John M. Branion, Jr. v. Richard B. Gramly

855 F.2d 1256

• Court: Court of Appeals for the Seventh Circuit
• Decided: September 21, 1988
• Docket: 87-3052, 87-3053
• Status: Published

Opinion

EASTERBROOK, Circuit Judge.

Donna Branion died on December 22, 1967. She was strangled and shot at least four times. She was not molested; there were no signs of forced entry into the apartment, from which nothing was stolen. This led the police to doubt that a stranger was responsible. A jury concluded that Donna’s husband, John M. Branion, Jr., did the deed. The evidence was circumstantial, but what circumstances!

• Branion called the police after finding his wife sprawled in a pool of blood. Although a physician, he did nothing to investigate her condition or assist her. He told the police that he knew from the lividity of Donna’s legs that she was dead. A pathologist testified that Donna Branion’s legs did not display lividity.

• Ballistics experts determined, from the rifling of the slugs and marks on the casings, that the murder weapon was a 9mm, .38 caliber Walther PPK, a rare gun. John Branion, a gun collector, owned a 9mm, .38 caliber Walther PPK. When the police asked whether Branion had a 9mm weapon, he said yes and gave the police a Luger. Later the police asked whether he had a .38 caliber weapon; he said yes, one, and turned over a Hi Standard pistol. He did not mention his Walther PPK, which was never found — yet could not have been stolen by an intruder on December 22, for the family’s weapons cabinet was locked when the police arrived.

• The cabinet contained a clip, target, and brochure for Branion’s Walther PPK together with two boxes of .38 caliber ammunition. One box was full. The other was short four shells. Four shell casings were found near Donna Branion’s body.

• Branion had a mistress, a nurse at the hospital where he worked, and the Branion home was not a model of domestic tranquility. Branion married his mistress shortly after his wife’s death.

The defense denied some of this and tried to cast the rest in a better light, but the jury was entitled to believe the prosecution’s evidence. Jackson v. Virginia, 443 U.S. 307, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979).

The death and trial attracted attention, common for family murders among the well-to-do and intensified by the nature of the defense in this one: impossibility. Dr. Branion held a position of responsibility at the Ida Mae Scott Hospital, in the Hyde Park district of Chicago. The defense was that he had been at the Hospital attending patients until 11:30 a.m. and that, on the way home, he made two stops — one to pick up his son Joby at a private school, the other to meet a friend with whom the Bran-ions planned to lunch. The police logged the call informing them of the murder at 11:57 a.m. There was not enough time to drive from the Hospital to his home, with two stops, and to kill his wife before calling the police, Branion insists — especially because Donna was strangled as well as shot, and the bruises on her neck took at least 15 minutes to form.

A claim of impossibility is not the only unusual feature of this case. Both sides of the family were prominent. Dr. Branion marched with Dr. Martin Luther King, Jr., in the 1960s. His father was a well-known attorney and the deputy chief public defender of Chicago. Donna Brown Bran-ion’s father was a wealthy banker, and other members of the family have been successful in other fields. Reginald Hol-zer, who presided over the trial, may have tried to induce Branion’s many friends to pay him off in exchange for a judgment notwithstanding the verdict — a disposition that in Illinois is not open to appellate review. Holzer has since been convicted of extortion in many other cases. See United States v. Holzer, 816 F.2d 304 (7th Cir.), vacated, — U.S. -, 108 S.Ct. 53, 98 L.Ed.2d 18 (1987), after remand, 840 F.2d 1343 (7th Cir.), cert. denied, — U.S. -, 108 S.Ct. 2022, 100 L.Ed.2d 608 (1988). On getting wind that Holzer had dictated to the court reporter an opinion absolving Branion, the prosecutor, Patrick A. Tuite— now a prominent criminal defense attorney —paid a private call on the judge. During this ex parte conversation Tuite insisted that Judge Holzer let the case proceed through appellate channels, but he left thinking that Branion would prevail.

Holzer later called Tuite at home, insisting that Tuite move for a week’s postpone ment of the impending ruling on Branion’s motion for a new trial. Tuite complied, and Holzer put on a little show, dressing down the prosecutor for begging for more time and cutting the extension back to two days. The theatrics (and, perhaps, the last effort to extract cash from Branion’s friends) over, Holzer denied the motion on condition that Branion remain free pending appeal. The Supreme Court of Illinois affirmed the conviction, People v. Branion, 47 Ill.2d 70, 265 N.E.2d 1 (1970), cert. denied, 403 U.S. 907, 91 S.Ct. 2213, 29 L.Ed.2d 683 (1971).

