chevron-down Created with Sketch Beta.

Litigation Journal

Winter 2021 | Proof

Federal Habeas Corpus Review Is Broken

Jonathan Potts, Charles A Weiss, and Stephen Robert Snodgrass


  • In any rational system of justice, innocence must be paramount.
  • In federal habeas corpus, however, the justice system has become an end in itself.
  • The system cannot worry itself with offending state judges who were wrong. 
  • The focus needs to return to the central purpose of that system: protecting the innocent.
Federal Habeas Corpus Review Is Broken
gorodenkoff via Getty Images

Jump to:

The instinct to correct wrongful convictions is part of our national conscience. Books like To Kill a Mockingbird have molded many Americans’ basic notions of right and wrong. Recently, popular media like the podcast Serial and the docuseries Making a Murderer have recaptured national attention and provoked public debate about the complexities—and grayer areas—of innocence.

Exoneration is a topic that would seem likely to focus on the facts more than on the law. Few non-lawyers would balk at the notion that a duly convicted prisoner who can prove his innocence should be freed at a moment’s notice. Yet the current system of federal habeas corpus review makes it exceptionally difficult—and in some cases literally impossible—for innocent people to regain their freedom.

Over the past 12 years, our pro bono practice has exonerated four men wrongfully convicted of murder: Josh Kezer (2009), George Allen Jr. (2013), and David Robinson (2018), and Donald Nash (2020).

Each of these men was innocent in the purest sense of the word: He simply didn’t do it. But none of these exonerations came with the cooperation of prosecutors after their review of the evidence. None came through an act of executive grace. Each exoneration was its own long and expensive battle against unnecessary legal obstacles. Throughout this piece, we’ll draw on Donald Nash’s and David Robinson’s cases to illustrate the shortcomings of the federal habeas system. Both cases are recent, both men are obviously innocent, and both men spent many years in prison unsuccessfully litigating in federal court.

The authors are not bleeding hearts. Outside of these cases, we are primarily defense-side business litigators (which is not to say we’re heartless either). The observations in this piece are informed by our background as civil litigators who take on innocence cases for free, and not criminal defense attorneys, public servants, or even nonprofit attorneys who spend their careers living and breathing this system. We took on these cases simply because each one of these exonerees was able to persuade us of his innocence, despite knowing we could toil for years with uncertain odds of success.

Habeas corpus is civil, although it “feels” criminal. But within the realm of civil litigation, habeas corpus plays by its own rules. Those rules do not feel civil or criminal. In “regular” civil cases, courts consider allegations with an open mind and allow the parties to conduct discovery and present evidence. In habeas cases, however, a hidden switch is flicked. Success is not simply about carrying a high burden of proof. Rather, in most cases, allegations of innocence, and even evidence of innocence, receive a systematically frosty reception of distrust. This sense of distrust pervades almost every case, even when the allegations involve new evidence that no one—the prisoner, the prosecution, the judge, or the jury—ever knew about. The result is that, during federal habeas review, evidence is handicapped, discovery is rare, and hearings are rarer.

True, some (or many) prisoners lie. But so do some civil plaintiffs suing for damages. And so do some witnesses called by those plaintiffs. In our daily litigation practices, we deal with dishonest plaintiffs and witnesses by using the standard weapons in a litigator’s arsenal. Sometimes ferreting out the truth takes longer than we or our clients would prefer, but we can trust that the system is well equipped for competent advocates to close the gap between lies and reality. In federal habeas cases, however, we find that even when we develop new, reliable evidence before entering a courtroom, that evidence may be greeted with automatic suspicion.

In our experience, the primary battleground during federal habeas review is not litigating the central question of innocence. The question of innocence is so well barricaded with procedural obstacles that innocence can resemble a technicality.

Faithfulness to the Constitution is not imperative during federal habeas review, and the legal standards for relief are dizzying. Congress and the Supreme Court have cloaked judicial review of wrongful convictions in layers of abstraction. These legal standards are totally unfamiliar to most lawyers and nonsensical to the rest of society. Instead of determining whether a constitutional violation did occur, federal courts must ponder whether an erroneous state court decision was so wrong that it was “objectively unreasonable.” “Wrong” isn’t enough.

Instead of doing what judges do, the federal judiciary must let slide obvious constitutional violations unless they exceed a vague level of “wrongness.” This process is in equal parts slow and ineffective.

