Richter’s Scale: Proving Unreasonableness Under AEDPA

Josiah Rutledge
Volume 32
,  Issue 2

Introduction

Imagine you are charged with stealing a pair of shoes from a charity auction.1This hypothetical is loosely based on Louis Sachar, Holes (1998), and its film adaptation, Holes (Walt Disney Pictures 2003). You persistently proclaim your innocence, but the evidence against you is strong and you are convicted. After trial, you learn that, during the investigation into the theft, another suspect confessed to the police. No one disclosed this confession to you or your lawyer before trial. You immediately file a petition for post‑conviction relief in state court, alleging a violation of your due process rights under Brady v. Maryland.2373 U.S. 83 (1963). The Brady rule provides that a prosecution’s suppression of material, exculpatory evidence in a criminal case violates a defendant’s due process. Id. at 87. To have your conviction reversed, you must show (among other things) that there is a reasonable probability that the outcome of your trial would have been different if the confession had been disclosed.3See, e.g., Turner v. United States, 137 S. Ct. 1885, 1893 (2017). The state‑court judge, however, is not so keen on engaging in that “factually complex” inquiry.4Id. So he decides to resolve your case using the simplest, most efficient method he knows: asking a fortune teller. Consulting the fortune teller’s tarot cards, the judge concludes that the undisclosed confession was not material and denies your request for relief in a written opinion explaining his decision‑making process (tarot cards and all).

Understandably perturbed, you file a petition for a writ of habeas corpus in federal court, raising the same Brady claim. The district court conducts a thorough review of your claim and issues a thorough opinion. First, the court expresses its firm conviction that your Brady rights were violated—the confession was favorable to you, it was suppressed by police, and it is material.5See Strickler v. Greene, 527 U.S. 263, 281–82 (1999). At this point, you start to get your hopes up, but you read on. Next, the court explains that the state court “adjudicated” your Brady claim “on the merits,” meaning that the Antiterrorism and Effective Death Penalty Act of 1996’s (“AEDPA’s”) limitations on relief—known as Section 2254(d)—apply.6Antiterrorism and Effective Death Penalty Act of 1996, Pub. L. No. 104‑132, § 104(3), 110 Stat. 1214, 1218–19 (codified at 28 U.S.C. § 2254(d)). Section 2254(d) provides:

An application for a writ of habeas corpus on behalf of a person in custody pursuant to the judgment of a State court shall not be granted with respect to any claim that was adjudicated on the merits in State court proceedings unless the adjudication of the claim—

(1) resulted in a decision that was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States; or

(2) resulted in a decision that was based on an unreasonable determination of the facts in light of the evidence presented in the State court proceeding.

28 U.S.C. § 2254(d). In this context, “adjudicated on the merits” is a term of art that is “best understood by stating what it is not: it is not the resolution of a claim on procedural grounds.” Muth v. Frank, 412 F.3d 808, 815 (7th Cir. 2005) (citing Sellan v. Kuhlman, 261 F.3d 303, 311 (2d Cir. 2001)). The objectionable way the state judge adjudicated your claim is irrelevant to this question; Section 2254(d) “applies regardless of the procedures employed or the decision reached by the state court, as long as a substantive decision was reached; the adequacy of the procedures and of the decision are addressed through the lens of § 2254(d), not as a threshold matter.” Teti v. Bender, 507 F.3d 50, 57 (1st Cir. 2007) (citing cases).
No problem, you think: Section 2254(d) “stops short of imposing a complete bar” to relief, and preserves a role for federal courts to “guard against extreme malfunctions in the state criminal justice systems.”7Harrington v. Richter, 562 U.S. 86, 102 (2011) (internal quotation marks omitted) (quoting Jackson v. Virginia, 443 U.S. 307, 332 n.5 (1979) (Stevens, J., concurring in the judgment)). Surely, you think, using fortune tellers and tarot cards to decide a constitutional claim is exactly the sort of malfunction that warrants relief under Section 2254(d).

The court next considers, as required by Section 2254(d), the “arguments or theories [that] supported” the state court’s decision—namely, the tarot cards—and concludes that “fairminded jurists” would unanimously agree that this rationale was inconsistent with clearly established federal law.8See id. Outstanding, you think: Section 2254(d) will not bar relief. But the court continues, explaining that, under circuit precedent, it must also consider reasons that “could have supported” the state court’s decision and ask whether they were similarly unreasonable—even if those hypothesized reasons in fact played no role in the state court’s decision.9See, e.g., Myers v. Superintendent, Ind. State Prison, 410 F. Supp. 3d 958, 991 (S.D. Ind. 2019) (“[T]he Court applies currently controlling . . . Circuit precedent requiring an analysis of what other grounds could have supported the [state] Court of Appeals’ decision.”), rev’d on other grounds sub nom. Myers v. Neal, 975 F.3d 611 (7th Cir. 2020). To that end, the court considers the possibility that the strength of the evidence against you rendered the confession immaterial.10See Smith v. Cain, 565 U.S. 73, 76 (2012) (“We have observed that evidence . . . may not be material if the State’s other evidence is strong enough to sustain confidence in the verdict.” (citing United States v. Agurs, 427 U.S. 97, 112–13, 112 n.21 (1976)). Although the court finds that argument clearly erroneous, it notes that even clear error “will not suffice” for purposes of Section 2254(d).11See White v. Woodall, 572 U.S. 415, 419 (2014) (citing Lockyer v. Andrade, 538 U.S. 63, 75–76 (2003)). The court concludes that a fairminded‑but­‑mistaken judge could find the evidence so overwhelming that the confession was not material, and thus that Section 2254(d) forbids it from granting your petition. Begrudgingly, it denies relief.

This result is shocking, but it is the law in at least some circuits.12See infra Sections II.B.1, II.B.2. The Seventh Circuit has even explicitly addressed the tarot card scenario (albeit in dicta):

[I]t is clear that a bad reason does not necessarily mean that the ultimate result was an unreasonable application of established doctrine. A state court could write that it rejected a defendant’s claim because Tarot cards dictated that result, but its decision might nonetheless be a sound one. If a state court’s rationale does not pass muster under the . . . standard for Section 2254(d)(1) cases, the only consequence is that further inquiry is necessary.13Brady v. Pfister, 711 F.3d 818, 827 (7th Cir. 2013) (emphasis added).

This result stems from a fundamental misreading of the Supreme Court’s decision in Harrington v. Richter.14562 U.S. 86 (2011). Although Richter shows that, in some cases, a federal court must search for hypothetical reasons that “could have supported” the state‑court decision, applying that approach broadly overreads Richter. Rather, federal courts should understand a finding that no “argument[] or theor[y] . . . could have supported” the state‑court decision as one nonexclusive path to satisfying Section 2254(d)’s requirements.15See id. at 102. On some sets of facts—such as Richter itself—it is, as a practical matter, the only path. But transforming that practical requirement for some cases into a legal requirement for all cases turns Richter on its head. On some facts—such as the tarot card hypothetical—courts can and should travel another path.

The scope of the “could have supported” framework16Adapting the Supreme Court’s terminology from Wilson v. Sellers, 138 S. Ct. 1188 (2018), this Article refers to this “‘could have supported’ approach” as the “‘could have supported’ framework.” Id. at 1193, 1195. Pre‑Wilson cases and commentary use a variety of other terms to describe the same approach. See, e.g., Noam Biale, Beyond a Reasonable Disagreement: Judging Habeas Corpus, 83 U. Cin. L. Rev. 1337, 1366–67 (2015) (“‘ultimate result’ approach”); Adam N. Steinman, Reconceptualizing Federal Habeas Corpus for State Prisoners: How Should AEDPA’s Standard of Review Operate After Williams v. Taylor?, 2001 Wis. L. Rev. 1493, 1510 (“result‑deference”); Dennis v. Sec’y, Pa. Dep’t of Corr., 834 F.3d 263, 283–84 (3d Cir. 2016) (en banc) (“gap‑filling”). is “of considerable consequence.”17See Dennis, 834 F.3d at 356 n.12 (Jordan, J., concurring in part and concurring in the judgment). For one thing, it strikes at a central issue in federal habeas corpus law: AEDPA is “the most significant habeas reform since 1867,”18Kenneth Williams, The Antiterrorism and Effective Death Penalty Act: What’s Wrong with It and How to Fix It, 33 Conn. L. Rev. 919, 923 (2001). Section 2254(d) is its “centerpiece,”19John H. Blume, AEDPA: The “Hype” and the “Bite”, 91 Cornell L. Rev. 259, 272 (2006). and the key underlying question—the extent to which Section 2254(d) “target[s] the state court’s ‘opinion’”— “goes to the heart” of Section 2254(d).20Steinman, supra note 16, at 1495. For another, the framework’s applicability is outcome determinative in cases where (1) the underlying claim is meritorious, (2) the state court’s written opinion fails to meet even Section 2254(d)’s minimal standards, and (3) a hypothetical reasonable‑but‑erroneous rationale for denying the claim exists that would meet those standards.21Cf. Patrick J. Fuster, Taming Cerberus: The Beast at AEDPA’s Gates, 84 U. Chi. L. Rev. 1325, 1329–30 (2017) (“[T]he approach [is] outcome‑determinative given . . . conditions [(2) and (3)].”). Fuster excludes condition (1), the merits of the claim, as a condition of outcome‑determinativeness, arguing that “find[ing] § 2254(d) satisfied . . . in most cases dictates that relief will be granted.” Id. at 1329 (comma omitted). He acknowledges that “[a]fter bypassing § 2254(d), a petitioner technically must still prevail on de novo review,” but explains that “[t]he process of demonstrating that a state decision was unreasonable . . . will almost always include the lesser showing that it was incorrect.” Id. at 1329 n.32 (emphasis added). Although there is likely a strong correlation between a petitioner’s ability to satisfy Section 2254(d) and the merits of their claim, the two questions are analytically distinct. See infra Section III.A. Because that distinction has significant theoretical implications, it cannot simply be glossed over.

Though a trilogy of recent cases has prompted a wave of scholarship on the Great Writ’s nature, history, and purpose,22The three cases are Shinn v. Ramirez, 142 S. Ct. 1718 (2022), Brown v. Davenport, 142 S. Ct. 1510 (2022), and Edwards v. Vannoy, 141 S. Ct. 1547 (2021). For examples of recent scholarship on the Great Writ, see generally Lee Kovarsky, The New Negative Habeas Equity, 137 Harv. L. Rev. 2222 (2024); William M.M. Kamin, The Great Writ of Popular Sovereignty, 77 Stan. L. Rev. (forthcoming 2025); Micah S. Quigley, What Is Habeas?, 173 U. Pa. L. Rev. 453 (2025); Anthony G. Amsterdam & James S. Liebman, Loper Bright and the Great Writ, Colum. Hum. Rts. L. Rev. (forthcoming Feb. 2025); David Kinnaird, Habeas Corpus and Void Judgments, 100 Notre Dame L. Rev. (forthcoming 2025); Brandon L. Garrett & Kaitlin Phillips, AEDPA Repeal, 107 Cornell L. Rev. 1739 (2022); Leah M. Litman, The Myth of the Great Writ, 100 Tex. L. Rev. 219 (2021). neither the Supreme Court nor any scholar has squarely addressed the scope of Richter’s “could have supported” framework or its effect on the role of written state‑court opinions.23The article that asks this sort of question most directly predates Richter by nearly ten years. See Steinman, supra note 16, at 1510–11. Thus, it does not account for more than two decades of Section 2254(d) jurisprudence, including nearly all the key cases discussed in this Article. This Article aims to fill that void. In Part I, the Article provides the necessary background on the framework and its history. Section I.A spells out how courts apply the framework, while Section I.B places it within the broader historical context of federal relitigation, including both Richter and Wilson v. Sellers,24138 S. Ct. 1188 (2018). the first case to address (and limit, albeit to an unclear extent) the framework’s scope.

In Part II, this Article examines the difficulties courts and commentators face in delineating the framework’s proper scope. Section II.A explains the apparent tension between three lines of cases in which the Supreme Court applied Section 2254(d) to claims requiring a prisoner to prove multiple elements. Section II.B then examines the approaches lower courts have taken, explaining that, while each has some appeal, none fully resolves the tension in the Supreme Court’s Section 2254(d) jurisprudence.

In Part III, this Article begins to sketch a theory of the “could have supported” framework by first articulating a broader theory of Section 2254(d). In Section III.A, this Article explains that Section 2254(d) is not a modification of the merits inquiry, but a separate and independent requirement for relief. To overcome Section 2254(d), a prisoner must prove the state court committed some “qualifying error.” In Section III.B, this Article identifies the components of a qualifying error. Such errors must be of the right type, be of sufficient severity, and have some potential effect on the state court’s decision.

Finally, in Part IV, this Article argues that the “could have supported” framework is best understood as a path to proving that a qualifying error occurred. It is a function of (1) the nature of the qualifying‑error inquiry, (2) the burden of proof, and (3) the quantum of available evidence. It is not a freestanding requirement for relief. Instead, it is a type of evidence that may, in appropriate circumstances, support the inference that a qualifying error has occurred.

I.      The “Could Have Supported” Framework

In Richter, the Supreme Court laid out the following framework:

Under § 2254(d), a habeas court must determine what arguments or theories supported or, as here, could have supported, the state court’s decision; and then it must ask whether it is possible fairminded jurists could disagree that those arguments or theories are inconsistent with the holding in a prior decision of this Court.25Harrington v. Richter, 562 U.S. 86, 102 (2011) (emphasis added).

This Part lays the groundwork for a proper understanding of this framework and its scope. It first discusses the analysis that a court must conduct when applying the “could have supported” framework. Then, this Part examines the historical and factual context in which Richter and Wilson arose to see what light they shed on the framework’s scope.

A.      The “Could Have Supported” Analysis

As the above‑quoted language suggests, the “could have supported” framework proceeds in two steps.26See Biale, supra note 16, at 1367 (“Since Richter, some courts have adopted this approach and framed it as a two‑step inquiry . . . .”); Matthew Seligman, Note, Harrington’s Wake: Unanswered Questions on AEDPA’s Application to Summary Dispositions, 64 Stan. L. Rev. 469, 497 (2012) (“This more realistic, nonlinear model of review requires two stages of analysis.”). First, the habeas court must identify every possible “argument[] or theor[y]” that could justify the state court’s denial of the petitioner’s claim.27Richter, 562 U.S. at 102. Second, it must evaluate each argument and theory to see if any surmount Section 2254(d)’s low bar. Together, these steps require a state court to “craft a story that makes the state result justifiable.”28See Brittany Glidden, When the State Is Silent: An Analysis of AEDPA’s Adjudication Requirement, 27 N.Y.U. Rev. L. & Soc. Change 177, 189 (2001).

There is debate about the scope of a court’s duties in the first step. Judge Calabresi describes it as requiring a federal court to “imagine” potential rationales, including ones the state courts “never in fact espoused.”29Hawthorne v. Schneiderman, 695 F.3d 192, 199 (2d Cir. 2012) (Calabresi, J., concurring) (emphasis omitted). The Fifth Circuit understood its task in precisely this way, describing its perceived duty to “invent possible avenues the state court could have relied upon.”30Evans v. Davis, 875 F.3d 210, 217 (5th Cir. 2017); accord Hittson v. Chatman, 576 U.S. 1028, 1029–30 (2015) (Ginsburg, J., concurring in the denial of certiorari) (describing the Eleventh Circuit’s application of Richter as “hypothesiz[ing] reasons”); see also Grueninger v. Dir., Va. Dep’t of Corr., 813 F.3d 517, 526 (4th Cir. 2016); Walker v. McQuiggan, 656 F.3d 311, 318 (6th Cir. 2011); Torres v. Bauman, 677 F. App’x 300, 302 (6th Cir. 2017); Pye v. Warden, Ga. Diagnostic Prison, 50 F.4th 1025, 1074 n.31 (11th Cir. 2022) (en banc) (Pryor, J., dissenting); Wilson v. Warden, Ga. Diagnostic Prison, 834 F.3d 1227, 1245 (11th Cir. 2016) (en banc) (Jordan, J., dissenting), rev’d and remanded sub nom. Wilson v. Sellers, 138 S. Ct. 1188 (2018); Hedlund v. Ryan, 750 F.3d 793, 836 n.11 (9th Cir. 2014) (Wardlaw, J., concurring in part and dissenting in part), withdrawn, 815 F.3d 1233 (9th Cir. 2016); Montgomery v. Bobby, 654 F.3d 668, 700 (6th Cir. 2011) (Clay, J., dissenting). The Supreme Court’s description of the inquiry as requiring analysis of “hypothetical reasons” that a “state court might have given” supports this view.31Brumfield v. Cain, 576 U.S. 305, 323 (2015) (emphasis added) (describing Richter parenthetically). But Justice Gorsuch disputes that the framework requires so much, calling such descriptions “caricature[s].”32Wilson, 138 S. Ct. at 1199 (Gorsuch, J., dissenting). In his view, “a federal court generally isn’t required to imagine or hypothesize arguments that neither the parties before it nor any lower court has presented.”33Id. Instead, it need only consider the state court’s “opinion (if there is one), any argument presented by the parties in the state proceedings, and any argument presented in the federal habeas proceeding.”34Id.

Regardless of the particulars, the result will be a list of “arguments or theories [that] supported or . . . could have supported, the state court’s decision.”35See Harrington v. Richter, 562 U.S. 86, 102 (2011). The court must then ask “whether it is possible fairminded jurists could disagree that those arguments or theories are inconsistent with the holding in a prior decision of [the Supreme] Court”36Id.—a straightforward application of the normal Section 2254(d) analysis to a particular argument.37See, e.g., Yarborough v. Alvarado, 541 U.S. 652, 664 (2004) (applying the “fairminded jurists could disagree” standard before Richter’s creation of the “could have supported” framework). There is a good argument that Richter’s “fairminded jurist” standard did break new ground. See Biale, supra note 16, at 1339, 1351–60; Judith L. Ritter, The Voice of ReasonWhy Recent Judicial Interpretations of the Antiterrorism and Effective Death Penalty Act’s Restrictions on Habeas Corpus Are Wrong, 37 Seattle U. L. Rev. 55, 68 (2013) (“Richter represents a shift in unreasonable application analysis.”). The Court had previously held that “[d]efining an ‘unreasonable application’ by reference to a ‘reasonable jurist’ . . . is of little assistance to the courts that must apply § 2254(d)(1) and, in fact, may be misleading.” Williams v. Taylor, 529 U.S. 362, 409 (2000) (majority opinion of O’Connor, J.); see also Brandon L. Garrett, Constitutional Reasonableness, 102 Minn. L. Rev. 61, 83–84 (2017). However, to the extent this was a new development, it does not appear to have been tied to the “could have supported” framework, as the Court has applied it in cases that did not invoke the framework—including at least one case that seemed to pointedly omit the “could have supported” language. See Wetzel v. Lambert, 565 U.S. 520, 524 (2012) (per curiam) (using an ellipsis to remove “could have supported” language from its Richter quotation). Taken together, the two steps require that, if any “plausible argument exists to support the ruling, [the court] defer[s].”38Sheppard v. Davis, 967 F.3d 458, 467 (5th Cir. 2020).

1.      Double Hypothesizing

Because habeas petitioners raise claims of ineffective assistance of counsel under Strickland v. Washington39466 U.S. 668 (1984). more often than any other claim (and because they are used below to illustrate some of the tensions in Section 2254(d) jurisprudence40See infra notes 169–77 and accompanying text.), it is worth pausing to consider how this framework applies to such claims.41See, e.g., Justin F. Marceau, Embracing a New Era of Ineffective Assistance of Counsel, 14 U. Pa. J. Const. L. 1161, 1161–62 (2012). One element of a Strickland claim is deficient performance.42Strickland, 466 U.S. at 687 (“First, the defendant must show that counsel’s performance was deficient.”). Not just any attorney error will do; the errors must be “so serious that counsel was not functioning as the ‘counsel’ guaranteed the defendant by the Sixth Amendment.”43Id. Similarly, Section 2254(d) requires more than an “ordinary error”; the writ may issue only “where there is no possibility fairminded jurists could” agree with the state court’s decision.44Harrington v. Richter, 562 U.S. 86, 102 (2011). Both standards are “highly deferential,”45Id. at 105 (first quoting Strickland, 466 U.S. at 689; and then quoting Lindh v. Murphy, 521 U.S. 320, 333 n.7 (1997)). and when combined they create what has been called “double deference”46See, e.g., Rogers v. Mays, 69 F.4th 381, 389 (6th Cir. 2023) (“[W]hen we apply the highly deferential AEDPA standard to the already deferential Strickland standard, we give the state‑court decision double deference.” (citing Cullen v. Pinholster, 563 U.S. 170, 190 (2011)). Although the Supreme Court has never used the term “double deference”—perhaps recognizing that what Section 2254(d) requires is not truly “deference,” see infra notes 298–300 and accompanying text—it has repeatedly described review as “doubly deferential.” E.g., Dunn v. Reeves, 141 S. Ct. 2405, 2410 (2021) (quoting Burt v. Titlow, 571 U.S. 12, 15 (2013)); Yarborough v. Gentry, 540 U.S. 1, 6 (2003) (per curiam).—the federal court must defer to the state court, which must in turn defer to counsel.47See Titlow, 571 U.S. at 15 (noting that double deference “gives both the state court and the defense attorney the benefit of the doubt” (citing Cullen, 563 U.S. at 190)). Similar “double deference” arises as to other claims (such as evidentiary sufficiency, discrimination in jury selection, and some double jeopardy claims) that require deference even when raised on direct appeal. See Daniel J. O’Brien, Heeding Congress’s Message: The United States Supreme Court Bars Federal Courthouse Doors to Habeas Relief Against All but Irrational State Court Decisions, and Oftentimes Doubly So, 24 Fed. Sent’g Rep. 320, 320 (2012).

