ALFRED R. DYER,
ARTHUR CALDERON, Warden,
of California State Prison
at San Quentin,
No. 95-99002 D.C. No.CV-89-03454-VRW
Appeal from the United States District Court
for the Northern District of California
Vaughn R. Walker, District Judge, Presiding
Argued and Submitted
December 18, 1997--San Francisco, California
Filed August 6, 1998
Before: Procter Hug, Jr., Chief Judge, and
James R. Browning, Betty B. Fletcher, Harry Pregerson,
Stephen Reinhardt, Melvin Brunetti, Alex Kozinski,
David R. Thompson, Diarmuid F. O'Scannlain,
Thomas G. Nelson and Andrew J. Kleinfeld, Circuit Judges.
Opinion by Judge Kozinski; Dissent by Judge Brunetti; Dissent by Judge O'Scannlain
Jon B. Streeter, Orrick, Herrington & Sutcliffe, San Francisco, California, for the petitioner-appellant.
Dane R. Gillette, Deputy Attorney General, San Francisco, California, for the respondent-appellee.
KOZINSKI, Circuit Judge:
Petitioner was convicted of murder and sentenced to death. We consider whether he was denied a fair trial because one of the jurors obtained her seat by lying during voir dire.
In 1980 Alfred Dyer and two friends took four hostages and drove them into the remote hills overlooking Oakland, California. They ordered the hostages out of the car and instructed them to lie down alongside the road. Dyer and an accomplice then shot all four; amazingly, two survived. Dyer admitted participating in the shootings but argued diminished capacity.
During voir dire the jurors had been asked the usual questions, including:
13. Have you or any of your relatives or close friends ever been the victim of any type of crime? . . .
15. Have you or any of your relatives or close friends ever been accused of any offense other than traffic cases?
Jessica Freeland, the last prospective juror to be questioned, answered "no" to both questions. She was named as an alternate and eventually was seated as a juror and helped decide Dyer's fate.
After the guilt-phase verdict, the defense learned that Freeland's brother Richard had been shot and killed some six years earlier. When questioned by the trial judge, Freeland explained that she had answered "no" to question 13 because she thought the shooting was an accident, not a crime. The trial judge accepted her explanation. The prosecutor--who had also prosecuted Richard Freeland's killer--had just turned over to the court the file in that case. A quick review of the file would have disclosed that the killing had none of the earmarks of an accident: Richard, who was just seventeen, was pistol-whipped four times and then shot in the back of the head. The prosecutor said nothing, the judge did not examine the file and therefore remained unaware of the facts which undermined Freeland's explanation. Freeland remained on the jury, which then sentenced Dyer to death. The California Supreme Court affirmed. See People v. Dyer, 45 Cal. 3d 26 (1988).
On federal habeas, Dyer's lawyers conducted an investigation of Jessica Freeland and discovered further evidence casting doubt on her veracity. The district court held an evidentiary hearing on the issue of juror bias but concluded that it was bound by the state court's finding of impartiality and denied relief. A divided panel of this court affirmed. See Dyer v. Calderon, 122 F.3d 720 (9th Cir. 1997). A majority of active judges subsequently voted to accept Dyer's suggestion for rehearing en banc.
A. The Sixth Amendment guarantees criminal defendants a verdict by impartial, indifferent jurors. The bias or prejudice of even a single juror would violate Dyer's right to a fair trial. See, e.g., United States v. Hendrix, 549 F.2d 1225, 1227 (9th Cir. 1977). One important mechanism for ensuring impartiality is voir dire, which enables the parties to probe potential jurors for prejudice. For voir dire to function, jurors must answer questions truthfully. Nevertheless, we must be tolerant, as jurors may forget incidents long buried in their minds, misunderstand a question or bend the truth a bit to avoid embarrassment. The Supreme Court has held that an honest yet mistaken answer to a voir dire question rarely amounts to a constitutional violation; even an intentionally dishonest answer is not fatal, so long as the falsehood does not bespeak a lack of impartiality. See McDonough Power Equip. v. Greenwood, 464 U.S. 548, 555-56 (1984).  Accordingly, we must determine whether Freeland's answers were dishonest and, if so, whether this undermined the impartiality of Dyer's jury. 
The state trial judge here held a brief hearing in chambers and found that Freeland was candid and impartial. The state clings to this finding, reminding us that juror impartiality is a question of historical fact. See Patton v. Yount, 467 U.S. 1025, 1038 (1984). Moreover, state court fact findings are presumed to be correct; we may set them aside only in limited circumstances specified by statute. See 28 U.S.C. §2254(d) (1)-(d)(8) (1994).  Dyer argues that one of these circumstances applies to him, namely that "the material facts were not adequately developed" by the state courts. 28 U.S.C. S 2254(d)(3) (1994).
B. Freeland's candor first came into question after the jury found Dyer guilty of murder but before the start of the penalty phase. Freeland's estranged husband, Melvin Provost, had been arrested for rape earlier that month. Although the details are murky, it seems that Dyer and Provost crossed paths in the courthouse jail and Provost told Dyer that Freeland's brother had been shot and killed some years earlier. According to Dyer's lawyer, Provost said that Freeland "had expressed strong views about that event and the circumstances surrounding it." The court turned down defense counsel's request that Provost be called as a witness but encouraged counsel to investigate.
The next morning, just before the start of the penalty phase, the court returned to the question of Freeland's bias. The prosecutor, Jerry Curtis, explained that he was familiar with the Richard Freeland homicide:
I remember the case of Richard Freeland as a homicide victim, I had the case myself, and it was a plea bargain case to a manslaughter. And it was a case where there was some kind of an argument in a block in front of some apartments and someone was killed.
Curtis then handed the file over to the judge, who called Freeland into chambers.
The judge asked Freeland if anyone in her family had been the victim of a homicide, and this time she answered yes. He then asked whether she was aware of any legal action taken, civil, criminal or otherwise. Freeland answered:"It was an accident. They didn't let us know when the trial was going on so we don't know what happened to the man that accidentally shot him." This, she felt, justified her voir dire answer: "Well, if it was an accident I don't think that's really a crime, is it?" Freeland then explained why she thought the shooting was accidental:
The guy is not some, you know, maniac or anything like that. As a matter of fact, I saw him once or twice and he didn't look like he did this to my brother on purpose. He was the manager of an apartment that my brother happened to be at. It was a party going on and he was just trying to clear everybody out of the party because there was a fight or something, and what supposedly happened is that the gun went off and it happened to shoot my brother. So you can't hate somebody or accuse him for doing it on purpose, or something like that. That was his--you know, that's what he said and, you know, we had no choice but to believe him. He didn't even know my brother, so he had no reason to just shoot him for nothing. So it was just an accident.
Dyer's lawyer asked Freeland if she had spoken to the police or to the District Attorney about the case, and she answered no. He then asked: "And no one from your family ever testified in court at all?" Freeland responded, "Oh, no." The trial judge did not ask Freeland about her husband's statement that she had "expressed strong views" about the shooting. Nor did he question Freeland about the fact that her husband was then in jail, which seemed to conflict with her answer to voir dire question 15. Provost himself was never questioned.
On the basis of this hearing, which seems to have lasted less than five minutes, the judge retained Freeland on the jury: "She has explained her answer [to the voir dire questions], and I would not characterize anything she did or did not do as demonstrating any lack of candor, as far as I'm concerned." Was it plausible that Freeland could be so confused about the circumstances of her brother's death? Yes, the judge explained:
There's no foundation establishing how well she knew the brother. We have brothers and we have brothers. We're making a whole series of assumptions. I don't know, you may not know, where the homicide occurred, how much information she had about it, except for her answers, and the record speaks for itself.
(Emphasis added.) Why the judge thought he could not verify the details of the crime--even where it occurred--when he had the casefile in his hands, is a mystery.
C. A court confronted with a colorable claim of juror bias must undertake an investigation of the relevant facts and circumstances. See 28 U.S.C. §2254(d)(3) (1994); Remmer v. United States, 350 U.S. 377, 379 (1956); Remmer v. United States, 347 U.S. 227, 230 (1954). An informal in camera hearing may be adequate for this purpose; due process requires only that all parties be represented, and that the investigation be reasonably calculated to resolve the doubts raised about the juror's impartiality. See Smith v. Phillips, 455 U.S. 209, 217 (1982); United States v. Boylan, 898 F.2d 230, 258 (1st Cir. 1990). So long as the fact-finding process is objective and reasonably explores the issues presented, the state trial judge's findings based on that investigation are entitled to a presumption of correctness. See Tinsley v. Borg, 895 F.2d 520, 526 (9th Cir. 1990).
As a predicate for his ultimate finding that Freeland was not biased, the state judge found that Freeland did not "demonstrate[ ] any lack of candor" in response to voir dire questions 13 and 15. The judge did not explain his reasoning fully, but his comments suggest the following chain of logic:
1. Freeland claims she did not mention her brother's murder in response to voir dire question 13 because she thought it was an accident.
2. She would not have been aware of the circumstances of her brother's death if they were not very close. ("We have brothers and we have brothers.")
3. There is no evidence that Freeland and her brother were close. ("There's no foundation establishing how well she knew the brother. . . . I don't know . . . where the homicide occurred, how much information she had about it.")
Freeland's relationship with her brother thus was the fulcrum of the judge's finding that Freeland was candid. If Jessica and Richard were siblings in name only--if they were separated at birth and lived in distant cities--Freeland's claim that she didn't know exactly how her brother died would be plausible.  But Jessica and Richard were not estranged; in fact, they lived under the same roof, with their mother. As one would expect, Freeland attended her brother's funeral. Their mother, moreover, testified at the preliminary hearing (there was no trial) of Richard's killer. At the time of the testimony, Jessica and her mother were still living together. Other readily available facts also suggest that the judge was wrong to infer that Jessica and her brother weren't close. The criminal casefile noted that Freeland was a plaintiff in a civil suit against Richard's killer. As a plaintiff in a wrongful death action, Freeland claimed she had suffered because of the death of her brother; this too suggests they were close enough for her to know that the killer was charged with a crime.
In any event, Freeland never claimed she was estranged from her brother or that she was unaware of how he died. Rather, she stated several times that she knew the details of the shooting and had reason to believe it was an accident. See supra at 8579-80. Whether her explanation was plausible--not whether she and her brother were close--was the key to determining if Freeland had lied during voir dire. The actual facts of the crime were remarkably different from Freeland's description: Richard was pistol-whipped four times and shot in the back of the head, much like Dyer's own victims. It is inconceivable that Freeland, living under the same roof as Richard, their mother and two other siblings, could have been so confused.