Habeas Corpus—The Basics

Habeas corpus, or the “Great Writ,” is meant to be one of the fundamental checks on governmental power. Roughly speaking, the writ traditionally ensured that government officials could not detain or incarcerate individuals in the absence of proper charges or a conviction. Beginning in the 1950s, however, the Supreme Court introduced the concept that habeas corpus could be used to examine the constitutionality of state criminal trials.

The current habeas regime dates back to the passage of the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA). AEDPA was a disaster from inception. Under the pretense of federal-state comity, AEDPA restricts the availability of federal habeas corpus relief to awarding relief based on state court decisions that are contrary to, or involve unreasonable applications of, clearly established federal law. The Supreme Court has interpreted these terms to protect criminal defendants from only “extreme malfunctions” of the state criminal justice system. Functionally, the standards (discussed below) largely render federal habeas review meaningless. AEDPA allows the mechanic to check the engine only when the hood is on fire.

Before presenting claims in federal court, the applicant must exhaust all state court remedies through any available statutory procedure, including direct appeal and statutory post-conviction remedies. The stated goal is to channel all constitutional claims into the appropriate state courts and grant those courts the opportunity to correct the errors for themselves. Practically, this requirement weeds out many meritorious claims without the need for federal assistance. But if an innocent person fails to identify and raise meritorious claims in state court, it generally means those claims will never be adjudicated, no matter how consequential the error. These claims are “procedurally defaulted.” A procedurally defaulted claim generally can never serve as a basis for federal habeas relief.

Innocence Is a Technicality

One of the few ways to circumvent a procedural default is a so-called “gateway” claim of actual innocence. The standard is high. The standard is also confusing: To establish a gateway claim of actual innocence, the habeas applicant must show that, based on all the evidence, old and new, exculpatory and inculpatory, without regard to rules of admissibility that may govern at trial, it is more likely than not that no reasonable juror would currently find the prisoner guilty beyond a reasonable doubt. Only a few dozen prisoners have satisfied this standard over the past few decades in an adversarial federal court proceeding.

Even if the prisoner satisfies this high standard, a “gateway” claim of innocence is not even a basis for overturning a wrongful conviction. Rather, upon this showing, the federal court will then only consider—consider—a procedurally defaulted claim. As a result, “actual innocence” cases form only a tiny percentage of the total exonerations in the United States.

In addition, some federal jurisdictions will not even consider all newly discovered evidence of innocence. Rather, these jurisdictions will disregard newly discovered evidence that could have been discovered previously through the exercise of due diligence. Consider irrefutable evidence like a newly discovered videotape of someone else committing the murder. If it fails the due diligence test, the videotape is not “new” evidence, and any meritorious procedurally defaulted claims will be summarily denied.

In this vein, the U.S. justice system’s dirty little secret is that demonstrating innocence is not a universally recognized legal basis for overturning a wrongful conviction. As Justice Scalia wrote in 2009, “[t]his Court has never held that the Constitution forbids the execution of a convicted defendant who has had a full and fair trial but is later able to convince a habeas court that he is ‘actually’ innocent.” That principle remains true today. A so-called “freestanding” claim of actual innocence is recognized by certain states, but no federal circuit court has ever upheld a writ of habeas corpus based on factual innocence alone.

I Have to Prove—What?

Putting innocence aside, consider claims that were not “procedurally defaulted.” In other words, the innocent person properly preserved the claims on direct appeal or during post-conviction review. AEDPA does allow federal courts to grant relief from these state court judgments if the decisions were either “contrary to” or “an unreasonable application of” clearly established federal law. At first glance, these standards might not seem that troubling. As interpreted by the Supreme Court, however, these standards render AEDPA mostly toothless.

In terms of court decisions, the phrase “clearly established Federal law” refers only to Supreme Court precedent. Immediately this reduces the possible grounds for relief because the volume of constitutional cases resolved by the circuit courts dwarfs that of the Supreme Court. But under AEDPA, if the Supreme Court has not spoken, no one has. Even decades-old circuit precedent is insufficient. Also, consider that the Supreme Court may not grant certiorari to address a legal question precisely because there is a consensus in the circuit courts. Even a consensus among the circuits is not “clearly established Federal law.”

The Supreme Court’s precedent is also frozen in time. Prisoners may rely only on case law that existed at the time of the state court’s decision upholding their convictions. If the Supreme Court later corrects course—even before the federal habeas case is over—no relief is available.

