IN THE EARLY AFTERNOON of May 14, 1982, a man carrying a gun, wearing a hooded gray sweatshirt, and holding a cloth in front of his face walked into a North Dallas apartment complex, confronted a woman in a bathing suit, and ordered her to take it off. She refused, and the man fled to a nearby complex. There, he found another sunbather and took her back to her apartment, where he exposed himself to her. Over the next twelve hours, he went to three more complexes, forcing women to disrobe, exposing himself, and in one instance, raping a woman while her two-year-old son played in a nearby room. The next morning a man armed and dressed in a similar manner went to three Elaine Powers Figure Salons in Garland, Carrollton, and Denton and made four dozen women take off their clothes. “Gunman Terrorizes 61 in 2-Day Spree,” read the headline in the Dallas Morning News.

Five days later, Steven Phillips, a 24-year-old Dallas roofer with a penchant for flashing and voyeurism, was arrested for exposing himself to two young women at two North Dallas apartment complexes. His picture was taken, his wife posted bond, and he was released. A week after that, a man wearing a hooded gray sweatshirt, carrying a gun, and holding a cloth over his face struck two health spas, a condominium complex, and an outdoor tennis court in the Kansas City area, forcing two women and five girls to disrobe while he exposed himself. Dallas and Kansas City police consulted with each other, and a photo of Phillips was put in a lineup in each city. One victim from Kansas City and two from Dallas picked Phillips out, including the rape victim. Dallas police issued a warrant for Phillips; after he read about it in the newspapers, he turned himself in, denying his guilt. Three more victims identified him from a live lineup. Several women from both the Denton and Kansas City crime scenes identified a second suspect.

No fingerprints, gun, or gray sweatshirt were ever found to connect Phillips to any of the crimes. A tiny amount of seminal fluid was taken from the rape victim, but not enough to identify anyone, at least not with the technology of 1982. Three body hairs were also found, but they were useless for identification. The evidence was stored away, and Phillips went to court, where the case against him was built almost exclusively on witness testimony. At trial, in August 1982, the rape victim told how she briefly got a look at his face when he struggled with the front door on his way out and then stopped for a few seconds to hide his gun. She described him as hairy, muscular, having a mustache and a receding hairline, and wearing running shoes. His eyes, she said, were dark blue. “I’ll never forget those eyes,” she said in court. A second victim from that first day identified him too, as she had done at a photo lineup. She said he wore work boots and had “piercing blue” eyes. The only other evidence was the testimony from the policeman who’d arrested Phillips for flashing, who said he remembered seeing a gray sweatshirt in Phillips’s car but could not say whether it had had a hood. “I did not look at the item,” he said at the trial. “I just observed the item.”

Phillips has green eyes, and he also had three alibis for the afternoon of the rape: his wife, plus her sister and her sister’s husband, all of whom lived with Phillips in a rented house in Garland. Each said that when the assault took place, he was at home sleeping after a morning roofing job. But he was found guilty in separate trials, first of burglary and then of aggravated sexual abuse, and got thirty years for each. In 1984, facing a third jury trial with similar witness testimony, he pled guilty to the remaining charges from the two-day spree in exchange for a ten-year sentence, knowing he’d be eligible for release in the mid-nineties. He went to prison and spent much of his time over the next thirteen years writing hundreds of letters to judges, policemen, lawyers, and the media, proclaiming his innocence. In 1996 he was released, but he quickly fell back into his bad habits. He was arrested again, in 1997, for peeping into an apartment. His parole was revoked, and he was sent back to prison, where he sits today, at the Stiles Unit, in Beaumont.

THE WAY WE SOLVE crimes has changed remarkably since 1982, when Phillips was first convicted. Forensic DNA testing, first used to solve a rape and murder in England in the mid-eighties, has been universally adopted in the United States, where it has nailed thousands of guilty criminals and exonerated some 166 falsely convicted men and women, including 20 Texans (see “Body of Evidence,”). In April 2001, on the heels of a spate of these exonerations, the Texas Legislature passed an emergency measure that became Chapter 64 of the Code of Criminal Procedure, giving convicts the means to ask for DNA testing; previously, they had to file a writ of habeas corpus. Texas inmates can now request testing on evidence that was collected either before DNA technology was available or in its early, less accurate days. It is up to a judge to decide whether to grant a test, based on whether there is existing doubt about the perpetrator’s identity and whether the evidence would have made a difference in the verdict. (Some DNA evidence, such as hair found on the floor after a liquor store robbery, wouldn’t necessarily prove anyone’s innocence.)

Steven Phillips was one of the first to file. In the spring of 2001, he requested testing of all existing evidence, which turned out to be a vaginal swab and three body hairs collected from the original rape case that were being stored at the Southwest Institute of Forensic Science, in Dallas. In a letter to texas monthly shortly after, he again proclaimed his innocence, even as he acknowledged his guilt of other things. “My credibility sucks,” he wrote. “I am a (recovering) sex addict, with a 30 year history of misdemeanor acting out—exhibition, peeking tom, porno, etc.” But he wasn’t, he insisted, a rapist. Nearly three years later, on March 9, 2004, district judge Mark Nancarrow turned Phillips down. “Identity is not, and was not at issue,” he wrote in his opinion, adding that Phillips “is not entitled to testing of the vaginal swab because overwhelming evidence existed of Phillips’ guilt.”