Are these facts the state trial judge should have uncovered? We think so. If the judge thought that the plausibility of Freeland's explanation turned on how well she knew Richard, he could have asked her. Or he could have asked Curtis, the prosecutor. Or he could have examined the casefile, which showed that Freeland's mother had testified at the preliminary hearing in the criminal proceeding, that Freeland herself was a plaintiff in the civil suit and that the shooter was charged with murder.  Or he could have ordered Provost brought in to testify as to what he knew about Freeland's relationship with her brother. Indeed, Provost had been reported as saying that Freeland had expressed strong feelings about her brother's killing--which was not only inconsistent with the nonchalance she displayed when questioned about it, but also bespoke bias on her part. 
The judge did none of these things, leaving the matter in doubt: "We're making a whole series of assumptions," he said. But a judge investigating juror bias must find facts, not make assumptions, and here the key facts were easily discernible. A few questions to Jessica about her relationship with her brother; an examination of the casefile; a request for confirmation from the prosecutor; an order to bring Provost into court--any or all of these would have disclosed that Freeland knew a lot more about her brother's death than she was letting on. Perhaps the judge was lulled into a false sense of security by the prosecutor's assurance that he was familiar with the Richard Freeland homicide; the judge might have understood the prosecutor's silence as confirming that Jessica's explanation was plausible.  In these circumstances, "it was the [trial] court's obligation to develop the relevant facts on the record, not merely presume them." United States v. Gaston-Brito, 64 F.3d 11, 13 (1st Cir. 1995). The judge's lack of verve in pursuing the matter casts doubt on his findings.
To warrant setting aside the presumption of correctness, facts left undeveloped by the state court must be "material," see 28 U.S.C. §2254(d)(3), not trivial or tangential to the issue presented. Indeed, something more than the usual meaning of "material" seems to be required: The missing facts must be "crucial," "vital" or "indispensable to a fair, rounded, development of the material facts." Townsend v. Sain, 372 U.S. 293, 321-22 (1963); see also Thomas v. Zant, 697 F.2d 977, 986 (11th Cir. 1983). What this means is that the evidence not discovered (that an adequate hearing would have discovered) must be so significant that a reasonable factfinder could have decided the bias question differently if armed with this knowledge. In other words, there must be a material possibility that the judge would have made a different decision if he had been aware of the facts in question.
Is there a material possibility the state trial judge would have found that Freeland was not impartial had he considered all the facts readily available to him? We believe there is. Had he questioned Freeland about her relationship with her brother, the judge would have discovered that the siblings were, in fact, very close; he would have learned that they lived in the same household and that she attended his funeral.
And, had the judge just cracked open the casefile, he would have learned that Freeland's mother testified in the criminal proceeding and that Freeland herself was a plaintiff in the wrongful death action. All this would have cast considerable doubt on Freeland's explanation that she thought the killing was an accident and could well have shaken the judge's confidence in Freeland's veracity. And, once the trial judge determined that Freeland lied about her brother's death, he might well have developed doubts about her impartiality.
The bias hearing was deficient in other ways as well. The trial judge left untouched Freeland's negative answer to voir dire question 15: "Have . . . any of your relatives or close friends ever been accused of any offense other than traffic cases?" The judge's failure to question Freeland about her husband's criminal history is remarkable because the issue of Freeland's potential bias arose when Dyer and Provost met in a jail cell during the trial. Situations where a juror's husband lands in the same jail cell as the defendant are sufficiently rare and troublesome to merit a close look by the trial judge.  There may have been a legitimate explanation for Freeland's failure to disclose that her husband was in jail, but the trial judge never put her to the test. Ten years later, in her deposition, Freeland claimed a fuzzy memory: "When I found out [Provost] was in trouble either the case was already over or we were already past voir dire." But Provost had been arrested a full month before the trial, and Freeland paid him a visit. Had she been asked about this at the time, the date of her visit could have been confirmed by consulting prison visitation records. Maybe Freeland knew of her husband's arrest during voir dire, maybe she didn't. The trial judge passed up the opportunity to find out.
The trial judge's most serious blunder was his failure to pursue Provost's statement that Freeland "expressed strong views" about the circumstances of her brother's killing. One potential juror had been excused for cause because she admitted to having "strong . . . feelings that people who are involved in drugs or have anything to do with drugs I just don't care for them." If Freeland had strong views about her brother's death--entirely plausible given Freeland's later admission that she was very close to her brother--this might have undermined her ability to render an impartial verdict; it certainly would have cast a sinister light on her failure to mention the killing in her response to question 13.
Dyer's lawyer, John Burris, went to the heart of the matter when he asked the court to bring Provost in for questioning. The trial judge refused because at that point defense counsel had no evidence to support the allegation; the whole thing might have been ginned up by Dyer to throw sand into the machinery of the trial. The judge thus reasonably directed Burris to gather basic information about "whether in fact there was a death and whether it was a death that occurred in such an instance, because we want to know the preliminary fact whether she has a brother by that name, and all of these things that have to be looked at." The judge held open the possibility of calling Provost later: "I don't know about subpoenaing somebody at this stage. It may be necessary to do that. It could become necessary, let me put it that way."
Yet, the next morning, when the prosecutor produced a casefile confirming Dyer's account of what Provost had told him, the judge did not call Provost. Richard's death was no invention; everything Dyer had reported that could be verified turned out to be true. The trial judge thus had no basis for shrugging off Provost's further statement that Freeland had expressed strong views about the shooting. Had Provost been called as a witness, he could have elaborated on his statement and reported any bias Freeland might have exhibited as a result of that traumatic experience. He most certainly could have shed light on whether Freeland was telling the truth when she claimed that she thought her brother was killed by accident. Since Provost was in jail, there would have been no delay in trying to locate and subpoena him. Had the judge ordered him to appear, he would have been brought in post-haste.
The California Supreme Court laid the blame for the inadequate factual development at the bias hearing on the shoulders of defense counsel. See People v. Dyer, 45 Cal. 3d 26, 59 (1988). See generally Keeney v. Tamayo-Reyes, 504 U.S. 1, 11-12 (1992) (applying cause-and-prejudice standard to inadequate factual development attributable to attorney error). Specifically, the court refused to take judicial notice that Freeland was a plaintiff in the civil action and that an opinion in the criminal case against Richard's killer stated that the attacker "struck the victim on the head four times with a pistol and then shot him in the back of the head." Dyer, 45 Cal. 3d at 59. It did so because "[e]ach of the facts of which defendant now requests us to take judicial notice could have been presented to the trial court." Id.
The California Supreme Court overlooked the fact that this evidence was before the trial court at the time of the hearing; it was all contained in the criminal casefile of Richard's killer. The court also did not take account of the fact that Curtis, the prosecutor, was very familiar with the case and had an obligation to share his information with the court and opposing counsel. And, of course, Freeland herself and Provost were available to be questioned on these points. This was not a case where defense counsel failed to discover relevant evidence or knowledgable witnesses.
Where Burris might have been more aggressive is in questioning Freeland and pointing out the conflict between her claim that she thought Richard's killing was an accident and the information available in the casefile. Burris, however, was seriously handicapped in preparing for the hearing. He could not contact Freeland and question her, so he had no clue as to how she would respond when confronted with her brother's murder. Even if Burris had access to the casefile before the hearing--which we assume though it's nowhere disclosed in the record--he could not have examined the file with an eye towards rebutting Freeland's as yet untold story. By the time Freeland testified, the file was with the judge. Nor could Burris be too aggressive in questioning Freeland, who would soon have his client's life in her hands. Freeland might well have taken umbrage at counsel's suggestion that she was lying, especially as it exposed her to charges of perjury. Burris recognized as much when he said, "[M]aybe I'm not in a position to really get a candid answer." Counsel also had his hands full preparing for the penalty phase, which began just minutes after the bias hearing.
Given the extremely delicate situation when a juror is suspected of prejudice or misconduct, the trial judge must assume the "primary obligation . . . to fashion a responsible procedure for ascertaining whether misconduct actually occurred and if so, whether it was prejudicial." United States v. Boylan, 898 F.2d 230, 258 (1st Cir. 1990). While a trial is ongoing, lawyers may not conduct the kind of aggressive investigation of jurors they would of other witnesses. In such circumstances the trial judge fulfills his duty only if he "erects, and employs, a suitable framework for investigating the allegation [of bias] and gauging its effects[.]" Id. Where juror misconduct or bias is credibly alleged, the trial judge cannot wait for defense counsel to spoon feed him every bit of information which would make out a case of juror bias; rather, the judge has an independent responsibility to satisfy himself that the allegation of bias is unfounded.
Here, Burris did quite a bit, given the constraints of time and circumstance. He promptly advised the court that one of the jurors may have failed to disclose highly relevant information during voir dire. At the trial judge's direction, he obtained corroboration for the information from the prosecutor, including the casefile. He requested that Provost be brought in for questioning, and when that request was denied, he had co-counsel interview Provost in jail and then reported what he said to the court. At the hearing, Burris asked Freeland the right questions and timely objected when the court ruled against him. And immediately following the hearing he drew the court's attention to the key fact contained in the file, namely that "the blue card from the prosecution . . . certainly suggests that her brother was shot in the back of the head and died in somewhat of a violent manner." Given the delicate circumstances and the fact that the penalty phase of the trial was moments away, it's hard to fault Burris for not doing more. It seems to us that counsel did enough to alert the trial court to the problem; it was then up to the judge to give the matter the attention it obviously deserved.
We are not unmindful that the trial judge too was in a difficult position. Through no fault of his own, the verdict in a five week murder trial was suddenly in jeopardy. Freeland was the last alternate juror; removing her would have required the court to set aside the guilty verdict, empanel a new jury and start from scratch.  Had Freeland's omissions been discovered during voir dire, another juror could have been selected. With five weeks of trial completed and a verdict in hand, the problem had no comfortable solution. No judge would be eager to discover bias in these circumstances, and we attribute the trial judge's complacency to an ostrich-like desire to avoid learning anything that would jeopardize the verdict.
Whatever the reason, the judge did not avail himself of evidence that was, almost literally, right under his nose. Instead, he made a finding that Freeland told the truth--a finding that is nearly inexplicable given what he knew, and positively irrational given what he could easily have learned. If what happened here is an adequate investigation into juror bias, we are at a loss to say what is not.