When the conviction became final Bran-ion neither reported to prison nor sought collateral review in state court. Instead he fled to Africa where, after a short stay in the Sudan, he journeyed to Uganda and became Idi Amin’s personal physician during 1972-79. New York Times, Oct. 15, 1983, § 1 p. 6 col. 2; Associated Press dispatch October 14, 1983. Escaping the ravages of Uganda’s civil wars and invasions, 1 Branion was unceremoniously put on a plane by a new regime and shipped back to the United States in October 1983. (Uganda has no extradition treaty with the United States.) Since his sudden return Branion has been serving his sentence of 20-30 years’ imprisonment. In 1986 he filed a petition for a writ of habeas corpus, 28 U.S.C. § 2254. The state has not argued that Branion’s flight and the consequent delay disqualifies him under Rule 9(a) of the Rules for Section 2254 cases.

I

We must first address a difficulty in our (and the district court’s) jurisdiction. The case was assigned to Judge Getzendanner, who issued three opinions. The first, reported at 664 F.Supp. 1149 (N.D.Ill.1987), held that Branion had exhausted all state remedies — exhausted them in the sense that his flight to Uganda coupled with his neglect to preserve points on direct appeal would lead Illinois to reject any effort to seek relief now. The second, 1987 U.S. Dist. LEXIS 8767 [available on WEST-LAW, 1987 WL 27], rejected two of Bran-ion’s four claims on the merits and held the others barred by the doctrine of Wainwright v. Sykes, 433 U.S. 72, 97 S.Ct. 2497, 53 L.Ed.2d 594 (1977). Judgment was entered in the state’s favor on August 27, 1987. Branion filed a timely motion for reconsideration under Fed.R.Civ.P. 59. Judge Getzendanner wrote a third opinion declining to reopen the subject and adding some reasons why the state was entitled to prevail. The order denying the Rule 59 motion was filed on September 22, 1987. Eight days later Judge Getzendanner resigned from the bench.

On October 20, 1987, Branion filed a notice of appeal from this decision. The clerk of the district court neglected to forward it to us until December 14. Meanwhile, on October 15, Branion had filed a second post-judgment motion, this time under Rule 52(b). The motion was assigned to Judge Plunkett, who neither knew about the notice of appeal nor recognized that this was a successive post-judgment motion. Judge Plunkett took briefs and evidence, heard argument, and wrote an opinion [available on WESTLAW, 1987 WL 20414], 1987 U.S. Dist. LEXIS 10899, denying Branion’s petition for the third time. Judge Plunkett entered a judgment on November 20, and Branion filed a second notice of appeal on November 25. This prompted the clerk’s office to transmit the first notice; the two were assigned sequential numbers in this court.

The appeal filed on October 20, No. 87-3052, transferred the case to us. Nothing that happened after that in the district court matters. Judge Plunkett would have lacked jurisdiction even had there been no appeal in October, for the losing party may file post-judgment motions under Rules 52 and 59 only within ten days. See Fed.R. Civ.P. 6(b). A district court may not act on an untimely motion. Bailey v. Sharp, 782 F.2d 1366 (7th Cir.1986). If a judge alters her judgment in response to the first motion, the party aggrieved by the change may file a motion directed to the difference. Charles v. Daley, 799 F.2d 343, 347- 48 (7th Cir.1986). But Judge Getzendanner reaffirmed rather than revised her judgment. This Rule 52 motion would have been untimely even if Judge Getzendanner had changed her disposition, for it was filed 23 days after the judgment became final, and the Rule allows only ten. Since weekends and holidays are excluded in the computation, Fed.R.App.P. 26(a), Branion would have had until October 6, 1987, to serve any further motion that might have been in order; he took until October 15. At oral argument Branion’s lawyer contended that the motion was served within ten days after Branion received Judge Get-zendanner’s order, but this ignores the language of Rule 52(b): “Upon motion of a party i,.ot later than 10 days after entry of judgment the court may amend its findings or make additional findings and may amend the judgment accordingly.” Branion is represented by a wealth of talent, which we would not have expected to have had so much difficulty with the civil rules. 2 The case has been in this court since October 20, 1987, and Judge Plunkett had no authority to act. We vacate the opinion and judgment of November 20 and dismiss appeal No. 87-3053 for want of jurisdiction, because it was filed more than 30 days after the judgment became final on September 22. See Browder v. Director, Department of Corrections, 434 U.S. 257, 98 S.Ct. 556, 54 L.Ed.2d 521 (1978).

