James Scott v. Ron Broomfield

• Court: Court of Appeals for the Ninth Circuit
• Decided: April 29, 2026
• Docket: 22-99000
• Status: Published

Opinion

FOR PUBLICATION

UNITED STATES COURT OF APPEALS FOR THE NINTH CIRCUIT

JAMES ROBERT SCOTT, No. 22-99000

Petitioner-Appellee, D.C. No. 2:03-cv-00978- v. ODW

RONALD BROOMFIELD, Warden, California State Prison at San Quentin, OPINION

Respondent-Appellant.

Appeal from the United States District Court for the Central District of California Otis D. Wright II, District Judge, Presiding

Argued and Submitted March 19, 2025 San Francisco, California

Filed April 29, 2026

Before: Jacqueline H. Nguyen, Ryan D. Nelson, and Bridget S. Bade, Circuit Judges.

Opinion by Judge Nguyen; Concurrence by Judge R. Nelson 2 SCOTT V. BROOMFIELD

SUMMARY*

Habeas Corpus / Death Penalty

The panel reversed the district court’s grant of habeas corpus relief to James Robert Scott on Scott’s guilt phase ineffective-assistance-of-counsel (IAC) claim, and remanded for consideration of remaining claims, in a case in which a California judge found Scott guilty, following a bench trial, of first degree murder with special circumstances and sentenced him to death. The district court concluded that defense counsel William Clark performed deficiently in numerous ways and these deficiencies cumulatively prejudiced Scott. Applying the deferential standards set forth in the Antiterrorism and Effective Death Penalty Act to the California Supreme Court’s decision denying relief, the panel concluded that Scott’s IAC claim fails. Scott argued that Clark was ineffective in failing to investigate and present a mental state defense. The panel did not need to decide whether Clark’s decisions related to a possible mental defense amounted to deficient performance because the California Supreme Court reasonably concluded that any deficient performance did not result in prejudice. Scott argued that Clark was ineffective in presenting a defense that the victim’s death was not caused by Scott but rather by medical malpractice, a defense Scott asserted was “a non-starter.” The panel held that the California Supreme

* This summary constitutes no part of the opinion of the court. It has been prepared by court staff for the convenience of the reader. SCOTT V. BROOMFIELD 3

Court reasonably held that Clark’s choice in relying primarily on medical malpractice as a cause of death was not deficient. Scott argued that Clark misunderstood the law and was ineffective for arguing that Carlos v. Super. Ct., 672 P.2d 862 (Cal. 1983), required premeditation and not just an intent to kill. The panel held that because Clark argued both that Scott lacked intent to kill, in part due to voluntary intoxication, and that Carlos should be interpreted to require more than intent, the California Supreme Court reasonably concluded that Clark’s advocacy on behalf of his client was effective. Scott argued that Clark was ineffective for informing the judge of his prior guilty plea. The panel held that the California Supreme Court reasonably determined that trial counsel made a tactical decision to introduce evidence of Scott’s guilty plea. Scott argued that Clark was ineffective for advising him to waive a jury trial. The panel held that the California Supreme Court’s finding that Clark’s advice was based on “valid tactical reasons” was not unreasonable. Scott argued that Clark was ineffective for failing to move to suppress his confessions on the grounds that his Miranda waiver was invalid and his statements were involuntary. The panel held that Scott failed to show that counsel would have prevailed on a suppression motion on either ground and thus the California Supreme Court could have reasonably concluded that trial counsel’s performance was adequate. Scott argued that Clark was ineffective for failing to adequately investigate the circumstances of the victim’s 4 SCOTT V. BROOMFIELD

attack and present a third-party culpability defense. The panel disagreed with the district court’s conclusion that hearsay declarations from Scott’s friends—claiming that a third party had admitted to the attack—made a prima facie showing of deficient performance. The panel instead concluded that the California Supreme Court could have reasonably denied this claim on the ground that Clark did not provide deficient performance in failing to investigate a third-party culpability defense. Scott argued that Clark was ineffective for failing to adequately consult with him. The panel concluded that the California Supreme Court could have reasonably denied this claim on the ground that trial counsel did not render deficient performance. Scott claimed cumulative prejudice based on multiple instances of IAC. The panel concluded that the California Supreme Court could reasonably have found no cumulative prejudice. The panel therefore reversed the district court’s grant of habeas relief and remanded for consideration of Scott’s remaining claims. Judge R. Nelson concurred. He wrote separately to explain why cumulative prejudice is never an appropriate basis to grant habeas relief to a state prisoner who asserts ineffective assistance of counsel. SCOTT V. BROOMFIELD 5

COUNSEL

Emily J.M. Groendyke (argued) and Jennifer L. Molayem, Deputy Federal Public Defenders; Kristina A. Harootun, Associate; Cuauhtémoc Ortega, Federal Public Defender; Office of the Federal Public Defender, Los Angeles, California; for Petitioner-Appellee. Gabriel K. Bradley (argued), Herbert S. Tetef, A. Scott Hayward, and Ana R. Duarte, Deputy Attorneys General; James W. Bilderback II, Senior Assistant Attorney General; Lance E. Winters, Chief Assistant Attorney General; Rob Bonta, California Attorney General; Office of the California Attorney General, Los Angeles, California; for Respondent- Appellant.

OPINION

NGUYEN, Circuit Judge:

In 1986, James Robert Scott (“Scott”) raped and beat Wanda Jensen (“Jensen”) in her home, set her on fire, and left her to die. Despite suffering burns on over a third of her body, Jensen survived for ten months. Scott pled guilty to her rape and attempted murder. When Jensen later died due to complications from her injuries, Scott was charged with first degree murder with special circumstances. Following a bench trial, the court found him guilty and sentenced him to death. The California Supreme Court (sometimes referred to as the “state court”) affirmed his conviction and sentence and denied his state habeas petitions. Scott claimed, among 6 SCOTT V. BROOMFIELD

other things, that his trial counsel provided ineffective assistance at both the guilt and penalty phases. A referee took evidence but found insufficient support for these claims, and the state court adopted the referee’s findings. In his federal habeas petition, Scott raised several claims, including ineffective assistance of counsel (“IAC”). In 2021, the district court granted relief on Scott’s guilt phase IAC claim. The district court concluded that defense counsel performed deficiently in numerous ways and these deficiencies cumulatively prejudiced Scott. Applying deference to the state court’s factual findings, see 28 U.S.C. § 2254(d)(2), we conclude that Scott’s IAC claim fails. The California Supreme Court reasonably determined that trial counsel’s performance was not deficient in most respects and that any errors did not prejudice Scott. We therefore vacate the district court’s grant of Scott’s petition and remand for the district court to consider Scott’s remaining claims in the first instance. I. Background A. The Attack Around 3:00 a.m. on April 22, 1986, Jensen’s friend and neighbor, Leah Varela, awoke to the sound of screaming. Varela opened her apartment door and saw Jensen, badly burned, holding her five-year-old daughter. Jensen had white foam “all over” her mouth, and her nightgown had melted and was stuck to her skin, which appeared to be sizzling or bubbling. Varela called 911 for assistance. Jensen told Varela that a man claiming to be “Rerun’s brother Tony” had raped and burned her when she answered a knock on her door around 2:00 a.m. Although Jensen did not know her assailant, she saw him in the parking lot with SCOTT V. BROOMFIELD 7

Rerun minutes before the attack. Rerun, whose real name is Glenn Johnson, Jr., lived in Jensen and Varela’s apartment building and was friends with Scott. When the police arrived, Jensen provided additional details: “Tony” held a screwdriver against her side and forced her into a back bedroom, telling her she would never see her daughter again unless she had sex with him. He grabbed a baseball bat that she kept for protection and began hitting her with it. He proceeded to rape, beat, and choke her until she passed out. When Jensen regained consciousness, she was on fire, and her daughter was screaming. Jensen told the deputies that her attacker used the name “Tony” or “Tony James.” She estimated that he was around 23 years old, was between 5 feet 7 inches and 5 feet 9 inches tall and weighed roughly 140 to 150 pounds. He had scars on the back of his hands and wore a baseball cap and a California Angels jacket. Later that morning, Rerun told a detective that he had given Scott a screwdriver before the fire. Scott came to his apartment around 2:00 a.m. for about 15 minutes, they smoked “coke” and “dope,” and Scott asked Rerun for a screwdriver and gloves. Thinking that Scott planned to steal a car stereo, Rerun gave him a screwdriver and some socks for his hands. Rerun told the detective that Scott was wearing an Angels jacket and baseball cap. Richard Vargas, a witness who interacted with Scott that same night, confirmed seeing Scott at Jensen’s apartment complex around the same time and in a California Angels Jacket and baseball cap. 8 SCOTT V. BROOMFIELD

