We must determine whether alleged judicial coercion or juror misconduct warrants habeas relief for a convicted double-murderer.

I

Officers William H. Freeman and Roy Paul Blecher were murdered on December 22, 1978 on the 1-80 Freeway in West Sacramento, California at approximately 3:40 a.m. At trial, the prosecution presented evidence that Rodriguez and his girlfriend Margaret Klaes had engaged in a cross-state crime-spree involving armed robberies, kidnapping, and auto-thefts leading up to the murders. Acquaintances testified that Rodriguez had said he would kill any officers who interfered with him. Rodriguez’s drivers’ license was found next to the officers’ patrol car and footprints at the scene matched his shoes. Klaes testified that she was in the stolen car Rodriguez was driving when they were stopped for speeding, and that Rodriguez had murdered the two officers who pulled them over. An other acquaintance testified that Rodriguez had boasted of “downing” two officers.

In his defense, Rodriguez suggested that Klaes, with her drug suppliers, had been involved in the murder, and that she had framed Rodriguez. He testified that on the night of the murder, he and Klaes had been drinking and taking cocaine at the home of Jeri Engel in Crockett, California. He claimed that they left Crockett at 1:30, arriving in West Sacramento at around 2:30, where he went directly to sleep in a motel room. He therefore could not have been at the murder scene an hour later. Rodriguez further testified that Klaes drove off in the Camaro to buy more cocaine, and that she acted in a suspicious manner the following morning when she returned to the motel. He claimed that Klaes carried his drivers’ license and gun in her purse, giving her the opportunity to plant incriminating evidence. The defense sought to undermine further Klaes’s credibility by pointing to the fact that she was testifying in exchange for a plea-bargained lesser crime, with a maximum sentence of three years, and that she was jealous because she had read a letter written by Rodriguez to another woman.

Both the defense and the prosecution presented evidence relating to the timing of events on the night of the murder. First, it was undisputed that the distance between Crockett and West Sacramento was about 57 miles, and that the travel time was approximately one hour. However, the time of departure from Crockett was sharply disputed. The State presented evidence that Rodriguez and Klaes left Crockett at around 2:30 a.m., placing them at the murder scene about an hour later. Rodriguez presented evidence suggesting that the couple left at 1:30 a.m. One witness, Wanda Hawthorne, testified that she was driving from San Francisco to Sacramento and noticed a brown Camaro around 2:00 a.m. which followed her car between Vacaville and Davis, California. Rodriguez claims that Hawthorne’s testimony supports his theory that the couple left Crockett around 1:30 a.m.

The prosecution and defense also presented testimony from several drivers passing near the scene of the murder. These witnesses provided widely conflicting statements about the color and model of the car that was stopped next to the officer’s patrol car.

The evidentiary phase of the trial spanned three months and included more than 100 witnesses and 200 exhibits. The jury began deliberations on January 22, 1981. On the sixth day of deliberations, one juror fell ill and an alternate was chosen with deliberations beginning anew. The newly comprised jury met for a total of 15 court days. During that time, the jury sent five separate notes on four different days indicating that they were deadlocked. They also made many requests for testimony to be read back and received three supplemental instructions. Judge Karesh, the presiding judge, refused five defense motions for a mistrial, commented on the evidence which the jury had requested while giving a supplemental instruction, and, after the jury repeatedly declared that it was deadlocked, learned that the jury’s split had remained at 11-1 for each of the five ballots taken. Finally, on February 25, the jury returned a unanimous verdict finding Rodriguez guilty of the double murder. The jury recommended the death penalty, which the judge imposed. A protracted appeals process in the State court system followed. The California Supreme Court affirmed Rodriguez’s conviction on direct appeal in 1986, but remanded to the trial court for the judge to make an independent ruling on an automatic application for modifi cation of the death sentence. People v. Rodriguez, 42 Cal.3d 730 , 230 Cal.Rptr. 667, 709 , 726 P.2d 113, 155 (1986). Among its many-other rulings, the California Supreme Court rejected Rodriguez’s claim that the verdict was the product of judicial coercion. Id. 230 Cal.Rptr. at 696 , 726 P.2d at 142 .

On remand, Judge Karesh went beyond the remand instructions and considered a habeas petition requesting a new trial on the grounds of juror misconduct. The defense, six years after the conviction, contacted the jurors and learned that one member of the jury had told the others about his calculations of the distance and driving time between Crockett and the scene of the murders.