Applying the “could have supported” framework leads to not only double deference, but also double hypothesizing. Because Strickland set out an objective standard, some courts ask only how “some reasonable lawyer could have conducted the trial,” rather than counsel’s actual reasons.48Chandler v. United States, 218 F.3d 1305, 1315 n.16 (11th Cir. 2000) (en banc). When combined with the “could have supported” framework, this requires federal courts to hypothesize what a state court could have hypothesized about how a reasonable lawyer could have conducted the trial.

Take the Eleventh Circuit’s decision in Hammond v. Hall,49586 F.3d 1289 (11th Cir. 2009). for instance. In Hammond, the prosecutor argued—in violation of a Georgia statute—that the defendant should not be given a life sentence because he could one day be paroled.50Id. at 1328–29. Under the Georgia statute, this entitled the defendant to an automatic mistrial.51Id. at 1329. Trial counsel, however, did not request a mistrial.52Id. at 1328. He later explained that he failed to do so for a very simple reason: He did not know his client was entitled to one.53Id. at 1329. The state court, in turn, had written a sixty‑seven page opinion explaining its conclusion that counsel’s performance was not deficient.54Id. at 1304–05. But rather than assess the reasonableness of the actual reasons provided by trial counsel and the state court, the Eleventh Circuit instead hypothesized four rationales that the state court could have concluded could have motivated trial counsel.55Hammond, 586 F.3d at 1333–34.

As Hammond illustrates, applying the “could have supported” framework to Strickland claims produces an even more powerful form of deference.

B.      The Framework in Context

A firm understanding of the “could have supported” framework and its scope requires not just a grasp of the framework itself, but also an understanding of the context—both legal and factual—that gave rise to it. In this Section, this Article examines (1) the pre‑AEDPA regime governing relitigation of claims previously rejected by state courts, (2) the adoption and early interpretations of Section 2254(d), and (3) the Court’s jurisprudence of the interplay between Section 2254(d) and unexplained state‑court decisions.

1.      The Pre‑AEDPA Regime

AEDPA does not exist in a vacuum. Rather, it is part of the broader scheme of statutory and judge‑made law that governs federal habeas. Understanding the change Section 2254(d) wrought requires first understanding the regime and debates that predated it.

a.      Brown v. Allen and Bator’s Critique

The scope of post‑conviction habeas before 1953 is hotly debated among both scholars56See Ann Woolhandler, Demodeling Habeas, 45 Stan. L. Rev. 575, 582–87 (1993) (comparing the two predominant views); James S. Liebman, Apocalypse Next Time?: The Anachronistic Attack on Habeas Corpus/Direct Review Parity, 92 Colum. L. Rev. 1997, 2055–94 (1992) [hereinafter Liebman, Apocalypse Next Time?] (offering a third, distinct view). The two most influential views are commonly associated with Paul Bator and Gary Peller, respectively. See Paul M. Bator, Finality in Criminal Law and Federal Habeas Corpus for State Prisoners, 76 Harv. L. Rev. 441 (1963); Gary Peller, In Defense of Federal Habeas Corpus Relitigation, 16 Harv. C.R.-C.L. L. Rev. 579 (1982). and jurists.57Compare Brown v. Davenport, 142 S. Ct. 1510, 1520–22 (2022) (Justice Gorsuch, joined by Chief Justice Roberts and Justices Thomas, Alito, Kavanaugh, and Barrett, presenting a narrow view of pre‑1953 habeas), and Wright v. West, 505 U.S. 277, 285 (1992) (plurality opinion) (Justice Thomas, joined by Chief Justice Rehnquist and Justice Scalia, presenting the same narrow view), and Fay v. Noia, 372 U.S. 391, 449–60 (1963) (Harlan, J., dissenting) (Justice Harlan, joined by Justices Clark and Stewart, presenting the same narrow view), with Davenport, 142 S. Ct. at 1531–35 (Kagan, J., dissenting) (Justice Kagan, joined by Justices Breyer and Sotomayor, presenting a broad view of pre‑1953 habeas), and West, 505 U.S. at 297–99 (O’Connor, J., concurring in the judgment) (Justice O’Connor, joined by Justices Blackmun and Stevens, presenting the same broad view), and Noia, 372 U.S. at 415–24 (Justice Brennan, joined by Chief Justice Warren and Justices Black, Douglas, White, and Goldberg, presenting the same broad view). But no one disputes that the Supreme Court’s decision in Brown v. Allen58344 U.S. 443 (1953). embodied “a regime of broad federal relitigation” of claims rejected by state courts.59Richard H. Fallon, Jr., John F. Manning, Daniel J. Meltzer & David L. Shapiro, Hart and Wechsler’s the Federal Courts and the Federal System 1274 (7th ed. 2015) [hereinafter Hart & Wechsler]. In Brown, the Court held that, while a federal habeas court could “look to the State proceedings for whatever light they shed on the historical facts,” it was “for the federal judge to assess” the claim’s legal merits, without giving “binding weight” to the state court’s decision.60Brown, 344 U.S. at 507–08 (opinion of Frankfurter, J.) (outlining a generalized six-step approach to a habeas case and considered a second majority opinion, alternatively known as Daniels v. Allen). Together with decisions recognizing broad authority for federal courts to hold their own evidentiary hearings,61See Townsend v. Sain, 372 U.S. 293, 310–12 (1963). decide claims not properly raised in state court,62See Noia, 372 U.S. at 428–34. and consider habeas petitions from prisoners who had already been denied relief in federal court,63See Sanders v. United States, 373 U.S. 1, 2–23 (1963). Brown produced a “Golden Age” of federal habeas.64See Jordan Steiker, Restructuring Post‑Conviction Review of Federal Constitutional Claims Raised by State Prisoners: Confronting the New Face of Excessive Proceduralism, 1998 U. Chi. Legal F. 315, 323–24 (1998).

That Golden Age was, like other parts of the Warren Court’s criminal docket, the subject of withering criticism from some corners.65See Liebman, Apocalypse Next Time?, supra note 56, at 2028 (“Brown met intense scrutiny, not only in Congress and in the lower federal courts, but also in the writings of the Federal Jurisdiction titans of the 1950s and 1960s.”). The most prominent critic was Professor Paul Bator, who argued in a 1963 article—acknowledged even by its detractors as a “tour de force66Id. at 1999, 2041–43 (describing “Professor Bator’s flawed tour de force”).—that federal review should focus not on the underlying merits of the petitioner’s claim, but rather on the adequacy of the state court’s processes for evaluating the claim.67See Bator, supra note 56, at 452–53. Bator’s argument was in part historical, critiquing Brown as a departure from previous practice.68Id. at 463 (“[I]t is at most doubtful whether any such principle [as federal courts re-opening final criminal judgments issued by competent state courts] existed before Brown v. Allen established it in 1952.”). But it was also part epistemological. Bator emphasized that, even “[a]ssuming that there ‘exists,’ in an ultimate sense, a ‘correct’ decision of a question of law, we can never be assured that any particular tribunal has in the past made it.”69Id. at 447. Therefore, he argued, “the notion of legality must at some point include the assignment of final competences to determine legality.”70Id. at 450–51. Thus, he concluded that, where a court with jurisdiction employed “processes fairly and rationally adapted to” deciding constitutional claims and ultimately denied relief, that denial should be final absent “functional or institutional requirements” to the contrary.71Id. at 462.

b.      Teague v. Lane and Debate over Its Scope

There is debate over whether Brown itself required de novo review of claims that were previously rejected by state courts.72Compare Wright v. West, 505 U.S. 277, 287 (1992) (plurality opinion) (arguing that Brown “had no occasion to explore in detail the question whether a ‘satisfactory’ conclusion was one that the habeas court considered correct, as opposed to merely reasonable”), with Liebman, Apocalypse Next Time?, supra note 56, at 2019–29 (arguing that Brown required de novo review). Regardless, de novo review became the norm in Brown’s aftermath.73West, 505 U.S. at 301–03 (O’Connor, J., concurring in the judgment) (collecting cases); see Wainwright v. Sykes, 433 U.S. 72, 87 (1977) (“[S]ince Brown v. Allen, it has been the rule that the federal habeas petitioner who claims he is detained pursuant to a final judgment of a state court in violation of the United States Constitution is entitled to have the federal habeas court make its own independent determination of his federal claim, without being bound by the determination on the merits of that claim reached in the state proceedings.” (citation omitted)). The Court breathed new life into arguments for more deferential review, however, when in Teague v. Lane74489 U.S. 288 (1989) (plurality opinion). it revamped its doctrine regarding the application and development of “new” rules of criminal procedure.75See id. at 310 (opinion of O’Connor, J.). Previous retroactivity doctrine relied on a multi‑factor balancing test that focused primarily on the nature of the rule in question.76See Stovall v. Denno, 388 U.S. 293, 296–97 (1967); Linkletter v. Walker, 381 U.S. 618, 629 (1965). Though Teague rejected the LinkletterStovall analysis for federal habeas, it still controls in some states’ post‑conviction proceedings. See Josiah Rutledge, With Great (Writ) Power Comes Great (Writ) Responsibility: A Modified Teague Framework for State Courts, 59 Crim. L. Bull. 480, 487–89 (2023). The test considered the reliance interests of “law enforcement authorities,” but not of state courts.77See Stovall, 388 U.S. at 297–98. One possible explanation for this emphasis is that both Linkletter and Stovall addressed the retroactivity of rules aimed at police. Linkletter held that the Fourth Amendment exclusionary rule of Mapp v. Ohio, 367 U.S. 643 (1961), did not apply retroactively. Linkletter, 381 U.S. at 639–40. Stovall held the same for the right to counsel at lineups under United States v. Wade, 388 U.S. 218 (1967) and Gilbert v. California, 388 U.S. 263 (1967). Stovall, 388 U.S. at 300. Thus, while Teague is in some respects akin to Section 2254(d), LinkletterStovall was more akin to the Court’s holding in Davis v. United States, 564 U.S. 229 (2011), that “searches conducted in objectively reasonable reliance on binding appellate precedent are not subject to the exclusionary rule.” Davis, 564 U.S. at 232. Under Teague, by contrast, the focus shifted to the position of the state court that reviewed (or would have reviewed) the petitioner’s claim at the time. The primary question is now whether the rule in question was “new.”78Teague, 489 U.S. at 310 (opinion of O’Connor, J.).

In subsequent cases, the Court defined a “new” rule using language strikingly similar to what it would later use in Richter: A new rule is one that would not have been “apparent to all reasonable jurists” at the time the conviction became final.79Lambrix v. Singletary, 520 U.S. 518, 527–28 (1997). The Court explained that the purpose of such a capacious definition was to “validate reasonable interpretations of existing precedents” by state courts.80Stringer v. Black, 503 U.S. 222, 237 (1992); see also Butler v. McKellar, 494 U.S. 407, 414 (1990) (“The ‘new rule’ principle therefore validates reasonable, good‑faith interpretations of existing precedents made by state courts even though they are shown to be contrary to later decisions.” (citing United States v. Leon, 468 U.S. 897, 918–19 (1984)). This explanation suggested that the emphasis had now shifted to the decisions of state courts—vaguely reminiscent of a standard of review. There are, however, problems with viewing Teague as a new standard of review. For one, Teague applies even when there has not been a state‑court decision on the merits.81Peter Bozzo, What We Talk About When We Talk About Retroactivity, 46 Am. J. Crim. L. 13, 23 (2019) (“In addition, if a state court never ‘adjudicated’ a federal habeas petitioner’s claims ‘on the merits,’ . . . Teague could still bar relief.”). For another, it requires application of some rules that were articulated after the last state‑court decision, such as those articulated while a petition for certiorari is pending.82See Caspari v. Bohlen, 510 U.S. 383, 390 (1994) (stating that a conviction does not become final until “the time for filing a petition for a writ of certiorari has elapsed or a timely filed petition has been finally denied” (citing Griffith v. Kentucky, 479 U.S. 314, 321 n.6 (1987)). Thus, although Teague’s rationale focuses on state courts, its analysis remains focused on federal precedent.83See Stringer, 503 U.S. at 237 (“[T]he ultimate decision whether Clemons [v. Mississippi, 494 U.S. 738 (1990)] was dictated by precedent is based on an objective reading of the relevant cases.” (emphasis added)); cf. Beard v. Banks, 542 U.S. 406, 413 (2004) (“[T]he Teague principle protects not only the reasonable judgments of state courts but also the States’ interest in finality quite apart from their courts.”).

Nevertheless, debates over Teague’s scope persisted, bubbling up to the Supreme Court in Wright v. West.84505 U.S. 277 (1992) (plurality opinion); id. at 291–94. Writing for a plurality, Justice Thomas maintained that Teague “implicitly questioned” the de novo standard of review by requiring a federal habeas court to “defer to the state court’s decision rejecting the claim unless that decision is patently unreasonable.”85Id. at 291 (internal quotation marks omitted) (quoting Butler, 494 U.S. at 422 (Brennan, J., dissenting)); accord Larry W. Yackle, Federal Courts: Habeas Corpus 120 (2003) (“Teague has always been a contrivance for forcing federal habeas courts to ‘defer’ to previous state court judgments on the merits.”). He acknowledged, however, that Teague was “not directly controlling” as to mixed questions of law and fact.86West, 505 U.S. at 294 (plurality opinion). In her concurrence, Justice O’Connor—Teague’s author—criticized Justice Thomas for “mischaracteriz[ing] Teague,” which in her view “is not the same as deference” and “did not establish a standard of review at all.”87Id. at 303–05 (O’Connor, J., concurring in the judgment). Justice Kennedy agreed, writing that “it would be a misreading of Teague to interpret it as resting on the necessity to defer to state‑court determinations.”88Id. at 307 (Kennedy, J., concurring in the judgment). Three years later, the Court in Thompson v. Keohane89516 U.S. 99 (1995). put some of West’s potential implications to rest, reaffirming the Court’s pre‑West holding that mixed questions of law and fact were subject to de novo review in habeas.90See id. at 112–13 (pertaining to criminal interrogation confessions); Miller v. Fenton, 474 U.S. 104, 112 (1985) (same).

On the eve of AEDPA, therefore, the scope of federal habeas review was “nearly identical” to the review exercised by the Supreme Court on direct appeal.91Liebman, Apocalypse Next Time?, supra note 56, at 1998. Courts reviewed questions of fact deferentially and questions of law and mixed questions de novo, subject to Teague’s nonretroactivity rule (a rule that reinforced, rather than undermined, parity between habeas and direct appeal92Id. at 2095–96.).93Id. at 2003–05.

2.      The Dawn of AEDPA

Having reached a “stalemate” in the courts, advocates for deferential review turned back to Congress.94See Kent S. Scheidegger, Habeas Corpus, Relitigation, and the Legislative Power, 98 Colum. L. Rev. 888, 889–90 (1998). Though previous calls for legislative intervention had been rebuffed,95Id. at 890 & nn.13–14; see also Larry W. Yackle, A Primer on the New Habeas Corpus Statute, 44 Buff. L. Rev. 381, 423–32 (1996) (tracing proposed legislation in the lead‑up to AEDPA). a great national tragedy would propel this one to success. In the aftermath of the Oklahoma City bombing, calls for new legislation finally prevailed.96Bryan A. Stevenson, The Politics of Fear and Death: Successive Problems in Capital Federal Habeas Corpus Cases, 77 N.Y.U. L. Rev. 699, 701 (2002); Williams, supra note 18, at 923; James S. Liebman, An “Effective Death Penalty”? AEDPA and Error Detection in Capital Cases, 67 Brook. L. Rev. 411, 412–13 (2001) [hereinafter Liebman, An “Effective Death Penalty”?]; see also Blume, supra note 19, at 270. But the statute is far from clear, and it is widely panned for its poor draftsmanship.97E.g., Steinman, supra note 16, at 1528; Blume, supra note 19, at 261; Yackle, supra note 85, at 57; Liebman, An “Effective Death Penalty”?, supra note 96, at 426; Yackle, supra note 95, at 381. As the Court put it, “in a world of silk purses and pigs’ ears, the Act is not a silk purse of the art of statutory drafting.”98Lindh v. Murphy, 521 U.S. 320, 336 (1997). Justice Sotomayor once opined that “[t]here’s nothing logical about this statute, or clear about this statute.”99Transcript of Oral Argument at 16, Cullen v. Pinholster, 563 U.S. 170 (2011) (No. 09‑1088). Justice Scalia apparently agreed, asking “who’s responsible for writing this[?]”100Transcript of Oral Argument at 14, Dodd v. United States, 545 U.S. 353 (2005) (No. 04‑5286). The legislative debates and reports had little to say about Section 2254(d), meaning that the legislative history is “unilluminating,”101Blume, supra note 19, at 273. even for those inclined to look to it.102But see Lindh v. Murphy, 96 F.3d 856, 868 (7th Cir. 1996) (en banc) (explaining that AEDPA’s House Conference Report “is a wonderful illustration why legislative history so often misleads”), rev’d and remanded on other grounds, 521 U.S. 320 (1997).

Larry Yackle nicely summarized how decades of debate culminated in such an enigmatic text:

The new law is not well drafted. It bears the influence of various bills that were fiercely debated for nearly forty years. Along the way, proponents of habeas legislation adjusted their initiatives in light of contemporaneous events and circumstances: the Powell Committee Report in 1989, for example, as well as shifting levels of political support for particular measures and new Supreme Court decisions on point. Proponents often kept abreast of the times by adding new elements to their bills without, at the same time, reexamining old formulations in order to maintain an intellectually coherent whole. The result, I am afraid, is extraordinarily arcane verbiage that will require considerable time and resources to sort out.103Yackle, supra note 95, at 381.

Courts were thus left with the difficult task of liquidating Section 2254(d)’s precise meaning.104Cf. The Federalist No. 37, at 236 (James Madison) (Jacob E. Cooke ed., 1982) (“All new laws . . . are considered as more or less obscure and equivocal, until their meaning be liquidated and ascertained by a series of particular discussions and adjudications.”). Unsurprisingly, lower courts did so in “sharply divergent” ways.1052 Randy Hertz & James S. Liebman, Federal Habeas Corpus Practice and Procedure § 32.3, at 1783 (6th ed. 2011); see also Biale, supra note 16, at 1351 (“Following the passage of AEDPA, interpretation of the meaning of ‘unreasonable application’ quickly produced a circuit split.”). One common interpretation viewed Section 2254(d)’s three clauses as effectively establishing appellate standards of review—with the “contrary to” clause governing pure questions of law, the “unreasonable application” clause governing mixed questions, and paragraph (d)(2) governing pure questions of fact.106See 2 Hertz & Liebman, supra note 105, § 32.3, at 1783 (attributing this interpretation of Section 2254(d)(1)’s “contrary to” and “unreasonable application” clauses to the Fifth, Seventh, and Eleventh Circuits). Some courts also viewed an “unreasonable application” as one “that all reasonable jurists would agree is incorrect.”107Biale, supra note 16, at 1351–52 (attributing this view to the Fourth, Fifth, and Eleventh Circuits). At least two circuits—the Fifth and the Eleventh—adopted both positions.108See Drinkard v. Johnson, 97 F.3d 751, 767–68 (5th Cir. 1996) (“Subsection (d)(2) thus supplies the applicable standard of review for the second type of question—a question of fact. . . . The second clause of subsection (d)(1), by its own language, refers to mixed questions of law and fact . . . . We read the first clause, on the other hand, as referring to questions of law.”); id. at 769 (“[W]e can grant habeas relief only if a state court decision is so clearly incorrect that it would not be debatable among reasonable jurists.”); Neelley v. Nagle, 138 F.3d 917, 923–24 (11th Cir. 1998) (“Giving the phrase ‘contrary to’ its plain meaning, we can readily think of two situations in which a state court decision would be ‘contrary to’ clearly established Supreme Court case law. . . . Both of these types of errors are errors of pure law . . . . By its very language, ‘unreasonable application’ refers to mixed questions of law and fact . . . .”); id. at 924 (noting the Fifth Circuit’s “reasonable jurists” formulation and “adopt[ing] the Fifth Circuit’s standard”).

