FOR PUBLICATION

UNITED STATES COURT OF APPEALS

FOR THE NINTH CIRCUIT

DOUGLAS GORDON BRADFORD, No. 21-55038

Petitioner-Appellant, D.C. No. 2:17-cv-05756- v. JAK-JC

DANIEL PARAMO, Warden,

OPINION

Respondent-Appellee.

Appeal from the United States District Court for the Central District of California John A. Kronstadt, District Judge, Presiding

Argued and Submitted November 15, 2021 Pasadena, California

Filed May 3, 2024

Before: Marsha S. Berzon and Johnnie B. Rawlinson, Circuit Judges, and John Antoon II,* District Judge.

Opinion by Judge Rawlinson; Partial Concurrence and Partial Dissent by Judge Antoon

* The Honorable John Antoon II, United States District Judge for the Middle District of Florida, sitting by designation. SUMMARY**

Habeas Corpus

The panel reversed the district court’s denial of Douglas Bradford’s habeas corpus petition challenging his first- degree murder conviction, and remanded with instructions to grant a conditional writ. This “cold case” culminated in a conviction, thirty-five years after a murder, based entirely on circumstantial evidence after the trial judge excluded exculpatory evidence of another viable suspect, Joseph Giarrusso, who dined with the victim on the evening of the murder and was the last known person to see her alive. Applying the standard of review required by the Antiterrorism and Effective Death Penalty Act of 1996, the panel held that in light of Holmes v. South Carolina, 547 U.S. 319 (2006), the decision of the California Court of Appeal was both contrary to and an unreasonable application of clearly established Supreme Court law. The panel held that the decision was also based on an unreasonable determination of facts. First, the California Court of Appeal did not acknowledge that under Holmes, the application of a state evidentiary rule to exclude defense evidence may violate the federal Constitution. Instead, the state court held that “[b]ecause the exclusion was consistent with the rules of

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

BRADFORD V. PARAMO 3

evidence, there was no constitutional violation.” The state court thereby misapprehended the constitutional rule. Second, the California Court of Appeal provided reasons for discounting the Giarrusso evidence, but never weighed its probative value against any “risk of harassment, prejudice, or confusion of the issues,” as is required under Holmes. Third, the California Court of Appeal’s conclusion that the trial court properly considered that Giarrusso “did not in any way” match a witness’s physical description of the suspect was an unreasonable determination of the facts, and one that affected the California Court of Appeal’s ability to access the relevance and probative value of the Giarrusso evidence. As the California Court of Appeal issued no ruling with respect to prejudice, the panel applied the actual-prejudice standard set forth in Brecht v. Abrahamson, 507 U.S. 619 (1993), and concluded that the errors likely had a substantial and injurious effect on the verdict. Concurring in part and dissenting in part, District Judge Antoon agreed with the majority that the state trial court improperly precluded Bradford from introducing evidence regarding Giarrusso, but disagreed with the majority on its determination that the error had substantial and injurious effect or influence in determining the jury’s verdict. He would conclude that the error was harmless and affirm the district court’s denial of habeas relief. Alan S. Yockelson (argued), Law Office of Alan S. Yockelson, San Diego, California, for Petitioner-Appellant. Nikhil Cooper (argued), Deputy Attorney General; Stephanie C. Brenan, Supervising Deputy Attorney General; Susan Sullivan Pithey, Senior Assistant Attorney General; Lance E. Winters, Chief Assistant Attorney General; Rob Bonta, California Attorney General; California Attorney General’s Office, Los Angeles, California, for Respondent- Appellee.

OPINION

RAWLINSON, Circuit Judge:

This “cold case” culminated in a conviction thirty-five years after a murder, based entirely on circumstantial evidence. Unfortunately, the trial judge erroneously excluded powerful exculpatory evidence of another viable suspect, namely the individual who dined with the victim on the evening of the murder and was the last known person to see her alive. Under the compelling facts of this case, we are persuaded that this is one of the rare instances when the state court’s determination that there was no constitutional error in excluding third-party evidence was unreasonable.

I. BACKGROUND

A. Facts In 2014, Douglas Bradford (Bradford) was convicted of the 1979 murder of Lynne Knight (Knight).

BRADFORD V. PARAMO 5

1. The Murder On August 30, 1979, at 3:00 a.m., Knight’s neighbor, Richard Rolleri (Rolleri), was awakened by a scream. He looked through the window, saw a light go out at Knight’s home, and noticed that the gate to the property was “wide open.” Moments later, Rolleri “saw a guy . . . [w]alking across the street,” and decided to check on Knight. When Knight failed to respond, Rolleri called the police. The responding officers discovered Knight’s body lying on a bed with multiple stab wounds and a garrote1 beneath her body. Next to Knight’s body was a broken clasp to a necklace and a gold medallion, but the rest of the necklace was never found. An invitation for Knight’s sister’s wedding was seen crumpled up “in or near a trash can.” Rolleri described the man he saw walking away as “a possible white male, approximately 5'9", with “curly Afro- type dark hair, wearing a beige jacket and pants” and “carrying a small black unknown-type bag.” With this description, officers surveyed the area and discovered Gerardo Juarez (Juarez), approximately a half-mile away. 2. Investigation of Juarez Officers described Juarez as “extremely nervous” and “very jittery” during their conversation. They noticed “what appeared to be fresh, moist blood on [Juarez’s] jacket” and abrasions on his knuckles. When asked about the blood, Juarez explained that he had been mad and hit a telephone

1 A garrote is a device used to accomplish strangulation. One form of a garrote “is a length of wire with wooden handles at the ends, held by the executioner.” Garrote, Britannica Encyclopedia https://academic.eb.com/levels/collegiate/article/garrote/36121 (last visited July 29, 2022). pole. As questioning continued, other officers transported Rolleri to the scene. Upon his arrival, Rolleri faced Juarez and expressed uncertainty about whether Juarez was the person he observed. But once officers directed Juarez to turn his back, Rolleri immediately said: “That looks like the guy.” He added that Juarez’s jacket color, pants color, and hair were “very similar” to that of the person he saw leaving the location of the murder. Juarez was taken to the station, where a criminalist took his jacket to compare its stains to the blood at the crime scene. The criminalist ultimately determined that the spots and stains on Juarez’s clothing were not blood. In a follow- up interview 24 years later, Rolleri recalled that “upon reflection immediately afterward,” he realized that Juarez “was too small to be the person he saw running away” and that “the only similarity” was his “coat and perhaps the general look of his hair.” “Rolleri emphasized that he never saw the face of the person running away and therefore would not be able to conclusively identify this person.” During the same follow-up interview, “Rolleri said that he recalled the events surrounding Knight’s murder ‘like it was yesterday.’” Rolleri related that after he had called the police on the night of the murder, he waited at his window and noticed at a nearby intersection the same “beat up” light blue Volkswagen that had been parked outside Knight’s home before he went to bed. 3. Investigation of Joseph Giarrusso (Giarrusso) Herlinda, Rolleri’s wife, stated that Knight told her she was having a friend over for dinner on the night of the murder. According to Herlinda, Knight stated that this friend had previously “punched her in the mouth,” but “the incident had been a misunderstanding.” On the evening of the

BRADFORD V. PARAMO 7

murder, between 6:00 and 7:00 p.m., Knight came to the Rolleris’ home to borrow cooking utensils, at which time Herlinda looked out her rear window and saw the man she assumed Knight was entertaining. She described the man as “white,” “a little older than Knight,” and “on the heavy side with dark hair that was balding in front.”2 During an interview with detectives, Giarrusso stated that he had a dinner date with Knight hours before she was murdered. He related that he arrived at Knight’s home at approximately 8:30 p.m. and left for his girlfriend’s home at approximately 11:50 p.m. Giarrusso informed detectives that he and Knight had lived together for more than two years, “broke up,” but remained “very, very, very close friends.” When asked if he had intercourse with her the night of the murder, he replied: “Never. I have not made love to Lynne Knight for at least three years . . . [t]wo and a half.” Upon learning of Knight’s murder, Giarrusso described himself as distraught. He also offered to do whatever he could to aid detectives in their investigation and consented to a full search of his car and home. During the interview, Giarrusso indicated that he knew what a garrote was, describing a garrote as “piano wire with two handles” that is “[p]laced around the neck,” resulting in “strangulation or severing of the jugular veins.” Although he denied ever making a garrote, he explained that if he were to make one, “I would make one big enough to loop around the individual’s neck,” whether it be “two feet” or “18

2 By contrast, Bradford was 6'1", 155 pounds, 27 years old (a year younger than Knight), and had a full head of brown hair at the time of the murder. Giarrusso was 5'9", 164 pounds, and 31 years old (3 years older than Knight), at the time of the murder, and according to Herlinda was “balding.” inches, I have no idea.” Giarrusso also told detectives that he had served in the military as an electronic technician during the Vietnam War. When detectives interviewed Giarrusso’s girlfriend, Marcia Collado (Collado), she told detectives that while Giarrusso had his own apartment, he frequently stayed with her. She responded that on the night of the murder, Giarrusso returned from Knight’s house sometime after 11:00 p.m. They discussed his evening with Knight, and eventually went to sleep on her waterbed. She assured detectives that Giarrusso did not leave the house until the next morning, explaining that she is a light sleeper and would have noticed even the slightest movement on the waterbed. Both Giarrusso and Collado agreed to take a polygraph test. The polygraph results were inconclusive for Giarrusso, but reflected that Collado “was not lying on any of the questions asked of her.” Long after Giarrusso passed away in 1994, Collado maintained that she was “100% certain” of the facts regarding his whereabouts. 4. Investigation of Bradford Bradford’s name came up in several interviews with Knight’s close associates. When asked if Knight told him about “trouble with any of the men in her life,” Knight’s boyfriend of five weeks responded that Knight “thought that Doug was getting too serious.” When the same question was asked of Knight’s “good friend” Judy Bradley, she stated that a year or so prior, Knight told her about a man named Doug Bradford who was “serious.” Knight’s friend Mary Coburn (Coburn) knew of eight men Knight was dating, but specifically described Bradford as “sending her roses and the roses would appear on the table in her apartment.” Coburn also recalled Bradford buying Knight a necklace. Giarrusso

BRADFORD V. PARAMO 9

recalled trouble between Bradford and Knight, and particularly described an incident in which Bradford showed up at Knight’s home and found her with another man. He reported that Bradford “whipped the screen door off, made a very horrendous scene and hurt Lynne’s emotional feelings very, very much, to the point she called [Giarrusso] up.” Giarrusso said Bradford left after “20 or 30 minutes of carrying on, calling her a slut, a tramp” and accusing her of giving him a sexually transmitted disease.3 Knight’s co- worker Rita Retort also mentioned Bradford, recalling that Knight told Bradford she did not want him to come over again unexpectedly because “it was embarrassing to her” that he had seen her with another man. Rolleri likewise recalled an incident in which Bradford had “found Knight in her apartment entertaining another boyfriend” and “angrily left Knight’s apartment.” When Rolleri later asked her about this incident, Knight “just laughed it off” and she “expressed no concern about the incident.” Detective Hilton interviewed Bradford, who maintained that he was sailing a boat the night of the murder from 10:00 p.m. to around 1:30 a.m. When asked by Detective Hilton to show him the boat, Bradford refused. Bradford said he dated Knight for several months until they decided in early June, 1979, to no longer see each other. He acknowledged that he had gone to Knight’s home to return some items to her the day after they broke up, and saw a male there, which caused him to become angry and make several unkind statements. Bradford identified this visit as the last time he saw Knight.