One more procedural matter. The appendix Branion filed in this court contains a good deal of material that was not before the district court. An entire section of Branion’s brief is devoted to arguments based on this material, which includes an affidavit from Branion telling his side of the story — for the first time in 20 years, since he did not testify at trial or in the district court — as well as affidavits from other witnesses, a report of a physician describing the nature of the wounds in Donna Branion's body in an effort to show that more than four shots must have been fired, notes from an interview with the prosecutor reflecting his assessment of the evidence and his disagreement with Judge Getzendanner’s analysis of the facts, and an affidavit from his current counsel narrating his own investigation. The state has filed a motion to strike this “evidence”. See Heirens v. Mizell, 729 F.2d 449, 455-56 (7th Cir.1984); United States v. Burke, 781 F.2d 1234, 1245-46 (7th Cir.1985). Branion replies that he is entitled to draw exculpatory information to the court’s attention at any time in light of Kuhlmann v. Wilson, 477 U.S. 436, 452-55, 106 S.Ct. 2616, 2626-27, 91 L.Ed.2d 364 (1986), which holds that in order to succeed on a successive petition for a writ of habeas corpus, the prisoner must make a colorable showing of innocence. Cf. Henry J. Friendly, Is Innocence Irrelevant? Collateral Attack on Criminal Judgments, 38 U.Chi.L.Rev. 142 (1970). Circuit Rule 10, like Fed.R.App.P. 10(a), excludes from the appendix any materials that were not on file in the district court, because “[otherwise they deprive the opposing party of an opportunity to comment on them and the district judge of an opportunity to evaluate their significance.” Henn v. National Geographic Society, 819 F.2d 824, 831 (7th Cir.1987). Branion replies:

Rule [10] is expansive in its terms and designed to facilitate the appellate process. There is certainly nothing in Rule [10] that says an exhibit included in an Appendix pursuant to a Supreme Court directive would violate Rule [10].

The requirement of Kuhlmann is thus transformed into a “directive” that the petitioner file untimely, non-record materials.

This passage reflects a set of mind, evident in Branion’s treatment of the jurisdictional questions as well: when innocence is at stake, no procedural rules should get in the way. We do not doubt the value of protecting the innocent. That is the principal reason why we have an elaborate criminal procedure rather than kangaroo courts. But nothing in Kuhlmann, which is designed to limit the scope of federal collateral relief, is either a license or a requirement to ignore procedures the judicial system has found important to the truth-seeking process. If we were sure someone is innocent, perhaps we should dispense with procedural folderol; if we were sure of guilt, a trial also would be a waste of time. Judges, however, are professional sceptics. If we entertain, as we ought, doubt about the facts, procedures are invaluable. Bran-ion’s lawyers may be sure he is innocent, but their confidence is not infectious. Materials significant to Branion’s guilt or in-' nocence should be presented to the district court, there to be tested in the usual adversarial ways. They may not be stored- up for an appeal or generated (as some of these documents were) expressly for the appellate panel; this court is in no position to cope with factual disputes. If Branion wanted to refer to these materials, he should have filed them in timely fashion in the district court; if that were impossible, he could have asked for enlargement of the record under Rule 10 or requested this court to remand the case to the district judge for further proceedings. Instead of doing either, he tried to smuggle the 'materials into the appendix and brief. That is not acceptable advocacy. We grant the state’s motion and strike exhibits K through Q of Branion’s appendix, and Argument III of Branion’s brief. Because exhibit J of the appendix was filed with Judge Plunkett, we strike it too. No document filed after mid-September 1987 is before us.

II

We start with Branion’s contention that rational jurors, taking the evidence in the light most favorable to the state, could not have found beyond a reasonable doubt that he murdered his wife. Jackson v. Virginia, 443 U.S. at 324, 99 S.Ct. at 2791. This question was pursued vigorously in staté court. The Supreme Court of Illinois held that the jury rationally could have concluded that Branion killed his wife with four bullets from his Walther PPK, which he disposed of before calling the police. Branion failed to investigate his wife’s prone body because, having murdered her, he knew her condition; and his lie to the police about owning only one .38 caliber weapon was as good as a confession.