The Supreme Court was mostly silent during this period. Although it granted certiorari in the Seventh Circuit’s key Section 2254(d) case, it decided only that the statute’s limitations on relief did not apply to cases filed before AEDPA was enacted.109Lindh v. Murphy, 521 U.S. 320, 322–23 (1997). The only substantive comment was a footnote observing that Section 2254(d) set forth a “new, highly deferential standard for evaluating state‑court rulings.”110See id. at 333 n.7. The dissenters agreed that Section 2254(d) “simply alters the standard under which . . . prior judgment[s are] evaluated.” Id. at 342 (Rehnquist, C.J., dissenting). Until the turn of the millennium, that was all the Court had to say.

As cases filed after AEDPA’s effective date moved their way through the judiciary, however, things began to heat up in the Court’s nascent Section 2254(d) jurisprudence. In January of 2000, the Court decided Weeks v. Angelone,111528 U.S. 225 (2000). the first case to invoke the new limitations on relief. Weeks established (without explanation) one potentially important doctrinal rule—that, if a state court decision is correct, “it follows a fortiori that the adjudication of the [state court] . . . neither was ‘contrary to,’ nor involved an ‘unreasonable application of,’” clearly established law.112Id. at 237. The implications of this “a fortiori doctrine” are discussed below.113See infra text accompanying notes 318–22. But its application in Weeks prevented the Court from considering when Section 2254(d) might bar relief to which a prisoner was otherwise entitled. Justice Souter did consider that possibility, though, in his dissent in another case decided that day.114See Smith v. Robbins, 528 U.S. 259, 303 (2000) (Souter, J., dissenting). Writing for himself and three others, he concluded that neither Section 2254(d) nor Teague precluded relief there because the rule at issue was clearly established, various state‑court holdings to the contrary notwithstanding.115Id. at 302–03. The majority in that case—unlike the Weeks majority—saw no need to mention Section 2254(d) at all, because it found no constitutional violation.116See id. at 265 (majority opinion).

The Section 2254(d) invocations in these cases proved only an appetizer for the blockbuster decision three months later in Williams v. Taylor,117529 U.S. 362 (2000). the Court’s “foundational” Section 2254(d) case.118Seligman, supra note 26, at 471. Like Brown before it,119See Liebman, Apocalypse Next Time?, supra note 56, at 2020 (“Brown has two majority opinions—one by Justice Reed, another by Justice Frankfurter.”). Williams produced two majority opinions—the portion of Justice O’Connor’s opinion that construed Section 2254(d) in the abstract and the portion of Justice Stevens’s opinion that applied it to the case at hand.120See Williams, 529 U.S. at 367, 399 (“Justice Stevens announced the judgment of the Court and delivered the opinion of the Court with respect to Parts I, III, and IV . . . . Justice O’Connor delivered the opinion of the Court with respect to Part II . . . .”). Justice O’Connor insisted on according “independent meaning to both the ‘contrary to’ and ‘unreasonable application’ clauses of the statute.”121Id. at 404 (majority opinion of O’Connor, J.) (criticizing Justice Stevens for failing to make this distinction). But in lieu of the “appellate standards of review” approach, she largely adopted the Fourth Circuit’s framework, which viewed the separate clauses of Section 2254(d) not as different inquiries for different situations, but as independent paths to overcoming the limitations on relief.122See id. at 405–07. She explained that a decision is “contrary to” clearly established law if it applies a legal rule that “contradicts” Supreme Court precedent or reaches a different result on “materially indistinguishable” facts.123Id. at 405–06. By contrast, a decision is an “unreasonable application” of clearly established law if it “unreasonably applies” the governing legal rule to the facts or if it extends (or fails to extend) precedent in unreasonable ways.124Id. at 407. In a later case, the Court dispensed with the “failure to extend” theory. See White v. Woodall, 572 U.S. 415, 426 (2014). If the state decision fell into any of these categories, federal review was “unconstrained by” Section 2254(d).125See Williams, 529 U.S. at 406 (majority opinion of O’Connor, J.).

She rejected, however, the Fourth Circuit’s application of the “all reasonable jurists” standard.126See id. at 409–10. Relying on Drinkard, the Fourth Circuit had adopted this test in Green v. French, 143 F.3d 865 (4th Cir. 1998), its own key Section 2254(d) case. See Green, 143 F.3d at 870 (“In other words, habeas relief is authorized only when the state courts have decided the question by interpreting or applying the relevant precedent in a manner that reasonable jurists would all agree is unreasonable.”). Although she found it “difficult to fault” lower courts for adopting “nearly identical terminology” to what the Court itself had employed under Teague, she emphasized the objective nature of the reasonableness standard.127Williams, 529 U.S. at 409–10 (majority opinion of O’Connor, J.). In her view, Congress drew the same distinction between “unreasonable” and “incorrect” that she and Justice Thomas had each drawn in West, siding with Justice Thomas on the implications of that distinction.128Id. at 410–11 (emphasis omitted). Williams thus made clear that Section 2254(d) “is a significant change from Brown.”129Steinman, supra note 16, at 1507. It did not, however, explore the unreasonable–incorrect dichotomy “in any helpful detail.”130Jordan Steiker, Habeas Exceptionalism, 78 Tex. L. Rev. 1703, 1728 (2000).

3.      The Unexplained Decisions Cases

For nearly a decade after Williams, explanations accompanied the state‑court determinations at issue in the Court’s Section 2254(d) cases.131See Amicus Curiae Brief of National Association of Criminal Defense Lawyers in Support of Respondent at 7, Knowles v. Mirzayance, 556 U.S. 111 (2009) (No. 07‑1315), 2008 WL 4580043, at *7 [hereinafter NACDL Mirzayance Brief] (noting that the Supreme Court had never decided a case involving both Section 2254(d) and a state court summary denial). That arrangement couldn’t last long, however. During this time, the California Supreme Court—the highest court in the most populous state in the union—was disposing of the vast majority (97%) of habeas petitions in unexplained summary denials,132See Seligman, supra note 26, at 506 & tbl.3 (compiling summary disposition data for state habeas petitions in the courts of appeal of California from 2006 to 2009). commonly known as “postcard denials.”133See Harris v. Superior Ct., 500 F.2d 1124, 1125 (9th Cir. 1974). The moniker comes from the California Supreme Court’s former practice of mailing such denials to prisoners on postcards. Id. That practice was destined to collide with Section 2254(d) jurisprudence.

a.      Knowles v. Mirzayance: A Crisis Delayed

The Court dodged that collision, however, in Knowles v. Mirzayance,134556 U.S. 111 (2009). the first summary denial case to reach the Court under Section 2254(d). After Alexandre Mirzayance was convicted of first‑degree murder, he raised an ineffective assistance of counsel claim in the California courts, which denied it without comment.135Id. at 114–15; NACDL Mirzayance Brief, supra note 131, at 6–7. When the case reached the Supreme Court, the National Association of Criminal Defense Lawyers (“NACDL”) flagged the potential difficulties with applying Section 2254(d) to such denials, urging the Court in an amicus brief to instead conclude that postcard denials do not constitute adjudications on the merits and thus are not covered by Section 2254(d).136See NACDL Mirzayance Brief, supra note 131, at 18–23. Perhaps heeding NACDL’s warnings, the Court dodged the issue, applying Section 2254(d) on the ground that Mirzayance had abandoned the issue.137Mirzayance, 556 U.S. at 121 n.2. The Court covered its bases, however, separately explaining why Mirzayance’s constitutional rights had not been violated,138Id. at 123–28. meaning he was not entitled to relief even if Section 2254(d) did not apply. NACDL’s warnings seemingly caught the attention of three Justices, however: Justices Scalia, Souter, and Ginsburg each declined to join the portion of the Court’s opinion applying Section 2254(d), instead joining only the portion explaining that Mirzayance’s claims lacked merit.139Id. at 113 (syllabus).

b.      Harrington v. Richter: Evaluating “Postcard Denials”

Just months after Mirzayance, the Court granted certiorari in Richter.140Harrington v. Richter, 559 U.S. 935 (2010) (granting certiorari). In addition to the question presented in the petition for certiorari, the Court asked the parties to brief and argue the postcard denial issue.141Id. Like Mirzayance, Joshua Richter raised an ineffective assistance claim in the California state courts, which rejected it without comment.142Harrington v. Richter, 562 U.S. 86, 96 (2011). The Court first considered whether such a denial is an adjudication on the merits that triggers Section 2254(d)’s limitations, concluding that they should be treated as such “in the absence of any indication or state‑law procedural principles to the contrary.”143Id. at 99 (citing Harris v. Reed, 489 U.S. 255, 265 (1989)); see Biale, supra note 16, at 1349 (describing this as “the main holding of Richter”).

In reaching that conclusion, the Court rejected NACDL’s warnings that unexplained orders do not provide federal courts “the resources necessary to make the assessments the statute requires.”144Amicus Curiae Brief of National Association of Criminal Defense Lawyers in Support of Respondent at 17–18, Richter, 562 U.S. 86 (2011) (No. 09-587), 2010 WL 2811206, at *17–18; see also NACDL Mirzayance Brief, supra note 131, at 18 (“The inquiries necessitated by the plain language of the statute cannot be made when a federal court has only a summary state court denial with which to work.”); Steinman, supra note 16, at 1516–17 (“As a practical matter, then, a state court that withholds its legal reasoning would thwart the very review that an opinion‑deference reading of § 2254(d)(1) would require . . . .”). Having rejected those warnings, the Court needed to answer the logical follow‑on question: How can a habeas petitioner show that the state court’s decision was sufficiently unreasonable without resort to the usual tool (i.e., the state court’s opinion)? According to the Court, a petitioner could do so only “by showing there was no reasonable basis for the state court to deny relief.”145Richter, 562 U.S. at 98. That burden could be met, the Court explained, only by satisfying the “could have supported” framework.146Id. at 102. Applying that framework, the Court denied relief.147Id. at 113.

c.      Hittson v. Chatman: A Rejoinder

As commentators recognized, Richter “raised more questions than it answered”148Eliza Beeney, Note, Why Silence Shouldn’t Speak So Loudly: Wiggins in a Post‑Richter World, 101 Cornell L. Rev. 1321, 1331 (2016).—including questions about the scope of the newly minted “could have supported” framework.149Other questions included, for instance, the applicability of the Richter presumption when the state court explains its decisions on some federal claims but does not mention others raised by the petition. See Johnson v. Williams, 568 U.S. 289, 298 (2013) (holding that the presumption applies). The first such question to reach the Court was how to handle cases where the last state‑court decision was an unexplained order, but a previous state court had issued an opinion explaining its own denial of the claim. That issue was first addressed in Justice Ginsburg’s concurrence in the denial of certiorari in Hittson v. Chatman.150576 U.S. 1028, 1028 (2015) (Ginsburg, J., concurring in denial of certiorari). In Hittson, the trial court denied the petitioner’s federal claims in a reasoned opinion, but the Georgia Supreme Court denied them without explanation.151See Hittson v. GDCP Warden, 759 F.3d 1210, 1228–29, 1232 (11th Cir. 2014). Relying on Richter, the Eleventh Circuit emphasized it was “not reviewing the reasoning announced by” the trial court, but rather reviewing the summary decision of the Georgia Supreme Court using the “could have supported” framework.152Id. at 1232 & n.25, 1233 n.26. As the concurring judge explained, whether the trial court’s rationale cleared Section 2254(d) was “irrelevant.”153Id. at 1273 (Carnes, C.J., concurring).

In Justice Ginsburg’s view, “[t]he Eleventh Circuit plainly erred” in applying the “could have supported” framework rather than the “look‑through” approach that the Court had previously established in the procedural default context, under which a federal court presumes a silent state court adopted the rationale of a previous state court that had also denied the claim.154Hittson, 576 U.S. at 1029 (Ginsburg, J., concurring in denial of certiorari); see Ylst v. Nunnemaker, 501 U.S. 797, 803 (1991). She distinguished such situations from the situation in Richter, where “no state court” issued an explanation, thus making “Richter’s hypothetical inquiry . . . necessary.”155Hittson, 576 U.S. at 1030 (citing Harrington v. Richter, 562 U.S. 86, 98 (2011)). Nevertheless, she agreed with the decision to deny certiorari, in part because a petition for rehearing en banc then pending in another case would “afford[] the Eleventh Circuit an opportunity to correct its error.”156Id. at 1030–31.

d.      Wilson v. Sellers: The Look‑Through Rule

Perhaps swayed by Justice Ginsburg’s rejoinder,157For a discussion of the Hittson concurrence as a potential “Supreme Court signal” to lower courts, see Richard M. Re, Another Supreme Court Signal: Hittson v. Chatman, PrawfsBlawg (June 15, 2015, 11:01 AM), https://perma.cc/8U9N-SLGX. the Eleventh Circuit granted the petition for en banc review, and Georgia “changed its position,” arguing that the Eleventh Circuit should evaluate the reasons given by the state trial court.158See Wilson v. Warden, Ga. Diagnostic Prison, 834 F.3d 1227, 1231–32 (11th Cir. 2016) (en banc), rev’d and remanded sub nom. Wilson v. Sellers, 138 S. Ct. 1188 (2018). Justice Ginsburg’s influence went only so far, however: The Eleventh Circuit ultimately sided with the amicus curiae it had appointed to defend its Hittson approach.159Id. at 1232, 1242. The Supreme Court granted certiorari in that case, now captioned Wilson v. Sellers.160137 S. Ct. 1203 (2017) (granting certiorari).

Relying in part on Justice Ginsburg’s Hittson concurrence, the Wilson Court likewise distinguished the summary denial in Wilson from the summary denial in Richter and declined to apply the “could have supported” framework, instead applying the “look through” approach.161Wilson, 138 S. Ct. at 1195. It concluded (quoting the Hittson concurrence) that a federal habeas court must “train its attention on the particular reasons—both legal and factual—why state courts rejected a state prisoner’s federal claims”162Id. at 1191–92 (internal quotation marks omitted) (quoting Hittson v. Chatman, 576 U.S. 1028, 1028 (2015) (Ginsburg, J., concurring in denial of certiorari)). and then “defer[] to those reasons if they are reasonable.”163Id. at 1192.

II.      Drawing the RichterWilson Line

After Richter and Wilson, two things are clear. First, in some circumstances, Section 2254(d) requires a federal habeas court to apply the “could have supported” framework. If that were not so, Richter would not have enunciated the framework. Second, in some circumstances, the actual reasons employed by the state court make a difference. If that were not so, the question presented in Wilson—whether to consider a lower state court’s reasoning—would be meaningless.164See Pye v. Warden, Ga. Diagnostic Prison, 50 F.4th 1025, 1066 (11th Cir. 2022) (en banc) (Pryor, J., dissenting) (If a federal habeas court can always consider hypothetical, unstated reasons, “then Wilson’s look‑through rule does no work.”); Dennis v. Sec’y, Pa. Dep’t of Corr., 834 F.3d 263, 355 n.11 (3d Cir. 2016) (en banc) (Jordan, J., concurring in part and concurring in the judgment) (“If we ‘look through’ an unreasoned state court decision, Ylst presumably requires that we then review the reasoning given in the lower state court. If not, then why bother ‘looking through’ at all?”). Between these two fixed points lies much uncertainty, however. Lower courts are divided on the proper place of each approach, and thus on the proper place of a state‑court opinion in the Section 2254(d) analysis.165See Brian R. Means, Federal Habeas Manual § 3:70, Westlaw (database updated June 2024) (describing the “confusion . . . over the proper review standard in cases where the state court does provide an explanation for rejecting a claim”). This Part surveys the obstacles lower courts face in drawing the line between Richter and Wilson and their valiant—though ultimately ill‑fated—attempts to elucidate it.

A.      Doctrines in Tension

Three doctrines involving the application of Section 2254(d) to multipart claims (i.e., those that require the petitioner to prove more than one element) that have been adjudicated on the merits in state court present crucial obstacles to a satisfactory analysis of the “could have supported” framework.

1.      The Doctrines

Although these doctrines apply to all such claims,166See, e.g., Brumfield v. Cain, 576 U.S. 305, 323 (2015) (applying partial adjudication rule to intellectual disability claim); Wetzel v. Lambert, 565 U.S. 520, 525 (2012) (per curiam) (applying alternative ground doctrine to suppression of evidence claim); Wogenstahl v. Mitchell, 668 F.3d 307, 327 (6th Cir. 2012) (applying presumption of full adjudication to prosecutorial misconduct claim). But see Holland v. Rivard, 800 F.3d 224, 237 (6th Cir. 2015) (suggesting that “the particular reasoning” behind the partial adjudication rule “is limited to the ineffective‑assistance‑of‑counsel context”). for simplicity’s sake this Article illustrates them using the two‑prong test for ineffective assistance of counselthe context in which “most disputes of this kind arise.”167Thomas v. Clements, 797 F.3d 445, 446 (7th Cir. 2015) (per curiam) (Easterbrook, J., concurring in the denial of rehearing en banc). To show a constitutional violation under Strickland, a defendant must show (1) deficient performance by his counsel and (2) resulting prejudice.168Id. A court has only one path to finding a constitutional violation: finding both prongs satisfied.169See, e.g., Smith v. Robbins, 528 U.S. 259, 289 (2000) (“In sum, Robbins must satisfy both prongs of the Strickland test in order to prevail on his claim of ineffective assistance of appellate counsel.”). In a narrow category of cases, prejudice is presumed; however, these cases fall outside of Strickland’s test. See Bell v. Cone, 535 U.S. 685, 695–96 (2002). See generally Sheri Lynn Johnson, Racial Antagonism, Sexual Betrayal, Graft, and More: Rethinking and Remedying the Universe of Defense Counsel Failings, 97 Wash. U. L. Rev. 57, 63–79 (2019). On the other hand, a court can choose between several paths to finding no violation: finding that the defendant (1) satisfied neither prong, (2) satisfied one prong but not the other, or (3) failed to satisfy one prong without deciding whether he satisfied the other.170See, e.g., Andrus v. Texas, 140 S. Ct. 1875, 1886 (2020) (per curiam) (observing that the state court’s order “does not conclusively reveal whether it determined that Andrus had failed to demonstrate deficient performance under Strickland’s first prong, that Andrus had failed to demonstrate prejudice under Strickland’s second prong, or that Andrus had failed to satisfy both prongs of Strickland”). And, under Richter, Section 2254(d) applies even if the state court denies the claim without specifying which path it has chosen.171See Harrington v. Richter, 562 U.S. 86, 98 (2011) (citing Early v. Packer, 537 U.S. 3, 8 (2002) (per curiam)).

Each path implicates a different doctrine under the Court’s Section 2254(d) jurisprudence. A decision to deny the claim on only one prong implicates the partial adjudication rule of Wiggins v. Smith.172539 U.S. 510 (2003); see infra Section II.A.1.a. A decision to deny relief under both prongs implicates the alternative ground doctrine of Wetzel v. Lambert.173565 U.S. 520 (2012) (per curiam); see infra Section II.A.1.b. And a court’s decision not to reveal which path it took implicates Richter’s presumption of full adjudication.174See infra Section II.A.1.c.

a.      The Partial Adjudication Rule

Under the partial adjudication rule, when a state court denies a claim on one element of a multipart claim and either does not decide the other elements or finds that they were satisfied, a federal habeas court evaluates the other elements unfettered by Section 2254(d).175See, e.g., Porter v. McCollum, 558 U.S. 30, 39 (2009) (per curiam) (citing Rompilla v. Beard, 545 U.S. 374, 390 (2005)). This rule appears first in Williams,176See Williams v. Taylor, 529 U.S. 362, 395–97 (2000) (majority opinion of Stevens, J.) (evaluating the performance prong de novo where the state court did not decide whether counsel had performed deficiently). though it was not explicitly stated until Wiggins. In Wiggins, both state courts found that counsel had not performed deficiently, and thus neither addressed Strickland’s prejudice prong.177Wiggins, 539 U.S. at 517–18. After concluding that counsel’s performance was deficient (and that the state‑court determination to the contrary was unreasonable), the Court reviewed the prejudice prong de novo. It observed that federal “review is not circumscribed by a state court conclusion with respect to prejudice” where “neither of the state courts below reached this prong of the Strickland analysis.”178Id. at 534.