3 Our colleague in dissent notes that Giarrusso “only relay[ed] what Knight told him.” See Dissenting Opinion, p. 51, n.4. However, it is undisputed that Giarrusso reported this incident to detectives. Detectives later showed Rolleri’s brother, who lived in the garage adjoining Knight’s home, a photograph of Bradford and asked when he last saw him. Rolleri’s brother reported to the police that he saw Bradford approximately one month prior to Knight’s death, walking on the sidewalk in front of her apartment.4 Detectives suspended the murder investigation in 1982 for lack of sufficient evidence. After the case was reopened in 2000, the police discovered new evidence, including Bradford’s stalking of a girlfriend following their 2009 breakup, and picture-hanging wire found at Bradford’s mother’s home 28 years after the murder “of the same class as” the wire used to create the garrote that was found beneath Knight after the murder. As the government concedes, the case against Bradford was entirely circumstantial. There was no eyewitness identification, no fingerprints, no DNA evidence, no video surveillance evidence, and no confession. B. Procedural History 1. The State Trial Court’s Exclusion of Third-Party Culpability Evidence Before trial, the prosecution moved to exclude evidence related to Giarrusso and Juarez due to insufficient connection to the murder. This motion encompassed evidence of Giarrusso’s presence at Knight’s home on the night of the murder. Citing People v. Hall, 41 Cal.3d 826 (1986) (in bank),5 the prosecution argued that the evidence failed to

4 This approximate date of July is not necessarily inconsistent with Bradford’s statement that he last visited Knight in June. 5 In Hall, the California Supreme Court considered the application of California Evidence Code § 352 to third-party culpability evidence. Section 352 permits trial courts to exclude evidence “if its probative

BRADFORD V. PARAMO 11

raise a reasonable doubt as to Petitioner’s guilt. The state trial court agreed with the prosecution, and excluded the evidence on the basis that there was “nothing to connect [Giarrusso] there at the time of the perpetration of the crime.” Following this ruling, and during its opening statement at trial, the prosecution declared: “Now, what’s very clear is that for several nights leading up to the murder there was one night when Lynne was home by herself. One night only. And that was the night that she was murdered.” This statement would have been directly contradicted if evidence regarding Giarrusso had not been excluded. Bradford challenged the prosecution’s statement as misleading the jury. Bradford argued that it was important for the jury to know that Knight was not alone on the night of the murder. The prosecution responded that who Knight had dinner with that evening was irrelevant, and contended that if the court were to find this fact relevant, the jury needed an instruction that “the person that Lynne Knight had dinner with is not a suspect in this case and you are not to draw any inferences that he was involved in the homicide.” The court determined that the prosecution’s statement was not inaccurate because

value is substantially outweighed by the probability that its admission will (a) necessitate undue consumption of time or (b) create substantial danger of undue prejudice, of confusing the issues, or of misleading the jury.” Cal. Evid. Code § 352. The California Supreme Court held that “[t]o be admissible, the third party evidence need not show ‘substantial proof of probability’ that the third person committed the act; it need only be capable of raising a reasonable doubt of defendant’s guilt.” Hall, 41 Cal.3d at 833. The Court clarified, however, that “evidence of mere motive or opportunity to commit the crime in another person, without more, will not suffice to raise a reasonable doubt about a defendant’s guilt: there must be direct or circumstantial evidence linking the third person to the actual perpetration of the crime.” Id. Knight was alone “during the relevant time period.” Ultimately, a jury convicted Bradford of first degree murder, and he was sentenced to 26 years to life in prison. 2. The Affirmance The California Court of Appeal affirmed Bradford’s conviction. The appellate court concluded that the trial court did not abuse its discretion in excluding evidence related to Juarez because “[his] proximity to the murder scene meant that Juarez had, at most, the opportunity to commit the crime,” which was not sufficient to constitute third-party culpability evidence required to be presented to the jury under Hall. The appellate court remarked that “[t]he only other evidence linking Juarez to the murder was [Rolleri’s] statement that Juarez ‘possibly could have been’ the person leaving Knight’s apartment.” The appellate court deemed Rolleri’s identification unreliable because it “was based entirely on seeing him from behind and was later recanted.” The court added that even if Rolleri’s identification could have been construed as linking Juarez to the actual perpetration of the crime, “it is of such minimal probative value that the trial court did not abuse its discretion in excluding it under [California Evidence Code] § 352 given the absence of any other evidence connecting Juarez to Knight’s murder.”6 The Court of Appeal also concluded that the trial court did not err in excluding the third-party culpability evidence related to Giarrusso “because it reflects, at best, that Giarrusso had the opportunity to commit Knight’s murder” but “opportunity is not enough.” The Court of Appeal

6 Because Juarez was ultimately cleared as a suspect, we do not further discuss the exclusion of evidence related to him.

BRADFORD V. PARAMO 13

determined that the trial court “properly considered that Giarrusso did not in any way match the neighbor’s physical description of the person the neighbor saw leaving Knight’s apartment.”7 In addition, the court concluded that Giarrusso’s knowledge of the side entrance to Knight’s apartment, his knowledge of what a garrote was, that he had a bandage on his thumb a few days after the murder, and that he once hit Knight while they dated did not forge a sufficiently strong link between Giarrusso and Knight’s murder to constitute third-party culpability evidence under Hall. Ultimately, the state appellate court held that “[b]ecause the [trial court’s] exclusion [of the evidence concerning Giarrusso] was consistent with the rules of evidence, there was no constitutional violation.” The California Supreme Court summarily denied Bradford’s petition for review. 3. Federal Court Proceedings Bradford filed a Petition for Writ of Habeas Corpus in the United States District Court for the Central District of California, asserting that the state courts unreasonably applied Holmes v. South Carolina, 547 U.S. 319 (2006). The magistrate judge concluded that the exclusion of third-party culpability evidence relating to Juarez and Giarrusso did not involve an unreasonable application of clearly established federal law, and agreed with the state appellate court’s reasoning. The magistrate judge emphasized that evidence of opportunity to commit the

7 As discussed, the record reflected that Giarrusso’s hair matched Rolleri’s description of the suspect, his height matched one of Rolleri’s initial descriptions, and Herlinda’s description of the person she observed dining with Knight. offense “must be coupled with substantial evidence tending to directly connect that person with the actual commission of the offense.” (citations and emphasis omitted). The district court adopted the magistrate judge’s report and recommendation, denied Bradford’s petition and dismissed the action with prejudice. The district court granted a certificate of appealability, see 28 U.S.C. § 2253(c), on a single issue.

Whether the California courts’ rejection of petitioner’s challenge to the exclusion of third party culpability evidence regarding Joe Giarrusso was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the United States Supreme Court.

Bradford filed a timely appeal to this Court.8

II. STANDARDS OF REVIEW

We review de novo a district court’s denial of a state prisoner’s federal habeas corpus petition. Smith v. Ryan, 813 F.3d 1175, 1178-79 (9th Cir. 2016), as corrected. Because Bradford filed his habeas petition after the 1996 effective date of the Antiterrorism and Effective Death Penalty Act (AEDPA), the AEDPA standard governs our review. See id. at 1179. Under AEDPA, a federal court may only grant habeas corpus relief when the state court’s ruling was (1) “contrary to, or involved an unreasonable application of, clearly established Federal law as determined

8 We deny Bradford’s request to expand the certificate of appealability to include uncertified issues.

BRADFORD V. PARAMO 15

by the Supreme Court of the United States,” or was (2) “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 Williams v. Taylor, 529 U.S. 362, 402-03 (2000). Although we give deference to the state court decision, deference does not equate to abdication. See Miller-El v. Cockrell, 537 U.S. 322, 340 (2003); accord Boyer v. Belleque, 659 F.3d 957, 965-66 (9th Cir. 2011) (“AEDPA requires that we treat the decisions of state courts with deference, but it does not insulate them totally from our review when federal constitutional rights are implicated. . .”). We review the last reasoned decision of state courts. See Curiel v. Miller, 830 F.3d 864, 869 (9th Cir. 2016). In this case, the last reasoned decision is the California Court of Appeal decision on direct appeal.

III. DISCUSSION

A. Unreasonable Application of Clearly Established Federal Law For purposes of 28 U.S.C. § 2254(d)(1), clearly established law refers to “the governing legal principle or principles set forth by the Supreme Court at the time the state court renders its decision.” Lockyer v. Andrade, 538 U.S. 63, 71-72 (2003) (citations omitted). A state court’s decision is contrary to clearly established federal law “if the state court applies a rule that contradicts the governing law set forth in [Supreme Court] cases or if the state court confronts a set of facts that are materially indistinguishable from a decision of [the Supreme Court] and nevertheless arrives at a result different from [Supreme Court] precedent.” Id. at 73 (citations and internal quotation marks omitted). “A state court’s decision is an ‘unreasonable application’ of clearly established federal law if the state court identifies the correct governing legal principle from the Supreme Court’s decisions but unreasonably applies that principle to the facts of the prisoner’s case.” Norris v. Morgan, 622 F.3d 1276, 1289 (9th Cir. 2010) (citations, alteration, and internal question marks omitted). In this case, the clearly established federal law concerns third-party culpability evidence. Third-party culpability evidence is some of the most powerful evidence a criminal defendant can offer to the jury. If believed, reasonable doubt is almost a foregone conclusion. See Lunbery v. Hornbeak, 605 F.3d 754, 762 (9th Cir. 2010). Exclusion of such evidence is often a considerable blow to the defense. See id.9 In Holmes, the Supreme Court considered when the exclusion of third-party culpability evidence violates a criminal defendant’s constitutional right to present a defense. Holmes addressed an appeal from a criminal defendant who had been convicted of murder and related crimes. See 547 U.S. at 322. At trial, the prosecution relied heavily on forensic evidence. The defendant attempted to refute that evidence by presenting evidence that another man was in the victim’s neighborhood on the morning of the assault and either acknowledged defendant’s innocence or admitted to committing the crimes himself. See id. at 323. The trial court excluded this third-party culpability evidence on the basis of a South Carolina Supreme Court case holding that third-

9 A statement to this effect by the investigating detective regarding the lack of forensic evidence was excluded by the trial court.

BRADFORD V. PARAMO 17

party culpability evidence “is admissible if it raises a reasonable inference or presumption as to the defendant’s own innocence, but is not admissible if it merely casts a bare suspicion upon another or raises a conjectural inference as to the commission of the crime by another.” Id. at 323-24 (citation, alterations, and internal quotation marks omitted). On appeal, the South Carolina Supreme Court found no error in the exclusion of the third-party culpability evidence, concluding that “where there is strong evidence of [a defendant’s] guilt, . . . the proffered evidence about a third party’s alleged guilt does not raise a reasonable inference as to the [defendant’s] own innocence.” Id. at 324 (citation omitted). On review, the United States Supreme Court rejected the reasoning of the South Carolina courts, because that reasoning permitted a court to focus on “the strength of the prosecution’s case” rather than “the probative value or the potential adverse effects of admitting the defense evidence of third-party guilt.” Id. at 329. The Supreme Court determined that the South Carolina “rule seem[ed] to call for little, if any, examination of the credibility of the prosecution’s witnesses or the reliability of its evidence.” Id. The Supreme Court characterized the issue to be decided as “whether a criminal defendant’s federal constitutional rights are violated by an evidence rule under which the defendant may not introduce proof of third-party guilt if the prosecution has introduced . . . evidence that, if believed, strongly supports a guilty verdict.” Id. at 321. The prosecution presented strong forensic evidence of the defendant’s guilt, including:

• the defendant’s palm print, found above the interior door knob of the victim’s home • fibers consistent with a black sweatshirt belonging to the defendant, found on the victim’s sheets • matching blue fibers, found on the defendant’s jeans and the victim’s nightgown • consistent fibers found on the victim’s nightgown and the defendant’s underwear • a mixture of the victim’s and the defendant’s DNA, found on the defendant’s underwear • a mixture of the victim’s and the defendant’s blood, found on the defendant’s tank top.

See id. at 322. The prosecution also introduced evidence that the defendant had been spotted in the vicinity of the victim’s home at the approximate time of the murder. See id. The defendant sought to introduce evidence that another man had committed the crime. Specifically, the defendant “proffered several witnesses who placed [the other man] in the victim’s neighborhood on the morning of the [crime],” and other witnesses “who testified that [the other man] had

BRADFORD V. PARAMO 19

either acknowledged that [the defendant] was innocent or had actually admitted to committing the crimes.” Id. at 323 (citation and internal quotation marks omitted). Despite the strong evidence presented by the prosecution, the Supreme Court focused on the constitutional guarantee to criminal defendants of “a meaningful opportunity to present a complete defense.” Id. at 324 (citation and internal quotation marks omitted). The Court explained that this constitutional right is impermissibly “abridged by evidence rules that infringe upon a weighty interest of the accused and are arbitrary or disproportionate to the purposes they are designed to serve.” Id. at 324-25 (citation, alteration, and internal quotation marks omitted). Holmes recognized that the third-party culpability evidence, to be admissible, need not be sufficient to sustain a guilty verdict against the third party. It need only have the potential, considered along with other evidence in the record, to raise a reasonable doubt as to the guilt of the defendant. See id. at 327. The Supreme Court identified the South Carolina rule as one that infringed upon the constitutional right of a criminal defendant to present a complete defense to the jury. The Court described the logic of the South Carolina rule as: “Where (1) it is clear that only one person was involved in the commission of a particular crime and (2) there is strong evidence that the defendant was the perpetrator, it follows that evidence of third-party guilt must be weak.” Id. at 330. The Supreme Court continued:

But this logic depends on an accurate evaluation of the prosecution’s proof, and the true strength of the prosecution’s proof cannot be assessed without considering challenges to the reliability of the prosecution’s evidence. Just because the prosecution’s evidence, if credited, would provide strong support for a guilty verdict, it does not follow that evidence of third-party guilt has only a weak logical connection to the central issues in the case. And where the credibility of the prosecution’s witnesses or the reliability of its evidence is not conceded, the strength of the prosecution’s case cannot be assessed without making the sort of factual findings that have traditionally been reserved for the trier of fact and that the South Carolina courts did not purport to make in this case.

