MEMORANDUM **

Burt was convicted in state court of second-degree murder, after a trial at which he claimed self-defense. He claims that his trial counsel was ineffective for failing to call two witnesses to testify in his defense. The district court denied his ha-beas petition, and we affirm.

1. We review under the Antiterrorism and Effective Death Penalty Act (“AED-PA”) the state court’s determination that Burt failed to make a prima facie showing of prejudice on his ineffectiveness claim. Burt is correct that, because the state court made no findings of fact, we need not defer under 28 U.S.C. § 2254(d)(2) or (e)(1) of AEDPA. But § 2254(d)(1) applies to purely legal rulings, and the state court’s determination that prejudice was not made out even if one assumes the *24truth of Burt’s factual submissions is such a ruling. That some additional, but not substantially different, evidence was revealed during the federal evidentiary hearing does not impact our standard of review where, as here, the relevant evidence was presented to the state court and presumed true. Cf. Killian v. Poole, 282 F.3d 1204, 1208 (9th Cir. 2002) (“AEDPA deference does not apply to Killian’s perjury claim in this case because the state courts could not have made a proper determination on the merits. Evidence of the perjury, after all, was adduced only at the hearing before the magistrate judge.”).

2. We consider the merits in a slightly different posture than that of the district court, but reach the same result. Because the relevant state court decision determined that Burt failed to state a prima facie claim of prejudice, and because Burt’s federal evidentiary hearing revealed no significant new evidence that was not before the state court as to the prejudice issue, we must first determine whether “it was objectively unreasonable for the state court to conclude on the record before it that no reasonable factfinder could believe that [Burt] had been prejudiced.” Nunes v. Mueller, 350 F.3d 1045, 1055 (9th Cir. 2003).

The question is a close one. Edward Lee would have testified that he saw the victim with a black gun in his waistband shortly before the shooting. This testimony would have corroborated, to some degree, Burt’s otherwise entirely uncorroborated testimony that the victim pulled a black gun out of his waistband immediately before Burt shot him. Moreover, the state’s case against Burt was quite weak. Only two eyewitnesses testified to witnessing the shooting, and one had made inconsistent statements to the police. More significantly, both testified that Burt also committed attempted robbery, but the jury acquitted Burt on the attempted robbery count. So the jury plainly did not credit their entire testimony. One additional witness heard a verbal altercation before the shooting, and one witness had made a pri- or statement that she so heard, but testified at trial that she heard nothing before the shooting.

But Lee’s testimony would have bolstered Burt’s self-defense testimony only slightly. Although the testimony provides support for the notion that the victim was armed, Burt also was armed, and Lee’s testimony has no bearing on who initiated the conflict. That was the key question, given Burt’s self-defense defense at trial.

This case is thus unlike Riley v. Payne, 352 F.3d 1313 (9th Cir. 2003), in which Riley claimed that a joking conversation with the victim turned sour and Riley shot him in self-defense. Id. at 1315-16. In Riley, the uncalled witness would have corroborated a critical element of Riley’s self-defense claim: that the victim (or his friend) threatened Riley. Id. at 1317-18. Here, in contrast, Lee’s testimony does not speak to the crucial question of whether the victim threatened or pulled a gun on Burt. Similarly, the corroboration provided by the uncalled witnesses’ testimony in Luna v. Cambra, 306 F.3d 954 (9th Cir. 2002), amended 311 F.3d 928, and Brown v. Myers, 137 F.3d 1154 (9th Cir. 1998), is significantly stronger than that which Lee’s testimony would have provided.

We thus hold that the state court’s decision that Burt failed to state facts which, if proved, would support a claim of prejudice was not an “unreasonable application of[] clearly established Federal law.” § 2254(d)(1). Because we conclude that the state court’s prejudice determination was not unreasonable, we need not decide whether trial counsel’s performance was deficient.

*253. Burt moved to expand the certifícate of appealability to encompass two claims of instructional error. Burt points to the trial court’s failure to properly instruct the jury on voluntary manslaughter and to instruct the jury at all on involuntary manslaughter. But “the failure of a state trial court to instruct on lesser included offenses in a non-capital case does not present a federal constitutional question.” Windham v. Merkle, 163 F.3d 1092, 1106 (9th Cir. 1998). And while Burt does have a constitutional right to instructions on his theory of the defense, see Clark v. Brown, 450 F.3d 898, 904-05 (9th Cir. 2006), the jury was thoroughly instructed on both perfect and imperfect self-defense. He has thus failed to make “a substantial showing of the denial of a constitutional right.” Slack v. McDaniel, 529 U.S. 473, 483, 120 S.Ct. 1595, 146 L.Ed.2d 542 (2000) (quotation marks omitted).