The “overwhelming evidence” was the testimony of the two witnesses, particularly the rape victim. But while eyewitnesses can offer the most compelling evidence a prosecutor puts in front of a jury, study after study has shown that victims and witnesses get it wrong as often as they get it right, for all kinds of reasons: stress, the inherent limitations of memory, even the subtle suggestions of the police. In fact, an astounding 75 to 80 percent of the 166 Americans who have been exonerated by a DNA test were put in prison mainly because of faulty eyewitness testimony. And multiple eyewitnesses, as in the case with Phillips, don’t necessarily mean more certainty. In 1980 Luis Diaz was convicted of being a serial rapist in Miami after eight victims identified him. In 1985 Kirk Bloodsworth was convicted in Maryland of murdering a nine-year-old girl and given the death penalty based mainly on the testimony of five eyewitnesses who placed him with the victim. In both cases, the eyewitnesses were wrong. DNA cleared Diaz and Bloodsworth, and they were freed.

Phillips isn’t the only Texas inmate who has been denied a DNA test. Over the past five years, an estimated 800 convicts have made a request, but only 83 have seen it granted. “It’s harder to get DNA testing in Texas than any other state,” says Nina Morrison, a lawyer at the New York–based nonprofit Innocence Project. “The Texas statutory language is generous; it strikes an appropriate balance between granting testing in cases where DNA is relevant to the identity of the perpetrator and not granting it in cases where there are no doubts.” The problem, says Morrison and other lawyers, is that many Texas prosecutors aggressively fight DNA motions no matter what the merits of the request, and many judges apply a strict interpretation of Chapter 64, one that can make it almost impossible to get a test. Even defense attorneys concede that as many as three quarters of the requests may be frivolous—cases in which DNA evidence was not left at the crime scene or wouldn’t make any difference anyway. But in cases like Phillips’s, where DNA evidence, however old, is the only physical evidence available, it’s hard to understand why a judge would refuse to allow a scientific test, one that takes less than 48 hours and only costs about $1,000.

William S. Sessions, a former Texas federal judge, was a pioneer in the use of forensic DNA during his tenure as director of the FBI, from 1987 to 1993. Under his leadership, the bureau began testing DNA from active cases; in the first one hundred, the samples from thirty suspects did not match samples taken from the crime scenes. This opened Sessions’s eyes, and time after time since then, he has seen DNA clear up doubt, especially in cases like Phillips’s. “With those people who have consistently maintained their innocence and consistently pursued it,” he told me during a phone interview in October, “and in whose cases violence was committed or sexual conduct was involved, if DNA might be of consequence in those cases, then it should be tested. Why should a court or DA not be willing for justice to prevail?”

ONE OF THE MOST PRESTIGIOUS DNA labs in the world sits in a large gray building amid dozens of modern warehouses in the North Dallas suburb of Farmers Branch. Orchid Cellmark was formed by a consolidation of four of the world’s first DNA labs. The company has worked on cases from O.J. to JonBenet and has helped identify World Trade Center victims whose remains were mere wisps of cellular matter.

On a cold December day, I visited the lab and met with Robert Giles, Orchid Cellmark’s executive director. A longtime veteran of forensic testing, Giles, 53, is a quiet and studious man. He also has plenty of patience, a virtue every DNA scientist needs, especially when dealing with people who never took classes in cellular biology or population genetics. Over the course of several interviews, he had explained—and explained again—the science of DNA testing, and we had discussed Phillips’s case.

Orchid Cellmark is no liberal defense group; though the company does pro bono work for indigent convicts seeking DNA testing, it also does a lot of work for police labs. Giles agreed to run a profile of Phillips’s DNA with the understanding that, since we were not following the standard protocols, it would not be admissible in court. Under normal circumstances, there would be a strict chain of custody (a licensed technician would obtain a sample from Phillips in the presence of a witness). But given that neither of us had legal access to the inmate, Phillips had instead simply chewed on a piece of toilet paper and sent it to me in the mail. I brought the unopened envelope with me to Orchid Cellmark. We had no way of knowing whether it had been contaminated, or even if it had really come from Phillips. But the goal of the exercise wasn’t to figure out whether the sample was really his. I wanted to understand how a DNA test works and, in the process, determine just how easy it would be to grant Phillips’s request.

