole evidence for the identification of a narcotic or drug of abuse.” Police officers were not chemists, and chemists themselves had long ago stopped relying on color tests, preferring more reliable mass spectrographs. By 1978, the Department of Justice had determined that field tests “should not be used for evidential purposes,” and the field tests in use today remain inadmissible at trial in nearly every jurisdiction; instead, prosecutors must present a secondary lab test using more reliable methods.

But this has proved to be a meaningless prohibition. Most drug cases in the United States are decided well before they reach trial, by the far more informal process of plea bargaining. In 2011, RTI International, a nonprofit research group based in North Carolina, found that prosecutors in nine of 10 jurisdictions it surveyed nationwide accepted guilty pleas based solely on the results of field tests, and in our own reporting, we confirmed that prosecutors or judges accept plea deals on that same basis in Atlanta, Boston, Dallas, Jacksonville, Las Vegas, Los Angeles, Newark, Philadelphia, Phoenix, Salt Lake City, San Diego, Seattle and Tampa.

This puts field tests at the center of any discussion about the justice of plea bargains in general. The federal government does not keep a comprehensive database of prosecutions in county and state criminal courts, but the National Archive of Criminal Justice Data at the University of Michigan maintains an extensive sampling of court records from the 40 largest jurisdictions. Based on this data, we found that more than 10 percent of all county and state felony convictions are for drug charges, and at least 90 percent of those convictions come by way of plea deals. In Tennessee, guilty pleas produce 94 percent of all convictions. In Kansas, they make up more than 97 percent. In Harris County, Tex., where the judiciary makes detailed criminal caseload information public, 99.5 percent of drug-possession convictions are the result of a guilty plea. A majority of those are felony convictions, which restrict employment, housing and — in many states — the right to vote.

Demand for the field tests is strong enough to sustain the business of at least nine different companies that sell tests to identify cocaine, heroin, marijuana, methamphetamine, LSD, MDMA and more than two dozen other drugs. The Justice Department issued guidelines in 2000 calling for test-kit packaging to carry warning labels, including “a statement that users of the kit should receive appropriate training in its use and should be taught that the reagents can give false-positive as well as false-negative results,” but when we checked, three of the largest manufacturers — Lynn Peavey Company, the Safariland Group and Sirchie — had not printed such a warning on their tests. (Lynn Peavey Company did not respond to our request for comment. A spokesman for the Safariland Group said the company provides law-enforcement agencies with extensive training materials that are separate from the tests and their packaging.

We asked John Roby, Sirchie’s chief executive, about the missing warnings and requested an interview in May. He responded in writing a month later saying that the boxes carrying Sirchie’s cocaine tests had been updated and now display a warning that reactions may occur with both “legal and illegal substances.” After our inquiry, Sirchie added another warning to its packaging, listing at the bottom of its printed instructions: “ALL TEST RESULTS MUST BE CONFIRMED BY AN APPROVED ANALYTICAL LABORATORY!”)

Even trained lab scientists struggle with confirmation bias — the tendency to take any new evidence as confirmation of expectations — and police officers can see the tests as affirming their decisions to stop and search a person. Labs rarely notify officers when a false positive is found, so they have little experience to prompt skepticism. As far as they know, the system works. By our estimate, though, every year at least 100,000 people nationwide plead guilty to drug-possession charges that rely on field-test results as evidence. At that volume, even the most modest of error rates could produce thousands of wrongful convictions.

After he arrested Amy Albritton, Officer Helms sent what remained of the crumb he found on the floor of her car to the Houston Police Department crime laboratory. He listed it as “.02 grms crack cocaine” and noted on the submission form that he was also sending a “syringe w/ unknown substance .01 gr” — presumably the needle Officer Nguyen reported pulling from the ceiling lining and that Albritton had not seen and still could not explain. (Helms’s submission form, which was separate from the arrest report, said it came not from the ceiling but from the “suspect visor.”) The last item Helms turned in was a ziploc bag of the “unknown wht powder” that had been removed from the BC Powder package.

“HOLD + ANALYZE FOR COURT,” Helms wrote by hand. And then, with no court case pending, the evidence sat on hold, one of several thousand samples in the laboratory’s backlog of untested pills, plants, powders and assorted crumbs and pebbles.