This is bad enough, but AEDPA’s constitutional standards are even worse. The “contrary to” standard does not mean what you might think. Basically, a state court must either (a) recite a completely wrong constitutional principle or (b) reach the opposite result from the Supreme Court when confronted with a set of either identical or materially indistinguishable facts. If a case is merely similar to a Supreme Court case, the state court’s decision is not “contrary to” that precedent. If the precise contours of a constitutional principle remain unclear, state courts enjoy “broad discretion” in their adjudication of a prisoner’s claims.

Let’s pause for a moment here. There is quite a bit to unpack. The Supreme Court’s docket is already small, but the annual number of criminal cases involving constitutional rights is in the single digits. Most of these cases are not foundational.

But what happens when the wrongfully convicted defendant’s arguments rely on critically important, but broad, foundational cases? Consider Brady v. Maryland, which requires the state to disclose material exculpatory information, or Jackson v. Virginia, which establishes the standard for constitutionally insufficient evidence of guilt. But to be “contrary to” Brady or Jackson under AEDPA, an innocent person needs identical or materially indistinguishable facts.

Does anyone even remember the facts in Brady? We’ll tell you: Brady’s co-defendant confessed to the murder, and the state suppressed the confession. Sure, this might happen in another case, but police and prosecutors can withhold countless other forms of exculpatory evidence, from eyewitness statements to fingerprints to DNA testing. Are these critical pieces of exculpatory evidence “materially indistinguishable” from a murder confession?

And in Jackson, the U.S. Supreme Court found that there was constitutionally sufficient evidence that the defendant committed the murder. Even if you had materially indistinguishable facts (which is essentially impossible in a totality-of-the-evidence review), reaching the “opposite” result from Jackson would mean that the wrongfully convicted prisoner was acquitted or his conviction was overturned.

Because the “contrary to” standard is usually unattainable, habeas cases almost always turn on whether the state court “unreasonably applied” clearly established federal law. The Supreme Court has made it quite clear that an “unreasonable” application of federal law is very different from an “incorrect” application of federal law. Even clear error is insufficient. The result must be “objectively unreasonable,” not merely wrong. In Harrington v. Richter, the Supreme Court crystallized its interpretation of this standard by announcing that habeas petitioners must show that the state court’s ruling “was so lacking in justification that there was an error well understood and comprehended in existing law beyond any possibility for fairminded disagreement.”

Thus, in asking whether a state court “unreasonably applied” the Constitution, federal judges do not inquire whether a constitutional violation resulted in a wrongful conviction or even whether a state court properly applied the Constitution. Instead, federal judges must imagine “fairminded jurists” and ask themselves whether those fairminded jurists could disagree about the correctness of the state court’s ruling. If there is room for disagreement, an innocent person loses, even if a constitutional violation occurred. Try explaining to an innocent client that, although his constitutional rights were violated, he must die in prison because imaginary “fair” judges—not real ones—could have an imaginary debate in which one of those imaginary judges reaches the wrong answer and everyone else knows it. If it sounds like this doesn’t make sense, it’s because it doesn’t.

Case Study—Donald Nash

This brings us to Donald Nash’s case. Nash’s girlfriend was murdered in 1982 at an abandoned schoolhouse in rural Missouri. The killer brutally strangled the victim with her own shoelace and then shot her with a shotgun. There were no eyewitnesses and no physical evidence that connected Nash to the crime. He did not even own a shotgun. He also tested negative for gunshot residue just hours after the murder. Fresh tire tracks were found at the scene of the crime; they did not belong to Nash’s truck.

For 26 years, Nash was never charged with his girlfriend’s murder. During that time, the two main suspects were men whose fingerprints were located on the victim’s stranded vehicle. The first man was a violent sex offender. The second man lived near the victim’s vehicle, denied ever seeing the vehicle, and then abandoned his family a few weeks after the murder.

In 2008, however, the Missouri Highway Patrol analyzed the victim’s fingernails and found a trace amount of Nash’s DNA. Infected by tunnel vision, the investigators arrested Nash immediately. In reality, it’s common for a romantic couple—especially a couple who live together—to share each other’s DNA underneath their fingernails. The tiny amount of Nash’s DNA was consistent with that type of relationship.

