McGill v. Shinn

Full title: Leroy McGill, Petitioner-Appellant, v. David Shinn, Director, Arizona…

Court: United States Court of Appeals, Ninth Circuit

Date published: Oct 21, 2021

16 F.4th 666 (9th Cir. 2021)

Opinion

No. 19-99002

10-21-2021

Leroy MCGILL, Petitioner-Appellant, v. David SHINN, Director, Arizona Department of Corrections; Walter Hensley, Warden, Arizona Department of Corrections - Eyman Complex, Respondents-Appellees.

Jennifer Y. Garcia (argued) and Sara Chimene-Weiss, Assistant Federal Public Defenders; Jon M. Sands, Federal Public Defender; Office of the Federal Public Defender, Phoenix, Arizona; for Petitioner-Appellant. Erin D. Bennett (argued), Assistant Attorney General; Lacey Stover Gard, Deputy Solicitor General/Chief of Capital Litigation; Mark Brnovich, Attorney General; Office of the Attorney General, Phoenix, Arizona; for Respondents-Appellees.

BYBEE, Circuit Judge

Jennifer Y. Garcia (argued) and Sara Chimene-Weiss, Assistant Federal Public Defenders; Jon M. Sands, Federal Public Defender; Office of the Federal Public Defender, Phoenix, Arizona; for Petitioner-Appellant.

Erin D. Bennett (argued), Assistant Attorney General; Lacey Stover Gard, Deputy Solicitor General/Chief of Capital Litigation; Mark Brnovich, Attorney General; Office of the Attorney General, Phoenix, Arizona; for Respondents-Appellees.

Before: Jay S. Bybee, Milan D. Smith, Jr., and Daniel P. Collins, Circuit Judges.

Concurrence by Judge Collins ;

Partial Concurrence and Partial Dissent by Judge Milan D. Smith, Jr.

BYBEE, Circuit Judge:

Petitioner Leroy McGill was sentenced to death in 2004 for the murder of his former housemate, Charles Perez. The Arizona Supreme Court affirmed McGill's conviction and sentence on direct review, and the state trial court denied post-conviction relief. McGill now appeals the district court's denial of his petition for habeas relief under 28 U.S.C. § 2254. The district court granted a certificate of appealability as to trial counsel's performance at the penalty phase of trial but denied a certificate as to McGill's remaining claims.

We evaluate McGill's claims under the Antiterrorism and Effective Death Penalty Act. Applying that deferential standard to his certified claim, we conclude that in denying McGill relief, the state court reasonably applied clearly established federal law and relied upon reasonable factual determinations. We further find that McGill has failed to make a substantial showing that he was denied a constitutional right to effective assistance of counsel at the guilt phase and therefore deny a certificate of appealability as to that claim. We grant a certificate of appealability on McGill's ex post facto claim but deny relief on the merits. Accordingly, we affirm the judgment of the district court.

I. FACTUAL AND PROCEDURAL BACKGROUND

A. The Crime

In July 2002, McGill and his girlfriend, Jonna (Angel) Hardesty, were nearly homeless and temporarily living with a friend in the Sunnyslope area of Phoenix. State v. McGill , 213 Ariz. 147, 140 P.3d 930, 933 (2006) (en banc) ( McGill I ). The couple had recently been ousted from a nearby duplex apartment owned by Jack Yates after Charles Perez, another occupant of the duplex, accused McGill and Hardesty of stealing his shotgun. Id. In the early morning hours of July 13, 2002, McGill walked to the Yates duplex to teach Perez and Yates "a lesson, that nobody gets away with talking about [McGill and Hardesty]." Id. (alterations in original). McGill was confronted by Eddie Keith who lived in the duplex with his wife and two daughters. McGill warned Keith to get his family out of the duplex, and Keith obediently fled. Id. at 933–34.

When McGill entered the duplex, he found Perez sitting on a couch in the living room with his girlfriend, Nova Banta. Id. at 934. McGill admonished the couple not to "talk behind other people's backs," and before either could respond, he doused the pair with gasoline and threw a lit match at them. Id. Flames engulfed Perez and Banta, who ran screaming from the apartment. Id. First responders transported Perez and Banta to the hospital with third-degree burns covering more than seventy-five percent of their bodies. Id. Perez died the following day. Id. Banta survived, identified McGill to her nurse as the man who set her on fire, and later identified McGill at trial. Id. Fortunately, the other residents of the duplex escaped the blaze without injury.

In March 2003, a Maricopa County grand jury indicted McGill on charges of first-degree murder for Perez's death, attempted first-degree murder for his attack on Banta, and several counts of arson and endangerment. Id. Shortly thereafter, the state gave notice of its intent to seek the death penalty under then- Arizona Revised Statute § 13-703.01(B) (2003).¹ The state raised three statutory aggravating factors under Ariz. Rev. Stat. § 13-703(F) (2003): McGill's convictions for armed robbery, a "serious offense" under the statute; McGill's knowing creation of a "grave risk of death" to others during the crime; and McGill's commission of the crime in an especially heinous, cruel, or depraved manner. See id. § 13-703(F)(2), (3), (6).

1 Arizona has since amended and renumbered this portion of its death penalty statutes. Except where specified, we refer to the statutes as they were codified at the time of the crime.

Attorney Maria Schaffer of the Office of the Legal Advocate was appointed to represent McGill, and Elizabeth Todd was appointed as second chair. The defense team also retained Mitigation Specialist Marianne Brewer and private investigator Mark Mullavey. Counsel submitted several expert-witness requests to her supervisor at the Office of the Legal Advocate, Susan Sherwin, including requests for a neuropsychologist, addictionologist, and a domestic violence specialist. Director Sherwin approved the requests for a neuropsychologist and addictionologist but denied Schaffer's request for a domestic violence expert. Ultimately, counsel retained neuropsychologist Dr. Richard Lanyon and addictionologist Dr. Mace Beckson. Schaffer had wanted to retain Dr. Lesley Hoyt-Croft as McGill's addictionologist, but Sherwin refused to authorize it because Hoyt-Croft was a psychologist, not a medical doctor. Only Dr. Lanyon would testify at trial, however, because Dr. Beckson felt he could not offer helpful testimony unless McGill accepted responsibility for his actions, which McGill refused to do.

B. The Trial

McGill was tried in October 2004.² At the guilt phase, the state presented evidence that witnesses saw McGill at the Yates duplex before the fire; McGill warned Keith to flee the duplex; and McGill mixed styrofoam pieces into the gasoline before dousing Perez and Banta, believing it would create a pasty substance that would be more difficult to extinguish. See McGill I , 140 P.3d at 934. After brief deliberations, the jury found McGill guilty on each charge in the indictment. At the aggravation phase, the jury determined beyond a reasonable doubt that sufficient aggravating factors existed under Ariz. Rev. Stat. § 13-703(F) to consider imposition of the death penalty.

2 Capital trials in Arizona are divided into three proceedings: the guilt phase, aggravation phase, and penalty phase. See Ariz. Rev. Stat. § 13-703(A)–(C). At the guilt phase, the jury determines the defendant's guilt or innocence. At the aggravation phase, the jury determines whether sufficient aggravating factors exist to warrant consideration of the death penalty. See id. § 13-703(B). If the state proves aggravating factors beyond a reasonable doubt, the proceeding moves to the penalty phase where the jury determines whether the death penalty is appropriate in light of any mitigating factors. Id. § 13-703(C).

Counsel presented mitigation evidence over the course of a four-day penalty-phase trial. The mitigation presentation focused on McGill's life history, including his dysfunctional childhood and removal from his home; early drug use and juvenile delinquency; drug and alcohol addiction later in life, including chronic methamphetamine use; a purported brain injury suffered in a car accident in the 1980s; and the dysfunctional and abusive relationship between McGill and his girlfriend, Jonna Hardesty. Mitigation Specialist Brewer and several of McGill's family members explained the dysfunction during McGill's childhood. The story was a difficult one. McGill's father was an alcoholic and was physically violent with McGill's mother, Ann. They divorced about the time McGill was born, and Ann took the five children from California to Arizona. McGill's father went to Arizona and took the children from Ann and returned to California. Ann regained custody and returned to Arizona, where she worked multiple jobs to support herself and the children. After another marriage, divorce, and a sixth child, Ann took her children to Texas. While she worked, McGill and his siblings were left to fend for themselves.

The children were removed from the home on several occasions because Ann was unable to care for them. In 1970, when McGill was eight years old, he and his brothers Cordell and Lonnie were placed briefly in foster care. Shortly thereafter, McGill and Cordell were transferred to Buckner's Boys Ranch, a harsh and structured environment, in San Antonio, Texas. Two years later, McGill was released to his mother. But by 1976, Ann was again unable to care for McGill and applied for his admission to Boysville, another all-boys reform school in San Antonio.

Witnesses testified that McGill's fractured home life affected him well into adulthood. McGill began using drugs and alcohol at a young age. By adulthood, McGill was a chronic, daily methamphetamine user. Methamphetamine affected McGill's behavior, sleep patterns, and decision-making ability. Counsel connected McGill's drug use to his criminal history, especially two armed robberies. The defense attempted to mitigate the effects of those convictions by framing them through the lens of substance abuse. To that end, the defense elicited testimony that McGill was intoxicated and nearly homeless at the time, and that he accepted responsibility for the robberies by pleading guilty and serving his sentence. After McGill completed his sentence, he reconnected with family and held a job. Unfortunately, any progress McGill made was undercut by his relationship with Hardesty, who was described by his family as "one of the most evil people" they had ever met. Hardesty also enabled McGill's chronic methamphetamine use, which further disrupted his life.

McGill's expert neuropsychologist, Dr. Richard Lanyon, provided useful context for McGill's cognitive function, history of substance abuse, and relationship with Hardesty. Dr. Lanyon administered a battery of cognitive tests, which revealed slight impairments to McGill's language and symbolic skills development but did not uncover any other noteworthy cognitive deficiencies. Dr. Lanyon also reviewed "quite a lot of records" from McGill's childhood and found that McGill's neglectful and emotionally distant mother made him particularly susceptible to manipulation from women—especially Hardesty. When Hardesty "said jump, [McGill] jumped." According to Dr. Lanyon, McGill's unhealthy dependence on Hardesty was further exacerbated by his chronic methamphetamine use, which likely impaired McGill's judgment leading up to the crime.

The jury was unpersuaded by McGill's mitigation presentation and ultimately returned a sentence of death. The Arizona Supreme Court affirmed the conviction and sentence in a published opinion, with one justice concurring in part and dissenting with respect to a question under the Confrontation Clause. McGill I , 140 P.3d 930 ; id. at 946 (Hurwitz, J., concurring in part and dissenting in part). McGill sought certiorari review of the Confrontation Clause issue, which the United States Supreme Court denied. McGill v. Arizona , 549 U.S. 1324, 127 S.Ct. 1914, 167 L.Ed.2d 570 (2007) (mem.).

C. Post-Conviction Relief

In June 2010, McGill sought post-conviction relief (PCR) in the Maricopa County Superior Court. Raising multiple ineffective assistance claims, McGill argued that his counsel failed, inter alia , to: retain necessary expert witnesses and prepare Dr. Lanyon for his mitigation testimony; subpoena material witnesses and effectively cross-examine others; obtain necessary mitigation evidence; and discover and present evidence that McGill was sexually abused at Boysville. In preparation for the PCR proceedings, McGill underwent a PET scan, which produced digital imaging of his brain function. PCR counsel also retained additional expert witnesses: psychiatrist and brain imagine expert Dr. Joseph Wu, who reviewed McGill's PET scan ; psychiatrist Dr. Richard Rosengard; and pharmacologist Dr. Edward French.

The PCR court summarily denied all but one of McGill's claims by written order in October 2010, but ordered an evidentiary hearing on McGill's challenge to trial counsel's failure to retain experts to explore the relationship between his purported brain injury and his crime. The PCR court held the evidentiary hearing the following October. It heard testimony from four witnesses: lead trial counsel Maria Schaffer, Dr. Wu, Dr. Rosengard, and Dr. Lanyon.³

3 Although Dr. French provided a report in support of McGill's PCR petition, he did not testify at the evidentiary hearing.

Schaffer testified that she had difficulty retaining the experts in addictionology and domestic violence that she felt were necessary to defend McGill and thought that "Dr. Lanyon did a horrible job of preparing for his testimony." She felt that Dr. Lanyon's lack of preparation coupled with her earlier difficulty retaining necessary expert witnesses prevented her from adequately presenting mitigating evidence of McGill's drug addiction, brain injury, and domestic violence at the hands of Hardesty. But Schaffer also admitted that she had deliberately withheld from Dr. Lanyon a pre-sentence report that the state used to discredit his testimony on cross-examination.

Dr. Wu, who was the director of the Brain Imagery Center at the University of California, Irvine, testified that he had not personally examined McGill, but that he had reviewed McGill's PET scan and Dr. Lanyon's report. He explained that a PET scan acts as a "thermometer" for gauging cognitive conditions but added that diagnosing cognitive disorders requires separate neuropsychological testing. Dr. Wu further testified that he would have ordered additional testing to assess the actual effect of McGill's brain injury on his behavior had he reviewed McGill's PET scan prior to trial. Nevertheless, he conceded that Dr. Lanyon had administered a "comprehensive battery" of tests and that the results of those tests were compatible with the abnormalities revealed in McGill's PET scan.

Dr. Rosengard's testimony provided a psychiatric perspective to McGill's cognitive functioning. It also detailed the effects of McGill's relationship with Hardesty, and, for the first time, disclosed that McGill was a victim of childhood sexual assault. He opined that McGill's turbulent upbringing made him especially susceptible to Hardesty's influence and that McGill suffered from Stockholm Syndrome as a result of her manipulations. Dr. Rosengard was not convinced, however, that McGill suffered from cognitive deficiencies as a result of a traumatic brain injury. He testified that his psychiatric evaluation independently revealed that McGill did not suffer any cognitive deficits—the same conclusion that Dr. Lanyon reached.

Dr. Lanyon stood by his trial testimony, stating that he had a "[v]ery good" working relationship with counsel and that neither Dr. Wu's nor Dr. Rosengard's findings would have significantly altered his testimony. He added that, although Dr. Wu's and Dr. Rosengard's reports indirectly supported his conclusions, neither doctor's report corroborated his findings or provided a "smoking gun" regarding McGill's brain damage. In hindsight, Dr. Lanyon conceded that McGill's PET scan would have been helpful in one way; it would have supported his finding that McGill's language functioning was deficient. Aside from that minor area, Dr. Lanyon was unpersuaded that Drs. Wu and Rosengard presented any evidence that would have altered his original analysis.

Following the evidentiary hearing, the PCR court denied McGill's final ineffective assistance claim in a written order. Applying Strickland , the PCR court held that counsel's performance at trial did not fall below an objective standard of reasonableness. The PCR court noted that defense counsel presented "a substantial amount of mitigation" evidence that covered McGill's "dysfunctional family background, his relationship with Ms. Hardesty, and his substance abuse." The PCR court also found that McGill's challenge to Dr. Lanyon's performance was flawed because Dr. Lanyon's "thorough and complete" evaluation of McGill simply did not reveal evidence of "brain related impairment" that counsel hoped it would. Thus, the state court concluded that counsel was not deficient for failing to better present evidence of cognitive deficiency to the jury.

The PCR court also held that even if counsel had been deficient, none of her alleged errors prejudiced McGill. Especially persuasive to the PCR court was Dr. Lanyon's testimony that neither Dr. Wu nor Dr. Rosengard aided his trial testimony in any significant way. Neither expert demonstrated that McGill suffered from any cognitive deficiencies apart from minor language and speech functions. The PCR court also noted that Dr. Wu's and Dr. Rosengard's testimony disagreed on the central issue of whether McGill's prior drug use influenced his PET scan results. At best, their testimony would have been cumulative to Dr. Lanyon's testimony; at worst, the inherent contradictions would have weakened the mitigation presentation. As a result, even if counsel erred in failing to secure their expert opinions, it was not reasonably probable that their testimony would have changed McGill's sentence.