Albritton served 21 days of her 45-day sentence. When she was released, she took a taxi to the motel where she had planned to stay with Wilson, whom she never saw again after the arrest. (Helms and Nguyen would not comment for this article; Wilson did not respond to requests.) The manager had kept her clothes, so she took a room again and waited for her friend Doug Franklin to fly in from Louisiana. The plan was that he would lend her the money to get her impounded car and keep her company on the drive home. When they retrieved the car, it had been sitting in the summer heat for more than three weeks. Albritton was overwhelmed by the smell of rotting hamburgers.

When Albritton pleaded guilty, she asked Franklin to explain the situation to her bosses at the rental-property firm, but Franklin decided it was safer to say nothing. She was going to be fired in any case, he reasoned, and alerting an employer about the drug felony would only hurt her future prospects. Albritton had managed the Frances Place Apartments, a well-maintained brick complex, for two years, and a free apartment was part of her compensation. But as far as the company knew, Albritton had abandoned her job and her home. She was fired, and her furniture and other belongings were put out on the side of the road. “So I lost all that,” she says.

Albritton’s older son, Adam, then 24, had been living on his own for years and learned of his mother’s arrest only after she had begun her sentence. While Albritton was incarcerated, her younger son, Landon, remained with his father, who had threatened in the past to seek custody but never followed through. Albritton’s father, Tommy Franklin (no relation to Doug), was openly skeptical about her claim of innocence. “If the law said you had crack, you had crack,” she recalls him telling her.

Albritton gave up trying to convince people otherwise. She focused instead on Landon. Using a wheelchair, he needed regular sessions of physical and occupational therapy, and Albritton’s career managing the rental complex had been an ideal fit, providing a free home that kept her close to her son while she was at work, and allowing her the flexibility to ferry him to his appointments. But now, because of her new felony criminal record, which showed up immediately in background checks, she couldn’t even land an interview at another apartment complex. With a felony conviction, she couldn’t be approved as a renter either. Doug Franklin allowed Albritton and Landon to move in with him temporarily, and Albritton took a minimum-wage job at a convenience store.

Through all of this, the crumb of evidence remained in storage in the Houston crime lab. It was a closed case, and the prosecutor, as was standard practice, had filed a motion to destroy the evidence. Only some final paperwork — a request from the lab and a judge’s signature — was needed. But this was an extremely low priority in a complex bureaucracy.

By 2010, the lab had been discredited by a decade of botched science and scandal. Thousands of untested rape kits were shelved from unsolved assaults. Errors in fingerprint matches were discovered in more than 200 cases. The lab had lost key blood samples; employees had tampered with or falsified other evidence. And it was continuing to struggle with a significant backlog of drug-test evidence — one that stemmed from what amounted to an epic experiment in field testing.

When Hurricane Katrina struck the Gulf Coast in August 2005, more than 250,000 mostly black refugees streamed into Houston, and local authorities openly anticipated a crime surge in which the refugees were portrayed as would-be perpetrators. Charles McClelland, who retired in February as Houston’s police chief and was then an assistant chief, says the department decided that pursuing drug-possession charges would also help suppress the number of predicted robberies and burglaries. “Anecdotally, it makes sense: Where does a person who has a substance-abuse problem get the money to buy drugs?” McClelland argues. “One could easily make the connection that they’re committing crimes.” The city distributed thousands more of the color field tests than usual to patrol officers, and drug evidence swamped the controlled-substances section of the lab. Even as the Katrina refugees gradually left Houston, the emphasis on low-level drug enforcement remained. By 2007, annual submissions to the lab had climbed to 22,000, even as budget cuts had reduced the staff, leaving the scientists with far more samples than they could competently analyze.

In 1972, the Department of Justice published a training guide for forensic chemists in the nation’s crime labs, emphasizing that they were “the last line of defense against a false accusation,” but 40 years later, that line had largely vanished. A federal survey in 2013 found that about 62 percent of crime labs do not test drug evidence when the defendant pleads guilty. But the Houston crime lab, for all its problems, would not be among them.