B. Jensen’s Medical Treatment The attack left Jensen with severe burns on approximately 35 percent of her body and a long horizontal bruise on her chest. Although she did not require immediate breathing assistance at the emergency room, burn damage to lungs often develops in the hours and days after fire exposure and is potentially fatal. Around 8:30 a.m., Jensen was transferred from the hospital to a burn center, where doctors made an incision into her burn tissue to reduce the risk of sepsis. Three days later, Jensen underwent a seven-hour surgery in which doctors removed her dead tissue and performed a skin graft. These procedures required a transfusion of six units of packed red blood cells and the administration of fluids and plasma to address significant blood loss. After the surgery, Jensen was transferred from the intensive care unit (“ICU”) to the burn ward, where her condition began to decline. She had concentrated urine in her bladder with no urine output, and doctors became concerned that she was dehydrated. Doctors administered 300 cc of fluid intravenously, and she drank an additional 4.6 liters. On the evening of April 27, 1986, Jensen suffered complications as a result of adult respiratory distress syndrome (“ARDS”), or “shock lung.” She developed severe shortness of breath, which was relieved with oxygen. Around 3:00 a.m. on April 28, 1986, she suffered cardiac arrest. Doctors resuscitated her for 40 minutes, but she was left in a vegetative state with irreversible hypoxic brain injury. She suffered a second cardiac arrest later that morning around 7:35 a.m. Doctors kept her alive for several months, during which time she contracted several SCOTT V. BROOMFIELD 9

pneumonic and septic infections due to her condition. On February 25, 1987, she died of pneumonia. C. Scott’s Confessions The day after the fire, Scott voluntarily appeared at the local sheriff’s station, accompanied by his mother and sister, and was taken into custody. In separate, unrecorded interviews with Deputy Sheriffs Glenn Snyder and Harmon Watters, Scott confessed to Jensen’s rape and attempted murder. Deputy Snyder advised Scott of his Miranda rights, see Miranda v. Arizona, 384 U.S. 436 (1966), and Scott stated that he understood, did not want an attorney, and was willing to talk. Scott admitted that he beat and raped Jensen in a manner consistent with her account. After initially denying knowledge of how the fire started, he stated that he used a pack of matches from his pocket to light the bedding. When Snyder asked if he set the fire to get rid of Jensen and destroy the evidence, Scott lowered his head and said “yes.” Scott then tensed, raised his head, and glared at Deputy Snyder. Scott stated that “he was now Tony and that Tony had taken charge of James’ body.” “Tony” said that “he had to destroy the girl because she was a shit bomb” and that “James was a wimp.” Speaking again as “James,” Scott repeatedly told Snyder “how bad Tony was.” When Snyder asked Scott if he wanted to talk to his mother, he calmed down and said yes, so Snyder brought her in, and she told her son that he needed psychiatric help. When questioned about Jensen’s daughter, Scott again became tense and glared at Snyder as though he was “ready to go back into [being] Tony.” Snyder believed that the “Tony” persona was just an act, not an alternate personality. 10 SCOTT V. BROOMFIELD

Deputy Watters, who was investigating the arson, also interviewed Scott, and observed scarring on the back of Scott’s hands. When Deputy Watters advised Scott of his Miranda rights and asked him to sign an acknowledgment card, Scott initially used the name “Tony Adman.” Scott changed the signature to James Scott after Watters threatened to end the interview unless he signed with his “true name.” However, Scott stated that he was speaking as “Tony” and that “James was a wimp.” Watters discounted this statement because Scott described his actions on the night of the fire logically and methodically, without further mention of “Tony” or “James.” Scott repeated to Deputy Watters the admissions he made to Deputy Snyder, adding that he had worn an Angels jacket that was damaged by the fire. He also claimed that he had purchased cocaine for Jensen that evening and they “did a couple lines” before he returned to attack her. D. The Charges Scott was charged with attempting to murder Jensen and her daughter, raping Jensen, and burglarizing and setting fire to the apartment. The trial court appointed William Clark (“Clark”) to represent him. Scott agreed to plead guilty to Jensen’s rape and attempted murder as well as to a similar rape he had committed against another victim three weeks earlier. The court sentenced him to 42 years’ imprisonment, which included 26 years for his crimes against Jensen. Nearly a year after Jensen’s death, Scott was charged with first-degree murder and the special circumstances of committing the crime while engaged in rape and burglary, rendering him eligible for the death penalty. See Cal. Penal Code § 190.2(a)(17)(iii), (vii) (1986). The trial court again SCOTT V. BROOMFIELD 11

appointed Clark to represent him. Scott pled not guilty and waived a jury trial. E. Scott’s Trial 1. Guilt Phase At the bench trial, Clark conceded Scott’s identity as the perpetrator. The prosecution proceeded solely on a theory of felony murder, while the defense strategy focused on two independent theories. First, Clark contested whether the prosecution had proven, as was then required, that Scott intended to kill Jensen. See Carlos v. Super. Ct., 672 P.2d 862, 877 (Cal. 1983) (“[T]he felony murder special circumstance . . . requires proof that the defendant intended to kill.”), overruled by, People v. Anderson, 742 P.2d 1306 (Cal. 1987). Clark also contested causation, arguing that Jensen would not have died but for the hospital’s intervening negligence. See People v. Roberts, 826 P.2d 274, 295 (Cal. 1992) (“[W]hen medical treatment is grossly improper, it may discharge liability for homicide if the maltreatment is the sole cause of death and hence an unforeseeable intervening cause.”). i. Intent to Kill Defense Clark argued that Scott had a diminished “actual ability . . . to formulate the intent to kill” due to his drug use the night he attacked Jensen. He elicited Rerun’s testimony that he and Scott “got high” shortly before the attack, that Scott came over “every now and then,” including earlier that day, and that using cocaine was “all [Rerun] did” at the time. In light of Scott’s drug use, Clark argued that Scott’s subsequent actions were “extremely rash, unpremeditated, [and] unconsidered.” In a “rash moment,” Scott “reached in his pocket, realized he had matches, and started a fire.” 12 SCOTT V. BROOMFIELD

Clark pointed out that it is common to carry matchbooks and that Jensen was located “off the bed and on the floor” when Scott set fire to the bedsheets. Clark also argued that Jensen’s statements that Scott “burned me” indicated that she believed he intended to burn her, but that he had no further purpose. Clark attempted to cast doubt on Detective Snyder’s testimony that Scott admitted to setting the fire “to get rid of the girl.” Clark argued that Scott had been “loosened up” by his mother’s presence in the room and that there was no way to know what exactly Scott had acknowledged because the interview was not recorded, Scott did not sign a statement acknowledging what he had said, Snyder’s notes were “extremely attenuated” and written “sometime after the conversation,” and Scott’s response to Snyder—“yes”—was ambiguous. ii. Medical Malpractice Defense Clark also argued that Scott did not proximately cause Jensen’s death because she would have survived the burn injuries had she not received negligent medical treatment. Among other things, the defense highlighted Jensen’s low mortality rate, the medical team’s mistakes that allegedly caused ARDS, and the medical team’s failure to anticipate and properly monitor for reabsorption of fluid. Several experts testified on this issue, including Dr. Stanley Drew, Dr. Christopher Swalwell, Dr. Martin Ginsburg, and Dr. Bruce Zawacki. The preliminary hearing testimony of Dr. Drew, who initially treated Jensen at the emergency room, was read into the record pursuant to the parties’ stipulation. According to Dr. Drew, Jensen suffered burns over 35 percent of her body, with 12 percent being third-degree burns. Dr. Drew SCOTT V. BROOMFIELD 13

explained that the mortality rate for burn victims like Jensen was substantial due to fluid loss, heightened risk of infection, and potential lung damage from inhaled heat and smoke. After receiving emergency treatment, Jensen was transferred that morning from the emergency room to the University of Southern California (“USC”) burn ward. Dr. Swalwell, who performed Jensen’s autopsy, testified that her cause of death was “acute pneumonia due to hypoxic encephalopathy which was a result of thermal burns.” According to Dr. Swalwell, the “critical point” was when she had a cardiopulmonary arrest” because of ARDS, which leads to fluid buildup in the lungs. Subsequently, Jensen developed irreversible brain damage, had a second cardiopulmonary arrest, and developed multiple infections, including a fatal bout of pneumonia. In Dr. Swalwell’s opinion, there was a causal relationship between Jensen’s burns and her ultimate death. Dr. Swalwell maintained that it was reasonable to assume that Jensen developed ARDS as a result of the fire given her severe burns and the likelihood she inhaled hot air, smoke, and chemical gases. Dr. Ginsburg, a cardiopulmonary specialist, was the sole witness called by the defense to argue that Jensen did not receive “the usual and customary standard of care.” He explained that Jensen was a poor candidate to receive an escharectomy, a procedure to remove burnt tissue from her body, on the day of that surgery. Dr. Ginsburg testified that Jensen had low levels of albumin, a necessary protein for maintaining oncotic pressure in the blood vessels and preventing fluid from leaking into the lungs. He opined that Jensen was “overloaded” with “fluid” during the escharectomy. Dr. Ginsburg believed that Jensen’s hypoalbuminemia and the fluid overload increased Jensen’s 14 SCOTT V. BROOMFIELD

risk for acute pulmonary edema, a complication that can cause ARDS if not properly treated. Dr. Ginsburg also testified that the medical team at USC failed to provide the “normal customary care” Jensen required after surgery. Despite clear indications of the potential for acute pulmonary edema, Jensen was not kept in the ICU for proper monitoring. Dr. Ginsburg noted that, the night before the first cardiopulmonary event, Jensen exhibited severe shortness of breath during an attempted bowel movement which was relieved after she was given “40 percent oxygen”—a significant red flag that should have prompted a medical consult. However, no physician was called to assess her oxygen levels, order necessary tests, or initiate treatment. Dr. Ginsburg opined that, at this stage, Jensen’s condition was correctable, and the failure to act contributed to her subsequent cardiopulmonary arrests. Dr. Ginsburg testified that after Jensen’s first cardiopulmonary arrest, she was transferred to the ICU and placed on a ventilator. However, her medical team failed to adequately monitor her, which exacerbated her condition. Approximately four-and-a-half hours later, Jensen suffered a second cardiopulmonary arrest, resulting in additional brain damage and a worsening of her vegetative neurological state. Dr. Ginsburg opined that the cardiopulmonary edema and ARDS were not caused by the burn injuries sustained during the attack. He emphasized that continuous monitoring with an electrocardiogram could have detected and addressed her heart irregularities before the arrests occurred. On recross examination, the prosecutor highlighted the anesthesiologist’s surgery notes, which Dr. Ginsburg had overlooked in the medical record. According to the notes, SCOTT V. BROOMFIELD 15