Citing Wiggins, the Court stated the rule even more clearly in Rompilla v. Beard:179545 U.S. 374 (2005). “Because the state courts found the representation adequate, they never reached the issue of prejudice, and so we examine this element of the Strickland claim de novo . . . .”180Id. at 390 (citation omitted). And in Porter v. McCollum,181558 U.S. 30 (2009) (per curiam). the Court relied on Rompilla to apply the same rule when performance rather than prejudice was the unadjudicated prong.182Id. at 39 (citing Rompilla, 545 U.S. at 390). Although the partial adjudication rule applies most frequently to Strickland claims,183See supra note 170. the Court applied the same rule to an intellectual disability claim in Brumfield v. Cain.184576 U.S. 305 (2015). In this case, the state court issued a ruling without finding that the defendant failed to produce evidence necessary to meet an age-of-onset requirement for intellectual deficiency. Id. at 323.

b.      The Alternative Ground Doctrine

The next doctrine, which the Fourth Circuit dubbed the “alternative ground” doctrine,185Long v. Hooks, 972 F.3d 442, 459 (4th Cir. 2020) (en banc). stems from Wetzel v. Lambert.186See id. Under the alternative ground doctrine, a state court’s unreasonable decision on one prong of a multipart claim does not by itself overcome Section 2254(d)’s barrier to relief; instead, that barrier is lifted only if “each ground supporting the state court decision is examined and found to be unreasonable under AEDPA.”187Wetzel v. Lambert, 565 U.S. 520, 525 (2012) (per curiam). In Wetzel itself, the state court rejected the defendant’s Brady claim by finding that the suppressed evidence was both “not exculpatory or impeaching” and not material.188Id. at 524. The Third Circuit granted relief, finding that the claim was meritorious and that the materiality determination rested on a “patently unreasonable” view of impeachment evidence.189Id. at 523 (internal quotation marks omitted) (quoting Lambert v. Beard, 633 F.3d 126, 134 (3d Cir. 2011), vacated sub nom. Wetzel v. Lambert, 565 U.S. 520 (2012) (per curiam)). The Supreme Court vacated, remanding for the lower courts to determine whether the state court’s conclusion that the suppressed material was not favorable to the defendant but rather “entirely ambiguous” was likewise unreasonable.190Id. at 524–26 (internal quotation marks omitted) (quoting Petition for Writ of Certiorari app. at 34, 36, Wetzel, 565 U.S. 520 (No. 13-874)).

The Court has applied the alternative ground doctrine in two cases since Wetzel. One was a relatively straightforward application of the rule to a Strickland claim: The state court denied the claim on both prongs, the prejudice determination was not unreasonable, and thus the alternative ground doctrine barred relief regardless of the reasonableness of the performance determination.191Shinn v. Kayer, 141 S. Ct. 517, 523–24 (2020) (per curiam). The other, Parker v. Matthews,192567 U.S. 37 (2012) (per curiam). was less straightforward and may suggest that the alternative ground doctrine applies outside of the multipart claim scenario, discussed in greater detail below.193See infra notes 279–88 and accompanying text.

c.      The Presumption of Full Adjudication

The third and final relevant doctrine for multipart claims is Richter’s rule that, when a state court denies a claim “without providing its rationale, it is presumed that the state court adjudicated [all] components of the claim.”194Means, supra note 165, § 3:22. As Richter explained:

Where a state court’s decision is unaccompanied by an explanation, the habeas petitioner’s burden still must be met by showing there was no reasonable basis for the state court to deny relief. This is so whether or not the state court reveals which of the elements in a multipart claim it found insufficient, for § 2254(d) applies when a “claim,” not a component of one, has been adjudicated.195Harrington v. Richter, 562 U.S. 86, 98 (2011).

This is a straightforward application of the “could have supported” framework in its natural habitat: unexplained state‑court decisions.

2.      The Tension

The tension between these three doctrines is best illustrated by comparing the protection Section 2254(d) affords a state‑court decision in each Strickland scenario. If a state court provides a reasoned rejection of neither prong, it receives deference on both—the same deference it would have received if it had provided a reasoned rejection of both. On the other hand, a state court that provides a reasoned rejection of one prong and stops there receives deference on only that prong. It is counterintuitive (to say the least) that a state court that explains its decision is entitled to less respect than a state court that does not196See Hodges v. Colson, 727 F.3d 517, 537 n.5 (6th Cir. 2013) (“[The partial adjudication case] also leave this court with the following peculiar rule: if the state court fails to given [sic] an explanation as to either prong, then full AEDPA deference is due to both prongs; but if the state court gives an explanation of one prong, then we do not give deference to the other. In other words, the more information the state court provides, the less deference we grant it.”); cf. Seligman, supra note 26, at 493 (“Summary dispositions thus provide a safe harbor—by writing nothing, the state court protects itself from reversal, with no guarantee that the state court even looked at the relevant evidence.”).—especially given that Strickland explicitly invites197Strickland v. Washington, 466 U.S. 668, 697 (1984) (“[T]here is no reason for a court deciding an ineffective assistance claim . . . even to address both components of the inquiry if the defendant makes an insufficient showing on one. . . . If it is easier to dispose of an ineffectiveness claim on the ground of lack of sufficient prejudice, which we expect will often be so, that course should be followed.”). (and principles of judicial restraint arguably require198Thomas v. Clements, 797 F.3d 445, 446 (7th Cir. 2015) (per curiam) (Easterbrook, J., concurring in the denial of rehearing en banc) (“Resolving an ineffective‑assistance claim on one of these grounds makes for a shorter opinion and also avoids what many judges consider to be dicta (others would call it an alternative holding).”); cf. Dobbs v. Jackson Women’s Health Org., 142 S. Ct. 2228, 2311 (2022) (Roberts, C.J., concurring in the judgment) (describing the “simple yet fundamental principle of judicial restraint” that “[i]f it is not necessary to decide more to dispose of a case, then it is necessary not to decide more”).) one‑prong‑only adjudications.

This Article is hardly the first to note this tension.199See, e.g., Rayner v. Mills, 685 F.3d 631, 637 (6th Cir. 2012) (noting “the argument that tension may exist between the cases”); Dennis v. Sec’y, Pa. Dep’t of Corr., 834 F.3d 263, 375 & n.7 (3d Cir. 2016) (en banc) (Hardiman, J., dissenting) (acknowledging “some tension” between the “could have supported” framework and the partial adjudication rule); Andrew L. Adler, The Non‑Waivability of AEDPA Deference’s Applicability, 67 U. Mia. L. Rev. 767, 775 n.48 (2013) (“While dicta, the language in Richter is difficult to reconcile with this aspect of Rompilla.”). In particular, commentary on the tension between the partial adjudication rule and the presumption of full adjudication has been prevalent, and the question of whether these “cases can co‑exist” has “generat[ed] some conflict” among the federal courts of appeals.200McBride v. Superintendent, SCI Houtzdale, 687 F.3d 92, 100 n.10 (3d Cir. 2012). Discussion of the alternative ground doctrine has been relatively absent from this debate. That is, in part, because the alternative ground doctrine is less a source of tension and more an obstacle to certain proposed resolutions of the tension between the partial adjudication rule and the presumption of full adjudication. See infra text accompanying notes 279–88. In a vacuum, Wetzel can nicely complement either Wiggins (it makes perfect sense that a decision on two prongs would receive deference on both, while a decision on only one prong would receive deference on only one) or Richter (it makes perfect sense that, if an unexplained decision receives deference on both prongs, an explained decision on both prongs would likewise receive deference on both prongs). The problem, in other words, is not that Wetzel cannot be reconciled with Wiggins or that it cannot be reconciled with Richter, but that it cannot be reconciled with both simultaneously. As the Third Circuit has noted, the effect of Richter on “the teachings from Wiggins” is a “complicated question,” because Richter “arguably undermines” the partial adjudication rule.201McBride, 687 F.3d at 100 n.10. The en banc Eleventh Circuit noted that Richter “suggests” the partial adjudication rule “may no longer be good law.”202Childers v. Floyd, 642 F.3d 953, 969 n.18 (11th Cir. 2011) (en banc), vacated on other grounds and remanded, 568 U.S. 1190 (2013). Judge Easterbrook agrees, arguing that “the Supreme Court ought to revisit” the partial adjudication rule in light of Richter203Thomas, 797 F.3d at 448. and hold that Section 2254(d) “governs both elements of Strickland once the state judiciary decides an ineffective‑assistance claim,” regardless of how the claim was decided.204Carter v. Duncan, 819 F.3d 931, 950 (7th Cir. 2016) (Easterbrook, J., concurring). In his view, Richter “impl[ies] that, when a state court gives one sufficient reason and stops, the claim has been fully adjudicated.”205See Thomas, 797 F.3d at 447. Nevertheless, the majority of courts that have considered the issue have concluded (and this Article takes the position206See infra text accompanying note 278.) that the partial adjudication rule survives Richter,207See Rayner v. Mills, 685 F.3d 631, 639 (6th Cir. 2012); Ferrell v. Hall, 640 F.3d 1199, 1226 (11th Cir. 2011); Sussman v. Jenkins, 642 F.3d 532, 534 (7th Cir. 2011) (Ripple, J., in chambers); see also Childers, 642 F.3d at 986 (Wilson, J., concurring in the judgment); id. at 990 n.3 (Barkett, J., dissenting); Beeney, supra note 148, at 1332–45; Means, supra note 165, § 3:22 (collecting cases). Other cases have continued to apply the partial adjudication rule without discussing the possible tension with Richter. See, e.g., Salts v. Epps, 676 F.3d 468, 480 & n.46 (5th Cir. 2012); Williams v. Cavazos, 646 F.3d 626, 637 n.6 (9th Cir. 2011), rev’d on other grounds and remanded sub nom. Johnson v. Williams, 568 U.S. 289 (2013). and the Supreme Court applied it four years after Richter in Brumfield.208Brumfield v. Cain, 576 U.S. 305, 323 (2015).

3.      The (Non)Explanations

This tension can be explained in two ways. First, accept that the doctrines are irreconcilable.209Cf. Walton v. Arizona, 497 U.S. 639, 673 (1990) (Scalia, J., concurring in part and concurring in the judgment) (“Since I cannot possibly be guided by what seem to me incompatible principles, I must reject the one that is plainly in error.”). This is an unattractive option: Lower courts are bound to apply all three until the Supreme Court says otherwise,210See Bryan A. Garner et al., The Law of Judicial Precedent 27–33 (2016). and the Court is loath to overrule itself, especially on issues of statutory interpretation.211Id. at 333–37. Moreover, it is unnecessary. As this Article explains, the task of reconciling these doctrines in a principled way is difficult, but not impossible. It is worth considering, however, why this knot is so difficult to untie. That leads to the second explanation: The relevant legal materials—AEDPA’s text and the Supreme Court’s decisions—have been spectacularly unclear.

Some of the blame lies at Congress’s feet, and some at the Court’s. On the former front, this Article has already explained the criticisms of AEDPA’s draftsmanship.212See supra notes 97–104 and accompanying text. Compounding that lack of clarity, Section 2254(d)’s language “had no prior habeas history or pedigree”; that is, it was not derived from case law or previous reform proposals.213Blume, supra note 19, at 261, 272–73. Thus, it is difficult to tell what, if any, “old soil” it brings with it.214See Felix Frankfurter, Some Reflections on the Reading of Statutes, 47 Colum. L. Rev. 527, 537 (1947) (“[I]f a word is obviously transplanted from another legal source, whether the common law or other legislation, it brings the old soil with it.”).

The Supreme Court also bears a portion of the blame, having given little-to-no explanation of the doctrines’ foundations or rationales. Take the development of the partial adjudication rule, for instance, which Judge Easterbrook described as “the judicial equivalent of a rumor chain”:215Thomas v. Clements, 797 F.3d 445, 448 (7th Cir. 2015) (per curiam) (Easterbrook, J., concurring in the denial of rehearing en banc).

 

* Williams applies the rule without even acknowledging that it is doing so.216See Williams v. Taylor, 529 U.S. 362, 395–97 (2000) (majority opinion of Stevens, J.).

 

* Wiggins states simply (and without citation to authority) that federal “review is not circumscribed by a state court conclusion with respect to prejudice” because “neither of the state courts below reached this prong of the Strickland analysis.”217Wiggins v. Smith, 539 U.S. 510, 534 (2003). It never explained why the state courts’ partial adjudications led to this conclusion.

 

* Rompilla similarly explained that the Court would “examine this element of the Strickland claim [(i.e., prejudice)] de novo” because the state courts “never reached the issue of prejudice.”218Rompilla v. Beard, 545 U.S. 374, 390 (2005) (citing Wiggins, 539 U.S. at 534). It cited only Wiggins and offered no further explanation.219Id.

 

* Porter (relying solely on Rompilla) explained that “we review [performance] de novo” “[b]ecause the state court did not decide whether Porter’s counsel was deficient.”220Porter v. McCollum, 558 U.S. 30, 39 (2009) (per curiam) (citing Rompilla, 545 U.S. at 390.

 

The only time the Court even hinted at an explanation of the partial adjudication rule was Brumfield’s opaque statement that, because the state court had “never made any finding” on the prong at issue, there was “no determination on that point to which a federal court must defer.”221Brumfield v. Cain, 576 U.S. 305, 323 (2015). Judge Easterbrook was right to call these “drive‑by statements.”222Thomas v. Clements, 797 F.3d 445, 448 (7th Cir. 2015) (per curiam) (Easterbrook, J., concurring in the denial of rehearing en banc).

The alternative ground doctrine has not been given any clearer footing. Wetzel—the case establishing the doctrine—cites only six cases, three of which were prior proceedings in the prisoner’s case.223See Wetzel v. Lambert, 565 U.S. 520, 522–23 (2012) (per curiam). Two others—Brady and Strickler v. Greene224527 U.S. 263 (1999).—were cited only for the basics of the prisoner’s specific constitutional claim.225See Wetzel, 565 U.S. at 521–22 (first citing Brady v. Maryland, 373 U.S. 83 (1963); and then citing Strickler, 527 U.S. at 281). That leaves only one citation to a Section 2254(d) case: Richter.226Id. at 524 (citing Harrington v. Richter, 562 U.S. 86, 102 (2011)). And the Court did not explain how Richter, the text of Section 2254(d), or any other authority or principle supported the alternative ground doctrine. Instead, it stated only that one reasonable conclusion by a state court would render unreasonable conclusions “beside the point.”227Id. The other two cases to apply the alternative ground doctrine likewise do not explain its basis, instead simply citing Wetzel.228See Parker v. Matthews, 567 U.S. 37, 42 (2012) (per curiam) (citing Wetzel, 565 U.S. at 524–25); see also Shinn v. Kayer, 141 S. Ct. 517, 524 (2020) (per curiam) (first quoting Wetzel, 565 U.S. at 525; and then citing Matthews, 567 U.S. at 42). And the Wetzel dissent (the only recorded dissent in this line of cases) does not mention the doctrine.229See Wetzel, 565 U.S. at 526–28 (Breyer, J., dissenting).

Perhaps the best‑explained of the three doctrines is the presumption of full adjudication. Even it, however, has been explained only in Richter’s “ironically terse” opinion.230Seligman, supra note 26, at 477. Like AEDPA’s drafting, see supra notes 97–104 and accompanying text, Richter’s reasoning has been sharply criticized. See, e.g., John H. Blume & Sheri Lynn Johnson, Gideon Exceptionalism?, 122 Yale L.J. 2126, 2141 (2013) (“The Court’s reasoning was almost as shoddy as trial counsel’s representation.”). Immediately after concluding that postcard denials were presumptively “on the merits,” the Richter Court explained that the presumption of full adjudication grows out of “the habeas petitioner’s burden,” which (under the “could have supported” framework) required the petitioner to show that “there was no reasonable basis for the state court to deny relief.”231Richter, 562 U.S. at 98.

These doctrines stem from a statutory “pig’s ear.” And the cases establishing them not only “fail to build the bridge between the authorities they cite and the results they decree,”232Alexander M. Bickel & Harry H. Wellington, Legislative Purpose and the Judicial Process: The Lincoln Mills Case, 71 Harv. L. Rev. 1, 3 (1957). but largely fail to cite any authorities at all. This may be in part because several of these cases (Porter, Wetzel, Matthews, and Kayer) were per curiam summary reversals—a disposition that occurs far more often in habeas cases than in any other context233Richard C. Chen, Summary Dispositions as Precedent, 61 Wm. & Mary L. Rev. 691, 707 (2020) (“Of the eighty‑eight summary reversals from the 2005 to 2018 terms [of the Roberts Court], forty‑one have come in federal habeas cases. Qualified immunity is a somewhat distant second at eleven.” (footnote omitted)). and does not lend itself to thorough explanations.234See Stephen L. Wasby, Steven Peterson, James Schubert & Glendon Schubert, The Per Curiam Opinion: Its Nature and Functions, 76 Judicature 29, 37 (1992) (“All types of summary dispositions by the Court, including per curiam rulings based only on appeals or certiorari papers (and thus without full briefing and oral argument), have been criticized for providing insufficient guidance to lower court judges and lawyers, who must try to figure out what the Court meant in its brief order.”).

B.      The Circuit Split

Reflecting the confusion, the lower courts have split into effectively three approaches—two categorical and one “hybrid.”235This circuit split is far from clean—numerous decisions are not clear which category they fit in. Compare Pye v. Warden, Ga. Diagnostic Prison, 50 F.4th 1025, 1037–38 (11th Cir. 2022) (en banc) (classifying decisions as applying an expansive approach), with id. at 1070–72 & nn.21–28 (Pryor, J., dissenting) (classifying the same decisions as ambiguous or applying a narrow approach). Each approach, however, either has significant theoretical problems or fails to account for the doctrines discussed above. This Section considers each in turn.

      1.       The Expansive Approach

Under one approach, which this Article refers to as “the expansive approach,”236In using the terms “expansive” and “narrow” to describe two approaches, see infra Section II.B.2, this Article refers to the expansiveness or narrowness of the Court’s reading of Richter. It would be equally accurate to reverse these labels and speak of expansive and broad readings of Wilson. Cf. Thompson v. Skipper, 981 F.3d 476, 485 (6th Cir. 2020) (Nalbandian, J., concurring) (referring to what this Article calls the “narrow approach” as “the broad reading of Wilson”). federal courts should apply the “could have supported” framework in all cases that were adjudicated on the merits in state court, regardless of any state‑court opinion on the claim. These courts have “extended Richter’s logic to allow”—indeed, require—“federal habeas courts to hypothesize, in the face of an unreasonable and erroneous state court ruling, an alternative ground of decision.”237Aziz Z. Huq & Genevieve Lakier, Apparent Fault, 131 Harv. L. Rev. 1525, 1553 (2018).

One particularly vivid example is Holland v. Rivard.238800 F.3d 224 (6th Cir. 2015). The petitioner claimed that his confession was inadmissible because it was obtained via a custodial interrogation that occurred after he invoked his right to counsel.239Id. at 233; see Edwards v. Arizona, 451 U.S. 477, 478, 487 (1981) (holding that the Fifth, Sixth, and Fourteenth Amendments require suppression of a post arrest confession obtained after a defendant invokes a right to counsel). The state court concluded that there was no interrogation—a conclusion the district court found unreasonable.240Holland, 800 F.3d at 235. The Sixth Circuit was willing to assume this conclusion was indeed unreasonable, but nevertheless felt bound by the “could have supported” framework to search for other reasonable justifications for denying the claim.241Id. It found that justification in the argument that the petitioner was not “in custody” at the time he gave his confession—a completely unrelated part of the constitutional test.242Id. at 237. Holland distinguished the partial adjudication cases by concluding that they are “limited to the ineffective‑assistance‑of‑counsel context.” Id. But see Brumfield v. Cain, 576 U.S. 305, 323 (2015) (applying partial adjudication rule to intellectual disability claim).

Rationales for this approach vary. Some courts have attempted to justify it as a matter of “the plain language of AEDPA,” arguing that Section 2254(d) “requires federal courts to examine the relevant state court ‘decision,’” rather than its reasoning.243Gill v. Mecusker, 633 F.3d 1272, 1288 (11th Cir. 2011) (quoting 28 U.S.C.A. § 2254(d)); see, e.g., Robinson v. Polk, 438 F.3d 350, 358 (4th Cir. 2006) (“In assessing the reasonableness of the state court’s application of federal law, [therefore,] the federal courts are to review the result that the state court reached, not whether [its decision] [was] well reasoned.” (alterations in original) (internal quotation marks omitted) (quoting Wilson v. Ozmint, 352 F.3d 847, 855 (4th Cir. 2003))); Dennis v. Sec’y, Pa. Dep’t of Corr., 834 F.3d 263, 371 (3d Cir. 2016) (en banc) (Hardiman, J., dissenting) (“By its terms, AEDPA applies to federal review of state‑court decisions—not to the specific explanations that support them.” (citing 28 U.S.C. § 2254(d)). Others have argued that a focus on the state court’s opinion would be “unduly formalistic,” given that federal courts are perfectly capable of analyzing the full record.244Neal v. Puckett, 286 F.3d 230, 246 (5th Cir. 2002) (en banc) (per curiam). Since Richter, justifications of this approach have focused on Richter’s articulation of the “could have supported” framework, suggesting that it would be incongruous to have “two divergent analytical modes—one when there is no previous reasoned decision below and another for when there is.”245Wilson v. Warden, Ga. Diagnostic Prison, 834 F.3d 1227, 1236 (11th Cir. 2016) (en banc), rev’d and remanded sub nom. Wilson v. Sellers, 138 S. Ct. 1188 (2018); see also Dennis, 834 F.3d at 371–73 (Hardiman, J., dissenting) (arguing that the expansive approach “is supported by notions of consistency and coherence as well”); Mann v. Ryan, 774 F.3d 1203, 1225 (9th Cir. 2014) (Kozinski, J., concurring in part and dissenting in part) (“[R]eversing a state court’s reasonable decision on the grounds of incorrect reasoning risks treating defendants inconsistently: Those who are given incorrect reasoning get relief while those who aren’t given any reasoning do not.”). And the approach has the advantage of apparently having three votes on the current Supreme Court: Justices Thomas and Alito both joined Justice Gorsuch’s dissent in Wilson, which argued that, even after Wilson, “a federal habeas court should look at all the arguments presented in state and federal court and examine the state court record” and “deny[] relief if those materials reveal a basis to do so reasonably consistent with” clearly established federal law.246See Wilson, 138 S. Ct. at 1197, 1204 (Gorsuch, J., dissenting).