Id. Thus, the state court had erred in excluding the third- party evidence based on the strong forensic evidence presented by the prosecution, rather than weighing the probative value of the third party evidence against its “potential adverse effects.” Id. at 329. In sum, Holmes held that “the Constitution . . . prohibits the exclusion of defense evidence under rules that serve no legitimate purpose or that are disproportionate to the ends that they are asserted to promote.” Id. at 326. At the same time, Holmes recognized that “well-established rules of evidence permit trial judges to exclude evidence if its probative value is outweighed by certain other factors such as unfair prejudice, confusion of the issues, or potential to mislead the jury.” Id. (citations omitted). As we recently summarized under Holmes, a trial court “may, consistent with the Constitution, exclude defense evidence through the proper application of evidentiary rules that serve a valid purpose in a given case, including when proposed evidence

BRADFORD V. PARAMO 21

is ‘only marginally relevant or poses an undue risk of harassment, prejudice, or confusion of the issues.’” Jones v. Davis, 8 F.4th 1027, 1036 (9th Cir. 2021) (quoting Holmes, 547 U.S. at 326-27). In light of Holmes, the decision of the California Court of Appeal was both contrary to and an unreasonable application of clearly established Supreme Court law. The court also unreasonably determined the facts. First, the court did not acknowledge that under Holmes, the application of a state evidentiary rule to exclude defense evidence may violate the federal Constitution. Instead, the court held that “[b]ecause the exclusion was consistent with the rules of evidence, there was no constitutional violation.” The state court thereby misapprehended the constitutional rule. Second, the Court of Appeal provided reasons for discounting the Giarrusso evidence, but never weighed its probative value against any “risk of harassment, prejudice, or confusion of the issues,” as is required under Holmes. Holmes, 547 U.S. at 326-27 (citations, alteration and internal quotation marks omitted); see also Jones, 8 F.4th at 1036. Specifically, the Court of Appeal decided that Giarrusso’s presence in Knight’s apartment on the night of the murder, his knowledge of the side entrance to Knight’s apartment, and his knowledge of what a garrote was, as well as the fact that he had a bandage on his thumb when police spoke with him a few days after the crime, and had once hit Knight while they dated, were facts that did not “forge[] a greater link between Giarrusso and Knight’s murder” than the link that would be formed from evidence of opportunity alone. That assessment is unreasonable: the prior violence, bandage, presence in the apartment rather than simply the opportunity to be there, and knowledge of relevant circumstances, taken together, add up to more than evidence of “opportunity alone.” Rather than allowing the jury to weigh the evidence, the Court of Appeal determined for itself that these facts did not constitute sufficient direct or circumstantial evidence linking Giarrusso to the “actual perpetration of the crime.” The Court of Appeal also provided explanations for Giarrusso’s knowledge of the side entrance and a garrote, as well as the bandage on his hand.10 This approach adjudged the strength of the evidence against Giarrusso under the wrong standard– which, again, is not whether Giarrusso should be found guilty were he on trial, but whether the evidence was probative – that is whether the jury could regard the evidence against Giarrusso, added to the other grounds for reasonable doubt, as a sufficient basis to find Bradford not guilty. The Giarrusso evidence was in no way marginal to the defense. It therefore may not be excluded unless its probative value is outweighed by identifiable adverse effects of admitting it. See Holmes, 547 U.S. at 326-27; Jones 8 F.4th at 1036. The California Court of Appeal identified no risk of undue prejudice, harassment, or confusion that would outweigh the potential benefit to the defense of admitting the evidence. In failing to weigh the potential value of the Giarrusso evidence against any identifiable adverse effects of admitting it, the Court of Appeal did not apply the analysis

10 Our colleague in dissent, in the context of determining prejudice, does the same regarding Giarrusso’s knowledge of a garrote and his bandaged thumb. See Dissenting Opinion, p. 67. But of course assessing such explanations is a question for the jury, not judges. See Holmes, 547 U.S. at 329. Nevertheless, it is interesting to note that Giarrusso had a cut on his thumb when a knife was used to murder Knight, and Bradford had no cut on his hands at all.

BRADFORD V. PARAMO 23

required under Holmes, and its approach was therefore contrary to and an unreasonable application of Holmes. See Holmes, 547 U.S. at 326-27. The evidence regarding Giarrusso was both relevant and probative, and so in no way marginal to the defense. The combination of facts Bradford sought to present had a tendency to make it possible that Giarrusso murdered Knight, and so to provide a basis for the jury to find a reasonable doubt as to Bradford’s guilt. Nor did the evidence have only a “weak logical connection to the central issues in the case,” Holmes, 547 U.S. at 330: the facts discounted by the state court are all circumstances bearing a significant connection to the crime. For example, Bradford was reportedly seen on the sidewalk near Knight’s apartment weeks before the murder, but Giarrusso was in her apartment the night of the murder, like Bradford, Giarrusso had knowledge of the side entrance to Knight’s apartment, and although Bradford had an angry encounter with Knight three months before the murder, he did not physically attack her, and there was evidence Giarrusso had assaulted Knight in the past. The Giarrusso evidence is thus closely analogous to evidence the prosecution was permitted to introduce against Bradford in terms of its probative value. Further, “[t]he jury would not have been confused by such evidence. It would probably have been led to a state of reasonable doubt.” Lunbery, 605 F.3d 762. Notably, the prosecution’s decision to argue to the jury that Knight was alone on the night of the murder, confirms that the undisputed evidence that Giarrusso was with the victim at her apartment in the hours leading up to her murder, was centrally connected to the issues before the jury. Yet the California Court of Appeal completely elided the trial court’s refusal to admit evidence that would have directly contradicted the prosecution’s assertion, in its opening statement, that Knight was alone on the night of the murder. Third, and relatedly, the California Court of Appeal’s conclusion that the trial court properly considered that Giarrusso “did not in any way” match Rolleri’s physical description of the suspect was an unreasonable determination of the facts, and one that affected the California Court of Appeal’s ability to access the relevance and probative value of the Giarrusso evidence.11 Although Rolleri gave varying descriptions of the person he saw running away from the murder scene, Giarrusso matched one consistent aspect of Rolleri’s initial description of the suspect. Rolleri repeatedly described the suspect as having “dark, curly hair,” “curly Afro-type dark hair,” or “black, curly hair.” A photograph of Giarrusso at the time depicted him with dark curly hair. Herlinda’s description of Giarrusso also referenced dark hair. Additionally, one of Rolleri’s initial descriptions stated that the fleeing suspect was approximately 5'9" and Giarrusso was 5'9." The Court of Appeal improperly ignored the similarities between these descriptions and Giarrusso. When combined with the evidence that, when the police were arriving on the night of the murder, Rolleri spotted Giarrusso’s car, the evidence that Giarrusso resembled the fleeing suspect obviously increased the possibility that Giarrusso was present at the time of the murder, and increased the probative value of the Giarrusso evidence as a whole.

11 An unreasonable determination of facts is a separate basis for granting habeas relief. See Harrington v. Richter, 562 U.S. 86, 98 (2011) (explaining that an application for a writ of habeas corpus may be granted if a decision “was based on an unreasonable determination of the facts in light of the evidence presented in the [s]tate court proceeding”).

BRADFORD V. PARAMO 25

Accordingly, the state appellate court’s decision was contrary to and an unreasonable application of clearly established federal law as set forth by the Supreme Court in Holmes, and also an unreasonable determination of the facts. See Bemore v. Chappell, 788 F.3d 1151, 1160 (9th Cir. 2015) (holding that AEDPA relief may be granted “if the last state court merits decision was . . . contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court” or “was based on an unreasonable determination of the facts in light of the evidence presented in the [s]tate court proceeding.”) (citation and internal quotation marks omitted). Applying the legal standard established in Holmes, we conclude that the exclusion of the Giarrusso evidence violated Bradford’s constitutional right to present a defense. B. Prejudice Bradford is “not entitled to habeas relief based on trial error unless [he] can establish that it resulted in actual prejudice.” Brecht v. Abrahamson, 507 U.S. 619, 637 (1993) (citation and internal quotation marks omitted). Actual prejudice is assessed by inquiring “whether the error had substantial and injurious effect or influence in determining the jury’s verdict.” Id. (citation and internal quotation marks omitted). It is at this stage that the strength of the case against Bradford or lack thereof becomes important. As the California Court of Appeal issued no ruling with respect to prejudice, we need only apply Brecht, and need not separately evaluate the issue of prejudice under the AEDPA standard. See Brown v. Davenport, 596 U.S. 118, 127 (2022). Illustrative of the application of Brecht in circumstances similar to this case is Lunbery, 605 F.3d at 762. Lunbery, similarly to Bradford, was prevented from presenting third- party culpability evidence. See id. We held that exclusion of the evidence was prejudicial under Brecht. See id. Lunbery’s husband was found dead inside his home, which was previously occupied by a known drug dealer. See id. at 755, 756-57. Three days after the murder, a confidential informant told a detective that “he felt the killing had been a mistake” and that “[t]he intended victim had been [the drug dealer] because he had ripped off several people in town over drug dealings.” Id. at 757 (internal quotation marks omitted). A neighborhood resident also reported seeing a car that was later linked to the drug dealer and his associate outside Lunbery’s home on the morning of the murder. See id. The detective was informed that an associate of the drug dealer was heard in a restaurant saying: “That’s a bummer. My partners blew away the wrong dude.” Id. In sum, “the police had [multiple] independent sources connecting the murder to drugs; [a] motive for the murder had been provided; [a] connection between the intended victim and [the drug dealer’s associate] had been established; [and the drug dealer’s associate] had admitted knowledge of the murderers and of their mistake.” Id. (cleaned up). Nevertheless, the investigation was closed. See id. Nine years later, the investigation was reopened. See id. Investigators visited Lunbery at her home, and informed her that “we think you did it.” Id. After “[t]he interview became intense” and investigators told Lunbery “a secret witness had inculpated her,” she confessed. Id. at 757-58. Lunbery subsequently repudiated her confession. See id. at 758. An expert witness interviewed Lunbery and reported that she had “a well-recognized type of confession-a stress

BRADFORD V. PARAMO 27

compliant false confession.” Id. Before trial, Lunbery sought to introduce evidence that the murder had been committed by the drug dealer’s associates. See id. The trial court denied introduction of the third-party culpability evidence. See id. Meanwhile, the prosecution’s “sole significant evidence against [Lunbery] was her confession.” Id. at 759. Lunbery appealed her conviction to the California Court of Appeal, challenging the trial court’s exclusion of the third-party culpability evidence. See id. The Court of Appeal discounted the strength of the statement that the associate’s “partners blew away the wrong dude” on the basis that the statement, “may have been nothing more than boasting or the mindless remark of someone under the influence of liquor,” and “deemed the statement inadmissible.” Id. After excluding that statement, the Court of Appeal also excluded the evidence of the neighbor seeing the car in front of Lunbery’s home on the morning of the murder. See id. The Court of Appeal ultimately determined that Lunbery’s proffered third- party culpability evidence “lacked sufficient probative value to raise a reasonable doubt of defendant’s guilt.” Id. On review of the federal district court’s denial of habeas relief, we first determined that exclusion of the third-party evidence violated Lunbery’s right to present a defense and that the trial court’s error had a “substantial and injurious effect on the verdict.” Id. at 762. We reasoned that:

Because [Lunbery] was prevented from presenting this evidence, she was prevented from offering any alternate theory as to who might have committed the crime. The jury would not have been confused by such evidence. It would probably have been led to a state of reasonable doubt. The murder called out for a murderer. The trial as conducted by the superior court and as approved by the court of appeal left only [Lunbery] in view as the murderer.