Jensen’s fluid levels were normal immediately before the escharectomy, and she lost much more blood during the procedure than Dr. Ginsburg had originally thought, which appears to justify giving Jensen more fluids. Based on this information, Dr. Ginsburg stated that it appeared that Jensen was given “the appropriate amount of fluid” during the surgery. However, Dr. Ginsburg maintained that Jensen’s low albumin level should have been addressed before the escharectomy and noted that post-operative fluid overload was still a possibility based on Jensen’s vital signs and severe shortness of breath. In rebuttal, the prosecutor called Dr. Zawacki, a burn specialist and director of the burn facility where Jensen was treated. Dr. Zawacki’s involvement in Jensen’s care was limited to advising residents during rounds. In his opinion, Jensen received the appropriate amount of fluid before, during, and after her escharectomy. He disagreed with Dr. Ginsburg that Jensen’s low albumin level rendered her surgery a mistake because postponing it or increasing her albumin level carried risks as well. He also explained that, while high blood pressure and elevated pulse and respiratory rates “might be associated with fluid overload,” such vital signs were typical for burn patients like Jensen who were “in a great deal of pain even w[ith] narcotics.” According to Dr. Zawacki, Jensen’s surgery “was an hour and a half too long probably,” but he could not attribute it to negligence because Jensen “must have just bled more than they anticipated.” Dr. Zawacki opined that the decision to move Jensen back to the burn ward after her surgery instead of keeping her in the ICU was reasonable considering her stable condition at the time. Nevertheless, Dr. Zawacki expected Jensen “to have a less than three 16 SCOTT V. BROOMFIELD

percent mortality rate when treated on our burn service,” and he “really d[id]n’t know what . . . led to her cardiac arrest.” Dr. Zawacki saw no evidence of inhalative smoke damage and ruled that out as a cause of ARDS. His “best guess” for Jensen’s complication was “delayed reabsorption of fluid” that she had received during the first “couple of days” of her treatment, which led to “volume overload to which she did not respond with diuresis leading to pulmonary edema.” In Dr. Zawacki’s 20 years of experience, it was unprecedented for a burn patient as young and healthy as Jensen to fail to urinate the fluids she had received. In retrospect, Jensen’s shortness of breath on the evening before her first cardiopulmonary arrest was “a little red flag.” Her low urine output, however, was “very remarkable and a clear sign that something was wrong.” Dr. Zawacki initially faulted the physician who responded to Jensen’s low urine output for “not going back and checking on the urinary output,” and delaying in placing a Foley catheter, initiating a fluid challenge, and moving Jensen to the ICU. Dr. Zawacki opined that, without the delay, it was “very likely” that Jensen’s cardiopulmonary arrests could have been prevented. However, later in his testimony, Dr. Zawacki realized that he had misread the medical records, and that Jensen had better urine output than he originally thought. Accordingly, while Jensen’s physician still “moved too slow,” Dr. Zawacki was “not as critical” of the physician’s performance. Dr. Zawacki further testified that, without medical treatment, there was a “very high degree of probability that [Jensen’s burn injuries] would be fatal.” The burn injuries “set in motion a chain of events that ultimately led to SCOTT V. BROOMFIELD 17

[Jensen’s cardiopulmonary] arrest, and without the burn [injuries], there would have been no arrest.” While Jensen’s medical team “could have done more to potentially prevent the arrest,” the medical team did not do anything to “affirmatively precipitate the arrest.” iii. Trial Court’s Findings The trial court, Judge Grignon, found that Scott had a clear intent to kill based on the totality of his statements to Deputies Snyder and Watters and the lack of any “other rational explanation for setting fire to the room that Miss Jensen was in.” Judge Grignon also rejected Clark’s argument that Carlos required a higher mental state than just an intent to kill. In her verdict, Judge Grignon found that “ordinary medical negligence occurred in this case,” and that Jensen “would probably have survived . . . in the absence of . . . [that] negligence.” However, Judge Grignon determined that the burn injuries Scott inflicted on Jensen “would have been lethal . . . if she had not received medical treatment,” and that the medical negligence was only a contributing, rather than a superseding, cause of Jensen’s death. Thus, the negligence did not relieve Scott of his responsibility for Jensen’s death. 2. Penalty Phase At the penalty phase, Judge Grignon returned a verdict of death. During the penalty phase proceedings, Clark highlighted Scott’s history of accepting responsibility as demonstrated by his voluntary surrender, confession, and guilty plea, and re-argued the medical malpractice defense as presented at the guilt phase. According to Judge Grignon, the only factors in mitigation were Scott’s acceptance of 18 SCOTT V. BROOMFIELD

responsibility by “turn[ing] himself in and confess[ing], [and] the contributory negligence of the physicians.” Judge Grignon explained that she “spent a considerable amount of time” assessing the medical negligence evidence, and that without this evidence she “would have had little difficulty in arriving at a verdict of death.” In her opinion, Clark “was very correct in his assessment that a court might deal with that issue somewhat differently than a jury . . . in evaluating its significance.” Nevertheless, Judge Grignon concluded that the aggravating evidence outweighed the medical negligence evidence because there were “lots of things that could have happened,” such as the smoke alarm failing to sound and Jensen’s daughter not rescuing her. F. State Habeas Petitions and Referee Hearing Scott appealed to the California Supreme Court, which affirmed the trial court. Scott filed his first state habeas petition in March 1997, alleging various IAC claims. In August 1998, following the conclusion of Scott’s direct appeal, the California Supreme Court issued an order to show cause as to why Scott was not entitled to habeas corpus relief based on his claims that Clark failed to present an adequate mental defense at the guilt phase, failed to present mitigating evidence at the penalty phase, and gave ineffective advice to waive a jury. In re Scott, 61 P.3d 402, 404 (Cal. 2003). The court appointed Los Angeles Superior Court Judge Howard J. Schwab as a “referee” to take evidence and make factual findings regarding those three issues. Id. at 404–05. The referee took testimony from Clark, Scott’s family members and friends, and four medical experts—Drs. Dale Watson, Roderick Pettis, Julie Kriegler, and Kaushal Sharma. Id. at 406–12. SCOTT V. BROOMFIELD 19

1. Testimony from Clark Clark testified about his representation of Scott in 1986 and 1988. He requested an evaluation of Scott’s competency to stand trial due to references to Scott’s “Tony” persona. Clark attempted to discuss the issue with Scott, who “‘giggle[d]’ whenever ‘Tony’ was mentioned.” In re Scott, 61 P.3d at 412. Because the two court-appointed doctors who examined Scott found that his asserted symptoms of mental distress were feigned and opined that he was competent, Clark did not conduct further investigation into a mental health defense. Clark had multiple conversations with Scott about negotiating a plea deal and the possibility of a subsequent prosecution if Jensen died. Clark noted that while the prosecution agreed that Scott’s plea could not be used against him in a later trial, the parties had no agreement about future capital charges. Based on the length of Scott’s negotiated 42-year sentence, as well as discussions “with other prosecutors in the office, and other people in general,” Clark believed that “it would be extremely unlikely . . . that [Scott] would be brought back” to stand trial. Although Scott initially was in “straight denial” about his guilt, Clark said Scott later “confirmed the veracity of the police reports” and made a “full admission” regarding his “involvement.” By the time he was re-appointed in 1988, Clark had gained additional experience and had handled multiple other capital and homicide cases. Clark could not recall how many times he talked with Scott but was certain he talked with him before and after any court appearances and he obtained removal orders to have Scott transferred from the jail to the courthouse for other meetings. However, Clark no longer had the notes of his conversations with Scott due to 20 SCOTT V. BROOMFIELD

“eliminating some of the clutter” in his case file at an unknown date. Clark recalled having multiple conversations about the jury waiver and stated that, in 1988, Scott was more open about his guilt and admitted that the “Tony” persona “was a con.” Clark hired an investigator, Walter Hill, to assist with the case. Hill wrote in his November 1998 report that Scott “never admitted to me that he had been or believed he was Tony and he laughed whenever . . . I broached the subject to him.” Hill’s report, however, noted that Scott had been prescribed the antipsychotic medication Mellaril while in jail after his arrest in 1986. Additionally, Scott told Hill that, a few hours before the attack on Jensen, he drank a pint of Southern Comfort, snorted “crystal mint, . . . a powdery . . . speed . . . type of illegal narcotic substance,” and smoked a marijuana cigarette laced with opium. Scott claimed that he was “loaded” and did not remember anything between the time he left his friend Derrick Tibbs’s apartment and when he arrived at the apartment of one of his sisters sometime after the fire started. Clark was aware that Dr. Sharma’s report noted that Scott described a history of PCP, cocaine, and marijuana use. However, Clark did not see any signs or symptoms that Scott suffered from substance abuse, and Scott’s 1986 probation report, which indicated that Scott was not a habitual drinker or user of marijuana, was consistent with Scott’s statements to him. Clark further testified that, after meeting first with Dr. James Ungar, a trauma specialist, and then with Dr. Ginsburg, he decided to pursue the medical malpractice defense because it was “extremely viable” and “consistent with what [Clark] felt had happened in the case.” Clark focused on this issue for both the guilt and penalty phases because it was the strongest defense he had and presenting a SCOTT V. BROOMFIELD 21