Nevertheless, the prosecution found a crime lab technician willing to testify that the victim’s act of hair washing the evening before her murder would have had a “great effect” on eliminating Nash’s preexisting DNA. This was junk science. The “expert” was unqualified. The scientific literature does not exist. You will not find a single scientific article in the world that offers this opinion. In the absence of real evidence, however, the prosecution stretched this ridiculous inference from hair washing to construct a fantastical narrative about running the victim’s car off the road, abducting her, strangling her, and shooting her with a nonexistent weapon, and cunningly staging the scene to make the crime “look” like it was sexually motivated.

By contrast, Nash’s defense sought to present evidence that the convicted violent sex offender who left his fingerprints on the dead woman’s abandoned car might have committed the murder and not Nash. The suspect had a lengthy rap sheet and had been fired from his job as a gas station attendant for drilling a peephole in the women’s restroom. This suspect was known to keep a shotgun in his trunk. The suspect denied ever visiting the town where the murder occurred, and the victim’s checkbook showed that she had filled up her gas multiple times in the past month but never at his gas station, which was in a different city. The suspect later committed suicide by shotgun. A joint task force had conducted a cold case review and listed the sex offender as the primary suspect.

The trial court excluded the evidence. The court ruled that a violent sex offender’s fingerprints on a dead woman’s abandoned car lacked a “direct connection” to the corpus delicti, despite the fact that the prosecution argued that she had been abducted from that car.

Nash, who had no criminal record, was convicted of premeditated murder. He was sentenced, at age 66, to life without the possibility of parole for 50 years. The Supreme Court of Missouri upheld the sufficiency of the evidence against Nash under Jackson and upheld the exclusion of the alternative perpetrator evidence against a challenge that the “direct connection” rule violated Nash’s right to a complete defense.

We took over the case after Nash lost on direct appeal and sought federal habeas review, not wanting to miss the precious one-year filing window under AEDPA. During federal habeas review, two new developments occurred. First, we engaged a highly experienced forensic scientist who was professionally horrified by the junk science used to convict Nash of murder. Second, we conducted additional DNA testing. We tested the shoe whose shoelace was removed. We reasoned that the killer must have handled the shoe during the strangulation and later disposed of the evidence by throwing it across a fence line. The shoe had not previously been tested by the state. We found male DNA. That DNA did not belong to Nash.

Triumphantly, we told the federal habeas court. We asserted actual innocence claims on top of our claim of constitutionally insufficient evidence under Jackson and the claim that the exclusion of the sex offender evidence under the “direct connection” rule violated Nash’s right to a complete defense.

There was no discovery and no evidentiary hearing. The district court issued an opinion that, under AEDPA, the two uncertain words of the expert’s opinion—“great effect”—amounted to constitutionally sufficient evidence that Nash committed premeditated murder because the state court ruling was not “objectively unreasonable.” The court also held that the claim about the exclusion of the sex offender evidence was procedurally defaulted because Nash’s original counsel had framed the argument to the Supreme Court of Missouri as a facial challenge instead of an as-applied challenge. The court held that the new DNA testing was not “newly discovered evidence” to overcome procedural default because the DNA on the victim’s shoe could have previously been discovered through the exercise of due diligence. The district court finally denied Nash’s freestanding claim of actual innocence because the claim was not recognized by the Supreme Court.

Despite these harsh rulings, the district court wrote: “Although the undersigned finds his hands tied . . . and cannot offer him any relief, the Court hopes that the State of Missouri may provide a forum, either judicial or executive, in which to consider the evidence that Petitioner may be actually innocent of the crime for which he was convicted.” On appeal, the Eighth Circuit echoed the district court, stating that “the newly presented evidence in this case deserves ‘serious consideration’ in the state courts” and that “state court would be a more appropriate forum for Nash’s claims.” In short, the federal courts are not equipped to handle federal constitutional problems during habeas review.

When we returned to state court and finally had the opportunity to conduct discovery, we deposed the prosecution’s original expert. It did not take much prodding. She retracted the “great effect” opinion on the sixth page of her deposition transcript. The Supreme Court of Missouri also appointed an excellent special master, the Hon. Richard Zerr. Nash received a full evidentiary hearing with a dozen witnesses. Judge Zerr then issued a 226-page report recommending Nash’s exoneration.

In Nash’s state court habeas proceedings, it did not matter whether the DNA testing could have been performed earlier. What mattered was the testing result. We thus were finally allowed to pursue and prove Nash’s claim of actual innocence. We also established a constitutional violation based on the exclusion of the sex offender evidence, which the federal courts had refused to consider. After 12 years behind bars, Nash went home on July 4, 2020—Independence Day. Five of those years had been lost under AEDPA.