Because the facts were not properly developed by the state court, its finding that Freeland was unbiased is not entitled to a presumption of correctness.  Under such circumstances, the federal courts must address the question de novo, considering not only the evidence developed in state court, but also other evidence which comes to light afterward. See Rhoden v. Rowland, 10 F.3d 1457, 1460 (9th Cir. 1993). While the district court made no findings, it nevertheless held an evidentiary hearing. We can therefore consider the question of whether Freeland was biased with the benefit of a fully developed record. This record conclusively answers our two key questions: First, Freeland plainly lied when she answered "no" to voir dire questions 13 and 15; no rational trier of fact could find otherwise. And she lied again when she was questioned by the state judge in chambers--presumably to cover her earlier lie and remain on the jury. Second, Freeland's lies give rise to an inference of implied bias on her part. Because implied bias is a mixed question of law and fact reviewable de novo, see Burton v. Johnson, 948 F.2d 1150, 1158 (10th Cir. 1991), there is no need to remand to the district court for consideration of this issue in the first instance.
A. The circumstances of Richard's killing were such that Freeland could not possibly have confused it with an accident. Freeland's account of what happened made it sound like the shooter was waving his gun around, and the gun happened to go off in the direction of her brother.  But the California Court of Appeal reviewing the indictment of Richard's killer described the event as "an angry confrontation between [the shooter] and the victim[.]" The court concluded that "the fact that [the shooter] struck the victim on the head four times with a pistol and then shot him in the back of the head suggests an inference of malice." This paints quite a different picture from Freeland's account of the incident and explains why Curtis charged the shooter with murder.
Did Freeland know what really happened? No doubt. In her deposition, Freeland acknowledged that she and Richard were "very close." It is inconceivable that the brutal circumstances of Richard's killing would not have been discussed within the Freeland household; indeed, Freeland's father testified that no one in the family had ever spoken of the killing as an accident. Nor is it believable that Freeland would have been unaware that her mother testified at the killer's preliminary hearing, which happened within ten weeks of the shooting.
Freeland's account of Richard's death was just the tip of Pinocchio's nose. Freeland also lied when she said she had never been a victim of crime. When she was a child, at a time when her parents were separated, Freeland's father kidnapped her and her siblings. Freeland explained that "the police finally caught up with us and he was arrested." When Freeland was a teenager she was attacked by her cousin, Buddy, who broke into her house in the dark, showed a knife and pulled her underwear off before she fought him off. Freeland's father called the police. Freeland downplayed the incident in her deposition, but admitted that it "scared the hell out of me."
Freeland also failed to mention the many burglaries to which she had fallen prey. Her cars had been broken into so many times she lost count. One car had been stolen outright. Her home had been burglarized on at least three occasions. During her deposition Freeland vividly recalled walking in on one robbery and watching as the burglar fled through the back door with a sack of loot thrown over his shoulder "like Santa Claus' bag." Several potential jurors had disclosed burglaries during voir dire; Freeland watched as most of them were picked off by peremptory challenge.  Being the last to be seated, Freeland was not taken by surprise; the trial judge had instructed the potential jurors to "[m]ake sure that while this process is going on that you hear every question that is asked and every answer that is given[.]" Lest the point be lost, the judge continued, "[A]s I've indicated to you before, you have a right to hear, and a duty to hear. So make sure that you do."
Freeland sat through the questioning of 74 potential jurors over the course of five days and had time to consider how she would answer the same questions when her turn came. When her turn did come, Freeland answered "no" to voir dire questions 13 and 15, even though her incidents were far more serious, both in number and magnitude, than those disclosed by others who had preceded her.
Equally inexplicable was Freeland's negative response when she was asked if any "relative" had been "accused of any offense other than traffic cases." Freeland's husband had been arrested on rape charges just a month before trial. In her deposition, Freeland explained that she did not consider her husband to be a relative, but other prospective jurors had offered information about in-laws and former spouses. Freeland and Provost were still legally married, and Freeland went to visit him shortly after his arrest.
Even if her nondisclosure about Provost could somehow be explained, the arrest record of other relatives cannot. Freeland's father was arrested for the kidnapping already mentioned. Freeland's uncle, Jason Caldwell, who lived with the Freelands from time to time, had been arrested for murder in Louisiana in 1970. Caldwell was arrested again in 1978 in connection with an armed robbery and pled guilty to being an accessory after the fact. Caldwell was paroled to Oakland a few years later. In 1980 Freeland's brother Billy had been convicted of possessing brass knuckles. And less than a month before the voir dire, Billy was arrested for possession of LSD and marijuana.
Freeland's family had a long history with the law, and not every event might have come to mind as Freeland sat in the jury box during voir dire. But her failure to mention that any of her relatives had been accused of crime defies an innocent explanation. Nearly every close relative of hers had been arrested: Her husband for rape; her father for kidnapping; her uncle Jason for murder, and again for armed robbery; her brother Billy for possession of brass knuckles, and again for drug possession; her cousin Buddy for attempted rape of Freeland herself. Freeland overlooked too many incidents for us to attribute her responses to mere forgetfulness.
Indeed, in her deposition Freeland did not claim to be forgetful.  Rather, she steadfastly maintained that she--not the court, not the defense, not the prosecutor--was the best and only judge of what information was relevant. As to the countless burglaries of her cars and homes, she explained that she thought that these events had no bearing on her impartiality:
I didn't have time to rack my brain to remember every time my car or house had been burglarized and it didn't have a bearing on the case to me because if you live in Oakland, I don't know of anyone who lives in Oakland who has never been burglarized.
(Emphasis added.) Freeland even shrugged off the suggestion that burglary could be considered a crime: "It's just a way of life and it's just the way things go. It was not a traumatic event that caused me to be so bitter that I would irrationally try to pay back someone like Alfred Dyer who has evidently done burglaries or robberies." Id. Freeland even belittled the significance of her uncle's arrest for murder: "Am I to reveal everything, the little information I know about other relatives? That's ridiculous. If that's what they wanted and you feel I have been unfair by not saying a distant relative has done something, then let Alfred Dyer get a new trial. " Later, Freeland stated forthrightly, "I dislike giving information, period. Information that to me is not relevant. . . . [E]ven if my brother was the victim of a violent crime, how can you take that little information and say, `Here she was unfair to Alfred Dyer.' "
B. Because the record conclusively establishes that Freeland lied, and lied repeatedly, we proceed to consider whether her lack of candor reflects an "[in]ability to render an impartial verdict." Smith, 455 U.S. at 220.
Whether Freeland was actually biased--i.e., whether she was disposed to cast a vote against Dyer--is difficult to figure out eighteen years later. Freeland acknowledged that she was very close to her brother, and added that the period following his death was painful. It's certainly possible that anger about her brother's killing drove Freeland to finagle a seat on the jury so she could lobby for a conviction and death sentence. Freeland herself was a victim of many crimes, and she may have wanted to use her jury service to send a message of deterrence. On the other hand, the fact that many of her relatives had been arrested suggests she could have harbored some empathy for criminal defendants. However, we need not resolve the actual bias question, which would first require factual findings by the district court, because the implied bias issue is dispositive here. Thus, we need not speculate further about what motivated Freeland to lie.
In extraordinary cases, courts may presume bias based on the circumstances. See McDonough, 464 U.S. at 556-57 (Blackmun, Stevens and O'Connor, JJ., concurring) (accepting doctrine of implied bias in exceptional circumstances); id. at 558 (Brennan and Marshall, JJ., concurring in the judgment) (same); Zerka v. Green, 49 F.3d 1181, 1186 n.7 (6th Cir. 1995); Amirault v. Fair, 968 F.2d 1404, 1405-06 (1st Cir. 1992); Tinsley v. Borg, 895 F.2d 520, 527 (9th Cir. 1990); Cannon v. Lockhart, 850 F.2d 437, 440 (8th Cir. 1988); United States v. Eubanks, 591 F.2d 513, 517 (9th Cir. 1979); United States v. Allsup, 566 F.2d 68, 71-72 (9th Cir. 1977).
What kind of circumstances give rise to a finding of implied bias? In Smith v. Phillips, Justice O'Connor gave some examples of situations where bias may be presumed: "a revelation that the juror is a close relative of one of the participants in the trial or the criminal transaction, or that the juror was a witness or somehow involved in the criminal transaction." 455 U.S. at 222 (concurring opinion). Of course, a juror could be a witness or even a victim of the crime, perhaps a relative of one of the lawyers or the judge, and still be perfectly fair and objective. Yet we would be quite troubled if one of the jurors turned out to be the prosecutor's brother because it is highly unlikely that an individual will remain impartial and objective when a blood relative has a stake in the outcome. Even if the putative juror swears up and down that it will not affect his judgment, we presume conclusively that he will not leave his kinship at the jury room door. The effect of this factor would be impossible to predict: Would the juror yield to his sympathies, or fight them and lean the other way? There is no way to know, but permitting such a juror to serve would introduce into the jury room an extraneous influence that could materially color the deliberations. The juror in question would be lacking the quality of indifference which, along with impartiality, is the hallmark of an unbiased juror.
Jessica Freeland was not involved in the crime which was the subject of the case, nor did she have a personal relationship with any of the trial participants. But there is every indication that she was not indifferent to service on the jury. After watching a number of potential jurors disclose relatively minor crimes and get dismissed, she chose to conceal a very major crime--the killing of her brother in a way that she knew was very similar to the way Dyer was accused of killing his victims.  She also failed to disclose many other facts that would have jeopardized her chances of serving on Dyer's jury. Later on, when she was questioned about her brother's death, she lied once again by pretending she thought it was an accident, and by telling the judge that no one in her family had testified about the killing. The inference we draw from all this is that Freeland lied in order to preserve her status as a juror and to secure the right to pass on Dyer's sentence.
We don't know why Jessica Freeland so cherished her seat on Dyer's jury.  Jury service is a civic duty that citizens are expected to perform willingly when called upon to do so. But there is a fine line between being willing to serve and being anxious, between accepting the grave responsibility for passing judgment on a human life and being so eager to serve that you court perjury to avoid being struck. The individual who lies in order to improve his chances of serving has too much of a stake in the matter to be considered indifferent. Whether the desire to serve is motivated by an overactive sense of civic duty, by a desire to avenge past wrongs, by the hope of writing a memoir or by some other unknown motive, this excess of zeal introduces the kind of unpredictable factor into the jury room that the doctrine of implied bias is meant to keep out. 
A juror, like Freeland, who lies materially and repeatedly in response to legitimate inquiries about her background introduces destructive uncertainties into the process. There is, of course, the possibility that she did so because of some personal bias against the defendant which she managed to hide from the court. But a perjured juror is unfit to serve even in the absence of such vindictive bias. If a juror treats with contempt the court's admonition to answer voir dire questions truthfully, she can be expected to treat her responsibilities as a juror--to listen to the evidence, not to consider extrinsic facts, to follow the judge's instructions--with equal scorn. Moreover, a juror who tells major lies creates a serious conundrum for the fact-finding process. How can someone who herself does not comply with the duty to tell the truth stand in judgment of other people's veracity? Having committed perjury, she may believe that the witnesses also feel no obligation to tell the truth and decide the case based on her prejudices rather than the testimony.