After I arrived, Giles presented the envelope with the sample to analyst Jody Hrabal, who would walk me through the testing process. Hrabal, 28, has been with Orchid Cellmark for six years. One of her duties is to testify before juries, a task that has required her to come up with a concise explanation of a very complicated science. “I always make it short and simple,” she told me. “I say, ‘DNA, or deoxyribonucleic acid, is a molecule that exists in almost every cell of your body—your blood, saliva, hair, bones—that you get it from your mom and dad, and that ninety-nine percent of your DNA is the same as everyone else’s, but the one percent is what makes you unique. What we do at the lab is look at thirteen locations, or loci, on your DNA, where there are markers that repeat themselves a set number of times. Those thirteen loci combined make a unique profile for every person.’”

DNA’s markers are made up of sequences of the molecule’s four building blocks: adenine, thymine, cytosine, and guanine (abbreviated A, T, C, and G). These link with one another to form the spiraling rope ladder of DNA’s familiar double helix. There are 3 billion rungs on this microscopic ladder, and if unwrapped, the molecule would be anywhere from six to nine feet long. What forensic DNA analysts do is compare the thirteen loci of a suspect’s DNA sample—his unique profile—with the thirteen from a sample taken from a crime scene. If the number of repeating markers at all thirteen loci matches, the analyst can safely conclude that the statistical likelihood of this same DNA profile reoccurring in the general population is something like 1 in 200 billion. In other words, either detectives have found their man or something very strange has taken place, along the lines of a monkey’s hammering out a sonnet on a typewriter. (It was the FBI that chose these thirteen loci, precisely because they are so unique. Every time a DNA sample is collected at a crime scene or taken from a convict, the profile is entered into CODIS, the Combined DNA Index System, a nationwide database, which is a great crime-fighting tool, since most criminals are repeat offenders.)

Phillips’ sample was first given a number, then Hrabal and I put on lab coats, masks, and disposable latex gloves. She swabbed the inside of my cheek for a reference sample. It would be one of several control tests for contamination, which can be caused by exposure to another person’s DNA, especially in the early stages of its collection at the crime scene or in the lab. In rare instances, contamination can cause markers to show up at loci that the suspect in truth doesn’t have. DNA survives heat sterilization, so no test tubes are reused. “Everything is disposable,” said Hrabal.

Orchid Cellmark is as spare on the inside as it is on the outside, a minimally adorned maze of offices, storage areas (for the thousands of files it has worked on), and labs. We first went to a preparatory lab to remove the sample. Hrabal explained how all samples are considered either “known” or “questioned,” and since we weren’t positive where Phillips’s came from, we would call it questioned. She then opened the envelope Phillips had sent and pulled out another sealed envelope. Inside that was a small piece of carefully folded toilet paper, and inside that was the hard clump Phillips had chewed on. “Terrible toilet paper for the prisoners,” joked Hrabal. She took it into a separate room, turned off the lights, and shone a blue light on the wad; it glowed slightly, indicating saliva. “Semen glows brighter,” she said. She cut a small piece of the clump, put it in a pink tube, and we went to the extraction lab next door. There, two chemicals were added, one to break down the walls of the cells and the other to eat up all the other proteins and release Phillips’s DNA. Hrabal capped the tube and put the same chemicals in a separate tube, called the reagent blank, which would go through the whole process and be used as a check for contamination.

For the next couple of hours, the sample in the tube was shaken on a vortex machine and spun in a centrifuge, put in warm water, and dosed with ethyl alcohol and sodium acetate. At every step, Hrabal took elaborate notes, entered figures on a computer, and cleaned up instruments and work benches with bleach. “It’s kind of repetitive,” she said at one point. I told her that you probably couldn’t be much of a DNA analyst if you weren’t detail oriented. She laughed. “I don’t think you’d last very long if you weren’t.”

Finally, the sample was ready to be copied. In the early nineties, DNA testing required a quarter-size sample of blood or saliva, which was often used up in the process. Nowadays, through a process called polymerase chain reaction, or PCR, just a handful of cells taken from the skin, ear wax, or dandruff can be used—even those taken from samples degraded by the elements or partially devoured by bacteria. From these cells, a billion copies of a DNA profile can be made, allowing a sample to be tested immediately or far into the future. To get the Phillips sample ready, Hrabal added primers, which would identify the thirteen loci, and enzymes, which would do the actual copying. The tube was next placed in a thermal cycler, which would heat up and separate the strands of DNA and then cool them for the next three to four hours.

THE FOLLOWING MORNING (in the real world, a DNA test at Orchid Cellmark takes several days to run), after more shaking, stirring, and, in Hrabal’s words, “chemical manipulation,” the sample was put in an ABI Prism Genetic Analyzer, where it traveled through superthin glass capillaries, was zapped by a laser, and then photographed by a camera. The results were sent to a computer, and then, less than a day after we started the whole process, Hrabal called up a graph showing Phillips’s genetic profile.

There they were, the thirteen loci, each represented by one or two tall, thin peaks. Under the peaks were figures showing the number of markers at each locus. Though each human has the same marker at each locus—say, AGAT—there are only a certain number of repeats possible at each one. Phillips had 15 and 17 at his first, 21 and 23 at his third, and so on. I asked Hrabal if it all looked normal. “Yes,” she replied. Anything weird? “No. All the controls and the reagent blank check out.”