In an affidavit, juror Shelby Martin admitted that he had been visiting friends in Crockett at some point during the trial and had driven past a bar Rodriguez visited on the night of the murder. He looked at his odometer and then drove to the freeway. When he exited the freeway at his destination, he saw a sign indicating the distance to Sacramento. He figured the time it would take to drive to the scene of the murder, and concluded that Rodriguez could have been at the scene at the time the murders occurred. He shared his distance and time calculations with other jurors when deliberations were centered on the time issue. He also related to the other jurors that during his drive he had been unable to recall with accuracy the makes and models of cars parked by the side of the road during his drive.

Judge Karesh granted the habeas petition, ordering a new trial. He also modified Rodriguez’s sentence from the death penalty to life in prison without parole because he had granted the Writ. The California Court of Appeal reversed in 1991, holding that the juror misconduct was harmless because the travel time between the two points was undisputed at trial, making Martin’s misconduct immaterial. The court once again remanded for Judge Karesh to conduct a proper consideration of Rodriguez’s automatic application for modification of the death sentence. The California Supreme Court refused to consider an appeal of this decision. On remand, Judge Karesh granted Rodriguez’s motion to modify the sentence to life in prison without parole. The California Court of Appeal affirmed in June 1993.

In May 1994, Rodriguez filed a petition for a writ of habeas corpus in federal district court which was assigned to a magistrate judge for findings. The magistrate issued a Report and Recommendation in October 1995, recommending that the court grant the petition because she concluded that Rodriguez had been deprived of his constitutional rights by Martin’s misconduct and by judicial coercion of the verdict.

The district court rejected this recommendation because the magistrate erred in her application of the law, and further held that Rodriguez had faded to establish that his conviction was substantially and injuriously affected by any constitutional errors. The district court agreed with the California Court of Appeal that the crucial question before the jury was a determination of the time Rodriguez left Crockett. Furthermore, it was undisputed at trial that it took about an hour to drive the 57 miles from Crockett to the scene of the crime. Since nothing Martin could have learned about time and distance between the two points could have any bearing on Rodriguez’s departure time, the court held that the misconduct was harmless.

The district court also rejected the allegations of judicial coercion, finding that each of Judge Karesh’s rulings was lawful, and that their cumulative effect was not coercive. He therefore denied the petition for habeas relief. Rodriguez filed this timely appeal requesting a new trial, almost twenty years after the two officers were murdered, on the grounds of juror misconduct and judicial coercion. Each claim will be considered in turn.

II

Rodriguez had a constitutionally protected right to have the jury decide the case based on evidence subject to confrontation, cross examination, and the assistance of counsel. See Thompson v. Borg, 74 F.3d 1571, 1574 (9th Cir.), cert. denied, Thompson v. White, — U.S.-, 117 S.Ct. 227 , 136 L.Ed.2d 159 (1996). On habeas review, Rodriguez bears the burden of establishing that a constitutional error had a “substantial and injurious effect or influence in determining the jury’s verdict.” Brecht v. Abrahamson, 507 U.S. 619, 637 , 113 S.Ct. 1710, 1722 , 123 L.Ed.2d 353 (1993). Whether an instance of juror misconduct was prejudicial to the defendant is a mixed question of law and fact which we review de novo. See Thompson, 74 F.3d at 1573 .

There is no bright line test for determining whether a defendant has suffered prejudice from an instance of juror misconduct. See Marino v. Vasquez, 812 F.2d 499, 505 (9th Cir.1987). In assessing prejudice claims in juror misconduct cases we place great weight on the nature of the extraneous information that has been introduced into deliberations. See Jeffries v. Wood, 114 F.3d 1484, 1490 (9th Cir.1997) (“[T]he appropriate focus should be on the nature of the information itself.”), petition for cert. filed (Aug. 11, 1997) (No. 97-289); see also, Lawson v. Borg, 60 F.3d 608, 612 (9th Cir.1995); Dickson v. Sullivan, 849 F.2d 403, 407-8 (9th Cir.1988). Juror misconduct which warrants relief generally relates “directly to a material aspect of the case.” United States v. Bagnariol, 665 F.2d 877, 885 (9th Cir.1981). However, the introduction of duplicative or cumulative extraneous material may render juror misconduct harmless. See Jeffries, 114 F.3d at 1491 , Hughes v. Borg, 898 F.2d 695, 700 (9th Cir.1990).