In 2013, McGill sought habeas relief under 28 U.S.C. § 2254. The district court denied his petition in January 2019, but granted McGill a certificate of appealability on his ineffective assistance of counsel claim arising out of trial counsel's investigation and presentation of mitigation evidence at the penalty phase of his trial. McGill timely appealed.

II. SCOPE AND STANDARD OF REVIEW

A. Scope of Review

McGill presents several claims for review, only one of which is certified for appeal. We may not review McGill's uncertified claims unless we grant a certificate of appealability (COA). 28 U.S.C. § 2253(c)(1)(A) ("Unless a ... judge issues a certificate of appealability, an appeal may not be taken to the court of appeals from ... the final order in a habeas corpus proceeding."). We will treat McGill's briefing of his uncertified issues as an application for a COA. Fed. R. App. P. 22(b)(1)–(2) ; Ninth Cir. R. 22-1(e) ; Slack v. McDaniel , 529 U.S. 473, 483, 120 S.Ct. 1595, 146 L.Ed.2d 542 (2000). We may issue a COA "only if the applicant has made a substantial showing of the denial of a constitutional right." 28 U.S.C. § 2253(c)(2).

McGill's lone certified claim arises out of counsel's performance at the penalty phase of his trial. We take up this certified claim in Part III. Of McGill's remaining uncertified claims, one—a claim that counsel was deficient by failing to present the mitigating circumstances of McGill's armed robbery convictions—also arises from counsel's performance at the penalty phase. Although the district court considered this a separate grounds for a COA, and denied it, we have explained that the right to counsel secured by the Sixth and Fourteenth Amendments "is a guarantee of effective counsel in toto ." Browning v. Baker , 875 F.3d 444, 471 (9th Cir. 2017). We must consider "counsel's conduct as a whole to determine whether it was constitutionally adequate." Id. In Browning we observed that separating counsel's alleged errors in a proceeding into different questions and considering whether to issue a COA as to each error "distort[s]" the ineffective assistance of counsel inquiry. The proper procedure is for a district court to consider whether to grant a COA "at a higher level of generality" so that we may consider counsel's performance in the context of the entire proceeding. Id. See White v. Ryan , 895 F.3d 641, 645 n.1 (9th Cir. 2018) (treating counsel's failure to investigate and present mitigating evidence as "a single claim regarding his right to the effective assistance of counsel at the penalty phase of resentencing"). Because the district court granted a COA with respect to other aspects of counsel's performance at the penalty phase, we will treat McGill's claim with respect to the circumstances of his armed robbery as if the district court had granted a COA and consider it in Part III.

McGill also seeks a COA for two other claims. First, McGill asks that we issue a COA to review counsel's deficient performance at the guilt phase at his trial. McGill argues that trial counsel failed to retain an arson expert to rebut evidence that McGill mixed styrofoam into the gasoline before dousing Perez and Banta. Because this alleged omission occurred during the separate and discrete guilt phase, we will consider it apart from McGill's penalty phase claim. For reasons we will explain in Part IV.A, we deny a COA as to his guilt phase claim. Second, McGill challenges his death sentence under the Ex Post Facto Clause in light of Ring v. Arizona , 536 U.S. 584, 122 S.Ct. 2428, 153 L.Ed.2d 556 (2002). Because we believe that "jurists of reason would find it debatable whether the petition states a valid claim of the denial of a constitutional right," we will grant a COA with respect to McGill's claim under the Ex Post Facto Clause. Slack , 529 U.S. at 484, 120 S.Ct. 1595. We consider this claim on the merits in Part IV.B.

B. Standard of Review

Although we review the district court's denial of a § 2254 petition de novo, Ramirez v. Ryan , 937 F.3d 1230, 1240 (9th Cir. 2019), our review of McGill's two certified claims is subject to the Antiterrorism and Effective Death Penalty Act of 1996 (AEDPA), which "guard[s] against extreme malfunctions in the state criminal justice systems, and [is] not ... a means of error correction." Greene v. Fisher , 565 U.S. 34, 38, 132 S.Ct. 38, 181 L.Ed.2d 336 (2011) (citation and internal quotation marks omitted). AEDPA provides that a federal court

shall not ... grant[ ] [a writ of habeas corpus] 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)(1)–(2) ; see Harrington v. Richter , 562 U.S. 86, 97–98, 131 S.Ct. 770, 178 L.Ed.2d 624 (2011). McGill challenges the PCR court's decision under both of § 2254(d) ’s prongs.

We may only grant relief under § 2254(d)(1)

if the state court arrives at a conclusion opposite to that reached by [the Supreme] Court on a question of law or if the state court decides a case differently than [the] Court has on a set of materially indistinguishable facts. Under the "unreasonable application" clause, a federal habeas court may grant the writ if the state court identifies the correct governing legal principle ... but unreasonably applies that principle to the facts of the prisoner's case.

Williams v. Taylor , 529 U.S. 362, 413, 120 S.Ct. 1495, 146 L.Ed.2d 389 (2000). "[C]learly established Federal law" is limited to "the holdings ... of [Supreme] Court[ ] decisions" that existed when the state court issued its decision. Id. at 412, 120 S.Ct. 1495. The "pivotal question" is whether the court's application of law was unreasonable. Richter , 562 U.S. at 101, 131 S.Ct. 770. A state court's application of federal law that is merely incorrect will not warrant relief, Williams , 529 U.S. at 410–11, 120 S.Ct. 1495 ; see White v. Woodall , 572 U.S. 415, 427, 134 S.Ct. 1697, 188 L.Ed.2d 698 (2014) ("[R]elief is available under § 2254(d)(1) ’s unreasonable-application clause if, and only if, it is so obvious that a clearly established rule applies ... that there could be no ‘fairminded disagreement’ on the question.") (citation omitted). Our review of state court factual determinations under § 2254(d)(2) is similarly deferential. We may not disturb the PCR court's factual findings unless they are "objectively unreasonable," Miller-El v. Cockrell , 537 U.S. 322, 340, 123 S.Ct. 1029, 154 L.Ed.2d 931 (2003) ( Miller-El I ), which is "a substantially higher threshold" than merely "believ[ing] the state court's determination was incorrect." Schriro v. Landrigan , 550 U.S. 465, 473, 127 S.Ct. 1933, 167 L.Ed.2d 836 (2007) (citing Williams , 529 U.S. at 410, 120 S.Ct. 1495 ).

III. CERTIFIED CLAIM

The Sixth Amendment provides that "[i]n all criminal prosecutions, the accused shall enjoy the right ... to have the Assistance of Counsel for his defence." U.S. Const. amend. VI. The Supreme Court has said that "the right to counsel is the right to the effective assistance of counsel," McMann v. Richardson , 397 U.S. 759, 771 n.14, 90 S.Ct. 1441, 25 L.Ed.2d 763 (1970), which means that "the accused is entitled to ‘a reasonably competent attorney,’ whose advice is ‘within the range of competence demanded of attorneys in criminal cases,’ " United States v. Cronic , 466 U.S. 648, 655, 104 S.Ct. 2039, 80 L.Ed.2d 657 (1984) (quoting McMann , 397 U.S. at 770–71, 90 S.Ct. 1441 ). In evaluating a claim of ineffective assistance of counsel, Strickland v. Washington , 466 U.S. 668, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984), supplies the "clearly established Federal law" for purposes of § 2254(d)(1). Strickland sets out a two-part test. First, "the defendant must show that counsel's performance was deficient. This requires showing that counsel made errors so serious that counsel was not functioning as the ‘counsel’ guaranteed the defendant by the Sixth Amendment." Id. at 687, 104 S.Ct. 2052. Second, "the defendant must show that the deficient performance prejudiced the defense. This requires showing that counsel's errors were so serious as to deprive the defendant of a fair trial, a trial whose result is reliable." Id. Strickland requires that a defendant prove that his "counsel's conduct so undermined the proper functioning of the adversarial process that the trial cannot be relied on as having produced a just result." Id. at 686, 104 S.Ct. 2052. In reviewing counsel's performance, our scrutiny "must be highly deferential," lest "a court, examining counsel's defense after it has proved unsuccessful, ... conclude that a particular act or omission of counsel was unreasonable." Id. at 689, 104 S.Ct. 2052. A "fair assessment," accordingly, "requires that every effort be made to eliminate the distorting effects of hindsight." Id.

McGill argues that, in denying his claim of ineffective assistance of counsel at the penalty phase, the PCR court misapplied Strickland under § 2254(d)(1) and unreasonably determined the facts under § 2254(d)(2). If so, we would no longer owe deference to the PCR court's determinations, and we would then resolve his ineffective assistance of counsel claims "without the deference [to the PCR court] AEDPA otherwise requires." Panetti v. Quarterman , 551 U.S. 930, 953, 127 S.Ct. 2842, 168 L.Ed.2d 662 (2007) (collecting cases); see also Johnson v. Williams , 568 U.S. 289, 303, 133 S.Ct. 1088, 185 L.Ed.2d 105 (2013) ("AEDPA permits de novo review in those rare cases when a state court decides a federal claim in a way that is ‘contrary to’ clearly established Supreme Court precedent."); Early v. Packer , 537 U.S. 3, 8, 123 S.Ct. 362, 154 L.Ed.2d 263 (2002) (per curiam) (holding that § 2254 does not apply where "the reasoning [or] the result" of the PCR court's decision contradicts Supreme Court precedent (citation omitted)). Whether McGill has satisfied § 2254(d) is thus critical to his claims. We will begin with McGill's claim that we owe no duty of deference under § 2254(d) to the PCR court's decision. Because we conclude that we do, we then review the merits of McGill's ineffective assistance of counsel claim through the lens of § 2254(d).

A. Whether We Owe a Duty of Deference to the PCR Court's Decision

1. Application of clearly established federal law under § 2254(d)(1)

We will first examine the PCR court's application of Strickland . Section 2254(d)(1) outlines two possible avenues to attack the state court's legal analysis: where the state court has misstated on-point Supreme Court precedent, and where the state court has correctly stated the standard, but its application of the precedent is unreasonable. The PCR court cited Strickland and Richter and correctly recited Strickland ’s two-step test. McGill does not assert otherwise; rather, McGill limits his challenge to the second avenue, arguing that the PCR court unreasonably applied Strickland . He raises four claims: (a) "[r]ather than assessing if counsel acted in accordance with prevailing professional norms in investigating and presenting evidence, the [PCR] court focused on the specifics of the un-presented evidence and other mitigation that was presented"; (b) the PCR court asked whether the evidence would have changed "the court's view," rather than "whether this evidence might have affected an objective sentencer"; (c) the PCR court imposed a causal-nexus requirement; and (d) the PCR court failed to assess counsel's cumulative failures.

a. Assessing professional norms

McGill argues that the PCR court failed to evaluate counsel's performance "under prevailing professional norms" as required by Strickland , 466 U.S. at 688, 104 S.Ct. 2052. In Strickland , the Court said that there were no "detailed rules for counsel's conduct" or "checklist for judicial evaluation," but that we should look to "American Bar Association standards and the like" and judge counsel based on "all the circumstances." Id. at 688, 690, 104 S.Ct. 2052. At this most elevated perch on the ladder of abstraction, we are unclear what errors McGill thinks the PCR court committed. The only case McGill points to in support of his claim is our decision in Milke v. Ryan , 711 F.3d 998 (9th Cir. 2013). That case does not advance McGill's claim. In Milke , we granted relief where a PCR court "applied the wrong legal authority" to the petitioner's claim that he was entitled to impeachment evidence under Giglio v. United States , 405 U.S. 150, 92 S.Ct. 763, 31 L.Ed.2d 104 (1972). Milke , 711 F.3d at 1006. We held that a state court unreasonably applies clearly established federal law when it applies the wrong legal standard. Id. at 1006–07.

McGill cannot argue that the PCR court set out the wrong standard. It did not. The PCR court cited Strickland and properly set out its two-step test. Under step one, the PCR court concluded that it could "not find that [Schaffer's] representation fell below an objective standard of reasonableness." The PCR court further held that Strickland ’s "second prong which concerns prejudice [was] not met." To the extent we understand McGill's argument, we find no merit in it.

b. Evaluating new mitigation evidence

McGill claims that the PCR court misapplied Strickland by evaluating whether the mitigation evidence would have affected the court's judgment rather than asking if the evidence could reasonably have changed the outcome for the trier of fact .⁴ In this context, prejudice under Strickland requires a "reasonable probability" that, but for counsel errors, one juror would have voted against the death penalty. Wiggins , 539 U.S. at 537, 123 S.Ct. 2527. The PCR court's duty is to "reweigh the evidence in aggravation against the totality of available mitigating evidence" and determine whether the result of the proceeding would have been different had the new evidence been presented. Id. at 534, 123 S.Ct. 2527 (citing Strickland , 466 U.S. at 694, 104 S.Ct. 2052 ); White v. Ryan , 895 F.3d 641, 670 (9th Cir. 2018). That review is objective, and the state court may not deny relief merely because "it would have imposed a death penalty if it had considered the mitigation evidence." White , 895 F.3d at 670.

4 In this context, McGill claims that the trier of fact was the jury in the first instance and the Arizona Supreme Court in the second. He argues that the PCR court erred because "it focused only on the trial, and ignored that the Arizona Supreme Court had to independently reweigh the aggravation and mitigation." We refuse to fault the PCR court for not conducting two separate inquiries. We are hard pressed to understand how any reviewing court could decide that new mitigation evidence would not persuade a jury, but would persuade a majority of a state supreme court. Cf. Shinn v. Kayer , ––– U.S. ––––, 141 S. Ct. 517, 525–26, 208 L.Ed.2d 353 (2020) (declining to address how the Arizona Supreme Court might have independently weighed the evidence because "the weighing of aggravating and mitigating evidence in a prior published decision is unlikely to provide clear guidance about how a state court would weigh the evidence in a later case.").

We think that McGill has mischaracterized what the PCR court did. McGill highlights two isolated statements from the PCR court's order that he claims are evidence that the PCR court impermissibly relied on its subjective view: that McGill's newly presented evidence was "not significant in the Court's view," and the evidence "would not have been a significant game changer with regard to the outcome." To start, the PCR court's statement that newly presented evidence was insignificant "in the Court's view," does not betray the subjective analysis that McGill claims. As courts we are always asked for "our" opinion. We can do no other than state "in the court's view" what the record supports and what the law requires. Sometimes we are tasked with giving a second-order opinion—determining, for example, whether a set of facts might have altered the jury's view. In this case, it is clear from the context that the PCR court's reference to "the Court's view" was a shorthand expression—the equivalent of saying "the court concludes." Nothing in the PCR court's opinion suggests that it had stepped outside its role to say what it would have decided if it had been the jury or the Arizona Supreme Court. The court fully explained its reasoning. There is no constitutional error here. Nor can we find any error in the PCR court's reference to whether the evidence was a "game changer."

c. Causal-nexus test

McGill argues that the PCR court impermissibly conditioned relief on his ability to prove a causal nexus between his brain injury and his actions. We have previously explained the troubled history of the causal-nexus test in Arizona:

Beginning in the late 1980s, [the] Arizona Supreme Court developed a "causal nexus" test for nonstatutory mitigation. Under this test ... evidence of a difficult family background or a mental condition was not in and of itself relevant mitigating evidence. As a matter of Arizona law, such evidence was relevant for mitigation purposes only if it had some causal effect contributing to the defendant's behavior in the commission of the

crime at issue. Thus, while the defendant could submit evidence of his difficult family background or mental condition, the sentencing court was prohibited from treating it as legally relevant mitigation evidence unless the defendant proved a causal connection between his background or disorder and the crime. In capital cases from the late 1980s to the mid-2000s, the Arizona Supreme Court repeatedly articulated this causal nexus test for nonstatutory mitigation. The test was "contrary to ... clearly established Federal law, as determined by the Supreme Court of the United States" in Eddings [v. Oklahoma , 455 U.S. 104 [102 S.Ct. 869, 71 L.Ed.2d 1] (1982) ].