James Miller, the lab’s controlled-substances manager, had long practiced a kind of evidentiary triage. Evidence tied to pending drug manufacturing, sale or possession cases — 50 a year on average — would receive immediate attention, because only laboratory analysis would be admissible in court. But evidence from cases in which the defendants pleaded guilty before going to trial — the overwhelming majority of the remaining thousands of submitted drugs samples — would also be tested. The city had no legal requirement to confirm that the substances were the illegal drugs the police claimed they were. But in Miller’s lab, everything would be checked, even if it took years. “All along, we’ve said we’re about the science,” he says — not securing convictions. So the evidence sat, waiting.

The forensic scientists in Miller’s lab keep untested samples in Manila envelopes locked in cabinets below their work benches. Some sat there for as long as four years, lab records show. Albritton’s evidence stayed locked up for six months. On Feb. 23, 2011 — five months after Albritton completed her sentence and returned home as a felon — one of Houston’s forensic scientists, Ahtavea Barker, pulled the envelope up to her bench. It contained the crumb, the powder and the still-unexplained syringe. First she weighed everything. The syringe had too little residue on it even to test. It was just a syringe. The remainder of the “white chunk substance” that Officer Helms had tested positive with his field kit as crack cocaine totaled 0.0134 grams, Barker wrote on the examination sheet, about the same as a tiny pinch of salt. Barker turned to gas chromatography-mass spectrometry analysis, or GC-MS, the gold standard in chemical identification, to figure out what was in Albritton’s car that evening. She began with the powder. First the gas chromatograph vaporized a speck of the powder inside a tube. Then the gas was heated, causing its core chemical compounds to separate. When the individual compounds reached the end of the tube, the mass spectrometer blasted them with electrons, causing them to fragment. The resulting display, called a fragmentation pattern, is essentially a chemical fingerprint. The powder was a combination of aspirin and caffeine — the ingredients in BC Powder, the over-the-counter painkiller, as Albritton had insisted.

Then Barker ran the same tests on the supposed crack cocaine. The crumb’s fragmentation pattern did not match that of cocaine, or any other compound in the lab’s extensive database. It was not a drug. It did not contain anything mixed with drugs. It was a crumb — food debris, perhaps. Barker wrote “N.A.M.” on the spectrum printout, “no acceptable match,” and then added another set of letters: “N.C.S.” No controlled substance identified. Albritton was innocent.

Inger Chandler oversees the small conviction-integrity unit of the Harris County district attorney’s office, where she has been a prosecutor for 12 years. Conviction-integrity units are a fairly new concept in law enforcement: Prosecutors re-examine convictions in light of new evidence, often in the form of previously unavailable DNA tests. Conviction-integrity units originally focused on murder and rape cases, but they also increasingly investigate drug convictions.

In early 2014, Chandler took a call while sitting at her desk, encircled by stacks of case files and pictures of her toddler twins. Eric Dexheimer, a reporter at The Austin American-Statesman, told her he had noticed a series of unusual exonerations coming out of the Texas Court of Criminal Appeals. He’d tracked 21 drug convictions across Texas that had been reversed because labs had found that the drugs in question weren’t really drugs. The laboratory results came after defendants had already pleaded guilty. Did Harris County have any other bad drug convictions beyond what the courts had overturned? Chandler didn’t know, but she said she would try to find out.

Chandler called Miller, the controlled-substances manager at the lab, and asked him if there was something wrong with any of their drug convictions. Miller was not surprised to hear from Chandler. He explained that the lab had indeed found problems with their drug convictions; when his forensic scientists found discrepancies in the evidence — officially labeled “variants” — they sent the details by email to the district attorney’s office, and they had been doing so for years. Chandler hadn’t known any of this. She found the email inbox for lab notices, and it did indeed contain hundreds of messages that were sent from the lab. One after another, the lab notices said, “No Controlled Substance.” In cases involving drug possession, that meant the defendants were not guilty. (Drug manufacturing and selling charges can hold even if the underlying substance is not illegal.)

It was unclear if anyone had ever followed up on the notices. When Chandler entered several of the court case numbers into the district attorney’s records-search system, however, she found that a majority of the convictions remained in place. She started a list. Over the course of the following year, she found that the district attorney’s office had failed to correct 416 “variants” between January 2004 and June 2015, all of them in cases that ended in guilty pleas. Some variants were legally ambiguous — the field test was positive, but for the wrong drug; the drug weights were incorrect; or there was too little of the evidence to analyze — but in 251 cases, the results were simple: “No Controlled Substance.”