Id. at 762. In this case, the trial court’s rejection of the proffered third-party culpability evidence likewise amounted to the total exclusion of any mention during trial of Giarrusso as an alternate suspect, and, despite the relative weakness of the case against Bradford, left Bradford as the only viable suspect. Again, Bradford was prevented from eliciting testimony that Rolleri saw Giarrusso at Knight’s home early in the evening on the night of the murder, that Rolleri saw Giarrusso’s car at Knight’s home before Rolleri went to bed, as well as at an intersection near Knight’s home just after the murder12, that Giarrusso had a past relationship with Knight that involved an incident of physical violence, and that Giarrusso fit one of the physical descriptions of the fleeing suspect provided by Rolleri. As we recently recognized, “[t]he excluded evidence could have provided the missing link to establish reasonable doubt for the jury: an actual individual who had knowledge, motive, and opportunity.” United States v. Espinoza, 880 F.3d 506, 519 (9th Cir. 2018). As in Espinoza, exclusion of the evidence regarding Giarrusso “precluded [Bradford] from answering the only question that mattered,” id. at 518. If Bradford didn’t kill

12 The dissent discredits this evidence partially due to Rolleri’s “24-year delay in mentioning it.” Dissenting Opinion, p. 59, n.8. However, the delay is similar to that of the detective in recalling the wedding invitation.

BRADFORD V. PARAMO 29

Knight, who did? The trial court’s erroneous exclusion of the exculpatory evidence proffered by Bradford pertaining to Giarrusso was “compounded and magnified,” Lunbery, 605 F.3d at 762, when the trial court permitted the prosecution to falsely state to the jury that Knight was alone on the night of the murder. That, contrary to what the jury was told, there was a plausible perpetrator seen at Knight’s home the night of the murder constituted powerful evidence for the defense that the jury never heard. The proffered third-party culpability evidence was particularly powerful in light of the lack of direct evidence incriminating Bradford, and the fact that Giarrusso was the last person known to have seen the victim alive. The prosecution’s evidence against Bradford consisted of reports that Knight had broken up with Bradford almost three months prior to the murder, a crumpled wedding invitation, statements describing Bradford’s violent reaction after finding Knight with another man, a missing necklace that Bradford helped Knight purchase, picture-hanging wire seized from Bradford’s mother’s home twenty-eight years after the murder that police determined “had the same general characteristics” as the wire used in the garrote, statements from a girlfriend years after the murder describing Bradford’s menacing (but not personally violent) behavior, a report from a behavioral crime scene analyst who opined that Knight’s murder was “motivated by a preexisting interpersonal conflict,” a photograph of Knight recovered from a desk drawer in Bradford’s home dated a month before Knight’s murder, and the fact that Bradford’s alibi about sailing a two-man boat in the dark seemed unlikely. The core of the prosecution’s case was that Bradford was a jealous boyfriend who lost his temper on one occasion – but did not physically attack Knight – nearly three months before the murder, when he saw Knight with another man one day after Bradford and Knight had ended their relationship; Bradford stalked another girlfriend three decades later; and he had a fairly weak alibi. But the fact that Bradford lost his temper under the circumstances (seeing his ex-girlfriend with another man one day after their break-up), on one occasion months before the murder, is not particularly probative of murder, even when combined with the other evidence in the case. As for his actions toward his girlfriend 30 years later, as the California Court of Appeal aptly observed, Bradford “did not kill [her] and was never violent with her; his conduct in stalking [his later girlfriend] thus provides, at most, tepid support for the argument that he had the propensity to kill Knight.” Moreover, there was no evidence that Bradford was present at Knight’s apartment the night of the murder, and the only possible connections between Bradford and the crime scene were some picture wire found twenty-eight years after the murder and the absence of a necklace. Unlike in Lunbery, in which the defendant confessed and later recanted, there was no confession; Bradford has always steadfastly maintained his innocence. The evidence against Bradford in itself was weak enough to raise some doubt in the minds of the jurors. “[W]here the government’s case is weak, a defendant is more likely to be prejudiced” by trial court error. United States v. Frederick, 78 F.3d 1370, 1381 (9th Cir. 1996), and especially by the exclusion of evidence of a different, plausible suspect. That the jury was prevented from learning this information about Giarrusso leaves us in “grave doubt” about whether the California Court of Appeal’s ruling in conflict with Holmes had a substantial influence on the guilty verdict. Kotteakos v. United States, 328 U.S. 750, 765 (1946).

BRADFORD V. PARAMO 31

IV. RESPONSE TO DISSENT

Our colleague in dissent agrees that the trial court erred in excluding evidence pertaining to Giarrusso, the last known person to see Knight alive on the evening of her murder. However, our colleague is of the view that this momentous error was harmless. The dissent, for the most part, does not grapple with the strength of the “proof of third-party guilt.” Holmes, 547 U.S. at 321. Rather the dissent focuses on what it perceives as “strong evidence of Bradford’s guilt.”13 The dissent catalogs the following evidence as “convincing” regarding Bradford’s guilt. Dissenting Opinion, p. 48. A. Bradford’s interview statements, including an alleged reference “to [Knight] as being dead.” Id., pp. 48-50. The evidence offered by the prosecution in support of its allegation that Bradford stated during his initial police interview that Bradford was “dead” was decidedly underwhelming. The defense denied that Bradford said Knight was “dead.” The audio recording was unclear, the detective who conducted the interview did not hear Bradford say “she” was dead, and the detective never included that statement in his report. The critical word “she” appeared in the transcript of the interview and, in one version of the transcript, was typed in a different font. Importantly, the jury was instructed that the transcript was not evidence in any

13 Even so, the prosecution’s evidence in this case was not strong, considering that there was no confession, no eyewitness, no DNA evidence, no blood evidence, no fingerprint evidence and no murder weapon. event.14 And Bradford’s use of the word “dead” was in the context of explaining that his relationship with Knight was over, he had moved on and was “busy . . . dating other people,” and Knight was “history.” The dissent also emphasized that during his first police interview, Bradford “sometimes referred to Knight in the past tense.” Id., p. 49. But the dissent omits that the police were asking Bradford about his past interactions with Knight, during the time they had been dating several months earlier. In that context, it is not surprising that Bradford sometimes referred to Knight in the past tense. He also referred to her in the present tense at times before he was told of her death – stating, for example, that “[s]he’s pretty independent. She . . . does things kind of on the spur of the moment,” and, just before the “dead” reference, “I don’t have any reason to . . . ever see her again or want to see her again.” In addition, the dissent asserts that Bradford equivocated during his police interview about whether Knight was seeing other men while he was dating her, whether he wanted to marry Knight, and whether the decision to end the relationship was his idea or Knight’s. Id., pp. 49-50. These are all highly personal, private topics. Indeed, at one point during the first police interview, Bradford explained that “some of the . . . personal things I think, you know that . . . it’s just kind of hard to talk about it. You know, it was kind of a personal relationship.” That Bradford may have

14 The Dissent says that the version with the different font was not presented to the jury. Dissenting Opinion, p. 49, n.2. But that compounds the error because it deprived the jury of important information reinforcing the lack of proof that this statement was actually made by Bradford.

BRADFORD V. PARAMO 33

provided a face-saving account of his relationship is hardly evidence, whether alone or in combination with the other circumstantial evidence in this case, that he committed murder.15 B. Motive The dissent discusses in some detail an incident when Bradford reacted angrily when he saw another man at Knight’s apartment one day after Bradford and Knight broke up. See id., pp. 50-52. However, it is undisputed that during the incident Bradford never engaged in any act of physical violence against Knight. And the record reflects that following this incident, Knight expressed that she was “embarrass[ed],” rather than fearful. In fact, Knight “laughed off the incident.” And of the two men, Giarrusso was the only one who had ever physically assaulted Knight. But of course, the jury never heard that evidence. The dissent also emphasizes that there was evidence at trial that Bradford was seen “in front of the property a week or two before the murder.”16 Id., p. 52. But that fact pales in comparison to the undisputed evidence that Giarrusso was inside Knight’s home having dinner with her hours before her

15 To the extent the dissent relies on Bradford’s demeanor during his police interviews, it is worth noting that Giarrusso appeared nervous at times during questioning by police. For example, when police noticed the bandage on Giarrusso’s thumb and asked him if they could examine his hands, they then asked him, “Are you a nervous individual?” Later in the interview, police told Giarrusso to “relax” and “calm down,” assuring him “[w]e’re not trying to back you into a corner or nail you to the wall.” 16 The same witness gave a varying account to the police, reporting that he saw Bradford at the property approximately one month before the murder. murder, and his car was spotted at an intersection outside her home immediately after the crime. Again, the jury never heard this powerful third-party evidence. In addition, the dissent references the crumpled wedding invitation. See id., pp. 52-53. But there was absolutely no contemporaneous record of a wedding invitation being found in Knight’s home. The investigating detective only remembered a crumpled wedding invitation “in retrospect” after Knight’s sister insisted that the wedding invitation must have been there. On a different occasion, the detective stated “he had no recollection of any crumpled wedding invitation.” The wedding invitation was never found, but a copy was sent to the detective “at a much later date” by the victim’s sister. The prosecution presented a replica of a crumpled invitation during its Opening Statement for “demonstrative purposes,” but no invitation was ever introduced into evidence because no foundation could be laid. The court ruled that the invitation was not an exhibit. There was no photograph of the invitation and no mention of the invitation in the reports of the investigation. From this sketchy context, the dissent speculates “that Bradford had crumpled [the invitation] up in anger on the night of the murder.” Id., p. 53. Further, the dissent makes much of a gold chain that was missing from the crime scene. See id., p. 53. Once again, the dissent speculates that “Bradford had given Knight the chain . . . and that he took it back on the night of the murder.” Id. There is absolutely no evidence in the record to support this speculation, as no chain was discovered during the search of Bradford’s vehicle, his home, or his mother’s home.

BRADFORD V. PARAMO 35

C. Consciousness of Guilt. The portion of the dissent concerning consciousness of guilt focuses primarily on Bradford’s alibi. See id., pp. 53- 55. However, the alibi evidence was not as one-sided as the dissent would have one believe. Bradford told police that he was out sailing that night (a Wednesday) and came back late, around 3 a.m.; there was no “wind,” so he had to paddle his way back. He said he had gone sailing on Monday night as well. He told police where the boat was moored and gave them driving directions. In the same interview, when police asked Bradford’s father whether he remembered Bradford coming home, he corroborated Bradford’s account in part:

“I’m sure he came home late. His mother was awake . . . . I’ll admit I’d gone to bed. . . .But [his mother] mentioned it to me the next morning that, that he was late. And I said what happened? He said well, I got becalmed and that he had to row that thing back by himself.”

Bradford also told police, during the same interview, that “I sign out on the calendar when [] . . . the boat’s going out.” He explained that “we keep a calendar” and “we fill in the name and about the time that we’re leaving and about the time we’re getting back.” When police recovered the sign- out calendar, the calendar reflected that for August 29 (the day in question), there was an entry indicating that someone named “Doug” had signed the boat out for 10 to 12. The calendar also indicated that on the previous Monday of the same week, “Doug” had signed a boat out from 8 to 12. Further, evidence was presented by the defense that there was no official rule regarding sailing at night, and the sign- out logs for 1979 indicated that several individuals had sailed alone at night. And Bradford’s expert stood by his testimony that the wind readings relied on by the prosecution were inaccurate. Of course, it is much easier to attack the alibi of a defendant who is the only suspect. However, if the evidence relating to Giarrusso had been admitted, Bradford’s alibi would have been considered together with the evidence pointing to Giarrusso as the murderer. Our colleague in dissent minimizes the fact that no garrote was found in Bradford’s home, because “it was found under Knight’s body.” Id., p. 56. But it is undisputed that the detectives were searching in 2007 for evidence of a garrote at Bradford’s home because they seized “wire and dowelling” from Bradford’s mother’s garage and home to compare “to the garrote recovered from the crime scene.” The most they could say after the comparisons was that some of the picture wire was “of the same class” as the wire used to make the garrote found at the crime scene. Finally, the dissent focuses on a photograph of Knight in an envelope found in a desk drawer, not hanging in a prominent position in the home. See id., pp. 56-57. But there is nothing incriminating about keeping a photograph of an ex- girlfriend in an envelope in a desk drawer, together with other keepsakes, especially when the desk was not one used regularly. More relevant is what was not found during the search of Bradford’s home. As with the search of his car, the search of Bradford’s home failed to uncover a garrote, wooden handles for a garrote, wire for a garrote, or the chain