Writing for a unanimous Court, Justice Cardozo concluded that a juror who lies his way into the jury room is not really a juror at all: "The judge who examines on the voir dire is engaged in the process of organizing the court. If the answers to the questions are wilfully evasive or knowingly untrue, the talesman, when accepted, is a juror in name only. " Clark v. United States, 289 U.S. 1, 11 (1933). Clark held that a juror who obtains that position by committing fraud on the court is no more entitled to the privileges of that position than a stranger who sneaks into the jury room: "His relation to the court and to the parties is tainted in its origin; it is a mere pretense and sham." Id. 
Were we to follow Clark to the letter, we would have to conclude that Dyer was not convicted by a jury of twelve, but by eleven jurors and one intermeddler. We need not go quite so far, because we can resolve the case on narrower grounds. Clark is nevertheless instructive because Justice Cardozo there equates a juror who lies his way onto the jury to a juror who is related to a litigant: "If a kinsman of one of the litigants had gone into the jury room disguised as the complaisant juror, the effect would have been no different. " Id. Just as we would presume bias if the brother of the prosecutor were on a jury, we presume bias where a juror lies in order to secure a seat on the jury. 
More is at stake here than the rights of petitioner; "justice must satisfy the appearance of justice. " Offutt v. United States, 348 U.S. 11, 14 (1954). An irregularity in the selection of those who will sit in judgment "casts a very long shadow." Cruz v. Abbate, 812 F.2d 571, 574 (9th Cir. 1987). A perjured juror is as incompatible with our truth-seeking process as a judge who accepts bribes. Cf. Bracy v. Gramley, 117 S. Ct. 1793, 1797 (1997).  In this regard we agree with Chief Judge Winter:
[C]ourts cannot administer justice in circumstances in which a juror can commit a federal crime in order to serve as a juror in a criminal case and do so with no fear of sanction so long as a conviction results. The government's brief exhibits no concern over the possible criminality of the juror's conduct and asks us to affirm without further inquiry. . . .[W]hether the government chooses to prosecute such cases is not for us to decide. We need not reduce its incentives to take such conduct seriously, however, by giving the government cause to believe that overlooking juror misconduct will preserve tainted convictions.
United States v. Colombo, 869 F.2d 149, 152 (2d Cir. 1989) (internal citations omitted).
Not all jurors may walk a perfectly straight line. A distracted juror might fail to mention a magazine he subscribes to. An embarrassed juror might exaggerate the importance of his job. Few voir dires are impeccable, and most irregularities can be shrugged off as immaterial to the fairness of the trial. But the magnitude of Freeland's lies and her remarkable display of insouciance--her expressed feeling that only she would decide what matters--fatally undermine our confidence in her ability to fairly decide Dyer's fate. The facts here add up to that rare case where we must presume juror bias.  See Burton v. Johnson, 948 F.2d 1150 (10th Cir. 1991).
Presumed bias dates back in this country at least to Aaron Burr's trial for treason, where Chief Justice Marshall, riding circuit, noted that an individual under the influence of personal prejudice "is presumed to have a bias on his mind which will prevent an impartial decision of the case, according to the testimony." Marshall explained, "He may declare that notwithstanding these prejudices he is determined to listen to the evidence, and be governed by it; but the law will not trust him." United States v. Burr, 25 F. Cas. 49, 50 (D. Va. 1807) (emphasis added). In the common law, implied bias can be traced all the way back to Sir Edward Coke's dictum in Bonham's Case that no man shall be judge in his own cause. See Dr. Bonham's Case, 77 Eng. Rep. 646, 652 (C.P. 1610). This pedigree neatly disposes of the state's argument that implied bias would be a "new rule" barred by Teague v. Lane, 489 U.S. 288 (1989). Implied bias may indeed be the single oldest rule in the history of judicial review, as Bonham's Case is often identified as the first case in which a court struck down a duly enacted legislative act. See, e.g., Suzanna Sherry, The Founders' Unwritten Constitution, 54 U. Chi. L. Rev. 1127, 1130 (1987).
Judge O'Scannlain nevertheless argues that implied bias is not an established rule of constitutional procedure because we can point to no Supreme Court case announcing it. But a rule needs to be announced for purposes of Teague only if it's new. What we have here is the antithesis of Teague--a rule so deeply embedded in the fabric of due process that everyone takes it for granted. This is precisely what the Supreme Court did in Clark v. United States, 289 U.S. at 11. Offering an obvious example of someone who would not be qualified to serve as a juror, it mentioned a relative of one of the parties: "If a kinsman of one of the litigants had gone into the jury room disguised as the complaisant juror, the effect would have been no different." Id. Such a juror, the Court said, would be a juror "in name only." Id. The Court there understood--as every court that has dealt with the question has understood--that prejudice must sometimes be inferred from the juror's relationships, conduct or life experiences, without a finding of actual bias.
In Tumey v. Ohio, 273 U.S. 510, 532 (1927), the Court made it clear that bias must sometimes be presumed, even though the decision maker may be completely unbiased. In Tumey, a man in a small Ohio town was tried for unlawful possession of liquor. The Mayor also served as judge. In addition to his salary as Mayor, he received a small cut of each fine he imposed as a judge. The Supreme Court unanimously held that a conviction so imposed violates due process. The Court first acknowledged that some fact-finders would not actually be prejudiced by a personal stake in the outcome: "There are doubtless mayors who would not allow such a consideration as $12 costs in each case to affect their judgment in it[.]" Id. But, the Court held, subjective state of mind is not dispositive when the circumstances create an inference of bias. "[T]he requirement of due process of law in judicial procedure is not satisfied by the argument that men of the highest honor and the greatest self-sacrifice could carry it on without danger of injustice." Id. Rather, the Court held, we must look to the fairness of the procedure and whether bias should be presumed from the circumstances: "Every procedure which would offer a possible temptation to the average man as a judge to forget the burden of proof required to convict the defendant, or which might lead him not to hold the balance nice, clear and true between the State and the accused, denies the latter due process of law." Id. As in Clark, the Court in Tumey never stopped to question whether bias can be presumed from circumstances. Rather, it accepted the doctrine of implied bias as a given and applied it to the facts of the case.
Nothing in Smith v. Phillips, 455 U.S. 209 (1982) or McDonough Power Equip., Inc. v. Greenwood, 464 U.S. 548 (1984) is to the contrary. In Smith, Justice Marshall dissented, worrying that the majority was doing away with the doctrine of implied bias. Justice O'Connor joined the opinion on the understanding that it did not abolish the doctrine, see id. at 221, and the majority said nothing to the contrary. Smith has no bearing on our case unless we assume the majority tacitly did what Justice Marshall accused it of doing. That it did not is confirmed by Greenwood, where the Court held that the bias could not be implied on the facts of that case, but assumed that the doctrine does, in fact, exist. See Greenwood, 464 U.S. at 556-57 (Blackmun, J., concurring, joined by Stevens and O'Connor, JJ.) (fifth, sixth and seventh votes of majority). Accord Leonard v. United States, 378 U.S. 544 (1964); Tumey, 273 U.S. at 532.
No opinion in the two centuries of the Republic--except the dissent in our case--has suggested that a criminal defendant might lawfully be convicted by a jury tainted by implied bias.  Under the dissent's logic, reasonable jurists could hold that Dyer would have been accorded due process even if he had been convicted by a jury comprised of the following twelve individuals: (1) the mother of Jerry Curtis, the prosecutor, (2) Curtis's former law partner, (3) Oakland's Chief of Police, (4) the Grand Dragon of the Bay Area KKK, (5) the sister of Floyd Murray, who died in the shooting, (6) Floyd Murray's mother, (7) the victim of Dyer's prior robbery, (8) Dyer's ex-wife, (9) the District Attorney, (10) an Oakland councilman running for re-election on a "tough-on-crime" platform, (11) Melvin Provost, Dyer's cellmate, and (12) Provost's wife, Jessica Freeland--so long as they had all sworn they would be fair. We, on the other hand, believe that no reasonable jurist would take that position. Rather, jurists of reason would all agree that each of these individuals, had they made their way onto the jury, should have been struck without stopping to inquire into their subjective state of mind.
The panel opinion is VACATED, the district court is REVERSED and the case is REMANDED for further proceedings.
 We do not condone any lying by jurors; perjury is perjury. We are concerned here, however, with the rights of the defendant, not with whether the juror may be prosecuted for a deliberate lie during voir dire. See, e.g., Clark v. United States, 289 U.S. 1, 11 (1933).
 The presence of a biased juror cannot be harmless; the error requires a new trial without a showing of actual prejudice. See United States v. Allsup, 566 F.2d 68, 71 (9th Cir. 1977). Like a judge who is biased, see Tumey v. Ohio, 273 U.S. 510, 535 (1927), the presence of a biased juror introduces a structural defect not subject to harmless error analysis. See generally Arizona v. Fulminante, 499 U.S. 279, 307-10 (1991).
 The AEDPA does not apply retroactively to Dyer's appeal. See Lindh v. Murphy, 117 S. Ct. 2059 (1997).
 We say this only because we give very wide berth to the state trial judge's findings. For our part, we find it hard to believe that a sister would be aware that her brother was killed by gunfire yet be so unmindful of the details as to believe it was an accident. This is especially so since Richard's mother was alive and no doubt discussed the details of the killing as part of the normal bereavement process. See Eric Schlosser, A Grief Like No Other, Atlantic Monthly, September 1997, at 52.
 As the case ended in a plea bargain, the file was not very big, so it would not have taken very long for the judge to examine it. Nevertheless, if the judge did not wish to postpone the sentencing hearing to make such an examination, he could have taken the matter under advisement and examined the file over the course of the next day or two.
 The judge was thus wrong when he said that "the record speaks for itself" as to Freeland's lack of bias. Provost's statement, as reported by Dyer's lawyer, was also part of the record and flatly contradicted Freeland. While Provost's statement, as reported, was multiple-level hearsay, it could have been made much more definite if Provost had been questioned about it at the bias hearing.
 The prosecutor had an ethical obligation to bring to the judge's attention evidence that would cast light on the question on which the judge focused--namely whether Jessica was in a position to know how her brother was killed. Had this incident occurred in federal court, we might have considered whether the prosecutor's silence amounted to misconduct. See, e.g., United States v. Kojayan, 8 F.3d 1315 (9th Cir. 1993).