The district court is AFFIRMED. The motion to expand the COA is DENIED.

dissenting:

I cannot agree with the majority that the failure on the part of Burt’s trial counsel to introduce testimony from Edward Lee was nonprejudicial. In my view, the majority’s assertion that “Lee’s testimony would have bolstered Burt’s self-defense testimony only slightly ” misreads the record in this case and misperceives the nature of circumstantial evidence. Maj. Op. at 24 (emphasis added). As the only person who would have provided support for Burt’s testimony that Timothy Orange-Jackson, the decedent, had a gun at the time of the shooting, Edward Lee was hardly an inconsequential witness and the failure to call him was surely prejudicial.

Burt’s sole defense was that he shot Orange-Jackson to protect himself after he saw Orange-Jackson reaching for a gun in his waistband. However, none of the trial witnesses, other than Burt himself, “made ... reference to Mr. Orange-Jackson holding a firearm.” In re Burt, No. B155787, slip op. at 3 (Cal.Ct.App. Dec. 5, 2002). Nor was any gun found at the scene.1 This gap in Burt’s defense was significant, and the prosecutor exploited it in closing argument, saying, “No one saw another gun. Why? Because the defendant is making that up.”

Had Edward Lee been called as a trial witness, the prosecutor’s assertion that “no one saw [a] gun” in Orange-Jackson’s possession would have been directly contradicted. As Lee’s deposition makes clear, he saw a black gun in Orange-Jackson’s waistband mere minutes before Orange-Jackson walked up to the apartment where the shooting took place. This corroborates not only Burt’s testimony that Orange-Jackson had a gun, but also his description of that gun and its location on Orange-Jackson’s person. To be sure, Lee made prior statements to the police that would have subjected him to impeachment regarding other aspects of the case had he been called as a witness, but this was true of virtually all of the witnesses in this prosecution: The government’s two eye witnesses had both made directly inconsistent prior statements and were also demonstrably biased cohorts of the decedent. These witnesses’ weaknesses explain why the jury disbelieved significant portions of their testimony, acquitting Burt of armed *26robbery. In light of the thinness of the prosecution’s case against Burt, Lee’s testimony could have made a crucial difference. Cf. Strickland v. Washington, 466 U.S. 668, 696, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984) (“[A] verdict or conclusion only weakly supported by the record is more likely to have been affected by errors than one 'with overwhelming record support.”).

The majority asserts that the failure to call Lee as a witness was not prejudicial because his potential testimony that he saw Orange-Jackson with a gun moments before the shooting “does not speak to the crucial question of whether [Orange-Jackson] threatened or pulled a gun on Burt.” Maj. Op. at 24. This is a perplexing argument on which to rest the outcome in this case. How can the question of whether Orange-Jackson possessed a gun “not speak to” whether he “pulled a gun”? Clearly evidence suggesting that Orange-Jackson did not have a gun — as implied by the government’s case in chief and argued by the prosecutor in closing argument— “speaks to” the question of whether he “pulled a gun”: If he didn’t have a gun to pull, then he could not have pulled one. How then can directly contradictory evidence that Orange-Jackson did have a gun “not speak to” the “crucial question” underlying Burt’s theory of self defense?

The majority incorrectly analyzes the prejudice issue in this case because it significantly mispereeives the fundamental nature of circumstantial evidence, of which Lee’s testimony is a particularly strong example. Circumstantial evidence is just as valid and useful as direct evidence,2 and the fact that it does not directly establish the precise point for which it is offered does not detract from its potency. To borrow from a seminal treatise, the majority’s argument that Lee’s testimony is nonprejudicial because it does not directly establish that Orange-Jackson pulled a gun on Burt improperly imposes “a standard of conclusiveness that very few single items of circumstantial evidence ever could meet.” McCormick on Evidence § 185 (Kenneth S. Broun, et al., eds., 6th ed., 2006). Under the majority’s approach, a wide swath of defective performance on the part of trial attorneys would be deemed nonprejudicial simply because it relates to evidence that would not directly and definitively establish a particular defense.