With respect to the defense that traveling the narrow, icy streets of Hyde Park on December 22 (it was about 10° F. and snowing lightly) would have taken too long, the court replied that the evidence conflicted: Branion told the police that he left the Hospital at 11:30 a.m. and picked up his son Joby at 11:35. Joyce Kelly, a teacher, said that Branion appeared between 11:45 and 11:50 to fetch his boy. 47 Ill.2d at 72, 265 N.E.2d 1. The implication: Branion went from the Hospital to his apartment, killed his wife and resumed his journey, showing up at the school as late as 11:50. He then stopped to see Maxine Brown, with whom he had made a luncheon date between 10:00 and 11:00 p.m. on December 21; Brown cancelled the engagement. Branion returned home and called the police. The jury would have been entitled to infer— from the timing of the request and the fact that this was the first time he had asked her to lunch — that Branion wanted Brown to be present as a witness to his “discovery” of the body. Judge Getzendanner had a different theory: Branion left the Hospital before 11:30 and killed his wife (a neighbor reported hearing a commotion in the Branion apartment before 11:30), returned to the Hospital to establish his presence away from the scene of the crime, and only then picked up his son.

Branion insists that these explanations are physically impossible. He could not have left the Hospital before 11:30 because time notations on patients’ records show that he met 13 spaced through the morning, the last at 11:28. On the way out he saw an emergency patient, consuming a few minutes. Branion contends that he could not have reached his car before 11:34 a.m. From there it would have taken four to six minutes to drive to his son’s nursery-school. Police officers repeatedly drove in 1968 the route Branion says he took, never exceeding the speed limit; they logged times between six and twelve minutes for the circuit, recording 4 to 5VÍ2 minutes for the stretch from the Hospital at 5027 South Prairie Street to the school at 5470 South Kenwood Avenue. 3 After finding a parking place, Branion could not have entered the school until 11:40. Now adopting Kelly’s testimony that he picked up Joby between 11:45 and 11:50, Branion submits that he left the school at about 11:45. From the school to 1369 East 53d Street, where Maxine Brown was employed, would have taken 1-2 minutes. The conversation with Brown, at which she begged off from the luncheon date, added another minute or two. At 11:49 Branion set off for home at 5054 South Woodlawn. He could not have arrived before 11:51; parking would have taken additional time. 4 There were only six minutes left before the call to the police at ll:57-not enough, Branion says, to kill his wife, clean his hands, and dispose of the gun. Not when part of that time was spent yelling to the next-door neighbor.

Especially not when Donna Branion was strangled as well as shot. Her neck was bruised; she had been garrotted with the cord from her iron. The pressure was not great enough to break any bones. Defense counsel asked the pathologist: “Would you tell me how long would be a long period of time to make a bruise like that if the pressure weren’t great?”, to which the pathologist replied: “I would say fifteen minutes to a half hour.” Neither side pursued the matter further at trial. Branion now insists that the pathologist’s testimony means that pressure was applied to his wife’s neck continuously for 15-30 minutes, which was impossible given the time sequence; the state believes that the testimony means only that a bruise would have taken 15-30 minutes to form after the application of pressure, and that the pathologist was in any event throwing off a rough estimate rather than giving the physical limits.

Time can be squeezed out of Branion’s account quite easily. Suppose the clock at the Hospital was a few minutes fast (digital watches were rare in 1967) — or suppose Dr. Branion entered the time on patients’ records just a little inaccurately. He could have started his journey at 11:30 if not a few minutes earlier. Suppose he was not punctilious about finding legal parking spaces. Then the time for the whole trip is only a minute or two longer than the minimum driving time: at the legal speed, six minutes. Add one minute to grab Joby and one minute to meet Maxine Brown and you still have only 8 minutes. If Branion started a little before 11:30, there is enough time left over to kill someone — even to detour home first, as the prosecution contended. As for the bruises: Branion, a physician, knew how to inflict this kind of injury in minimum time. The jury could have concluded that Branion strangled his wife with a cord briefly at 11:40 and then shot her, and that bruises continued forming until the police arrived at noon. If Branion killed his wife, he planned to kill her; the assumption of planning helps to evaluate the time sequence. A physician planning murder would have inflicted a kind of injury that looks like it took longer to cause than it did. He would have fudged the time records at the Hospital, exceeded the speed limit and parked illegally to save minutes; he would have whisked Joby out of school and barely nodded at Maxine Brown. He would not have come to a full stop at stop signs. The entire sequence can be thrown off on the assumption of planning, an assumption the jury was entitled to indulge.