 For one thing, one would have to wonder whether Provost told Freeland anything about Dyer's statements and conduct that might bear on her evaluation of his guilt or innocence.
 We express no view as to whether such a juror would have to be excused as a matter of due process. We note this only as an authoritative determination of how the state law would operate in these circumstances.
 Because the verdict was in, the trial judge would probably not have been free to reopen jury deliberations, even if another alternate had been available.
 State appellate court findings are normally entitled to a presumption of correctness as well. See Sumner v. Mata, 449 U.S. 539, 546 (1981). But the California Supreme Court here did not cure the inadequacy of the trial court's hearing. It explicitly refused to consider any of the circumstances surrounding Richard's shooting. See People v. Dyer, 45 Cal. 3d 26, 59 (1988). Because the California Supreme Court's findings were grounded in the same incomplete record as the trial court's, its findings are no more entitled to a presumption of correctness.
 Because we conclude that Freeland lied, we need not decide whether dishonesty is a necessary predicate to a finding of juror bias. Compare McDonough, 464 U.S. at 556, with id. at 556-57 (Blackmun, Stevens and O'Connor, JJ., concurring); id. at 558 (Brennan and Marshall, JJ., concurring in the judgment); Zerka v. Green, 49 F.3d 1181, 1186 n.7 (6th Cir. 1995); Amirault v. Fair, 968 F.2d 1404, 1405-06 (1st Cir. 1992); Cannon v. Lockhart, 850 F.2d 437, 440 (8th Cir. 1988).
 That anyone would think such a reckless killing is not a crime itself stretches credulity.
 No doubt these events were embarrassing to Freeland, and she may not have wanted to air her family's dirty linen in public. Yet she could have asked that she talk about the events in chambers, as other potential jurors had done during voir dire.
 Courtney Coleman stated that "I've had my house--when I was living in Union City when I was married I had my house broken into several times. I also had my wife's car stolen twice." Barbara Covarrubias stated, "My house was broken into three times in one year." Dale Gritton stated, "My brother has had his apartment broken into several times." John Braucht stated that a burglar, "I guess supposedly under the influence of drugs, broke into my parents' home while they were there[.]"
 The district court did not hear Freeland testify because she evaded twenty-one separate attempts by the petitioner to subpoena her. Because the district court deferred to the trial court under section 2254, we do not have the benefit of an independent finding of impartiality to review. The district court did, however, review Freeland's deposition testimony, and was not convinced that Dyer had shown enough to overcome the presumption of correctness. This, of course, is not the same as an independent finding of lack of bias.
 At the time of voir dire the judge had already informed the venire that Dyer was on trial for murder and attempted murder, with a handgun, and that the state had asked for the death penalty.
 In yet another bizarre twist, after Dyer's trial Freeland became an officer with the California Department of Corrections and served as a guard on death row in San Quentin, where Dyer is held awaiting execution. Freeland later became a parole agent, and on at least two occasions--in violation of prison regulations--she reviewed and photocopied portions of Dyer's confidential Department of Corrections Central File to check up on the status of his appeal. While it is hard to know what to make of these facts--they are too few in number to prove that Freeland was on a vendetta or had a particular interest in seeing Alfred Dyer executed--they certainly are not consistent with the picture of a model indifferent juror.
 For example, a juror who lies his way on because he secretly plans to write a memoir of the experience might then vote differently to provide drama, or he might inject personal prejudice into the jury room in an attempt to jazz up the deliberative process. Similarly, a juror who lies because he disputes the importance of following the judge's instructions may also walk to his own beat when considering the meaning of reasonable doubt. See Jeffrey Rosen, One Angry Woman, New Yorker, Feb. 24- Mar. 3, 1997, at 54-64 (noting erosion of public confidence in unanimous jury system caused by irrational holdout jurors).
 The question in Clark was whether the juror's statements in the jury room could be considered in convicting her of having lied during voir dire. In holding that the statements could be considered, the Court had to overcome the privilege then accorded to communications between jurors during deliberations. The Court held that the privilege does not apply because a juror who obtains his position by fraud is not in fact a juror. While this statement was made in a context very different from ours, it is clearly a holding and not dicta.
 Having committed perjury, a juror in Freeland's position might realize that the likelihood of being prosecuted diminishes significantly if she votes to convict. Compare Clark, 289 U.S. at 9, with United States v. Colombo, 869 F.2d 149, 152 (2d Cir. 1989).
 The doctrine of implied bias has a counterpart in the canons of judicial ethics which require judges to disqualify themselves if they have even the smallest stake in the proceedings, or if they believe that their impartiality might reasonably be questioned. See Bracy, 117 S. Ct. at 1797; ABA Code of Judicial Conduct, Canon 3E (1990). Both rules safeguard both the existence and the appearance of justice.
 We appear to be unanimous that if Freeland knew the circumstances of her brother's death, we must presume bias. See Brunetti dissent, infra, at 8609 ("In order for Freeland to have the necessary relationship to create the emotional involvement she would have had to know and understand the circumstances and facts of her brother's death to be a brutal murder."). We part company only as to whether such knowledge can be imputed to Freeland.
 The dissent cites United States v. Malloy, 758 F.2d 979 (4th Cir. 1985), for the proposition that "the federal courts of appeals were split on the issue of implied bias." O'Scannlain dissent at 8622. Malloy does nothing more than refuse to find implied bias on a particular set of facts. See Malloy, 758 F.2d at 981 ("Malloy urges us to adopt a per se rule [of implied bias] in joint service cases. We decline the invitation."). Courts disagree (e.g., Smith) about when the doctrine applies, not whether it exists.
Brunetti, Circuit Judge, dissenting, with Judges Thompson, O'Scannlain, and Kleinfeld, joining.
Today's opinion sets forth a new rule of law that is a substantial departure from our precedent and deals a serious blow to the power of the trial court. Our system of law mandates deference to the trial court on certain issues regarding facts and credibility of witnesses. Today, that deference is pushed aside so that the appellate court may not only review the applicable laws, but also decide the facts. Our position has never been to sit as a finder of fact and I cannot support the extension of our power to do so. For the reasons that follow, I respectfully dissent.
I. Actual Bias Standard
The Sixth Amendment "guarantees to the criminally accused a fair trial by a panel of impartial, `indifferent' jurors." Irvin v. Dowd, 366 U.S. 717, 722 (1961). The Constitution "does not require a new trial every time a juror has been placed in a potentially compromising situation. " Smith v. Phillips, 455 U.S. 209, 217 (1982). As the majority points out, due process requires only that the defendant be tried by a "jury capable and willing to decide the case solely on the evidence before it, and a trial judge ever watchful to prevent prejudicial occurrences and to determine the effect of such occurrences when they happen." Id. For Dyer to be entitled to a new trial, he would have to "first demonstrate that a juror failed to answer honestly a material question on voir dire, and then further show that a correct response would have provided a valid basis for a challenge for cause." McDonough Power Equipment v. Greenwood, 464 U.S. 548, 556 (1984).
In reviewing a petition for habeas, we presume the correctness of state court findings of "basic, primary, or historical facts" unless one of the eight statutory exceptions applies. Thompson v. Koehane, 516 U.S. 99,109-10 (1995). Because resolution depends heavily on the trial court's appraisal of witness credibility and demeanor, juror impartiality is a factual issue that falls within the statutory presumption of correctness. Id. The majority contends that the trial court was so lax in its investigation of Freeland that its determination that Freeland was impartial is not entitled to the presumption of correctness. I disagree.
The state trial court made findings that Freeland's answers on voir dire were not dishonest or intentionally misleading. At the in camera hearing, the trial court had an adequate opportunity to witness Freeland's demeanor and questioned her about her understanding of the circumstances surrounding her brother's death and whether she believed it would affect her partiality. The trial court refused to "characterize anything [Freeland] did or did not do as demonstrating any lack of candor." Also, the trial court found that Freeland's allegedly false answers to the two voir dire questions at issue were "inadvertent" and were given in "good faith."
Absent application of a statutory exception, we must give presumptive weight to the finding of Freeland's impartiality. The majority refuses to apply the presumption because, they believe, (1) material facts were not developed, and (2) clear and convincing evidence demonstrates that the state court's determination as to Freeland's honesty was erroneous. I disagree.
The majority concludes that material facts about Richard's death were not developed adequately at the trial court proceeding. However, the trial court found Freeland honestly believed that her brother's death was an accident, and therefore was not a crime. This finding is not inconsistent with Freeland and Richard being close, the fact that the family was seriously affected by his death, or even institution of a wrongful death suit and eventual recovery. Similarly, that Freeland's mother testified in the criminal proceedings is of no import since it was not shown that Freeland knew about this testimony. Therefore, like the majority of the original Ninth Circuit panel, I would not disrupt the state courts' determination that Freeland answered honestly based on evidence unknown to the state trial judge that is wholly consistent with such a finding.
The majority assumes that Freeland obviously lied by failing to reveal that she was attacked by her cousin, that her father had been arrested for kidnapping, that her uncle had been arrested for murder, that her brother had been arrested on drug charges, that Freeland's cars and home had been burglarized, and that her former husband had been arrested for rape. However, as to each of these additional facts, Freeland has explained why she originally failed to disclose them. The district court explicitly credited Freeland's testimony and found that the new evidence did not detract from the state courts' findings about Freeland's credibility.
The procedural history of this case illustrates that Freeland's credibility was thoroughly scrutinized on multiple occasions. First, the state trial court held an in camera hearing between the guilt and penalty phases and concluded that Freeland was not biased. A motion for mistrial was then denied. Second, the state trial court held a hearing on this issue in response to Dyer's motion for a new trial. The trial court found that the voir dire questions were ambiguous and that Freeland's responses were inadvertent. Third, on direct appeal, the California Supreme Court agreed that the questions were ambiguous and found no evidence that Freeland was lying when she said that no member of her family had been a crime victim. Fourth, on federal habeas, the district court conducted an evidentiary hearing accepting new evidence and still affirmed the trial court's finding that Freeland had answered honestly, refusing to presume bias.
Accordingly, we have before us not just the state courts' findings of impartiality, but also the district court's findings that are based on the additional facts that Dyer had since proffered and on which the majority now focuses its attention. As the district court found, none of these additional facts provides any reason to disturb the state courts' finding that Freeland was honest in her responses. See Nix v. Williams, 467 U.S. 431, 449-50 (1984) (newly discovered evidence failed to demonstrate that facts were not adequately developed in state court). None of the district court's findings as to Freeland's credibility were clearly erroneous. Although the number of crimes committed by members of her family seems high, and the number of incidents in which she or members of her family were victims of crimes also seems high, these facts do not establish that Freeland was dishonest. Freeland is from a crime-ridden community -- a fact which plausibly both affects her understanding of what qualifies as a crime and influenced her answers to Questions 13 and 15. The trial court and the state supreme court both found the voir dire questions ambiguous and that Freeland's answers in camera and in deposition were honest. The record as a whole, including Freeland's deposition testimony and her responses at the in camera hearing, is adequate to support the findings that she was not being dishonest during voir dire. Therefore, the additional facts do not destroy the presumption that the state courts' findings of fact are correct.