Furthermore, we consider whether there is a “direct and rational connection between the extrinsic material and a prejudicial jury conclusion, as distinguished from a connection that arises only by irrational reasoning.” Bagnariol, 665 F.2d at 885 . We are not to consider evidence concerning the subjective impact of extrinsic evidence on the deliberation process, and instead focus on an objective inquiry into the potential for prejudice from the extraneous information. See Dickson, 849 F.2d at 406 .

Typically, juror misconduct which warrants relief involves the introduction into deliberations of matter not in evidence or in the instructions. See Thompson, 74 F.3d at 1574 . We have granted a new trial where the jury receives extraneous information that is ordinarily excluded from trial as inflammatory or unduly prejudicial. See, e.g., Jeffries, 114 F.3d at 1490 (concerning extraneous information of prior armed robbery conviction); Lawson, 60 F.3d at 612 (finding prejudice where jurors learned that defendant had a reputation as a violent man); Dickson, 849 F.2d at 408 (involving a deputy’s statement to jurors that defendant had “done something like this before.”). On the other hand, we have refused to grant a new trial where the extraneous information is cumulative of information presented at trial or immaterial to the ultimate issues. See Bagnariol, 665 F.2d at 888-89 (“Reviewing courts will not disturb jury verdicts on appeal when extraneous information relates only to issues not material to the guilt or innocence of the defendant.”) (gathering cases).

Rodriguez claims that Martin’s declaration regarding his presentation of extrajudicial information to other jurors reveals two separate grounds for reversal: first, Martin’s “memory experiment” in which he noted that he could not recall the makes and models of cars parked on the side of the road with accuracy, and secondly the evidence which Martin presented regarding the travel time to the site of the murder.

A

Martin informed the other jurors that he had attempted to discern and recall the makes and colors of ears on the side of the freeway and that he could not do so with accuracy. As an initial matter, we note that the difficulty of discerning and recalling objects while driving at freeway speeds is the kind of common knowledge which most jurors are presumed to possess. See Bagnariol, 665 F.2d at 888 (discounting claim of prejudice where extraneous information was something “any reasonable juror already knew.”); see also Hard v. Burlington Northern R. Co., 870 F.2d 1454, 1461 (9th Cir.1989) (denying new trial where one juror used personal knowledge of x-ray interpretation to sway others because “[i]t is expected that jurors will bring their life experiences to bear on the facts of the case.”). Nor is the matter the kind of prejudicial material which is routinely kept from the jury.

Furthermore, since many different witnesses testified about the ears they saw stopped on the freeway, all with different recollections, it is elementary that most of these witnesses could not have been recalling the scene with accuracy. See Rodriguez, 230 Cal.Rptr. at 676 , 726 P.2d at 122 (summarizing testimony). Martin’s conclusions were thus merely cumulative of evidence presented at trial. See Hughes, 898 F.2d at 700-1 .

Nor has Rodriguez explained how Martin’s memory experiment could be considered more prejudicial to his case than to the State, since both presented evidence about the kind of car stopped next to the patrol car. We conclude that Rodriguez has failed to demonstrate any prejudicial effect-of Martin’s discussion of his inability to recall the makes and models of cars on the freeway.

B

As to the timing matter, both sides agree that it was misconduct for Martin to conduct his own inquiry and to present to other jurors extrajudicial evidence regarding the distance and travel time between Crockett and the scene of the murder. The only question before us is whether Rodriguez can establish that this misconduct had a substantial arid injurious effect or influence on the jury’s verdict. See Brecht, 507 U.S. at 637-39 , 113 S.Ct. at 1722 .

Rodriguez first argues that Martin was himself influenced by his extrajudicial timing experiment, and that his subsequent bias is sufficiently prejudicial to require habeas relief. It is true that the Sixth Amendment right to an impartial jury is violated by the presence of a single improperly influenced juror. See Lawson, 60 F.3d at 613 . However, in order to meet his burden on habeas review Rodriguez still must demonstrate that the extrajudicial information had a substantial and injurious effect on the jury’s verdict. See id. (“[R]eversible error commonly occurs where there is a direct and rational connection between the extrinsic material and a prejudicial jury conclusion, and where the misconduct relates directly to a material aspect of the ease.”) (quoting Marino, 812 F.2d at 506 ).

