On a February afternoon in 1994, Maria Montalvo’s Volkswagen Jetta caught fire in a driveway in Long Branch, New Jersey. Strapped into the back were her two children — 28-month-old Rafael-Louis and 18-month-old Zoraida-Angelin. Neighbors and Montalvo’s husband pulled her from the driver’s seat, her coat in flames. They could not reach the kids. Both died. Montalvo, a registered nurse, was charged with murder. She has been in prison ever since 1.
The state’s case rested on a New Jersey arson investigator who told jurors he could read the geometry of the burn — a “V-pattern” on the seat, “low and deep” charring, a survival-window claim about who could and could not have escaped a fully involved vehicle fire. He demonstrated his methodology in trainings by setting a pizza on fire in a garage and inviting trainees to interpret the marks 2. On that testimony, the jury convicted. The judge sentenced her to 100 years.
Almost everything that investigator told the jury has, in the three decades since, been thoroughly repudiated by the fire-science profession itself. The “V-pattern” doctrine. The “lowest and deepest burn” rule. The “pour patterns.” The “no one could survive that” certainty. A 2021 federal report concluded that pinpointing fire origin from burn patterns in a fully involved fire is no better than random chance. New Jersey Governor Phil Murphy commuted Montalvo’s sentence in 2024. The parole board denied her release in 2025. In February 2026, a trial judge rejected her motion for a new trial — adding a note that calling the fire an accident would “dishonor the memory of those two children” 3.
That sentence — doubt as dishonor — is the entire problem in miniature. And it is not a New Jersey problem. Georgia has its own Maria Montalvos. Some of them have been freed. Most have not. And in Georgia, more than in almost any other state, the legal architecture for correcting the record when the science collapses barely exists.
The Pizza in the Garage
For most of the twentieth century, American arson investigation was not a science. It was a craft, passed from one generation of fire investigators to the next as a set of rules of thumb: alligator-pattern char on wood meant accelerant; spider-web “crazed glass” meant rapid heating; low burns on baseboards meant the fire’s point of origin; “pour patterns” on floors meant gasoline; fires that moved fast meant arson. Investigators testified to these indicators with the certainty of priests reading entrails.
The collapse of that orthodoxy began in October 1990, when fire chemist John Lentini and Kirk’s Fire Investigation co-author John DeHaan obtained permission to burn down a condemned house next door to a Jacksonville, Florida home where six people had died. The condemned house was an architectural twin of the Lewis residence. They furnished it identically — same couch, same sheetrock, same carpet — and lit the couch with a single match and no accelerant. Within four minutes, the room reached flashover. After the burn, the floor showed “pour patterns.” It showed deep charring. It showed every marker investigators had used for decades to convict thousands of Americans of arson 4. Lentini later wrote that he “had come within 24 hours of giving testimony that could well have sent an innocent person to Florida’s electric chair” 5.
In 1992, the National Fire Protection Association published NFPA 921, the consensus guide that required fire-cause determinations to be made using the scientific method. Subsequent editions explicitly disclaimed the indicators on which a generation of arson convictions had been built. Yet investigators trained in the older orthodoxy continued to testify to it for decades. They had no incentive not to.
Cameron Todd Willingham was executed by Texas on February 17, 2004, for the 1991 deaths of his three young daughters in a Corsicana house fire. The state’s deputy fire marshal told jurors the burn patterns were the “fingerprints” of arson, that the fire “spoke” to him, that Willingham had used a liquid accelerant to trap his children. Days before the execution, Cambridge-trained chemist Gerald Hurst submitted a report to Governor Rick Perry’s office concluding that “there’s nothing to suggest to any reasonable arson investigator that this was an arson fire. It was just a fire.” Perry’s office did not stay the execution.
The Texas Forensic Science Commission later retained fire engineer Craig Beyler to review the case. Beyler concluded the deputy fire marshal’s testimony was “hardly consistent with a scientific mind-set and is more characteristic of mystics or psychics” 6. Two days before Beyler was scheduled to testify, Governor Perry replaced the Commission’s chairman. The eventual final report acknowledged “flawed science” but found no misconduct.