Branion disagrees, saying that any alteration of the time sequence would require the jury to disbelieve the nurse who said that she saw Branion at the Hospital after 11:30 and the teacher who said she saw Branion at the school between 11:45 and 11:50. The jury did not need to question the' teacher’s times if Branion killed his wife earlier. At all events, the jury was entitled to disbelieve anyone it pleased. Most persons’ sense of time is not so acute that they can distinguish 11:25 a.m. from 11:35 a.m. with perfect accuracy a year after the event — especially when, at 11:25 a.m. on December 22, 1967, the nurse had no reason to note the time precisely. Permitting such disbelief would make a shambles of Jackson, Branion insists, because it would relieve the state of its obligation to prove the crime. Through selective disbelief, the jury could create the state’s case out of thin air. Not so. A trier of fact may disbelieve or reject any evidence. Disbelief is not evidence of the opposite of the thing discredited, however. Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 256-57, 106 S.Ct. 2505, 2514-15, 91 L.Ed.2d 202 (1986); LHLC Corp. v. Cluett, Peabody & Co., 842 F.2d 928, 934-36 (7th Cir.1988). A jury that disbelieved everything would be compelled to acquit the defendant. Selec-five disbelief, however, is an ordinary incident of trial; a jury could credit the objective evidence (the ballistics and the proof that Branion owned a Walther PPK) while doubting the ability of witnesses to pin Branion’s movements down to the minute.

Even if this were so, Branion insists, it would still be exceedingly improbable that he could have killed his spouse — so unlikely that no sane person could find guilt beyond a reasonable doubt. Branion starts from the ranges of time offered for driving (6 to 12 minutes) and forming bruises (15 to 30 minutes) and submits that the likely times were in the middle of those ranges. A time at the low end of each range — which Bran-ion thinks necessary to make the crime possible — was correspondingly unlikely: a probability of “less than 0.01” in each case, Branion says. The probability that the low end of each range would occur back-to-back is 0.01 x 0.01 = 0.0001. Proof beyond a reasonable doubt means a probability much greater than the 0.51 more-likely-than-not standard; 5 a probability of 0.0001 that the accused did it is so far away from “beyond a reasonable doubt” that the federal court must issue the writ. We attach, as an appendix, Branion’s full argument (omitting footnotes).

We shall not become mired in the debate, set off by People v. Collins, 68 Cal.2d 319, 68 Cal.Rptr. 497, 438 P.2d 33 (1968), about the proper use of statistical inference in criminal litigation. 6 Statistical methods, properly employed, have substantial value. Much of the evidence we think of as most reliable is just a compendium of statistical inferences. Take fingerprints. The first serious analysis of fingerprints was conducted by Sir Francis Galton, one of the pioneering statisticians, and his demonstration that fingerprints are unique depends entirely on statistical methods. See Gal-ton, Finger Prints (1892). See also Stephen M. Stigler, The History of Statistics: The Measurement of Uncertainty before 1900 297-98 (1986). Proof based on genetic markers (critical in rape and paternity litigation) is useful though altogether statistical, United States v. Green, 680 F.2d 520, 523 (7th Cir.1982). So too evidence that, for example, the defendant's hair matched hair found at the scene of the crime. United States ex rel. DiGiacomo v. Franzen, 680 F.2d 515, 517-19 (7th Cir.1982). None of these techniques leads to inaccurate verdicts or calls into question the ability of the jury to make an independent decision. Nothing about the nature of litigation in general, or the criminal process in particular, makes anathema of additional information, whether or not that knowledge has numbers attached. After all, even eyewitnesses are testifying only to probabilities (though they obscure the methods by which they generate those probabilities)— often rather lower probabilities than statistical work insists on.

The lesson of Collins is not that statistical methods are suspect but that people must be sure of what they are looking for, and how they can prove it, before they start fooling with algebra. The crime had been committed by a blonde woman with a ponytail and a bearded black man with a mustache, driving a yellow car. The defendants had these characteristics. An “expert” determined the probability that any particular person is black, blonde, bearded, etc., and multiplied these, coming up with the probability of one in 12 million that any randomly selected persons would have matched the characteristics of the perpetrators. That, the prosecution maintained, showed that the defendants were the culprits: how improbable that there should be two such couples. But all these figures showed is that the defendants had not been randomly selected! The state scrounged up two persons who matched the description of the offenders; the expert “proved” that if the police had arrested people at random they would not have nabbed two such persons, but not that the persons actually arrested had anything to do with the offense. The figures did not even show that such couples are scarce; every person has some distinguishing, uncommon characteristics, and the likelihood of their occurrence can be multiplied to yield some very small number. Any two numbers less than one, when multiplied, give a still smaller number. So it is easy to provide the kind of “proof” in Collins for every person and crime, without doing anything to help figure out whether the state caught the offender. This error should put us in mind of an example suggested by Condorcet: a lottery may have a million possible number combinations, so when the proprietor of the lottery announces that the winning combination is such-and-such, the probability that this is indeed the winning combination is one in a million. Conclusion: the proprietor is lying. See Isaac Todhunter, A History of the Mathematical Theory of Probability from the Time of Pascal to That of Laplace 400 (1865).