Parsimonious as it is, the expansive approach encounters serious problems. For one, the “plain language” rationale falls apart upon a closer reading of the text, which does not ask whether the “decision” was itself unreasonable, but whether it “involved an unreasonable application of” clearly established federal law.24728 U.S.C. § 2254(d)(1) (emphasis added). Even more concerningly, it is flatly inconsistent with Wilson, which rejected the application of the “could have supported” framework in situations “where there is a reasoned decision by a lower state court.”248Wilson, 138 S. Ct. at 1195; see Means, supra note 165, § 3:70 (“With this observation, the Supreme Court apparently settled the matter: the ‘fill the gaps’ aspect of Richter—considering grounds that could have supported the state court’s decision—does not extend beyond unexplained rulings to reasoned state court decisions.”). And it cannot account for the partial adjudication rule: If the state court’s reasons don’t matter, how can Section 2254(d)’s applicability turn on the reasons the state court gave?

2.       The Narrow Approach

A second approach—which this Article refers to as “the narrow approach”—is nearly as categorical. Rather than viewing the “could have supported” approach as the rule, the narrow approach views it as a limited exception to the general rule that “a federal habeas court simply reviews the specific reasons given by the state court and defers to those reasons if they are reasonable.”249Wilson, 138 S. Ct. at 1192. In the ordinary case, therefore, a court applying the narrow approach must “review the actual grounds on which the state court relied,”250Coleman v. Bradshaw, 974 F.3d 710, 719 (6th Cir. 2020) (emphasis added) (citing Wilson, 138 S. Ct. at 1192). and then ask “whether that explanation was reasonable thereby requiring [a federal court’s] deference.”251Winfield v. Dorethy, 956 F.3d 442, 454 (7th Cir. 2020) (citing Wilson, 138 S. Ct. at 1192). Thus, under the narrow approach, “if a state court articulates its reasoning, it is only that reasoning that receives deference.”252Ford v. Peery (Ford I), 976 F.3d 1032, 1044 (9th Cir. 2020), withdrawn on reh’g, Ford v. Peery (Ford II), 999 F.3d 1214 (9th Cir. 2021). The Ford cases are a particularly good demonstration of the effect the state court’s opinion can have under the narrow approach. In Ford I, the Ninth Circuit granted relief because it interpreted the state court’s opinion as applying only the “harmless beyond a reasonable doubt” standard of Chapman v. California, 386 U.S. 18, 87 (1967), and not the native prejudice standard under Darden v. Wainwright, 477 U.S. 168 (1986). Ford I, 976 F.3d at 1041, 1045. On rehearing, however, the court interpreted the state‑court opinion as applying “the functional equivalent of the Darden harmlessness test,” and thus denied relief under Section 2254(d). Ford II, 999 F.3d at 1225–27. This does not mean, however, that a state court’s deficient reasoning is sufficient to warrant relief: While courts applying the narrow approach “afford . . . no deference” to unreasonable state‑court explanations,253Rogers v. Superintendent Greene SCI, 80 F.4th 458, 463 (3d Cir. 2023). they may nevertheless reject a claim on the merits even after concluding that Section 2254(d) does not bar relief.254See, e.g., Gish v. Hepp, 955 F.3d 597, 604, 607 (7th Cir. 2020); Dennis v. Sec’y, Pa. Dep’t of Corr., 834 F.3d 263, 349 n.6 (3d Cir. 2016) (en banc) (Jordan, J., concurring in part and concurring in the judgment) (explaining “the interplay between §§ 2254(a) and 2254(d)”). The relationship between the merits of the claim and Section 2254(d) is discussed in greater detail below. See infra Section III.A.

The primary justification for the narrow approach mirrors the primary problem with the expansive approach: consistency with Wilson. Nearly every court to adopt the narrow approach post‑Wilson has cited Wilson in so doing.255See, e.g., Wooten v. Lumpkin, 113 F.4th 560, 566–67 (5th Cir. 2024) (citing Wilson, 138 S. Ct. at 1192); Porter v. Coyne‑Fague, 35 F.4th 68, 75 (1st Cir. 2022) (same); Ford I, 976 F.3d at 1044 (same); Coleman, 974 F.3d at 719 (same); Winfield, 956 F.3d at 454 (same); Gibbs v. Adm’r N.J. State Prison, 814 F. App’x 686, 689 & n.6 (3d Cir. 2020) (same); accord Means, supra note 165, § 3:70. Some have also argued that Richter is limited “by its terms.”256Grueninger v. Dir., Va. Dep’t of Corr., 813 F.3d 517, 525 (4th Cir. 2016) (“But ‘[b]y its terms,’ Richter is limited to cases ‘where a state court’s decision is unaccompanied by an explanation.’” (alteration in original) (internal quotation marks omitted) (quoting Woolley v. Rednour, 702 F.3d 411, 422 (7th Cir. 2012)); Woolley, 702 F.3d at 422. Others have argued that it would be illogical to defer to a rationale that we know did not exist, “disregard[ing] a state court’s expressed rationale . . . and presum[ing] instead that” state courts adopted a different rationale.257Woolley, 702 F.3d at 422; Dennis, 834 F.3d at 353 (Jordan, J., concurring in part and concurring in the judgment) (“We would do real damage to [comity and federalism] principles were we to begin re‑writing state court opinions to save them.”). Like the expansive approach, the narrow approach has multiple votes on the Supreme Court: Citing Wilson, Justice Sotomayor has explained, in an opinion joined by Justice Kagan, that “[w]hen a state court gives a reasoned explanation for its decision, federal habeas courts must review that decision on its own terms.”258Dunn v. Reeves, 141 S. Ct. 2405, 2413, 2420 (2021) (Sotomayor, J., dissenting) (per curiam) (citing Wilson, 138 S. Ct. at 1192). The majority does not address the applicability of the “could have supported” framework.

Ultimately, the narrow approach produces similar results to the approach this Article advocates below. However, it has its own share of struggles. Perhaps the foremost is explaining why the “could have supported” approach would ever apply if state‑court decisions are to be reviewed on their own terms. If bad reasons mean no deference, why shouldn’t no reasons mean no deference?259See Dennis, 834 F.3d at 371 (Hardiman, J., dissenting) (describing the narrow approach’s distinction between explained and unexplained orders as “unprincipled” since “AEDPA does not distinguish” along those lines); see also Holland v. Rivard, 800 F.3d 224, 236 (6th Cir. 2015) (“Richter suggests that this is not a meaningful distinction . . . .”). The narrow approach also has trouble explaining another category of underexplained state‑court decisions: cases where the state court does not cite (or even exhibit awareness of) the controlling Supreme Court precedent. Under Early v. Packer,260537 U.S. 3 (2002) (per curiam). such decisions can still bar federal relief.261Id. at 8 (“Avoiding these pitfalls does not require citation of our cases—indeed, it does not even require awareness of our cases, so long as neither the reasoning nor the result of the state‑court decision contradicts them.”). The narrow approach seemingly does not allow for that result. A state‑court decision that articulates reasons for denying a federal claim without reference to the governing legal standard nevertheless “explain[s] the rejection of a claim,”262Grueninger v. Dir., Va. Dep’t of Corr., 813 F.3d 517, 525 (4th Cir. 2016). thus limiting deference to that explanation under the narrow approach. Yet it certainly cannot stand “on its own terms”;263Reeves, 141 S. Ct. at 2420 (Sotomayor, J., dissenting) (per curiam) (citing Wilson, 138 S. Ct. at 1192). the habeas court could find the decision reasonable only by supplying its own reason that the decision is consistent with federal law—a version of the “could have supported” framework.

3.      The Eleventh Circuit’s Hybrid Approach

Perhaps recognizing the weaknesses of both more categorical approaches, the Eleventh Circuit recently adopted a sort of “hybrid” approach. Sitting en banc in Pye v. Warden, Georgia Diagnostic Prison,26450 F.4th 1025 (11th Cir. 2022) (en banc). the court distinguished the “specific reasons given by the state court” from the “particular justifications that the state court provided.”265Id. at 1035–36. Under Pye, a court “evaluate[s]” only “the reasons offered by the [state] court.”266King v. Warden, Ga. Diagnostic Prison, 69 F.4th 856, 867 (11th Cir. 2023) (citing Pye, 50 F.4th at 1036). However, in justifying those reasons, the habeas court is “not limited by the particular justifications the state court provided for its reasons, and . . . may consider additional rationales that support the state court’s determination.”267Jennings v. Sec’y, Fla. Dep’t of Corr., 55 F.4th 1277, 1292 (11th Cir. 2022) (citing Pye, 50 F.4th at 1036). This approach thus applies a modified “could have supported” approach to all claims that were denied on the merits. Rather than asking what arguments could have supported any hypothetical reason to deny the claim, it asks what arguments could have supported denying the claim for the reason given by the state court.

The distinction between “specific reasons” and “particular justifications” is fuzzy, however—the Eleventh Circuit has never defined either, and has noted only that “reasons” are defined “at a relatively high level of generality.”268See Pye, 50 F.4th at 1040 n.9. In Pye itself, the “reason” attributed to the state‑court decision seems to have been the bottom‑line conclusion that counsel’s performance did not prejudice the defendant.269See id. at 1041–42 (“Applying AEDPA to Strickland’s prejudice standard, we must decide whether the state court’s conclusion that [counsel]’s performance at the sentencing phase of Pye’s trial didn’t prejudice him . . . was ‘so obviously wrong that its error lies beyond any possibility for fairminded disagreement.’” (quoting Shinn v. Kayer, 141 S. Ct. 517, 523–24 (2020) (per curiam)); see also Jennings, 55 F.4th at 1293–96; cf. Washington v. Att’y Gen., No. 21-13756, 2023 WL 7391693, at *4 (11th Cir. Nov. 8, 2023) (searching for—and ultimately finding—a “potential justification” not discussed by the state court for a conclusion that counsel’s performance was not deficient) (emphasis omitted); Heidler v. Warden, GDCP, No. 20-13752, 2023 WL 4927253, at *24 (11th Cir. Aug. 2, 2023) (treating state court decisions on both prongs of Strickland as reasons under Pye). And in a Batson case,270Referring to Batson v. Kentucky, a landmark Supreme Court decision which held that prosecutors may not strike jurors through peremptory challenges in a racially discriminatory manner. 476 U.S. 79, 89 (1986). an Eleventh Circuit panel defined the state court’s reason as its conclusion that “it was not an abuse of discretion for the trial court to accept [the prosecutor’s] explanation of the strike.”271King, 69 F.4th at 872.

The Pye court defended its approach primarily by reference to “AEDPA’s plain language and the logic of . . . Richter.”272Pye, 50 F.4th at 1037. In light of these arguments, it reasoned that “[t]he lone question” was whether “Wilson instituted an entirely new and different AEDPA regime.”273Id. at 1038–39. It concluded that Wilson did no such thing, but instead “confronted only a very narrow question” about which state‑court opinions are relevant, chiding the dissent (which adopted the narrow approach) for “overread[ing] (and thus misread[ing]) Wilson.”274Id. at 1039.

On first blush, the hybrid approach seems largely consistent with the doctrines discussed above. Because Pye applies a modified version of the “could have supported” approach across the board, it is plainly consistent with Richter. And if different prongs of a multipart claim constitute distinct “reasons,” as Pye and its progeny suggest,275See supra note 269. then a court applying the hybrid approach would not supply hypothetical reasons for prongs the state court did not decide, resulting in the de novo review that Wiggins requires. It would, however, supply hypothetical reasons for both prongs if the state court decided both prongs—just as Wetzel requires.276In Heidler v. Warden, the Eleventh Circuit applied Pye to essentially recreate the alternative ground doctrine without citing Wetzel. Heidler v. Warden, GDCP, No. 20-13752, 2023 WL 4927253, at *24 (11th Cir. Aug. 2, 2023). And, though it gives Wilson a somewhat miserly reading, it leaves some room for the look‑through presumption as the method for determining the state court’s “reasons”—a role consistent with (though not dictated by) the Court’s only post‑Wilson application of the look‑through presumption.277See Shinn v. Kayer, 141 S. Ct. 517, 524 & n.1 (2020) (per curiam) (applying Wilson to conclude that the state court rejected both prongs of a Strickland claim).

Nevertheless, this approach falls short. First, the reason-justification distinction is highly artificial. Indeed, part of the reason the hybrid approach seems so consistent with Supreme Court case law is that the reason-justification distinction is seemingly reverse‑engineered to create that consistency. As a result, it is also difficult to draw the line between reasons and justifications.278The Eleventh Circuit has never delineated the distinction, remarking only that “reason[s]” are defined “at a relatively high level of generality.” Pye, 50 F.4th at 1040 n.9. In Pye itself, the “reason” attributed to the state court was simply its conclusion that the petitioner had not shown Strickland prejudice. See id. at 1041–42. In another case, the court classified a state court’s conclusion that the trial court did not abuse its discretion by accepting the prosecutor’s explanation of a juror strike as its “reason” for rejecting a Batson claim. King v. Warden, Ga. Diagnostic Prison, 69 F.4th 856, 872 (11th Cir. 2023). To be sure, coming up with any theory that accounts for the seemingly conflicting decisions in this area is quite the feat, and sometimes vertical stare decisis requires distinctions that don’t make much sense. But the need for such an inexplicable and unworkable distinction is a sign that something has gone awry.

Moreover, the hybrid approach falls just short of accounting for all of the Supreme Court decisions: Although it explains the application of the alternative ground doctrine in Wetzel and Kayer, it struggles to explain Matthews. In Matthews, the Court considered the state court’s rejection of the defendant’s claim that the trial court impermissibly shifted the burden of proof as to his extreme emotional disturbance defense.279Parker v. Matthews, 567 U.S. 37, 41–45 (2012) (per curiam). The Court separately considered the state court’s rejection of the prisoner’s due process claim arising from the prosecutor’s closing argument. Id. at 45–49. The state court gave two rationales for rejecting this claim. First, it argued that one of its recent decisions required rejecting the claim.280Id. at 41. Second, it explained that the jury instructions accurately represented the burden of proof.281Id. at 42. The Sixth Circuit found the first rationale unreasonable and thus held that Section 2254(d) did not preclude relief.282Id. The Supreme Court reversed, applying Wetzel to hold that, because the second argument was “sufficient to reject [the] claim,” the possible unreasonableness of the first argument was “irrelevant.”283Id. However, the Court explained that the first argument’s potential unreasonableness “would be relevant if [it] formed the sole basis for denial of [the] claim.”284Id. at 42.

Treating these arguments as reasons would seem to violate Pye’s instruction to define reasons at a “high level of generality,”285Pye v. Warden, Ga. Diagnostic Prison, 50 F.4th 1025, 1040 n.9 (11th Cir. 2022) (en banc). especially when contrasted with decisions (like Pye itself) defining reasons in terms of a particular prong of a multipart claim.286See supra note 278. They are much more naturally described as justifications. But if that is so, why would it matter whether a particular justification was or was not “the sole basis for” the state court’s decision?287Matthews, 567 U.S. at 42. Under Pye, the federal court would be obliged to supply its own justifications, and the state court’s justifications would have been irrelevant.

*  *  *

In sum, lower courts have struggled to draw the line between Richter and Wilson. Their attempts to do so have been frustrated by three doctrines governing multipart claims that not only make understanding Richter and Wilson difficult but are also unclear in their own right.

III.      Defining Unreasonableness

Having explained why each prevailing approach falls short, this Article now discusses an alternative theory of Section 2254(d) that borrows from each of the approaches above. It offers a principled reconciliation of the entire body of Supreme Court jurisprudence under Section 2254(d). Under this theory, Section 2254(d) does not ask about the merits of a claim. Instead, it asks whether a “qualifying error”—that is, an error that triggers one of Section 2254(d)’s enumerated exceptions—infected the state court’s adjudication.

A.      The Nature of Section 2254(d)

Understanding the distinction between the Section 2254(d) inquiry and the merits requires first understanding Section 2254(d)’s place within the labyrinthine structure of federal habeas review for state prisoners. For nearly a century, federal courts could generally issue writs of habeas corpus only for prisoners held in federal custody.288See 1 Hertz & Liebman, supra note 105, § 2.4[d][i], at 47–49; see also Judiciary Act of 1789, ch. 20, § 14, 1 Stat. 73, 81–82 (“And be it further enacted . . . [that federal judges] shall have power to grant writs of habeas corpus for the purpose of an inquiry into the cause of commitment.—Provided, That writs of habeas corpus shall in no case extend to prisoners in [state jail], unless where they are in custody, under or by colour of the authority of the United States . . . .”). An exception existed for bankruptcy, and Congress twice “authorized limited issuance of the writ in response to two crises it viewed as sufficiently pressing to warrant a federal response.” Cent. Va. Cmty. Coll. v. Katz, 546 U.S. 356, 374 & n.11 (2006). During Reconstruction, however, Congress extended this authority to those held in state custody289Act of Feb. 5, 1867, ch. 28, 14 Stat. 385, 385–86.—an authority that lives on in one of Section 2254(d)’s neighboring paragraphs, Section 2254(a).290See 28 U.S.C. § 2254(a).

Recognizing the breadth of this power, courts soon began developing doctrines to circumscribe its use.291See, e.g., Ex parte Royall, 117 U.S. 241, 252–53 (1886) (requiring exhaustion of state remedies); see also Lee Kovarsky, The New Negative Habeas Equity, 137 Harv. L. Rev. 2222, 2228 (2024) (recognizing “a set of limits that are creatures of judicial making”). Congress, too, acted to “offset the broad § 2254(a) power with a variety of rules restricting its use.”292Habeas Assistance & Training Counsel Project, Practitioner’s Guide to 28 U.S.C. § 2254(d), at 2 (2021) (on file with the George Mason Law Review) [hereinafter Practitioner’s Guide to § 2254(d)]. These barriers can be sorted into two categories: (1) those directing how federal courts should decide the merits of a claim,293Barriers in this category include limitations on when a federal court may hold an evidentiary hearing, see 28 U.S.C. § 2254(e)(2), and restrictions on the availability of appeal, see id. § 2253. and (2) those that prohibit relief regardless of the merits.294Barriers in this category include the prohibition on relief—but not on consideration of the merits of the claim—where the petitioner has not “exhausted the remedies available in the courts of the State,” see id. § 2254(b), and the general rule against granting relief on procedurally defaulted claims, see, e.g., House v. Bell, 547 U.S. 518, 536 (2006).