If this had been a valid forensic DNA test, Hrabal would have next compared Phillips’s profile with any male profile taken from the vaginal swab being stored at the lab in Dallas. Hrabal explained what she’d do. “Assuming that it was a mixture, we would first account for the victim’s profile, which we would know, and subtract it. What would be left is the male profile.” The number of repeats at each locus of the two samples would be compared, and while the odds that two males would have the same number of repeats at one locus aren’t so bad, they decrease when two locations are compared, then three, on up to thirteen. If all thirteen matched, one could reasonably conclude that Phillips is a liar as well as a rapist. If, on the other hand, just one locus was off, Phillips would be excluded, even if twelve others matched. “Your DNA is going to match your DNA,” Giles had told me, “no matter where it’s from, no matter when you took it. One group of cells from your body looks like every other.”

Most of the time with PCR testing, thirteen loci show up, and there’s either a match or a non-match. The problems come in a third area, when only six or eight or twelve can be read, usually because the sample is a hard-to-analyze mixture, such as one with more than two profiles. Unfortunately, whether there is in fact a gray area can also be a matter of interpretation. Giles says labs have been known to “call” matches at loci when they really shouldn’t. “Sometimes the loci don’t meet the lab’s own standards for reporting, but the analyst still calls them. He’s speculating. It’s usually because of poor training.” Giles says that analysts will also force the data for other reasons. “I think analysts at some labs, without training and under a lot of pressure to get something done—to get it to the DA’s office—are really squeezing the data, rerunning the tests, saying, ‘I’ve got to get this to work. I think I can call this.’ You have to go by your protocols. You can’t be pressured.”

In criminal cases, it’s the analyst who goes to court to give testimony in front of a jury of nonscientists, and here, says Giles, who has testified some two hundred times, the expert has to be cautious. “The idea is to try to be as conservative as you can in interpretation,” he says. “If you have some loci that match but others that can’t be read, it might be best to say the suspect ‘can’t be excluded,’ which is a very conservative way of saying, ‘His type is present.’ But you have to be careful about what you say on the stand. Some analysts get caught up in being an expert and interpret everything as being clear even when it’s not. You can’t overinterpret the data.”

IF YOU WANT TO MAKE a forensic scientist wince—particularly one who has proselytized about the power and accuracy of DNA—just mention three other letters: HPD. DNA, for all the crimes it has solved, is still struggling to overcome the damage its reputation received from the Houston Police Department Crime Laboratory, which was closed in 2002. An independent audit of the lab’s DNA section had uncovered a mess, the worst investigators had ever seen: analysts who didn’t know what they were doing, supervisors who knew even less, no quality-control system, and very few standard operating procedures. Other HPD crime lab sections had issues too—there were problems in the ballistics and toxicology labs, and analysts in the controlled-substances section were actually making up results—but it was the DNA section that caused the most alarm, casting doubt on thousands of DNA tests done over the years in Harris County, which has sent more men than any other county in the U.S. to death row. Retests in cases in which DNA had played a part began almost immediately. One of the first only seemed to confirm just how screwed-up things were: Josiah Sutton, who had been convicted of rape in 1998 at age sixteen and given 25 years largely on the basis of DNA evidence, was exonerated by a new test. “Worst Crime Lab in the Country,” read the headline in the New York Times the next day.

There were a lot of reasons for the Houston crime lab disaster, but most had to do with people, not science. James Bolding, who would become the scapegoat, was a crime lab drug chemist who was given the go-ahead to start the DNA section in 1989, but he had a limited budget to pay and train analysts. Of the first five, hired by 1991, only one had taken all of the college classes necessary for a basic understanding of DNA, such as statistics, genetics, and molecular biology. Bolding himself attended two FBI courses on DNA testing but never did the sample analyses required to get full course credit. Indeed, the head of the DNA section never once actually worked on a DNA sample at the crime lab. As DNA technology changed through the nineties, the HPD’s analysts got no training in the newer methods. They were overworked and inconsistently following protocols. And they learned some bad habits. Analysts didn’t know how to handle mixtures—how to isolate and identify the sperm from a vaginal swab, for example, and then how to analyze the results. In the Sutton case, the analyst claimed to have found the suspect’s profile in a car-seat stain and also in a vaginal swab; the profile was so unique to Sutton, she wrote in her report, that it could be expected to occur in only “one out of 694,000 people among the black population.” In fact, Sutton’s profile wasn’t in the car-seat stain at all, and although it was consistent with the one found in the swab, so were the profiles of one in every 15 black men, not one in every 694,000. It was a serious statistical error.