Every event, if specified in detail, is extremely improbable; indeed, with enough detail it is unique in the history of the universe. It is always possible to take some probabilities, small to start with, and multiply them for effect. In order to avoid the errors produced by mindless multiplication, the statistician must specify with care what we should expect to find if the event in which we are interested has occurred, and what we should expect to find if it has not. Next we must develop criteria that might differentiate the one from the other. Only then can we begin to assess probabilities. This can be a daunting task. See Mister v. Illinois Central Gulf R.R., 832 F.2d 1427 (7th Cir.1987). It is not, however, an impossible task.

Branion’s lawyers did not attempt it. They simply multiplied two small numbers to get a smaller one, without describing why these were plausible numbers or why we ought to multiply them. Imagine a defendant seeking a writ of habeas corpus with this argument: “Twenty witnesses testified for me, and although the probability that any one of them was lying may have been 0.9, the probability that every one of them was lying was (0.9)20, or 0.12, so I must be innocent.” Such a maneuver, like the one in Collins, assumes that the events are independent, when they are not. This is not precisely the error committed here, but it is close. The first task is to specify the minimum time Branion needed, not the minimum possible time. The jury could have found that 30 minutes lapsed between Branion’s leaving the Hospital and his call to the police. We therefore should like to know the probability that the combination of travel and murder times came to 30 minutes or less; Branion has offered only his view of the probability that the time came to 21 minutes or less. He produced even this figure by a method that is proper only if the probability of the 6-min-ute drive and the 15-minute bruise are independent events. Yet on the state’s hypothesis of a planned murder, they are anything but independent.

That’s not all. Branion assumed that the distribution of driving and bruise-forming times is Gaussian (that is, characterized by a normal bell-shaped curve centered on the mean of the distribution). He derived the probability of a six-minute drive beginning with the calculation of a mean of 9 minutes, from the average of the extreme times (6 and 12). He then added the assumption that the standard deviation is one minute, leading to the conclusion that a travel time of six minutes, three standard deviations from the mean, will happen less than one time in a hundred. Where did this mean and standard deviation come from? The range of 6 to 12 minutes is from a series of six trials in 1968. We know that a driving time of 6 minutes was achieved at least once in six runs, not (as Branion’s calculation implies) once in a hundred. For all this record reveals, the six trials came out with times of 6, 6, 7, 7, 8, and 12 minutes. A calculation based on a mean driving time of nine minutes, with a standard deviation of one minute, would produce a bizarre conclusion. Nothing suggests a Gaussian distribution or the absence of skewness. As for the pathologist’s range of 15 to 30 minutes for bruise formation: this was a number from the air, and we have no idea what the mean time or standard deviation might be.

That’s still not all. Even if the time sequences are independent, even if we are interested in the probability that the driving plus choking time is 21 minutes or less, even if the distributions are Gaussian, the probability is very sensitive to the assumed standard deviation. On Branion’s assumptions, the probability is 0.1% rather than 0.01% as Branion believes; on more plausible assumptions, the probability is 10%. The proof appears in the margin. 7 Since this probability reflects a series of assumptions generally favorable to Branion (though implausible individually and collectively), the statistical argument that the jury was compelled to find him innocent collapses.

It would be nonsense to pretend that this is a simple case or that Branion’s guilt is obvious. There is no direct evidence of guilt — no powder residue on Branion’s hands or blood under his fingernails, for example. Doubtless a physician knows how to clean his hands thoroughly and quickly, yet the time is tight even on the state’s theory. The two district judges who looked at this record came up with different, and inconsistent, theories of how and when Branion committed the crime, if he did. Prosecutor Tuite seems to believe today that Branion hired an assassin and may not have fired the fatal shots. Bran-ion submits that more than four bullets were used, which, if true, undercuts one of the inferences used to convict him. Yet even “direct” evidence does not ensure guilt; eyewitness testimony is often unreliable. See Elizabeth F. Loftus, Eyewitness Testimony: Psychological Research and Legal Thought, 3 Crime and Justice: An Annual Review of Research 105 (1981). A federal court reviewing the record of a state trial cannot ensure accurate verdicts; no human agency can do that, and a tribunal reviewing a cold record 20 years after the events is likely to err more frequently than a jury that had a chance to listen to contemporaneous evidence. All we can ask —under Jackson all we may ask — is whether, assuming the jury resolved all disputes in the state’s favor and drew all inferences from that evidence, it would have been rational to convict. Not whether we would convict, but whether thoughtful people could convict. Careful, rational jurors could have believed that the evidence about the Walther PPK and Branion’s failure to approach his wife’s body, coupled with his lie to the police about the lividity of his wife’s legs, swamped any doubts about the time sequence. The evidence in this case is sufficient under Jackson.