Congress doubtless erected a barrier when it adopted Section 2254(d). It is not immediately clear, however, into which category that barrier falls. Courts often describe it as requiring deference to the state court’s judgment,295See, e.g., Kernan v. Hinojosa, 578 U.S. 412, 413 (2016) (per curiam) (“If the state courts adjudicate the prisoner’s federal claim ‘on the merits,’ then AEDPA mandates deferential, rather than de novo, review . . . .” (citation omitted) (quoting 28 U.S.C. § 2254(d))); Coleman v. Johnson, 566 U.S. 650, 656 (2012) (per curiam) (“The state court . . . determination in turn is entitled to considerable deference under AEDPA.”); Cavazos v. Smith, 565 U.S. 1, 9 (2011) (per curiam) (describing “this Court’s opinions highlighting the necessity of deference to state courts in § 2254(d) habeas cases”); Knowles v. Mirzayance, 556 U.S. 111, 123 (2009) (describing the “doubly deferential judicial review that applies to a Strickland claim evaluated under the § 2254(d)(1) standard” (citing Yarborough v. Gentry, 540 U.S. 1, 5–6 (2003) (per curiam))). and the House Conference Report suggests a deference‑based reading.296 H.R. Rep. No. 104‑518, at 111 (1996) (Conf. Rep.). This report is the only official document present in AEDPA’s legislative history. Claudia Wilner, “We Would Not Defer to That Which Did Not Exist”: AEDPA Meets the Silent State Court Opinion, 77 N.Y.U. L. Rev. 1442, 1458 n.82 (2002). As Judge Easterbrook observed, however, “the word ‘deference’ does not appear in the statute,” and Section 2254(d) “does not tell us to ‘defer’ to state decisions, as if the Constitution means one thing in Wisconsin and another in Indiana.”297Lindh v. Murphy, 96 F.3d 856, 868 (7th Cir. 1996) (en banc), rev’d and remanded on other grounds, 521 U.S. 320 (1997). Indeed, a statute that required actual deference would pose difficult constitutional concerns about infringing on “federal courts’ independent interpretive power.” Id. at 868–69; see 2 Hertz & Liebman, supra note 105, § 32.3, at 1831–32 (“[T]he only arguable basis for sustaining the constitutionality of section 2254(d)(1)’s apparent encroachment upon the powers of Article III courts is a view of section 2254(d)(1) as constituting merely a limitation on relief—and not a congressional dilution of the federal courts’ ability to reach independent conclusions as to whether a constitutional violation has occurred . . . .” (footnotes omitted)). “Deferential” thus works better as an adjective than “defer” does as an imperative.

Rather than focusing on the claim and whether it is strong enough to overcome deference, Section 2254(d) “focuses on what a state court knew and did.”298Cullen v. Pinholster, 563 U.S. 170, 182 (2011). As one court put it, Section 2254(d) “exalt[s] the role that a state court’s decision plays in a habeas proceeding by specifically directing the habeas court to make the state court decision the cynosure of federal review.”299See O’Brien v. Dubois, 145 F.3d 16, 20 (1st Cir. 1998); see also Adler, supra note 199, at 789 n.126 (“[T]he focal point of AEDPA deference is the state court’s decision rather than the petitioner’s claim.”). This is clear from the statutory text, which focuses on the “adjudication of the claim,” not on the claim itself. It references the state‑court “decision,” “determination,” and “proceeding,” but refers to the “claim” only to ask whether it was the subject of said “adjudication.”30028 U.S.C. § 2254(d).

In addition to squaring with the statutory text, this understanding makes sense in light of the debates and developments that presaged AEDPA. First, it recognizes that, though Congress did not enact Bator’s process model,301See Alan K. Chen, Shadow Law: Reasonable Unreasonableness, Habeas Theory, and the Nature of Legal Rules, 2 Buff. Crim. L. Rev. 535, 604 (1999) (“Section 2254(d)(1) clearly does not further a pure process model.”). it heeded his fundamental insight that relitigation should “address itself . . . not so much to the substantive question whether truth prevailed but to the institutional or functional one.”302Bator, supra note 56, at 449. Although Congress did not limit the “institutional question,” as did Bator, to the fairness of the state court’s processes, it did limit relitigation to situations where certain red flags call into question “the presumption that state courts know and follow the law.”303Woodford v. Visciotti, 537 U.S. 19, 24 (2002) (per curiam). Second, it harmonizes Section 2254(d) with Teague, its spiritual forbear. Under Teague, retroactivity is a “threshold question,”304Teague v. Lane, 489 U.S. 288, 300–01 (1989) (plurality opinion). meaning it must be decided before and separate from the merits, even though the two are “obviously interrelated.”305See Penry v. Lynaugh, 492 U.S. 302, 352 (1989) (Scalia, J., concurring in part and dissenting in part). Accordingly, the Court has described Teague as a limitation on relief independent of the merits.306See Danforth v. Minnesota, 552 U.S. 264, 291 (2008) (“Teague does not imply that there was no right and thus no violation of that right at the time of trial—only that no remedy will be provided in federal habeas courts.”). In recent years, the Court tightened that limitation, eliminating one of Teague’s two exceptions. See Edwards v. Vannoy, 141 S. Ct. 1547, 1560 (2021). Although I have previously disagreed with that characterization,307Rutledge, supra note 76, at 512. Instead, I view Teague as a “temporal choice of law” issue. Id.; see also Liebman, Apocalypse Next Time?, supra note 56, at 2032; Danforth, 552 U.S. at 307 (Roberts, C.J., dissenting). it reflects the Court’s understanding.

The Court has recognized this Section 2254(d)-merits distinction. In Richter, it chastised the Ninth Circuit for “treat[ing] the unreasonableness question as a test of its confidence in the result it would reach under de novo review.”308Harrington v. Richter, 562 U.S. 86, 102 (2011). Instead, the Court explained, “it is a necessary premise that the two questions are different.”309Id. at 101. Because of this, Section 2254(d) “sets forth a precondition to the grant of habeas relief[,] . . . not an entitlement to it.”310Fry v. Pliler, 551 U.S. 112, 119 (2007). This means that in some cases Section 2254(d) may be satisfied even where the constitutional claim is not viable.311See Rose v. Lee, 252 F.3d 676, 689–91 (4th Cir. 2001) (criticizing the district court for granting relief upon a finding that the state‑court decision was contrary to clearly established law without also examining whether the underlying claim was meritorious).

There are two arguments for nevertheless treating Section 2254(d) as modifying the merits analysis rather than creating an additional requirement. One is textual; it emphasizes that the statutory language focuses on the state court’s “decision”—which can be read as referring only to its result.312See supra note 243 and accompanying text; cf. Johnson v. Williams, 568 U.S. 289, 310 (2013) (Scalia, J., concurring in the judgment) (“[W]hat is accorded deference is not the state court’s reasoning but the state court’s judgment . . . .”). But if that argument were applied to the “contrary to” clause, it would directly contradict Williams, which held that a decision can be contrary to clearly established law either in its outcome or in the reasoning it employs.313Williams v. Taylor, 529 U.S. 362, 405–06 (2000) (majority opinion of O’Connor, J.). Relying on Williams, Judge Hardiman distinguishes between the “unreasonable application” clause, which he says is subject to the “could have supported” analysis in all cases, and the “contrary to” clause, which is “not amenable” to that analysis. Dennis v. Sec’y, Pa. Dep’t of Corr., 834 F.3d 263, 372 n.3 (3d Cir. 2016) (en banc) (Hardiman, J., dissenting) (emphasis omitted). But if anything, the text of Section 2254(d) supports just the opposite distinction. See infra text accompanying notes 314–15. This is especially so if, as Judge Hardiman asserts, “AEDPA applies to federal review of state‑court decisions—not to the specific explanations that support them.” Dennis, 834 F.3d at 371. And the argument is even weaker for the “unreasonable application” clause. That clause asks whether the decision “involved an unreasonable application of clearly established Federal law.”31428 U.S.C. § 2254(d)(1) (emphasis added) (comma omitted). Answering that question requires looking at what the decision “involved” and how it “appli[ed]” federal law315Id.—questions about the means rather than merely the result.

The other argument relies on dicta from Berghuis v. Thompkins,316560 U.S. 370 (2010). where the Court suggested that a decision that is “correct under de novo review” is “therefore necessarily reasonable under” Section 2254(d).317Id. at 389 (citing Knowles v. Mirzayance, 556 U.S. 111, 123–24 (2009)). That language is imprecise, however. A better formulation is Judge Arthur Tarnow’s:

Most times a finding in favor of the Petitioner under § 2254(d) means that a fortiori Petitioner is held in violation of his constitutional rights under § 2254(a). It is usually the case that where a state court unreasonably rejects a constitutional claim it can also be immediately determined that the constitutional right was violated. But this is not always true.318Ballinger v. Prelesnik, 844 F. Supp. 2d 857, 868 n.1 (E.D. Mich. 2012) (emphasis added), rev’d and vacated, 709 F.3d 558 (6th Cir. 2013).

Indeed, the possibility that a petition may fail to state a ground for relief under Section 2254(a) even if Section 2254(d) is no bar to relief gives life to Cullen v. Pinholster’s319563 U.S. 170 (2011). assurance that its reading of Section 2254(d) (which limits the analysis to the record before the state court) does not render Section 2254(e)(2)’s limitations on evidentiary hearings “superfluous.”320See id. at 185. In his Pinholster concurrence—published just months after Thompkins—Justice Breyer identified several scenarios where an evidentiary hearing would be needed to determine the merits even where Section 2254(d) does not bar relief.321Id. at 205 (Breyer, J., concurring in part and dissenting in part). Specifically, Justice Breyer pointed to scenarios where the state court accepts the petitioner’s facts as true, “deciding [unreasonably] that, even if those facts were true, federal law was not violated,” and scenarios where “the state‑court rejection rested on only one of several related federal grounds.” Id. In context, Thompkins is best read as establishing only that a correct result is necessarily a reasonable result, not that a correct result guarantees a reasonable decision. As the Court recently reiterated, oblique language suggesting that one inquiry subsumes another should not trump a careful analysis of Section 2254(d)’s text.322See Brown v. Davenport, 142 S. Ct. 1510, 1528 (2022).

B.      Qualifying Errors

The question under Section 2254(d) is thus not whether the claim is meritorious, or even whether it is obviously meritorious. Instead, it is whether the state court committed an error that pierces the general bar on relief. To do so, an error must have three features: it must be the right type of error, it must be sufficiently severe, and it must have affected the state court’s decision.

      1.       Type of Error

The first requirement—that the error be of the right type—simply reflects the errors enumerated by Section 2254(d). Three types of errors qualify.323See Yackle, supra note 85, at 109 (“Section 2254(d) bars federal habeas relief unless the state court decision in question fails one of the three tests stated in paragraphs (1) and (2) [of the Section].”). Some courts consider both errors described in paragraph (d)(1) together, resulting in statements that there are only two types of qualifying errors. See, e.g., Valentino v. Clarke, 972 F.3d 560, 575 (4th Cir. 2020) (“For these already‑adjudicated claims, § 2254(d) permits only two paths to federal habeas relief.”). Mindful, however, of paragraph (d)(1)’s disjunctive structure and the Supreme Court’s admonition to “give independent meaning to both the ‘contrary to’ and ‘unreasonable application’ clauses of [§ 2254(d)(1)],” Williams v. Taylor, 529 U.S. 362, 404 (2000) (majority opinion of O’Connor, J.), this Article follows Yackle’s lead and treats the two species of (d)(1) error as separate ways of satisfying this requirement. The first two are listed in paragraph (d)(1): “a decision” that is “contrary to . . . clearly established Federal law,” and “an unreasonable application of” that clearly established law.32428 U.S.C. § 2254(d)(1). The third is found in paragraph (d)(2): “an unreasonable determination of the facts in light of” the state‑court record.325Id. § 2254(d)(2). The most important point is that the list of errors is exclusive.326See Practitioner’s Guide to § 2254(d), supra note 292, at 16 (“Once it is determined that a claim was ‘adjudicated on the merits in State court proceedings,’ § 2254(d) forbids a federal court from granting habeas relief ‘unless’ the state court’s ‘decision’ on the claim is shown to have been defective in one or more of the ways contemplated by subsections (d)(1) and (d)(2).”); Davenport, 142 S. Ct. at 1520 (“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.”). Other errors—such as unreasonable interpretations of state law327See, e.g., Wilson v. Corcoran, 562 U.S. 1, 5 (2010) (per curiam) (“But it is only noncompliance with federal law that renders a State’s criminal judgment susceptible to collateral attack in the federal courts.”). or lower‑court precedent,328See, e.g., Kernan v. Cuero, 138 S. Ct. 4, 9 (2017) (per curiam) (“[A]s we have repeatedly pointed out, ‘circuit precedent does not constitute “clearly established Federal law, as determined by the Supreme Court.”’ Nor, of course, do state‑court decisions, treatises, or law review articles.” (citations omitted) (quoting Glebe v. Frost, 135 S. Ct. 429, 430 (2014) (per curiam)). unreasonable failure to extend Supreme Court precedent,329See White v. Woodall, 572 U.S. 415, 424–26 (2014). or lack of citation to (or awareness of) of controlling cases330Early v. Packer, 537 U.S. 3, 8 (2002) (per curiam).—do not qualify.

2.       Severity of Error

The second requirement—the error’s severity—likewise simply reflects Section 2254(d)’s language and the Supreme Court’s interpretations of it. The general requirement is that Section 2254(d) bars relief “absent an error that lies ‘beyond any possibility for fairminded disagreement.’”331Mays v. Hines, 141 S. Ct. 1145, 1146 (2021) (per curiam) (quoting Shinn v. Kayer, 141 S. Ct. 517, 520 (2020) (per curiam)). Under the “contrary to” clause, an error is only sufficiently severe if the decision is “‘diametrically different,’ ‘opposite in character or nature,’ or ‘mutually opposed’” to clearly established law.332Williams v. Taylor, 529 U.S. 362, 405 (2000) (majority opinion of O’Connor, J.) (emphasis added) (quoting Webster’s Third New International Dictionary 495 (1976)). Under the “unreasonable application” clause, the application must be “objectively unreasonable,” not merely wrong; even ‘clear error’ will not suffice”333Woodall, 572 U.S.at 419 (internal quotation marks omitted) (quoting Lockyer v. Andrade, 538 U.S. 63, 75–76 (2003)).—the error must be “well understood and comprehended in existing law.”334Harrington v. Richter, 562 U.S. 86, 103 (2011). And under paragraph (d)(2), the barrier to relief remains if “evidence in the state‑court record can fairly be read to support” the state court’s factual finding, even if the federal court “would have reached a different conclusion in the first instance.”335Wood v. Allen, 558 U.S. 290, 301–02 (2010). Overall, the severity requirement “reflects the view that habeas corpus is a ‘guard against extreme malfunctions in the state criminal justice systems,’” not ordinary errors.336Richter, 562 U.S. at 102–03 (emphasis added) (quoting Jackson v. Virginia, 443 U.S. 307, 332 n.5 (1979) (Stevens, J., concurring in the judgment)); see also Dennis v. Sec’y, Pa. Dep’t of Corr., 834 F.3d 263, 370 (3d Cir. 2016) (en banc) (Hardiman, J., dissenting).

3.       Effect of Error

The third requirement—that the error affect the decision to deny the claim—is more intricate. In the context of paragraph (d)(1)’s “unreasonable application” clause, it derives from the verb “involved.”33728 U.S.C. § 2254(d)(1). In the context of paragraph (d)(2), it is derived from the phrase “based on.”338Id. § 2254(d)(2). The key language in either context must be distinguished from the passive verb “was” in paragraph (d)(1)’s “contrary to” clause, which does not incorporate an effect requirement.339Id. § 2254(d)(1); see infra text accompanying notes 343–48.

When used as a verb, “base” means “to find a foundation or basis for.”340Base, Merriam‑Webster Online, https://perma.cc/TN7C-E5EF. A foundation is, of course, meant to bear weight—an architect who designed one not suited for that purpose wouldn’t be an architect for long. For an unreasonable determination of facts to be the “foundation” or “basis” of a state‑court decision, therefore, the judicial equivalent of a non‑load‑bearing wall will not do. Instead, the state‑court decision must have relied on the unreasonable determination in some way.

Paragraph (d)(1)’s use of the verb “involve” is similar. Though “involve” can, in some contexts, mean simply “to have within or as part of itself,” it more often “suggests inclusion by virtue of the nature of the whole.”341Involve, Merriam‑Webster Online, https://perma.cc/LVZ2-PWYP. In other words, “involve” usually suggests not mere inclusion, but inclusion that in some way changes or affects the whole. Consider a football analogy: It would be quite natural for an underutilized receiver to complain that he is not “involved” in the team’s passing offense, even though he is both “within” and “part of” it. Everyone would understand that he is complaining about not being sufficiently integral to the offense, not about not being a member of it. Such an interpretation also avoids making the word “involve” superfluous; reading “involve” to denote mere inclusion would give the statute the same meaning as if the word were excluded altogether.342Although it might be argued that redundancy‑based arguments are not applicable to poorly drafted statutes—a category AEDPA is generally considered to belong, see supra notes 97–103 and accompanying text—such objections are rarely persuasive. See Antonin Scalia & Bryan A. Garner, Reading Law: The Interpretation of Legal Texts 179 (2012).

In contrast to both paragraph (d)(2) and the “unreasonable application” clause in paragraph (d)(1), the language of paragraph (d)(1)’s “contrary to” clause does not suggest an effect requirement.343But cf. Williams v. Taylor, 529 U.S. 362, 418 (2000) (Rehnquist, C.J., concurring in part and dissenting in part) (arguing that, because the state court “did not rely” on its erroneously legal standard, its decision was not contrary to clearly established law). That is because the “contrary to” clause, unlike the other two clauses, asks about the state‑court decision itself, not about the applications of law or determinations of fact included therein.344See supra text accompanying notes 314–15. In addition to the textual indications, the distinction between the “contrary to” clause and the other two clauses makes good practical sense. When considering the effect of the other two types of error, the habeas court must place the error within the legal framework underlying the state court’s decision.345See infra text accompanying notes 355–56 (discussing the effect of the legal framework in the unreasonable application analysis under Wetzel). But with one of the two species of “contrary to” error delineated by the Supreme Court—those where the state court “applies a rule that contradicts the governing law”346Williams, 529 U.S. at 405 (majority opinion of O’Connor, J.).—the framework is absent from the state‑court decision. Thus, like structural errors in criminal trials, these “contrary to” errors “affect[] the framework within which the [decision] proceeds” and thus any evaluation of their effects would be nothing more than speculation.347See Arizona v. Fulminante, 499 U.S. 279, 310 (1991); Williams, 529 U.S. at 414 (O’Connor, J., concurring) (“It is impossible to determine, however, the extent to which the [state court’s application of the wrong legal standard] affected its ultimate finding . . . .”). And with the second species of “contrary to” error—which occurs when a state court reaches the opposite result from a Supreme Court precedent on “materially indistinguishable” facts—there is no need to evaluate the effect because we already know there will have been one; the cases are “indistinguishable.”348Williams, 529 U.S. at 405 (majority opinion of O’Connor, J.).

Besides textual grounding and practicality, applying an effect requirement to “unreasonable application” error and to paragraph (d)(2) error has other virtues. First, it preserves the presumption of correctness when a state court denies a claim despite being unreasonably generous to a defendant on certain points in the analysis.349For instance, without an effect requirement, a state court that unreasonably found that a defendant did not knowingly and intelligently waive his rights under Miranda v. Arizona, 384 U.S. 436 (1966), would, by making a finding that was unreasonably generous to the defendant, thereby lose Section 2254(d) protection for its otherwise reasonable determination that no custodial interrogation occurred. Second, it avoids asking federal courts to “flyspeck” a state‑court’s reasoning,350See Meders v. Warden, Ga. Diagnostic Prison, 911 F.3d 1335, 1349 (11th Cir. 2019); see also Rashad v. Walsh, 300 F.3d 27, 45 (1st Cir. 2002) (“It is not our function, however, to grade a state court opinion as if it were a law school examination.”); Wright v. Sec’y for the Dep’t of Corr., 278 F.3d 1245, 1255 (11th Cir. 2002) (“Requiring state courts to put forward rationales for their decisions so that federal courts can examine their thinking smacks of a ‘grading papers’ approach that is outmoded in the post‑AEDPA era.” (citing Hennon v. Cooper, 109 F.3d 330, 335 (7th Cir. 1997)); Cruz v. Miller, 255 F.3d 77, 86 (2d Cir. 2001) (“[W]e are determining the reasonableness of the state courts’ ‘decision,’ not grading their papers.” (citation omitted) (quoting 28 U.S.C. § 2254(d)(1))); Hennon, 109 F.3d at 335 (focusing on actual reasoning “would place the federal court in just the kind of tutelary relation to the state courts that [AEDPA was] designed to end”). instead requiring them to focus only on errors that might have been consequential.351Cf. Liebman, Apocalypse Next Time?, supra note 56, at 2023 (“It is hard to imagine less deference to an erroneous state court legal determination than a decision to reverse the determination though it made no difference to the outcome.”).

Finally, it accounts for several features of Section 2254(d) jurisprudence that otherwise remain unexplained. As early as Williams, the Court seemed to ask whether the state‑court decision “turned on” the error.352Williams, 529 U.S. at 397 (majority opinion of Stevens, J.). The Court appeared to get at something similar—though the lack of citation or explanation makes it hard to tell353See supra notes 223–28.—when it asked whether the “unreasonable application” error in Wetzel was “beside the point.”354Wetzel v. Lambert, 565 U.S. 520, 524 (2012) (per curiam). An effect requirement nicely explains Wetzel and its progeny: If the state court independently (and reasonably) rejected the claim on an alternative ground, the error cannot have affected the decision. In such a case, the otherwise‑qualifying errors are essentially “quarantined” from the rest of the decision, preventing them from infecting the outcome. That is true whether or not the alternative argument is framed at a high enough level of generality to be a separate “reason” under the hybrid approach,355Cf. supra notes 279–88 and accompanying text. so long as the alternative argument is (1) sufficient by itself to deny the claim and (2) analytically separate from the infected argument. Thus, it explains not only Wetzel and Kayer, but also Matthews. By contrast, when a state court chooses not to adjudicate other prongs of a claim, it leaves no uninfected argument to prove that the error had no effect. This explains the partial adjudication rule.356See supra Section II.A.1.a.