The majority rejects the state courts' factual findings and instead concludes that Freeland's responses were patently implausible. To reach this conclusion they continually recite their assumption that `Freeland lied and lied repeatedly.' Essentially, the majority is re-finding the facts and, through speculation and assumption regarding these newly-found facts, concluding that Freeland's explanations defy common sense. That is not our proper function. The trial court saw Freeland testify and simply did not believe that she lied. Our judicial system relies heavily on the trial judge to make this very sort of determination. Trial judges, viewing the testimony first hand and, using all their senses, are in the best position to determine credibility. For this reason, appellate judges properly defer to the trial judge's factual findings of credibility under our laws. See Knaubert v. Goldsmith, 791 F.2d 722, 727 (9th Cir. 1986) (stating that "[w]e can think of no sort of factual finding that is more appropriate for deferential treatment than is a state court's credibility determination."). As the majority of the prior three judge panel concluded, "[s]imply put, while there are substantial questions raised as to her credibility, the findings by the state trial court, added to by the findings of the district judge, have not been demonstrated by Dyer to be clearly erroneous -- given the special credibility deference we are required to apply: she believed what she said was true at that time." Dyer v. Calderon, 122 F.3d 720, 730 (9th Cir. 1997). Actual bias is the issue here; not whether she lied on the juror questionnaire or voir dire. See McDonough, 464 U.S. at 556.
II. Implied Bias
Not only does the majority find actual bias, based on the assumed lie, but they also find implied bias based on the circumstances. The Supreme Court has never explicitly held that we may infer or presume bias based on the totality of the circumstances. We should not do so now.
The majority relies on case law where we have said that "bias could be implied or presumed from the `potential for substantial emotional involvement, adversely affecting impartiality,' inherent in certain relationships." Tinsley v. Borg, 895 F.2d 520, 527 (9th Cir. 1990) quoting United States v. Allsup, 566 F.2d 68, 71 (9th Cir. 1977). However, these cases are inapplicable to today's case. In those cases the court found that the relationship of a juror to the subject of the trial was too close. In Allsup, we held that two jurors in a bank robbery trial were partial despite the district court's finding of impartiality because the jurors were employees of a different branch of a bank that was robbed. Allsup, 566 F.2d at 71. Similarly, in United States v. Eubanks, 591 F.2d 513 (9th Cir. 1979), a heroin conspiracy trial, we presumed bias where a juror did not disclose that two of his children were in prison for heroinrelated crimes. Id. at 516-17. In this case, there is no such relationship. In order for Freeland to have the necessary relationship to create the emotional involvement she would have had to know and understand the circumstances and facts of her brother's death to be a brutal murder. To the contrary, Freeland believed her brother's death was an accident. Accordingly, she would have no impermissible substantial emotional involvement to Dyer's case.
The majority contends that this is such an extreme and extraordinary case that we must presume bias. They claim that Freeland must be presumed partial because Freeland's brother was shot several times in the head, just as the victims in this case. It is true that on rare occasions, "[c]ourts have been willing to presume bias where a juror or his close relatives have been personally involved in a situation involving a similar fact pattern." Tinsley, 895 F.2d at 528 (citing cases). However, as stated in our discussion of actual bias, we accept the trial court's finding that Freeland honestly believed that her brother had been killed accidentally. That said, the remaining similarity -- death by gunshot -- does not render this case extraordinary or extreme.
While at first glance it may appear extraordinary that Freeland has experienced such tragedy and so much crime, considering the alleged rate of crime in Freeland's neighborhood, her circumstances, though lamentable, are not surprising or extraordinary. Indeed, Freeland has experienced crime not only as a victim but also as the family member of criminals. Thus, it would be pure speculation to say that there was an inherent "potential for substantial emotional involvement, adversely affecting impartiality." Id. at 527 (internal quotation omitted).
The majority would go further, however. Now, sixteen years after the trial, they would conduct in the appellate court an inquiry similar to a perjury trial to impeach the former juror Freeland when in fact the state trial court conducted a face-to-face hearing at the time that the issue of Freeland's bias was raised.
Finally the majority's holding requiring this "perjury inquiry" to sustain an implied bias finding is clearly a new rule barred by Teague, and I join and concur in Judge O'Scannlain's dissent which ably sets forth the Teague application in this case.
I would affirm the district court's denial of the writ of habeas corpus.
O'Scannlain, Circuit Judge, with whom Brunetti, Thompson, and Kleinfeld, Circuit Judges, join, dissenting.
I respectfully dissent from the majority's holding that we are not foreclosed by Teague v. Lane, 489 U.S. 288 (1989), from declaring that the entire hierarchy of California courts committed constitutional error by not implying juror bias. 
State court judges are our co-equal partners in the protection of federal constitutional rights.  Although Congress has granted us the authority to grant habeas corpus relief to state prisoners, due consideration of our circumscribed role in the federal system counsels prudent restraint in exercising that extraordinary power to second-guess state courts. Lest we forget, in Teague, the Supreme Court instructed us not to impose constitutional rules on our state court brethren that were not compelled by existing precedent when a habeas petitioner finished raising his claims on direct review. As the Court explained in Gilmore v. Taylor, 508 U.S. 333 (1993), Teague " `validates reasonable, good-faith interpretations of existing precedents made by state courts,' and thus effectuates the States' interest in the finality of criminal convictions and fosters comity between federal and state courts." Id. at 340 (quoting Butler v. McKellar, 494 U.S. 407, 414 (1990)) (internal citation omitted).
Under Teague, the reviewing federal court must inquire whether " `a state court considering [the petitioner's] claim at the time his conviction became final would have felt compelled by existing precedent to conclude that the rule [he] seeks was required by the Constitution.' " Lambrix v. Singletary, 117 S. Ct. 1517, 1524 (1997) (quoting Saffle v. Parks, 494 U.S. 484, 488 (1990)) (emphasis added). That is to say, in order to conclude that application of a rule is not barred by Teague, the federal court must determine not only that the rule was "a reasonable interpretation of prior law," or even the "most reasonable" interpretation, but also that "no other interpretation was reasonable." Id. at 1530. Because state courts would not have felt compelled by precedent (existing on the date Dyer's conviction became final) to conclude that the "implied-bias rule" developed by the majority was required by the Constitution, application of that rule is barred by Teague.
In Teague, the Supreme Court held, "[s]ubject to two narrow exceptions," Gilmore v. Taylor, 508 U.S. 333, 339 (1993), that "new constitutional rules of criminal procedure will not be applicable to those cases which have become final before the new rules are announced." Teague, 489 U.S. at 310. The federal court's inquiry under Teague must be conducted in three steps. First, the federal court must determine the date on which the petitioner's conviction became final. See Caspari v. Bohlen, 510 U.S. 383, 390 (1994). Second, it must "[s]urvey the legal landscape as it then existed," Graham v. Collins, 506 U.S. 461, 468 (1993), to "determine whether a state court considering [the petitioner's] claim at the time his conviction became final would have felt compelled by existing precedent to conclude that the rule [he] seeks was required by the Constitution." Saffle, 494 U.S. at 488. Finally, "if the court determines that the habeas petitioner seeks the benefit of a new rule, the court must consider whether the relief sought falls within one of the two narrow exceptions to nonretroactivity." Lambrix, 117 S. Ct. at 1524- 25.
Dyer's conviction became final on October 31, 1988, the date on which the United States Supreme Court declined to review the California Supreme Court's decision affirming Dyer's conviction on direct review. See People v. Dyer, 45 Cal. 3d 26, 246 Cal. Rptr. 209, cert. denied, 488 U.S. 934 (Oct. 31, 1988).
In surveying the legal landscape as it existed on that date, it is necessary to determine whether "precedent " would have "compelled" a state court to conclude that the Constitution "required" the implied-bias rule applied by the majority. Saffle, 494 U.S. at 488. The majority purports to rely on three sources of precedent in support of its conclusion that juror bias may be implied: (1) decisions of the United States Supreme Court; (2) decisions of the lower federal courts; and (3) the common law.
Our survey of the decisions of the Supreme Court begins with Smith v. Phillips, 455 U.S. 209 (1982). In Smith, the Court examined whether a habeas petitioner's right to an impartial jury had been violated because a juror submitted an application for employment as an investigator in the District Attorney's Office during the petitioner's trial. See id. at 212. The petitioner argued that "[g]iven the human propensity for self-justification . . . the law must impute bias to jurors in [such a] position." Id. at 215. The Supreme Court "disagree[d]," explaining that it had "long held that the remedy for allegations of juror partiality is a hearing in which the defendant has an opportunity to prove actual bias. " Id.; see also id. at 215-17 (discussing Chandler v. Florida, 449 U.S. 560 (1981), Remmer v. United States, 347 U.S. 227 (1954), and Dennis v. United States, 339 U.S. 162 (1950)).
Justice O'Connor concurred in the Court's opinion in Smith. Justices Marshall, Brennan, and Stevens dissented. The concurring and dissenting Justices disagreed as to whether the Smith majority had in fact precluded the possibility that an implied-bias rule is required by the Constitution. In her concurring opinion, Justice O'Connor wrote "to express [her] view that the [majority] opinion does not foreclose the use of `implied bias' in appropriate circumstances." Id. at 221 (O'Connor, J., concurring). However, Justice Marshall, joined by Justices Brennan and Stevens, expressed a different view of the majority opinion: "According to the majority, the Constitution requires only that the defendant be given an opportunity to prove actual bias." Id. at 228 (Marshall, J., joined by Brennan and Stevens, JJ., dissenting) (emphasis added).
For Teague purposes, it matters little whether Justice O'Connor or the dissenting Justices in Smith arrived at the correct interpretation of the majority opinion; what does matter, quite simply, is that reasonable jurists could disagree. I posit what should be considered a most unremarkable point: Justices Marshall, Brennan, and Stevens are reasonable jurists. Because they could read a majority opinion of the United States Supreme Court as precluding the possibility of a constitutional implied-bias rule, I cannot conclude, unlike my majority colleagues, that all reasonable jurists would be compelled to locate such a rule in the Constitution.