The lab was never inspected by an outside agency, nor did it aim for voluntary accreditation, both of which are required to meet FBI standards. In 1996 a man named Lynn Jones was excluded as a rape suspect by a DNA test after spending nine months in jail waiting for the results. During that time, three confused analysts had passed the sample to one another. In response, the HPD audited the section and found little supervision and no case-management system. Still, department higher-ups did nothing to fix things, which was not surprising, given their attitude about more-obvious problems. The next year, the HPD moved into a 26-story, sixties-era downtown skyscraper with a bad roof. When it rained, water leaked onto the top floor, where several sections of the crime lab, including DNA, sat. The leaks were never patched, and in May 2001 Tropical Storm Allison poured water through the roof, soaking DNA evidence in some three dozen murder and rape cases; bloody water was seen seeping out of boxes.

Elizabeth Johnson, a molecular biologist who set up and ran the DNA lab at the Harris County Medical Examiner’s office from 1992 to 1996, was one of the HPD’s earliest critics. “It always boils down to the people,” she says. “They have to be honest and careful. At HPD, they were ignorant and incompetent. Plus, the city allowed it to go on. Nobody slapped anyone’s hands—and this was a validation to them.” She says that analysts at the crime lab had a stubborn, cop-in-a-lab-coat law-enforcement mentality. “There was this ingrained bias—‘The guy is guilty,’” she says, “and if the data was anywhere close, and sometimes even when it wasn’t close, they were going to push it through.” She saves special scorn for Bolding, who, it turned out, had a direct role in the second exoneration to come out of the HPD mess. George Rodriguez had been convicted of rape in 1987 after Bolding, then the head of the serology section, testified that blood-typing tests on crime scene semen proved another suspect in the case could not have been the donor; therefore, prosecutors said, it had to have been Rodriguez. The jury agreed and gave Rodriguez sixty years. DNA testing in 2004 of a crime scene hair incriminated that other suspect, and Rodriguez was released. A 2004 panel of experts said Bolding’s 1987 testimony contained “egregious misstatements.”

Prosecutors, judges, and defense attorneys were also to blame for the false convictions. “Everyone in the system was ignorant about DNA evidence and how to spot sloppy work,” says Marie Munier, the Harris County assistant DA in charge of all the retests. Johnson puts some of the blame on defense attorneys, who should have understood not only the basics of DNA but also the problems associated with mixtures and partial profiles and what “can’t be excluded” really meant. Sutton’s attorney, for example, didn’t object when the crime lab analyst didn’t even use statistics in her testimony but said only that Sutton’s profile could be found in three different crime scene samples. “They didn’t squawk enough,” says Johnson. “They needed to raise holy hell, and they didn’t.” Munier adds, “The prosecutor allowed it, the defense attorney didn’t say, ‘Wait a minute,’ and the judge didn’t recognize it either. It’s a failure of the entire system.”

THE DNA SECTION IS still closed, and the entire lab remains under two different investigations. The first, being conducted by the Harris County DA and the HPD, has completed 404 out of 422 DNA retests. Sutton’s was the first and so far the worst for the lab. The rest have mostly corroborated the lab’s original work: 334 have been confirmed; 34 are still being tested; 18 cannot be confirmed at present; 11 more are, in Munier’s words, “funky cases” that fit no category; and 6 have had major statistical problems, but ones that don’t put the verdict in doubt, at least not yet. The second investigation is being done by a team led by Michael Bromwich, a former Department of Justice inspector general, who finished a first phase in June. “We do not yet know,” he wrote at the end of his report, “whether the well-publicized cases of the Crime Lab’s failures are isolated analytic breakdowns or only the tip of an iceberg of widespread analytic failures, incompetence, or worse.” He hopes to find out in the second phase, in which investigators are currently analyzing, but not retesting, 2,799 cases.

With the bloody, waterlogged boxes and all the other terrible things that went on at the HPD crime lab, one might assume that there should have been a lot more Suttons found in the retests. Perhaps this is evidence of DNA’s inherent soundness, even when processed by a bunch of overworked, poorly paid biology majors. Indeed, studies done on DNA have found that simple negligence won’t contaminate a sample; for that to happen, analysts have to do really stupid things, such as spray a mist of foreign DNA into an open test tube. Giles agrees that the science of forensic DNA testing itself is sturdy—“Most of the mistakes seen to date have been really big ones, sloppy ones like switching samples”—though he’s not so certain there won’t be more bad news out of the HPD. Neither is Bromwich. In a November interview, he told me, “The book’s not closed yet. We’re in the middle of reviewing the retests, as well as many other DNA cases, and we don’t know yet whether there will be any other cases like Sutton’s. We won’t know until our case reviews are complete.” That should be sometime this year.