Ill

Prosecutor Tuite implored Judge Holzer, in an ex parte conversation, to en ter judgment on the jury’s verdict, so that the final decision about the sufficiency of the evidence would be made by the Supreme Court of Illinois. The judge ultimately agreed — whether because he thought this the right course or because he was foiled in extorting money from Bran-ion’s friends we may never know. Tuite allows that he suspects, and one of Bran-ion’s friends insists, that money changed hands. 8

Ex parte contacts of the sort to which Tuite admits are forbidden. We were dismayed to learn that an experienced prosecutor, now a respected private attorney, would drop in on a judge to hold a private conversation about the merits of a case— especially when, the prosecutor believed, he had received confidential information about the judge’s impending decision. We assume, as other courts have held, that such lobbying violates the Due Process Clause of the fourteenth amendment. Jones v. Hess, 681 F.2d 688, 692 (10th Cir.1982); Haller v. Robbins, 409 F.2d 857, 859-61 (1st Cir.1969). Cf. United States v. Napue, 834 F.2d 1311, 1316 (7th Cir.1987); United States v. Solomon, 422 F.2d 1110, 1119-21 (7th Cir.1970). Condemning ex parte contacts and issuing a writ of habeas corpus are different matters, however. No objection to this exchange has been presented to the state courts. Judge Get-zendanner concluded, and the state effectively concedes, that state courts are closed to Branion on this (and every other) subject, so that he has no state remedies to exhaust. She also concluded that Branion had not raised this matter in a timely fashion in state court, so that it may be presented in federal court only if Branion can establish both “cause” and “prejudice” under Wainwright v. Sykes. She assumed that Branion could establish cause but held that he had not established prejudice. We agree with both the approach and the conclusion.

Whether Branion could establish “cause” is a close question, turning on facts that have not been fully aired. He says that, not until 1986, when Tuite made out an affidavit, was he aware of the ex parte contacts or the judge’s attempt at extortion. An event unknown and unknowable because- of the state’s machinations is “cause”. Amadeo v. Zant, — U.S. -, 108 S.Ct. 1771, 100 L.Ed.2d 249 (1988). But an affidavit filed by one of Branion’s friends says that Nelson Brown, a member of the defense camp, knew in 1968 that Holzer wanted money and paid him $10,-000; and that Holzer begged off from acquitting Branion (releasing him instead on trivial bond, with no travel restrictions) because the prosecutor had found out about the plan. See note 8 above. If that is so, then Branion either knew or readily could have found out in 1968 just what had spooked Judge Holzer, and the subject could have been raised on direct appeal to the Supreme Court of Illinois. If this is the best characterization of what happened, then Branion has not established cause. There are no affidavits about what Bran-ion’s friends told him in 1968; Branion did not request an evidentiary hearing on the subject in the district court; when the sub ject came up at oral argument Branion’s counsel insisted that the record contains all the information he wishes to have before the court. Perhaps, then, we should infer that Branion did not pursue in 1968 the investigation that might have led to the Holzer-Tuite meeting and phone call because he could not have presented it to the Supreme Court of Illinois without opening up the subject of paying off a judge— which does not always cast in the best light a defendant who is protesting innocence. The district judge did not decide whether Branion had established “cause”, in light of her conclusion that he had not established “prejudice”.