Of course, the effect requirement is not a pure counterfactual analysis asking how the state court actually would have ruled if not for the error. Were that so, the partial adjudication rule would not require de novo review of unadjudicated prongs. Instead, it would require, at best, something akin to an “Erie guess,”357An “Erie guess” occurs when a federal court sitting in diversity, required by Erie R.R. v. Tompkins, 304 U.S. 64 (1938), to apply state law to a claim, is “confronted with a state law issue of first impression” and thus “must attempt to predict how the state’s highest court would decide the issue.” Baldwin v. Express Oil Change, LLC, 87 F.4th 1292, 1301 n.7 (11th Cir. 2023). and at worst something akin to psychoanalysis. Not only would that analysis be impossible,358See Steinman, supra note 16, at 1516 (“[W]here a lower court’s actual reasoning fails to pass muster, there is no way to know whether it would have reached the same decision had it based its decision on the correct factors.”). it would also violate Williams’s directive to view Section 2254(d) through an objective lens rather than a subjective one.359See Williams v. Taylor, 529 U.S. 362, 409–10 (2000) (majority opinion of O’Connor, J.). Instead, a federal court must ask the (admittedly nebulous) question of whether the error was “beside the point.”360Wetzel v. Lambert, 565 U.S. 520, 524 (2012) (per curiam).

IV.      Proving Unreasonableness

The previous Part discussed what a federal court must conclude to overcome Section 2254(d)’s barrier: that the state court committed a “qualifying error” that is of the right type, is sufficiently severe, and may have affected the outcome. The remaining question, then, is how a federal court can reach that conclusion.

Scrutinizing a state court’s opinion in some cases but only its result in others seems at first like applying divergent standards. But, as is often the case, things are not as they appear. Having established that Section 2254(d) adds an independent requirement of its own, rather than merely dictating the lens through which the constitutional claim must be evaluated, things come sharply into focus. As this Part demonstrates, the seemingly divergent standards are just two evidentiary paths to proving the same ultimate fact: that the state court committed a qualifying error.

A.      Allocating the Burden of Proof

Formally speaking, a federal petition for a writ of habeas corpus is an original, civil action.361See, e.g., Banister v. Davis, 140 S. Ct. 1698, 1702–03 (2020). But see Liebman, Apocalypse Next Time?, supra note 56, at 2038–40 (criticizing this characterization). As the plaintiff in that action, a prisoner seeking relief bears the burden of proof on most issues.362See Garlotte v. Fordice, 515 U.S. 39, 46 (1995) (“[T]he habeas petitioner generally bears the burden of proof . . . .”). Indeed, in the only notable exception to this rule—the assessment of whether a constitutional error prejudiced the defendant—the Court shied away from “burden of proof” terminology. See O’Neal v. McAninch, 513 US. 432, 436 (1995) (“[W]e deliberately phrase the issue in this case in terms of a judge’s grave doubt, instead of in terms of ‘burden of proof.’”). Nevertheless, the effect is the same. See 2 Hertz & Liebman, supra note 105, § 31.2[b], at 1686 (“The Court’s answer [in O’Neal] was that the habeas corpus petitioner should prevail in that situation, thus effectively giving the state the burden of proof . . . .”). Section 2254(d)’s bar on relief is no exception.363See Cullen v. Pinholster, 563 U.S. 170, 181 (2011) (citing Woodford v. Visciotti, 537 U.S. 19, 25 (2002) (per curiam)). As a result, habeas petitioners must “affirmatively demonstrate that the state court was unreasonable.”364Seligman, supra note 26, at 494. This point seems simple enough, but failure to recognize and account for it has undermined previous scholarship in this area.

Take Professor Adam Steinman’s (pre‑Richter) approach, for instance. His view is largely similar to this Article’s. Both reject the expansive approach discussed in Section II.B.1, supra, instead opting to evaluate state‑court opinions largely on their own terms.365See Steinman, supra note 16, at 1528 (“Therefore, in light of Williams, § 2254(d)(1) must be read to bar habeas relief only where the state court articulates a legal basis for its ruling that reasonably applies the established rule of constitutional law.”). However, he subtly shifts the burden of proof from the petitioner to the state.366In fairness, Steinman’s article was published the year before the Court first indicated that petitioners bear the burden of proof under Section 2254(d). See Visciotti, 537 U.S. at 25. In his view, Section 2254(d)’s applicability is “contingent on whether the state court’s actual reasoning passes analytical muster.”367Steinman, supra note 16, at 1522. Thus, he considers Section 2254(d) a “standard” that a state court’s reasoning must “satisfy” rather than a barrier that a petitioner must overcome.368See id. at 1516. This is clearest in his discussion of unexplained decisions. In two separate sections, he follows out the natural conclusion of his combination of a focus on the state‑court opinion and allocation of the burden of proof to the state: “[A] state court that fails to write an opinion articulating how it interpreted and applied federal law should not receive deference under § 2254(d)(1).”369Id. at 1522; see id. at 1517 (“[T]he safest course would be to withhold deference where the state court fails to write an opinion setting forth its legal reasoning.”). Although this view had many other proponents (including the NACDL briefs discussed supra Section I.B.3.a), it was squarely rejected in Richter.370See Harrington v. Richter, 562 U.S. 86, 99 (2011). See generally Wilner, supra note 296 (arguing against applying Section 2254(d) to unexplained decisions); Glidden, supra note 28 (same).

Professor Steven Semeraro’s approach (likewise articulated before Richter) meets a similar fate. His approach, like this Article’s, incorporates the state court’s rationale into the Section 2254(d) inquiry.371Steven Semeraro, A Reasoning‑Process Review Model for Federal Habeas Corpus, 94 J. Crim. L. & Criminology 897, 900 (2004). However, he establishes a series of hurdles that the state court must overcome before the state can invoke Section 2254(d)’s limitations on relief. First, it must “cite[] all applicable federal law—including statutes, Supreme Court cases, and federal appellate court cases from the circuit in which the state is located—that the federal habeas court would have cited had it been charged with the responsibility to decide the claim on the merits.”372Id. at 927–28. Thus, “opinions wholly failing to cite significant federal authority” would not do.373Id. at 928. Indeed, neither would “[r]ote citation of federal precedents.”374Id. Rather, the state court would have to “demonstrate[] a thorough understanding of federal law” (a standard to be evaluated by the federal habeas court).375Id. Semeraro even seems to view a state court’s failure to meet his requirements as adequate grounds for a federal court to grant a measure of relief—sending the case back to the state court for reconsideration—without finding that a federal right was violated.376Id. at 929 (“When a state court does not engage in sufficiently thorough reasoning, the federal habeas court should identify the weakness—i.e., the cases not cited or the factors left unaddressed—and return the case to the state system for appropriate analysis.”). Underscoring his misplacement of the burden of proof, he advocates that sort of treatment in one case where the Supreme Court plurality agreed with the state court on the merits,377Semeraro, supra note 371, at 930–33 (analyzing Ramdass v. Angelone, 530 U.S. 156 (2000) (plurality opinion)). In Ramdass, the plurality explicitly stated that “the Constitution does not require the instruction that [the petitioner] now requests.” Ramdass, 530 U.S. at 178 (plurality opinion). To make matters worse, the primary basis for the argument that the state court in Ramdass acted unreasonably is that it “cited only one federal case.” Semeraro, supra note 371, at 931. But the one case it did cite was the case that caused the Supreme Court to vacate the state court judgment for reconsideration. See Ramdass v. Virginia, 512 U.S. 1217 (1994) (remanding for reconsideration in light of Simmons v. South Carolina, 512 U.S. 154 (1994)). Before Simmons, the state court had “repeatedly” rejected claims like the one at issue in Ramdass. Ramdass v. Commonwealth, 437 S.E.2d 566, 573 (Va. 1993). While a state court opinion can fail to warrant deference under Section 2254(d) even where the claim is not ultimately meritorious, see supra note 311 and accompanying text, application of the doctrine of stare decisis to a federal claim is surely not by itself the type of error that would warrant such a result. and one that this Article uses below as an example of the correct approach.378Compare Semeraro, supra note 371, at 933–36 (criticizing Yarborough v. Alvarado, 541 U.S. 652 (2004)), with infra text accompanying notes 416–27 (discussing Alvarado with approval).

This approach obviously runs afoul of Richter.379See King v. Warden, Ga. Diagnostic Prison, 69 F.4th 856, 869 (11th Cir. 2023) (noting that Eleventh Circuit Section 2254(d) precedent does not “require[] state courts to show their work . . . by mentioning every relevant circumstance”). It just as obviously runs afoul of Packer, which rejected a requirement that state courts cite (or even exhibit awareness of) the controlling Supreme Court precedents.380Early v. Packer, 537 U.S. 3, 8 (2002) (per curiam). Presumably, if a state court does not need to cite Supreme Court precedent, which they are required to follow on questions of federal law as a matter of vertical stare decisis, see Garner et al., supra note 210, at 679, they also have no need to cite lower federal court precedents, which “are not controlling authority in state court[s],” id. at 691, and which “do[] not constitute ‘clearly established Federal law’” under AEDPA, Kernan v. Cuero, 138 S. Ct. 4, 9 (2017) (per curiam) (quoting Glebe v Frost, 135 S. Ct. 429, 430 (2014) (per curiam)). But its flaws are more fundamental than being at odds with Supreme Court precedent—an accusation to which Semeraro would likely plead guilty as charged.381Semeraro’s project was not descriptive, but rather prescriptive, offering “an alternative to the objective reasonableness standard.” Semeraro, supra note 371, at 900. Rather, it is the epitome of a “grading papers” approach, requiring state courts not only to show their work, but to use the same citations the federal court would use and to demonstrate (to the federal court’s satisfaction) a “thorough understanding of federal law.”382Id. at 928. Semeraro anticipates this objection, asserting in response that “state judges, who are obligated to follow federal law, can fairly be required to understand and apply that law in a sophisticated way.”383Id. at 929. That may be so, but it is surely not what Section 2254(d) actually requires. While not all approaches that focus on the state court’s reasoning fall victim to Judge Posner’s charge against restoring the pre‑AEDPA “tutelary relation” between the state and federal courts,384Hennon v. Cooper, 109 F.3d 330, 334–35 (7th Cir. 1997). Semeraro’s certainly does.385The approach advocated by Claudia Wilner, which would treat unexplained orders as inherently unreasonable, see Wilner, supra note 296, at 1444, falls prey to a similar burden‑shifting trap. See Seligman, supra note 26, at 494–95 (criticizing Wilner on this ground).

The same misapprehension of the burden of proof (or perhaps of the fact that there is something to be proved) implicitly underlies arguments based on purported “asymmetry” between the “could have supported” framework and the scrutiny that courts employing the narrow approach give to reasoned decisions.386See supra note 245 and accompanying text. In these views, it is problematic that a petitioner’s entitlement to relief can turn on whether the state court explained its decision or kept silent.387See Seligman, supra note 26, at 474. But allocating the risk of evidence insufficient to establish truth is among the specific functions of a burden of proof.388Cf. Joseph M. Livermore, Absent Evidence, 26 Ariz. L. Rev. 27, 36 (1984) (arguing that inferences from the absence of evidence “[g]enerally . . . should be used against the party having the burden of proof”). As the Supreme Court has said, Section 2254(d) requires giving state courts “the benefit of the doubt.”389Woodford v. Visciotti, 537 U.S. 19, 24 (2002) (per curiam). This must mean that when there is “doubt” about whether a qualifying error has occurred—whether because of an unexplained order or otherwise—the state court must receive the “benefit” of that doubt. Given the presumption of state‑court competence, it’s no surprise that a state court—even an unreasonable one—that “keeps silent is considered wise.”390Cf. Proverbs 17:28 (English Standard Version) (“Even a fool who keeps silent is considered wise; when he closes his lips, he is deemed intelligent.”); The Yale Book of Quotations 466 (Fred R. Shapiro ed., 2006) (“Better to remain silent and be thought a fool than to speak and remove all doubt.”).

How, then, can an opinion‑focused approach avoid the burden‑shifting problem? Packer suggests the solution: Its language indicates that a state‑court opinion becomes unreasonable by what it does, not by what it doesn’t do.391See Early v. Packer, 537 U.S. 3, 8 (2002) (per curiam) (using the active verbs “applies” and “confronts” to describe what a state court must do to lose Section 2254(d)’s protections (quoting Williams v. Taylor, 529 U.S. 362, 405–06 (2000) (majority opinion of O’Connor, J.))). So long as the court does not slip into the “pitfalls” described by Section 2254(d)—that is, so long as it does not commit a qualifying error—its decision was reasonable and a federal court may not grant relief.392Id. Unlike Steinman’s and Semararo’s approaches, which permit relief unless the state court demonstrates its reasonableness, the correct approach bars relief unless the petitioner demonstrates the state court’s unreasonableness.

B.      Meeting the Burden: Two Paths

The Section 2254(d) inquiry is one of objective reasonableness (or, more precisely, objective unreasonableness).393Williams, 529 U.S. at 409 (majority opinion of O’Connor, J.). As one scholar put it, such standards “pervade the law.”394Biale, supra note 16, at 1350. In Williams, Justice O’Connor explicitly invoked this legal backdrop, relying on the “familiar” meaning of the term “unreasonable.”395Williams, 529 U.S. at 410 (majority opinion of O’Connor, J.). As this invocation suggests, objective reasonableness tests in other areas may shed light on Section 2254(d).396Biale, supra note 16, at 1376 (“Williams’ citation to the ‘common,’ ‘familiar’ nature of the objective reasonableness standard suggests that it is entirely appropriate to look to other areas of law for the meaning of the standard in the habeas context.”).

Perhaps the most prominent and well‑theorized of these objective reasonableness tests comes from the law of negligence, where reasonableness “has long been the touchstone.”397Sanchez v. State, 784 N.E.2d 675, 679 (N.Y. 2002); see Restatement (Third) of Torts: Liability for Physical and Emotional Harm § 3 (Am. L. Inst. 2010) (“A person acts negligently if the person does not exercise reasonable care under all the circumstances.”). To be sure, the analysis under Section 2254(d) does not mimic Judge Learned Hand’s famous cost‑benefit balance.398See United States v. Carroll Towing Co., 159 F.2d 169, 173 (2d Cir. 1947). As has been recognized, “the analysis of or test for reasonableness often varies with the context.” Ritter, supra note 37, at 80. Instead, prisoners typically must demonstrate “an extraordinary measure of fault akin to gross negligence or recklessness.”399Aziz Z. Huq, Habeas and the Roberts Court, 81 U. Chi. L. Rev. 519, 528 (2014). But one feature of negligence law (besides the presence of an objective reasonableness test) nevertheless makes it a promising analogue: its clear placement of the burden of proof on the plaintiff.400See Dan B. Dobbs, Paul T. Hayden & Ellen M. Bublick, The Law of Torts § 165 (2d ed.), Westlaw (database updated Apr. 2024).

One way a negligence plaintiff can meet that burden is through the doctrine of res ipsa loquitur, under which negligence can be inferred if “the injury occur[s] under such circumstances that in the ordinary course of events it would not have occurred if ordinary care had been observed.”401Hake v. George Wiedemann Brewing Co., 262 N.E.2d 703, 705 (Ohio 1970); see Dobbs et al., supra note 400, § 169; 1 Stuart M. Speiser, The Negligence Case: Res Ipsa Loquitur § 2:4 (1972). It is typically also required that the defendant is probably the person responsible. See Dobbs et al., supra note 400, § 173. In other words, a negligence plaintiff can prevail without identifying what specific unreasonable thing the defendant did, so long as she can show that the accident would not have occurred unless the defendant made some unspecified unreasonable act or omission. That is not the only way a negligence plaintiff can meet her burden of proof, however—nor is it even the most common. Instead, plaintiffs typically meet their burden by pointing out “precisely what the defendant did or didn’t do.”402Dobbs et al., supra note 400, § 168.

Negligence law thus demonstrates a sort of two‑path approach to proving unreasonableness. A plaintiff can either point to a specific unreasonable act or omission or to a result so unlikely that an unreasonable act or omission must have occurred. Habeas follows suit: A prisoner can likewise prove unreasonableness either by pointing to a specific qualifying error or by pointing to a result so unreasonable that the state court must have committed a qualifying error. Simply comparing Williams and Richter supports this proposition. In Williams, as Judge Katzmann observed, at least six Justices endorsed examining the state court’s actual reasoning to determine whether Section 2254(d) barred relief.403Washington v. Schriver, 255 F.3d 45, 53–54 (2d Cir. 2001). In Richter, by contrast, the Court clearly indicated that a petitioner could overcome Section 2254(d) using only the state court’s result.404See supra Section I.A (describing the “could have supported” framework adopted by Richter). Therefore, unless Richter overruled Williams sub silentio, despite citing it favorably—and the “prevailing consensus” is that it did not405Fuster, supra note 21, at 1342.—both paths are open.

But the two‑path approach has roots much deeper than just Richter and Williams. As this Article demonstrates in the following sections, the two‑path approach has left its fingerprints in every corner of Section 2254(d) jurisprudence. And, as Wetzel and Wilson indicate, it has not yet worn out its welcome at One First Street—some federal circuit court suggestions to the contrary notwithstanding.

      1.       The Two‑Path Approach and the Contrary To Clause

Proving the existence of a two‑path approach is easiest in the context of the “contrary to” clause, because the Williams Court explicitly required it. Under Williams, a state‑court determination can be contrary to clearly established federal law in two ways. First, it can “appl[y] a rule that contradicts the governing law set forth in” the controlling cases.406Williams v. Taylor, 529 U.S. 362, 405 (2000) (majority opinion of O’Connor, J.). This is an example of the first path—pointing to a specific error—because it requires ascertaining precisely what rule the state court applied, which cannot be done without reference to its opinion. Williams itself illustrates this, which found the state court’s decision “contrary to” Strickland because it applied the wrong prejudice standard—something the Court demonstrated the only way it could: by quoting the state court’s opinion.407Id. at 393–94 (majority opinion of Stevens, J.). Seven years later, in Abdul‑Kabir v. Quarterman,408550 U.S. 233 (2007). the Court similarly examined the state court’s precise reasoning, down to its choice of citations, before concluding that its “formulation of the issue”—not its outcome—was contrary to clearly established law.409Id. at 257–58. And five years after that—this time post‑RichterLafler v. Cooper410566 U.S. 156 (2012). found the state court’s decision “contrary to” clearly established law because, though presented with an ineffective assistance claim, it “failed to apply Strickland.”411Id. at 173. Again, the Court backed this accusation with citations to the state court’s opinion.412See id.; see also Practitioner’s Guide to § 2254(d), supra note 292, at 23 & n.56 (collecting similar examples from federal circuit courts). Clearly, then, the “governing law” prong of the “contrary to” test requires—or at minimum permits—a federal court to examine a state court’s actual reasoning.

But that is not the exclusive path to establishing that a state court decision is “contrary to” clearly established law. That task can also be accomplished by showing the state court “confront[ed] a set of facts that are materially indistinguishable from” a controlling case, yet arrived at a different result.413Williams v. Taylor, 529 U.S. 362, 406 (2000) (majority opinion of O’Connor, J.). This is an example of the second path; determining that the “facts” a state court “confronts” in a particular case are “materially indistinguishable” from Supreme Court precedent requires only a knowledge of the facts and the precedent, not a knowledge of the state court’s reasoning. From those facts and precedents, it can be inferred that the state court either applied a rule that was contrary to clearly established law or unreasonably applied that law (such as by manufacturing an unreasonable distinction), since no other explanation can account for the different outcomes.414See Steinman, supra note 16, at 1526 n.129. The Supreme Court has never clearly permitted relief on this ground. One possible explanation is that lower federal courts handle cases in this category satisfactorily, which have permitted relief on this ground. See, e.g., Cockerham v. Cain, 283 F.3d 657, 663 (5th Cir. 2002). Whatever the cause, “[s]uch cases . . . [are] rare.” Seligman, supra note 26, at 492. This inference can be drawn even in some cases where the facts are “significantly different,” so long as the legal consequences of those facts “clearly are the same.”4152 Hertz & Liebman, supra note 105, § 32.3 n.30 (citing Ramdass v. Angelone, 530 U.S. 156, 180 (2000) (O’Connor, J., concurring in the judgment)); see Ramdass, 530 U.S. at 180 (O’Connor, J., concurring in the judgment) (observing that, if the entry of judgment “was foreordained,” then the case would have been “materially indistinguishable” from clearly established law and thus relief would be permitted under the “contrary to” clause). For an example of a lower federal court applying this rationale, see Budder v. Addison, 851 F.3d 1047, 1055–60 (10th Cir. 2017).