Rodriguez seeks to short-circuit this inquiry by arguing that we are bound by a factual determination of the California Court of Appeal that Martin was biased as a result of his timing experiment. See 28 U.S.C. § 2254 (d). He bases his argument on the following statement in the opinion of the California Court of Appeal: “When he was on the freeway he noticed a sign showing the mileage to Sacramento, and he calculated how long it would take to drive there. He determined that defendant could have been at the crime scene at the time the crime was committed.” Rodriguez asserts that we must grant habeas relief because this “finding” demonstrates as a factual matter that Martin’s own impartiality was destroyed.

We reject Rodriguez’s claim that the statement represents a factual finding that Martin was indeed prejudicially swayed by the extraneous information. First, it is clear from the statement’s context that the Court was simply reciting the content of Martin’s declaration, and was not making a binding factual determination of the prejudicial impact of the historical facts. Regardless, our inquiry can not permit such a subjective statement of Martin’s internal thought processes to control our inquiry into the objectively prejudicial effect of the extraneous information. See Dickson, 849 F.2d at 406 (reviewing courts may not investigate the subjective effects of any extrinsic evidence upon the jurors but must conduct only an objective inquiry) (gathering cases). We consider Martin’s statement as relevant only as an item of historical fact that he made the statement in his declaration, not as dispositive of our prejudice inquiry.

Furthermore, whether or not Martin’s impartiality was destroyed by his extrajudicial inquiry is a mixed question of law and fact which we review de novo. See Thompson, 74 F.3d at 1573 . We would therefore owe no deference to a state court finding on this question. We note too that the Court of Appeal concluded that the State had rebutted a presumption of prejudice fi*om the misconduct, which it could not have done had the Court factually determined that Martin was prejudiced by the extraneous information. The Court held that Martin’s misconduct was immaterial to the central question of arrival and departure times and therefore was “not of such a character as is likely to have influenced the verdict improperly.” We conclude that we are not bound by any factual finding of the State court that Martin’s impartiality was destroyed.

Next, Rodriguez argues that the very fact that Martin took it upon himself independently to inquire into the driving time between the two points indicates that he must have had a lingering doubt about the evidence on this point, and we may thus infer that Martin was unduly influenced by the extrajudicial information. See Gibson v. Clanon, 633 F.2d 851, 855 (9th Cir.1980) (“[T]he fact that at least [one juror] believed that it was necessary to obtain more evidence is, by itself, an indication that there may have been a need to resolve some lingering hesitation or uncertainty.”). To bolster this claim, Rodriguez even suggests that Martin deliberately took it upon himself to go to Crockett for the very purpose of resolving the question when jury deliberations were focused on the timing evidence.

Martin’s own declaration refutes this theory, since he states that he was in Crockett by happenstance and that he cannot recall whether the trip took place during the three months of the trial’s evidence phase or during deliberations. Nor did Martin report driving the entire distance between the two points, synchronizing his watch at various landmarks and analyzing the testimony of various witnesses, as Rodriguez suggests. Instead he drove to his own destination and merely noted the freeway sign indicating the additional distance and estimated travel time to Sacramento. This kind of “investigation” is so casual as to undermine the contention that Martin was compelled to resolve a lingering doubt in his own mind by seeking out additional information. Furthermore, since the travel time and distance were undisputed at trial it is not at all obvious how one would come to have a nagging doubt over this matter. In any case, not all instances of juror misconduct require reversal, see Marino, 812 F.2d at 505 , and there is no authority for the proposition that prejudice is presumed from the mere existence of juror misconduct. Instead, Rodriguez has the burden of demonstrating the substantial and injurious effect of the extraneous information on the jury’s verdict.

Finally, Rodriguez seeks to establish the material and prejudicial effect of Martin’s extrajudicial information by arguing that the jury was focused on the “time” issue when Martin discussed his own calculations. Since the time issue was material to the question of guilt, and Martin’s extraneous information related to time, Martin’s information was necessarily material to the jury’s verdict. However, as noted by the district court and the California appeal court, a trip’s duration cannot reveal an arrival time without reference to a time of departure. It was this departure time, rather than the trip’s duration, which was the focus of dispute at trial. At oral argument for this appeal, Rodriguez’s attorney repeatedly asserted that Martin calculated the travel time at an hour and a half and that the jury set the departure time at 2:00 a.m.: the jury therefore prejudicially concluded on the basis of Martin’s information that Rodriguez was at the crime scene at 3:30 a.m., the time of the murders.