“cafeteria-type” defense would have risked harming his credibility, especially at a bench trial. Clark explained that the purpose of the jury waiver was to present the medical malpractice defense specifically to Judge Grignon. Clark had “years of experience with her as a judicial officer,” considered her “a skyrocket in the legal community,” and knew that both her husband and father-in-law were physicians. Therefore, he was certain that she would be “better able to grasp the intricacies of a medical malpractice defense than a jury.” Moreover, based on his experience practicing in Antelope Valley, Clark was concerned about how the typically conservative jury would react to both the “heinous” and “cross-racial” nature of Scott’s multiple sexual assaults. Clark testified that he discussed the medical malpractice defense and the jury waiver with Scott and allowed Scott to think about it for days, if not weeks. 2. Testimony from Scott’s Family and Friends Scott’s sisters Carrie MacMurray, Cynthia Ford Geiggar, and Candius Ford all testified to alibi accounts that were inconsistent with one another. Scott’s sisters, parents, and friends testified to Scott’s unstable and abusive upbringing, which included domestic violence, sexual abuse, neglect, poverty, drug use, suicide attempts, and bizarre behavior. Some of the testimony regarding Scott’s home life, however, was rebutted by his stepfather. Several witnesses, including Vargas, also claimed that Scott was intoxicated the night of Jensen’s attack. 3. Medical Testimony Drs. Watson, Pettis, and Kriegler testified on behalf of Scott. Dr. Watson testified that he conducted a neuropsychological evaluation of Scott in 1996 and concluded that Scott operated in the low average to 22 SCOTT V. BROOMFIELD

borderline range of intelligence. Dr. Watson stated that there was “evidence” that Scott suffered from mental illness and post-traumatic stress disorder (“PTSD”). Dr. Pettis also testified that Scott suffered from severe PTSD, depression, and drug-induced psychosis. Dr. Pettis opined that Scott was likely psychotic, in a dissociative state, and disorganized in his thinking when he attacked Jensen. Dr. Pettis believed that his diagnosis was corroborated by the fact that Scott, upon being jailed, was prescribed Mellaril and placed in four-point security restraints. Dr. Kriegler testified that she diagnosed Scott with PTSD, multiple mood disorders, and chemical dependency, but that Scott’s mental state at the time of the attack was “beyond the scope of [her] investigation.” Dr. Sharma testified for the prosecution regarding his 1986 evaluation of Scott’s competence. At that time, Dr. Sharma found that Scott “was competent to stand trial” and exaggerated his mental impairment. For example, though Scott “act[ed] as if he was confused” during the interview, he “clear[ed] up in his thinking” when Dr. Sharma “briefly confronted him.” Dr. Sharma also concluded that Scott’s “Tony” persona was fabricated, as his medical history and behavior showed no signs of a psychiatric condition that could cause a personality change. Dr. Sharma opined that one of the hallmarks of multiple-personality disorder is a clear separation between the different personalities. However, during Scott’s interviews with the deputies, Scott indicated that Scott and “Tony” knew about each other and thus were not separate. According to Dr. Sharma, the opinions of Scott’s post- conviction medical experts were flawed because they were based on the inconsistent accounts provided by Scott and his friends and relatives which lacked credibility and contained SCOTT V. BROOMFIELD 23

multiple levels of hearsay. Dr. Sharma opined that Scott did not suffer from psychosis, PTSD, or mood disorders, and instead had antisocial personality disorder and knew right from wrong. In 1986, Mellaril was “essentially . . . used as a tranquilizer,” including “as a chemical restraint in institutional settings,” and Scott was administered the medication after participating in a jailhouse fight. 4. Referee’s Findings The referee found Clark “totally credible” based on the “demeanor and the content of his testimony.” In contrast, the referee found the testimony of Scott’s family and friends to have “very little credible value” because it “appeared to be exaggerated,” “histrionic,” and “recently fabricated after the sentence of death had been imposed.” The referee also accorded “very little weight” to the opinions of Scott’s experts—Drs. Watson, Pettis, and Kriegler—because they were primarily based on “examinations done years after [Scott’s] crimes,” the non-credible statements of Scott’s family and friends, and Scott’s own statements, which conflicted with Clark’s credible testimony. The referee rejected Dr. Pettis’s “diagnosis of dissociative disorganized state,” in part because “the crimes of rape and the burning were committed in a logical sequence.” The referee concluded that, while Clark’s “investigation of potential mental defenses . . . was minimal,” Scott “did not proffer any significant credible evidence that would have been discovered if there had been a more thorough investigation.” Additionally, Clark would not have been able to gather “the bulk of the matters testified to by family and friends” because this information was “recently fabricated and/or greatly exaggerated,” and Scott “himself requested the curtailment of the investigation and 24 SCOTT V. BROOMFIELD

presentation of evidence.” The referee also concluded that Clark’s advice to Scott to waive a jury “was based upon his concerns about the nature of the jury panels in the Antelope Valley,” the “desire to try the case before a certain highly- skilled jurist,” and the “belief that a defense of intervening medical malpractice would be best presented before that particular judge rather than before a jury.” Moreover, “no additional investigation would have affected the advice to waive a jury” because “there was nothing relative to the investigation that would have changed the dimension of the trial.” 5. California Supreme Court’s Decisions The California Supreme Court adopted the referee’s factual findings, concluded that Scott failed to establish IAC, and denied his first state habeas petition. The court summarily denied other claims that were not at issue in the reference hearing on the merits. Scott filed three other state petitions that included various IAC claims, many of which were repeated from his first petition. The state court summarily denied these claims on their merits, as successive, or as untimely. G. Federal Habeas Proceedings Scott filed a federal habeas corpus petition on January 27, 2004, and an amended petition on May 4, 2010. Scott largely raised the same unsuccessful IAC claims that he had asserted in his state petitions. The district court found that Scott received “constitutionally deficient and prejudicial ineffective assistance of counsel during his 1989 capital murder trial” due to (1) Clark’s failure to challenge the admissibility of Scott’s confession, (2) Clark’s failure to investigate the viability of mental defenses, (3) Clark’s failure to investigate the circumstances of the offense, SCOTT V. BROOMFIELD 25

(4) Clark’s failure to ascertain that the law precluded a defense of intervening causation based on medical negligence, (5) Clark’s mistaken belief that premeditation and deliberation were necessary for a first-degree murder conviction, (6) Clark’s recommendation that Scott waive a jury trial, (7) Clark’s introduction of Scott’s 1986 guilty plea at his 1989 capital trial, and (8) Clark’s failure to adequately consult with Scott. Due to the “cumulative impact of [Clark’s] multiple deficiencies,” the district court found that Scott established prejudice, and therefore, had demonstrated that trial counsel was constitutionally ineffective. The State appealed. II. Standards of Review We review de novo a district court’s grant of habeas relief, including relief on IAC claims, which raise mixed questions of law and fact. Dickinson v. Shinn, 2 F.4th 851, 857 (9th Cir. 2021). The Anti-Terrorism and Effective Death Penalty Act (“AEDPA”) applies to federal habeas petitions, like Scott’s, filed after April 24, 1996. Murray v. Schriro, 745 F.3d 984, 996 (9th Cir. 2014). Under AEDPA, habeas relief is barred unless the state court’s denial of the claim “was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States,” or “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). We review the last reasoned state court decision. Hibbler v. Benedetti, 693 F.3d 1140, 1146 (9th Cir. 2012). If there is no reasoned state court decision addressing a claim of ineffective assistance of counsel, we consider “what arguments or theories could have supported the state court’s summary denial, and then 26 SCOTT V. BROOMFIELD

ask whether it is possible that fair-minded jurists could conclude that those arguments or theories are consistent with Strickland.” Montiel v. Chappell, 43 F.4th 942, 958 (9th Cir. 2022) (citing Harrington v. Richter, 562 U.S. 86, 96, 102 (2011)). We apply a hybrid of the deference owed under AEPDA and the deference owed under Strickland when the state court provides a reasoned decision on a narrow IAC claim but summarily denies a broader IAC claim. Bemore v. Chappell, 788 F.3d 1151, 1161 (9th Cir. 2015). III. Discussion “To establish ineffective assistance under Strickland, a [petitioner] must demonstrate that: (1) counsel’s ‘performance was deficient’; and (2) counsel’s ‘deficient performance prejudiced the defense.’” Andrews v. Davis, 944 F.3d 1092, 1107–08 (9th Cir. 2019) (en banc) (quoting Strickland v. Washington, 466 U.S. 668, 687 (1984)). We need not address both prongs if the petitioner fails to establish either element. Strickland, 466 U.S. at 697. To be deficient, a “defendant must show that counsel’s representation fell below an objective standard of reasonableness.” Id. at 688. Counsel’s performance must be judged “on the facts of the particular case, viewed as of the time of counsel’s conduct.” Id. at 690. “Counsel in a death-penalty case has a duty to make reasonable investigations or to make a reasonable decision that makes particular investigations unnecessary.” Andrus v. Texas, 590 U.S. 806, 814 (2020) (per curiam) (internal quotation marks omitted). “[S]trategic choices made after thorough investigation of law and facts relevant to plausible options are virtually unchallengeable; and strategic choices made after less than complete investigation are reasonable precisely to the extent that reasonable professional SCOTT V. BROOMFIELD 27

judgments support the limitations on investigation.” Strickland, 466 U.S. at 690–91. Because we must give deference to both counsel’s strategic decisions and a state court decision, our analysis of Strickland’s deficient performance prong is “doubly deferential.” Dunn v. Reeves, 594 U.S. 731, 739 (2021) (per curiam). “With respect to prejudice, a petitioner must demonstrate that, ‘but for counsel’s unprofessional errors,’ there is a ‘reasonable probability’ that the ‘result of the proceeding would have been different.’” Andrews, 944 F.3d at 1108 (quoting Strickland, 466 U.S. at 694). “A reasonable probability is a probability sufficient to undermine confidence in the outcome.” Strickland, 466 U.S. at 694. “In the guilt phase, then, ‘the question is whether there is a reasonable probability that, absent the errors, the factfinder would have had a reasonable doubt respecting guilt.’” Hendricks v. Calderon, 70 F.3d 1032, 1036 (9th Cir. 1995) (quoting Strickland, 466 U.S. at 695). We address each of Scott’s IAC claims below. We start with his claim that Clark was ineffective in failing to investigate and present a mental state defense, which the California Supreme Court denied on the prejudice prong of Strickland. It also denied Scott’s claim that Clark was ineffective for failing to investigate and present mitigating evidence at the penalty phase on the prejudice prong. As to the remaining claims, the state court either found no deficient performance or summarily denied the claims on the merits or on procedural grounds. Specifically, the state court found that Clark was not deficient regarding Scott’s claims that Clark improperly raised a medical malpractice defense, misunderstood the law, improperly introduced Scott’s prior guilty plea, and improperly advised Scott to waive a jury. The state court also summarily denied, on the merits and as 28 SCOTT V. BROOMFIELD