BRADFORD V. PARAMO 37

that the dissent speculates Bradford took from the crime scene. D. Bradford’s lack of cooperation. Despite the dissent’s characterization of Bradford as not “entirely cooperative,” id., p. 48, the dissent acknowledges, as it must, that Bradford consented to a search of his car. See id., pp. 55-56. It is notable that the search did not uncover any incriminating evidence — no garrote, no wire for the garrote, no wooden handles for the garrote, no hair, no fibers, and no blood, despite the pathologists’ opinion that the assailant would “have a whole lot of blood on him as a result of the attack.” The best the dissent can muster is that the car “‘reeked of Armor All’ and appeared to have been freshly cleaned.” Id., p. 56. But this observation was made decades after the fact rather than when the search was conducted. And despite the dissent’s focus on Bradford’s purported attempt “to pull off an ‘air swab’” when providing a DNA sample, id., it is undisputed that a DNA sample was gathered from Bradford, and there was no match to Bradford’s DNA at the crime scene. E. Giarrusso. The dissent ends its factual recitation of the “convincing evidence” by attempting to minimize the fact that the description Herlinda gave of Giarrusso, Knight’s dining companion, more closely matched one of Rolleri’s initial descriptions of the suspect than Bradford did. Id., pp. 59-65. The night of the murder, shortly after Rolleri saw the suspect fleeing, he told the police that the suspect was “a possible white male, approximately 5'9", with “curly, Afro-type dark hair.” Police immediately used this description to search the surrounding area for the suspect shortly after 3 a.m.. If the Giarrusso evidence were presented to the jury, the jury could conclude that Giarrusso fit the height and hair color of the suspect described by Rolleri more than Bradford did.17 Bradford was described by one trial witness as 5'10" or 5'11" with a “slight build” and a “lot of hair” that was “sandy” in color. Bradford was similarly described at trial as 5'10" or 5'11" with a “slight, medium to slight” build, and “sandy features.” Bradford’s mother described him as 6'1" and “slender.” Bradford weighed 155 pounds. In contrast, Giarrusso was described by Herlinda as “on the heavy side with dark hair that was balding.” He was also 5'9", the height Rolleri gave early on for the suspect. Giarrusso weighed 164 pounds. Herlinda’s description of Giarrusso confirms that Giarrusso’s physical characteristics matched her husband’s description of the fleeing suspect. This is especially true in view of our colleague’s reliance on the existence of “several descriptions of the suspect” in the record. Id., p. 60. Our colleague in dissent questions why the majority relies on the description in the police report. See id., pp. 61- 62. The answer is easy: because that is the description the police elected to document. By referencing other descriptions in an attempt to discount this third-party evidence the defense could have presented, the dissent once again strays from our task. If the prosecution elected to challenge the defense evidence on the bases identified by the

17 The dissent criticizes the description in the police report because it was linked to the investigation of Juarez. See id., pp. 60, 62. But it is undisputed that this report was part of the record and reflected the facts uncovered by the police while investigating Knight’s murder, including the height and hair color of the fleeing suspect, which did not match Bradford.

BRADFORD V. PARAMO 39

dissent, the jury would have been called upon to weigh this conflicting evidence. Moreover, the fact that Rolleri provided varying descriptions of the suspect, some aspects of which matched Giarrusso more than Bradford, reinforces that the admission of the Giarrusso evidence quite likely could have tipped the scales toward reasonable doubt. As the dissent notes, an officer on the scene relayed to dispatch a description of the fleeing suspect as being “a white male adult, 25 years, 6 foot, medium build . . . dark curly hair.” Id., p. 60. The next morning, Rolleri explained to police that he saw the suspect from behind and he was unsure if the suspect was white or black. He again told police that the person he saw fleeing had “black curly hair.” When asked about the suspect’s “build,” Rolleri said that the man “wasn’t big,” but he was “bigger than me.” Rolleri said the man looked “built” or “strong.” When asked by police, Rolleri responded that his own height was 5'9". Although Rolleri’s descriptions of the suspect he saw from behind varied somewhat, he repeatedly described the suspect as having dark, curly hair, a key detail that, if credited by the jury, would exclude Bradford but could inculpate Giarrusso in combination with other evidence. The dissent also questions whether the majority opinion fairly suggests that Giarrusso had a motive to murder Knight. See id., pp. 65-66. In fact, Giarrusso’s motive was the same as that imputed to Bradford. He too had been relegated to the “friend zone.” The dissent speculates that Giarrusso’s circumstance was different because “the majority apparently assumes that Knight broke off the relationship with Giarrusso rather than the other way around.” Id., p. 66. However, we make no such assumption. Giarrusso simply stated that he and Knight “broke up.” And Bradford stated that he and Knight “decided not to see each other any longer.” Regardless of who ended each relationship, the result was the same. And as previously mentioned, Giarrusso, not Bradford, had physically assaulted Knight before. In an effort to avoid the critical impact of the Holmes opinion, the dissent interprets Holmes and Lunsberry as requiring Bradford to submit evidence that would “squarely prove” Giarrusso’s culpability. Id., p. 69.18 Although that was the quality of the excluded evidence in Holmes and Lunsberry, that is not the test for harmless error. Rather, the test for harmless error is whether the wrongfully excluded evidence likely had a substantial influence on the guilty verdict. See Kotteakos, 328 U.S. at 765. “If so, or if one is left in grave doubt, the conviction cannot stand.” Id. At a minimum, the unrefuted evidence that Giarrusso was dining with Knight mere hours before she was murdered would have given the lie to the prosecutor’s outrageous misrepresentation that Knight was murdered on the only night when she was alone. Given the choice between Bradford, who was seen in front of the property a week or two before the murder and Giarrusso, who was seen inside Knight’s home mere hours before her murder, the jury could well have concluded that Giarrusso was the more likely suspect (or at least a plausible suspect). Given the choice between Bradford, who did not meet some of the descriptions of the fleeing perpetrator, and Giarrusso, who more closely matched some of those descriptions, Giarrusso

18 The dissent also observes that Holmes was not a habeas case. See Dissenting Opinion, p. 68. But the precepts and principles set forth by the Supreme Court in Holmes were not limited to the direct-review context.

BRADFORD V. PARAMO 41

could have been seen as the more likely suspect. Given the choice between Bradford, who had never physically attacked Knight and Giarrusso, who had physically attacked Knight in the past, Giarrusso could have been seen as the more likely suspect. Even the investigating detective described the lack of forensic evidence against Bradford as a “blow.”19 The conclusion is inescapable that Bradford’s “conviction cannot stand.” Id. Try as he might, our dissenting colleague cannot muster strong enough evidence of guilt to succeed in minimizing the evidence in favor of the defense that Giarrusso could have been the murderer. See Holmes, 547 U.S. at 329. At bottom, the majority and dissent’s differing views of the strength of the evidence against Bradford proves the point that all of the evidence should have been presented to the jury for consideration in determining Bradford’s guilt or innocence in this case based entirely on circumstantial evidence. Try as he might, our colleague in dissent cannot marshal from this record sufficient proof of guilt for us to say with any confidence that the state court’s error in excluding evidence of third-party guilt should not leave us in “grave doubt” of “substantial and injurious effect or influence in determining the jury’s verdict.” Kotleakos, 328 U.S. at 765, 776; Brecht, 507 U.S. at 638.

V. CONCLUSION

In a murder case built entirely on circumstantial evidence, preventing a defendant from presenting evidence of another possible suspect who admitted to being at the

19 The investigative detective described the complete lack of forensic evidence tying Bradford to the crime as a “severe blow” but the statement was excluded by the trial court. victim’s home on the night she was killed, and who had previously subjected the victim to physical violence at his hands violated the defendant’s right to present a defense. The California Court of Appeal’s decision otherwise was contrary to and an unreasonable application of the Supreme Court’s decision in Holmes. In addition, the California Court of Appeal unreasonably determined the facts. See Taylor v. Maddox, 366 F.3d 992, 1005, 1018 (9th Cir. 2004) (granting habeas relief after a state court admitted a confession but ignored testimony explaining that the confession was a result of intimidation), overruled on other grounds by Murray v. Schriro, 745 F.3d 984, 999-1000 (9th Cir. 2014); see also Lunbery, 605 F.3d at 762 (commenting that deprivation of the right to present a defense “was compounded and magnified by factual mistakes made by the court of appeal in the course of disposing of the evidence offered”). As we have explained, the errors likely had a substantial and injurious effect on the verdict. Accordingly, we reverse the decision of the district court and remand with instructions to grant a conditional writ of habeas corpus, ordering Bradford’s release unless the state of California notifies the district court within thirty days of the issuance of this court’s mandate that it intends to retry Bradford without excluding the evidence pertaining to Giarrusso, and commences Bradford’s retrial within seventy days of issuance of the mandate. See Taylor, 366 F.3d at 1018; see also Lunbery, 605 F.3d at 763. REVERSED and REMANDED.

BRADFORD V. PARAMO 43

Antoon, District Judge, concurring in part and dissenting in part: I agree with the majority that the state trial court improperly precluded Douglas Bradford from introducing evidence regarding Joe Giarrusso. This violated Bradford’s constitutional right to present a defense, and the California Court of Appeal’s approval of the exclusion was contrary to clearly established federal law. But I respectfully part company with the majority on its determination that the error “had substantial and injurious effect or influence in determining the jury’s verdict.” Brecht v. Abrahamson, 507 U.S. 619, 637 (1993) (quoting Kotteakos v. United States, 328 U.S. 750, 776 (1946)). Based on my review of the record, I conclude that the error was harmless under Brecht, and for that reason I would affirm the district court’s denial of habeas relief.

I. Harmless Error Principles

The determination that a state court committed constitutional error does not, of course, complete the required analysis in a § 2254 case. If the error was harmless, no relief is to be granted. And an error is not harmful on habeas review unless it “resulted in actual prejudice”— meaning that it “had substantial and injurious effect or influence in determining the jury’s verdict.” Brecht, 507 U.S. at 637 (quoting Kotteakos, 328 U.S. at 776). It is not enough that “there is a ‘reasonable possibility’ that [the] trial error contributed to the verdict.” Id. (distinguishing the standard established in Chapman v. California, 386 U.S. 18, 24 (1967), that applies on direct review). Assessing whether an error had such an effect is a serious and sometimes difficult fact-intensive undertaking, but reviewing courts are not without guidance. As the Supreme Court explained in Brecht, “[t]rial error . . . is amenable to harmless-error analysis because it ‘may . . . be quantitatively assessed in the context of other evidence presented in order to determine [the effect it had on the trial].’” Id. at 629 (all but first and second alterations in original) (emphasis added) (quoting Arizona v. Fulminante, 499 U.S. 279, 307–08 (1991)). Courts thus should examine harmlessness “in light of the record as a whole.” Id. at 638 (emphasis added). In doing so, judges “will often make numerous independent evaluations about the weight and sufficiency of the various items of evidence, the inferences to be drawn, and the different theories of the case.” Inthavong v. Lamarque, 420 F.3d 1055, 1060–61 (9th Cir. 2005). Two years after Brecht, the Supreme Court instructed that a judge evaluating harmlessness in a habeas case should ask, “Do I, the judge, think that the error substantially influenced the jury’s decision?” O’Neal v. McAninch, 513 U.S. 432, 436 (1995). If the answer to that question is “yes,” or if the judge “is in grave doubt as to the harmlessness of an error,” then there is “actual prejudice” under the Brecht standard and habeas relief is warranted. Id. at 436–37. If the answer is “no,” the error is harmless and relief must be denied. The question whether the exclusion of the Giarrusso evidence was harmless does not turn on how off base the state court’s evidentiary ruling was. Instead, we must engage in the arduous task of assessing the excluded evidence in the context of all the evidence of record, including the trial transcript and evidence proffered by both sides. Having done so, I disagree with the majority’s conclusion that the Giarrusso evidence was “powerful.” And I do not think that its exclusion “substantially influenced the jury’s decision,” nor am I in “grave doubt”

BRADFORD V. PARAMO 45

about whether it did so. Indeed, I am convinced that it did not. Thus, I would find the error harmless under Brecht.

II. Discussion

As the majority notes, the state’s case against Bradford was circumstantial. But this does not mean that the evidence of guilt was unconvincing. Cf. Brecht, 507 U.S. at 639 (finding error harmless where “the State’s evidence of guilt was, if not overwhelming, certainly weighty”). It is not unusual for direct evidence of a crime to be lacking, and many defendants are convicted based on circumstantial evidence alone. As the trial court instructed the jury here using a California pattern instruction: “Both direct and circumstantial evidence are acceptable types of evidence to prove or disprove the elements of the charge including intent and mental state and acts necessary to a conviction and neither is necessarily more reliable than the other. Neither is entitled to any greater weight than the other.” Continuing with another California pattern instruction, the judge cautioned the jury:

[B]efore you may rely on circumstantial evidence to find the defendant guilty you must be convinced that the only reasonable conclusion supported by the circumstantial evidence is that the defendant is guilty. If you can draw two or more reasonable conclusions from the circumstantial evidence and one of those reasonable conclusions points to innocence and another to guilt you must accept the one that points to innocence. . . . [W]hen considering circumstantial evidence you must accept only reasonable conclusions and reject any that are unreasonable.