This scenario is nothing more than a flight of fancy. By the time the defense interviewed them six years after the fact, none of the jurors could recall Martin’s conclusion about the travel time, nor is there any indication that the jury set the departure time at 2:00 a.m. Furthermore, calculating how long a 57 mile freeway trip would take is not so complex a matter that one would expect that a person of ordinary intelligence would inexplicably come up with a figure that was off by half. Rodriguez cannot meet his burden of showing prejudice by mere suppositions and “what ifs”, but must rather clearly demonstrate how he suffered a constitutional deprivation. “We are not called upon to isolate, examine, and negative each of the possible and irrational constructions on which a jury conceivably could rely.” Bagnariol, 665 F.2d at 888 .

Finally, had the jury actually concluded that the travel time took an hour and a half as Rodriguez claims, this conclusion would have been as damaging to the government’s case as it was to himself. If the couple left Crockett at 2:30 as the State argued, an hour and a half trip would have placed them at the murder scene half an hour after the crime was committed. We therefore find support for our conclusion that Rodriguez has not met his burden of demonstrating the prejudicial effect of Martin’s misconduct from the fact that both sides’ cases would be equally weakened by the introduction of any time calculation inconsistent with the one hour estimate agreed upon at trial.

We agree with the district court that Rodriguez has failed to demonstrate that Martin’s extraneous information was material or prejudicial to the question of the defendant’s guilt. Rodriguez’s claim that it had a substantial and injurious effect on the jury’s verdict must therefore fail.

Ill

Next, Rodriguez asks for a new trial because he alleges that Judge Karesh’s actions during deliberations deprived him of due process by coercing the jury into a eon viction. Whether a judge has coerced a jury’s verdict is a mixed question of law and fact which we review de novo. Jiminez v. Myers, 40 F.3d 976, 979 (9th Cir.1993). We consider whether the court’s actions were coercive under the totality of circumstances. Id. As the California Supreme Court on direct review included an exhaustive description of the events which occurred during deliberations, we will treat them only briefly here. See Rodriguez, 230 Cal.Rptr. at 687-88, 692-97 , 726 P.2d at 133-134, 138-143 .

Rodriguez points to three circumstances during deliberations which he contends establishes judicial coercion. First, the jury declared itself deadlocked on four occasions over the course of fifteen deliberation days, and yet the judge refused repeated defense requests to grant a mistrial. Secondly, in one of three supplemental instructions to the jury, Judge Karesh commented on evidence which the jury had requested. In this comment he remarked upon weaknesses in the testimony of one defense witness, while reminding the jurors that they were the sole fact finders and that they should keep in mind all of the evidence in considering their verdict. Finally, on the eleventh day of deliberations, Judge Karesh inquired into the numerical split of the jurors on the balloting. He learned that the vote had remained at 11-1 over five ballots, and he again denied a motion for a mistrial.

A

We agree with the district court that none of these actions taken in isolation requires a new trial. We leave the question of whether to declare a mistrial to the sound discretion of the trial court, which is in the best position to determine the likelihood that a jury will be able to reach a verdict. See United States v. Sommerstedt, 752 F.2d 1494, 1497-98 (9th Cir.), modified, 760 F.2d 999 (9th Cir.1985). Here, the jury declared itself to be deadlocked on four occasions, and yet when the judge offered supplementary instructions and ordered the jury to continue deliberations the jury continued to request additional instructions, new testimony to be read back, and “remained properly focused on the evidence.” Rodriguez, 230 Cal.Rptr. at 696 , 726 P.2d at 142 . There is no indication here that the judge’s decisions to require the jury to continue deliberations was an abuse of discretion.