Han Tak Lee lost twenty-four years to the same junk in Pennsylvania 7. Daniel Dougherty was sentenced to death in Pennsylvania on V-pattern testimony then-debunked by NFPA 921. Kristine Bunch served seventeen years in Indiana. Adam Gray was convicted in Illinois at age fourteen and served twenty-four. Louis Taylor served forty-two years for the 1970 Pioneer Hotel fire in Tucson. Ernest Ray Willis was on death row in Texas for seventeen years before fire-science review exonerated him in 2004.
Each of those names is a measurement of how slowly the legal system absorbs what its own forensic profession has already conceded.
“Indeed and Without a Doubt”: The Bite Mark Scandal Comes to Georgia
In 2006, a Ware County jury convicted Sheila Denton of murdering her partner Eugene Garner. The case was ambiguous on almost every front — except for the bite mark on Garner’s right arm, which an American Board of Forensic Odontology Diplomate testified was “probably made by Sheila Denton” to “a reasonable degree of scientific certainty.” The jury believed him. Denton went to prison.
She stayed there for fifteen years and eight months 8.
In February 2020, Ware County Superior Court Chief Judge Dwayne Gillis vacated the conviction. His order is one of the bluntest documents any Georgia judge has produced on the subject of forensic science. The bite mark evidence used at trial, Gillis wrote, “is now known to be unsupported by science.” Bite mark testimony, he added, “will seldom, if ever, be probative” — and “the future of admissibility of such evidence is dubious at best” 9. The state declined to retry. Denton walked out of prison on April 8, 2020.
What the Denton record makes plain is that the prosecutor knew the science was thin even at the time of trial. Reading the closing argument back today is its own kind of confession: the prosecutor told the jury that the bite mark testimony was “not like DNA” and “not as definitive” — and asked them to convict on it anyway 10.
By the time Denton was freed, the scientific reckoning over bite mark analysis was already complete. The 2009 National Academy of Sciences report concluded that the underlying science was “insufficient.” The 2016 report from the President’s Council of Advisors on Science and Technology was sharper still, finding bite mark analysis “scientifically unreliable” and stating that “additional research in this area is not warranted” 11. The Texas Forensic Science Commission became the first government scientific body in the country to recommend a courtroom moratorium on the discipline. The Innocence Project has now catalogued at least 36 wrongful convictions involving bite mark testimony nationally.
The names form a roll call. Keith Allen Harward — Virginia, six ABFO Diplomates testified the marks on the victim’s leg matched him to a “practical impossibility” of being anyone else; thirty-three years lost; exonerated by DNA in 2016. Eddie Lee Howard — Mississippi, sentenced to death in 1994 on testimony from Dr. Michael West that the bite marks matched “indeed and without a doubt”; twenty-six years on death row before DNA cleared him. Levon Brooks and Kennedy Brewer — Mississippi, both convicted on West’s testimony, both exonerated when DNA identified the actual killer. Robert Lee Stinson — Wisconsin, twenty-three years lost. Robert DuBoise — Florida, thirty-seven years lost 12.
Michael West, the Mississippi forensic odontologist whose testimony anchored several of those convictions, was eventually filmed pressing a defendant’s dental mold into a victim’s body — footage that surfaced in the 2020 Netflix documentary The Innocence Files. He resigned from the International Association for Identification under threat of suspension and from the American Academy of Forensic Sciences. By then, he had already helped imprison men for a combined seventy-five years.
Sheila Denton’s 2006 conviction was secured the same year Eddie Lee Howard was sitting on Mississippi’s death row on the same kind of testimony. The science was already dead. The prosecutions were not.
“An Article of Faith”: Shaken Baby Syndrome and the Smith Decision
Shaken Baby Syndrome was a hypothesis published by a British pediatric neurosurgeon in 1971. By the 2000s, it had hardened into a diagnostic doctrine: a child presenting with the so-called triad — subdural hematoma, retinal hemorrhage, and brain swelling — was presumed to have been violently shaken. By 2001, the American Academy of Pediatrics instructed clinicians to presume abuse whenever the triad appeared. No controlled biomechanical or epidemiological study ever validated the inference. The neurosurgeon who first proposed the hypothesis, Norman Guthkelch, spent the last years of his life working with the Arizona Justice Project on innocence cases — what he described as doing “what I can so long as I have a breath to correct a grossly unjust situation.”