McKinney v. Ryan , 813 F.3d 798, 813 (9th Cir. 2015) (en banc). Accordingly, after Eddings , it is constitutional error for a trial court to exclude mitigation evidence solely because the defendant cannot show a causal nexus between the evidence and his crime. See 455 U.S. at 113–14, 102 S.Ct. 869 (holding that a sentencing court cannot "refuse to consider, as a matter of law , any relevant mitigating evidence"). But Eddings does not hold that evidence of a causal nexus is irrelevant to the trier of fact. As we said in McKinney , "[o]nce the jury has heard all the defendant's mitigation evidence, there is no constitutional prohibition against the State arguing that the evidence is not particularly relevant or that it is entitled to little weight." McKinney , 813 F.3d at 817 (quoting with approval State v. Anderson , 210 Ariz. 327, 111 P.3d 369, 392 (2005) ). Moreover, "the failure to establish such a causal connection may be considered in assessing the quality and strength of the mitigation evidence ." Id. at 818 (emphasis added) (quoting with approval State v. Newell , 212 Ariz. 389, 132 P.3d 833, 849 (2006) ). Thus, after Eddings , the sentencing court may not exclude mitigation evidence because of a lack of a causal nexus, but the prosecutor may argue to the jury that such evidence is not deserving of any weight. See 455 U.S. at 114–15, 102 S.Ct. 869.

Here, McGill argues that the PCR court—not the trial court—erred by referring to the lack of a "causal nexus." In addressing that issue, the PCR court wrote:

The nexus between the defendant's mental condition and his actions on the night of the murder would not have been significantly strengthened by the testimony of Dr. Wu and Dr. Rosengard, and thus the Court concludes that there would not have been a reasonable probability that the result of the proceedings would have been different.

Schaffer testified at the PCR hearing that she was trying to connect the dots between McGill's head injury and his crime: "she wanted more mitigation to give a nexus to explain that the defendant had brain damage[,] which might explain his violent behavior on the night in question." The PCR court's findings are consistent with the record:

Dr. Lanyon was clear in his testimony that Dr. Wu's report would have only helped him in one small minor way, that is[,] being consistent with his finding that [McGill] has some speech and language deficiencies. He testified that Dr. Wu's report was no "smoking gun" by any means but was simply not inconsistent with his findings. He also testified that Dr. Rosengard's findings did not assist him in any way and would not have added to the findings that he presented to the jury.

... Further, had Dr. Wu and Dr. Rosengard testified at trial, they would have completely contradicted one another with regard to the effect of substance abuse upon the brain and whether or not

the damage is permanent and irreversible. This would have been an opening for the State to poke holes in the mitigation and would have hurt the defense's presentation. The Court does not find that had Dr. Wu and Dr. Rosengard testified during the mitigation phase of the trial that there likely would have been a different result. At best, their testimony would have been cumulative and, at worst, would have contradicted each other and weakened the mitigation presentation.

Read in context, the PCR court's brief conclusion that "[t]he nexus between [McGill's] mental condition and his actions on the night of the murder would not have been significantly strengthened by the testimony of Dr. Wu and Dr. Rosengard" clearly goes to its weight and not to its admissibility. The court's order discounted the persuasive value—not the relevance—of McGill's evidence. Schaffer's trial strategy was to create a nexus between McGill's brain damage and the crime—evidence that would have been far more persuasive than simply proving that McGill had suffered a head injury. It was thus reasonable for the PCR court to comment on the new evidence related to such a "nexus." And the PCR court reasonably concluded that even if McGill's PCR-stage evidence was admitted, its lack of nexus to his crime rendered it less persuasive in light of the contradictory evidence. Such an assessment is consistent with Eddings .

d. The cumulative effect of counsel's errors

Finally, McGill argues that the PCR court improperly applied Strickland ’s prejudice prong by evaluating his PCR-stage evidence piecemeal instead of considering whether its cumulative effect would have sufficiently undermined confidence in the jury's death sentence. To assess prejudice, the PCR court was required to reweigh the aggravation evidence against the newly presented mitigation evidence, whether "adduced at trial ... [or] in the habeas proceeding." Williams , 529 U.S. at 397–98, 120 S.Ct. 1495 ; see Wiggins , 539 U.S. at 534, 123 S.Ct. 2527. The purpose is to determine whether, considering counsel's errors, there is a reasonable probability that the result would have been different. Strickland , 466 U.S. at 694, 104 S.Ct. 2052.

A PCR court, however, need only assess prejudice if counsel's performance was deficient. A court "cannot consider the cumulative effect of non -errors." Williams v. Filson , 908 F.3d 546, 570 (9th Cir. 2018) ; see Strickland , 466 U.S. at 697, 104 S.Ct. 2052 ("[T]here is no reason for a court deciding an ineffective assistance claim to approach the inquiry in the same order or even to address both [the performance and prejudice] components of the inquiry if the defendant makes an insufficient showing on one."). McGill faults the PCR court for limiting its prejudice determination to the effect that Dr. Wu's and Dr. Rosengard's testimony would have had on the jury during the penalty phase. But, for the reasons discussed, the PCR court reasonably determined in two separate orders that counsel's representation at trial satisfied Strickland ’s objective standard of reasonableness. At that point, the prejudice evaluation was unnecessary, although the PCR court proceeded to step two of Strickland anyway. It would make little sense to require the PCR court to consider the cumulative effect of deficient decisions when the PCR court did not find any particular deficiency; such a requirement asks the court to take a pointillist view of counsel's performance—to see if the court can assemble a picture from indistinct impressions. To be sure, Strickland requires a reviewing court to consider "the totality of the evidence," but that holistic inquiry is a means of "taking due account of the effect of the errors ." Strickland , 466 U.S. at 695–96, 104 S.Ct. 2052 (emphasis added). If there are no errors, there is no need to consider their cumulative effect. The PCR court therefore reasonably applied Strickland ’s objective standard to trial counsel's performance.

2. Determinations of fact under § 2254(d)(2)

Section 2254(d)(2) imposes a "daunting standard" to disrupt a state court's factual findings, which precludes relief in all but "relatively few cases." Taylor v. Maddox , 366 F.3d 992, 1000 (9th Cir. 2004), abrogated on other grounds by Murray v. Schriro , 745 F.3d 984, 999–1000 (9th Cir. 2014). We have recognized "several flavors" of unreasonable factual determinations that may satisfy § 2254(d)(2). Id. These include a state court plainly misapprehending or misstating the record, id. at 1001 (citing Wiggins v. Smith , 539 U.S. 510, 528, 123 S.Ct. 2527, 156 L.Ed.2d 471 (2003) ); failing to consider key aspects of the record, id. at 1008 (citing Miller-El I , 537 U.S. at 346, 123 S.Ct. 1029 ); and ignoring "highly probative" evidence supporting the petitioner's claim, id. at 1001. McGill identifies three PCR-court factual findings that, he claims, meet those "flavors" of erroneous fact finding: (a) that counsel's decision not to call addictionologist Dr. Beckson was tactical; (b) that McGill failed to substantiate his claims of childhood sexual assault; and (c) that Dr. Lanyon was a qualified expert witness upon whom Schaffer was entitled to rely. We conclude that these findings were reasonable.

a. Addictionologist testimony

The record supports the PCR court's determination that counsel's "tactical decision not to call Dr. [Beckson] was reasonable and does not constitute ineffective assistance of counsel." McGill challenges this finding with two arguments: first, that McGill's lead counsel did not make a "tactical" decision not to call an addictionologist because she was denied her addictionologist of choice; and, second, that when Schaffer decided not to call Dr. Beckson, she failed to seek additional funding for a different addictionologist.

As Schaffer was preparing her case, she approached Office of Legal Affairs Director Sherwin for funding to retain an addictionologist. Counsel's preferred addictionologist was Dr. Lesley Hoyt-Croft, but Sherwin denied Schaffer's request because Dr. Hoyt-Croft was not a medical doctor.⁵ Instead, Schaffer hired Dr. Mace Beckson. McGill provides no explanation to support why Director Sherwin's preference to retain a medical doctor was unreasonable. Sherwin was not counsel of record, but as Schaffer's supervisor, she had some responsibility for managing the cases handled by the Office of Legal Affairs. When Dr. Beckson informed Schaffer that he could not testify unless McGill admitted his involvement in the crime, Schaffer decided against calling Beckson as a witness. That left her with Dr. Lanyon as McGill's only expert witness. The PCR court found that Schaffer's decision was "tactical."

5 The record is not precise on Sherwin's reasons. Nothing in the record suggests that Sherwin thought that Hoyt-Croft was not qualified.

McGill argues that the PCR court's finding is unreasonable. The PCR court found that "[Schaffer] did not call Dr. [Beckson] to testify because she ‘ultimately opined that he could not assist in the case because Mr. McGill would not admit his guilt.’ " McGill argues that the finding is clearly erroneous because Schaffer did not "opine" that Dr. Beckson could not testify; it was Dr. Beckson who "opined" he could not testify. According to McGill, the PCR court's finding should read "he" rather than "she." Assuming that this is indeed an error, we cannot tell whether this is an error in perception or transcription by the PCR court. But, in either case, it does not matter. Whether Dr. Beckson told Schaffer that he would not testify as an addictionologist without McGill admitting that he committed the crime, or whether Schaffer determined that she would not put Beckson on the stand, the context for the PCR court's finding is clear: Schaffer knew she could not put Beckson on the stand because he would not be an effective witness. The PCR court found the obvious: "[counsel] conceded that she did get an addictionologist but that she did not want to use that particular expert." In that sense, her decision not to force an ineffective expert witness to testify was "tactical."

That finding was especially reasonable in light of counsel's purpose for calling an addictionologist in the first place, which was to demonstrate the central role that McGill's drug abuse played in his behavior at the time of Perez's murder. In essence, the addictionologist was there to portray McGill's drug use as a mitigating factor for his crime instead of an aggravating factor. But McGill's refusal to admit any involvement in Perez's murder hindered Dr. Beckson's ability to tie McGill's drug use to his actions leading up to the crime. Dr. Beckson was, therefore, justified in his professional assessment that "he could not assist in the case because Mr. McGill would not admit his guilt." From there, counsel's options were limited. Schaffer could subpoena Dr. Beckson to testify despite his stated reservations or she could decline to call Dr. Beckson at all. Because Dr. Beckson's testimony would not have been helpful to the defense, it was reasonable for the PCR court to conclude that counsel's decision was tactical.

McGill argues that once Schaffer realized she could not call Dr. Beckson, she should have secured a different addictionologist and that it was ineffective assistance of counsel to fail to do so. Nothing in Schaffer's decision to proceed without Dr. Beckson suggests she was ineffective because she did not procure a different addictionologist. A different addictionologist might have been willing to testify anyway, but that is not Schaffer's failing. She had sought funding for a qualified addictionologist, and Sherwin had approved funding for Dr. Beckson. Nothing in the Sixth Amendment guarantees McGill his choice of addictionologists, or an addictionologist who will testify favorably to him. See Ake v. Oklahoma , 470 U.S. 68, 83, 105 S.Ct. 1087, 84 L.Ed.2d 53 (1985). Although Schaffer made clear that she wanted additional resources, in practice, attorneys must often decide how to use limited resources when confronted with these evidentiary "dead ends." Carter v. Davis , 946 F.3d 489, 524 (9th Cir. 2019) (per curiam); see Bobby v. Van Hook , 558 U.S. 4, 11, 130 S.Ct. 13, 175 L.Ed.2d 255 (2009) (per curiam).

The PCR court also found that McGill had not shown "how any other addiction specialist would have testified without [McGill's] admission." McGill argues that this finding is objectively unreasonable and points to Dr. French's PCR-stage report that an addictionologist could have testified regardless of McGill admitting guilt. Dr. French's report does not change our view of the PCR court's findings. Even if we accepted that some other addictionologist could have testified for McGill, the PCR court's decision rested on the tactical nature of counsel's decision not to call Dr. Beckson and not whether another addictionologist could have testified. At that point, McGill relied on Dr. Lanyon to put his drug use in context. Dr. Lanyon testified at the penalty phase of the trial about McGill's chronic methamphetamine use and that such use generally "makes [the user] paranoid, actively paranoid, and seriously impairs their judgment." Indeed, he said, such drug use "would remove any remaining fragment of ability to reason." The PCR court addressed Dr. French's report and found that Dr. French's report "would add nothing to Dr. Lanyon's testimony" because Dr. French "did not evaluate [McGill] so any testimony regarding [McGill's] substance abuse history, including amounts used on the night in question, history of abuse and brain damage is speculative." This finding is supported in the record, as Dr. French's report provided similar statements regarding chronic methamphetamine use as did Dr. Lanyon's. Like Dr. Beckson, neither Dr. French nor Dr. Lanyon could connect McGill's drug use to his actual behavior. Thus, regardless of whether a different addictionologist could have testified absent McGill's admission of guilt, the only evidence in the record suggests that testimony would have been substantially similar to Dr. Lanyon's. The PCR court reasonably determined that Dr. French's report added nothing to Dr. Lanyon's testimony and that counsel's tactical decisions did not constitute ineffective assistance.

b. Unsubstantiated sexual abuse

McGill claims that counsel was ineffective for failing to investigate a sexual assault McGill said he experienced at Boysville. At the PCR stage, McGill alleged for the first time that he suffered two sexual assaults while at Boysville and that those assaults would have been persuasive mitigating evidence if presented to the jury. The PCR court rejected McGill's allegations of sexual assault as unsubstantiated. McGill now contends that the PCR court ignored credible evidence of the sexual abuse, including Dr. Rosengard's PCR-stage report and the account of McGill's brother Lonnie, explaining his own sexual abuse at Boysville.

Evidence of sexual abuse can be powerful evidence "relevant to assessing a defendant's moral culpability." Wiggins , 539 U.S. at 535, 123 S.Ct. 2527 ; see Wharton v. Chappell , 765 F.3d 953, 978 (9th Cir. 2014). At the § 2254(d)(1) stage, the inquiry is not whether counsel reasonably investigated the sexual abuse, but whether the PCR court's conclusion that McGill failed to substantiate the abuse was reasonable in light of the new evidence. McGill offered no evidence at trial that he was sexually abused as a child and did not disclose sexual abuse to anyone until his 2009 evaluation with Dr. Rosengard—nearly a decade after Perez's murder. Yet, McGill had the opportunity to disclose the purported abuse as early as 2003, when Dr. Lanyon performed his pre-trial evaluation. Dr. Lanyon's report noted:

Mr. McGill was questioned regarding abuse as a child. He stated that his mother's second husband (his step-father) physically abused him on one occasion. ... Apparently this man frequently threatened physical violence and was very intimidating, but actually hit Mr. McGill only once. Mr. McGill denied any sexual abuse and has never considered that he was emotionally abused.