Case Study—David Robinson

David Robinson spent 18 years in prison under a life-without-parole sentence for a murder he did not commit. There was no physical evidence implicating him, and he had numerous, credible alibi witnesses. Robinson’s conviction rested solely on the perjured testimony of two jailhouse snitches.

The first snitch was a crack user and paid police informant, Albert Baker. After Baker provided a statement against Robinson, the detective then arranged for Baker’s release from jail on a personal recognizance bond. Baker also went home with a payment of $50. The detective continued to deliver cash payments of $12, $25, $50, $10, $55, $70, $20, $10, $100, and $50, respectively, to Baker at the home where he was staying with another crack user. Later, under the guise of a “witness protection” program, Baker received his own apartment through trial, paid for by the state.

The second jailhouse snitch was a paranoid schizophrenic who literally “heard voices.” He had just returned from a stint in a psychiatric ward after a suicide attempt. The schizophrenic snitch told the same detective that Robinson had confessed in their jail cell. The only problem was that jailhouse records established that the two men had never shared a jail cell. Still, the schizophrenic testified at trial for the prosecution. The prosecutor who tried the case later admitted that she knew they had never been in the same cell.

Robinson was not involved with the murder or anywhere near the crime scene. The murder was committed by Romanze Mosby, who confessed to the murder on audiotape and to numerous other people after Robinson’s conviction. Mosby’s cousin confirmed that he witnessed Mosby commit the murder when Mosby panicked during a drug transaction.

Mosby’s repeated confessions were not bragging; he was concerned. When our firm took on the Robinson case, Mosby, who was in prison, read a newspaper article about it. That day, Mosby confessed to his stepfather and stated he did not want to spend the rest of his life in prison. The two men took a prison photograph together, and Mosby asked his stepfather to take care of Mosby’s mother. That night, Mosby hanged himself in his cell.

None of this was sufficient for the state to cooperate in Robinson’s exoneration. It took another nine years after Mosby’s suicide to free Robinson.

Baker, the crack addict, had first recanted in 2004. Baker later testified, however, that shortly before he was scheduled to testify at Robinson’s state post-conviction hearing, a police officer had driven up next to Baker while he was walking on the street. By that time, Baker was living in a city over 100 miles away. The officer handed Baker a cell phone with the original detective on the other end, who told him, “We’re keeping an eye on you.” Baker stuck to his original story when he testified at the post-conviction hearing, and Robinson spent another 14 years in prison.

In 2015, we deposed Baker. Before we could ask Baker a single question, here is what the state had to say:

Mr. Baker, I am an assistant prosecuting attorney of the attorney general of the state, and I’d like to make you aware obviously you have concerns about perjury. I’d like to make you aware that for statements made during the course of a murder trial perjury is a class A felony. There’s no statute of limitations. We can bring that charge at any time. The exposure that you’re looking at is 10-year mandatory minimum to 30 years to life. Perjury is charged per false statement, every time you’ve made a false statement under oath, and I expect that if these two gentlemen would like to elicit some sort of confession of perjury or otherwise get out that you may have made some false statements during the trial of David Robinson, the state intends to prosecute you to the fullest extent of the law.

Notwithstanding the state’s threat, Baker still confessed to perjury. Yet, it took another three years to exonerate Robinson, including yet another loss in trial court. Finally, the Supreme Court of Missouri appointed another special master, who wrote a scathing 95-page report to exonerate Robinson, finding that not only was Robinson innocent but Mosby was guilty.

The startling thing is that most of the evidence used to exonerate Robinson had been available for years. The federal habeas court, however, had viewed Baker’s original recantation with suspicion. The court said that Baker’s “history as a drug abuser and criminal leaves his new, differing account of the events tainted with unreliability.” But Baker was also a drug abuser and criminal when he testified on behalf of the prosecution; why was he a liar this time? Perhaps a better solution than blaming prisoners for the lack of reliability of recantations for the witnesses they’re stuck with is for prosecutors to stop using this type of witness to secure convictions in the first place.

Despite the fact that Robinson’s habeas application sat in federal court for four years, the district court never held a hearing to ask Baker about his flip-flopping. When Baker did testify before a court in 2018, he not only admitted to perjury (again), but apologized to Robinson and asked for forgiveness. Robinson, who weighs over 300 pounds and is built like Paul Bunyan, started crying. Baker also revealed the police intimidation that led him to un-recant—testimony that the detective did not deny and that the special master described as “bone-chilling.” But the federal court had never asked.