A growing body of biomechanical, neuropathological, and neuroradiological research has identified short falls, rebleeds of birth-related subdurals, infections, hypoxia, seizures, vitamin deficiencies, metabolic disorders, and venous thrombosis as conditions that can mimic the triad. A 2016 systematic review by Sweden’s national assessment body concluded that the scientific evidence supporting shaking as the sole cause of the triad was of “very low quality.”
Robert Roberson, an autistic Texas man, has been on death row since 2003 for the death of his two-year-old daughter Nikki — a death his attorneys argue was caused by severe pneumonia (a 104.5°F fever two days earlier), a bedside fall, and a prescription for Phenergan now carrying an FDA black-box warning against pediatric use. Police read his autistic flat affect as guilt. The lead detective, Brian Wharton, has since reversed his position publicly. In October 2024, Roberson was hours from becoming the first person in U.S. history executed for a conviction tied to Shaken Baby Syndrome before the Texas Supreme Court issued an emergency stay following a bipartisan legislative subpoena 13. A second execution date was set, and stayed again, in October 2025 14.
The federal judge who reviewed the Shaken Baby conviction of Illinois daycare worker Jennifer Del Prete in 2014 wrote that aspects of the diagnosis amounted to “an article of faith” — “abundant doubt, not merely reasonable doubt.” That phrase, “article of faith,” is the most accurate description of how the doctrine functioned for decades inside American courtrooms.
The Innocence Network estimates roughly 2,500 Shaken Baby prosecutions a year in the United States, with at least forty parents and caregivers across twenty states formally exonerated. The structural problem is that pediatric child-abuse specialists — a credentialing track established in 2009 — testify in nearly every case, often with strong institutional incentives that align with prosecution narratives.
In October 2025, the Georgia Supreme Court issued a quiet but significant opinion in Smith v. State. The petitioner had been convicted of felony murder in 2003 based on Shaken Baby testimony. He filed an extraordinary motion for new trial. The trial court denied it. The Supreme Court vacated that denial and sent the case back, emphasizing that the discovery of new evidence “that materially affects the question of the defendant’s guilt or innocence” remains a proper subject of an extraordinary motion. The Court also flagged something that is rarely said out loud in Georgia opinions: it noted concerns that the trial court had “uncritically adopted drafts prepared by the State” and instructed the lower court “to proceed with care” 15.
It is the closest thing Georgia has produced to a usable framework for changed-science claims. It is also a long way from a remedy.
Cotton Candy and Bird Droppings: Georgia’s Field-Test Crisis
In December 2016, sheriff’s deputies in Monroe County, Georgia, pulled over a car carrying Dasha Fincher. They found a bag of what looked like blue cotton candy. They dropped a sample into a $2 NIK roadside drug-test pouch, watched the chemicals turn a color the field guide associated with methamphetamine, and arrested Fincher. A magistrate set bond at $1 million. She sat in the Monroe County jail for three months — losing her job, her housing, and time with her family — before a Georgia Bureau of Investigation lab confirmed what she had said all along. The substance was cotton candy.
Dasha Fincher’s case is not unusual. It is canonical. The roadside drug test — typically a NIK Public Safety pouch manufactured by the Safariland Group, or a competing kit from Sirchie or Lynn Peavey — is a chemical color-change reaction interpreted by an officer at the side of a highway against a paper key. Substances that produce false positives include chocolate, sage, oregano, basil, soap, drywall compound, BC headache powder, air fresheners, vitamin powder, Krispy Kreme donut glaze (a 2017 Florida arrest), and bird droppings (a 2020 Georgia college football player’s arrest). A 2017 court filing in Imperial County, California, revealed that Safariland keeps an internal list of more than fifty legal substances known to produce false positives 16.
In 2018, Fox 5 Atlanta’s I-Team obtained every negative drug-test report the GBI Crime Lab issued in 2017 and confirmed at least 145 false positives in a single year — eleven for heroin, twenty-four for ecstasy, forty for cocaine, sixty-four for methamphetamine. In every one of those cases, the original charges were ultimately dropped after lab confirmation. By then the defendants had already lost jobs, housing, custody, and weeks or months of their liberty 17.
Manufacturers acknowledge in fine print that field-test results are “presumptive” and require lab confirmation. Across most U.S. jurisdictions, that fine print is observed enough that the test results stay out of the trial itself — they generate probable cause and indictment and pretrial detention and plea pressure, but they do not reach the jury.