Dr. Lanyon's report and trial testimony demonstrate that McGill had no reservations about disclosing physical abuse he suffered at the hands of his step-father. Yet, there is no indication that McGill suffered the type of sexual abuse that McGill now alleges. The PCR court was not required to accept McGill's delayed allegation of abuse in light of his earlier denial that such abuse ever happened. Lonnie's PCR-stage affidavit describing his own sexual abuse at Buckner's Boys Ranch does not render the state court's determination unreasonable. McGill asserts that Lonnie's affidavit was sufficiently corroborative of his own sexual abuse to have prompted an evidentiary hearing. But Lonnie's account does not support McGill's allegations of abuse. Lonnie claimed that the sexual assaults occurred at Buckner's Boys Ranch, while McGill claims that the abuse occurred at Boysville. McGill attempts to dismiss the discrepancy due to Dr. Rosengard's limited interaction with McGill. But the inconsistencies between the two brothers’ reports support the PCR court's conclusion that McGill did not substantiate the sexual abuse. Even setting aside those discrepancies, counsel could not have known about the abuse Lonnie suffered because Lonnie was generally unhelpful to the investigation. As Schaffer testified, and the PCR court accepted, Lonnie "absolutely refused to cooperate in coming to court." Indeed, Schaffer testified that Lonnie threatened her and "his brother's case should he be forced to appear." Under those circumstances, it might have been ineffective assistance of counsel to have called Lonnie as a witness; at the very least, it was a strategic decision. We agree with the PCR court that her decision not to call Lonnie as a witness was "tactically sound." In light of McGill's earlier denials that any sexual abuse occurred and the equivocal nature of the new evidence, it was not "an unreasonable determination of the facts," 28 U.S.C. § 2254(d)(2), for the PCR court to conclude that McGill "failed to substantiate his claim that he was sexually abused."

c. Expert witness qualifications

McGill claims that Schaffer was ineffective because she allowed Dr. Lanyon to testify despite her concerns with his preparation and testimony. During counsel's PCR testimony, she testified that Dr. Lanyon "did a horrible job of preparing for his testimony" and that he "was not qualified for the tasks presented in Mr. McGill's case." The PCR court disagreed, finding that Dr. Lanyon was "a qualified expert and [that counsel] was entitled to rely on the competency of his evaluation." McGill now argues that the PCR court's focus on whether counsel was reasonably entitled to rely upon Dr. Lanyon's testimony misses the point. McGill claims that because Schaffer was forced to use Dr. Lanyon, she could not have made a tactical decision to rely on his testimony. McGill also argues that the PCR court's finding that Dr. Lanyon was qualified to testify was unreasonable because Dr. Lanyon was an "all-purpose" expert, not tailored to McGill's case.

The PCR court's finding that Dr. Lanyon was qualified to testify was reasonable. Dr. Lanyon was a professor of psychology at Arizona State University, a position he had held since 1975. He was board certified in clinical and forensic psychology with a specialty in psychological assessment, all of which were important to McGill's defense. He also had extensive experience testifying in capital and other cases for both the state and the defense, but more often for the defense. See, e.g. , Sansing v. Ryan , 997 F.3d 1018, 1037 (9th Cir. 2021) ; Bible v. Ryan , 571 F.3d 860, 869 (9th Cir. 2009) ; Boggs v. Shinn , No. CV-14-02165-PHX-GMS, 2020 WL 1494491, at *41 (D. Ariz. March 27, 2020) ; Morris v. Ryan , No. CV-17-00926-PHX-DGC, 2019 WL 1858137, at *5 (D. Ariz. April 25, 2019) ; Newell v. Ryan , No. CV-12-02038-PHX-JJT, 2019 WL 1280960, at *5 (D. Ariz. March 20, 2019) ; State v. Young , No. 1 CA-CR 13-0429, 2014 WL 6790746, at *1 (Ariz. Ct. App. Dec. 2, 2014) ; State v. Carr , No. 1 CA-CR 07-1046, 2009 WL 1879494, at *2 (Ariz. Ct. App. June 30, 2009). McGill's classification of Dr. Lanyon as an "all-purpose" witness "not tailored" to his specific case rings hollow in light of Dr. Lanyon's experience and credentials.

Furthermore, the record supports a conclusion that Schaffer made a tactical decision to go forward with Dr. Lanyon's testimony even though she had reservations about the strength of his conclusions. Schaffer made clear in her PCR hearing testimony that she was not happy with his performance on the stand. She had had reservations at the time of trial about his forthcoming testimony because he had told her that the case was "challenging," "he didn't like it," and "the facts of the case [were] horrendous; they're overwhelming." This is not the language of incompetence; it is an expert witness who was not impressed with the results of his testing. Dr. Lanyon's written report cited a series of tests he conducted in the approximately seven hours he spent with McGill. He concluded that "Mr. McGill's scores on these tests were all in the average range. These results suggest that he does not currently suffer any cognitive deficits (that is, his ability to use his thinking processes) as a result of brain impairment." When Schaffer was asked by the state about this conclusion, she admitted that "the problem for [her] as a defense lawyer, reading that kind of assessment, [was] that it's not very useful ... to make Mr. McGill sympathetic to the jury during the penalty phase. ... It was a very difficult case to defend." It may be an understatement to say that Dr. Lanyon's report was not as favorable as counsel wished. That counsel went forward with the only expert witness she had is evidence of the weakness of her case and not her ineffective assistance.⁶

6 We address in Part III.B.3 infra the claim that Schaffer was ineffective in her own preparation of Dr. Lanyon.

* * *

We conclude that McGill has not met § 2254(d) ’s high bar. The PCR court correctly identified and reasonably applied clearly established federal law, and its conclusions did not rely on unreasonable determinations of fact. McGill's claims remain subject to AEDPA's deferential standard of review.

B. AEDPA Review of McGill's Ineffective Assistance of Counsel Claim

Having determined that AEDPA review is appropriate, we turn to the merits of McGill's multifaceted ineffective assistance claim. To assess the constitutional sufficiency of counsel's performance, we compare counsel's actions at trial with the prevailing professional norms of the time. This is an objective test, and we must be cautious not to allow hindsight to color our evaluation. Strickland , 466 U.S. at 680, 689, 104 S.Ct. 2052. The standard is "necessarily a general one" and reflects counsel's complex responsibility to "take account of the variety of circumstances faced by defense counsel." Van Hook , 558 U.S. at 7, 130 S.Ct. 13 (quoting Strickland , 466 U.S. at 688–89, 104 S.Ct. 2052 ). We therefore begin our analysis with a "strong presumption" that counsel's decisions reflect "reasonable professional judgment." Cullen v. Pinholster , 563 U.S. 170, 190, 196, 131 S.Ct. 1388, 179 L.Ed.2d 557 (2011) (quoting Strickland , 466 U.S. at 689–90, 104 S.Ct. 2052 )). Moreover, because our review under both AEDPA and Strickland ’s standards are highly deferential to the PCR court's underlying decision, our review here must be "doubly deferential." Id. at 190, 131 S.Ct. 1388 (citation omitted). Surmounting such a "high bar is never an easy task." Padilla v. Kentucky , 559 U.S. 356, 371, 130 S.Ct. 1473, 176 L.Ed.2d 284 (2010).

McGill's ineffective assistance of counsel claim arises from counsel's purported failure at the penalty phase to (1) develop a "relationship of trust" with McGill and his family; (2) obtain "classic sources" of mitigation evidence; (3) prepare Dr. Lanyon for his testimony; and (4) uncover and present evidence of McGill's substance abuse, sexual assault, and domestic violence.

1. Development of a relationship of trust

McGill claims that "Schaffer and her team failed to develop a relationship of trust or a rapport with McGill and his family," and attributes McGill's refusal to share mitigating information to that failing. The PCR court concluded that it could "not find that Ms. [Schaffer's] representation fell below an objective standard of reasonableness."

The PCR court's conclusion is not contrary to clearly established Supreme Court precedent. Nothing in the Sixth Amendment suggests that an accused is entitled to "rapport" with his attorney, and McGill has not directed us to any case from the Supreme Court establishing such a proposition. Indeed, McGill's argument is not only not supported by Supreme Court authority, it is contrary to Morris v. Slappy , 461 U.S. 1, 14, 103 S.Ct. 1610, 75 L.Ed.2d 610 (1983). There, the Supreme Court held that the Sixth Amendment right to counsel did not guarantee criminal defendants "a meaningful attorney-client relationship." Id. at 13, 103 S.Ct. 1610 (emphasis omitted). The Court explained that "[n]o court could possibly guarantee that a defendant will develop the kind of rapport with his attorney" necessary to meet such a standard. Id. Accordingly, the Court "reject[ed] the claim that the Sixth Amendment guarantees a ‘meaningful relationship’ between an accused and his counsel." Id. at 14, 103 S.Ct. 1610 (footnote omitted).

Perhaps sensing the lack of any Supreme Court authority, McGill pivots to challenge the amount of time counsel spent with McGill and his family. He claims that Schaffer met with McGill for less than ten hours. He provides no further context, cites no cases establishing a standard for measuring time with a client, and refers only to the broadest of ABA guidelines for the criminal defense bar. That is not sufficient to satisfy his burden under AEDPA. The Court has consistently cautioned us against imposing a mechanical standard by which counsel may be evaluated. See Van Hook , 558 U.S. at 7, 130 S.Ct. 13 (recognizing that Strickland ’s standard was "necessarily a general one"); see also Strickland , 466 U.S. at 689, 104 S.Ct. 2052 (finding that formulaic rules or duties "interfere with the constitutionally protected independence of counsel"). We must consider whether the time counsel spent in consultation reflects "reasonable professional judgment." Strickland , 466 U.S. at 690, 104 S.Ct. 2052. There is no evidence that ten hours of consultation was unreasonable or that it undermined McGill's defense. Counsel's strategy was to focus on mitigation, which might not have required extensive time with McGill. The defense team traveled to interview McGill's family, obtained hundreds of pages of records, and presented four days of mitigation testimony.

Nor is there any evidence that the defense team violated McGill's trust or that of his family. McGill cannot escape the fact that, as Schaffer put it, much of his family was "either dishonest or not cooperative" in answering counsel's questions. Given the depth and breadth of mitigation evidence counsel presented, the PCR court reasonably concluded that counsel's performance did not fall below prevailing professional norms.

2. Investigation of additional mitigation records

McGill claims that Schaffer failed to obtain "classic" sources of mitigation evidence, such as McGill's juvenile adjudications, police reports, marriage and divorce records, school test scores, and his siblings’ arrest records. The PCR court noted that McGill's mitigation specialist "testified for three days on the witness stand when she told [McGill's] life story including his difficult childhood, drug abuse, and alleged domestic violence. All these areas were explored in detail." The court also concluded that counsel "present[ed] to the jury a substantial amount of mitigation" evidence with respect to "substance abuse .... [his] dysfunctional family background, [and] his relationship with Ms. Hardesty." That included "evidence that [McGill] had an abusive childhood; that he was psychologically immature and, as a result, his girlfriend had greater than normal influence over him; that he suffered from some degree of mental impairment ; that he performed well in institutional settings; and that his family cares about him." The court also found that Dr. Lanyon "reviewed substantial relevant information concerning Defendant McGill's background." From this, the PCR court concluded that it could "not find that Ms. [Schaffer] was ineffective in presenting these issues to the jury."

"[M]itigation evidence [can] complete, deepen, or contextualize the picture of the defendant presented by the prosecution [and] can be crucial to persuading jurors that the life of a capital defendant is worth saving." Allen v. Woodford , 395 F.3d 979, 1000 (9th Cir. 2005). To that end, penalty-phase counsel must conduct a thorough investigation into the relevant mitigation evidence. See Williams , 529 U.S. at 396, 120 S.Ct. 1495. Counsel need not "scour the globe on the off chance something will turn up," Rompilla v. Beard , 545 U.S. 374, 385, 125 S.Ct. 2456, 162 L.Ed.2d 360 (2005), but it is nonetheless obligated "to conduct a thorough investigation of the defendant's background." Williams , 529 U.S. at 396, 120 S.Ct. 1495 ; see Boyde v. California , 494 U.S. 370, 382, 110 S.Ct. 1190, 108 L.Ed.2d 316 (1990).

The record amply supports the PCR court's findings. McGill's counsel presented extensive mitigation evidence and diligently attempted to discover much more. Mitigation Specialist Brewer obtained documents spanning much of McGill's life, including McGill's elementary and middle school records, foster care placement records, records from Buckner's and Boysville, records of his convictions and incarcerations, and post-incarceration employment records. Brewer also attempted to recover other records but was thwarted by factors outside her control. For example, a hospital record-retention policy prevented her from obtaining medical records to corroborate McGill's head injury, and Buckner's Boys Ranch provided all of McGill's available records, which happened to be incomplete. McGill's public school records also proved difficult to obtain because the schools often did not keep full records and regularly forwarded records to new schools when McGill and his family moved. Nevertheless—and despite the difficulty in obtaining complete records—Brewer filled the gaps in the mitigation record with interviews and testimony from McGill's family and friends. Counsel and the defense team as a whole endeavored to uncover as much mitigation evidence as possible, and it is unclear what else they could have done to supplement the mitigation records they did obtain. We are not persuaded that their efforts were deficient.⁷

7 Dr. Lanyon's penalty-phase testimony further illuminates the breadth of mitigation evidence that the defense team recovered. He testified that counsel provided "quite a lot of school records, junior high school records and records from the institutions [McGill] was put in as a child ... and ... interview notes from interviews of a variety of people who knew [McGill]." The Arizona Supreme Court echoed Dr. Lanyon's sentiment two years later, finding the amount of mitigation evidence presented was "not insignificant." McGill I , 140 P.3d at 945.

Nor is there any evidence that the mitigation records McGill now seeks were readily available. Without citations to the record, McGill alleges that there are juvenile records, boys’-home records, and elementary school records that would have aided the defense. We do not know what these documents are, and he offers no explanation for how counsel overlooked or otherwise should have discovered them. McGill's speculation that these records would have aided his mitigation effort is unpersuasive given the extensive testimony presented during the penalty phase.

3. Preparation of Dr. Lanyon

McGill also claims that Schaffer did not adequately prepare Dr. Lanyon. The PCR court found that counsel was not deficient for failing to hire additional mental-health experts and that she adequately prepared Dr. Lanyon for his penalty-phase testimony. Both conclusions were reasonable.⁸

8 To the extent that McGill renews his argument that it was unreasonable for counsel to rely on Dr. Lanyon's testimony because she was forced to use him as an expert witness, that argument fails for the reasons discussed in Part III.A.2.c. Dr. Lanyon was qualified to evaluate McGill and had extensive experience testifying in capital cases. In any event, criminal defendants do not have a constitutional right "to choose [an expert] of his personal liking or to receive funds to hire his own." Ake , 470 U.S. at 83, 105 S.Ct. 1087.

We will accept as a general proposition that counsel is under a duty to prepare witnesses for their testimony. At the same time, "there is no expectation" that an attorney prepare for every possible contingency; she need not be a "flawless strategist or tactician," and "may not be faulted for a reasonable miscalculation or lack of foresight or for failing to prepare for what appear to be remote possibilities." Richter , 562 U.S. at 110, 131 S.Ct. 770.

Much of McGill's frustration with Dr. Lanyon arises out of the state's effective cross-examination during his penalty-phase testimony. Before he submitted his written report, Dr. Lanyon examined McGill, conducted various psychological tests, and reviewed extensive records provided to him by counsel. At trial, counsel asked Dr. Lanyon about a head injury McGill suffered in a car accident about eight months before he committed two armed robberies, for which he served time in prison. Dr. Lanyon testified that McGill described the symptoms he suffered after the accident. Dr. Lanyon testified such symptoms were consistent with frontal lobe damage. Dr. Lanyon added that McGill told him "he had no recollection of these robberies at all." Dr. Lanyon then stated that his "opinion of this, having written the report and sort of reflected on it, [is] that it's more likely that the head injury itself wiped out his memory[,] which is a common occurrence with head injuries."

On cross-examination, the state jumped on McGill's statement to Dr. Lanyon that he had no recollection of the robberies and Dr. Lanyon's conclusion that such memory loss was consistent with frontal lobe damage. The state introduced two police reports and a pre-sentence report in which McGill described the robberies in great detail. What was unknown to Dr. Lanyon at the time was that defense counsel had previously obtained the reports used to impeach his testimony but chose not to disclose them to Dr. Lanyon—a decision that set Dr. Lanyon up. After being shown the reports counsel withheld from him, Dr. Lanyon had the following exchange with the prosecutor:

Q: Does it appear from my reading of that portion of the police report, Dr. Lanyon, that Mr. McGill, in fact, had a memory of the events which occurred during the second robbery?

