NOTICE: This opinion is subject to formal revision before publication in the preliminary print of the United States Reports. Readers are requested to notify the Reporter of Decisions, Supreme Court of the United States, Washington, D. C. 20543, of any typographical or other formal errors, in order that corrections may be made before the preliminary print goes to press.
SUPREME COURT OF THE UNITED STATES
_________________
No. 20–826
_________________
MIKE BROWN, ACTING WARDEN, PETITIONER
v. ERVINE DAVENPORT
on writ of certiorari to the united states court of appeals for the sixth circuit
[April 21, 2022]
Justice Gorsuch delivered the opinion of the Court.
After a state court determines that an error at trial did not prejudice a criminal defendant, may a federal court grant habeas relief based solely on its independent assessment of the error’s prejudicial effect under
Brecht v.
Abrahamson,
507 U.S. 619 (1993)? Or must a federal court also evaluate the state court’s decision under the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA)? The Sixth Circuit ruled that an individual who satisfies
Brecht alone is entitled to habeas relief. This was mistaken. When a state court has ruled on the merits of a state prisoner’s claim, a federal court cannot grant relief without first applying both the test this Court outlined in
Brecht and the one Congress prescribed in AEDPA.
I
A
One evening in 2007, Annette White attended a gathering with Ervine Davenport. On the drive home, Mr. Davenport killed Ms. White. At trial, the only questions concerned why and how. Mr. Davenport claimed self-defense and testified to that effect. On his account, Ms. White grew angry during the trip and tried to grab the steering wheel from him while he was driving. Then she pulled out a box cutter and cut his arm. Mr. Davenport responded by extending one arm and pinning Ms. White against the passenger side of the car, with his hand under her chin. Eventually, she stopped struggling. On discovering that Ms. White was no longer breathing, Mr. Davenport panicked and left her body in a field.
The prosecution offered a very different version of events. It stressed that Ms. White was 5’2” tall, 103 pounds, and had a broken wrist, while Mr. Davenport was 6’5” tall and weighed nearly 300 pounds. The prosecution presented evidence that Mr. Davenport had bragged to others before the killing that, if he had a problem with someone, he would choke the person. Days before Ms. White’s death, Mr. Davenport had done just that—strangling another woman until she lost consciousness and urinated on herself. Nor, on the prosecution’s account, were Mr. Davenport’s actions after Ms. White’s death consistent with his claim of self-defense. Instead of contacting the police, he not only abandoned his victim’s body. He also fled the scene and later visited Ms. White’s home where he stole electronics and food. He told a witness, too, that he “had to off ” Ms. White.
The prosecution offered additional proof. When police questioned Mr. Davenport, he gave differing accounts and initially denied any involvement in Ms. White’s death. While authorities did locate a box cutter in the car, they did not find it inside the cab of the vehicle but in the trunk and untainted by blood. Also, a forensic pathologist testified that Ms. White died of manual strangulation. The pathologist explained that a victim of strangulation may lose consciousness after 30 seconds, but that death does not occur until the victim is without air for at least four to five minutes. After Mr. Davenport testified that he merely extended his arm across Ms. White’s neck to keep her from cutting him, the forensic pathologist offered his view that this account was not plausible. Ms. White’s injuries, found on both sides of her neck, were consistent with strangulation—but inconsistent with the application of broad force across the front of her neck.
After a 7-day trial, a jury convicted Mr. Davenport of first-degree murder.
B
On direct appeal in state court, Mr. Davenport sought to have his conviction set aside in light of
Deck v.
Missouri,
544 U.S. 622 (2005). In
Deck, this Court held that the
Fourteenth Amendment’s Due Process Clause generally forbids shackling a criminal defendant at trial absent “a special need.”
Id., at 626. Mr. Davenport noted that during his trial (but not his testimony) officials shackled one of his hands, his waist, and his ankles. Those shackles may not have been visible to many in the courtroom because of a “privacy screen” around the table where Mr. Davenport sat. But the trial court did not articulate on the record any special need for its security measures.
Ultimately, the Michigan Supreme Court agreed that the trial court’s actions violated
Deck. At the same time, the court sought to apply
Chapman v.
California,
386 U.S. 18 (1967). In
Chapman, this Court held that a preserved claim of constitutional error identified on direct appeal does not require reversal of a conviction if the prosecution can establish that the error was harmless beyond a reasonable doubt.
Id., at 24. To answer
Chapman’s question, the Michigan Supreme Court remanded the case to the trial court with instructions to determine whether “the jury saw the defendant’s shackles” and, if so, “whether the prosecution can demonstrate beyond a reasonable doubt that the shackling error did not contribute to the verdict against the defendant.”
People v.
Davenport, 488 Mich. 1054, 794 N.W.2d 616 (2011).
On remand, the trial court conducted an evidentiary hearing in which it heard from all 12 jurors. Five remembered seeing Mr. Davenport’s restraints; the remaining seven did not. All 12 testified that Mr. Davenport’s shackles did not enter into their deliberations or influence their unanimous verdict. Based on this evidence, the trial court found that the State had carried its burden to show harmlessness beyond a reasonable doubt.