Under the 1963 Supreme Court opinion in Brady v. Maryland, prosecutors must provide defendants with exculpatory evidence, even after a conviction. Chandler could have met that mandate simply by alerting the convicting court and the defense attorneys to the lab reports — “Every other Brady situation, as long as I give notice, I’m done,” she says — but in these cases, Chandler says, she knew very few of the wrongful drug convictions would be reversed if she let the system handle each of them individually. The exoneration effort needed to be centralized, so that someone would become responsible for finding the defendants themselves. Chandler took the list to Devon Anderson, Harris County’s district attorney.

Anderson, a former district-court judge, had been the top prosecutor for only seven months. Her husband, Mike Anderson, who took office as district attorney in January 2013, died of cancer eight months into his term, and Gov. Rick Perry appointed her to replace him. Now, as Chandler described the problem, Anderson felt sickened. The litany of wrongful convictions was not just enormous — it was still growing. Her office, she says, was to blame for “a breakdown at every point in the system.” She hired a former prosecutor to research the cases and find the defendants. “It may sound corny, but it’s true: Our duty under Texas law is to seek justice,” she says. “A lot of people think it’s convictions, but it’s justice.”

In April 2014, The American-Statesman published Dexheimer’s story, which focused on 21 wrongful drug convictions across Texas caused by lab delays. But prosecutors in Harris County were still uncovering the scale of their own problem.

Based in part on the information gathered by Marie Munier, the former prosecutor Anderson hired to examine the drug convictions, we determined that 301 of the 416 variants began as arrests by the Houston Police Department, with the rest coming from surrounding municipalities, and that 212 of those 301 arrests were based on evidence that lab analysis determined was not a controlled substance, or N.C.S.

In our own examination of those 212 cases — thousands of pages of arrest reports, court filings and laboratory-testing records, along with interviews of prosecutors, police executives, officers, defense attorneys and innocent defendants who pleaded guilty — we saw a clear story about both who is being arrested and what is happening to them. The racial disparity is stark. Blacks made up 59 percent of those wrongfully convicted in a city where they are 24 percent of the population, reflecting a similar racial disparity in drug enforcement nationally. Patrol units, not trained narcotics detectives, appeared to be the most prolific field-test users.

The kits, or the officers interpreting them, got it wrong most often when dealing with small amounts of suspected drugs. Sixty-three percent of the N.C.S. cases involved less than a gram of evidence. The smallest possession cases are the ones in which a field test can be of greatest consequence; if officers find larger quantities of white powder in dozens of baggies or packaged in bricks, they have sufficient probable cause to make an arrest regardless of what a color test shows. (Though in those cases, too, they are generally required to test the drugs.) It’s widely assumed in legal circles that these wrongfully convicted people are in fact drug users who intended to possess drugs. Barry Scheck, a founder of the Innocence Project, a nonprofit group that seeks to overturn wrongful convictions, says some who work toward exoneration have complained to him that those exonerated of drug charges often are just accidentally not guilty, and shouldn’t be added to the National Registry of Exonerations. The assumption is not entirely without basis — 162 of the 212 N.C.S. defendants had criminal histories involving illegal drugs. However, 50 had no criminal history involving drugs at all.

All of the 212 N.C.S. defendants struck plea bargains, and nearly all of them, 93 percent, received a jail or prison sentence. Defendants with no previous convictions have a legal right in Texas to probation on drug-possession charges, even if they’re convicted at trial. But remarkably, 78 percent of defendants entitled to probation agreed to deals that included incarceration. Perhaps most striking: A majority of those defendants, 58 percent, pleaded guilty at the first opportunity, during their arraignment; the median time between arrest and plea was four days. In contrast, the median for defendants in which the field test indicated the wrong drug or that the weight was inaccurate — that is, the defendants who actually did possess drugs — was 22 days. Not only do the innocent tend to plead guilty in these cases, but they often do so more quickly.

On July 29, 2014, Munier sent a letter to Amy Albritton. It was a form letter, one of hundreds Munier was sending to exonerated defendants, opening with the salutation “Dear Sir or Madam,” but the contents were highly personal. It stated that the Harris County district attorney’s office had learned that the drug evidence in Albritton’s case was not a controlled substance: “Accordingly, you were prosecuted for a criminal drug offense and convicted in error.” Munier mailed the letter to the address on Albritton’s driver’s license, but Albritton did not receive it. She had long since moved on.