By its verdict, the jury in this case obviously determined that the only reasonable conclusion from the circumstantial evidence was that Bradford was guilty. And I am convinced that had the Giarrusso evidence been introduced, the only reasonable conclusion would have remained the same. A. The Evidence Against Bradford 1. The Murder Scene Shortly before 3 a.m. on Thursday, August 30, 1979, Lynne Knight, a 28-year-old neonatal nurse, was brutally murdered in her Torrance, California backlot apartment. The trial evidence showed that the murderer first attempted to strangle Knight with a homemade garrote fashioned from two dowel-like pieces of wood (the handles) connected by metal picture-hanging wire. The assailant did not succeed in killing Knight with the garrote but sawed through her trachea in attempting to do so. The murderer then repeatedly stabbed Knight with a large knife, eventually severing the femoral artery in her left leg. At least one of the stab wounds was ten inches deep. Knight quickly bled to death. The police discovered the garrote beneath Knight’s body in a pool of blood; the knife was never found. Knight’s nude body was found on her bed. “Both [of] her legs were splayed with the knees pointing outward” and “[h]er right hand was positioned directly above her vaginal area.” A behavioral crime scene analyst explained that the assailant “may have posed her into this sexually demeaning position” because “[h]e wanted her humiliated even in death.” Knight also suffered several cuts to her left breast that the medical examiner explained were consistent with

BRADFORD V. PARAMO 47

intentional post-mortem mutilation by the assailant— “getting her where it hurts.” See Test. of Christopher Rogers. It was also the opinion of the crime scene analyst that the post-mortem breast wounds “further validate the level of resentment [the assailant] felt for [Knight].” Screams had awakened Knight’s landlord, Richard Rolleri, who lived with his family in the adjacent main house. A few minutes later, Rolleri looked out his front window and saw a man running away from the property, though he was unable to see the man’s face. Rolleri called the police at 2:53 a.m. The officers who responded to the scene relayed to dispatch the description that Rolleri provided of the fleeing man: “a white male adult, 25 years, 6 foot, medium build, wearing light brown coat, light brown pants, dark curly hair.” Although not disclosed to the jury, within half an hour the police detained eighteen-year-old, five-foot-eleven Gerardo Juarez, but they later ruled him out as a potential perpetrator of the crime.1

1 The police took Rolleri to the location where Juarez was detained, and Rolleri was unable to identify him from the front. But when the police turned Juarez around so that Rolleri could view him from the back, Rolleri said, “that looks like the guy.” Rolleri later retracted this identification, saying that he realized that Juarez was “too small” to be the man he saw at his house and that the only similarity was “the general look” of Juarez’s hair. Spots on Juarez’s clothing that officers initially thought were blood were determined not to be blood, and the police were unable to connect Juarez to Knight or the crime in any way other than Rolleri’s recanted identification. In his habeas petition, Bradford challenged—in addition to the exclusion of the Giarrusso evidence—the exclusion of evidence regarding Rolleri’s identification of Juarez. He also raised several other issues. The district court rejected the petition in full and granted a certificate of appealability only on the issue of the excluded Giarrusso 2. Bradford’s Statements and Demeanor Because of the manner in which Knight was killed and the path used by the murderer—an 18-inch-wide passageway leading to the back of Knight’s apartment that is not obvious from the street—it appeared that the murderer both knew Knight and was familiar with how to access her apartment without having to venture through the Rolleris’ backyard, where they kept their Siberian husky, Rocky. Bradford, an engineer who was 27 years old in 1979, had met Knight while skiing in January or February 1979 and dated her until June 3 of that year, when Knight ended the relationship. Bradford was one of the initial suspects in the murder based on statements that he and others provided to investigators. But the case went cold in the early 1980s, and Bradford was not charged with the murder until thirty years after Knight’s death. Knight was known to date several men at the same time, and detectives explained at trial that all of the boyfriends they contacted after the murder were entirely cooperative except one—Bradford. Bradford was willing to answer some questions, but he repeatedly cut interviews short when he did not like what he was being asked. During trial, the jurors heard tape recordings of investigators’ September 3 and September 6, 1979 interviews of Bradford, and the prosecution provided them with transcripts of those interviews as aids in their understanding of the recordings. The investigators did not tell Bradford until after the first interview that Knight had been murdered. Yet during that interview, Bradford

evidence. This appellate panel has not expanded the certificate to include the Juarez identification or any other issue raised in the petition.

BRADFORD V. PARAMO 49

sometimes referred to Knight in the past tense, (“she was a nurse”); (“I just kind of put the whole thing out of my mind. She was uh—you know, really uh—just uh something I wanted to put out of my mind entirely.”), and at one point he apparently referred to her as being dead: “I don’t have any reason to—to ever see her again, or want to see her again. You know, she’s just dead2 and something I want to put out of my mind. . . . I’m busy uh with dating other people um. She’s history.” On the very next page of the interview transcript, Bradford asked the investigators, “you still haven’t found her?” Bradford also equivocated during the interview about whether Knight was dating other men (or at least about whether he was aware of it) during their relationship. When asked, he said, “not to the best of my knowledge,” “I guess she might have gone out with some men,” “I don’t think she would . . . really dat[e] anybody else during the time that uh she was seeing me,” and “she may have been seeing other people.” And he described the man he saw at Knight’s apartment the day after their breakup as “just uh one of her boyfriends, I guess.” He also equivocated about whether the breakup was Knight’s idea or a mutual decision. He said, “we decided that we uh weren’t getting along very well and uh we decided that we should go our own ways” but then stated, “I think that Lynne was more decisive” about breaking off the relationship and he “was just kind of going along uh with it and . . . kind of saw that the relationship was

2 The parties vigorously debated at trial—and argued to the jury about— whether Bradford said “she’s just dead” or “it’s just dead” at this point in the interview. The jurors listened to the tapes and were told to determine for themselves what was said. The “she’s just dead” quotation is from the transcript provided to the jurors, which was not the “different font” transcript described by the majority, see Maj. Op. at 31–32. gonna come to an end. And so that was just a convenient stopping point for it.” On the latter point, several witnesses testified that Knight broke up with Bradford because he wanted to get married and was “getting too serious,” while she wanted to date other people.3 Bradford was inconsistent in the interviews about whether he ever considered marrying Knight. (“Q. [D]id you uh ask her to marry you at any time? A. No. No, I had no intention. She . . . certainly didn’t want to get married and I didn’t.”); (“[I]f things had worked out, we might have uh, you know. We talked about it a little bit, in the future, but there was never any real serious talk about it. . . . [W]e both talked. She said, yeah, well, some day, yeah. But uh, no, there was never any real serious talk about getting married.”). 3. Motive and Anger The state presented evidence contradicting Bradford’s description of his reaction to the breakup. Several witnesses testified at trial about an incident that occurred at Knight’s apartment on June 4, 1979—the day after Knight and Bradford broke up. Another suitor, Richard Frank, was visiting Knight at her apartment that day when Bradford showed up. Upon seeing Frank, Bradford reacted violently, tearing the screen off the door to Knight’s apartment, cursing at Knight, throwing a lamp at her, and calling her vile

3 For example, Jan Olsen testified that Knight told him that Bradford “wanted to be serious and . . . she did not[,] [s]he wanted to date other people,” and Knight’s sister testified that Knight told her that she broke up with Bradford because “Doug had become too serious, wanted to get married, and she didn’t.”

BRADFORD V. PARAMO 51

names.4 See, e.g., Test. of Richard Frank; Bradford interviews. One of Rolleri’s brothers, who was a teenager in 1979, testified that he “was outside and [his] mom told him to go inside” when she observed Bradford’s demeanor. And when Bradford departed the property, he kicked over trash cans on his way back to his Datsun 280Z, which he drove away recklessly, with “tires screeching.”5 Rolleri testified that he thought Bradford “was going to kill himself the way he drove away the car” and that Bradford “almost hit the pole” on the corner of a nearby intersection as he left. Bradford acknowledged this incident during his September 3 and September 6 interviews, admitting that he had “called [Knight] a god damn whore,” and “left[] in a big roar,” but backpedaling about being angry, (“[S]he was uh there with another guy. And uh I got mad at her because uh they uh—well I don’t know if I was really mad. I was just more upset about it.”). Discussing the June 4 incident, Bradford stated: “I didn’t make any threats. I, you know, I had too much to lose to . . . you know, to go—she was worried, I guess, that I’d come back and beat her up, or something. . . . I had no—no wish to do anything more with her, you know. I was thoroughly disgusted with her at that point. Just uh, you know, go away. I don’t want to hear from you again.” Bradford himself thus opined in his interview statements that Knight was afraid of him. This is consistent with Richard Frank’s trial testimony that Bradford

4 The majority opinion describes this as an incident that was relayed by Giarrusso—suggesting, perhaps, that he had a reason to shift suspicion toward Bradford. See Maj. Op. at 8–9. Although Giarrusso did describe this incident in an interview with investigators, he did not witness it and was only relaying what Knight told him. 5 See Test. of Richard Frank; Test. of Richard Rolleri; Test. of Carlos Vasquez; Test. of Maximino Vasquez. “was incredibly upset” during the June 4 incident, that “it was scary,” and that Knight “was scared” that day. And of course, Bradford’s conduct provided a rational basis for Knight’s fearing him. And although Bradford denied ever seeing Knight again or going back to her apartment after June 4, one of Rolleri’s brothers who witnessed Bradford’s stormy June 4 departure testified that he saw Bradford again in front of the property a week or two before the murder. See Test. of Carlos Vasquez. Bradford also saw him, and when they made eye contact Bradford walked “really fast . . . to his car” and “took off.” A detective who arrived on the scene within an hour of the murder testified that he observed a “wrinkled up” wedding invitation near a wastepaper basket in Knight’s apartment. It was an invitation to the wedding of Knight’s sister, Donna, which was scheduled for mid-September 1979 in Canada. Knight and Donna were very close, and Knight was to be Donna’s maid of honor. While Knight and Bradford were dating, Knight had told Donna that Bradford would be attending the wedding with her, but after they broke up Knight planned to attend alone. Donna testified that Knight never would have crumpled up the invitation and thrown it away. In light of the cancelled plans for Bradford to attend the wedding with Knight, the state’s theory was that Bradford had crumpled it up in anger on the night of the murder. Additionally, Donna had given Knight a gold medallion from Greece as a gift, and Knight habitually wore it on a gold chain as a necklace. The medallion and a broken clasp to a necklace were found at the crime scene, but the chain itself was not. See Maj. Op. at 5. According to one of Knight’s

BRADFORD V. PARAMO 53

friends, Bradford had bought Knight a necklace, see id. at 8, and during a police interview Bradford acknowledged having done so, though he was unable to describe it. The prosecution’s theory was that Bradford had given Knight the chain she wore with the medallion and that he took it back on the night of the murder. A crime scene analyst opined after assessing descriptions and photos of the horrific crime scene and other evidence “that Knight’s murder was ‘motivated by a preexisting interpersonal conflict.’” Maj. Op. at 29. The missing gold chain fit into that theory because, said the analyst, removal of it (but not the medallion) from the murder scene suggested that the chain had “significance to the offender” and that the offender “may have given [it] to the victim.” 4. Consciousness of Guilt Perhaps the strongest evidence of consciousness of guilt is Bradford’s far-fetched alibi. The majority opinion states that the story Bradford told investigators about his whereabouts on the evening of Knight’s murder “seemed unlikely” and was “fairly weak.” Maj. Op. at 29–30. But this is an understatement because that alibi was so absurd that it severely undermined Bradford’s credibility. The alibi—that Bradford went sailing alone in Long Beach on a 30-foot, 4600-pound sailboat at 10:00 p.m. at low tide without any wind, returning home to Costa Mesa, where he lived with his parents and several of his siblings, at 3:00 a.m., only minutes after the murder and five hours before he was due to be at work—was proved at trial to be absolutely unworthy of credence, even by Bradford’s own expert witnesses. At trial, one of those experts ultimately explained that Bradford’s sailing story was completely incredible and that no sensible sailor would reserve a sailboat for 10:00 p.m. in the first place or, if so, proceed without wind and with an incoming tide. See Garth Hobson Test.6 Bradford first provided his sailing alibi on September 6, 1979, during his second police interview, with his parents present. When asked whether anyone saw him sailing that night, Bradford unequivocally responded, “Not a soul,” and stated that he didn’t think anybody was awake when he got home “a little after 3.” But Bradford’s mother was indeed awake, and she stated that “it must have been about 1:30” when Bradford got home. When Bradford—seemingly uncomfortable with that answer—asked his mother, “It was later than that, wasn’t it?” Mrs. Bradford responded, “I don’t know, Doug.” There is nothing in the interview transcripts to suggest that the detectives told Bradford exactly when Knight was killed, yet Bradford seemed intent on establishing that he was at his parents’ Costa Mesa house at 3:00 a.m. and could not have been in Torrance at the time of the murder. A detective called Bradford three days after the second interview and asked Bradford to meet him in Long Beach to show him the boat and sailing club’s reservation calendar,

6 Even assuming that it is possible for someone to sail a Shields boat— the type of sailboat Bradford allegedly sailed—alone at night, it was not plausible that Bradford happened to do so on the night of the murder. Bradford’s expert explained that it is well-known that there is typically very little wind in Alamitos Bay at nighttime in the summer, and thus it made no sense to him that a sailor would plan to go out at 10:00 at night. And the expert was unable to think of any reason other than a competition or an emergency that a qualified Shields skipper—which Bradford was presumed to be—would choose to take the boat out alone. The expert’s position is summed up by his agreement on cross- examination that if someone told him, “Last night I decided to take my Shields boat out at night by myself,” he would “yell B.S.” on that “pretty quick.”