Plenty of the state’s crime labs (there are 56—13 run by the Department of Public Safety, plus other city, county, and private labs) have had similar problems as the HPD’s, and most were also the result of incompetence and poor training. In a 2002 Fort Worth case reminiscent of the HPD case of Lynn Jones, a man named Jonathan Byner was arrested for rape and sat in jail for eighteen months awaiting DNA tests from the city police department’s lab. They didn’t come until the sample was sent to the Tarrant County Medical Examiner’s office, where, three days before his trial, they proved he didn’t do it. The reason for the wait: The police department’s lab had been closed after a 2002 retest raised questions about an analyst’s work. Audits done between 1999 and 2003 at DPS labs, which do about half of all forensic testing in Texas, found many of the same problems the HPD had had, from analysts who didn’t know how to interpret mixtures to analysts who didn’t know they should do tests on blank control samples to detect contamination. The worst non-HPD screwup by far is the case of Brandon Moon, who was convicted of rape in El Paso in 1988 based on an ID from the victim and faulty serology tests done on evidence that had been sitting in a police officer’s car for two days in the West Texas heat. In 1996 a DPS analyst reviewing the evidence wrote an internal memo in which she raised serious doubts about the serology test results; she also requested that blood be taken for DNA testing. Her concerns were ignored. When Moon finally did get a DNA test, in 2004, it showed he was not the rapist. He had spent seventeen years in prison for a crime he didn’t commit.

Sutton, Jones, Moon—time and again over forensic DNA testing’s short lifetime, the serious problems in DNA labs have proven to be problems of human error. The technology itself has only gotten more and more accurate and probative, to the point where police are now using DNA to solve burglaries, and crime scene evidence from all felonies, not just murders and sexual assaults, is entered into CODIS. In 2001 the Texas Legislature finally took notice of what a powerful tool DNA was, passing Chapter 64. Then, in 2003, it passed a law requiring that all state crime labs be accredited by September 1, 2005, at least if they wanted evidence from that field to be usable in court. Accreditation requires that labs meet basic standards for equipment, procedures, analysts, and supervisors and then undergo yearly internal inspections and external ones every five years. It won’t solve every problem; as University of California, Irvine, professor of criminology William Thompson points out, the organizations that do the accrediting are not government groups but professional associations, or, in his words, “old boys’ networks.” Bromwich is more sanguine: “Accreditation is an important threshold, a baseline. It doesn’t tell you as much as you want to know about the quality or accuracy of work done, but it’s a necessary and important first step.”

Or, as Irma Rios, a nineteen-year DPS veteran hired to oversee the HPD lab in 2003, says, “It’s a beginning.” In May every section of the crime lab except for DNA was accredited, and she thinks the DNA section will follow soon. And she says she and the department are determined to make things right again: “If somebody should be in prison who isn’t, we want to help put him there. And if anybody has been wronged, we want to make it right. ”

DON UMPHREY IS AN associate professor of advertising at Southern Methodist University. A religious man and a recovering alcoholic (sober 32 years), he spends time helping others with, in his words, “aspects of Christian recovery.” In 1992 Umphrey was preaching with his church group at the Coffield Unit, in Tennessee Colony, when he met Steven Phillips. The inmate later wrote to the professor, and they began a correspondence. When Phillips got out in 1996, Umphrey helped him get a job, and they became friends. “We hung out together,” remembers Umphrey. “We’d go to church together, go bowling. I treated him as a peer. He’s got a lot of interesting ideas.” Umphrey introduced Phillips to a friend who got him to go to a couple of Sex Addicts Anonymous meetings, but Phillips, who at the time didn’t think he had a real problem, quit. Not long after that, he was arrested for trying to peek into an apartment, and he was sent back to prison. Umphrey has stayed in touch, even if he is occasionally wearied by his friend. After Judge Nancarrow turned down Phillips’s request for DNA testing, Umphrey even offered to pay for it (the judge said no). Umphrey knows his friend is a sex offender. He doesn’t think he’s a rapist, but he also admits that he isn’t sure. “Either he did it or he didn’t do it,” Umphrey says. “But let’s find out.”

In order to qualify for post-conviction DNA testing under Chapter 64, a case must meet two criteria: First, identity must be an actual issue (if, for example, as in many rape cases, the victim knew the suspect, it isn’t), and second, a positive DNA test would have had to have made a difference at trial. When Chapter 64 first became law, in 2001, it was given a restrictive reading by the superconservative Texas Court of Criminal Appeals (CCA), which in the 2002 Kutzner decision wrote that inmates had to “show [that] a reasonable probability exists that exculpatory DNA results would prove their innocence.” In other words, in order to get tests to prove one’s innocence, first one had to prove one’s innocence. In response, the Legislature amended Chapter 64 in 2003, taking out the statute’s “reasonable probability” language. An inmate no longer has to prove his innocence; he just has to prove that if he’d had a favorable DNA test (one that excluded him), there’s a 51 percent chance that he wouldn’t have been convicted. Still, defense attorneys complain that some courts continue to apply the tougher Kutzner standard. “Kutzner should no longer be good law,” says Philip Wischkaemper, a lawyer with the Texas Criminal Defense Lawyers Association. Lawyers have other Chapter 64 complaints: There are no deadlines, so motions like Phillips’s can languish for years. There is no remedy given if the findings are favorable. And, if you’ve had testing once, it’s almost impossible to get it again. Bob Wicoff, an attorney for Josiah Sutton, says that because the wrongly convicted rapist had already had a test back in 1998, he would have almost certainly been turned down under Chapter 64: “If it hadn’t been for the media and the retesting program, he’d still be in jail.”