Likewise, it is constitutionally permissible for a State trial judge to inquire into the numerical division of the jury, so long as the context of the inquiry does not suggest that the jurors were coerced into changing their views. See Locks v. Sumner, 703 F.2d 403, 406-7 (9th Cir.1983); Jiminez, 40 F.3d. at 981 . Here, the judge sent a questionnaire to the jury, which responded in writing with the numerical split. From the record it appears that the judge did not address the jury from that point until they returned a verdict four days later. Judge Karesh did not know whether the majority favored conviction, he did not know the identity of the hold-out juror, nor did he exhort the jury to reach a unanimous verdict, and the jury was dismissed for the day after responding to the court’s inquiry. See Locks, 703 F.2d at 407 (considering these factors). We conclude that it is not likely that the inquiry had a coercive impact on the jury.

While the practice is not without some dangers, a trial judge may also comment on the evidence so long as he makes it clear that it is the duty of the jury to determine the facts themselves. See Quercia v. United States, 289 U.S. 466, 469 , 53 S.Ct. 698, 699 , 77 L.Ed. 1321 (1933); see also, United States v. James, 576 F.2d 223, 228 (9th Cir.1978) (holding that where judge comments on the evidence the “bottom-line question ... is whether the judge has made it clear to the jury that all matters of fact are submitted to their determination”) (internal citation and quotation marks omitted). Here, the court repeatedly reminded members of the jury that they were to determine the facts of the case based on their own views of the evidence. We conclude that, given the judge’s cautionary admonitions regarding the jury’s duty, the comment was permissible.

Rodriguez also argues that the judge’s comment was unduly prejudicial because he called into question the testimony of a defense witness while failing to point out the weaknesses of prosecution witnesses. We disagree. A judge may call the juror’s attention to parts of the evidence when he believes it will assist the jury so long as he neither distorts the evidence nor adds to it. See Quercia, 289 U.S. at 469-70 , 53 S.Ct. at 699 (affirming that judge may draw jury’s attention to parts of the evidence he thinks important). Judge Karesh mentioned that all of the witnesses were testifying about the time of events within an estimated range, and that the estimates may or may not be determinative of the ultimate issues before the jury. See generally Rodriguez, 230 Cal. Rptr. at 694 , 726 P.2d at 140 (discussing the comment in the context of evidence given at trial). The comment did not distort the evi dence, nor was it so one-sided as to amount to an abuse of discretion. See United States v. Laurins, 857 F.2d 529, 537 (9th Cir.1988), cert. denied 492 U.S. 906 , 109 S.Ct. 3215 , 106 L.Ed.2d 565 (1989) (explaining that even on direct review a judge’s actions “justifies a new trial only if the record shows actual bias or leaves an abiding impression that the jury perceived an appearance of advocacy or partiality.”).

B

Rodriguez next contends that the combination of the following actions amounted to judicial coercion: (1) the judge inquired into the jury’s numerical split on the eleventh day of deliberations and knew that the vote had remained at 11-1 for each of the five ballots taken; (2) the jurors had declared themselves to be deadlocked on four occasions and; (3) the judge repeatedly required the jury to continue deliberations. Rodriguez argues that this combination of events amounted to impermissible judicial pressure on the holdout juror to surrender his minority view or remain forever a captive of the deliberations process.

In support, Rodriguez cites to Jiminez v. Myers, 40 F.3d 976 (9th Cir.1993). In Jiminez we held that under the totality of circumstances, a trial judge’s actions amounted to judicial coercion where the judge twice inquired into the jury’s numerical split after they had declared that they were deadlocked, twice noted that he was looking for movement toward unanimity, and refused to grant a mistrial where both the defense and prosecution agreed that further deliberation would not be useful. We held that under all the circumstances these actions were tantamount to the delivery of a flawed Allen charge. Id. at 980 .

An Allen charge is traditionally understood as an instruction to work towards unanimity by considering the views of others when a jury has reached an impasse in its deliberations. See United States v. Wills, 88 F.3d 704, 716 (9th Cir.), cert. denied, — U.S. -, 117 S.Ct. 499 , 136 L.Ed.2d 390 (1996). Since Rodriguez’s claim amounts to an allegation that Judge Karesh’s actions amounted to a flawed Allen charge, we note that we will not grant relief unless it is “clear from the record” that an Allen charge had an impermissibly coercive effect on the jury. See United States v. Lorenzo, 43 F.3d 1303, 1307 (9th Cir.1995); Wills, 88 F.3d at 717. An Allen charge is, on occasion, a legitimate and highly useful reminder to a jury to do its duty.