BRADFORD V. PARAMO 55

which was at a separate location. Bradford told the detective that he did not want to do that and that he did not like all the questions he was being asked. When the detective was nevertheless able to obtain the sign-out calendar the next day, the calendar did include notations of “Doug 10–12” on August 29 and “Doug 8–12” on Monday, August 27. Neither entry indicated whether it was for morning or evening, and as explained at trial by both the detective and the man who kept the calendar on his unlocked back porch in August 1979, Bradford would have had enough time to write in these “reservations” after the detective inquired about them on September 9 and before he obtained them on September 10. See Test. of Detective Hilton; see also Test. of Arthur Harrison that “[i]t would be easy” to backdate the calendar because anyone who knew where the calendar was kept could access it at any time. Bradford did consent to a search of his car about a week after the murder, and a detective who was present during the September 6, 1979 search testified at trial that although no blood was found in the car, it “reeked of Armor All” and appeared to have been freshly cleaned. The majority notes that no garrote was found in Bradford’s car or house. See Maj. Op. at 36. But as explained earlier, the garrote was not missing; it was found under Knight’s body. The majority also discounts the trial testimony about the smell and appearance of the car, characterizing these as “observation[s that were] made decades after the fact rather than when the search was conducted.” Id. at 37. But the detective testified that he was present when the car was searched, and thus he observed the odor and appearance of the car using his senses of smell and sight in 1979, not thirty-five years later. And when the police searched Bradford’s house and his mother’s house—where Bradford lived at the time of the murder— they were not looking for a garrote. They instead were searching for materials that could have been used to make the garrote found at the murder scene. At the mother’s house, the police did find wire consistent with that used to construct the garrote. Another detective recounted Bradford’s refusal to voluntarily provide a DNA mouth swab in 2007. When the police then presented Bradford with a search warrant compelling compliance, Bradford tried to pull off an “air swab” by moving the swab around the inside of his mouth without making any contact. Although investigators did ultimately obtain Bradford’s DNA, his attempt to avoid providing a true swab is yet more evidence of consciousness of guilt. And despite having attested in his September 3, 1979 interview that Knight was just “something [he] wanted to put out of [his] mind entirely,” a 2009 search of Bradford’s residence turned up an envelope containing a 5x7 enlargement of a photograph of Knight standing in snow wearing skiing attire. The order form for the enlargement was with the photograph, and the photograph had a date stamp on the back indicating that it was printed in July 1979—a month after the breakup and a month before the murder. And although the majority claims that “there is nothing incriminating about keeping a photograph of an ex- girlfriend in an envelope in a desk drawer,” Maj. Op. at 36, Bradford’s purchase of the enlarged photograph after the breakup was not consistent with his statement that he “wanted to put [Knight] out of [his] mind entirely” and it further undermined his credibility. Still more of Bradford’s untruthfulness was made known to the jury through the testimony of Jerilyn Seacat, a nurse

BRADFORD V. PARAMO 57

Bradford dated from 1996 until 2009. Seacat ended her relationship with Bradford after a detective revealed to her that Bradford was suspected of Knight’s murder and had married another woman several years earlier. Seacat recounted that Bradford told her early in their relationship that “in the late seventies” he almost married “a neonatal nurse” who “was the love of his life” but that she had “died of a terminal illness.” And although Seacat met Bradford’s mother early in their relationship, Bradford later told Seacat that his mother had died, though she had not. Seacat also testified that after she broke up with Bradford, he repeatedly showed up at her house—even inside her house— unannounced, attempting to reconcile. In an effort to do so, he assured Seacat that “he had hired O.J. [Simpson]’s murder team” and “this case was going to go away.”7 The California Court of Appeal determined that Bradford’s “conduct in stalking Seacat” was not admissible to show “propensity to kill Knight” but was “strong evidence of how he fixates on a woman after rejection and thus provides compelling evidence of his intent, common scheme or plan, and thus identity.” The court also explained that “the fact that [Bradford] lied to Seacat about the mutuality of affection between himself and Knight is what gives rise to the inference of a consciousness of guilt.” B. The Excluded Giarrusso Evidence and Harmlessness Giarrusso and Knight had lived together a few years before Knight’s murder and remained very close friends after ending their romantic relationship. Knight invited

7 The first part of this statement was partially true in that Robert Shapiro represented Bradford at trial. Giarrusso over to her apartment for dinner on the evening of August 29, 1979. Knight had told Rolleri’s wife, Herlinda, about the dinner plans, and the Rolleris saw Giarrusso there with Knight that evening. When the police questioned Giarrusso the next morning, he told police that he left Knight’s apartment shortly before midnight and then spent the night at the home of his then-girlfriend, Marcia Rubin— arriving there during The Tonight Show with Johnny Carson. Rubin corroborated this alibi both in 1979 and in 2008; in the interim, Rubin and Giarrusso married in 1981 but divorced in 1986, and Giarrusso died in 1994. In 2003— twenty-four years after the murder—Rolleri for the first time reported having seen what “looked to him like” Giarrusso’s VW Beetle at a nearby stoplight in front of an arriving police car just after Rolleri called the police on the night of the murder.8 The defense was precluded from presenting this and other evidence regarding Giarrusso to the jury. 1. Physical Characteristics In finding the error here harmful, the majority opinion places much stock in two physical characteristics of Giarrusso and Bradford, contending that Giarrusso’s height and hair matched Rolleri’s descriptions of the fleeing suspect more closely than Bradford’s did. But in view of the record as a whole, I cannot agree. On the date of the murder Giarrusso was 5' 9" tall and weighed 164 pounds, and Herlinda—who observed him dining with Knight on the night of the murder—described him as “on the heavy side with dark hair that was balding in

8 The trial court was troubled by Rolleri’s 2003 report regarding the car, due not only to the 24-year delay in mentioning it but also to the fact that Giarrusso’s car was “an old blue beat up Volkswagen,” “a pretty common car, especially back in the late seventies.”

BRADFORD V. PARAMO 59

front.” See Maj. Op. at 7 & n.2. And “[a] photograph of Giarrusso at that time depicted him with dark curly hair.” Maj. Op. at 24. Indeed, as Bradford states in his opening brief, that “[p]hotograph[] shows[] [Giarrusso] with dark, curly hair down to his collar.” See Opening Brief at 8. Bradford described himself in September 1979 as “about 6 feet” tall and weighing “about 160.” Other descriptions of him in the record are similar regarding his height and weight. See Maj. Op. at 38. And Bradford “had a full head of brown hair at the time of the murder,” Maj. Op. at 7 n.2, and a trial witness described his hair as “sandy,” id. at 38. The majority maintains that “Giarrusso fit the height and hair color of the suspect described by Rolleri more than Bradford did,” Maj. Op. at 38, and asserts that “dark, curly hair . . . would exclude Bradford but could inculpate Giarrusso in combination with other evidence,” Maj. Op. at 39. But in reaching its conclusion about whose height matched better, the majority—while eventually acknowledging that Rolleri gave several descriptions of the suspect—relies solely on an outlier description that appears in Juarez’s police report. The record as a whole, however, does not support the conclusion that Rolleri’s hair descriptions “inculpate Giarrusso” and “exclude Bradford.” Within minutes of the murder, Rolleri provided a description of the fleeing suspect to the responding officers at the murder scene that the officers then relayed to dispatch. That dispatch recording was transcribed: “a white male adult, 25 years, 6 foot, medium build, . . . dark curly hair.”9 Detective Hilton, who arrived at the scene about an hour

9 I have omitted here the clothing description, which is not at issue in this discussion. after the murder, documented in his report that Rolleri told investigators that the fleeing man “was approx 6-0” and “had dark curly hair, the hair was not long and . . . the back of the hair was trimmed to the neck above the collar.” A few hours later, detectives interviewed Rolleri and at that time he described the fleeing man as having “black, curly hair,” being “bigger than” Rolleri, who is 5' 9", and looking “strong” and “built.” Two days later, on September 1, 1979, Rolleri again spoke to the police, who documented his description of the suspect as “approx 5-11 to 6-0, possibly black curly hair, trimmed close to head, not bushy.” After the case was closed and reopened decades later, the police again interviewed Rolleri in October 2003. On that occasion, Rolleri described the suspect as “tall, 6 feet or more, with an athletic type build (not fat or skinny) and bushy or curly hair.” He also remembered that Juarez—who is listed in his booking report as 5' 11" tall—was “too small to be the person he saw running away” and stated that the only similarity between Juarez and the fleeing man was “perhaps the general look of his hair.” And in another interview in 2010, “Rolleri agreed that it was dark and he only saw the suspect from behind while he was running.” “Rolleri also agreed that there was not much more he could say about the suspect aside from the fact that he had curly hair . . . and was approximately 6 feet tall.” Thus, in interview after interview over the years, Rolleri repeatedly described the suspect as taller than 5' 9" and as six feet tall. Yet the majority chooses to focus on the lone 5' 9" description in Juarez’s booking report. The majority explains that it does so “because that is the description the police elected to document.” Maj. Op. at 38. But as just recounted, the investigators repeatedly documented their

BRADFORD V. PARAMO 61

interviews of Rolleri, and again and again he described the fleeing suspect as six feet tall or as “bigger than” 5' 9". I do not dispute that the 5' 9" description appears in Juarez’s booking report, but it is not clear how it made it into that report, which was typed up approximately seven hours after the murder.10 And although Rolleri initially identified Juarez at a field showup, Rolleri later realized that Juarez— listed in the booking report as 5' 11" tall—was “too small” to be the man he saw fleeing. This further undercuts the notion that Rolleri ever stated that the fleeing man was 5' 9". Moreover, even without the Giarrusso or Juarez evidence having been admitted, defense counsel could have asked Rolleri at trial how tall the fleeing man appeared to be or how tall Rolleri had reported him to be.11 But defense

10 The majority also claims that the 5' 9" description was the one that police dispatch distributed to officers out in the field immediately after the murder so that they could look for the murderer. See Maj. Op. at 5 (“With this description, officers surveyed the area and discovered Gerardo Juarez . . . .”); Maj. Op. at 37 (“Police immediately used this description to search the surrounding area for the suspect shortly after 3 a.m.”). Although the way the booking report is written may suggest this, it cannot be squared with the fact that the officers at the scene relayed to dispatch a “six foot” description. Why, and how, then, would a 5' 9" description have been broadcast? That would have required that sometime between 3:00 a.m.—when Rolleri provided the “six foot” description—and 3:20 a.m.—when Juarez was detained—Rolleri changed his description to 5' 9" and officers at the scene transmitted that new description to dispatch. There is no record evidence of that occurring. 11 Just before Rolleri testified, defense counsel asked the trial judge outside the jury’s presence “whether or not [he] c[ould] ask Mr. Rolleri the description he gave the police of” the fleeing suspect. The prosecutor agreed that “any description that [Rolleri] gave to the police that doesn’t involve Mr. Juarez’[s] showup they have a right to present that.” counsel—doubtless aware of the multiple recorded instances of Rolleri describing the fleeing suspect as six feet tall and as taller than 5' 9"—opted not to pursue that line of questioning and instead focused on the suspect’s hair.12 The majority asserts that “[b]y referencing other descriptions,” I “stray[] from our task” and that if the prosecution had challenged the 5' 9" description on the bases I identify, “the jury would have been called upon to weigh this conflicting evidence.” Maj. Op. at 38–39. While I am certainly keenly focused on the task at hand—applying the Brecht harmless error analysis—I agree that “the jury would have been called upon to weigh this conflicting evidence.” But in my view, upon doing so the jury would have had little trouble rejecting a purported 5' 9" description—the source of which is unknown—in the face of the overwhelming evidence to the contrary. I reach the same conclusion regarding Rolleri’s descriptions of the killer’s hair. Although Bradford was described as having “sandy” hair, witnesses—including Rolleri—also described him as having “bushy/curly” hair, “curly hair” and “a lot of hair”—attributes that did match Rolleri’s descriptions of the suspect and arguably were easier for Rolleri to discern in the dark as he observed the fleeing man from a distance of 65–165 feet while the suspect was between two distant street lights. See report describing October 9, 2003 interview of Rolleri; Trial Test. of Richard Frank; report describing December 7, 2010 interview of

12 And during a sidebar later in the trial, Bradford’s lawyer did not claim that Rolleri ever described the fleeing man as 5' 9". Rather, counsel stated: “Rolleri gives three different descriptions. One is the person is six-feet tall. One is the person is taller than I am, which is five-foot-nine. And one is the person is five-foot-eleven to six-feet tall.”