In the spring of 2001 hundreds of inmates, some of whom had found out about Chapter 64 from signs in prison libraries, filed for DNA testing. Many wrote obviously frivolous requests—for example, an HIV-positive inmate who wanted to find out if he’d gotten the disease from the woman he had raped. Other motions were made by inmates for whom the courts said it wouldn’t have made any difference: A death row inmate whose DNA had been found in a hair clutched in the victim’s hand was not now entitled to testing on the blood found under her fingernails. For some, there was nothing to test; evidence collected had been discarded long ago. Phillips was one of the lucky ones. He requested “testing on any and all biological evidence involved,” having no idea whether any DNA from his case still existed. It did.

Then his luck ended, like that of most inmates who have asked for a test. Nobody keeps track of how many tests have been requested since Chapter 64 became law, but based on rejection estimates from Dallas, Harris, and Travis counties, it’s fair to say that there have been about 800. Michelle Moore is an attorney with the Dallas County Public Defender’s office who represented Phillips for his motion. She says that the 83 tests that the DPS has granted so far are not nearly enough. “How pitiful,” she says. “That’s way too low.” Some defense attorneys, like Houston’s Randy Schaffer (who has used DNA to free three wrongly convicted men), blame the state’s law-and-order prosecutors for the low numbers. “Most big metropolitan DAs will fight tooth and nail over DNA. They are so afraid of wrongful convictions because they call into question the integrity of the process. The fallout is in the jury panels. After DNA exonerations, juries say, ‘We can’t convict on eyewitness testimony; we need DNA.’ District attorneys would rather see an innocent person stay in jail than run the risk of one hundred guys getting acquitted in the future.” John Rolater, a deputy chief of the Dallas County DA’s appellate division, disagrees. “We don’t want innocent people in prison,” he says; his office’s concern is frivolous requests. “A lot of our cases involve people with preexisting relationships with each other, such as intrafamily sexual assaults. Also, we have a lot of people trying to use DNA tests to cast doubt on other parts of the case.”

The ultimate arbiters in Chapter 64 requests, of course, are the judges. David Dow, of the Texas Innocence Network, at the University of Houston Law Center, says, “The statute promises a lot but delivers little. Unfortunately, it places so much authority in the trial judge.” And while some judges believe in DNA, others apparently do not. In front of Judge John Creuzot, of the 4th Criminal District Court, in Dallas, a defendant is likely to get a test. In front of Judge Nancarrow, of the 204th, just down the hall, he is not. Just how individual judges arrive at their decisions is a mystery to lawyers like Wischkaemper. “Judges go through all sorts of mental gymnastics to deny relief,” he says.

We don’t know why Judge Nancarrow turned Phillips’s Chapter 64 request down (he did not respond to interview requests), but we can guess at his motivations. Perhaps he was giving in to the very human impulse to want to aid the victim, to take her side against that of a pervert. Or maybe it’s a question of politics and the risks an elected law-and-order judge like Nancarrow would have to run in taking Phillips’s word against hers. But the fact is, rape victims, like other eyewitnesses, can be unreliable witnesses. Of the Texas men wrongly incarcerated for rape and freed by a DNA test, every single one was convicted mainly on the ID of the victim. The Innocence Project’s Morrison says, “One of the most consistent patterns we’ve seen in DNA cases is rape victims who, because of the enormously traumatic circumstances of the crime, are often honestly but tragically mistaken when attempting to identify their assailant.” Of course, it’s understandable why they would make mistakes. Memory is not objective and linear, like a movie; it is subjective and disjointed, like a dream, easily influenced from within the mind and without, especially considering the physical and emotional horror rape victims go through.

The 1982 Dallas victim in Phillips’s case never saw her rapist’s face close-up; he wore a hood over his head and kept a cloth over the bottom half of his face. She saw his profile for a few seconds as he was leaving, when she was in shock. She made a composite sketch with the police that prosecutors thought looked a lot like Phillips. But sketches are not perfect; the first American to be exonerated by DNA, Gary Dotson, had also been convicted after the victim made a composite sketch with help from the police and then identified him from photos and a lineup. “Most judges know that eyewitness testimony can be inaccurate,” says one-time Phillips attorney Moore.

Perhaps Nancarrow, as judges often do, merely followed the recommendations of the prosecution, which based its opposition to Phillips’s request using the outmoded 2001 statutory language—with its impossible standard of having to first prove one’s innocence to get a test. “Maybe,” Judge Sessions speculated, “it’s the mentality: ‘He may not have done this, but he probably did something else.’” Maybe it was an issue of money. Even though DNA testing is cheap now, Moore says, “If judges don’t have to spend money out of their budgets, they won’t.”