Georgia is the outlier. It is the only U.S. state in which unconfirmed field drug test results remain admissible at trial under standard practice 18. Florida Department of Law Enforcement data have shown that 21 percent of police-identified methamphetamine evidence was not actually methamphetamine, and that half of those false positives contained no controlled substance at all. Houston prosecutors have moved to vacate the convictions of at least 212 defendants who pleaded guilty between 2004 and 2015 on field-test results the Houston Forensic Science Center later determined were not controlled substances. More than half pleaded within a week of arrest.
In Georgia, those convictions are not being vacated. They are being secured.
“Erroneous in 96 Percent of Cases”: Hair, Firearms, and the FBI’s Reckoning
On April 20, 2015, the FBI and the Department of Justice issued a joint statement that should have been a national emergency.
The Bureau and the Innocence Project, working with the National Association of Criminal Defense Lawyers, had reviewed FBI hair-microscopy testimony in capital and other serious cases. Of the first 268 trial cases reviewed in which examiners gave inculpatory testimony, erroneous statements appeared in 257 — a 96 percent error rate. Twenty-six of the twenty-eight FBI hair examiners involved had given erroneous testimony. At least thirty-five of the affected defendants had been sentenced to death. Errors appeared in thirty-three of those thirty-five capital cases. Nine of the death-sentenced defendants had already been executed. Five had died of other causes on death row 19.
The capital states implicated included Texas, Florida, Tennessee, Pennsylvania, Ohio, Indiana, Missouri, Oklahoma, Arizona, and California. The FBI did not retract microscopic hair comparison as a discipline. It conceded only that examiners had “overstated” their findings 20. Most of the roughly 500 to 1,000 state-level examiners trained at the FBI’s Quantico facility were never audited at all. As of 2026, the bulk of state-level hair-microscopy testimony in the United States — including in Georgia — remains unreviewed.
Firearms identification has its own version of the same story. Patrick Pursley spent nearly twenty-four years in Illinois prisons after an examiner testified that bullets and cartridge cases at a 1993 Rockford murder scene matched his Taurus 9mm “to the exclusion of all other firearms.” There was no eyewitness, no DNA, no fingerprint. Two retired examiners, working pro bono after Pursley’s pro se litigation finally cracked open the evidence, independently concluded that none of the bullets or casings came from his gun 21. The 2016 PCAST report concluded that firearms-and-toolmark identification fell short of the criteria for “foundational validity” because only one appropriately designed black-box study existed at the time. The Department of Justice declined to adopt PCAST’s recommendations. The National District Attorneys Association called the findings “scientifically irresponsible.”
Bloodstain pattern analysis put Indiana state trooper David Camm on trial three times for the murder of his wife and two children — eight blood specks on his t-shirt that a prosecution expert called “high-velocity backspatter” turned out to be transfer stains from cradling his dying daughter 22. DNA on the victims’ clothing matched another man. Camm was acquitted on his third trial. Julie Rea was convicted of stabbing her ten-year-old son to death in Illinois on bloodstain testimony from the same expert; she was acquitted at retrial. A 2021 study funded by the National Institute of Justice found alarming inter-examiner disagreement in bloodstain analyses of identical samples.
Even fingerprints, the closest thing forensic science has to a scientifically grounded discipline, fail at the rates examiners testify against. Three FBI examiners, plus a court-appointed defense expert, all matched a print on a bag of detonators from the 2004 Madrid commuter-train bombing to Oregon attorney Brandon Mayfield “100 percent.” The Spanish National Police told the FBI in writing that the print did not match. The FBI continued its investigation until Spanish authorities identified the actual source — an Algerian national. The Department of Justice Office of the Inspector General later concluded that confirmation bias and circular verification had compounded the error, and that “if Mayfield had not been Muslim and had not represented a convicted terrorist in a child-custody case, the laboratory might have revisited the identification with more skepticism” 23.
“Subjective and Unscientific”: Why It All Persists
In 2009, the National Research Council released Strengthening Forensic Science in the United States: A Path Forward — the most consequential American review of forensic science ever conducted. Its core findings were brutal. Most pattern-matching disciplines lacked rigorous scientific validation. Error rates were unknown for nearly every discipline outside DNA. Forensic-science training was inadequate. Cognitive bias and “context effects” pervaded examiner work. The bloodstain analyses on which prosecutors had built murder cases were, in the report’s words, “more subjective than scientific.” The Council recommended creating an independent National Institute of Forensic Science. Congress did not act 24.