procedurally barred, Scott’s claims for IAC regarding Clark’s failure to file a suppression motion, failure to present a third-party culpability defense, and failure to adequately consult with Scott. Finally, we address the state court’s summary denial of Scott’s cumulative prejudice claim. We conclude that the California Supreme Court reasonably denied Scott’s IAC claims and reasonably found no cumulative prejudice. A. IAC Regarding Mental State Defense Scott argues that Clark was ineffective in failing to adequately investigate and present a mental state defense. Scott raised this claim in his first state habeas petition. In re Scott, 61 P.3d at 405. In denying relief, the California Supreme Court stated that it was a “close question” as to whether Clark’s performance was deficient. Id. at 426. But it determined that it “need not decide whether competent counsel should have investigated further” because there was “no prejudice in the failure to do so.” Id. The court explained that the only possible mental health defenses were legal insanity or that Scott did not intend to kill Jensen. Id. The court found that Scott had not established prejudice because he had not produced any credible evidence of “insanity or absence of intent to kill.” Id. The court further noted that the trial court heard evidence of Scott’s drug use and defense counsel argued that the government failed to establish intent to kill. Id. (citing Scott, 939 P.2d at 372). Scott argues that the state court’s decision “was based on unreasonable findings of fact” because the referee: (1) excluded evidence about abuse suffered by Scott’s siblings outside Scott’s presence; (2) “excluded expert testimony concerning the impact of Scott’s impaired mental state upon his purported Miranda waiver”; and (3) “cited no SCOTT V. BROOMFIELD 29

evidence to support th[e] theory of recent fabrication” by Scott’s witnesses. See 28 U.S.C. § 2254(d)(2). Scott also argues that the state court unreasonably focused on Dr. Pettis’s opinion about Scott’s “disorganized and dissociative state” without considering Dr. Pettis’s additional opinion that Scott was “psychotic.” Scott claims that if Clark had investigated and presented the entirety of his mental health history at trial, it would have shown that he was psychotic at the time of his attack on Jensen and lacked the intent to kill. Like the California Supreme Court, we need not decide whether Clark’s decisions related to a possible mental defense amounted to deficient performance because the state court reasonably concluded that any deficient performance did not result in prejudice. See Strickland, 466 U.S. at 697. As an initial matter, Clark could not have presented a general defense of diminished capacity under California law at the time of trial, which is what Scott appears to present on appeal. See Noguera v. Davis, 5 F.4th 1020, 1048 (9th Cir. 2021) (“[B]ecause [the] murder took place after June 1982, a diminished capacity defense was unavailable to [petitioner] under California law.”). Instead, Clark could have presented mental health evidence only to show diminished actuality (i.e., that Scott did not form the specific intent to kill Jensen). Cal. Penal Code § 28(a). Here, Scott’s only relevant post-conviction evidence concerning diminished actuality was Dr. Pettis’s testimony that, at the time of the attack on Jensen, Scott was psychotic and his thinking was in a dissociative and disorganized state. However, Dr. Pettis only opined that it was “not even clear” whether Scott had the intent to kill. Further, the state court, adopting the referee’s conclusion, reasonably rejected Dr. Pettis’s opinion because it was based on information from Scott’s non-credible lay witnesses and was rebutted by the 30 SCOTT V. BROOMFIELD

intentional nature of Scott’s attack on Jensen and his confessions. In re Scott, 61 P.3d at 426. Therefore, the state court reasonably concluded that Scott failed to show prejudice. We also reject Scott’s challenges to the state court’s factual findings. First, Scott does not explain how the exclusion of additional background evidence affected the state court’s determination that he was not prejudiced by trial counsel’s failure to adequately investigate and present a mental defense. Scott does not explain how such evidence impacted the presentation of his diminished actuality defense, especially considering that Dr. Pettis opined that Scott was mentally impaired. Second, while Scott argues that the referee erred in excluding expert testimony concerning the impact of his mental state upon his Miranda waiver, the issue of Scott’s Miranda waiver was not within the scope of that hearing. Id. at 423. Third, the referee cited evidence to support his findings that Scott’s lay witnesses lacked credibility. For example, the referee discussed the fact that Scott’s sisters offered inconsistent alibis as to Scott’s whereabouts on the night of the attack on Jensen, but Scott himself admitted his involvement in the incident. The referee also discussed that Stewart falsely claimed that Scott’s admitted rape of her mother was a “consensual affair.” Finally, the state court’s determination of Scott’s claim of ineffective assistance of counsel would not have been affected by a focus on Dr. Pettis’s diagnosis of psychosis instead of, or in addition to, his diagnosis of a disorganized and dissociative state. Dr. Pettis’s psychosis diagnosis was subject to the same criticism as the disorganized and dissociative state diagnosis—it was based on non-credible lay testimony and was rebutted by contrary evidence that was more reliable. See id. at 426. SCOTT V. BROOMFIELD 31

As the state court noted, the evidence at trial showed Scott’s voluntary intoxication on the night of the attack on Jensen and Clark relied on this evidence in arguing that Scott lacked the ability to “formulate the intent to kill” and urging Judge Grignon to find that Scott lacked the intent to kill. Id. Judge Grignon was unpersuaded and found that Scott had the intent to kill based on the nature of his attack on Jensen and his confessions to the deputies. Scott does not explain how additional evidence of his drug use or mental health background would have altered the outcome of the trial. As such, even if Clark’s performance was deficient in failing to further investigate and present a mental defense, the state court reasonably held that Scott failed to show prejudice. B. IAC Regarding Medical Malpractice Defense Scott argues that Clark was ineffective in presenting the defense that Jensen’s death was not caused by Scott but rather by medical malpractice. In denying this claim on direct appeal, the California Supreme Court held that, even assuming “the defense was a forlorn hope,” that assumption “d[id] not establish that counsel was incompetent for trying it” because Clark did not have “a better alternative available.” People v. Scott, 939 P.2d 354, 371 (Cal. 1997). The court reasoned that, “given the 10 months between the crime and the victim’s death, and the strong evidence of malpractice,” Scott likely would have faulted Clark if he had not presented the causation defense. Id. The California Supreme Court also rejected Scott’s argument that the prosecutor’s damaging cross-examination of Dr. Ginsburg based on the anesthesiologist’s surgery notes demonstrated “counsel’s incompetent representation.” Id. The state court explained that: (1) Scott was not arguing that “better preparation might have caused [the defense] to 32 SCOTT V. BROOMFIELD

succeed”; and (2) “[t]he prosecutor’s effective cross- examination d[id] not demonstrate that a better alternative [defense] existed.” Id. Additionally, the medical malpractice defense “had a spillover effect on the penalty phase,” allowing counsel to argue that medical malpractice was at least mitigating and counsel supplemented that defense at the guilt phase with a defense of “diminished actuality” based on Scott’s voluntary intoxication due to cocaine use. Id. at 371–72. As such, the California Supreme Court held that Clark’s performance was not deficient. On appeal, Scott argues that the medical malpractice defense was “a non-starter.” However, as the state court reasonably explained, the evidence of malpractice was strong, and Scott did not have a more compelling alternative defense. Id. at 371. Dr. Ginsburg’s and Dr. Zawacki’s combined testimony occupies over 400 pages in the trial transcript. While Scott notes that Dr. Ginsburg did not have all the “essential records,” and focuses on Dr. Ginsburg’s concession that the anesthesiologist’s notes refuted his initial opinion that Jensen’s fluid overload arose during the surgery, the medical record was complex and Dr. Zawacki also conceded to overlooking a relevant portion of those records. Significantly, both experts maintained that there was likely some negligence, and both disagreed with the coroner’s opinion that Jensen developed ARDS from the fire. Moreover, as the California Supreme Court explained, the medical malpractice defense “had a spillover effect on the penalty phase.” Id. This is evident from Judge Grignon’s expressed belief that Jensen would have survived with proper medical treatment, her conclusion that “the contributory negligence of the physicians” was a mitigating factor, and her statement that she spent a “considerable SCOTT V. BROOMFIELD 33

amount of time” considering that mitigating factor before she reached a death verdict. While Clark could have chosen to reserve that same evidence for the penalty phase, “the choice of guilt phase defense involves strategic decisions as to both the guilt and penalty phases.” Bemore, 788 F.3d at 1165. “‘[C]omplete innocence’ guilt phase defenses are often risky,” and a defense attorney may present an alternative “longshot” defense if it helps maintain “credibility in advocating for a client’s life at the penalty phase.” Id. at 1168 & n.17. In fact, Clark testified that this was his exact strategy. Finally, Scott’s contention that Clark “failed to reasonably explore any alternatives” to the medical- negligence defense is belied by the record. As the California Supreme Court correctly noted, Clark also presented a mental state defense based on voluntary intoxication. Scott, 939 P.2d at 371–72. Therefore, the California Supreme Court reasonably held that Clark’s choice in relying primarily on medical malpractice as a cause of Jensen’s death was not deficient. C. IAC Regarding Misunderstanding of the Law Scott argues that Clark misunderstood the law and was ineffective for arguing that Carlos required premeditation and not just an intent to kill. See Carlos, 672 P.2d 862. The California Supreme Court rejected this claim on direct appeal, finding that even though Clark’s “forceful legal argument for his client” lacked merit, “[a] good advocate often argues the law is, or should be, more favorable to the client than it actually is.” Scott, 939 P.2d at 372. Additionally, Clark “did not solely argue the absence of premeditation,” but “also argued the prosecution had not 34 SCOTT V. BROOMFIELD