The hardest thing to reconcile about Nancarrow’s decision is that, ultimately, it wasn’t going to set Phillips free or even give him a new trial. It was only going to give him a test. If the test had excluded him, he would still have had to file a writ of habeas corpus or ask the governor for a pardon, and the prosecutor still could have fought him. And the fact is, most of the time, post-conviction DNA tests basically reconvict the convict. Of those 83 tests that have been granted, 51 have done just that. One Dallas defense attorney who says she has filed ten Chapter 64 requests, all of which were denied, says, “My philosophy is: Give it to him so he’ll shut up.”

But we also know that 6 of those 83 led to an innocent person’s going home again. That’s a pretty good batting average for DNA. Judge Jon Wisser, of the 299th District Court, in Austin, says, “It’s not right to pick people over science. My feeling is, if there’s any evidence that even remotely indicates we need DNA evidence, let’s test it. The threshold should be very low. The cost is so minimal. Why not do it? I like to sleep well at night.” Wisser, who has okayed one request and turned another down (a third is pending), says one of the problems is that judges have to balance finality with fairness. “Finality is so important to the system. There’s a constant pressure to move these cases. There’s almost a sense that it’s okay to let innocent people go to prison—it’s more important to have faith in the system. I don’t think anyone would say that, but a lot of people believe that.”

Tammy Ardolf was the Dallas County assistant DA who made the official response that opposed Phillips’s DNA request. She couldn’t talk about the case, which is still pending before the CCA, but during a phone interview in November, she explained to me her office’s philosophy on a Chapter 64 “spectrum” for DNA testing requests. On one end are requests her office would oppose: cases in which the victim knew the assailant or was caught in the act. On the other end are requests the office would not oppose: stranger-on-stranger crimes in which nothing except an eyewitness tied the person to the crime scene. In between is a gray area, such as cases like Phillips’s. “It’s difficult,” she says, “especially when you’re looking at the specific facts of a specific case. A lot of time it’s a judgment call, which in larger counties often is made by higher-ups rather than the individual prosecutor. But in my opinion, if it fits the statute, you should agree to DNA testing.” Ardolf now works as a staff attorney with the county criminal district courts, but while working for the DA, she persuaded prosecutors to agree to two DNA testing requests, those of Donald Wayne Good and Keith Turner, both of whom had been misidentified by eyewitnesses—and both of whom were excluded by DNA. “People make mistakes,” she acknowledges.

But Dallas County contested the 2001 Chapter 64 request of Wiley Fountain, who had been convicted of rape in 1986 with what the prosecutor called “unequivocal eyewitness testimony” from the victim but little else. Luckily for Fountain, Judge Vickers Cunningham granted the test anyway, it excluded him, and in 2002 he was freed. He had spent fifteen years in prison for a crime he didn’t commit.

IN EARLY NOVEMBER I visited Phillips at the Stiles Unit. He’s muscular, completely bald, and he wears wire-rim glasses, through which peer his pale-green eyes. He has a wide gap between his front teeth. He’s a grandfather now, though his wife and two children never visit. He stays in touch with two people, really: his mother and Don Umphrey. Phillips wrote his friend in October about how he had lost everything: “Well if it’s God’s plan that I should reach a point where all that’s left is my prayer, that is, my dependence on him, then I believe I have arrived.”

He’s smart, though some I spoke with think he’s manipulative. He got a correspondence bachelor’s degree from Stephen F. Austin State University in the nineties and spends a lot of time reading classic literature. He quotes from Melville and likes The Scarlet Letter. “Hester Prynne wears her sin, her A, for all the world to see,” he told me excitedly. “Thereby, she’s truly saved. The preacher keeps his inside.” He says he’s finally taken responsibility for his sin, his sex addiction, the “SO,” for sex offender, stamped on every piece of paper in his life. “I’m a sex addict, man. I always knew something wasn’t right, but I didn’t think it was an addiction. Once I was talking to a counselor and I said, ‘It’s not a big deal,’ and he said, ‘What about the victims?’ And I said, ‘There are no victims.’ I was so much in denial. I wasn’t trying to get well. I was a one-dimensional monomaniac. Sin’s fun, dude.” He says he didn’t realize until he came back to prison that he was an addict.

His appeal of Nancarrow’s denial is before the CCA. He’s up for parole now, though he says he stands no chance of getting it. He’ll get a mandatory release at the end of next year, but what he really wants is a DNA test to clear him of the rape. I asked if he has any doubt that the semen stain would prove to not be his. “No doubt at all.”

We know how easy it would be to give Phillips a test. If the samples match, that’s the end of the story, and DNA has once again reinforced a guilty verdict. But if they don’t match, there would be two tragedies—that of a man locked in prison for more than two decades for something he didn’t do and that of a victim who never got justice and a rapist who could still be free.

Maybe Phillips is manipulative. We know he’s a pervert, a sex addict. Five women say he’s a dangerous pervert, and one fervently believes he’s a rapist. Is he? Not the judge, not the DA, not his defense attorney, not even his best friend knows for sure. But there is one way to find out.