She had struggled to rebuild her life as a felon. The hours at the convenience store were erratic, so she started waiting tables and tending bar as she tried to find work in property management again. In 2013, she heard about a small set of rentals in Baton Rouge that needed someone to run them day to day. The pay was low compared with what she had made at Frances Place, and there was no free apartment. But the owner agreed to interview Albritton, even with her drug felony, and quickly hired her. She had almost nothing to pack besides her clothes and Landon’s before relocating to the state capital. The reason this property owner was willing to hire a drug felon became apparent soon enough. The apartments were in disrepair, with broken heaters and plumbing, and the owner forced his property manager to deal with angry tenants. She had gone to work for a slumlord.

Albritton quit and took a bartending position at the restaurant attached to a Holiday Inn near Louisiana State University. Tips included, she was earning about $15,000 a year, but she liked her co-workers and impressed her bosses. One of them tried to promote her to shift supervisor, Albritton recalls, but the promotion was denied when a criminal-background check by the hotel chain’s corporate office flagged the Houston conviction. She could pour drinks and do nothing more. She remembers how desperate she had been to leave her jail cell, naïvely believing that the punishment for pleading guilty would end with her sentence. “No,” she says. “You’re not ever free and clear of it. It follows you everywhere you go.”

In the two years since the efforts to overturn wrongful convictions began at the Harris County district attorney’s office, Inger Chandler and her colleagues at the integrity unit have struck 119 N.C.S. convictions from the record. At least 172 remain. They haven’t been able to locate all of the wrongly convicted, at times even after hiring private investigators, and some defendants they have reached have declined to interact with the courts, even to clear their record. Last year, as we examined records in Harris County, we came upon Albritton’s file and decided to search for her ourselves to find out what had happened to one representative figure out of hundreds. Her case fit the larger pattern of convictions for no controlled substance: It moved rapidly, with Albritton pleading guilty within 48 hours of her arrest, and it involved an exceedingly small amount of supposed drugs. We searched for Albritton in public databases, finding likely relatives but no phone numbers or a current address. We called her sister, who said that Albritton was in Baton Rouge and provided a cellphone number. It was disconnected. But knowing where Albritton lived now, we found a Facebook profile she had been updating regularly with details of her life, including her work. Interestingly, we also found that Albritton had pleaded guilty to a 2008 misdemeanor, a D.U.I. conviction in Louisiana, despite breathalyzer results showing her blood-alcohol level at 0.0. When we asked her about this, she said that she had caused a collision by pulling onto the wrong side of a two-lane highway, and because she was guilty of that, she did not protest the other charges; she’s still unable to explain why she confessed to a crime there was no evidence she committed.

In August, we called and left a brief message for Albritton at the Sporting News Grill. She returned the call a couple of hours later, her voice small, wondering what this was about. When we described the details from the lab report and the letter from the district attorney that she never received, Albritton gasped. She didn’t make a sound for several seconds before shouting into the phone: “I knew it! I told them!”

If Albritton’s case is one of hundreds in Houston, there is every reason to suspect that it is just one among thousands of wrongful drug convictions that were based on field tests across the United States. The Harris County district attorney’s office is responsible for half of all exonerations by conviction-integrity units nationwide in the past three years — not because law enforcement is different there but because the Houston lab committed to testing evidence after defendants had already pleaded guilty, a position that is increasingly unpopular in forensic science.

Crime labs have been moving away from drug cases to focus on DNA and evidence from violent crimes. In some instances, the shift has been extreme. The Las Vegas Metropolitan Police Department’s forensic laboratory analyzes the evidence in, on average, just 73 drug cases a year, internal records show. Nearly all of its 8,000 annual possession arrests rest exclusively on field-test results.

The United States Department of Justice was once among the leading voices of caution regarding field tests, and encouraged all drug evidence go to lab chemists. But in 2008, the Justice Department funded a program developed by the National Forensic Science Technology Center, a nonprofit that provides crime-lab training, to reduce drug-evidence backlogs. Titled Field Investigation Drug Officer, the program consisted of a series of seminars that taught local police officers how to administer color field tests on a large scale. In its curriculum, the technology center states that field tests help authorities by “removing the need for extensive laboratory analysis,” because “the field test may factor into obtaining an immediate plea agreement.” The Justice Department declined repeated interview requests.