Again, Mr. Davenport appealed. This time, Michigan’s appellate courts declined to disturb the judgment. For its part, the Michigan Court of Appeals held that “the prosecution proved beyond a reasonable doubt that the shackling error did not affect the verdict.”
People v.
Davenport, 2012 WL 6217134, *3 (Dec. 13, 2012) (
per curiam). In doing so, the court relied on both the jurors’ testimony and that “the evidence at trial overwhelmingly established defendant’s guilt and belied his contention that he killed the 103-pound victim in self-defense, a theory that was explicitly disputed by expert medical testimony.”
Id., at *2, n. 2. The Michigan Supreme Court denied Mr. Davenport’s request for discretionary review.
People v.
Davenport, 494 Mich. 875, 832 N.W.2d 389, 390 (2013).
C
Mr. Davenport next sought relief in federal district court, filing a habeas petition in the Western District of Michigan. Under AEDPA, however, a federal court may disturb a final state-court conviction in only narrow circumstances. As relevant here, the statute provides that, when a state court has already ruled on the merits of the habeas petitioner’s claim, he must show that decision was either (1) “contrary to” or an “unreasonable application of ” clearly established federal law, as determined by the decisions of this Court, or (2) based on an “unreasonable determination of the facts” presented in the state-court proceeding. 28 U. S. C. § 2254(d).
The District Court found relief unwarranted under this standard. The Michigan state courts had ruled on the merits of Mr. Davenport’s claim of error. In doing so, they correctly identified this Court’s controlling harmless-error rule from
Chapman. And their conclusions involved neither an unreasonable application of
Chapman nor an unreasonable determination of the facts. To the contrary, the District Court agreed with a Magistrate Judge’s assessment that the state-court record contained no evidence “that the jurors were influenced” by his restraints and “overwhelming evidence of [Mr. Davenport’s] guilt.”
Davenport v.
MacLaren, 2016 WL 11262506, *4 (WD Mich., Nov. 7, 2016); see also
Davenport v.
MacLaren, 2017 WL 4296808, *1–*2 (WD Mich., Sept. 26, 2017) (citing 28 U. S. C. § 2254(d)(1)).
D
After that loss, Mr. Davenport appealed to the Sixth Circuit, where a divided panel reversed.
Davenport v.
MacLaren, 964 F.3d 448 (2020).
Unlike the District Court, the Sixth Circuit declined to analyze the case under AEDPA. Instead, it held, only this Court’s decision in
Brecht v.
Abrahamson governed its review. Handed down before Congress adopted AEDPA,
Brecht sought to adapt
Chapman’s harmless-error rule, developed for cases on direct appellate review, for use in federal habeas proceedings.
Brecht, 507 U. S., at 633–635. Citing the need to afford appropriate respect to final state-court decisions that have already endured direct appeal, including potential review in this Court,
Brecht effectively inverted
Chapman’s burden. 507 U. S., at 635. Rather than require the prosecution to prove that a constitutional trial error is harmless,
Brecht held that a state prisoner seeking to challenge his conviction in collateral federal proceedings must show that the error had a “ ‘substantial and injurious effect or influence’ ” on the outcome of his trial.
Id., at 637. Persuaded that Mr. Davenport could satisfy his burden under
Brecht, the panel majority ordered Michigan to retry or release him promptly. 964 F. 3d, at 464–468.
Judge Readler dissented. He argued that
Brecht and AEDPA set forth independent tests, and that both must be satisfied before habeas relief becomes permissible. In Judge Readler’s view, too, the District Court correctly rejected Mr. Davenport’s petition under AEDPA because the state courts hearing his case had not acted contrary to, or unreasonably applied, this Court’s decisions. 964 F. 3d, at 469, 478.
The Sixth Circuit denied rehearing en banc by a vote of 8 to 7.
Davenport v.
MacLaren, 975 F.3d 537 (2020). Judges Griffin and Thapar issued dissenting opinions. They expressed agreement with Judge Readler and observed that the panel majority’s decision conflicted with those of other circuits where petitioners are required to satisfy both
Brecht and AEDPA before becoming eligible for habeas relief. 975 F. 3d, at 552 (Thapar, J., dissenting) (citing decisions from the Third, Seventh, Tenth, and Eleventh Circuits). We granted Michigan’s petition for certiorari to resolve the conflict in the federal courts of appeals about the proper interaction between these two tests. 593 U. S. ___ (2021).
II
When Congress supplies a constitutionally valid rule of decision, federal courts must follow it. In AEDPA, Congress announced such a rule. It instructed that a federal court “
shall not . . .
gran[
t]” relief with respect to a claim that has been adjudicated on the merits in state court “
unless” the state court’s decision was (1) “contrary to” or an “unreasonable application of ” clearly established federal law, as determined by the decisions of this Court, or (2) based on an “unreasonable determination of the facts” presented in the state-court proceeding. § 2254(d) (emphasis added).