A: If that is accurate as you say, it appears that he did.

....

Q: And it appears that he had no blackout or amnesia at all which could be attributed to a head injury ?

A: Not at that time, correct.

The result was devastating to the conclusion counsel was trying establish, and Schaffer later admitted "the lack of corroboration contributed to Dr. Lanyon being discredited by the State."

The consequences of counsel's decision to withhold the pre-sentence reports from Dr. Lanyon may entice us to find counsel's preparation of Dr. Lanyon deficient. But we may not second-guess counsel's decision based on "the distorting effects of hindsight." Strickland , 466 U.S. at 689, 104 S.Ct. 2052. If counsel strategically withheld the reports after considering reasonable alternatives, Strickland prohibits us from critiquing counsel's decision after the fact. Id. Such tactical decisions represent "sound trial strategy" and are "virtually unchallengeable." See id. at 689, 690, 104 S.Ct. 2052 (quoting Michel v. Louisiana , 350 U.S. 91, 101, 76 S.Ct. 158, 100 L.Ed. 83 (1955) ). Despite the consequences of counsel's decision to withhold the pre-sentence reports from Dr. Lanyon, it was a reasonable and tactical decision at that time.

At her PCR-stage testimony, Schaffer explained that she purposefully withheld various pre-sentence reports to protect McGill's credibility with Dr. Lanyon. In particular, at the time, counsel believed that, in connection with one of the robberies, McGill had lied to a probation officer about having a child that he needed to care for, an allegation that counsel could not confirm. Counsel was concerned that if Dr. Lanyon learned that McGill had been untruthful during the presentence process that it would harm his credibility with Dr. Lanyon. As counsel saw it, she had two options, neither of which was particularly good. She could disclose the presentence reports to Dr. Lanyon and risk destroying McGill's credibility with his only expert witness, or she could withhold the reports in an effort to protect McGill's credibility. Counsel chose the latter, a decision that she characterized as a "strategic decision on [her] part."

This is precisely the type of tactical decision that Strickland protects. 466 U.S. at 690–91, 104 S.Ct. 2052. It is clear that counsel did not take lightly the decision to withhold information from Dr. Lanyon. In fact, counsel testified that the decision to withhold the reports was not unanimous among the defense team and that as lead counsel, she made the ultimate decision to do so. That the defense team discussed the benefits and potential consequences of not disclosing the pre-sentence reports is strong evidence that counsel's decision was sound trial strategy. Although counsel's decision may have ultimately harmed the mitigation effort, we cannot say that the decision fell outside "the wide range of reasonable professional assistance." Id. at 689, 104 S.Ct. 2052. Had Schaffer decided to provide Dr. Lanyon with the pre-sentence reports, McGill may have had an even less effective case for mitigation.

4. Evidence of substance abuse, sexual assault, and domestic violence

McGill claims that counsel was ineffective in investigating evidence of substance abuse, sexual assault and domestic violence. Under Williams , counsel's investigation falls short if it was insufficient to uncover evidence that reasonably should have been uncovered. See 529 U.S. at 396, 120 S.Ct. 1495 ; see also Porter v. McCollum , 558 U.S. 30, 40, 130 S.Ct. 447, 175 L.Ed.2d 398 (2009) (granting relief where counsel ignored avenues of potential mitigation evidence "of which he should have been aware"). The PCR court concluded that the defense presented "a substantial amount of mitigation" with respect to McGill's substance abuse, and his abusive relationship with Hardesty was "thoroughly explored." As for McGill's claim that he was sexually abused at Boysville, that was carefully reviewed by the PCR court, which concluded that he "ha[d] failed to substantiate his claim that he was sexually abused as a child. Therefore his trial counsel was not ineffective for failing to further investigate this potential mitigation." In particular, the PCR court found, as we have discussed, that McGill himself "denied any such abuse." The PCR court found that counsel adequately investigated evidence of McGill's substance abuse history, prior sexual abuse, and domestic violence. Here, the PCR court reasonably concluded that counsel's investigation met Williams ’s standard.

a. Substance abuse history

We previously discussed PCR counsel's interactions with addictionologist Dr. Mace Beckson through the lens of whether the PCR court relied on unreasonable determinations of fact under § 2254(d)(2). We must now evaluate whether counsel's investigation into McGill's drug-related mitigation evidence was sufficient and whether her decision not to retain a different addictionologist was reasonable. Although there is some overlap between this claim and the § 2254(d)(2) claim, here we focus on counsel's performance rather than the PCR court's factual determinations. We start with counsel's investigation into McGill's substance abuse history.

Counsel's investigation into McGill's substance abuse history was thorough and expansive. The defense team discovered and presented evidence of drug use spanning decades of McGill's life. Dr. Lanyon testified that McGill began using alcohol at nine years old and progressed to marijuana by thirteen. McGill eventually progressed to chronic, daily methamphetamine use, which often caused active paranoia, impaired his judgment, and interfered with "any remaining fragment of ability to reason." Dr. Lanyon's psychological evaluation also shed light on McGill's drug use in the weeks leading up to the crime. McGill actively used methamphetamine and "had been awake for several days at the time of the events with which he [was] charged." McGill's brother, Cordell, supported Dr. Lanyon's testimony, stating that McGill and Hardesty used crystal methamphetamine "all the time."

Nevertheless, McGill contends that counsel's investigation into his prior drug use was objectively deficient. He likens this case to Porter , where counsel "had only one short meeting with Porter regarding the penalty phase" and "did not obtain any of Porter's school, medical, or military service records or interview any members of Porter's family." 558 U.S. at 39, 130 S.Ct. 447 ("counsel did not even take the first step"). The Court had little difficulty concluding that a "decision not to investigate did not reflect reasonable professional judgment." Id. at 40, 130 S.Ct. 447. The investigation in McGill's case bears no resemblance whatsoever to the deficient investigation in Porter .

Nor was counsel deficient for failing to retain an alternate addictionologist. It is unclear exactly what additional evidence McGill believes counsel would have uncovered had she done so. After reviewing Dr. French's PCR-stage report, we are unconvinced that an alternate addictionologist would have more effectively investigated or presented McGill's substance abuse history. McGill assumes that an addictionologist would have more precisely attributed McGill's behavior to his drug use by testifying to the effects that methamphetamine had on McGill. However, Dr. French's report suffers from the same ambiguity and generalization as Dr. Lanyon's trial testimony. Although Dr. French listed more potential symptoms of methamphetamine use than Dr. Lanyon, he did not link McGill's behavior to any particular one. Nor did he opine on whether the type, amount, or concentration of methamphetamine McGill was using could have influenced McGill's decision-making on the night of the crime. In the end, the PCR court found that Dr. French's conclusions were "speculative" because, unlike Dr. Lanyon, Dr. French never examined McGill. The PCR court, thus, reasonably concluded that counsel conducted a thorough investigation of McGill's background and presented substantial evidence of substance abuse to the jury.

b. Childhood sexual abuse

The PCR court similarly held that counsel adequately investigated whether McGill suffered childhood sexual abuse. Evidence of sexual assault can be particularly powerful at the penalty stage due to its devastating and long-lasting effect on the victim. See Wiggins , 539 U.S. at 534–35, 123 S.Ct. 2527. Such evidence is especially powerful when the abuse occurs regularly over an extended period of time. Id. ; Wharton , 765 F.3d at 977. However, evidence of sexual assault does not always "tip the scales" toward leniency. Wharton , 765 F.3d at 978. For instance, we have denied relief where allegations of sexual abuse were bare or unsupported, Schurz v. Ryan , 730 F.3d 812, 815 (9th Cir. 2013), or where it was unclear whether the alleged sexual abuse actually occurred, Samayoa v. Ayers , 649 F.3d 919, 929 (9th Cir. 2011). We have also recognized that even if evidence of sexual abuse may have made a petitioner more sympathetic to a jury, it may also have the unwanted effect of making him seem less likely to be rehabilitated. Benson v. Chappell , 958 F.3d 801, 833 (9th Cir. 2020).

The problem for McGill is that there is no reliable evidence that he was sexually abused. The investigators did not turn up any well-grounded evidence, and McGill did not produce any such evidence for the PCR court. As the PCR court found, McGill came forward with evidence that his brother Lonnie had been abused at Buckner's Boys Ranch, but no one could testify that McGill had. And given that McGill had denied to Dr. Lanyon that he was sexually abused, there was nothing for counsel to investigate. Under any standard, "counsel is not deficient for failing to find mitigating evidence if, after a reasonable investigation, nothing has put the counsel on notice of the existence of that evidence." Babbitt v. Calderon , 151 F.3d 1170, 1174 (9th Cir. 1998) (quoting Matthews v. Evatt , 105 F.3d 907, 920 (4th Cir. 1997), abrogated on other grounds by Miller-El v. Dretke , 545 U.S. 231, 125 S.Ct. 2317, 162 L.Ed.2d 196 (2005) ). Nevertheless, McGill argues that two red flags should have prompted reasonable counsel to investigate further: Lonnie's childhood sex abuse and McGill's placement in foster care. But neither piece of evidence renders counsel's investigation unreasonable. Evidence that Lonnie was sexually assaulted in a boy's home did not mean that McGill was also sexually assaulted there. For instance, a second brother, Cordell, testified during the penalty phase that he had not suffered any type of sexual abuse at either Boysville or Buckner's and was unaware of whether McGill suffered such abuse. Further, investigating sexual abuse in McGill's past on the basis of Lonnie's allegations of abuse would have been particularly difficult because Lonnie was uncooperative and "absolutely refused" to come to court.

Second, the fact of foster-care placement is not itself indicative of sexual abuse. McGill points to Wiggins for support, but it does not stand for such a broad proposition. There, the Court granted relief in part based on counsel's failure to uncover reasonably available evidence of Wiggins's "repeated rape during his ... years in foster care." Wiggins , 539 U.S. at 535, 123 S.Ct. 2527. Nothing in Wiggins suggests a categorical imperative that counsel thoroughly investigate the possibility of sexual abuse for every capital defendant who has been placed in state care. Moreover, even if Wiggins did create some duty of inquiry, it does not stand for the proposition that counsel must continue to investigate sexual abuse after the defendant's express denial that such abuse ever occurred. Thus, Wiggins is not the "clearly established Federal law" that would require counsel to perform additional investigation into a claim of abuse that McGill's family did not disclose and that he actively denied.

In any event, we are unsure what more Schaffer should have done to investigate McGill's alleged sexual abuse. The defense team performed several interviews with McGill's family, but no one mentioned sexual abuse. Counsel retained Dr. Lanyon who probed McGill on any childhood abuse, but McGill denied suffering sexual abuse. Counsel requested several records from the boy's homes where McGill was placed, but the records did not include details of sexual abuse. After performing such a thorough investigation, we see nothing that should have placed counsel on notice of the abuse or demanded further investigation. Thus, the PCR court reasonably concluded that counsel was not deficient for failing to conduct additional investigation into sexual abuse.

c. Domestic violence

The PCR court concluded that counsel presented adequate evidence of McGill's relationship with Hardesty to portray her effect on him and that counsel was not ineffective for failing to retain a domestic violence expert to explain the McGill-Hardesty relationship. These conclusions were reasonable.

Counsel left the jury little doubt that Hardesty was psychotic and abusive. During closing arguments, counsel characterized the relationship as "codependent" and "sick" and opined that Hardesty was "crazy." Counsel also explained McGill's apparent weakness for Hardesty, stating that McGill "simply cope[d] with her psychoses, her addictive behaviors and her serious mental and physical abuse." Dr. Lanyon and several of McGill's family members echoed those statements during the penalty phase. Dr. Lanyon added that McGill was uniquely susceptible to Hardesty's manipulation as a result of the deep-seated psychological wounds left by his mother. And even McGill's mother testified that Hardesty was "one of the most evil people" she had ever met. Indeed, the PCR court aptly noted that Hardesty's influence on McGill was, in part, "the focus of the mitigation [case]."

There is no indication that a domestic violence expert would have better conveyed that mitigation evidence to the jury. McGill speculates that an expert would have more effectively explained the effects of Hardesty's domestic violence, but he does not provide any expert report, declaration, or other evidence to support that speculation. That none of McGill's witnesses used the words "domestic violence" does not detract from the clarity with which they described Hardesty's effect on McGill. The PCR court reasonably found that counsel performed an adequate investigation into Hardesty's abuse and McGill's unhealthy reliance on her.

5. Failure to Explain Adequately McGill's Prior Armed Robbery Convictions

McGill argues that Schaffer should have presented the circumstances surrounding his prior armed robbery convictions in a way that would have softened the convictions to the jury and that her failure to do so was ineffective assistance of counsel. The PCR court saw "no merit" in the claim and held that McGill "[did] not contest the validity of this prior conviction" as an aggravator. His claim that "his alleged brain damage" affected its commission was "speculative and not substantiated by any of the exhibits."

McGill's armed robbery convictions constituted statutory aggravating factors under Ariz. Rev. Stat. § 13-703(F)(2). Counsel attempted to soften the impact of the robberies as aggravators. Dr. Lanyon also gave the jury insight into McGill's state of mind at the time, testifying that he was essentially homeless, "living hand to mouth," and drinking heavily.⁹

9 The PCR court did commit one clear error in its explanation of who communicated the details of McGill's robbery to the jury. The court found:

Further, [McGill] recounted his prior crimes in detail on cross-examination (as he did to the presentence writer), undermining any claim that he blacked out or could not recall his crimes.

McGill neither testified, nor was he cross-examined at trial, rendering that portion of the PCR court's factual findings verifiably false. McGill seizes on that error to argue that the PCR court's ultimate conclusion relied upon an unreasonable determination of fact under 28 U.S.C. § 2254(d)(2). Nothing, however, turned on whether McGill testified in court that he could not recall the robberies or told the presentence report writer that he could not recall the crimes.
The PCR court's error merely misrepresented the medium by which the jury received McGill's statements—not the statements themselves. Rather than McGill making those statements on cross-examination as the PCR court incorrectly stated, the prosecutor introduced McGill's prior statements during Dr. Lanyon's cross-examination. That does not change the fact that the jury heard McGill's own statements recounting the details of the robberies. The PCR court acknowledged as much when it noted that McGill also recounted the details to a pre-sentence writer. Thus, the PCR court's conclusion was reasonable despite its slip of the pen.

McGill now argues that counsel failed to investigate his brain damage and explore how it might have contributed to his robberies. He points to PCR testimony from Dr. Wu and Dr. Rosengard. We have previously discussed their testimony in connection with the causal-nexus test. See Part III.A.1.c. We will not repeat our analysis, except to note that the PCR court concluded that, at best, their testimony would have added little to Dr. Lanyon's testimony and, at worst, "would have been an opening for the State to poke holes in the mitigation and would have hurt the defense's presentation." These findings are not unreasonable. * * *

McGill has not shown that counsel performed deficiently under Strickland at the penalty phase of his trial. The PCR court reasonably concluded that counsel's preparation, investigation, and presentation of mitigation evidence was thorough and reasoned. As a whole, the defense team uncovered a "not insignificant" amount of mitigation evidence that spanned decades of McGill's life and presented a comprehensive picture to the jury. See McGill I , 140 P.3d at 945. Conversely, there is no evidence that counsel failed to uncover any reasonably available mitigation records. We therefore conclude that counsel's performance was not objectively deficient in light of the prevailing professional norms. Because counsel's trial performance was adequate, we do not reach McGill's claims of prejudice. See Strickland , 466 U.S. at 697, 104 S.Ct. 2052.