In 2014, the Department of Justice and NIST established the National Commission on Forensic Science to implement the NAS recommendations. In April 2017, Attorney General Jeff Sessions disbanded it.
The 2016 PCAST report — Forensic Science in Criminal Courts: Ensuring Scientific Validity of Feature-Comparison Methods — examined six pattern-matching disciplines and concluded that only DNA single-source analysis and latent fingerprint comparison met “foundational validity” criteria. Bite marks, complex DNA mixtures interpreted by older statistical methods, firearms identification, footwear, and microscopic hair did not. Then-Attorney General Loretta Lynch declined to adopt the findings. The Trump administration in 2017 effectively withdrew federal engagement with the report’s framework altogether.
The Federal Rules of Evidence ostensibly require trial judges to act as gatekeepers for unreliable expert testimony. In practice, judicial gatekeeping in criminal cases has been, as Brandon Garrett, Jennifer Mnookin, and Erin Murphy have repeatedly documented, almost nonexistent. Trial courts admit forensic testimony as a matter of long judicial precedent. Appellate courts almost universally affirm. A 2023 amendment to Federal Rule 702 may incrementally strengthen gatekeeping, but the criminal courts have been slow to adjust.
Most American forensic labs, meanwhile, are housed inside law enforcement agencies — state police, sheriff’s offices, district attorneys, the FBI itself. The 2009 NAS report flagged this as a structural conflict of interest: the labs whose mission is forensic truth-finding are managerially answerable to the institutions whose mission is prosecution. The structure has not changed.
No Way Out: Georgia’s Post-Conviction Trap
In 2013, Texas became the first state to enact a dedicated “changed science writ.” Codified at Article 11.073 of the Texas Code of Criminal Procedure, the statute permits habeas relief when relevant scientific evidence was unavailable at trial, or when the science has changed since trial in a way that contradicts what the state relied upon. Article 11.073 is what put Steven Mark Chaney on the path to exoneration on bite-mark testimony. It is what produced the 2025 emergency stay in Robert Roberson’s case. California, Connecticut, Wyoming, Michigan, and Nevada have followed with variants.
Georgia has not.
A Georgia defendant whose conviction rests on forensic evidence that has since been scientifically repudiated has, as a practical matter, one vehicle: the extraordinary motion for new trial under O.C.G.A. § 5-5-41. The Georgia Supreme Court has called it “disfavored” — disfavored, the courts say, because it undermines finality. The petitioner must establish that the evidence is genuinely new, that diligence could not have surfaced it sooner, and that it would probably produce a different verdict. Most petitions fail.
Smith v. State in October 2025 was a partial opening. The Georgia Supreme Court reaffirmed that the discovery of new evidence that “materially affects the question of the defendant’s guilt or innocence” is a proper subject of an extraordinary motion, and the Court’s pointed observation about trial courts adopting State-drafted orders signaled at least some appetite for closer review 25. But the framework remains narrow. The burden is on the petitioner — incarcerated, often without counsel, almost always without independent expert resources — to surface the science the state should have surfaced before the conviction was ever entered.
The arithmetic is simple. The 2009 NAS report, the 2015 FBI hair review, the 2016 PCAST report, the 2020 Sheila Denton order, the 2025 Smith decision: every one of those documents is, in part, a record of how many people were already in Georgia prisons when the science underneath their convictions collapsed. There is no statewide review. There is no statewide audit. There is no statewide conviction integrity unit with the authority to act. There is no junk science writ.
A statewide Conviction Integrity Unit with real independence is one of the six post-conviction reforms in GPS’s Vision 2027 campaign — alongside repeal of Georgia’s four-year habeas deadline, retroactive review of constitutional violations, ineffective-assistance reform, the right to legal access, and plea bargain reform. None of those reforms exist as state law today. Until they do, the Georgia post-conviction record will continue to read the way it has read for two decades: a record of cases that the science cannot reopen because the law will not permit it.
The Burning Car, Revisited
Maria Montalvo has now been incarcerated for thirty-two years. The fire science that put her there has been demolished by the profession that produced it. The Lime Street experiment was conducted four years before her car burned. NFPA 921 was published two years before. Cameron Todd Willingham was executed eight years after her trial on the same kind of testimony. Han Tak Lee was exonerated in 2015 on the same kind of testimony. Daniel Dougherty’s death sentence was vacated on the same kind of testimony.