The upshot of these directions for our case is straightforward. No one questions that a state court’s harmless-error determination qualifies as an adjudication on the merits under AEDPA. See
Davis v.
Ayala,
576 U.S. 257, 269 (2015);
Fry v.
Pliler,
551 U.S. 112, 119 (2007);
Early v.
Packer,
537 U.S. 3, 10–11 (2002) (
per curiam). No one disputes that such a decision exists here. Nor does Mr. Davenport pursue any claim to relief under § 2254(d)(2). From this, it follows that he must satisfy § 2254(d)(1) to secure federal habeas relief. To be sure, where
Brecht is implicated a federal court must also ensure a habeas petitioner has carried his burden under its terms before granting relief. But in cases like ours satisfying
Brecht is only a necessary, not a sufficient, condition to relief. AEDPA too must be satisfied. The Sixth Circuit erred in holding otherwise.
A
Some background helps explain this arrangement. From the founding, Congress authorized federal courts to issue habeas writs to federal custodians. § 14,
1Stat.
81–82. After the Civil War, Congress extended this authority, allowing federal courts to issue habeas writs to state custodians as well. See Act of Feb. 5, 1867, ch. 28, § 1,
14Stat.
385. But these statutes used permissive rather than mandatory language; federal courts had the “power to” grant writs of habeas corpus in certain circumstances. That same structure lives on in contemporary statutes, which provide that federal courts “may” grant habeas relief “as law and justice require.” 28 U. S. C. §§ 2241, 2243;
Wright v.
West,
505 U.S. 277, 285 (1992) (plurality opinion).
Over the centuries, a number of writs of habeas corpus evolved at common law to serve a number of different functions. See
Ex parte Bollman, 4 Cranch 75, 97–98 (1807); 3 W. Blackstone, Commentaries on the Laws of England 129–131 (1768). But the most notable among these writs was that of
habeas corpus ad subjiciendum, often called the “Great Writ.”
Id., at 131. When English monarchs jailed their subjects summarily and indefinitely, common-law courts employed the writ as a way to compel the crown to explain its actions—and, if necessary, ensure adequate process, such as a trial, before allowing any further detention. See Petition of Right, 3 Car. 1, ch.1, ¶¶ 5, 8 (1628). The Great Writ was, in this way, no less than “the instrument by which due process could be insisted upon.”
Hamdi v.
Rumsfeld,
542 U.S. 507, 555 (2004) (Scalia, J., dissenting).
At the same time, even this writ had its limits. Usually, a prisoner could not use it to challenge a final judgment of conviction issued by a court of competent jurisdiction. See,
e.g.,
Opinion on the Writ of Habeas Corpus, Wilm. 77, 88, 97 Eng. Rep. 29, 36 (K. B. 1758). If the point of the writ was to ensure due process attended an individual’s confinement, a
trial was generally considered proof he had received just that. See,
e.g.,
Bushell’s Case, Vaugh. 135, 142–143, 124 Eng. Rep. 1006, 1009–1010 (C. P. 1670).
This traditional understanding extended from England to this country and persisted through much of our history. Asked to apply the Nation’s first habeas statute to a duly convicted prisoner, Chief Justice Marshall invoked the common-law rule that a judgment of conviction after trial was “conclusive on all the world.”
Ex parte Watkins, 3 Pet. 193, 202–203 (1830). Acknowledging that Congress had authorized the Court to “inquire into the sufficiency of ” the cause of the petitioner’s detention, Marshall asked rhetorically, “is not that judgment in
itself sufficient cause?”
Id., at 202 (emphasis added); see also
Ex parte Parks,
93 U.S. 18, 21–22 (1876); P. Bator, Finality in Criminal Law and Federal Habeas Corpus for State Prisoners, 76 Harv. L. Rev. 441, 465–469 (1963) (Bator).
If the answer was nearly always yes, an important exception existed in both English and American law: A habeas court could grant relief if the court of conviction lacked jurisdiction over the defendant or his offense. See
Watkins, 3 Pet., at 202–203; Bator 471–472. A perceived “error in the judgment or proceedings, under and by virtue of which the party is imprisoned, constitute[d] no ground for” relief.
Ex parte Siebold,
100 U.S. 371, 375 (1880). Instead, a habeas court could “examin[e] only the power and authority of the court to act, not the correctness of its conclusions.”
Harlan v.
McGourin,
218 U.S. 442, 448 (1910). To be sure, the line between mere errors and jurisdictional defects was not always a “luminous beacon” and it evolved over time. Bator 470;
Edwards v.
Vannoy, 593 U. S. ___, ___ (2021) (Gorsuch, J., concurring) (slip op., at 5). But this Court generally sought to police the doctrine’s boundaries in cases involving federal and state prisoners alike.[
1]
By 1953, however, federal habeas practice began to take on a very different shape. That year in
Brown v.
Allen this Court held that a state-court judgment “is not
res judicata” in federal habeas proceedings with respect to a petitioner’s federal constitutional claims.