Would reasonable jurists have been compelled to change their minds between 1982, when Smith was decided, and October 31, 1988, when Dyer's conviction became final? To answer this question, we first look to the Supreme Court's decision in McDonough Power Equip., Inc. v. Greenwood, 464 U.S. 548 (1984). In McDonough, a four-Justice plurality stated that, in order to obtain a new trial on account of juror bias, "a party must first demonstrate that a juror failed to answer honestly a material question on voir dire, and then further show that a correct response would have provided a valid basis for a challenge for cause." Id. at 556. In a concurring opinion, Justice Blackmun, joined by Justices Stevens and O'Connor, observed that the plurality opinion did not "foreclose" the possibility that bias may be implied. Id. at 556 (Blackmun, J., joined by Stevens and O'Connor, JJ., concurring). In yet another concurring opinion, Justice Brennan, joined by Justice Marshall, expressed disagreement with the plurality's analysis, arguing that courts may consider whether there are "any facts in the case suggesting that bias should be conclusively presumed." Id. at 558 (Brennan, J., joined by Marshall, J., concurring).
State courts would not have felt compelled by these concurring opinions to conclude that bias may be implied as a matter of federal law. The McDonough Court reversed the Tenth Circuit's holding that a new trial was required on account of juror bias, see McDonough, 464 U.S. at 549; therefore, the statements in both concurring opinions regarding implied bias were dicta not necessary to the holding. See United States v. Boatwright, 822 F.2d 862, 864 (9th Cir. 1987) (Kennedy, J.) ("The requirement [in United States v. Echegoyen, 799 F.2d 1271 (9th Cir. 1986)], that two independent searches be in progress [for the challenged evidence to be admitted] is dictum, as the case admits the challenged evidence."); Hutchinson v. Amateur Elec. Supply, 42 F.3d 1037, 1047 (7th Cir. 1994) ("Any authority for denying prejudgment interest in [Donnelly v. Yellow Freight Sys., Inc., 874 F.2d 402 (7th Cir. 1989)] thus is dicta, since the Donnelly court reversed the district court's denial of interest."); United States v. Helmsley, 985 F.2d 1202, 1207 (2d Cir. 1993) ("The statement in [Mills v. Scully, 826 F.2d 1192 (2d Cir. 1987), regarding potential deprivations of due process] was dictum since Mills reversed the grant of a writ of a habeas corpus."). Dicta in Supreme Court opinions are not binding, see McDaniel v. Sanchez, 452 U.S. 130, 141 (1981); Ayala v. United States, 550 F.2d 1196, 1200 (9th Cir. 1977), and are certainly insufficient to compel state courts to conclude that a rule is required by the Constitution under Teague. See Lambrix, 117 S. Ct. at 1525 (authority that supports point "in dictum" does not " `control[ ]' or `dictate[ ]' the result" for Teague purposes).
Between its decision in McDonough and 1988, the Supreme Court did not directly address the issue of implied bias. Thus, the decisions of the Supreme Court could not have "compelled" state courts to conclude that the Constitution requires the majority's implied-bias rule. 
The majority would apparently hold that, irrespective of whether there is a "Supreme Court case announcing " a constitutional rule of criminal procedure, the rule is not barred by Teague if it is "so deeply embedded in the fabric of due process that everyone takes it for granted." Maj. Op. at 8601. I am unable to agree. A rule is "new" under Teague unless "a state court considering [the petitioner's] claim at the time his conviction became final would have felt compelled by existing precedent to conclude that the rule [the petitioner] seeks was required by the Constitution." O'Dell, 117 S. Ct. at 1973 (holding that rule is not "new" only if this standard is met); Lambrix, 117 S. Ct. at 1524 (using same language to describe standard); Caspari, 510 U.S. at 390 (same); Saffle, 494 U.S. at 488 (same). As the Supreme Court's repeated formulation of the Teague inquiry has made clear, Teague asks not only (1) whether a rule is required by (or deeply embedded in) the Constitution, but also (2) whether existing precedent compelled that conclusion at the time the petitioner's conviction became final.
By "existing precedent," the Court could not have meant the Constitution itself. Otherwise, the Teague inquiry would not only be rendered circular (that is, Teague would ask whether "the Constitution compelled the conclusion that the rule the petitioner seeks is required by the Constitution"), but also toothless. Even if a rule was not compelled by Supreme Court decisions, a federal court would always be able to circumvent Teague simply by concluding that the rule had always been "deeply embedded in the fabric" of the Constitution. (Indeed, it is the rare case in which a federal court applies a constitutional rule of criminal procedure that the court does not believe to be "deeply embedded in the fabric" of some constitutional provision, or combination thereof.) If the Supreme Court had intended the result reached by the majority, it would simply have omitted the words "by existing precedent" from the Teague inquiry, and would have stated that a rule is not barred by Teague if state courts would "have felt compelled [ ] to conclude that the rule the petitioner seeks was required by the Constitution." But see Saffle, 494 U.S. at 488 (Teague asks whether state courts would "have felt compelled by existing precedent to conclude that the rule [the petitioner] seeks was required by the Constitution.") (emphasis added).
Even if there were a "deeply-embedded-in-the-fabric" exception to Teague, the Supreme Court's summary of its juror-bias decisions in Smith v. Phillips makes it clear that it would not apply here. The Smith Court stated, in pertinent part:
In argument before this Court, respondent . . . contends that a court cannot possibly ascertain the impartiality of a juror by relying solely upon the testimony of the juror in question. Given the human propensity for self-justification, respondent argues, the law must impute bias to jurors in Smith's position. We disagree.
This court has long held that the remedy for allegations of juror partiality is a hearing in which the defendant has the opportunity to prove actual bias. For example . . . .
Smith, 455 U.S. at 215 (emphasis added). Significantly, the Smith Court then proceeded to support this view by summarizing the Court's juror-bias jurisprudence, including its decisions in Dennis v. United States, 339 U.S. 162 (1950), Remmer v. United States, 347 U.S. 227 (1954), and Chandler v. Florida, 449 U.S. 560 (1981).  Even if the majority's historical sketch of juror-bias cases were somehow more accurate than that provided by the Smith Court (I respectfully submit that it is not), reasonable state judges sitting in 1988 were entitled to rely on the Smith majority's interpretation of precedent for the proposition that implied bias is not constitutionally required. Such state judges certainly could not have been expected to foresee our own interpretation of that precedent, decreed a decade later in 1998!
The majority's contention that "[n]o opinion in the two centuries of the Republic--except the dissent in our case--has suggested that a criminal defendant might lawfully be convicted by a jury tainted by implied bias," Maj. Op. at 8603, is beside the point for three reasons. First, even assuming that the Supreme Court had not said that "the Constitution does not require an implied-bias rule," it certainly does not follow that the Court has said that "the Constitution does require an implied-bias rule"; only the second proposition is important under Teague. Second, while no court has uttered the precise words used by the majority, many reasonable jurists have concluded, in light of existing precedent, that defendants must demonstrate actual bias to obtain a new trial. See, e.g., Irons v. Lockhart, 741 F.2d 207, 208 (8th Cir. 1984) ("In Smith v. Phillips . . . the Supreme Court held that a petitioner who seeks habeas corpus relief based on an allegation of juror bias must prove actual bias, either in a state court or federal court hearing, and that a court cannot impute bias based on the petitioner's bare allegations."); Rogers v. McMullen, 673 F.2d 1185, 1189 (11th Cir. 1982) ("The Supreme Court [in Smith v. Phillips] rejected the implied bias argument and held that due process requires only that a defendant have the opportunity at a post-trial hearing to prove actual bias."); United States v. Whiting, 538 F.2d 220, 223 (8th Cir. 1976) ("Where an attack is made upon the integrity of the trial by reason of alleged misconduct on the part of a juror in failing to disclose information pertinent to the issue of prejudice, the defendant's burden of proof must be sustained not as a matter of speculation, but as a demonstrable reality. No demonstration of intentional or knowing withholding of information by [the juror] is made here."). Finally, contrary to what the majority suggests, nowhere does this dissent take a position on whether the Constitution actually requires an implied-bias rule; rather, it simply observes that reasonable jurists could conclude that such a rule was not required at the time Dyer's conviction became final.
Buried beneath the majority's "fabric of due process" lies one relevant enduring fact. That is, Supreme Court precedent did not compel the conclusion that an implied-bias rule was required by the Constitution at the time Dyer's conviction became final. Cf. Tinsley v. Borg, 895 F.2d 520, 527 (9th Cir. 1990) ("The Supreme Court has never explicitly adopted or rejected the doctrine of implied bias.").
We next examine whether, as of the date Dyer's conviction became final, the decisions of the lower federal courts would have compelled state courts to conclude that an implied-bias rule was constitutionally required. Because a state court could reasonably have concluded -- and perhaps should have concluded -- that it was not bound by any lower federal court's interpretation of the United States Constitution on any matter, it follows a fortiori that lower federal court decisions would not have bound that state court with respect to the specific question of implied bias.
As of 1988, at least three federal courts of appeals had concluded that, with respect to the interpretation of federal law, state courts are bound only by the decisions of the United States Supreme Court, and not by the decisions of the lower federal courts.  See Bromley v. Crisp, 561 F.2d 1351, 1354 (10th Cir. 1977) ("[T]he Oklahoma Courts may express their differing views on the retroactivity problem or similar federal questions until we are all guided by a binding decision of the Supreme Court."); United States ex rel. Lawrence v. Woods, 432 F.2d 1072, 1074 (1972) ("The federal Circuit Courts of Appeals and, in respect to federal law, the state courts of last resort, are subject to the supervisory jurisdiction of the Supreme Court of the United States. They are, however, as to the laws of the United States, co-ordinate courts.") (quoting Iowa Nat'l Bank v. Stewart, 232 N.W. 445, 454 (1930)); Owsley v. Peyton, 352 F.2d 804, 805 (4th Cir. 1965) ("Though state courts may for policy reasons follow the decisions of the Court of Appeals whose circuit includes their state, they are not obliged to do so.") (internal citation omitted). Although the Supreme Court had, as of 1988, neither adopted nor rejected this position, several individual Justices had expressed a similar view. See Steffel v. Thompson, 415 U.S. 452, 482 n.3 (1974) (Rehnquist, J., joined by Burger, C.J., concurring); Perez v. Ledesma, 401 U.S. 82, 125 (1971) (Brennan, J., joined by White and Marshall, JJ., dissenting).