We found impermissible judicial coercion in Jiminez principally because the judge failed to remind the jurors of their duty not to surrender their sincerely held beliefs while at the same time making it clear that he wished them to return a unanimous verdict. Jiminez, 40 F.3d at 980-81 . The judge learned that over the two days of deliberation, the gap in the vote had continuously narrowed until there was only a single holdout juror.

Id.

The present case is clearly distinguishable from Jiminez . First, Jiminez involved a much simpler issue for the jury to resolve: whether or not the defendant acted with intent to kill when he fired a gun through a door. Both the prosecution and the defense conceded that this was the kind of case where further deliberations would not be helpful where deadlock occurred on the first day of deliberations and that a mistrial was in order. Id. at 979 .

In contrast, in Rodriguez’s trial the evidence phase spanned three months, the jury heard forty days of testimony from 124 witnesses, and about 250 exhibits had been entered. Rodriguez, 230 Cal.Rptr. at 687 , 726 P.2d at 133 . As the California Supreme Court explained:

Rodriguez, 230 Cal.Rptr. at 696 , 726 P.2d at 142 (internal citations and quotations omitted).

Additionally, our holding in Jiminez relied heavily on the fact that the judge never cautioned the jurors not to surrender their sincerely held beliefs under majoritarian pressure. Jiminez at 981 . In contrast, Judge Karesh so instructed the jurors on four separate occasions. In Jiminez , we noted that the judge commented approvingly on movement toward unanimity both times he sent the jury back to deliberate after inquiring into their numerical split. Judge Karesh, on the other hand, made no comment whatsoever after learning of the numerical split and instead sent the jury home for the night. The jury deliberated for four more court days before returning a verdict, called for the reading of new testimony, and remained properly focused on the evidence.” As we have previously noted, the fact that the jury is calling for the re-reading of testimony is a “pretty good indication” that jurors are probably not surrendering their sincere views because of judicial pressure, but are instead moving forward with deliberations. See United States v. Ajiboye, 961 F.2d 892, 894 (9th Cir.1992).

Nor are any of the other indicia of coercion which we have previously relied upon present here: the judge did not know whether the majority favored conviction or acquittal; nor did he know the identity of the hold-out juror; and the four day lapse of time between the judge’s last contact with the jury and their return of a verdict suggests that the jury did not simply succumb to any pressure for a unanimous verdict. See, e.g., Ajiboye, 961 F.2d at 894 ; Wills, 88 F.3d at 717-718.

Finding that the record before us does not clearly establish that Judge Karesh’s actions resulted in a coerced guilty verdict, we reject this claim for relief.

IV

The decision of the district court is AFFIRMED.