344 U.S. 443, 458 (1953). A state court may reject the petitioner’s claims after a fair hearing. No appellate court, including this one, may see fit to reverse that final judgment. Yet still,
Brown suggested, a federal district court approaching the same case years later should be free to decide
de novo whether the state-court proceedings “resulted in a satisfactory conclusion” and to issue habeas relief if that conclusion is found wanting.
Id., at 463; see also
Wright, 505 U. S., at 287–288 (plurality opinion). The traditional distinction between jurisdictional defects and mere errors in adjudication no longer restrained federal habeas courts. Full-blown constitutional error correction became the order of the day.
This shift did not go unnoticed. Concurring only in the result, Justice Jackson contended that the Court’s decision “trivializ[ed] . . . the writ” and was inconsistent with the presumption of finality that traditionally attached to criminal convictions.
Brown, 344 U. S., at 536, 543. He warned, too, that the Court’s ruling threatened “haystack[s]” of new habeas petitions—and that federal courts would struggle to identify the meritorious “needle[s]” among them.
Id., at 537. Over the ensuing years, that prediction proved prescient: Federal courts struggled with an exploding caseload of habeas petitions from state prisoners. See,
e.g.,
Schneckloth v.
Bustamonte,
412 U.S. 218, 274, n. 37 (1973) (Powell, J., concurring) (“In 1971 . . . state prisoners alone filed 7,949 petitions for habeas in federal district courts, over 14 times the number filed when Mr. Justice Jackson voiced his misgivings”); B. Garrett & L. Kovarsky, Federal Habeas Corpus 135–136 (2013) (documenting the rise of habeas filings by state prisoners).[
2]
B
Eventually, this Court responded to the post-
Brown habeas boom by devising new rules aimed at separating the meritorious needles from the growing haystack. The habeas statutes themselves provided the starting place for these efforts. Recall that Congress invested federal courts with discretion when it comes to supplying habeas relief—providing that they “may” (not must) grant writs of habeas corpus, and that they should do so only as “law and justice require.” 28 U. S. C. §§ 2241, 2243. This language, the Court recognized, serves as “authorization to adjust the scope of the writ in accordance with equitable and prudential considerations.”
Danforth v.
Minnesota,
552 U.S. 264, 278 (2008); see also
Withrow v.
Williams,
507 U.S. 680, 716 (1993) (Scalia, J., concurring in part and dissenting in part). Foremost among those considerations is the States’ “powerful and legitimate interest in punishing the guilty.”
Calderon v.
Thompson,
523 U.S. 538, 556 (1998) (internal quotation marks omitted). Granting habeas relief to a state prisoner “intrudes on state sovereignty to a degree matched by few exercises of federal judicial authority.”
Harrington v.
Richter,
562 U.S. 86, 103 (2011) (internal quotation marks omitted).
Exercising its equitable discretion, and informed by these concerns, the Court began to develop doctrines “aimed at returning the Great Writ closer to its historic office.”
Edwards, 593 U. S., at ___ (Gorsuch, J., concurring) (slip op., at 8). The Court established procedural-default standards to prevent petitioners from evading independent and adequate state-law grounds sustaining their convictions.
Wainwright v.
Sykes,
433 U.S. 72, 86–87 (1977). The Court held that some claims are not cognizable in federal habeas if state courts provide a mechanism for review.
Stone v.
Powell,
428 U.S. 465, 494–495 (1976). The Court also applied new rules to prevent cycles of repetitive filings.
McCleskey v.
Zant,
499 U.S. 467, 486–493 (1991).
Brecht was part of this effort. In
Chapman, this Court held that, when a defendant demonstrates on direct appeal that a constitutional error occurred at his trial, his conviction cannot stand unless the government proves the error’s harmlessness “beyond a reasonable doubt.” 386 U. S., at 24. In
Brecht, the Court resolved that this same standard was inappropriate for use in federal habeas review of final state-court judgments. 507 U. S., at 633–634. Instead, the Court reasoned, a state prisoner should not receive federal “habeas relief based on trial error unless” he can show the error had a “substantial and injurious effect or influence” on the verdict.
Id., at 637 (internal quotation marks omitted). In reaching its judgment, the Court stressed that undoing a final state-court judgment is an “extraordinary remedy,” reserved for only “ ‘extreme malfunctions in the state criminal justice system’ ” and different in kind from providing relief on direct appeal.
Id., at 633–634. To allow a federal habeas court to set aside a conviction based on nothing more than “speculation that the defendant was prejudiced by trial error” would be to give short shrift to the State’s “sovereign interes[t]” in its final judgment.
Calderon v.
Coleman,
525 U.S. 141, 146 (1998) (
per curiam). Much as the Court had “filled the gaps of the habeas statute with respect to other matters,” it found it “necessary to do so” again, in a by-now familiar exercise of its equitable discretion.
Brecht, 507 U. S., at 633.