The judge who denied Maria Montalvo’s motion for a new trial wrote that suggesting the fire was an accident would “dishonor the memory of those two children.” The sentence is worth pausing over. It treats doubt — the scientific posture, the constitutional posture, the posture every American court is supposed to take before depriving a citizen of liberty — as a moral betrayal. When the courts treat doubt as betrayal, they have abandoned the only thing that distinguishes them from the mob.
Sheila Denton was freed in Georgia in 2020 because a single Ware County Superior Court judge was willing to write the words “the bite mark evidence used at trial is now known to be unsupported by science.” There is, somewhere in a Georgia prison today, someone like Sheila Denton whose case has not yet found its judge. There is, somewhere, someone like Patrick Pursley, whose firearms-match testimony will not be reexamined until the petitioner himself somehow forces the issue from a cell. There is someone whose 2017 conviction rests on a GBI field-test reading that no laboratory ever confirmed.
The science changed. The legal architecture did not. Closing that gap is the unfinished work of American post-conviction practice. In Georgia, it has barely begun.
Call to Action: What You Can Do
You just read about people convicted on science the FBI itself called wrong 96% of the time, on bite marks courts now say were 'unsupported by science,' on a $2 test that flagged cotton candy as meth. Georgia's courts won't reopen these cases. The least you can do is make sure people know. https://gps.press/burned-by-the-state-junk-forensic-science-and-the-georgia-cases-the-courts-wont-reopen/
Spread the Word — It Takes One Click
Awareness without action changes nothing. Here’s how you can help push for accountability and real reform:
Join the GPS Advocacy Network — Sign up at https://gps.press/become-an-advocate/ and we’ll advocate on your behalf every week. GPS identifies your state legislators, crafts personalized letters on the most pressing prison issues, and sends them directly to the representatives who represent you. You receive a copy of every letter. It takes two minutes to sign up — we handle the rest.
Tell My Story — Are you or a loved one affected by Georgia’s prison system? GPS publishes first-person accounts from incarcerated people and their families. Submit your story at https://gps.press/category/tellmystory/ and help the world understand what’s really happening behind the walls.
Contact Your Representatives — Your state legislators control GDC’s budget, oversight, and the laws that created these failures. Find your Georgia legislators at https://gps.press/find-your-legislator/ or call Governor Kemp at (404) 656-1776 or the GDC Commissioner at (478) 992-5246.
Demand Media Coverage — Contact newsrooms at the AJC, local TV stations, and national criminal justice outlets. More coverage means more pressure.
Amplify on Social Media — Share this article and tag @GovKemp, @GDC_Georgia, and your local representatives. Use #GAPrisons, #PrisonReform, #GeorgiaPrisonerSpeak.
File Public Records Requests — Georgia’s Open Records Act gives every citizen the right to request incident reports, death records, staffing data, medical logs, and financial documents at https://georgiadcor.govqa.us/WEBAPP/_rs/SupportHome.aspx.
Attend Public Meetings — The Georgia Board of Corrections and legislative committees hold public meetings. Your presence is noticed.
Contact the Department of Justice — File civil rights complaints at https://civilrights.justice.gov. Federal oversight has forced abusive systems to change before.
Support Organizations Doing This Work — Donate to or volunteer with Georgia-based prison reform groups fighting for change on the ground. The Georgia Innocence Project and the Southern Center for Human Rights have driven many of the bite mark and forensic-science exonerations described in this article.
Vote — Research candidates’ positions on criminal justice. Primary elections often determine outcomes in Georgia.
Contact GPS — If you have information about a Georgia conviction built on forensic evidence that has since been scientifically repudiated — arson, bite marks, hair, firearms, bloodstain, Shaken Baby, or unconfirmed field drug tests — reach us securely at GPS.press.
Further Reading
Every Door Locked: Innocent People Trapped in Georgia Prisons
The companion piece to this investigation — a closer look at the Georgians whose innocence claims have nowhere to go in a state without a junk-science writ or independent conviction-integrity mechanism.
Blackstone Is Dead: Georgia Abandoned American Justice
“Better that ten guilty persons escape than that one innocent suffer.” How Georgia’s post-conviction architecture inverted the foundational principle of Anglo-American criminal law.