IV. UNCERTIFIED CLAIMS

We now turn to the two claims that the district court did not certify for appeal: that counsel was ineffective at the guilt phase by failing to retain an expert arson investigator; and that his death sentence violated the Ex Post Facto Clause.

We must consider whether McGill has made the necessary showing of a constitutional deprivation to warrant a certificate of appealability on any of his claims. As we have previously set out, habeas petitioners cannot appeal the denial of a § 2254 petition without first obtaining a certificate of appealability. See 28 U.S.C. § 2253(c)(1) ; Miller-El I , 537 U.S. at 327, 335–36, 123 S.Ct. 1029. To proceed, McGill must demonstrate a "substantial showing" that he suffered a deprivation of a constitutional right. 28 U.S.C. § 2253(c)(2) ; Slack v. McDaniel , 529 U.S. 473, 483–84, 120 S.Ct. 1595, 146 L.Ed.2d 542 (2000). He meets that standard if reasonable jurists "could disagree with the district court's resolution of his [case] ... or that ... the issues presented are adequate to ... proceed further." Miller-El I , 537 U.S. at 327, 123 S.Ct. 1029. Applying that standard, we deny a certificate of appealability on McGill's ineffective assistance claim at the guilt phase. We grant a certificate of appealability on McGill's ex post facto claim but deny relief on the merits.

A. Ineffective Assistance at the Guilt Phase for Failure to Hire an Arson Expert

This claim is unique among McGill's ineffective assistance claims as it is the only claim challenging counsel's performance at the guilt stage. McGill argues that counsel was deficient for failing to retain an arson expert at trial to rebut the state's evidence that McGill mixed styrofoam into the gasoline before dousing Perez and Banta. McGill did not raise this claim in his state PCR petition, however, resulting in the claim being procedurally defaulted. See Martinez v. Ryan , 566 U.S. 1, 9, 132 S.Ct. 1309, 182 L.Ed.2d 272 (2012) ; Coleman v. Thompson , 501 U.S. 722, 753–54, 111 S.Ct. 2546, 115 L.Ed.2d 640 (1991). To cure the default, McGill also argues that PCR counsel performed deficiently by failing to raise this claim before the PCR court.

Martinez announced a possible, narrow exception in cases, like this one, where the alleged cause of a procedural default is the ineffective assistance of PCR counsel. To cure the default, a petitioner must show that PCR counsel was ineffective under Strickland and that the underlying ineffective assistance claim is "substantial ... which is to say that ... the claim has some merit." Martinez , 566 U.S. at 14, 132 S.Ct. 1309 (citation omitted). The analysis of a claim's substantiality mirrors the standard for issuing a certificate of appealability, see id. –namely, whether "reasonable jurists would find" the denial of relief "debatable or wrong." See Miller-El I , 537 U.S. at 338, 123 S.Ct. 1029 (citation omitted). Martinez thus created a layered analysis in which we must determine (1) whether PCR counsel was ineffective under Strickland —with its own two-part test—for failing to raise the underlying claim and (2) whether the underlying claim "has some merit." 566 U.S. at 14, 132 S.Ct. 1309.

Reasonable jurists would find it undebatable that McGill has not satisfied Martinez because counsel's decision not to call an arson witness was reasonable under Strickland . McGill makes much out of witness testimony that he mixed styrofoam into the gasoline to make it more difficult for Perez and Banta to extinguish the flames. He assumes that, had counsel more effectively rebutted that testimony, the jury would not have found that he committed the crime in an especially cruel manner. But the state's reliance on the styrofoam testimony was inconsistent at best. Two state witnesses testified that McGill told them that he mixed styrofoam into the gas, which created a sticky "napalm-like" substance. On the other hand, the state's own experts discounted the styrofoam testimony. The state's expert arson investigator, for example, testified that he found no evidence of styrofoam at the scene and was unsure whether the mixture McGill described was even possible. The state's expert criminologist similarly testified that he would expect to find high concentrations of styrene at the scene had McGill used styrofoam but that he did not find above-average amounts of styrene in his testing. Whether McGill actually put styrofoam in the container of gasoline or merely said he did so is unclear as evidenced by the state's own experts. It was reasonable for counsel not to retain an arson expert to rebut that testimony. Such rebuttal would only have emphasized the disputed fact to the jury.

Moreover, reasonable jurists would not find it debatable that counsel's failure to retain an arson expert did not reasonably influence whether the jury would find that McGill committed this crime in an especially cruel manner. As both the Arizona Supreme Court and the district court noted, a crime is especially cruel "if the victim consciously experienced physical or mental pain prior to death, and the defendant knew or should have known that suffering would occur." See McGill I , 140 P.3d at 938 (citation omitted). Any reasonable defendant would know that dousing two people in gasoline and lighting them on fire would "necessarily cause[ ] ... tremendous suffering." Id. Indeed, before Perez succumbed to his burns, he "scream[ed] in pain" and pleaded with nurses to "[g]et the pain away." Id. at 934. No reasonable jurist would conclude that the essential cruelty of McGill's crime in this case would turn on whether he did or did not mix styrofoam into the gasoline before igniting his victims. We therefore deny a certificate of appealability on this issue.

B. McGill's Death Sentence and the Ex Post Facto Clause

We turn, finally, to whether McGill's death sentence violated the Ex Post Facto Clause. In Ring v. Arizona , 536 U.S. 584, 122 S.Ct. 2428, 153 L.Ed.2d 556 (2002) ( Ring I ), the Supreme Court invalidated Ariz. Rev. Stat. § 13-703(C) (2001) because it required the sentencing judge—not the jury—"to find an aggravating circumstance necessary for imposition of the death penalty." 536 U.S. at 609, 122 S.Ct. 2428. Perez's murder fell within the brief period between Ring I and Arizona's amendment of § 13-703. On direct appeal to the Arizona Supreme Court, McGill argued that Arizona lacked a valid procedure to sentence him to death at the time he committed his crime, in violation of the Ex Post Facto Clause. The Arizona Supreme Court denied relief, stating only that it had already rejected the argument. See McGill I , 140 P.3d at 945 (citing State v. Ring , 204 Ariz. 534, 65 P.3d 915, 928 (2003) (en banc) ( Ring II )). We believe that reasonable jurists could find debatable whether Arizona's death-sentencing process violates the Ex Post Facto Clause, and we therefore grant a certificate of appealability as to this issue. We conclude, however, that the Arizona Supreme Court reasonably applied clearly established federal law when it determined that Arizona had only made a procedural change to its death penalty process, and that change did not violate the Ex Post Facto Clause. See Ring II , 65 P.3d at 926–28.

A brief review of the timeline is helpful to frame McGill's argument. In the years leading up to Perez's murder, two separate statutory provisions codified the state's use and imposition of the death penalty: "availability" provisions and "procedural" provisions. The availability provisions specified what offenses were eligible for death and what aggravating factors were necessary to make that penalty available as punishment in a given case. Thus, Ariz. Rev. Stat. § 13-1105 (2001), stated that "[f]irst degree murder is a class 1 felony and is punishable by death or life imprisonment as provided by [Ariz. Rev. Stat.] section 13-703," and Ariz. Rev. Stat. § 13-703(G) (2001) listed the statutory aggravating circumstances that would make the death penalty available for a particular offender convicted of first degree murder. The procedural provisions, which were contained in the remaining provisions of § 13-703, outlined the procedure by which a death sentence was imposed. They allocated the burden of proof and provided for the broad admissibility of mitigating evidence. Id. § 13-703(B)–(F). Importantly, and relevant here, § 13-703(C) reserved for the "court alone" the duty to "make all factual determinations" regarding aggravation and mitigation evidence to determine whether to impose a capital sentence. Ring I , 536 U.S. at 592–93, 122 S.Ct. 2428 (citing Ariz. Rev. Stat. § 13-1105(C) ); see also Ariz. Rev. Stat. § 13-703(F) (2001) ("court ... shall impose a sentence of death if the court finds one or more of the aggravating circumstances enumerated in subsection G of this section and then determines that there are no mitigating circumstances sufficiently substantial to call for leniency").

In June 2002, the Court held unconstitutional the relevant portions of § 13-703(C) and § 13-703(F) under the Sixth Amendment, concluding that the right to a jury trial required the jury "to find an aggravating circumstance necessary for imposition of the death penalty." Id. at 609, 122 S.Ct. 2428 (overruling Walton v. Arizona , 497 U.S. 639, 110 S.Ct. 3047, 111 L.Ed.2d 511 (1990), in light of Apprendi v. New Jersey , 530 U.S. 466, 120 S.Ct. 2348, 147 L.Ed.2d 435 (2000) ). Thirty-eight days after Ring I , Arizona amended § 13-703 to require the "trier of fact" to "find[ ] one or more of the aggravating circumstances enumerated in subsection F" and to weigh any such aggravating factors against the mitigating circumstances in deciding whether to impose a capital sentence. See Ariz. Rev. Stat. § 13-703(E) (2002). McGill murdered Perez during that thirty-eight-day interregnum. He argues that Arizona lacked a valid death penalty when he committed his crime, and that Arizona's law curing the Ring I defect violates the Ex Post Facto Clause.

The Ex Post Facto Clause provides that "No State shall ... pass any ... ex post facto Law." U.S. Const. art. I, § 10, cl. 1. It prohibits a state from retroactively changing the definition of a crime to make formerly innocent behavior illegal or increasing the punishment for criminal acts. See Collins v. Youngblood , 497 U.S. 37, 42–43, 110 S.Ct. 2715, 111 L.Ed.2d 30 (1990) ; see also Dobbert v. Florida , 432 U.S. 282, 292, 97 S.Ct. 2290, 53 L.Ed.2d 344 (1977) ; Beazell v. Ohio , 269 U.S. 167, 169–70, 46 S.Ct. 68, 70 L.Ed. 216 (1925) ; Calder v. Bull , 3 U.S. (3 Dall.) 386, 390, 1 L.Ed. 648 (1798) (opinion of Chase, J.). The Supreme Court has set out three factors to determine whether a challenged law was issued ex post facto. First, the law must be retrospective, meaning that it applies to actions that pre-date its enactment. Miller v. Florida , 482 U.S. 423, 430, 107 S.Ct. 2446, 96 L.Ed.2d 351 (1987) (citing Weaver v. Graham , 450 U.S. 24, 29, 101 S.Ct. 960, 67 L.Ed.2d 17 (1981) ). Second, application of the new law must disadvantage the defendant. Id. Third, the law must affect the defendant's substantial rights such that it alters "the quantum of punishment." Weaver , 450 U.S. at 33, 101 S.Ct. 960 (quoting Dobbert , 432 U.S. at 293–94, 97 S.Ct. 2290 ). Mere procedural changes that "simply alter[ ] the methods employed in determining [a sentence]" do not affect substantial rights because they do not change the "quantum of punishment attached to the crime." Dobbert , 432 U.S. at 293–94, 97 S.Ct. 2290 (citing Hopt v. Utah , 110 U.S. 574, 589–90, 4 S.Ct. 202, 28 L.Ed. 262 (1884) ). This is so even if the procedural change disadvantages the defendant. Id. at 293, 97 S.Ct. 2290.

In Ring II , the Arizona Supreme Court held that Arizona's amended procedural statute did not violate the Ex Post Facto Clause of the U.S. Constitution. In that case, the Arizona high court consolidated the appeals of the thirty-one death-row inmates who were unconstitutionally sentenced under the state's prior version of § 13-703. See Ring II , 65 P.3d at 925. The court found two U.S. Supreme Court cases "particularly instructive." Id. at 927.

The first of these was Dobbert v. Florida , 432 U.S. 282, 97 S.Ct. 2290, 53 L.Ed.2d 344 (1977). In that case Dobbert was accused of brutally murdering his own children in 1971 and 1972. Under the law in effect at the time, Florida required imposition of the death penalty for capital crimes unless the jury recommended mercy. The jury's recommendation was binding on the judge. Id . at 287–88, 97 S.Ct. 2290. After Dobbert committed his crimes, but before he could be tried, the U.S. Supreme Court struck down Georgia's death penalty statute in Furman v. Georgia , 408 U.S. 238, 92 S.Ct. 2726, 33 L.Ed.2d 346 (1972). Weeks later the Florida Supreme Court, citing Furman , held that Florida's death penalty statute was similarly unconstitutional. Dobbert , 432 U.S. at 288, 97 S.Ct. 2290 (citing Donaldson v. Sack , 265 So. 2d 499 (Fla. 1972) ). Later that same year, the Florida legislature amended its death penalty statute to comply with Furman . The new statute provided that once a defendant was found guilty of a capital felony, the court must hold a separate hearing to consider aggravating and mitigating evidence. The jury would make a recommendation, which was not binding on the judge, and the judge would issue written findings of fact if imposing a death sentence. Id. at 290–92, 97 S.Ct. 2290. Dobbert was tried under the new statute. The jury recommended a life sentence, but the trial court—under its authority under the revised Florida statute—overruled the jury and sentenced Dobbert to death. Id. at 287, 97 S.Ct. 2290.

Dobbert argued that his rights under the Ex Post Facto Clause were violated by the change in Florida law. Under the Florida law in effect when he committed his crimes, the jury's recommendation of life would have been binding on the trial judge; under the new law, Dobbert got the death penalty. Dobbert raised two claims under the Ex Post Facto Clause that are relevant here. First, he alleged that Florida had altered the roles of the judge and jury and that he was entitled to proceed under the prior law. The Court rejected this argument. "Even though it may work to the disadvantage of a defendant, a procedural change is not ex post facto ." Id. at 293, 97 S.Ct. 2290. The Court found that "the change in the [Florida] statute was clearly procedural. The new statute simply altered the methods employed in determining whether the death penalty was to be imposed; there was no change in the quantum of punishment attached to the crime." Id. at 293–94, 97 S.Ct. 2290. The Court also took into account that "not only was the change in the law procedural, it was ameliorative"—the change was to deal with Furman —and it was not clear that the new law was "more onerous than the prior law." Id. at 294, 97 S.Ct. 2290.

Dobbert also argued that because the statute in effect in Florida when he murdered his children was unconstitutional, "there was no death penalty ‘in effect’ in Florida." Id. at 297, 97 S.Ct. 2290. The Court termed this argument "sophistic" and "highly technical," "mock[ing] the substance of the Ex Post Facto Clause." Id. For the Court, "[w]hether or not the old statute would in the future, withstand constitutional attack, it clearly indicated Florida's view of the severity of murder and of the degree of punishment which the legislature wished to impose upon murderers." Id. ¹⁰ The statute's "existence on the statute books provided fair warning" to Dobbert. Id. ; see also id. at 303, 97 S.Ct. 2290 (Burger, C.J., concurring) ("Petitioner was at least constructively on notice that this penalty might indeed follow his actions.").

10 We note that the U.S. Supreme Court initially upheld the constitutionality of Florida's 1972 revised sentencing process, see Proffitt v. Florida , 428 U.S. 242, 96 S.Ct. 2960, 49 L.Ed.2d 913 (1976), but it later invalidated the procedure under the Sixth Amendment, Hurst v. Florida , 577 U.S. 92, 98–99, 136 S.Ct. 616, 193 L.Ed.2d 504 (2016) (citing Ring I , 536 U.S. at 597, 122 S.Ct. 2428 ). Florida's procedure was deficient because it did not require the jury to make "specific factual findings with regard to the existence of mitigating or aggravating circumstances[,] and its recommendation [was] not binding on the trial judge." Hurst , 577 U.S. at 99, 136 S.Ct. 616 (citation omitted). Arizona's amended sentencing procedure, on the other hand, required the jury to make those crucial findings of fact and makes the jury's determination final. See Ariz. Rev. Stat. § 13-703(E) (2002).