established intent to kill” and “[m]aking multiple layers of arguments is generally effective, not incompetent.” Id. Because Clark argued both that Scott lacked intent to kill, in part due to voluntary intoxication, and that Carlos should be interpreted to require more than intent, the state court reasonably concluded that Clark’s advocacy on behalf of his client was effective. Id. To the extent that Scott argues that Clark’s alleged misunderstanding of the law affected his other trial decisions—such as his decision to recommend a bench trial and introduce Scott’s prior guilty plea—his argument fails for the same reason those claims fail. D. IAC Regarding Introduction of Prior Guilty Plea Scott argues that Clark was ineffective for informing Judge Grignon of his prior guilty plea. During his opening statement, Clark stated that Scott “was committed to a sentence of 42 years in state prison, representing an aggregate number of charges in this case, and it is our position that th[is] was the appropriate measure for a sentence, given the circumstances and given what I hope will develop in the course of the defense testimony.” The prosecutor objected, explaining that the prior plea was “by agreement . . . not admissible,” and that omitting references to it “has to be a two-way street.” Counsel responded that, although it was “very apparent, inasmuch as we are not before a jury, this court is aware of Mr. Scott’s stature before this court,” he would refrain from “argu[ing] the point.” On direct appeal, the California Supreme Court held that there was “a tactical reason for counsel’s actions,” and “[c]ounsel succeeded in having the best of both worlds, an agreement that the plea would not be used in deciding guilt of the capital charge, and the ability to argue that defendant had already been sufficiently punished.” Scott, 939 P.2d SCOTT V. BROOMFIELD 35

at 369. Moreover, “[t]he court . . . understood that the negotiated plea to the noncapital charges was just that, a negotiated plea, and was not to be used in considering the murder charge.” Id. As a result, “[a]lthough counsel’s tactics were unusual, under the equally unusual circumstances, and given the limited options he faced,” the strategy was not unreasonable. Id. at 369–70. The district court rejected the state court’s holding. It concluded that the California Supreme Court unreasonably made the “internally inconsistent” determinations “that it was a reasonable strategy to introduce the guilty pleas in the capital proceeding,” but “that the trial court could not consider the pleas.” Contrary to the district court’s determination, however, there is a distinction between Clark’s attempt to use the plea to argue for leniency, and the parties’ stipulation that the plea could not be used as evidence of guilt. Judge Grignon understood this point—she acknowledged that she could not consider Scott’s prior plea to the noncapital charges and explained how her verdict was in fact based on the evidence at trial. Therefore, we conclude that the state court reasonably determined that trial counsel made a tactical decision to introduce evidence of Scott’s guilty plea. E. IAC Regarding Advice to Waive Jury Scott asserts that Clark was ineffective for advising him to waive a jury trial. The California Supreme Court denied this claim on direct appeal and then denied it again in the first state habeas proceeding “[a]fter reviewing the evidentiary hearing and the referee’s report.” In re Scott, 61 P.3d at 428. The court determined that counsel had the following “valid tactical reasons” for waiving a jury: (1) “the desire to have a particular highly regarded judge decide the 36 SCOTT V. BROOMFIELD

case”; (2) “the nature of the particular jury pool from which a jury would have been selected”; (3) “the nature of the proffered medical malpractice defense”; and (4) “the horrific facts of the crime.”1 Id. The California Supreme Court reasonably applied Strickland’s attorney-performance standard because Clark’s advice to waive a jury trial was “a tactical decision about which competent lawyers might disagree.” Bell v. Cone, 535 U.S. 685, 702 (2002). As explained above, Clark’s decision to present a medical malpractice defense was a reasonable strategic decision, not deficient performance; therefore, Clark reasonably considered the nature of this defense in his recommendation that Scott waive a jury. Further, Clark had tried another capital murder case before a jury in Antelope Valley before he tried Scott’s capital case. See People v. Cudjo, 863 P.2d 635 (Cal. 1993). Accordingly, Clark’s reluctance to place Scott’s “cross-racial” sexual assault history in front of Antelope Valley jurors was likely supported by his firsthand experience with racial bias during his previous trial. See id. at 661. Accordingly, the state court’s finding that Clark’s advice was based on “valid tactical reasons” was not unreasonable. See In re Scott, 61 P.3d at 428.

1 Writing separately, Justice Kennard expressed her view that this issue was “close and difficult” because of “the inherent advantages of jury trial, in which verdicts for guilt and penalty require the unanimous agreement of 12 individual jurors rather than the opinion of a single judge.” In re Scott, 61 P.3d at 430 (Kennard, J., concurring and dissenting). Nevertheless, she agreed with the denial of relief due to lack of prejudice. SCOTT V. BROOMFIELD 37

F. IAC Regarding Suppression Motion Scott argues that Clark was ineffective for failing to move to suppress his confessions on the grounds that his Miranda waiver was invalid and his statements were involuntary. In his original state habeas petition, Scott only raised an IAC claim faulting Clark for failing to challenge his competence to waive his Miranda rights. The state court denied the claim summarily on the merits. In re Scott, 61 P.3d at 423. In his second and third state petitions, Scott raised both grounds. The state court summarily denied the claim on the merits, as successive, and as untimely under In re Robbins, 959 P.2d 311 (Cal. 1998), and In re Clark, 855 P.2d 729 (Cal. 1993). We review the court’s merits ruling and consider “what arguments or theories could have supported the state court’s summary denial, and then ask whether it is possible that fair- minded jurists could conclude that those arguments or theories are consistent with Strickland.” Montiel, 43 F.4th at 958 (citing Harrington, 562 U.S. at 96). “[T]he relevant question under Strickland” is whether “no competent attorney would think a motion to suppress would have failed.” Premo v. Moore, 562 U.S. 115, 124 (2011). Here, Scott fails to show that counsel would have prevailed on a suppression motion on either ground and thus, the state court could have reasonably concluded that trial counsel’s performance was “adequate under Strickland.” Id. “Only if the ‘totality of the circumstances surrounding the interrogation’ reveal both an uncoerced choice and the requisite level of comprehension may a court properly conclude that the Miranda rights have been waived.” Colorado v. Spring, 479 U.S. 564, 573 (1987) (quoting Moran v. Burbine, 475 U.S. 412, 421 (1986)). The latter 38 SCOTT V. BROOMFIELD

element requires a “showing that the accused understood the[] rights” being waived. Berghuis v. Thompkins, 560 U.S. 370, 384 (2010). Scott argues that Clark should have filed a suppression motion due to Scott’s mental health issues, proclamations of innocence, and the unreliable circumstances of the interviews with the deputies. But AEDPA deference applies to the state court’s determination of Scott’s mental state, which is relevant to whether Scott was able to understand his waiver. The California Supreme Court noted that the “strongest evidence of a mental defense [that] petitioner presented” was Dr. Pettis’s opinion and noted that the “reasons” that “the referee found Dr. Pettis’s testimony not credible . . . would have applied to any testimony regarding the Miranda waivers.” In re Scott, 61 P.3d at 423, 426. Further, Clark testified at the hearing that Scott confirmed the accuracy of the deputies’ accounts of his confessions by admitting his involvement in Jensen’s attack and discussing the details of that event, and Scott declined the opportunity to testify. Clark’s testimony was uncontroverted and we agree with the state court’s determination that it “fully support[ed] the referee’s finding that [Scott] did, in fact, confess . . . to trial counsel” and that “[t]his confession confirmed the confessions to the two [deputies].” Id. at 422. Clark also testified that Scott admitted that he fabricated the “Tony” persona, and Scott’s own experts denied that there was evidence that Scott had multiple-personality disorder. Therefore, Clark lacked credible grounds to file a suppression motion. Scott’s argument about the voluntariness of his confession also fails. “A necessary predicate to finding a confession involuntary is that it was produced through coercive police activity.” Cook v. Kernan, 948 F.3d 952, SCOTT V. BROOMFIELD 39

968 (9th Cir. 2020) (internal quotation marks omitted). This predicate does not exist here. Although the interviews were unrecorded and “took place at the sheriff’s station, when Scott was in custody and with no exigency,” these circumstances alone do not prove coercion. Id. (explaining that the court must consider the totality of the circumstances). Accordingly, the state court reasonably could have denied this claim on Strickland’s performance prong on the ground that Clark’s performance was not deficient for failing to file a suppression motion. See Dunn, 594 U.S. at 739. G. IAC Regarding Third-Party Culpability Scott argues that Clark was ineffective for failing to adequately investigate the circumstances of Jensen’s attack and present a third-party culpability defense. Scott raised the claim in his first state habeas petition, and the California Supreme Court denied it summarily on the merits. Scott raised the claim again in his second and third state habeas petitions, and the California Supreme Court summarily denied it on the merits, as untimely, and as successive. In March 2010, Scott filed a fourth state habeas petition based on newly discovered evidence, including Deputy Watters’s 2009 deposition. Scott alleged that the prosecution had improperly withheld information regarding Deputy Watters’s handwritten investigative notes and that Clark was ineffective for failing to discover it. Scott also “incorporate[d] . . . by reference the briefing on” all his previous IAC claims. The California Supreme Court summarily denied the petition on the merits. In Deputy Watters’s deposition, he recalled that Jensen told him that she had used cocaine at 6:00 p.m. on the evening of her attack. Deputy Watters could not remember 40 SCOTT V. BROOMFIELD

whether Jensen said “she had done the cocaine with James Scott,” but did recall her saying that “James purchased the cocaine for her.” Scott argues that this testimony raises doubts about the identity of Jensen’s attacker, because Jensen claimed that she did not “know” her attacker, yet claimed that Scott had purchased cocaine for her. He suggests the attacker could have been Johnson and that Clark was ineffective for failing to further investigate that possibility. Scott asks the court to assume that he could not have been Jensen’s attacker because “the two of them snorted cocaine together earlier on the night of the assault” but Jensen said that she did not “know” her attacker. However, such an assumption is predicated on the false dichotomy that if Jensen did not personally know her attacker, she could not have recognized him as someone who had purchased cocaine for her. The witness testimony at trial supports the conclusion that personally “knowing” Scott is different than just recognizing him from previous encounters. For example, Varela testified that Jensen told her that “she didn’t know [the attacker], but she had seen him.” Jensen’s boyfriend testified that he had “seen [Scott] in and around the apartments, but didn’t know him.” Further, Jensen’s claim that she did not “know” the attacker makes it even less likely that Johnson was the attacker compared to Scott, because Jensen and Johnson were close neighbors. Additionally, Jensen’s physical description of the attacker did not match Johnson. The evidence alleging that Johnson was the real attacker was based on hearsay declarations from Scott’s friends claiming that Johnson had admitted to the attack, but these declarations were not sufficient to make out a prima facie case under California law. See People v. Madaris, 175 Cal. SCOTT V. BROOMFIELD 41