C
Three years after
Brecht, and apparently finding the Court’s equitable doctrines insufficient, Congress introduced its own reforms in AEDPA.
In many ways, the statute represented a sea change in federal habeas law. As we have seen, Congress instructed that, if a state court has adjudicated the petitioner’s claim on the merits, a federal court “shall not” grant habeas relief “unless” certain conditions are satisfied. § 2254(d). Some of these conditions were new to the law at the time of their adoption; all are demanding. See
Richter, 562 U. S., at 102.
Still, Congress did not wash away everything that came before. While AEDPA announced certain new conditions to relief, it did not guarantee relief upon their satisfaction. Instead, Congress left intact the equitable discretion traditionally invested in federal courts by preexisting habeas statutes. So even a petitioner who prevails under AEDPA must still today persuade a federal habeas court that “law and justice require” relief. § 2243. See
Fry, 551 U. S., at 119;
Horn v.
Banks,
536 U.S. 266, 272 (2002) (
per curiam). And whatever else those inquiries involve, they continue to require federal habeas courts to apply this Court’s precedents governing the appropriate exercise of equitable discretion—including
Brecht. See
Banks, 536 U. S., at 272;
Johnson v.
Acevedo, 572 F.3d 398, 404 (CA7 2009); see also
Edwards, 593 U. S., at ___, n. 5 (Gorsuch, J., concurring) (slip op., at 9, n. 5).
Today, then, a federal court must
deny relief to a state habeas petitioner who fails to satisfy either this Court’s equitable precedents or AEDPA. But to
grant relief, a court must find that the petitioner has cleared both tests. The Sixth Circuit erred when it held Mr. Davenport to just one of these burdens. It granted relief after finding for him on
Brecht. But it failed to ask the further question whether he satisfied AEDPA. In doing so, the court disregarded Congress’s instruction that habeas relief “shall not be granted” unless AEDPA’s terms are satisfied. § 2254(d).
III
Mr. Davenport advances two arguments—one logical, one doctrinal—in defense of the Sixth Circuit’s decision. We consider them in turn.
A
Mr. Davenport first suggests the Sixth Circuit’s failure to discuss AEDPA amounted to no more than a forgivable peccadillo. On his account, the AEDPA inquiry represents a logical subset of the
Brecht test. So even though the Sixth Circuit did not formally find that he satisfied AEDPA, it implicitly did so when it found his case cleared
Brecht.
This theory is mistaken. Proof of prejudice under
Brecht does not equate to a successful showing under AEDPA. Instead, the inquiries are “entirely different in kind.” J. Greabe, The Riddle of Harmless Error Revisited, 54 Houston L. Rev. 59, 113, n. 297 (2016) (emphasis deleted). They pose courts with different questions to resolve and require courts to answer those questions based on different legal materials.
Take the questions the two tests pose. When a state court has applied
Chapman, § 2254(d)(1) requires a habeas petitioner to prove that the state court’s decision was unreasonable.
Cullen v.
Pinholster,
563 U.S. 170, 181 (2011);
Fry, 551 U. S., at 119. To accomplish that, a petitioner must persuade a federal court that no “fairminded juris[t]” could reach the state court’s conclusion under this Court’s precedents.
Ayala, 576 U. S., at 269 (internal quotation marks omitted). Similarly, if a petitioner alleges the state court’s decision “was based on an unreasonable determination of the facts” under § 2254(d)(2), it is not enough to show that “reasonable minds reviewing the record might disagree about the finding in question.”
Brumfield v.
Cain,
576 U.S. 305, 314 (2015) (internal quotation marks and alteration omitted). By contrast, under
Brecht a petitioner may prevail by persuading a federal court that it alone should harbor “grave doubt”—not absolute certainty—about whether the trial error affected the verdict’s outcome.
O’Neal v.
McAninch,
513 U.S. 432, 435 (1995). In sum, where AEDPA asks whether
every fairminded jurist would agree that an error was prejudicial,
Brecht asks only whether a federal habeas court
itself harbors grave doubt about the petitioner’s verdict.
Next, consider the legal materials a court may consult when applying the two tests. Section 2254(d)(1) limits habeas relief to cases where a state-court decision contravenes or unreasonably applies “clearly established Federal law, as determined by the Supreme Court of the United States.” It is not enough that the state-court decision offends lower federal court precedents. See,
e.g.,
Glebe v.
Frost,
574 U.S. 21, 24 (2014) (
per curiam). This Court’s dicta cannot supply a ground for relief. See,
e.g.,
White v.
Woodall,
572 U.S. 415, 419 (2014). Nor can holdings that speak only at a high level of generality. See,
e.g.,
Lopez v.
Smith,
574 U.S. 1, 6 (2014) (
per curiam);
Yarborough v.
Alvarado,
541 U.S. 652, 664 (2004). Under AEDPA too, “[s]tate-court decisions are measured against this Court’s precedents as of ‘the time the state court renders its decision’ ” and cannot be held unreasonable only in light of later decided cases.
Pinholster, 563 U. S., at 182 (quoting
Lockyer v.