The second case the Arizona Supreme Court looked to was Collins v. Youngblood , 497 U.S. 37, 110 S.Ct. 2715, 111 L.Ed.2d 30 (1990). In Collins , the defendant was convicted of aggravated sexual abuse. He was sentenced to life and fined $10,000. Collins filed in state court for a writ of habeas corpus on the grounds that the statute did not authorize a fine, and thus his sentence was void and he should be given a new trial. Before his habeas petition was reviewed, the Texas legislature adopted a statute allowing an appellate court to reform an improper sentence assessing a fine not authorized by law. Id . at 39–40, 110 S.Ct. 2715. The Texas Court of Criminal Appeals ordered his fine deleted from his sentence and refused his request for a new trial. The U.S. Supreme Court held that the Texas statute did not run afoul of the Ex Post Facto Clause: it "[did] not punish as a crime an act previously committed, which was innocent when done; nor make more burdensome the punishment for a crime, after its commission." Id. at 52, 110 S.Ct. 2715.

Relying on Dobbert and Collins , the Arizona Supreme Court thought it clear that "rights secured by the Sixth Amendment jury trial right, the right at issue here, are inherently procedural" and that "[t]he new sentencing statutes alter the method used to determine whether the death penalty will be imposed but make no change to the punishment attached to first degree murder." Ring II , 65 P.3d at 928. As a consequence, Arizona's new sentencing statute did not violate the Ex Post Facto Clause of the U.S. Constitution. Id.

McGill's case was not part of the cases consolidated in Ring II because unlike those defendants, McGill was sentenced under a constitutional version of Ariz. Rev. Stat. § 13-703. Nevertheless, the Arizona Supreme Court reviewed his case separately and found that it was governed by its decision in Ring II. McGill I , 140 P.3d at 945. McGill argues that he is in a different position from the defendants in Ring II. The defendants sentenced before Ring I had notice of both the availability of the death penalty and the procedure by which the sentence would be imposed, even though that procedure was later held inconsistent with the Sixth Amendment. McGill, on the other hand, claims he had only half the puzzle, as § 13-703 had been invalidated by the U.S. Supreme Court just three weeks before McGill committed his crime. McGill claims that he was in limbo. According to McGill, § 13-1105(C) put him on notice that "[f]irst degree murder ... [was] punishable by death ... as provided by section 13-703," Ariz. Rev. Stat. § 13-1105 (2001), but there was no procedure by which he could be tried and sentenced to death until § 13-703 was re-implemented weeks after his crime. It is on that basis that McGill argues there was no functioning death penalty statute in place at the time he committed murder.

Although we recognize the different position that McGill is in, we are not persuaded that McGill is entitled to relief under AEDPA. And we wish to be very clear about our role here. On AEDPA review, we may not grant relief unless the Arizona Supreme Court's application of federal law was flawed "beyond any possibility for fairminded disagreement." Richter , 562 U.S. at 103, 131 S.Ct. 770. Even if we thought the Arizona Supreme Court's conclusion was wrong under the Ex Post Facto Clause as a matter of first impression, we could not issue relief. Rather, we can only review the decision to determine if it is an unreasonable application of Dobbert and Collins .

As the Court held in Dobbert , the existence of § 13-1105(C) when McGill murdered Perez was likewise an "operative fact" that gave fair warning to McGill that Arizona could seek the death penalty if he was convicted of first-degree murder. 432 U.S. at 298, 97 S.Ct. 2290. McGill was on full notice that Arizona could punish him for murder with death or life imprisonment. The fact that § 13-1105(C) cross-referenced § 13-703 ’s procedural requirements does not detract from Arizona's "view of the severity of murder" and "the degree of punishment which the legislature wished to impose upon murderers." Dobbert , 432 U.S. at 297, 97 S.Ct. 2290. Moreover, the 2002 amendments did not alter the pre-existing enumeration of statutory aggravating factors that was contained in § 13-703(G) (which was reclassified as § 13-703(F) ). McGill thus received "fair warning as to the degree of culpability which the State ascribed to the act of murder." Id. There was no retroactive change in the penalty for his crime.

Furthermore, the Arizona Supreme Court reasonably concluded in light of Dobbert that the amendments to § 13-703 are plainly procedural, not substantive. Ring I did not invalidate the death penalty in Arizona. It held that Arizona could not confer on the judge the duty to find sufficient aggravating factors necessary for the imposition of the death penalty. Ring I , 536 U.S. at 609, 122 S.Ct. 2428. Arizona's change to § 13-703 affected the allocation of responsibility between judge and jury, and that makes it analogous to the change Florida made to its system in Dobbert . See 432 U.S. at 293–94, 97 S.Ct. 2290. And because Arizona was responding to Ring I , its change was "ameliorative," just as Florida's amendment responded to Furman . Id. at 294, 97 S.Ct. 2290. Arizona did not impose a penalty that was previously unavailable, nor did the state criminalize innocent conduct after the fact. The state "simply altered the method[ ] employed in determining whether the death penalty was to be imposed." Id. at 293–94, 97 S.Ct. 2290. As a consequence, McGill's substantial rights have not been affected. Such procedural changes fall outside the protections of the Ex Post Facto Clause. Id. at 296–97, 97 S.Ct. 2290 ; see Styers v. Ryan , 811 F.3d 292, 297 (9th Cir. 2015) ("The rule announced in Ring [I] , under which capital defendants are entitled to a jury determination of any fact on which the legislature conditions an increase in their maximum punishment ... is a procedural rule that applies to capital defendants on direct review." (cleaned up)).

Our dissenting colleague takes a different view of the privilege or immunity against ex post facto laws. We have three observations. First, although the dissent briefly refers to AEDPA, Dissenting Op. at 709-10, 710-11, the dissent makes no serious effort to engage with AEDPA's standard. The dissent cited only the general test for ex post facto inquiries, id. at 709-10, conducted a de novo review for that test, id. at 709-11, and pronounced the Arizona Supreme Court's decision unreasonable, id . at 709-10, 710-11. The Supreme Court has previously reversed us for conducting precisely this kind of analysis. Sexton v. Beaudreaux , ––– U.S. ––––, 138 S. Ct. 2555, 2560, 201 L.Ed.2d 986 (2018) (per curiam) ("The Ninth Circuit essentially evaluated the merits de novo , only tacking on a perfunctory statement at the end of its analysis asserting that the state court's decision was unreasonable."). Such analysis is "fundamentally inconsistent with AEDPA." Shinn v. Kayer , ––– U.S. ––––, 141 S. Ct. 517, 523, 208 L.Ed.2d 353 (2020) (per curiam) (citing Beaudreaux , 138 S. Ct. at 2560 ). Second, the Court has made clear that we are to conduct AEDPA review of state court judgments, not at the highest level of abstraction, but at the lowest. It is not sufficient for purposes of AEDPA to cite a general rule; rather, the Supreme Court must have "squarely addresse[d]" the issue. Wright v. Van Patten , 552 U.S. 120, 125, 128 S.Ct. 743, 169 L.Ed.2d 583 (2008) ; see Yarborough v. Alvarado , 541 U.S. 652, 664, 124 S.Ct. 2140, 158 L.Ed.2d 938 (2004) ("The more general the rule, the more leeway [state] courts have in reaching outcomes in case-by-case determinations."). We identified two cases— Collins and Dobbert , which were also identified and addressed by the Arizona Supreme Court—as the closest factually to McGill's case. The dissent does not really grapple with either case, but dismisses Collins and Dobbert perfunctorily. Dissenting Op. at 710-11, 711-12. With all respect, that is not AEDPA review.¹¹ Third, the dissent's analysis of the Ex Post Facto Clause is interesting and novel—to be sure, not grounds for granting habeas under AEDPA—but, we think, not sound. The dissent's principal point is that "McGill could not have been sentenced to death for murder when he committed his crimes because at that time there was no statute implementing the death penalty in Arizona." Dissenting Op. at 709 (emphasis added); see also id. at 709 ("There was no law that permitted McGill to be punished with a death sentence during the thirty-eight days between Ring I and the re-enactment of § 13-703."); id. at 710 ("[McGill was] on notice that he could not be sentenced to death for first-degree murder."); id. at 711 ("There was no risk that McGill would be sentenced to death for his crimes at the time they were committed ...."). And the dissent spins an interesting, creative counterfactual in which the Arizona legislature delays for years in fixing the Ring I problem. Dissenting Op. at 711. We are puzzled by this analysis, because we are not aware of anything in Ex Post Facto Clause jurisprudence that turns on whether, at the time the crime was committed, the proper procedures are in place to impose, at trial, the sentence of death that the substantive statute then authorizes as a punishment for the conduct that the defendant committed. McGill was plainly on notice that his conduct was punishable by death when he murdered Perez on July 13, 2002. The dissent's whole case for issuing the writ rests on the premise that, due to the declared invalidity of Arizona's capital sentencing procedure on the day he committed his crime, McGill could not have received his sentence if (in a remarkable and likely unconstitutional act of very swift justice) his trial had taken place later that same day. The dissent thus assumes that the protection of the Ex Post Facto Clause extends to the particular procedures that would be lawfully available to any sentencing conducted on the date of the offense, but that view seems difficult to square with Dobbert ’s statement that "[e]ven though it may work to the disadvantage of a defendant, a procedural change is not ex post facto." 432 U.S. at 293, 97 S.Ct. 2290. Here, as in Dobbert , "[t]he new statute simply altered the methods employed in determining whether the death penalty was to be imposed; there was no change in the quantum of punishment attached to the crime." Id. at 293–94, 97 S.Ct. 2290.

11 The dissent states, for example, that "Here, the issue is whether the change from no possibility of the death penalty to possibility of the death penalty was procedural or substantive." Dissenting Op. at 710. The best answer to that question comes from Dobbert , in which the Court said that a statute that "simply altered the methods employed in determining whether the death penalty was to be imposed" was procedural because there was "no change in the quantum of punishment attached to the crime." Dobbert , 432 U.S. at 293–94, 97 S.Ct. 2290. The dissent ignores the Court's analysis.
The dissent's argument that "By the time of McGill's sentencing, the Arizona legislature had increased the risk of a death sentence for conduct that had already taken place," Dissenting Op. at 711, has likewise been answered by Dobbert . Dobbert argued that when he committed his crime, "there was no death penalty ‘in effect’ in Florida." Dobbert , 432 U.S. at 297, 97 S.Ct. 2290. The Court thought this argument "sophistic." Id.

We are not sure why the question of sentencing is the relevant inquiry. The dissent claims we have "invented [a] distinction between a law authorizing a ‘punishment’ and a law authorizing a ‘sentence.’ " Dissenting Op. at 711 n.4. The dissent claims that there is no difference: "the punishment is the sentence." Id .¹² The line Arizona has drawn between the punishment prescribed for a crime and the procedure by which one can be convicted and sentenced is one that has been around for a long time. See Wayne R. LaFave, Jerold H. Israel & Nancy J. King, Criminal Procedure 24 (4th ed. 2004) (" ‘Legal procedure,’ Roscoe Pound long ago noted, ‘is a means, not an end; it must be made subsidiary to the substantive law as a means of making that law effective in action.’ "). The dissent's argument marries the procedure and the substantive law for Ex Post Facto purposes—it concludes that because Arizona lacked a constitutional statutory scheme for imposing the sentence, there was no death penalty.¹³

12 The dissent is playing fast and loose with its terms. Nothing in the substantive law changed for McGill—his crime was defined as punishable by death on the day he committed it and on the day he was sentenced. All that changed was the process by which his sentence was to be determined, and McGill was tried, convicted, and sentenced consistent with the U.S. Constitution. That puts McGill's case squarely within Dobbert : "not only was the change in the law procedural, it was ameliorative" and it was "[not] more onerous than the prior law." Dobbert , 432 U.S. at 294, 97 S.Ct. 2290. To paraphrase the dissent, "the later penalty imposed [on McGill] was the same penalty that could have been imposed at the time of [McGill's] offense." Dissenting Op. at 712.

13 The dissent assumes that if the Arizona legislature had not responded so promptly to Ring I , the Arizona courts could not have complied with the dictates of the Sixth Amendment for death-eligible defendants, consistent with Ring I . See Dissenting Op. at 711. What principle of Arizona law would prevent Arizona courts from complying with Ring I , even if the legislature had not amended the statute? The dissent has made a huge assumption about how Arizona law works, including state separation of powers principles. We are not sure why the dissent gets to decide such matters.

The Supreme Court might well agree with our colleague in the future, but on AEDPA review, we do not get to create new ex post facto law and then issue habeas on that basis. Edwards v. Vannoy , ––– U.S. ––––, 141 S. Ct. 1547, 1554, 209 L.Ed.2d 651 (2021) ; Teague v. Lane , 489 U.S. 288, 310, 109 S.Ct. 1060, 103 L.Ed.2d 334 (1989) (plurality opinion).

* * *

We need not go so far as to decide that Arizona's new scheme is consistent with the Ex Post Facto Clause. We are not called upon to predict whether the Supreme Court would have upheld McGill's conviction and sentence had it granted certiorari review from the judgment of the Arizona Supreme Court on direct appeal. We only decide that the Arizona Supreme Court's understanding of the Ex Post Facto Clause, which it based on its review of Dobbert and Collins , is not unreasonable.¹⁴ If we are to take seriously the Court's declaration that "an unreasonable application of federal law is different from an incorrect application of federal law," Richter , 562 U.S. at 101, 131 S.Ct. 770 (citation omitted), we cannot say that the Arizona Supreme Court's decision was an unreasonable one.

14 We note that the standard for granting the COA on this issue requires only that we determine that "jurists of reason could disagree with the district court's resolution of his constitutional claims or that jurists could conclude the issues presented are adequate to deserve encouragement to proceed further." Miller-El I , 537 U.S. at 327, 123 S.Ct. 1029. We are fully satisfied of the fair-mindedness of our dissenting colleague, and so we have granted the COA.
In order to grant habeas, however, our colleague must be persuaded that no "fairminded jurist could take a different view," Kayer , 141 S. Ct. at 525, that is, that the constitutional error was "well understood and comprehended in existing law beyond any possibility for fairminded disagreement," Richter , 562 U.S. at 103, 131 S.Ct. 770. Our colleague has not returned the favor.

V. CONCLUSION

The judgment of the district court is AFFIRMED .

COLLINS, Circuit Judge, concurring:

I concur in Judge Bybee's opinion, including its conclusion that, under Browning v. Baker , 875 F.3d 444 (9th Cir. 2017), the district court erred in limiting its grant of a certificate of appealability ("COA") to only one claim concerning whether McGill's trial counsel was ineffective at the penalty phase. See Opin. at 678. Browning held that, in deciding whether to grant a COA with respect to challenges to the adequacy of counsel at a particular phase, a district court should not "separat[e]" the ineffective assistance "argument into individual ‘claims’ of [ineffective assistance] corresponding to particular instances of [the attorney's] conduct." 875 F.3d at 471. Rather, the ineffective assistance "portion of the COA should [be] crafted at a higher level of generality." Id . We have construed Browning to require a COA concerning an ineffective assistance claim to extend to all other properly preserved challenges to counsel's effectiveness at the same phase. See White v. Ryan , 895 F.3d 641, 645 n.1 (9th Cir. 2018) (following Browning in holding that, because "White has but a single claim regarding his right to the effective assistance of counsel at the penalty phase of resentencing," the district court committed "error" by failing to issue a COA covering additional portions of White's ineffective assistance claim concerning sentencing). Judge Bybee's opinion faithfully applies this precedent, and I join it in full.

I write separately only to note that, on this point, Browning seems to me to be plainly incorrect. Browning based its holding on two premises: (1) under the Antiterrorism and Effective Death Penalty Act ("AEDPA"), a COA may issue "only if the applicant has made a substantial showing of the denial of a constitutional right," 28 U.S.C. § 2253(c)(2) ; and (2) because the underlying constitutional right at issue—the effective assistance of counsel—requires consideration of counsel's performance "as a whole," all asserted errors of counsel at a particular phase of the proceedings must be considered. 875 F.3d at 471 (emphasis omitted). The second premise is, I think, contrary to the statute and to common sense.