      1.       The Two‑Path Approach and the Unreasonable Application Clause

Proving that the “unreasonable application” clause incorporates a two‑path approach is slightly trickier—if the Court had said so outright, there would be no circuit split and no need for this Article. However, that the “contrary to” clause requires a two‑path approach lends credence to the idea that the “unreasonable application” clause requires the same. Moreover, that is precisely what the Court does in its “unreasonable application” cases.

Perhaps the most illustrative example is Yarborough v. Alvarado,416541 U.S. 652 (2004). a case evaluating the reasonableness of the state court’s conclusion that a juvenile was not in custody when he gave an un‑Mirandized confession.417See Miranda v. Arizona, 384 U.S. 436, 444 (1966).  In granting relief, the Ninth Circuit found the state court’s decision unreasonable primarily because it failed to consider the defendant’s youth and inexperience with police.418See Alvarado v. Hickman, 316 F.3d 841, 852–54 (9th Cir. 2002), rev’d sub nom. Yarborough v. Alvarado, 541 U.S. 652 (2004). Some language in the Ninth Circuit’s opinion also suggests they found the state court’s ultimate conclusion unreasonable. See id. at 854–55 (“After identifying these relevant circumstances, it is simply unreasonable to conclude that a reasonable 17‑year‑old, with no prior history of arrest or police interviews, would have felt that he was ‘at liberty to terminate the interrogation and leave.’” (quoting Thompson v. Keohane, 516 U.S. 99, 112 (1995))). In the Ninth Circuit’s view, this represented an unreasonable failure to extend the principle “that juvenile defendants are accorded heightened procedural safeguards commensurate with their age and experience”—clearly established in the context of evaluating the voluntariness of a confession or of a waiver of Miranda rights—to the determination of whether the defendant was “in custody.”419Id. at 853.

The Supreme Court reversed.420Alvarado, 541 U.S. at 655. In doing so, it proceeded in two steps. First, it analyzed the state court’s outcome.421Id. at 663–66. Relying on the “general” nature of the custody analysis and the “differing indications” presented by the facts of the interrogation, the Court concluded that “fairminded jurists could disagree over whether [the prisoner] was in custody.”422Id. at 664–65. This analysis mirrors the second path in the two‑path approach: It considered only the facts presented to the state court and the outcome it reached on those facts. These, the Court concluded, were not enough to show unreasonableness. In this Article’s terminology, the state court’s outcome was not enough, by itself, to prove it had committed a qualifying error.

But the Court did not end its analysis there. It next considered the possibility that the state court erred (as the Ninth Circuit held) by failing to consider the prisoner’s age and inexperience with law enforcement.423Id. at 666–69. The Court concluded, however, that this was not unreasonable, largely relying on “an important conceptual difference between the Miranda custody test and the line of cases from other contexts considering age and experience”—namely, the purely objective nature of the former—which made it reasonable for the state court not to consider the defendant’s age and inexperience.424Id. at 667. In Alvarado, the Court entertained a “failure to extend” theory, which a plurality had endorsed four years prior. See Ramdass v. Angelone, 530 U.S. 156, 166 (2000) (plurality opinion) (“A state determination may be set aside under this standard if, under clearly established federal law, the state court was unreasonable in refusing to extend the governing legal principle to a context in which the principle should have controlled.”). Later, however, the Court adopted the rule—advocated by the state in Alvarado, see 541 U.S. at 666—that failure to extend precedent cannot constitute an unreasonable application of clearly established law. See White v. Woodall, 572 U.S. 415, 426 (2014). However, since Alvarado, the Court has itself extended the principle that a juvenile defendant’s age should be considered in the Miranda custody analysis, rendering it part of the clearly established law that a federal court must consider in its Section 2254(d) analysis. See J.D.B. v. North Carolina, 564 U.S. 261, 264–65 (2011). “For these reasons”—but only, apparently, for these reasons—the Court concluded that “the state court’s failure to consider [the defendant’s] age [did] not provide a proper basis for finding that the state court’s decision was an unreasonable application of clearly established law.”425Alvarado, 541 U.S. at 668. Though the Court rejected the asserted qualifying error (apparently for lack of severity426See supra Section III.B.2 (discussing severity requirement).), its analysis reflects the first path of the two‑path approach: carefully considering whether a specific misstep by the state court vitiated Section 2254(d)’s limitations.

The penultimate sentence in Alvarado provides the strongest support for applying the two‑path approach under the “unreasonable application” clause. The prisoner failed to surmount Section 2254(d)’s barrier only because “[t]he state court considered the proper factors and reached a reasonable conclusion.”427Alvarado, 541 U.S. at 669 (emphasis added). If the “could have supported” framework was the exclusive path to proving an unreasonable application, there would have been no need to consider whether the state court considered the proper factors.

Although Alvarado found no qualifying error under either path, other cases demonstrate the first path’s vitality. Take Wiggins for instance.428Wiggins v. Smith, 539 U.S. 510 (2003). There, the Court examined the state court’s “application of Strickland’s governing legal principle” by closely examining (and quoting) its opinion.429Id. at 527–28. It concluded that this application “was objectively unreasonable” because it “merely assumed that [counsel’s] investigation was adequate.”430Id. at 527. Thus, it was the state court’s “assumptions,” not its outcome, that removed Section 2254(d)’s protections. Similarly, in Rompilla, the Court examined “the position taken by the state postconviction courts”—namely, that defense counsel’s other efforts to find mitigating evidence excused their otherwise unreasonable failure to look at the case file from the defendant’s prior prosecution.431Rompilla v. Beard, 545 U.S. 374, 388–89 (2005). It held that this rationale “fail[ed] to answer the considerations” set out in Strickland and its progeny “to the point of being an objectively unreasonable conclusion.”432Id. at 389. Again, it was the reasoning, not just the outcome, that doomed the state court’s decision.

Though these examples suffice to prove the point, such instances are numerous433See Brief of Petitioner at 25 n.12, Wilson v. Sellers, 138 S. Ct. 1188 (2018) (No. 16‑6855), 2017 WL 2472080, at *25 n.12 [hereinafter Wilson Petitioner Brief] (collecting nine examples); Practitioner’s Guide to § 2254(d), supra note 292, at 25–27 (collecting six).—as are examples of the Court examining the state court’s reasoning before finding it reasonable.434Wilson Petitioner Brief, supra note 433, at 26 n.13.

      1.       The Two‑Path Approach and Paragraph (d)(2)

Although the Court’s decisions applying paragraph (d)(2) are cryptic, they nevertheless provide examples of both paths to overcoming Section 2254(d)’s barrier to relief. Following the first path, Brumfield “train[ed] [the Court’s] attention on the two underlying factual determinations on which the [state] court’s decision was premised”—namely, that the defendant’s IQ score was “inconsistent with” a finding of significantly subaverage intellectual function and that “he had presented no evidence of adaptive impairment.”435Brumfield v. Cain, 576 U.S. 305, 313 (2015). Significantly subaverage intellectual functioning and adaptive deficits are, alongside onset during the developmental period, the defining criteria of intellectual disability. Hall v. Florida, 572 U.S. 701, 710 (2014). People with intellectual disabilities are categorically ineligible for the death penalty under Atkins v. Virginia, 536 U.S. 304, 321 (2002). In rejecting the first determination as unreasonable, the Court zeroed in on the state court’s “apparent[] belie[f]” that an IQ score of 75 “precluded any possibility” of subaverage intellectual functioning.436Brumfield, 576 U.S. at 314. In other words, the Court honed in on a specific qualifying error the state court committed in its factfinding.437Although the Court also considered whether there was “evidence of any higher IQ test score that could render the state court’s determination reasonable,” id. at 316, this is best understood as the Court seeking to falsify its inference that the state court had in fact made this assumption, not as hypothesizing alternative bases that could have justified the finding. In likewise rejecting the second determination, the Court zeroed in on the state court’s failure to “take[] into account that the evidence before it was sought and introduced at a time when [the defendant’s] intellectual disability was not at issue”—another specific qualifying error.438Id. at 322; see Ybarra v. Filson, 869 F.3d 1016, 1025 (9th Cir. 2017) (“The [Brumfield] Court relied on these contradictions to conclude that ‘the two underlying factual determinations on which the trial court’s decision was premised’ . . . were unreasonable under § 2254(d)(2).” (quoting Brumfield, 576 U.S. at 313)). Brumfield thus demonstrates that the first path of the two‑path approach is viable under paragraph (d)(2) just as it is under paragraph (d)(1).

Miller‑El v. Dretke,439545 U.S. 231 (2005). by contrast, demonstrates the second path at work. Miller‑El involved a claim of racial discrimination in the prosecutor’s exercise of peremptory strikes.440Id. at 235–36; see Batson v. Kentucky, 476 U.S. 79, 89 (1986). The Court limited its exposition of the state court’s decision to the opinion’s background section, which notes only conclusions, not rationales.441Miller-El, 545 U.S. at 236–37. In its analysis, the Court began by noting that the state court had made a factual finding that the prosecutor’s race‑neutral explanations were true and that this finding triggered Section 2254(d)’s barrier to relief.442Id. at 240. In the following twenty‑five page analysis, the Court does not mention the state court’s decision or analysis even once.443See id. at 240–66. The state court makes its next appearance in the last paragraph of the opinion, which concludes simply that “the state court’s conclusion was unreasonable as well as erroneous.”444Id. at 266 (emphasis added). This prolonged absence is especially striking given the Court’s repeated engagement with the lower federal court’s reasoning.445See id. at 246, 250, 251 n.10, 252, 255 n.14, 257, 265.

Instead of taking issue with the state court’s reasoning, the Court emphasized the “remarkable” evidence446Id. at 240. of the prosecutors’ “consistent pattern of opposition” to Black jurors,447Miller-El, 545 U.S. at 250. which could not “be explained away.”448Id. at 263. It also took aim at the prosecutor’s “incredible” explanations,449Id. at 265. which were “far at odds with the evidence”450Id. and “reek[ed] of afterthought.”451Id. at 246. The Court concluded that “when this evidence . . . is viewed cumulatively its direction is too powerful to conclude anything but discrimination,” which was the “undeniable explanation.”452Id. at 265–66. Accordingly, no court could “reasonably”453Miller-El, 545 U.S. at 247. reach any other conclusion; to do so would simply “blink[] reality.”454Id. at 266.

Though Miller‑El was decided six years before Richter and thus did not invoke its language, there could hardly be a clearer example of the “could have supported” framework in operation. The Court concluded that a qualifying error had occurred solely based on the overwhelming facts and the outcome reached by the state court, implicitly rejecting as unreasonable any hypothetical argument that “could have supported” denying the claim.455Cf. Cash v. Maxwell, 565 U.S. 1138, 1139 (2012) (Sotomayor, J., respecting the denial of certiorari) (describing “an avalanche of evidence demonstrating that the state court’s factual finding was unreasonable”). In the res ipsa loquitur terminology introduced earlier,456See supra notes 401–02 and accompanying text. the Court thought that the rejection of the claim “in the ordinary course of events . . . would not have occurred”457Hake v. George Wiedemann Brewing Co., 262 N.E.2d 703, 705 (Ohio 1970). absent the state court’s unreasonableness.

The combination of Brumfield and Miller‑El thus demonstrates that the two‑path approach is as viable for paragraph (d)(2) as it is for the “contrary to” and “unreasonable application” clauses.

C.       The Two‑Path Approach in Richter and Beyond

What, if anything, did Richter change about the analysis described above? Did its enunciation of the “could have supported” framework render that path—the second in my taxonomy—exclusive? The answers, as it happens, are “very little” and “not at all.” Instead, Richter represents the application of the two‑path approach to a novel evidentiary setting: unexplained orders. As subsequent developments confirm, the first evidentiary path—close examination of the state‑court opinion—remains alive and well. Rumors of its death were, as they say, greatly exaggerated.

      1.       The Two‑Path Approach and Unexplained Decisions

As noted above, every Section 2254(d) case to reach the Court until Mirzayance had a written state‑court opinion to analyze.458See supra note 131 and accompanying text. Richter was thus primarily concerned with the question Mirzayance left open: whether Section 2254(d) applies at all when the state court did not explain its decision.459See Biale, supra note 16, at 1349 (describing this as “the main holding of Richter”). Only after answering this question did the Court turn to the logical follow‑on question—which neither party briefed460Ritter, supra note 37, at 68.—of how to determine whether Section 2254(d)’s requirements were satisfied. Some commentators, focused on the first path to proving a qualifying error, saw this problem as intractable. Professor Judith Ritter, for example, remarked shortly after Richter that “without any information about what precedent the state court applied or the reasoning used when applying precedent to the facts of a specific case, there is no way to decide whether the state court used the correct Supreme Court precedent and/or whether its application of precedent was unreasonable.”461Id. at 67 n.92.

The Court, by contrast, seemed unbothered. Instead, it reminded readers that habeas petitioners bear the burden of proof and calmly pointed to the second path: the “could have supported” analysis.462See Harrington v. Richter, 562 U.S. 86, 98 (2011). In doing so, the Court created nothing new. This Article has already examined one pre‑Richter case applying that framework: Miller‑El.463See supra notes 439–57 and accompanying text. And the Court had granted relief after applying that framework in other cases, including at least one decided under the “unreasonable application” clause.464See Brewer v. Quarterman, 550 U.S. 286, 292–96 (2007) (discussing “the narrowest possible reading” of Supreme Court cases and rejecting all suggested rationales as unreasonable). Thus, Richter did not create the “could have supported” framework, but rather offered a canonical formulation of it and acknowledged that it would be, as a practical matter, the exclusive path in cases where a qualifying error is not clear from the face of the opinion.

Wilson confirmed that Richter’s enunciation of the “could have supported” framework was—even in the summary denial context—a recognition of practical (rather than legal) limitations. The first path to proving a qualifying error requires not just identifying a specific error, but also proving it has occurred.465See Woodford v. Visciotti, 537 U.S. 19, 22 (2002) (per curiam) (acknowledging that Strickland “specifically rejected” the standard denounced by the Ninth Circuit but finding that the Ninth Circuit’s attribution of that standard to the state court was “a mischaracterization of the state‑court opinion”). In Richter, that could not be done because there was no relevant evidence. In Wilson, by contrast, there was relevant evidence: the lower state court’s decision.466In addition to opinions from various levels of the state courts, other evidence might be relevant to determining whether qualifying error occurred. This could include, for instance, arguments presented to the state court and contemporaneous state‑court decisions in other cases. See Dennis v. Sec’y, Pa. Dep’t of Corr., 834 F.3d 263, 353–54 (3d Cir. 2016) (en banc) (Jordan, J., concurring in part and concurring in the judgment); cf. Wilson v. Sellers, 138 S. Ct 1188, 1199 (2018) (Gorsuch, J., dissenting) (“To determine if a reasonable basis ‘could have supported’ a summary denial of habeas relief under Richter, a federal court must look to the state lower court opinion (if there is one), any argument presented by the parties in the state proceedings, and any argument presented in the federal habeas proceeding.”). If Richter’s framework was based not on practical realities but legal requirements, those same requirements would have applied in Wilson, because in both cases the state‑court decision at issue was unexplained. By rejecting the “could have supported” framework under the facts presented there, Wilson thus demonstrated that the framework is an evidentiary reality, not a legal requirement. Richter did not abolish the first path, but merely recognized a practical limitation it always carried.

      1.       The Future of the Two‑Path Approach

Since Richter, the Court has on several occasions signaled that the first path to proving a qualifying error is alive and well. The first such indication came in Wetzel, the case that established the alternative ground doctrine.467See Wetzel v. Lambert, 565 U.S. 520, 520–24 (2012) (per curiam). As noted above, Wetzel cited to only one Section 2254(d) case: Richter.468See supra text accompanying note 226. Although the state court in Wetzel explained its decision, the Court nevertheless quoted Richter’s language describing the “could have supported” framework. But when it did so, it elided the “could have supported” language: “Under § 2254(d), a habeas court must determine what arguments or theories supported . . . the state court’s decision . . . .”469Wetzel, 565 U.S. at 524 (first omission in original) (internal quotation marks omitted) (quoting Harrington v. Richter, 562 U.S. 86, 102 (2011)). This pointed ellipsis suggests that a pure “could have supported” analysis is inappropriate when there is sufficient evidence to reach a conclusion about the state court’s actual reasoning.470Dennis, 834 F.3d at 283; id. at 355 (Jordan, J., concurring in part and concurring in the judgment).

Wilson points in the same direction. The Court explained that “when the last state court to decide a prisoner’s federal claim explains its decision on the merits in a reasoned opinion . . . a federal habeas court simply reviews the specific reasons given by the state court and defers to those reasons if they are reasonable.”471Wilson v. Sellers, 138 S. Ct 1188, 1192 (2018). As the Court explained:

Deciding whether a state court’s decision “involved” an unreasonable application of federal law or “was based on” an unreasonable determination of fact requires the federal habeas court to “train its attention on the particular reasons—both legal and factual—why state courts rejected a state prisoner’s federal claims,” and to give appropriate deference to that decision.472Id. at 1191–92 (citations omitted) (first quoting Hittson v. Chatman, 576 U.S. 1028, 1028 (2015) (Ginsburg, J., concurring in denial of certiorari); and then citing Harrington v. Richter, 562 U.S. 86, 101–02 (2011)).

This language is such strong support for continued scrutiny of state courts’ actual reasoning that one treatise writer thought it “settled the matter.”473Means, supra note 165, § 3:70.

Although not explicit, the strongest indication that the first path remains viable is that the Court has continued to employ it even after Richter. Two of the cases discussed above as examples of the first path—Lafler and Brumfield—postdate Richter. In addition, the two‑path approach unifies and explains the Court’s Section 2254(d) jurisprudence, including lines of that jurisprudence that the Court itself has never explained. First, it explains why Williams examined (and, with regard to the “contrary to” clause, explicitly required lower courts to examine) both the reasoning and the result of the state court: because either can be used to prove that the state court committed a qualifying error. The Court’s repeated statement in the years between Williams and Richter that “either ‘the reasoning [or] the result of the state‑court decision’” can render the decision unreasonable474Panetti v. Quarterman, 551 U.S. 930, 954 (2007) (quoting Early v. Packer, 537 U.S. 3, 8 (2002) (per curiam)); accord Mitchell v. Esparza, 540 U.S. 12, 16 (2003) (per curiam) (“We have held that a state court need not even be aware of our precedents, ‘so long as neither the reasoning nor the result of the state‑court decision contradicts them.’” (quoting Packer, 537 U.S. at 8)); Packer, 537 U.S. at 8 (“Avoiding these pitfalls does not require citation of our cases—indeed, it does not even require awareness of our cases, so long as neither the reasoning nor the result of the state‑court decision contradicts them.”). can be explained on the same ground. Second, it explains why Richter applied a “could have supported” framework: because the second path had always been a feature of Section 2254(d) jurisprudence, and the circumstances in Richter made it, as a practical matter, the exclusive path in that case. Third (and perhaps most importantly), it reconciles Richter and Wiggins. If the state court said that it rejected a multipart claim on one element, a federal court can rule out any explanations for its decision that rely on another element. As a result, the court can make the res ipsa loquitur inference if every remaining explanation involves a qualifying error. If the state court does not say so, the federal court cannot rule out any explanations and thus can make the res ipsa loquitur inference only if all explanations involve qualifying errors. Fourth, the two‑path approach explains Wetzel: Where a reasonable alternative ground exists, the error certainly did not affect the state court’s decision. Finally, it explains Wilson: Unlike the situation in Richter, the situation in Wilson gave some evidence of the state court’s reasoning. Thus, it was not necessary to examine only the state court’s result; courts can (and therefore must) also examine the state court’s reasoning.

Conclusion

For more than a century, federal habeas has featured relitigation of claims already rejected by state courts.475See Bator, supra note 56, at 467 (“[I]n 1873, the Court first announced the rule that habeas corpus may be used to reexamine . . . alleged illegality in the sentence . . . .”). For most of that time, federal courts turned a blind eye to the state court’s decision—much to critics’ consternation. Congress responded by enacting AEDPA, which forced the state courts into the spotlight. Yet federal courts have construed AEDPA to require just what it was designed to prohibit: “disregard” for “a state court’s expressed rationale.”476Woolley v. Rednour, 702 F.3d 411, 422 (7th Cir. 2012). Such constructions are misguided. In Justice Frankfurter’s words, “there is no need for the federal judge, if he could, to shut his eyes to the State consideration.”477Brown v. Allen, 344 U.S. 443, 508 (1953) (opinion of Frankfurter, J.).

Share this article

Twitter
Facebook
LinkedIn