Andrade,
538 U.S. 63, 71–72 (2003)). None of these restrictions applies under
Brecht. There, a federal habeas court may consult and draw on the whole body of law. So, for example, a petitioner might be able to prevail under
Brecht thanks to favorable circuit case law but still lose under AEDPA because no comparable holding exists in this Court’s precedents.
Today’s case illustrates how these differences matter. The Sixth Circuit granted relief to Mr. Davenport after concluding that it harbored grave doubts about the jury’s verdict. It did not claim that every reasonable jurist would share its doubts. Nor did it purport to hold that the Michigan state courts had acted contrary to or unreasonably applied a decision of this Court. Instead, the Sixth Circuit said only that the state-court decisions in this case could not be reconciled with a roughly analogous precedent from the Ninth Circuit. 964 F. 3d, at 467. Even assuming the Sixth Circuit’s analysis was enough to permit relief under
Brecht, none of its reasoning was enough to warrant relief under AEDPA. Nor can any of this come as a surprise. As we have seen, if AEDPA makes winning habeas relief more difficult, it is because Congress adopted the law to do just
that.[
3]
B
Failing in his first argument, Mr. Davenport offers an alternative. Even if all we have said is true as a matter of logic, he suggests, we should rule for him anyway as a matter of precedent thanks to
Fry v.
Pliler,
551 U.S. 112, and
Davis v.
Ayala,
576 U.S. 257.
Here, too, we cannot agree. Start with
Fry. Because no state court had ruled on the merits of the petitioner’s
Chapman claim, everyone in
Fry agreed that AEDPA did not apply to his federal habeas petition. Seeking to leverage that fact to his further advantage, the petitioner argued that Congress implicitly swept away this Court’s equitable habeas precedents when it adopted AEDPA. 551 U. S., at 119. The upshot? On the petitioner’s view, this meant a federal habeas court had to apply
Chapman (not
Brecht or AEDPA) to his case. Ultimately, the Court rejected this argument, confirming instead that our equitable precedents remain applicable “whether or not” AEDPA applies. 551 U. S., at 121.
None of this answers the question we face today. Where the petitioner in
Fry sought to suggest that AEDPA rendered
Brecht a dead letter, the petitioner here argues nearly the opposite. And the Court’s ruling in
Fry is, if anything, affirmatively unhelpful to Mr. Davenport. Not only did
Fry hold that this Court’s equitable precedents like
Brecht coexist side-by-side with AEDPA, it expressly recognized that AEDPA “sets forth a precondition to the grant of habeas relief . . . not an entitlement to it.” 551 U. S., at 119–120. Rather than suggest
Brecht duplicates AEDPA or vice versa,
Fry thus stands as a reminder that the two tests impose analytically distinct preconditions to relief.
Mr. Davenport offers no persuasive reply. Instead, he invites us to overlook all of this and train our attention to a brief passage at the end of
Fry’s relevant analysis. There, he notes, the Court said this: “[I]t certainly makes no sense to require formal application of
both tests (AEDPA/
Chapman and
Brecht) when the latter obviously subsumes the former.” 551 U. S., at 120. On Mr. Davenport’s telling, whatever else
Fry did or said, this language means it adopted his theory that a court applying
Brecht necessarily applies AEDPA along the way.
We do not see how we might read so much into so little. Doubtless, there are some cases “when”
Brecht does “subsum[e]” AEDPA, just as
Fry says. 551 U. S., at 120. As we have seen, and as was the case in
Fry, if a state court has not adjudicated the petitioner’s claim on the merits, AEDPA falls away. Likewise, if a federal court determines that a habeas petition fails because of
Brecht, there is no need to prolong the matter by “formal[ly] appl[ying]” AEDPA as well. 551 U. S., at 120. But none of this means, and
Fry never said, that a
Brecht inquiry
always subsumes an AEDPA inquiry. Nor did
Fry even have reason to consider the relationship between
Brecht and AEDPA in cases like ours—where a state court has issued a decision on the merits of the petitioner’s claim and AEDPA does apply. To the extent
Fry had anything to say about
that scenario, all it said was consistent with what we say today: In such cases, the Court recognized, relief “may not be granted unless the state court’s adjudication” was “unreasonable” under AEDPA. 551 U. S., at 119.[
4]
That leaves
Davis v.
Ayala, where a similar story unfolds. There, the California Supreme Court did rule on the merits of the petitioner’s claim: It determined that any federal constitutional errors in his trial-court proceedings were harmless under
Chapman.
Ayala, 576 U. S., at 263–264. In later federal habeas proceedings, the Ninth Circuit applied
Brecht, found prejudice, and granted relief without pausing to consider AEDPA directly.
Ayala v.
Wong, 756 F.3d 656, 674 (CA9 2014). Instead, it asserted, much as Mr. Davenport does, that a favorable finding for a petitioner under
Brecht necessarily answers AEDPA’s distinct inquiry. 756 F. 3d, at 674, n. 13. In the end, however, this Court reversed. We held that the Ninth Circuit misapplied
Brecht and that a proper
Brecht analysis precluded relief. Along the way, we indicated that the petitioner could not prove eligibility for relief under AEDPA’s demanding standard either.