Why a $2 field drug test, a high cash bond, and a few weeks in a county jail produce guilty pleas from people who did nothing wrong — the upstream pipeline that feeds the wrongful convictions described in this article.
The Sleeping Giants: Two Georgia Statutes That Could Unlock Post-Conviction Justice
The first article in the No Way Out series — two laws Georgia already passed that, if enforced, would give the state a real mechanism for reopening cases when the science underneath them collapses.
How to Use GPS Data with AI Tools
A step-by-step guide for researchers, journalists, and advocates to interrogate Georgia’s prison system using GPS’s machine-readable data and modern AI tools.
Research Explainers
GPS Research Explainers distill complex data and legal research into accessible briefings. These explainers are directly relevant to the issues covered in this article:
A $2 Drug Test That Gets It Wrong Up to 38% of the Time — and Why Georgia Must Act
GPS’s Research Library briefing on Colorado’s 2025 ban on arrests based on field drug tests, and what it would take for Georgia — the only state where these tests remain admissible at trial — to follow.
The six post-conviction reforms — including a statewide Conviction Integrity Unit and habeas reform — that would give Georgia a real mechanism for correcting wrongful convictions built on discredited forensic evidence.
Explore the Data
GPS makes GDC statistics accessible to the public through several resources:
- GPS Statistics Portal — Interactive dashboards translating complex GDC reports into accessible formats, updated within days of official releases.
- GPS Lighthouse AI — Ask questions about Georgia’s prison system and get answers drawn from GPS’s investigative archive and data analysis.
- Machine-Readable Pages for Researchers — GPS maintains AI-optimized pages for data analysis:
- AI Content Index — https://gps.press/ai-index/
The AI Content Index has links to numerous machine readable pages, but this is all that is needed by an AI to fully understand all the data. You can learn more about using GPS Data with AI in our article on the topic:
How to Use GPS Data with AI Tools (https://gps.press/how-to-use-gps-data-with-ai-tools/)
A step-by-step guide showing researchers, advocates, families, and journalists how to use GPS’s machine-readable data pages with AI tools like ChatGPT, Claude, and Gemini to analyze Georgia prison conditions, statistics, and policy.
Contact GPS at media@gps.press for access to underlying datasets used in this analysis.
About Georgia Prisoners’ Speak (GPS)
Georgia Prisoners’ Speak (GPS) is a nonprofit investigative newsroom built in partnership with incarcerated reporters, families, advocates, and data analysts. Operating independently from the Georgia Department of Corrections, GPS documents the truth the state refuses to acknowledge: extreme violence, fatal medical neglect, gang-controlled dorms, collapsed staffing, fraudulent reporting practices, and unconstitutional conditions across Georgia’s prisons.
Through confidential reporting channels, secure communication, evidence verification, public-records requests, legislative research, and professional investigative standards, GPS provides the transparency the system lacks. Our mission is to expose abuses, protect incarcerated people, support families, and push Georgia toward meaningful reform based on human rights, evidence, and public accountability.
Every article is part of a larger fight — to end the silence, reveal the truth, and demand justice.