The majority’s opinion asks too much of a piece of evidence before the failure to present that evidence can be deemed prejudicial; by setting the bar so high the majority fails to appreciate just how significant Lee’s potential testimony would have been. For one thing, Lee’s testimony would have directly contradicted the government’s contention that Orange-Jackson was unarmed, and would have impeached the already weak government witnesses who suggested as much. Further, Orange-Jackson’s simple possession of a gun is significant evidence in its own right because it corroborates the direct evidence in Burt’s own testimony that Orange-Jackson actually did “threaten! ] or pull[ ] a gun on Burt.” Maj. op. at 24. Burt’s theory of self-defense did not rely on a vague implication nor would the jury have had to deduce a defense from Lee’s testimony. To the contrary, Lee’s testimony was important because it would have confirmed a critical aspect of Burt’s own account of what happened — an account that, without Lee, “was supported solely by [Burt’s] own testimony.” Cf. Riley v. Payne, 352 F.3d *271313, 1319, 1320 (9th Cir. 2003) (granting writ for ineffective assistance of counsel where trial attorney failed to call “a critical witness to lend credibility to [the defendant’s] story of [a] threat by [the decedent];” “without any corroborating witnesses, [the defendant’s] bare testimony left him without any effective defense.”) (quoting Brown v. Myers, 137 F.3d 1154, 1158 (9th Cir. 1998)). Just as trial counsel’s failure to pursue similar evidence was deemed prejudicial in Riley v. Payne it ought to be deemed prejudicial here.

Finally, Lee’s testimony would have been significant independent of the question of whether Orange-Jackson ultimately pulled the gun from his waistband because the mere fact that Orange-Jackson had the gun could have supported an imperfect self-defense theory. See, e.g., People v. Cruz, 44 Cal.4th 636, 664, 80 Cal.Rptr.3d 126, 187 P.3d 970 (Cal. 2008) (“[W]hen a defendant kills in the actual but unreasonable belief that he or she is in imminent danger of death or great bodily injury, the doctrine of ‘imperfect self-defense’ applies to reduce the killing from murder to voluntary manslaughter.”). It is worth noting that a recent analysis of shootings by police officers in New York City reveals that “[i]n 77 percent of all shootings since 1998 when civilians were the targets, police officers were not fired upon.” A Bill to Require the New York City Police Department to Report Certain Informat’um About Police Shootings: Hearing on Int. 516-A Before N.Y. City Council Comm, on Public Safety (May 5, 2008) (statement of Christopher Dunn, Associate Legal Director, New York Civil Liberties Union), available at http://www.nyclu.org/node/ 1756. Furthermore, data from the seven most recent years available indicate that in forty percent of those incidents, the police shot at individuals who had not even “pointed firearms at officers.”3 'And of course there are the highly publicized cases of individuals such as Amadou Diallo, who was killed by police officers who “thought he had a gun[, which] turned out to be a wallet.”4 This is not unusual, as statistics show that one out of five persons shot at by New York police officers were not armed. An infinitesimally small number of police shootings result in prosecutions, let alone convictions,5 which suggests that shooting an individual who seems threatening and simply possesses a weapon is often considered justifiable, at least *28when law enforcement officers are doing the shooting.6 Of course the law, and more importantly prosecutors and juries, views police shootings very differently than other instances of self-defense, but surely not so differently that an alleged aggressor’s possession of a gun is reduced to an irrelevant piece of information that “does not speak to” the issue of self-defense. To the contrary, the fact that the purported aggressor in this case was carrying a gun at the time of the shooting is important information that would have strongly supported Burt’s theory of self-defense, imperfect or otherwise. Trial counsel’s failure to put forward such evidence, if it constituted deficient performance, would certainly be the type of error that is “sufficient to undermine confidence in the outcome” of Burt’s trial. Strickland v. Washington, 466 U.S. 668, 694, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984).

Because I disagree with the majority on the issue of prejudice, and because the district court did not reach the issue of deficient performance, I would remand for a finding on the latter question with the instruction that, if the performance was in fact deficient, the writ of habeas corpus be issued.