. The prosecution noted that in order to arrive in Vacaville from San Francisco (about a 50 mile distance) in half an hour as she claimed, Hawthorne would have had to drive at a sustained speed of 100 miles an hour. Hawthorne testified that she was careful not to speed during her drive, as she had received a speeding ticket earlier that evening. . According to the California Supreme Court, the following testimony was presented: One witness claimed to have seen a light colored Ford Galaxie, two others claimed to have seen a light colored Ford Fairlane, and another saw a white Nova stopped next to the patrol car around the time of the murders. These four witnesses were out-numbered by others who testified that the patrol car was parked next to various vehicles of entirely different descriptions. People v. Rodriguez, 42 Cal.3d 730 , 230 Cal.Rptr. 667, 676 , 726 P.2d 113, 122 (1986). . By the time the jurors were contacted six years after the fact, none of them could recall Martin’s conclusions regarding the time and distance calculations. Rodriguez’s attorney argued that the time and distance were "necessarily incorrect” because Martin began his trip from the bar rather than from Engel’s house. However, it is undisputed that the bar is located about three or four blocks from Engel’s home, making any error immaterial. At oral argument, Rodriguez’s attorney raised a new factual question: whether Martin calculated the driving time to Sacramento or West Sacramento, and what difference this distinction might make. In the ten years that Rodriguez has litigated over the issue of Martin’s misconduct, he never raised this factual question in the State court, nor before the magistrate, the district court, or in his briefs before this court, with the result that the State has had no opportunity to respond. We decline to consider this new issue raised for the first time at oral argument. See Campbell v. Kincheloe, 829 F.2d 1453 , 1456 n. 1 (9th Cir. 1987) cert. denied, 488 U.S. 948 , 109 S.Ct. 380 , 102 L.Ed.2d 369 (1988) (refusing to consider a new claim raised for the first time at the oral argument in a habeas proceeding); see also United States v. Sitton, 968 F.2d 947, 959 (9th Cir.1992), cert. denied, 506 U.S. 979 , 113 S.Ct. 478 , 121 L.Ed.2d 384 . . In supplemental briefing, the parties addressed the impact of our en banc court's decision in Jeffries v. Wood, 114 F.3d 1484 (1997) on this question. We reject Rodriguez's contention that Jeffries establishes a rule that any and all incidents of internal juror misconduct create a strong presumption of prejudice. We made clear in Jeffries that our traditional inquiry into the prejudicial effect of the extrajudicial information remains at the heart of our inquiry into whether habeas relief is warranted on the facts of a particular case. See Jeffries, 114 F.3d at 1490-1492 . . Rodriguez argues that on habeas review we must accept as presumptively correct the state trial court’s assertion that "Martin personally discounted the testimony of the ‘passers-by’ as a result of the [car recollection] experiment.” Rodriguez is incorrect because such a conclusion is not a factual finding, and is rather a determination of the prejudicial effect of the misconduct which is not owed deference on collateral review. Thompson, 74 F.3d at 1573 . Furthermore, the state Court of Appeal, basing its conclusions on the same paper record as the trial court had before it, explicitly rejected the trial court’s determination that the juror misconduct resulted in prejudice to Rodriguez. . The California courts next sentence reads: "Martin’s affidavit goes on to state that on the same trip he noticed that he could not accurately observe or recall the color and make of cars ...” . For purposes of comparison, Martin's own declaration reads: "I was able to calculate the time it would have taken Rodriguez to drive to Sacramento. I determined that he could, in fact, have been at the crime scene at the time the crime was committed." . Both sides agreed that it took about an hour to drive the 57 miles on the freeway. . To the extent that Rodriguez suggests that Martin or the jury may have determined arrival time based only on the trip’s duration, we note that relief will not be granted if the only connection between the extraneous information and a prejudicial conclusion relies on an irrational leap. See Bagnariol, 665 F.2d at 888 (rejecting a claim of prejudice from juror misconduct where such a conclusion “would require an assumption that the jury members reached an irrational conclusion, lacking in common sense.”). .The government argued that Rodriguez left Crockett at 2:30, and Rodriguez contended that he left at 1:30. . The California Supreme Court includes a lengthy description of the events during deliberations. People v. Rodriguez, 42 Cal.3d 730 , 230 Cal.Rptr. 667, 687-88 , 726 P.2d 113, 133-34 (1986). The following table is based on that summary. [[Image here]] . The comment which Rodriguez complains of is reproduced in full in the California Supreme Court's opinion. Rodriguez, 230 Cal.Rptr. 667 , 726 P.2d at 139 n. 10. In brief, Judge Karesh in the course of giving a supplemental instruction noted that there had been conflicting evidence given regarding the time that Rodriguez left Crockett, and he reminded the jury that most of the witnesses were giving only estimates of a time range and that they should keep in mind all of the evidence bearing on the ultimate issues. He went on to add: As you have singled out the testimony of Wanda Hawthorne for re-reading twice, you may wish to decide whether her testimony that she-left San Francisco at 1:30 a.m., the same time that the defendant said he left Crockett, and that the Camaro came from behind her is accurate. As she said, she was more sure of her time of leaving San Francisco than her arrival time, or whether her estimates of seeing a Camaro near Vacaville at about 2:00 a.m. and of arriving home at 4:00 a.m. are more accurate. You should also consider whether or not you believe from the evidence that the car that Wanda Hawthorne saw was the defendant’s car, and if so, the effect, if any, of her testimony that she did not see the car again. Now, again, you are the exclusive judges of the facts. My comments have been intended to be advisory only and are not binding on you as you are the exclusive judges of the questions of fact submitted to you and of the credibility of the witnesses. You should disregard any and all of my comments to you if they do not agree with your views of the evidence and the credibility of the witnesses. I also remind you again that both the people and the defendant are entitled to the individual opinion of each juror. It is the duty of each of you to consider the evidence for the purpose of arriving at a verdict if you can do so. Each of you must decide the case for yourself.... Rodriguez, 230 Cal.Rptr. at 692 n. 10, 726 P.2d at 139 n. 10.