Footnotes
- Deseret News, 1994 contemporaneous reporting on the Long Branch fire, https://www.deseret.com/1994/2/23/19093649/mom-accused-of-killing-2-tots-in-burning-car/ [↩]
- The Appeal, “Maria Montalvo, Arson, and Junk Science in New Jersey,” https://theappeal.org/maria-montalvo-arson-junk-science-new-jersey/ [↩]
- The Appeal, https://theappeal.org/maria-montalvo-arson-junk-science-new-jersey/ [↩]
- Wikipedia, “Lime Street fire,” compiling primary documents, https://en.wikipedia.org/wiki/Lime_Street_fire [↩]
- ABC News, “Fire Expert: How I Nearly Sent an Innocent Man to the Electric Chair,” https://abcnews.go.com/2020/john-lentinis-fire-arson-investigation/story?id=10562869 [↩]
- Texas Forensic Science Commission, Final Report on the Willingham/Willis Investigation, https://www.txcourts.gov/media/1454542/fr-willingham-mka-new-version-03162022.pdf [↩]
- NBC News, “After 24 Years of Wrongful Imprisonment, Han Tak Lee Finally Has Some ‘Fresh Air,'” https://www.nbcnews.com/news/asian-america/after-24-years-wrongful-imprisonment-han-tak-lee-finally-has-n537856 [↩]
- Innocence Project, “Sheila Denton is Freed After 15 Years of Wrongful Imprisonment,” https://innocenceproject.org/sheila-denton-is-freed-after-15-years-of-wrongful-imprisonment/ [↩]
- Forensic Resources, “State of Georgia v. Sheila Denton — Order for New Trial, February 7, 2020,” https://forensicresources.org/resources/state-of-georgia-v-sheila-denton-order-for-new-trial/ [↩]
- Atlanta Journal-Constitution, “Georgia court: Flawed bite mark evidence wrongly convicted woman,” https://www.ajc.com/news/local/judge-flawed-bite-mark-evidence-used-wrongful-conviction/bFrSsteVPOXwumxFXnu5dL/ [↩]
- PCAST, “Forensic Science in Criminal Courts: Ensuring Scientific Validity of Feature-Comparison Methods,” https://obamawhitehouse.archives.gov/sites/default/files/microsites/ostp/PCAST/pcast_forensic_science_report_final.pdf [↩]
- Innocence Project, “Description of Bite Mark Exonerations,” October 2023, https://innocenceproject.org/wp-content/uploads/2023/10/DESCRIPTION-OF-BITE-MARK-EXONERATIONS-Updated-10.18.23.pdf [↩]
- Innocence Project, “What to Know About Robert Roberson,” https://innocenceproject.org/news/what-to-know-about-robert-roberson-on-texas-death-row-for-a-crime-that-never-occurred/ [↩]
- PBS NewsHour, “Texas appeals court again pauses execution of Robert Roberson in shaken baby case,” https://www.pbs.org/newshour/politics/texas-appeals-court-again-pauses-execution-of-robert-roberson-in-shaken-baby-case [↩]
- Justia, “Smith v. State,” S25A0548, Supreme Court of Georgia, October 2025, https://law.justia.com/cases/georgia/supreme-court/2025/s25a0548.html [↩]
- ProPublica / New York Times Magazine, “Busted,” https://www.propublica.org/article/common-roadside-drug-test-routinely-produces-false-positives [↩]
- Fox 5 Atlanta I-Team, “Look how often field drug tests send innocent Georgians to jail,” https://www.fox5atlanta.com/news/look-how-often-field-drug-tests-send-innocent-georgians-to-jail [↩]
- Criminal Legal News, “Roadside Drug Tests: Failed Technology From the Failed War on Drugs,” https://www.criminallegalnews.org/news/2023/aug/1/roadside-drug-tests-failed-technology-failed-war-drugs/ [↩]
- FBI Press Release, “FBI Testimony on Microscopic Hair Analysis Contained Errors in at Least 90 Percent of Cases,” https://www.fbi.gov/news/press-releases/fbi-testimony-on-microscopic-hair-analysis-contained-errors-in-at-least-90-percent-of-cases-in-ongoing-review [↩]
- Washington Post, “FBI overstated forensic hair matches in nearly all criminal trials for decades,” https://www.washingtonpost.com/local/crime/fbi-overstated-forensic-hair-matches-in-nearly-all-criminal-trials-for-decades/2015/04/18/39c8d8c6-e519-11e4-b510-962fcfabc310_story.html [↩]
- Northwestern Pritzker Law, Center on Wrongful Convictions, “Patrick Pursley,” https://cwc.law.northwestern.edu/freed-exonerated/patrick-pursley/ [↩]
- ProPublica, “Bloodstain Analysis Convinced a Jury She Stabbed Her 10-Year-Old Son,” https://www.propublica.org/article/bloodstain-pattern-analysis-jury-wrongful-conviction-acquitted-exonerated [↩]
- DOJ Office of the Inspector General, “A Review of the FBI’s Handling of the Brandon Mayfield Case,” https://oig.justice.gov/sites/default/files/legacy/special/s0601/exec.pdf [↩]
- National Academy of Sciences, “Strengthening Forensic Science in the United States: A Path Forward,” 2009, https://www.ojp.gov/pdffiles1/nij/grants/228091.pdf [↩]
- FindLaw, “Smith v. State, S25A0548,” https://caselaw.findlaw.com/court/ga-supreme-court/117826676.html [↩]