Browning failed even to mention § 2253(c)(3), which states that any COA that is issued "shall indicate which specific issue or issues satisfy the showing required by paragraph (2)." 28 U.S.C. § 2253(c)(3) (emphasis added). This aspect of the COA process, although not jurisdictional in the strict sense, "screens out issues unworthy of judicial time and attention." Gonzalez v. Thaler , 565 U.S. 134, 145, 132 S.Ct. 641, 181 L.Ed.2d 619 (2012) (emphasis added). The statute itself thus refutes Browning ’s suggestion that a COA must be granted with respect to every single "issue" concerning alleged ineffective assistance of counsel at a particular phase of the proceedings. Under § 2253(c)(3), a COA must be limited to only those particular issues relating to the alleged deprivation of effective assistance of counsel that "reasonable jurists could debate" or that are "adequate to deserve encouragement to proceed further." Slack v. McDaniel , 529 U.S. 473, 483–84, 120 S.Ct. 1595, 146 L.Ed.2d 542 (2000) (citation and internal quotation marks omitted); see also Strickland v. Washington , 466 U.S. 668, 690, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984) ("A convicted defendant making a claim of ineffective assistance must identify the acts or omissions of counsel that are alleged not to have been the result of reasonable professional judgment.").

More broadly, Browning ’s in-for-a-penny-in-for-a-pound approach to ineffective assistance claims is divorced from the reality of habeas litigation. It is quite often the case—as it is here—that habeas counsel themselves carve up the ineffective assistance claim for a single phase of the proceeding into multiple separate claims. McGill's counsel thus separated the alleged ineffective assistance of counsel at the penalty phase into several different claims, with "Claim One" relating to multiple closely related alleged deficiencies in developing and presenting mitigating evidence and "Claim Two" relating to the asserted failure to challenge the weight to be given to the "prior serious offense aggravator." I can see no reason in law or logic why, if a COA is granted on Claim One, it must also be granted on Claim Two. Even McGill did not see these two points as raising the same claim—much less the same "specific issue," 28 U.S.C. § 2253(c)(3) —and nothing in § 2253(c) says that if a COA is granted for one issue, it must also be granted for the other. On the contrary, McGill's own division of his attack on the effectiveness of his counsel into multiple separate claims reflects the common-sense notion that these separate aspects of counsel's behavior are sufficiently distinct that they may be addressed and considered separately.

Moreover, the procedural rules governing habeas litigation may require separating out the distinct challenges to trial counsel's performance. For example, an ineffective assistance claim based on a particular aspect of trial counsel's conduct may be procedurally defaulted if it was not presented to the state courts, meaning that an ineffective assistance claim based on that aspect may be unable to proceed in federal court. See , e.g. , Landrum v. Mitchell , 625 F.3d 905, 918–19 (6th Cir. 2010) (holding that, "[a]lthough Landrum did raise an ineffective assistance of trial counsel claim in his [state] post-conviction petition, he did not include the allegation about introducing [a particular witness's] testimony in the guilt phase" and that claim was therefore procedurally defaulted). Indeed, a COA cannot issue with respect to a procedurally defaulted claim unless the petitioner makes a sufficient showing "directed at the underlying constitutional claims and ... directed at the district court's procedural holding." Slack , 529 U.S. at 484–85, 120 S.Ct. 1595 (emphasis added). That further confirms that claims concerning distinct aspects of a counsel's alleged ineffective assistance may warrant separate consideration for COA purposes.^{1 1} See , e.g. , Roberts v. Dretke , 356 F.3d 632, 641 & n.6 (5th Cir. 2004) (granting a COA only with respect to two particular aspects of the petitioner's ineffective assistance claims, while denying it as to two others, one of which was procedurally defaulted).

1 I do not mean to suggest, however, that Browning would require COA expansion in such a scenario, in direct contravention of Slack . Browning did not address the question of whether a COA should be expanded to cover other ineffective assistance claims that were procedurally defaulted, and its rule therefore applies only to additional ineffective assistance claims involving the same phase that have not been procedurally defaulted. Here, McGill's Claim Two (which is the claim that is the beneficiary of Browning ’s rule) was not procedurally defaulted, and so Slack ’s rule for COAs involving procedurally defaulted claims provides no obstacle to applying Browning here.

1 The State conceded multiple times during oral argument that if the Arizona legislature had never re-enacted § 13-703, McGill could not have been sentenced to death. This quickly disposes of the majority's speculation that the Arizona courts could have, on their own, imposed the death penalty in compliance with Ring I absent the legislature's revision of § 13-703. Op. at 706 n.13. Moreover, the Supreme Court has made clear that the death penalty cannot be imposed without an implementing statute, since "Furman mandates that where discretion is afforded a sentencing body on a matter so grave as the determination of whether a human life should be taken or spared, that discretion must be suitably directed and limited so as to minimize the risk of wholly arbitrary and capricious action." Gregg v. Georgia , 428 U.S. 153, 189, 96 S.Ct. 2909, 49 L.Ed.2d 859 (1976) (opinion of Stewart, Powell, and Stevens, JJ.). And if all that were not enough, § 13-1105 allows the death penalty only "as provided by" § 13-703. Clearly, the plain text of the statute would not permit a court to impose the death penalty outside of the requirements of § 13-703.

I am not entirely optimistic that Browning ’s error will ever be fixed, because its only effect is to create unnecessary work, rather than to change the outcome. Any case in which the Browning rule makes a difference—i.e. , a case in which, but for Browning , a COA would have been denied by the relevant judges of this court as to an additional ineffective assistance claim—is necessarily one in which that additional claim will fail on the merits under AEDPA's standards after the COA is wrongly granted. Indeed, I would have denied a COA as to Claim Two here, but for that very reason I obviously agree with Judge Bybee's conclusion, in his opinion, that Claim Two fails on the merits under AEDPA. But because Browning ’s rule is clearly wrong, defeats the screening purpose of § 2253(c)(3), and creates unnecessary work and delay, I would hope that, perhaps in the next en banc case in which that rule has played a role, we will take the time to revisit it.

M. SMITH, Circuit Judge, concurring in part and dissenting in part: LeRoy McGill could not have been sentenced to death for murder when he committed his crimes because at that time there was no statute implementing the death penalty in Arizona. Yet because the Arizona legislature passed a law thirty-eight days later that purported to allow his execution, McGill now sits on death row. For the reasons hereafter noted, I believe that McGill's death sentence is unconstitutional. I respectfully dissent.

I

Up until June 2002, imposition of the death penalty in Arizona was governed by Arizona Revised Statutes §§ 13-1105 and 13-703 (2001). As the majority notes, § 13-1105 stated, "First degree murder ... is punishable by death or life imprisonment as provided by § 13-703." Ariz. Rev. Stat. § 13-1105 (2001). The part of § 13-703 that permitted the death penalty required the judge in a criminal case to determine whether aggravating factors for a particular crime warranted execution. See id. § 13-703(C) (2001). But on June 24, 2002, § 13-703 was struck down as unconstitutional in Ring v. Arizona (Ring I) , 536 U.S. 584, 609, 122 S.Ct. 2428, 153 L.Ed.2d 556 (2002), when the Supreme Court held that the jury (not the judge) must determine whether aggravating factors exist to impose the death penalty. Thirty-eight days later, on August 1, 2002, the Arizona legislature enacted a new capital punishment statute that passed constitutional muster. On July 13, 2002, during the time period between Ring I and re-enactment of § 13-703, McGill committed the murder underlying this appeal, for which he was sentenced to death.

II

To determine whether a criminal law is ex post facto , the Supreme Court has instructed us to apply a three-prong test. "[F]irst, the law must be retrospective, that is, it must apply to events occurring before its enactment." Miller v. Florida , 482 U.S. 423, 430, 107 S.Ct. 2446, 96 L.Ed.2d 351 (1987) (internal quotation marks omitted). Second, the law "must disadvantage the offender affected by it." Id. (internal quotation marks omitted). Finally, "no ex post facto violation occurs if a change does not alter ‘substantial personal rights,’ but merely changes ‘modes of procedure which do not affect matters of substance.’ " Id. (quoting Dobbert v. Florida , 432 U.S. 282, 293, 97 S.Ct. 2290, 53 L.Ed.2d 344 (1977) ). Ultimately, "[t]he touchstone of [the Supreme] Court's inquiry is whether a given change in law presents a ‘sufficient risk of increasing the measure of punishment attached to the covered crimes.’ " Peugh v. United States , 569 U.S. 530, 539, 133 S.Ct. 2072, 186 L.Ed.2d 84 (2013) (quoting Garner v. Jones , 529 U.S. 244, 250, 120 S.Ct. 1362, 146 L.Ed.2d 236 (2000) ) (some internal quotation marks omitted).

McGill's claim plainly satisfies the first two prongs of the ex post facto test. The remaining question is whether the reenactment of the death penalty via § 13-703 altered McGill's "substantial personal rights," or whether it was merely a procedural change. There was no law that permitted McGill to be punished with a death sentence during the thirty-eight days between Ring I and the re-enactment of § 13-703.¹ I therefore believe that a retroactive death sentence for crimes committed during this period is unconstitutional. Furthermore, the Arizona Supreme Court's decision to the contrary was an unreasonable application of clearly established Supreme Court case law. See 28 U.S.C. § 2254(d)(1).

The State argued, and the majority agrees, that § 13-1105 gave McGill notice of the possibility of a death sentence even after the Supreme Court struck § 13-703 from the books. I believe the opposite is true. After the Supreme Court prohibited application of the portion of § 13-703 that allowed the judge to make factual findings concerning aggravating factors, a look at § 13-1105 would actually put a potential offender on notice that he could not be sentenced to death for first-degree murder. If, as § 13-1105 says, "[f]irst degree murder ... is punishable by death or life imprisonment as provided by § 13-703," and the death penalty prescribed in § 13-703 is a legal nullity, the statute is crystal clear that life imprisonment alone is the punishment for first degree murder committed during the period between Ring I and the reenactment of § 13-703.

In reviewing McGill's ex post facto claim, the Arizona Supreme Court gave it one sentence: "We rejected this argument in State v. Ring [(Ring II)] , 204 Ariz. 534, 547 ¶ 23, 65 P.3d 915, 928 (2003)." State v. McGill , 213 Ariz. 147, 140 P.3d 930, 945 (2006). But the Arizona Supreme Court did not, in fact, reject McGill's argument in Ring II . Instead, in Ring II , the Arizona Supreme Court considered whether the ex post facto clause prohibited the state from resentencing defendants sentenced under the old (unconstitutional) death penalty statute to new death sentences under the new death penalty statute. 204 Ariz. 534, 65 P.3d 915, 926 (2003). The court held that "[t]he new sentencing statutes do not place the defendants in jeopardy of any greater punishment than that already imposed under the superseded statutes. Accordingly, applying the new sentencing statutes does not violate the federal or state Ex Post Facto Clause[s]." Id. at 928 (emphasis added). Ring II examined whether the change from the old § 13-703 to the new § 13-703 was procedural or substantive. That is not the question here. Here, the issue is whether the change from no possibility of the death penalty to possibility of the death penalty was procedural or substantive. It is a very different question from the one answered in Ring II and, thus, has a very different answer.

For the same reason, the cases that the majority relies upon (the same cases that Ring II relies upon) simply do not apply. Dobbert v. Florida , 432 U.S. at 291, 97 S.Ct. 2290, and Collins v. Youngblood , 497 U.S. 37, 38, 110 S.Ct. 2715, 111 L.Ed.2d 30 (1990), addressed situations in which a law in effect at the time of the crime permitted the punishment that was eventually imposed (in Dobbert , the death penalty, and in Collins , a sentence of imprisonment without a monetary fine). That is not the situation here, and I have just explained why the distinction is relevant. (In fact, one might say that the most important distinction in evaluating these ex post facto claims is whether a law in effect at the time of the offense punished that offense.) Treating this case as if it is like Dobbert and Collins was precisely the unreasonable application of Supreme Court precedent that would lead me to grant relief under AEDPA's deferential standard of review.²

2 I believe my colleagues misapply AEDPA's standard of review. I do not think that my colleagues’ opinion that the Arizona court acted reasonably is itself unreasonable . See op. at 706 n.14. Where I differ with my colleagues is that they "need not go so far as to decide that Arizona's" decision was "consistent with the Ex Post Facto Clause," op. at 706, while I believe that the state court decision was inconsistent with the Ex Post Facto Clause—and unreasonably so.

A hypothetical illustrates the error of the majority's position that McGill is not entitled to relief. Between 2003 and 2006, Arizona courts sentenced nineteen individuals to death.³ But imagine that, rather than thirty-eight days, the Arizona legislature waited five years after Ring I to re-enact a death penalty implementation statute. Without reenactment of § 13-703, those nineteen people—everyone acknowledges—would have been sentenced to life in prison. If the Arizona legislature had not passed a new version of § 13-703 until 2006, would the State then be able to retroactively execute the nineteen individuals who had been sentenced to life in prison for murder convictions during those five years? What if the legislature had taken fifteen years to re-enact § 13-703, or thirty? The Ex Post Facto Clause provides that "No State shall ... pass any ... ex post facto Law." U.S. CONST. art. I, § 10, cl. 1. The Constitution does not state: "No State shall pass any ex post facto Law, unless such a law is passed within thirty-eight days." Lengthening the gap between valid death penalty statutes shows the unworkability of the standard that the majority endorses and imputes a "reasonableness" inquiry into the Ex Post Facto Clause when the text mandates a bright line rule.⁴

3 See https://corrections.az.gov/public-resources/death-row.

4 My colleagues in the majority make much of an invented distinction between a law authorizing a "punishment" and a law authorizing a "sentence." Op. at 704-06. The plain text of § 13-1105 puts this argument to bed. That section states that first degree murder "is punishable by death" "as provided by § 13-703." Section 13-1105 is dependent on § 13-703 alone, which provides for the imposition of a death sentence. Thus, the punishment is the sentence. The operative fact for McGill's claim is that at the time of his offense, no law allowed him to be punished with a death sentence.

My colleagues in the majority do quite a bit more to defend the Arizona Supreme Court's ruling on McGill's ex post facto claim than the Arizona Supreme Court did itself (though the majority stops short of endorsing it as correct). However, even the majority's reasoning requires incorrectly assuming that the analysis in Ring II applies with equal force to McGill's situation. That assumption contradicts Supreme Court precedent requiring courts to examine "whether a given change in law presents a ‘sufficient risk of increasing the measure of punishment attached to the covered crimes.’ " Peugh , 569 at 540, 133 S.Ct. 2072 (quoting Garner , 529 at 250, 120 S.Ct. 1362 ) (some internal quotation marks omitted). There was no risk that McGill would be sentenced to death for his crimes at the time they were committed, as such a sentence would run afoul of Gregg . See supra , n.1. By the time of McGill's sentencing, the Arizona legislature had increased the risk of a death sentence for conduct that had already taken place. Thus, McGill could not constitutionally be sentenced to death. The state court failed to reasonably apply Peugh and the three-prong test articulated in Miller when evaluating McGill's federal ex post facto claim. Instead, the state court relied on Dobbert and Collins , cases in which the later penalty imposed was the same penalty that could have been imposed at the time of the offense. That makes Dobbert and Collins not just different from McGill's case but different in the one outcome-determinative way . Therefore, I believe McGill is entitled to habeas relief.

III

I concur in my colleagues’ decision resolving McGill's challenges to the guilt phase of his trial. However, I would grant McGill's petition for the writ of habeas corpus with respect to the penalty phase because I believe sentencing McGill to death is unconstitutional pursuant to the Ex Post Facto Clause. I respectfully dissent.