Rptr. 869, 872–73 (Cal. Ct. App. 1981) (a petitioner may not rely on hearsay evidence to make out a prima facie case of IAC), overruled on other grounds by People v. Barrick, 654 P.2d 1243, 1250 (Cal. 1982). Thus, we disagree with the district court’s conclusion that these declarations made a prima facie showing of deficient performance. We instead conclude that the California Supreme Court could have reasonably denied this claim on the ground that Clark did not provide deficient performance in failing to investigate a third-party culpability defense. See Harrington, 562 U.S. at 102. H. IAC Regarding Lack of Consultation Scott argues that Clark was ineffective for failing to adequately consult with him. Scott originally raised this claim in his third state habeas petition, and the California Supreme Court summarily denied it on the merits, as successive, and as untimely. Although Clark could not recall the number of times he talked with Scott, he testified that he met with Scott each time he came to court and that he also had Scott brought to court so he could talk with him on other occasions. Clark was unable to recall the specific “words” used during his conversations with Scott, but he recalled that they discussed various critical aspects of the case. The California Supreme Court accepted the referee’s findings that Clark’s testimony was credible, and Scott chose not to testify at the state court hearing and did not present any other testimony to refute Clark’s. See In re Scott, 61 P.3d at 406, 422. Applying AEDPA’s deferential standard, we conclude that the California Supreme Court could have reasonably denied this claim on the ground that the trial counsel did not render deficient performance. See Cullen v. Pinholster, 563 U.S. 42 SCOTT V. BROOMFIELD

170, 188 n.12 (2011) (noting that “wholly conclusory allegations” do not state a prima facie case for relief under California law); Ross v. Davis, 29 F.4th 1028, 1054 (9th Cir. 2022) (holding that, “to the extent the California Supreme Court accepted the Referee’s findings of fact, we afford them a presumption of correctness under AEDPA.”). I. Cumulative Prejudice Scott claims cumulative prejudice based on multiple instances of IAC. Scott originally raised this argument in his third state habeas petition, and the California Supreme Court summarily rejected it on the merits, as successive, and as untimely. The district court concluded that the state court’s summary denial was unreasonable under AEPDA because, “in light of the lack of evidence against [Scott],” counsel’s “multiple deficiencies were cumulatively prejudicial” and “combined to effectively fail to subject the prosecution’s case to adversarial testing.” See United States v. Cronic, 466 U.S. 648, 659 (1984). The district court relied on the alleged deficiencies, as detailed above, to find cumulative prejudice on the ground that had counsel not rendered ineffective representation, there was a reasonable likelihood of a different result. As an initial matter, the State argues that Scott’s cumulative prejudice claim is procedurally barred because the state court alternatively denied it as untimely. Because we find that Scott’s cumulative prejudice claim fails on the merits, we need not address the state court’s procedural ground. See Ayala v. Chappell, 829 F.3d 1081, 1096 (9th Cir. 2016) (noting that “where claims are clearly not meritorious, appeals courts are empowered to, and in some cases should, reach the merits of habeas petitions despite an SCOTT V. BROOMFIELD 43

asserted procedural bar . . . .” (internal quotation marks and ellipsis omitted)); 28 U.S.C. § 2254(b)(2) (“An application for a writ of habeas corpus may be denied on the merits, notwithstanding the failure of the applicant to exhaust the remedies available in the courts of the State.”). “IAC claims that sufficiently demonstrate counsel’s deficient performance under Strickland but are insufficient to establish Strickland prejudice under AEDPA review are considered for their prejudicial effect in a cumulative error analysis.” See Michaels v. Davis, 51 F.4th 904, 930 (9th Cir. 2022). As we previously explained, we conclude that the California Supreme Court reasonably determined that counsel’s performance was not deficient with the exception of trial counsel’s failure to present a mental state defense, which the state court reasonably denied on the ground that Scott failed to show prejudice. As such, the state court could have reasonably found no cumulative prejudice. See McGill v. Shinn, 16 F.4th 666, 685 (9th Cir. 2021) (“If there are no errors, there is no need to consider their cumulative effect.”). The state court also could have reasonably determined that there was no cumulative prejudice given the strong evidence of Scott’s guilt, including Scott’s confessions and the corroborating physical evidence. See Clark v. Chappell, 936 F.3d 944, 994 (9th Cir. 2019), amended by, 948 F.3d 1172 (9th Cir. 2020). Accordingly, the district court erred in granting relief. IV. Conclusion Because Scott has failed to satisfy AEDPA’s standard for granting habeas corpus relief, we REVERSE the district court’s grant of habeas corpus relief and REMAND for consideration of Scott’s remaining claims. 44 SCOTT V. BROOMFIELD

R. Nelson, J., concurring:

James Scott forced Wanda Jensen onto her bed and raped her, choked her into unconsciousness, lit her bed on fire, and left her to die. Scott confessed to murder and was sentenced to death. Thirty-nine years later, a federal district court held that Scott’s counsel was ineffective and ordered California to release or retry him. Not because Scott is innocent. Nor because Scott’s counsel committed a serious error—the district court couldn’t identify a single error that, on its own, likely affected the verdict. The district court concluded that counsel committed minor errors whose “cumulative impact” undermined “confidence” in the state court’s decision, so it granted habeas relief. We rightly reverse. Even under Ninth Circuit precedent, the district court’s holding is easily dismissed. See Maj. Op. at 42–43. Congress demands that federal courts give more deference to state court judgments. The district court ignored overwhelming evidence establishing Scott’s guilt and substituted its own view of the trial record for the state court’s—thirty-nine years after the fact. I write separately to explain why cumulative prejudice is never an appropriate basis to grant habeas relief to a state prisoner who asserts ineffective assistance of counsel. Start with first principles. State courts “are fully competent to decide federal constitutional issues.” Swain v. Pressley, 430 U.S. 372, 383 (1977). In fact, the Founders viewed states as the key safeguards of the People’s rights. See, e.g., Federalist No. 44 (Madison); Jeffrey S. Sutton, 51 Imperfect Solutions: States and the Making of American Constitutional Law 10–21, 197–202, 214–16 (2018). State SCOTT V. BROOMFIELD 45

courts ensure the constitutionality of their own criminal proceedings. And while the United States Supreme Court may correct state courts’ resolution of federal issues on direct review, there is little reason for inferior federal courts to get involved. Implementing these principles, Congress commands lower federal courts to take a back seat as states resolve the constitutional questions that arise in their proceedings. See 28 U.S.C. § 2254. Generally, we may review constitutional claims through habeas only if those claims were litigated in state court. Id. § 2254(b). But see Martinez v. Ryan, 566 U.S. 1, 10 (2012) (discussing procedurally defaulted claims). And, as relevant here, we may reach the merits of those claims only if the state court’s resolution was “contrary to” or an “unreasonable application of” clearly established Supreme Court precedent. § 2254(d)(1). The state court’s disposition must be so erroneous that no “fairminded jurist[]” could deem it consistent with Supreme Court precedent. Harrington v. Richter, 562 U.S. 86, 101 (2011) (quotation omitted). Otherwise, we defer to the state court’s resolution of the constitutional claim. Under this standard, only Supreme Court precedent counts. § 2254(d); Marshall v. Rodgers, 569 U.S. 58, 63–64 (2013) (per curiam). State courts are not bound by the decisions of the federal courts of appeals. If state courts reasonably apply Supreme Court precedent, habeas is off the table. In addition, even when a state court departs from a principle laid out in Supreme Court precedent, we cannot grant habeas unless that principle was “clearly established.” § 2254(d)(1). That is, only Supreme Court “holdings” matter; a state court’s disagreement with Court “dicta” is not enough for habeas. Williams v. Taylor, 529 U.S. 362, 412 (2000). 46 SCOTT V. BROOMFIELD

With this in mind, habeas is never appropriate when a state court rejects a Sixth Amendment claim that relies on a theory of cumulative prejudice. Under Supreme Court precedent, the Sixth Amendment is violated when state- provided counsel commits an unreasonable error that likely affects the outcome of trial. Strickland v. Washington, 466 U.S. 668, 687, 691 (1984). But the Supreme Court “has never explicitly addressed” whether counsel violates the Sixth Amendment by committing a series of small errors that, while not individually prejudicial, could have impacted the outcome of the trial in the aggregate. See Marshall, 569 U.S. at 62 (quotation omitted). And since the Supreme Court has never endorsed cumulative prejudice, a state court cannot “unreasonably apply” or act “contrary to” Court precedent by denying a cumulative prejudice claim. See Knowles v. Mirzayance, 556 U.S. 111, 121–22 (2009). The Sixth and Eighth Circuits reached this conclusion long ago. See Middleton v. Roper, 455 F.3d 838, 851 (8th Cir. 2006) (collecting cases); Moreland v. Bradshaw, 699 F.3d 908, 931 (6th Cir. 2012). Both the Fifth and the Eleventh Circuits have suggested in unpublished opinions that they would follow suit. See Hill v. Davis, 781 F. App’x 277, 280–81 (5th Cir. 2019) (per curiam); Wood v. Dep’t of Corr., 793 F. App’x 813, 818 (11th Cir. 2019) (per curiam). And no circuit to consider whether Supreme Court precedent clearly mandates cumulative prejudice has gone the other way.1 The Court “has never squarely held that the