Further, a substantial majority of the state courts that had addressed this issue, including California's, had concluded that they were not bound by the decisions of the lower federal courts on federal questions. Compare, e.g., Cowan v. Myers, 187 Cal. App. 3d 968, 985, 232 Cal. Rptr. 299 (1986) ("[T]he decisions of the lower federal courts, even on federal questions, are not binding on this court."); State v. Webster, 114 Wis. 2d 418, 426 n.4, 338 N.W.2d 474 (1983); State v. Glover, 60 Ohio App. 2d 283, 287, 396 N.E.2d 1064 (1978); Greene v. State, 11 Md. App. 106, 110, 273 A.2d 830 (1971), with Handy v. Goodyear Tire & Rubber Co., 230 Ala. 211, 160 So. 530 (1935); Kuchenmeister v. Los Angeles & S.L.R., 52 Utah 116, 172 P. 725 (1918). In light of this line of authority, a state court could reasonably have concluded that it was not bound by the decisions of any lower federal court on any federal constitutional issue. Accordingly, such lower federal court decisions could not "compel" state courts within the meaning of Teague. Compare Clemmons v. Delo, 124 F.3d 944, 955 n.11 (8th Cir. 1997) (assuming without deciding that "when the [Supreme] Court says `firmly dictated by precedent,' it means Supreme Court precedent"); Glock v. Singletary, 65 F.3d 878, 885 (11th Cir. 1995) (en banc) (courts of appeals do not "dictate" particular rule to state courts for Teague purposes), with Jiminez v. Myers, 40 F.3d 976, 979-81 (9th Cir. 1994) (using Ninth Circuit decisions to support conclusion that "totality of the circumstances" rule was not "new" under Teague). For this reason, the majority errs in relying upon United States v. Burr, 25 F. Cas. 49, 50 (D. Va. 1807), a federal district court opinion, for the proposition that Teague is inapplicable.
Indeed, even assuming that all reasonable state court judges would have concluded, contrary to the weight of the authority, that they were bound to follow lower federal court precedent, Teague would still dictate that such judges were not compelled to recognize an implied-bias rule. At the time Dyer's conviction became final, the federal courts of appeals were split on the issue of implied bias. Compare United States v. Eubanks, 591 F.2d 513, 517 (9th Cir. 1979) (holding that bias could be implied as a matter of law), with United States v. Malloy, 758 F.2d 979, 982 n.6 (4th Cir. 1985) (rejecting implied-bias theory and collecting cases in which other lower federal courts had done so); see also Debtor Reorganizers, Inc. v. State Bd. of Equalization, 58 Cal. App. 3d 691, 696, 130 Cal. Rptr. 64 (1976) ("As between the decisions of the Ninth Circuit and [those] of the Fifth Circuit[on federal law], no primacy inheres in the former, so the persuasiveness of the conflicting views must depend upon the validity of the arguments made therein."). A state court judge sitting in 1988 could reasonably have followed those lower federal courts that had held that a defendant must establish actual bias to obtain a new trial; perforce, state court judges were not compelled to adopt the implied-bias rule. See Lambrix, 117 S. Ct. at 1530 ("[Teague asks] whether no other interpretation [of existing precedent] was reasonable.").
As of the time Dyer's conviction became final, the conclusion that an implied-bias rule was required by the Constitution was not compelled by either: (1) the decisions of the Supreme Court; or (2) the decisions of the lower federal courts.
Today the court holds that there is a third source of compelling authority: Sir Edward Coke's dictum in Dr. Bonham's Case, 77 Eng. Rep. 646, 652 (C.P. 1610).  See Maj. Op. at 8601. Not likely! Our own Supreme Court has stated that "authority that supports the point in dictum" does not " `control[ ]" or `dictate[ ]' the result" under Teague.See Lambrix, 117 S. Ct. at 1525. The Court could hardly have intended an exception to this Teague-based rule, which applies even to the Court's own opinions, for the dictum of a seventeenth-century English common-law judge.
I respectfully dissent.
 For the reasons expressed in Judge Brunetti's dissent, I would also accord the presumption of correctness to the factual findings made by the state trial court, and conclude that Juror Freeland was not shown to be actually biased. I would, therefore, affirm Judge Walker's denial of the petition for writ of habeas corpus.
 See Robb v. Connolly, 111 U.S. 624, 637 (1884) ("Upon the state courts, equally with the courts of the Union, rests the obligation to guard, enforce, and protect every right granted or secured by the constitution of the United States . . . .").
 The first "limited" exception to Teague applies to new rules " `forbidding criminal punishment of certain primary conduct [and] rules prohibiting a certain category of punishment for a class of defendants because of their status or offense.' " O'Dell v. Netherland, 117 S. Ct. 1969, 1973 (1997) (quoting Penry v. Lynaugh, 492 U.S. 302, 330 (1989)). The second, "even more circumscribed," exception applies to " `watershed rules of criminal procedure implicating the fundamental fairness and accuracy of the criminal proceeding.' " Id. (quoting Graham, 506 U.S. at 478). As the Supreme Court observed in Graham, "[w]hatever the precise scope of this [second] exception, it is clearly meant to apply only to a small core of rules requiring observance of those procedures that . . . are implicit in the concept of ordered liberty." Graham, 506 U.S. at 478. The majority does not suggest that either of these exceptions applies here.
 I strongly disagree with the majority's contention that "Smith has no bearing on our case unless we assume the majority tacitly did what Justice Marshall accused it of doing." Maj. Op. at 8602. Rather, Smith "has no bearing on our case" only if Justice Marshall's position was unreasonable. That it was not unreasonable is demonstrated by the fact that many other reasonable jurists have since joined Justice Marshall in reading Smith to hold that the Constitution requires only that defendants be given an opportunity to prove actual bias. See, e.g., United States v. Malloy, 758 F.2d 979, 981-82 (4th Cir. 1985); Irons v. Lockhart, 741 F.2d 207, 208 (8th Cir. 1984); Rogers v. McMullen, 673 F.2d 1185, 1189 (11th Cir. 1982).
 This conclusion is not affected by the Supreme Court's statement in Stringer v. Black, 503 U.S. 222 (1992), that the application of a rule is not barred by Teague when it emerges not "from any single case . . . but from [a] long line of [Supreme Court] authority . . . ." Id. at 232. No such "long line" of Supreme Court authority compelled jurists to recognize the majority's implied-bias rule at the time Dyer's conviction became final. Id. If such a "long line" ever existed (it did not exist, see Smith, 455 U.S. at 215-17 (collecting cases)), it certainly was severed by Smith. See Saffle, 494 U.S. at 488 (Teague asks whether state court would have felt compelled by "existing precedent.") (emphasis added).
 It is unclear whether the majority would create a separate Teague inquiry for rules that are "deeply embedded in " the Constitution, as opposed to those that are merely "required by " the Constitution. Such a distinction is entirely without support in the Supreme Court's Teague jurisprudence.
 The majority's argument that no reasonable jurist would fail to impute bias automatically to the majority's parade of horrible jurors (the mother of one of the victims, Dyer's ex-wife, the District Attorney, etc.), see Maj. Op. at 8603, is flawed for two reasons. First, irrespective of whether they had all "sworn to be fair," id., these hypothetical jurors (unlike Juror Freeland) clearly would not have survived application of California's own juror bias laws. See Weathers v. Kaiser Found. Hosp., 5 Cal. 3d 98, 110 & n.5 (1971) (intentional concealment on voir dire constitutes grounds for new trial). Lest we forget, we sit as a federal court in habeas review of the California courts, which found no bias (as a matter of either state or federal law) in this case.
Even in the unlikely event that, as the majority suggests, the federal Constitution itself requires courts to impute bias to jurors (in the absence of a showing of intentional concealment), we are nonetheless barred by Teague from applying an implied-bias rule unless existing precedent compelled that rule at the time Dyer's conviction became final. The majority simply fails to point to any such compelling precedent.
 As its discussion of these juror-bias cases demonstrates, the Smith Court's statement that it had "long held that the remedy for allegations of juror partiality is [an actual bias hearing], " id. at 215, was not limited to any particular factual situation.
 We must determine whether a state court would have felt bound by lower federal court decisions based upon a reasonable interpretation of existing precedent at the time Dyer's conviction became final. To hold a decade after the fact that state courts were bound by precedent that reasonable jurists would not have viewed as binding would be to turn Teague on its head. As the Supreme Court stated in O'Dell v. Netherland, 117 S. Ct. 1969 (1997), "[a]t bottom, . . . the Teague doctrine `validates reasonable, good-faith interpretations of existing precedents made by state courts even though they are shown to be contrary to later decisions.' " Id. at 1973 (quoting Butler v. McKellar, 494 U.S. 407, 414 (1990)). It follows that Teague "validates reasonable, good-faith interpretations of existing precedents" concerning the binding effect of "existing precedents." Butler, 494 U.S. at 414.
 In Yniquez v. Arizona, 939 F.2d 727 (9th Cir. 1991), decided three years after Dyer's conviction became final, we stated in dicta that "[d]espite the authorities that take the view that the state courts are free to ignore decisions of the lower federal courts on federal questions, we have serious doubts as to the wisdom of this view. " Id. at 736. However, in vacating our later decision in that case, the Supreme Court characterized our discussion of the binding effect of lower federal court decisions on state courts as "remarkable." Arizonans for Official English v. Arizona, 117 S. Ct. 1055, 1064 n.11 (1997) (citing Lockhart v. Fretwell, 506 U.S. 364, 375-76 (1993) (Thomas, J., concurring) (Supremacy Clause does not require state courts to follow rulings of federal courts of appeals on questions of federal law)).
 The majority opinion states:
In the common law, implied bias can be traced all the way back to Sir Edward Coke's dictum in Bonham's Case that no man shall be judge in his own cause. See Dr. Bonham's Case, 77 Eng. Rep. 646, 652 (C.P. 1610). This pedigree neatly disposes of the state's argument that implied bias would be a "new rule " barred by Teague v. Lane, 489 U.S. 288 (1989). Implied bias may indeed be the single oldest rule in the history of judicial review, as Bonham's Case is often identified as the first case in which a court struck down a duly enacted legislative act. See, e.g., Suzanna Sherry, The Founders' Unwritten Constitution, 54 U. Chi. L. Rev. 1127, 1130 (1987).
Maj. Op. at 8601. No matter how "old[ ] " such a "rule," id., state courts could not have been compelled by precedent to conclude that it was required by the Constitution. See Lambrix, 117 S. Ct. at 1524. The list of "old rules" whose application by state courts is not compelled by precedent (holding that such rules are constitutionally required) may be limitless, e.g.: The Ten Commandments, see Exodus 20:7-17; "If a man destroy the eye of another man, they shall destroy his eye," Hammurabi's Code §196 (c. 1700 B.C.); and "Do unto others as you would have them do unto you," Matthew 7:12.