Field tests provide quick answers. But if those answers and confessions cannot be trusted, Charles McClelland, the former Houston police chief, says, officers should not be using them. During an interview in March, McClelland said that if he had known of the false positives Houston’s officers were generating, he would have ordered a halt to all field testing departmentwide. Police officers are not chemists, McClelland said. “Officers shouldn’t collect and test their own evidence, period. I don’t care whether that’s cocaine, blood, hair.”

Judges, too, have the power, and a responsibility, some argue, to slow down the gears of the system. Patricia Lykos, the Harris County district attorney from 2009 to 2013, says that when she served as a criminal-court judge in the 1980s and 1990s, she would ask the defendants questions about their lives and the crimes they were accused of committing. If she wasn’t satisfied that the defendant was guilty of the charge, Lykos says, she wouldn’t accept the plea. At times the situation is even easier to decipher, says David LaBahn, president of the Association of Prosecuting Attorneys. The defendant can be heard arguing his or her innocence to the appointed attorney. In such drug-possession cases, when the prosecutor doesn’t have a lab report, “if I’m that judicial officer, this case is continued” — adjourned — “until everybody can do their job,” LaBahn says.

But that means the defendant, depending on his or her custody status, could go back to jail until the case proceeds, presenting a significant dilemma. Last year, Devon Anderson, the current Harris County district attorney, prohibited plea deals in drug-possession cases before the lab has issued a report. The labs issue reports in about two weeks, but defendants typically wait three before they can see a judge — enough time to lose a job, lose an apartment, lose everything. And yet since Anderson implemented the rule, case dismissals have soared 31 percent, primarily because the lab has proved defendants not guilty. People plead guilty when they’re innocent because they see no alternative. People who have just been arrested usually don’t know their options, or even that they have an option. “There’s a fail-safe in there, and it’s called the defense lawyer,” says Rick Werstein, the attorney now representing Albritton as she seeks to finalize her exoneration. Defense lawyers can demand a lab analysis, and they exist to help defendants navigate the consequences of the jail time while they wait, even as they explain the even higher costs of a felony conviction. They are fully authorized to pursue alternative deals.

In fact, Richardson, Albritton’s original court-appointed lawyer, says the prosecutor offered her a deferred adjudication, in which she may have been able to wait for the results of a lab test outside the walls of a jail cell. Richardson, who first said he had no memory of their conversations, says he told her about the offer but she refused it. Albritton says she has never heard of anything called deferred adjudication. Neither could explain what actually happened. Perhaps they simply accepted that the field test, with its promise of scientific inevitability, would eventually convict her. “The entire country works on these field-test kits, right?” Richardson asks.

In the past three years, people arrested based on false-positive field tests have filed civil lawsuits in Sullivan County, Tenn.; Lehigh County, Pa.; Atlanta, Ga.; and San Diego, Calif. Three of the four cases also named the manufacturers Safariland Group or Sirchie as defendants. Three of the cases have already been settled. In one of them, the Sullivan County case, Safariland secured a gag order on the plaintiff, explicitly to prevent media coverage, before entering settlement negotiations. The plaintiffs in each of the suits were people who were arrested, refused to plead guilty and were detained for a month or longer. So far, we have been unable find anyone who pleaded guilty based on field-test results and later filed suit, though Werstein said he and Albritton are considering their additional legal options.

The Texas Criminal Court of Appeals overturned Albritton’s conviction in late June, but before her record can be cleared, that reversal must be finalized by the trial court in Houston. Felony records are digitally disseminated far and wide, and can haunt the wrongly convicted for years after they are exonerated. Until the court makes its final move, Amy Albritton — for the purposes of employment, for the purposes of housing, for the purposes of her own peace of mind — remains a felon, one among unknown tens of thousands of Americans whose lives have been torn apart by a very flawed test.

To Learn More:

            Busted (by Ryan Gabrielson and Topher Sanders, ProPublica)

Tests Used in 6 States to Identify Marijuana-Impaired Drivers Called Unscientific by AAA (by Joan Lowy, Associated Press)