The federal habeas court also concluded, without taking evidence, that Mosby’s confession was unreliable because Mosby had originally tried to blame his cousin for the murder. Normally, a court might find it strange that two cousins with no connection to the charged defendant had been simultaneously accusing each other—and not Robinson—of committing the murder. Again, the federal habeas court never even held a hearing to ask Mosby, or his cousin, what was going on. Robinson’s actual innocence claim was denied without discovery or an evidentiary hearing, and the federal habeas court never reached his procedurally defaulted claims.

Many of Robinson’s federal claims were procedurally defaulted. One of his non-defaulted claims was the prosecution’s presentation of the schizophrenic’s testimony despite the knowledge that he had never shared a cell with Robinson. In his 2018 report, the special master concluded that the prosecution had knowingly presented perjured testimony to the jury by allowing the schizophrenic to testify about a confession that could not have possibly occurred.

In 2010, however, the federal habeas court had denied relief without ever mentioning the schizophrenic’s false testimony about being in the same cell. Instead, the district court focused on other allegations that the schizophrenic’s testimony was coerced and that officers provided the schizophrenic with details about the murder and pressured him to testify against Robinson. Remarkably, the district court held that the “prosecutor could not have committed a Brady violation in failing to disclose at trial that Mr. Richison was abused and his testimony was false when Mr. Richison’s allegations of abuse were first made known almost three years after the trial.” This remarkable statement ignored Robinson’s point—which is that police had coerced that false testimony. Ultimately, the district court not only denied habeas relief but also denied Robinson a certificate of appealability, stating that “[t]he undersigned is not persuaded that the issues raised in his petition are debatable among reasonable jurists, that a court could resolve the issues differently, or that the issues deserve further proceedings.”

Fortunately for Robinson, an excellent special master, the Hon. Darrell Missey, was appointed by the Supreme Court of Missouri, but it still took another eight years after the federal court’s denial of habeas relief to free Robinson.

What Next?

Two centuries ago, the fictional Edmond Dantès spent 14 years tunneling his way out of the island prison of Château d’If. Two centuries later, the National Registry of Exonerations reports that the U.S. justice system takes an average of 11 years to exonerate an innocent person. We can do better.

With the benefit of hindsight, wrongful convictions can be embarrassingly obvious.

But often the same evidence rejected by prior courts will serve as the clear basis for a later exoneration. The current federal system angles away from conducting discovery and hearing evidence in favor of reviewing a cold record with a dubious eye and almost impossibly high legal burdens of proof. When a future court finally makes the right decision, a transformation occurs overnight. We applaud the judge and question how such an obvious injustice could have ever occurred. The answer, however, is that we’ve created a system of federal habeas review that worries more about hurting the feelings of state judges than protecting the constitutional rights of citizens.

Even against the backdrop of AEDPA, the system would also be more meaningful if more prosecutors would work cooperatively during these proceedings. Psychologically, however, some prosecutors appear to struggle to apprehend the possibility that duly convicted people within their jurisdictions may nevertheless be innocent. They comprehend that wrongful convictions occur in other jurisdictions, just not their own. On a human level, it is understandable that prosecutors resist believing that they, or their peers, may have sent an innocent person to prison, especially when they genuinely believe the trial prosecutor did nothing wrong. But logic tells us that proof beyond a reasonable doubt is not the same as proof to a mathematical certainty. Innocent people will always slip through the cracks.

In the past few election cycles, a small surge of reform-minded prosecutors has won elected office in major metropolitan areas. Some of these prosecutors have established conviction integrity units that review potential cases of innocence from prior administrations. If a wrongfully convicted prisoner can persuade these conviction integrity units of their innocence, the prosecutor’s office will assist or join the prisoner in filing a motion to vacate the conviction in state court. These jurisdictions are a trending minority, but a meaningful one. There is still work to be done, both by shifting perspectives and by fixing poorly conceived laws.

In any rational system of justice, innocence must be paramount. In federal habeas corpus, however, the justice system has become an end in itself. The system cannot worry itself with offending state judges who were wrong. When it comes to the Constitution, wrong is wrong, whether you miss by an inch or you miss by a mile. And the focus needs to return to the central purpose of that system: protecting the innocent.