1 The Tenth Circuit applies cumulative prejudice on habeas review but has declined to decide whether Supreme Court precedent clearly establishes that approach. See Hooks v. Workman, 689 F.3d 1148, 1194 n.24 (10th Cir. 2012); Hancock v. Trammell, 798 F.3d 1002, 1025 n.17 (10th Cir. 2015); see also Saranchak v. Pa. Dep’t of Corr., 802 F.3d 579, 590 n.7 (3d Cir. 2015) (declining to resolve a similar question). SCOTT V. BROOMFIELD 47

cumulative error doctrine governs ineffective assistance of counsel claims.” Hill, 781 F. App’x at 280–81. Arguing for a different outcome, Scott grasps at straws. In Strickland, the Supreme Court stated that a defendant must show that counsel’s “errors”—plural—likely affected the outcome of trial. 466 U.S. at 687. Scott takes the plural as an implicit endorsement of cumulative prejudice. But Strickland never discussed cumulative prejudice. Nor did it suggest that anything of substance—let alone an entire prejudice doctrine—hinged on the plural form of “errors.” We cannot take the Court’s unexplained decision to use the plural as an implicit endorsement of an otherwise undiscussed doctrine. See St. Mary’s Honor Ctr. v. Hicks, 509 U.S. 502, 515 (1993) (courts should not “dissect the sentences of the United States Reports as though they were the United States Code”); see also Steel Co. v. Citizens for a Better Env’t, 523 U.S. 83, 91 (1998) (matters assumed but not discussed are not precedential). If the Supreme Court’s dicta is not enough for habeas, the Court’s unexplained grammar isn’t, either. See Williams, 529 U.S. at 412. A “fairminded jurist[]” could conclude that the plural does not “clearly establish[]” cumulative prejudice. Harrington, 562 U.S. at 101. Scott’s remaining Supreme Court cases fare no better. When counsel fails to present mitigation evidence, Supreme Court precedent requires courts to “reweigh” the aggravating evidence against “the totality of the available mitigation evidence.” Williams, 529 U.S. at 397–98; Wiggins v. Smith, 539 U.S. 510, 536 (2003); Porter v. McCollum, 558 U.S. 30, 41 (2009). This “totality” includes both the mitigation evidence that counsel admitted at trial and the evidence that counsel ineffectively omitted. Williams, 529 U.S. at 397–98 (criticizing a court for ignoring the omitted evidence). After 48 SCOTT V. BROOMFIELD

all, to determine whether the omitted evidence could have affected the trial, courts must determine whether a jury would have credited the mitigation evidence—all of it—over the aggravating factors. Id. Not whether the jury would credit the aggravating factors over one slice of the available mitigation evidence. Id. This straightforward proposition does not clearly establish that courts must aggregate prejudice when counsel commits many small, unrelated errors. The Court’s mitigation cases do not discuss the cumulative prejudice doctrine. Nor do they address whether courts must aggregate prejudice when counsel fails to commit two distinct errors— say omitting mitigation evidence and offering faulty jury instructions. See Middleton, 455 F.3d at 851. Thus, at the very least, a “fairminded jurist[]” could conclude that these cases do not “clearly establish[]” that the Sixth Amendment is violated every time counsel commits several small errors that are individually harmless but collectively prejudicial. See Harrington, 562 U.S. at 101. In fact, our colleagues on the Eighth Circuit have analyzed the Court’s mitigation cases and reached this very conclusion. Middleton, 455 F.3d at 851. And no circuit has concluded otherwise. If federal circuits disagree with Scott’s reading of Supreme Court precedent, California’s courts cannot be faulted for doing the same. The Supreme Court has never clearly established that cumulative prejudice is part of the Sixth Amendment analysis. Unlike the Supreme Court, we have incorporated cumulative prejudice into the Sixth Amendment analysis. E.g., Doe v. Ayers, 782 F.3d 425, 460 n.62 (9th Cir. 2015). Standing alone, this extension beyond Supreme Court precedent is not problematic. As lower federal courts, we must follow Court precedent, but we are not limited to that SCOTT V. BROOMFIELD 49

precedent. That said, when it comes to habeas, we cannot rely on circuit precedent to “refine or sharpen a general principle of Supreme Court jurisprudence into a specific legal rule that th[e] Court has not announced.” Marshall, 569 U.S. at 64. While we may apply our cumulative prejudice doctrine on direct review of federal criminal trials, we cannot apply the doctrine when reviewing state trials through habeas. Id.; § 2254(d)(1). Again, when it comes to habeas, only Supreme Court precedent counts. Unfortunately, we have not always recognized that limit on our habeas authority. We have granted habeas relief when state courts fail to analyze cumulative prejudice. See Michales v. Davis, 51 F.4th 904, 930 (9th Cir. 2022); Williams v. Filson, 908 F.3d 546, 570 (9th Cir. 2018); Sanders v. Ryder, 342 F.3d 991, 1001 (9th Cir. 2003). That is error. Without Supreme Court precedent endorsing cumulative prejudice, we have no authority to grant habeas relief based on that theory. Worse, we have not even purported to tie this practice to Court precedent. See, e.g., Noguera v. Davis, 5 F.4th 1020, 1051 (9th Cir. 2021) (identifying Court precedent that clearly establishes cumulative prejudice in the due process context, but not in the Sixth Amendment context).2 When we first invoked cumulative prejudice in habeas after Strickland, we transplanted the doctrine from our prior caselaw, not Strickland. See Mak v. Blodgett, 970 F.2d 614,

2 We interpret Supreme Court precedent as mandating cumulative prejudice in the due process context. See Parle v. Runnels, 505 F.3d 922, 928 (9th Cir. 2007) (citing Chambers v. Mississippi, 410 U.S. 284, 298 (1973)). Other circuits disagree. See Williams v. Anderson, 460 F.3d 789, 817 (6th Cir. 2006). In any event, that the Court mandates cumulative prejudice in the due process context does not clearly establish that doctrine in the distinct Sixth Amendment context. 50 SCOTT V. BROOMFIELD

622 (9th Cir. 1992) (citing Cooper v. Fitzharris, 586 F.2d 1325, 1333 (9th Cir. 1978) (en banc)). We never suggested that the doctrine flowed from Court precedent. See id. Later, AEDPA limited our habeas authority to correcting clear deviations from Court precedent. But rather than rethinking our practice of enforcing cumulative prejudice on habeas, we merely continued the practice without discussion. E.g., Boyde v. Brown, 404 F.3d 1159, 1176 (9th Cir. 2005) (citing Cooper, 586 F.2d at 1333); see also Alcala v. Woodford, 334 F.3d 862, 883 & n.7 (9th Cir. 2003) (citing Mak, 970 F.2d at 622). Not once did we try to tie our practice to a Supreme Court opinion. Two decades later, in a mitigation-evidence case, we finally remembered that Supreme Court precedent is the only precedent that matters. In White v. Ryan, a defendant claimed that his counsel failed to investigate “any mitigating circumstances.” 895 F.3d 641, 667 (9th Cir. 2018). Rejecting that claim, the state court separated the mitigation evidence into twelve categories and asked whether each category—in isolation—could have impacted trial. Id. at 672. Because the state court did not “consider[] the mitigation evidence cumulatively,” we held that it “applied a test for prejudice contrary to Strickland.” Id. at 671. We cited the Court’s mitigation cases, which require us to weigh the “totality” of the mitigation evidence against the aggravating factors. Id. at 670–71 (quoting Williams, 529 U.S. at 397–98, and Wiggins, 539 U.S. at 534); see also id. at 671 (citing Strickland’s use of the plural “errors”). Because the state court failed to follow that instruction, habeas was appropriate. Even in White, however, we stopped short of holding that Court precedent clearly establishes cumulative prejudice. While we read Supreme Court precedent to require courts to SCOTT V. BROOMFIELD 51

weigh the “totality” of the mitigation evidence, we did not hold that Court precedent requires courts to aggregate prejudice across counsel’s separate errors. See White, 895 F.3d at 667. As discussed above, those two propositions differ. See Middleton, 455 F.3d at 851. And since the state court overlooked the totality of the mitigation evidence in White, we had no occasion to consider whether Supreme Court precedent establishes cumulative prejudice more broadly. See 895 F.3d at 667. In other words, though we have applied cumulative prejudice on habeas for two decades, never once have we paused to tie that theory to Supreme Court precedent. That matters. When we hold that a principle is clearly established by Supreme Court precedent, later panels are bound by that holding. See Marshall, 569 U.S. at 64. But we are not bound by propositions that are “simply assumed” in our precedent. United States v. Kirilyuk, 29 F.4th 1128, 1134 (9th Cir. 2022). And since we have never “squarely address[ed]” whether Supreme Court precedent mandates cumulative prejudice in the Sixth Amendment context, our assumptions on that question “do[] not bind later panels.” Id. (quoting Brecht v. Abrahamson, 507 U.S. 619, 631 (1993)); Steel Co., 523 U.S. at 91. That is especially true for habeas, where not even our substantive holdings are relevant to the statutory analysis. See Knowles, 556 U.S. at 121–22. Thus, in a future case, we should clarify that petitioners cannot rely on cumulative prejudice to secure habeas relief on a Sixth Amendment claim. And if our panels continue to unwittingly apply cumulative prejudice on habeas review, we should take those cases en banc. As our sister circuits recognize, the Supreme Court has never addressed—let alone clearly held—that the Sixth Amendment requires analysis of cumulative prejudice. A state court therefore 52 SCOTT V. BROOMFIELD

never acts “contrary to” or “unreasonably appl[ies]” Court precedent by rejecting an ineffective-assistance claim that relies on cumulative prejudice.