With nothing in this holding to help him, Mr. Davenport again asks us to focus on a carefully curated snippet. Borrowing language from
Fry,
Ayala observed that “a prisoner who seeks federal habeas corpus relief must satisfy
Brecht, and if the state court adjudicated his claim on the merits, the
Brecht test subsumes the limitations imposed by AEDPA.”
Ayala, 576 U. S., at 270 (citing
Fry, 551 U. S., at 119–120). Again, though, it is hard to see how this much might carry the day. When a federal court determines, as we did in
Ayala, that a petitioner has
failed to carry his burden under
Brecht, that conclusion subsumes (or perhaps more precisely, obviates the need for) a separate AEDPA inquiry; relief must be denied. But none of this resolves the distinct question we face today—whether a petitioner who
can satisfy
Brecht also necessarily secures a victory under AEDPA. The
Ayala Court had no occasion to address that question. And to the extent it spoke to it, it spoke much as
Fry had, taking pains to reject any suggestion “that
Brecht somehow abrogates the limitation on federal habeas relief that § 2254(d) plainly sets out.”
Ayala, 576 U. S., at 268.[
5]
In the end, Mr. Davenport’s appeals to
Fry and
Ayala (echoed by the dissent) rest on a misunderstanding of
stare decisis. At its best, that doctrine is a call for judicial humility. It is a reminder to afford careful consideration to the work of our forbearers, their experience, and their wisdom. But respect for past judgments also means respecting their limits. This Court has long stressed that “the language of an opinion is not always to be parsed as though we were dealing with [the] language of a statute.”
Reiter v.
Sonotone Corp.,
442 U.S. 330, 341 (1979). Yet that is exactly what Mr. Davenport and the dissent ask of us. They would have us override a lawful congressional command—that no federal habeas relief should issue “unless” AEDPA’s applicable conditions are satisfied. § 2254(d). And they would have us do so on the basis of a handful of sentences extracted from decisions that had no reason to pass on the argument Mr. Davenport presents today. We neither expect nor hope that our successors will comb these pages for stray comments and stretch them beyond their context—all to justify an outcome inconsistent with this Court’s reasoning and judgments and with Congress’s instructions. Such an exalted view of this Court’s every passing remark would turn
stare decisis from a tool of judicial humility into one of judicial hubris.
IV
Having concluded that the Sixth Circuit erred by failing to apply AEDPA before granting habeas relief, one question remains: Assuming Mr. Davenport can satisfy
Brecht as the Sixth Circuit held, can he satisfy AEDPA? The answer helps illustrate how the two inquiries are distinct and why a federal court must answer both before overturning a state-court conviction.
Under the statute’s terms, we assess the reasonableness of the “last state-court adjudication on the merits of ” the petitioner’s claim.
Greene v.
Fisher,
565 U.S. 34, 40 (2011). In this case, that is the decision of the Michigan Court of Appeals. To be sure, after that intermediate court ruled against Mr. Davenport he sought discretionary review in the Michigan Supreme Court, which denied his request. See 494 Mich., at 875, 832 N. W. 2d, at 390. But a discretionary denial of leave to appeal does not typically entail an “adjudication” of the underlying claim’s “merits” under AEDPA’s terms. Instead, it usually represents “a decision by the state supreme court not to hear the appeal—that is, not to decide at all.”
Greene, 565 U. S., at 40; cf.
Ylst v.
Nunnemaker,
501 U.S. 797, 805–806 (1991) (“[T]he discretionary denial of review on direct appeal by the California Supreme Court is not even a ‘judgment’ ”).
In this respect, the Michigan Supreme Court follows a familiar practice. “The denial of a writ of certiorari” in this Court “imports no expression of opinion upon the merits of the case.”
United States v.
Carver,
260 U.S. 482,
490 (1923). And the Michigan Supreme Court has long described its denials of applications for leave to appeal in the same terms. See,
e.g.,
Malooly v.
York Heating & Ventilating Corp., 270 Mich. 240, 246–247, 258 N.W. 622, 624 (1935); see also Mich. Ct. Rules 7.303(B)(1), 7.305(B)(1)–(3) (2021). In the past, too, this Court has treated lower Michigan court decisions as the relevant AEDPA adjudication despite discretionary denials of review by the State Supreme Court.
Woods v.
Donald,
575 U.S. 312, 314–315, 317 (2015) (
per curiam);
Burt v.
Titlow,
571 U.S. 12, 20 (2013);
Lafler v.
Cooper,
566 U.S. 156, 161, 173 (2012).
Turning to the decision of the Michigan Court of Appeals, Mr. Davenport principally argues that it was contrary to or an unreasonable application of this Court’s decision in
Holbrook v.