BRADFORD V. PARAMO 63

Rolleri. Furthermore, defense counsel cross-examined Rolleri at trial about the fleeing man’s hair. Rolleri did not recall ever stating “that the person clearly had dark hair.” When questioned about his September 1, 1979 description, Rolleri acknowledged saying “black curly hair” at that time but stated that it was “very hard for [him] to really see.” He “tried” to give an accurate description of the hair “but it was too dark. . . . It was too dark for [him] to see exactly whether [the hair] was light or dark.” Thus, the issue of “sandiness” was presented to and considered by the jury. Additionally, although not revealed to the jury, Rolleri pointed out to police in his 2003 and 2010 interviews that his general description of the suspect matched Bradford, with whom Rolleri was familiar because he frequently saw Bradford at Knight’s apartment while the two were dating. The hair descriptions thus do not “exclude” Bradford. On the other hand, Rolleri’s “not long” and “trimmed to the neck above the collar” descriptions would exclude Giarrusso, whose hair was long and down to his collar rather than trimmed above it. Additionally, Rolleri described the fleeing man as “strong,” “built,” “athletic,” and “fast.” These adjectives fit Bradford13 much better than Giarrusso, who was described by Rubin as being “out of shape,” “not athletic,” “not a muscular person,” and having a “bit of a belly (perhaps as much as a 36 inch waist),” and by Herlinda as “on the heavy side.” In sum, I do not agree with the majority that the physical characteristics of Bradford and Giarrusso, when compared

13 For example, Bradford explained during the September 3, 1979 interview that his hobbies included snow skiing, bicycling, tennis, sailing, and scuba diving. to Rolleri’s descriptions, support a finding of harmful error under Brecht. 2. Other Evidence The majority opinion suggests that Giarrusso had a motive to kill Knight. See Maj. Op. at 28 (“As we recently recognized, ‘[t]he excluded evidence could have provided the missing link to establish reasonable doubt for the jury: an actual individual who had knowledge, motive, and opportunity.’” (alteration in original) (quoting United States v. Espinoza, 880 F.3d 506, 519 (9th Cir. 2018))).14 But the majority’s suggestion that Giarrusso had the same motive as Bradford because “[h]e too had been relegated to the ‘friend zone,’” id. at 39, is not supported by my reading of the record. First of all, multiple trial witnesses testified, and Bradford acknowledged in an interview, that Knight was the one who ended their relationship. And yet the majority apparently assumes that Knight broke off the relationship with Giarrusso rather than the other way around.15 I find no evidence supporting such an assumption, and there is a 2010 police report to the contrary. That report recounts a statement by Giarrusso’s friend and coworker that Knight “wanted a firmer commitment” but Giarrusso “was not

14 Espinoza was a methamphetamine-importation case before the Ninth Circuit on direct appeal rather than on habeas review. The Espinoza court rejected the government’s harmless error argument there because the defense theory that was excluded was “certainly plausible” under a much lower standard than Brecht’s “actual prejudice” standard. Espinoza, 880 F.3d at 519 (quoting United States v. Liera, 585 F.3d 1237, 1244 (9th Cir. 2009)). 15 I also disagree with the majority’s suggestion that “[r]egardless of who ended each relationship, the result was the same.” Maj. Op. at 40.

BRADFORD V. PARAMO 65

interested in marriage, and [Knight] eventually moved out.” See Summary of Witness Statement of Michael Nihill. There simply is nothing in the record to suggest any reason for Giarrusso to want to viciously murder Knight and pose her in a sexually demeaning position. Moreover, it is beyond dispute that Knight and Giarrusso had—like Knight and some of her other ex-boyfriends— remained close friends for years after ending their romantic relationship, but there is no evidence that Knight and Bradford shared any such friendship after their June 3 breakup and Bradford’s June 4 rampage. And although Herlinda did state in a 2003 interview that Knight told her Giarrusso had once hit her—apparently years earlier when they were dating—and brushed it off as “a misunderstanding,” Giarrusso was described by witnesses as a dear friend of Knight who was “very distraught” over her death, and he was completely cooperative with investigators. The same cannot be said of Bradford. Additionally, much of the evidence regarding Giarrusso that the majority finds supportive of Bradford’s petition is refuted by innocent explanations in the record. For example, Giarrusso’s “knowledge of what a garrote was,” Maj. Op. at 13, came from watching James Bond movies, and Giarrusso explained the reason for the “bandage on his thumb,” Maj. Op. at 13, to interviewing detectives hours after the murder—while rolling up his shirtsleeves and holding up his hands for the detectives’ inspection—as a days-old cut from a test tube in the lab where he worked. Moreover, Giarrusso’s girlfriend described him as “not ‘handy’ around the house or with tools,” and “not mechanically inclined as a builder,” and he “possessed no major tools (such as a hand drill)” like what was used to construct the garrote. Bradford, on the other hand, was an engineer who, according to his mother, “had an aptitude . . . for making and building things” in 1979. And it is highly improbable that the jury would have disbelieved Giarrusso’s alibi, which was corroborated by a woman who eventually married him—not a likely course of action for someone who thinks her beau murdered an ex-girlfriend in cold blood. She repeated the same account after she and Giarrusso divorced. 3. Prosecution’s Opening Statement During his opening statement, the prosecutor stated, “Now, what’s very clear is that for several nights leading up to the murder there was one night when [Knight] was home by herself. One night only. And that was the night that she was murdered.” The majority contends that “[t]his statement would have been directly contradicted if evidence regarding Giarrusso had not been excluded,” Maj. Op. at 11, and that the error in excluding the Giarrusso evidence “was ‘compounded and magnified’” when the trial court allowed the prosecution to make this statement, Maj. Op. at 29 (quoting Lunbery, 605 F.3d at 762). But in making the statement that Knight was alone that one night, the prosecutor was arguing that the killer took advantage of the only night she was at home by herself without an overnight guest. And as the trial court found, the prosecutor’s statement was not entirely inaccurate because Knight was alone “during the relevant time period here.” The fact that Giarrusso had dinner with her a few hours before the murder does not alter this fact. When Knight was killed, there was no one at the apartment but Knight and the murderer. Thus, I disagree that the prosecutor’s statement “compounded” error here.

BRADFORD V. PARAMO 67

4. Application of the Brecht Harmlessness Standard The majority relies in part on Holmes v. South Carolina, 547 U.S. 319 (2006), to support its conclusion that the error here was not harmless. But Holmes was not a habeas case, and the Court did not analyze the harmlessness of the error there under the Brecht standard.16 Moreover, in Holmes, the petitioner proffered third-party culpability evidence “that, if believed, squarely proved that [another man], not petitioner, was the perpetrator.” 547 U.S. at 330. The same is true of the other case discussed by the majority, Lunbery v. Hornbeak, 605 F.3d 754 (9th Cir. 2010). In contrast, the Giarrusso evidence does not “squarely prove” anything. This is not to say that Holmes and Lunbery “requir[e] Bradford to submit evidence that would ‘squarely prove’ Giarrusso’s culpability.” Maj. Op. at 40. Indeed, the Supreme Court cautions that judges should not “try to put the [harmlessness] question in terms of proof burdens,” O’Neal, 513 U.S. at 436–37, and instead must assess it “without benefit of such aids as presumptions or allocated burdens of proof that expedite fact-finding at . . . trial,” id. at 437 (quoting R. Traynor, The Riddle of Harmless Error 26 (1970)). But the third-party culpability evidence excluded in Holmes and Lunbery was far stronger than the evidence excluded in this case in both caliber and effect. So was the evidence in this court’s two other published post-Brecht opinions awarding § 2254 relief in this context. See Cudjo v. Ayers, 698 F.3d 752 (9th Cir. 2012) (third party’s confession); Gable v. Williams, 49 F.4th 1315 (9th Cir.

16 I concede that because of the strength of the third-party culpability evidence in Holmes, the Court likely would have found the error harmful even if it had applied Brecht. 2022) (same).17 In contrast to the facts in these cases, the Giarrusso evidence is weak, and I disagree with the majority’s reference to it as “powerful,” see, e.g., Maj. Op. at 4, 29 & 34.

III. Conclusion

In sum, the evidence against Bradford is stronger—in my view, much stronger—than the majority opinion suggests, especially considering the overwhelming evidence of Bradford’s consciousness of guilt. And although I agree that Bradford was entitled to present evidence regarding Joe Giarrusso, exclusion of that evidence did not have “substantial and injurious effect or influence in determining the jury’s verdict.” Brecht, 507 U.S. at 637 (quoting Kotteakos, 328 U.S. at 776). The prosecution presented extensive evidence of Bradford’s guilt, including but not limited to his lack of cooperation, his motive, his hair-trigger violent temper, his untruthfulness, and his suspicious presence in front of Knight’s apartment a week or two before the murder. Bradford’s alibi was completely incredible, and, like other evidence, its utter implausibility demonstrates Bradford’s consciousness of guilt and calls into question exculpatory statements he gave investigators. See, e.g., United States v. Holbert, 578 F.2d 128, 129–30 (5th Cir. 1978) (“[F]alse exculpatory statements may be used . . . as substantive evidence tending to prove guilt. When a defendant voluntarily and intentionally offers an explanation and this

17 The court also decided Alcala v. Woodford, 334 F.3d 862 (9th Cir. 2003), which affirmed the district court’s conditional grant of a § 2254 petition based on cumulative error that included ineffective assistance of counsel, exclusion of third-party culpability evidence, and other evidentiary rulings.

BRADFORD V. PARAMO 69

explanation is shown to be false, the jury may consider whether the circumstantial evidence points to a consciousness of guilt . . . .” (citations omitted) (citing, among other cases, United States v. Pistante, 453 F.2d 412, 413 (9th Cir. 1971))); see also United States v. Obi, 239 F.3d 662, 665 (1st Cir. 2001). Most, if not all, of the reasons that the majority discounts the various pieces of evidence against Bradford and credits his alibi were presented to the jury. Nevertheless, the jury disbelieved Bradford’s alibi and was persuaded by the state’s case against him. Even if Bradford had presented evidence regarding Giarrusso’s presence at Knight’s apartment on the evening of the murder, that evidence would have been countered by the prosecution. Giarrusso had an alibi, was completely cooperative with authorities, was a dear friend of Knight’s, and lacked any motive to kill her. I am convinced that even if the jury had been told about Giarrusso, it still would have reached the only reasonable conclusion supported by the evidence—that Bradford was guilty of Knight’s murder. “The writ of habeas corpus is an ‘extraordinary remedy’ that guards only against ‘extreme malfunctions in the state criminal justice systems.’” Shinn v. Ramirez, 142 S. Ct. 1718, 1731 (2022) (quoting Harrington v. Richter, 562 U.S. 86, 102 (2011)). And “[t]he Brecht standard reflects the view that a ‘State is not to be put to th[e] arduous task [of retrying a defendant] based on mere speculation that the defendant was prejudiced by trial error; the court must find that the defendant was actually prejudiced by the error.’” Davis v. Ayala, 576 U.S. 257, 268 (2015) (second and third alterations in original) (quoting Calderon v. Coleman, 525 U.S. 141, 146 (1998)). As stated above, even “a ‘reasonable possibility’ that [the] trial error contributed to the verdict” is not enough for habeas relief. Brecht, 507 U.S. at 637. With the Brecht standard in mind, my esteemed colleagues and I have reviewed the extensive record in this case, each asking, “Do I think the error substantially influenced the jury’s decision?” Having done so, we have come to different conclusions. I do not believe Bradford is entitled to the extraordinary habeas corpus relief he seeks because the exclusion of the Giarrusso evidence did not result in actual prejudice under Brecht. I therefore respectfully dissent and would affirm the district court’s denial of Bradford’s habeas petition.