Flynn,
475 U.S. 560 (1986). But it is hard to see how that could be the case. For one thing,
Holbrook was a decision about whether a constitutional trial error occurred at all, not whether the alleged error was prejudicial. For another,
Holbrook rejected the defendant’s claim that he “was denied his constitutional right to a fair trial whe[n] . . . the customary courtroom security force was supplemented by four uniformed state troopers sitting in the first row of the spectator’s section.”
Id., at 562. Nothing in that analysis is inconsistent with the Michigan Court of Appeals’ disposition of Mr. Davenport’s shackling claim.
Unable to make use of
Holbrook’s holding, Mr. Davenport once more asks us to turn our attention elsewhere. In particular, he notes that the trial court in
Holbrook asked potential jurors at the outset of trial whether the presence of state troopers would affect their ability to consider fairly the defendant’s case; they said no. This Court indicated that such questions and answers cannot alone “dispositive[ly]” resolve the question whether security measures prejudice a defendant’s right to a fair trial.
Id.,
at 570. Instead, the Court reasoned
that “jurors will not necessarily be fully conscious of the effect [such measures] will have on their attitude toward the accused. This will be especially true when jurors are questioned at the very beginning of the proceedings; at that point, they can only speculate on how they will feel after being exposed to a practice daily over the course of a long trial.”
Ibid. Mr. Davenport highlights that the Court in
Deck favorably quoted some of these passages too. 544 U. S., at 635.
Even bearing all this in mind, however, we cannot see how the Michigan Court of Appeals acted contrary to or unreasonably applied clearly established federal law. The Michigan court found the shackling in Mr. Davenport’s case harmless for two reasons—both because of the “overwhelmin[g]” evidence against him, and because jurors testified that his shackling did not affect their verdict. 2012 WL 6217134, at *1–*2, and n. 2.
Holbrook does not compel a different ruling on the first score. It addressed a different evidentiary record and affirmed the defendant’s conviction.
Nor does the decision compel a different ruling on the second score.
Holbrook cast doubt only on attempts to assess prejudice based on testimony from prospective jurors speculating about how the security measures might affect their perceptions over the course of an upcoming trial. Nothing in the decision purported to forbid courts from considering post-trial testimony about how trial security measures
actually affected juror deliberations. Nor may this or any federal court use an AEDPA case as an opportunity to pass on the wisdom of extending old precedents in new ways. AEDPA permits relief only when a state court acts contrary to or unreasonably applies this Court’s preexisting and clearly established rules. See,
e.g.,
Marshall v.
Rodgers,
569 U.S. 58, 64 (2013) (
per curiam);
Woodall, 572 U. S., at 424–426;
Lopez, 574 U. S., at 6.
As a backup, Mr. Davenport suggests that the Michigan Court of Appeals committed an independent error by unreasonably applying
Chapman. But
Chapman merely announced the default burden of proof for evaluating constitutional errors on direct appeal: The prosecution must prove harmlessness beyond a reasonable doubt. 386 U. S., at 24. And this Court has repeatedly explained that, when it comes to AEDPA, “the more general the [federal] rule[,] . . . the more leeway [state] courts have in reaching outcomes in case-by-case determinations” before their decisions can be fairly labeled unreasonable.
Renico v.
Lett,
559 U.S. 766, 776 (2010) (internal quotation marks and alteration omitted).
The case before us does not come close to exceeding that leeway. The Michigan Court of Appeals properly identified the controlling standard. 2012 WL 6217134, at *1. It then proceeded to find that the prosecution had established Mr. Davenport’s shackling was harmless beyond a reasonable doubt in light of the jurors’ testimony and the “overwhelmin[g]” evidence that “established defendant’s guilt and belied his contention that he killed the 103 pound victim in self-defense, a theory that was explicitly disputed by expert medical testimony.”
Id., at *1–*2, and n. 2. Even if
some fairminded jurist applying
Chapman could reach a different conclusion, we cannot say that
every fairminded jurist must.
Finally, in a variation on his
Chapman argument, Mr. Davenport faults the Michigan Court of Appeals for unreasonably focusing on a “false choice between first-degree murder and self-defense.” Brief for Respondent 45. In his view, the state court improperly ignored the possibility that Mr. Davenport’s shackling might have influenced the jury toward a conviction for first-degree murder rather than second-degree murder. But Mr. Davenport never presented this theory to the Michigan Court of Appeals. See Brief for Appellant in
People v.
Davenport, No. 306868, pp. 22–24 (arguing only that Mr. Davenport’s shackling influenced jurors’ receptiveness to his self-defense theory). Nor does it directly respond to what the state court called the “overwhelmin[g]” record evidence he committed murder in the first degree. In these circumstances we cannot say that every fairminded court would have both identified and adopted Mr. Davenport’s forfeited theory.
*
Even assuming Mr. Davenport met his burden under
Brecht, he cannot do so under AEDPA. And a federal court cannot grant habeas relief unless a state prisoner like Mr. Davenport satisfies both this Court’s equitable precedents and Congress’s statute. The judgment of the Court of Appeals is
Reversed.