There are at least two senses in which Branion may have suffered prejudice in fact. He may have lost an important right guaranteed by state law — the right to have the trial judge act as “thirteenth juror” and substitute his judgment for that of the jury, a decision that is not reviewable in Illinois. Under Illinois law, the judge must treat fairly with post-trial motions. People v. Yarbrough, 93 Ill.2d 421, 67 Ill.Dec. 257, 444 N.E.2d 493 (1982). And he may have lost the opportunity to receive an unjust, capricious, perhaps even paid-for acquittal. Neither is prejudice within the meaning of Sykes, a term the Court interpreted in United States v. Frady, 456 U.S. 152, 170, 102 S.Ct. 1584, 1595, 71 L.Ed.2d 816 (1982) (emphasis in original), as requiring a demonstration “not merely that the errors at ... trial created a possibility of prejudice, but that they worked to his actual and substantial disadvantage, infecting [the] entire trial with error of constitutional dimensions.” Neither of the two kinds of injury Branion may have suffered meets this test. The entitlement to have the judge act as an unreviewable thirteenth juror is not a constitutional right at all; in federal courts, district judges may set aside a jury’s verdict of guilt only if the evidence is legally insufficient, and acquittals notwithstanding the verdict may be reviewed on appeal, United States v. Scott, 437 U.S. 82, 91 & n. 7, 98 S.Ct. 2187, 2194 & n. 7, 57 L.Ed.2d 65 (1978); United States v. Allison, 555 F.2d 1385, 1386-87 (7th Cir.1977). If the ex parte exchange deprived Branion of an opportunity to have the judge play thirteenth juror, it deprived him only of an entitlement under state law, which is not sufficient to support a writ of habeas corpus. Jones v. Thieret, 846 F.2d 457, 459-60 (7th Cir.1988) (collecting cases); Watson v. Camp, 848 F.2d 89 (7th Cir.1988). The possibility that the meeting swindled Branion out of a corrupt acquittal may show that Judge Holzer did not deliver what was paid for, but it hardly establishes constitutional prejudice. Cf. Segura v. United States, 468 U.S. 796, 815-16, 104 S.Ct. 3380, 3390-91, 82 L.Ed.2d 599 (1984).

Did the ex parte contact work to Bran-ion’s “actual and substantial disadvantage, infecting [the] entire trial with error of constitutional dimensions”? No. It did not affect the trial at all. Branion had a fair trial, after which the jury rendered its verdict of guilt. He did not lose any trial opportunity — and he did not lose the opportunity to test the state’s evidence on appeal. The ex parte conversation induced Judge Holzer to send the case on its way through appellate channels — something he could have done summarily after the trial without violating any of Branion’s rights. Seven impartial arbiters reviewed the evidence after a full appellate presentation. Six justices of the Supreme Court of Illinois concluded that the evidence was sufficient; the seventh dissented on unrelated grounds. Unless every violation of the rules is “prejudice” — which cannot be right, for that would be tantamount to abolishing the prejudice component of Ny/ces-Branion has not met his burden.

IV

Branion presents two more contentions: That his representation at trial was constitutionally deficient, and that the prosecutor’s closing argument implied a comment on his decision not to testify in his own defense.

The latter is insubstantial. The prosecutor said, during a lengthy argument:

He [Branion] would have you believe, through hearsay, not through testimony on the witness stand under oath and subject to cross examination, but through hearsay, through Nelson Brown, said that he told him that night that he was missing a PPK.

This was a legitimate comment on the weakness of Branion’s defense. Hearsay is not the equivalent of direct testimony. Griffin v. California, 380 U.S. 609, 85 S.Ct. 1229, 14 L.Ed.2d 106 (1965), forbids asking the jury to draw an inference of guilt from the exercise of the privilege against compulsory self-incrimination, but the prosecutor did not ask the jury to draw such an inference. See generally United States v. Robinson, — U.S. -, 108 S.Ct. 864, 868-69, 99 L.Ed.2d 23 (1988); United States v. Sblendorio, 830 F.2d 1382, 1391-92 (7th Cir.1987).

Whether counsel was deficient is a harder question. Such contentions require a thorough appreciation of the opportunities open to counsel, and the use made of them. Judge Getzendanner held that Bran-ion had not shown “cause” for his failure to present this subject to state court. Defects in representation usually are presented immediately following the conclusion of the direct appeal. (Since the supposedly deficient counsel may be handling that appeal, he cannot be expected to expose his own shortfalls—even if he appreciates them.) Branion fled to Uganda instead of pursuing his case in the ordinary way. He had no cause to do this. The state did not hamper him from gathering and presenting the information he now deploys against his lawyer’s strategy and tactics; indeed, most of the necessary information was in his hands before the end of the trial. Branion insists, for example, that his counsel did not interview or call some witnesses whose names he suggested, did not interview additional pathologists to obtain evidence that more than four bullets must have passed through Donna Branion, and that against the judgment of an adviser, Eugene Pine-ham (now a judge of the Appellate Court of Illinois), did not call Branion as a witness in his own defense. These things were evident before and during trial. Branion’s own flight was the cause of the default in state court, and this is not a good excuse for a 20-year delay.

Murray v. Carrier, 477 U.S. 478, 495-97, 106 S.Ct. 2639, 2649-50, 91 L.Ed.2d 397 (1986), said that the cause-and-prejudice rule does not apply when the prisoner demonstrates that his incarceration is “fundamentally unjust”, citing Engle v. Isaac, 456 U.S. 107, 135, 102 S.Ct. 1558, 1576, 71 L.Ed.2d 783 (1982). Branion seeks the benefit of this exception. To the extent this is the claim of innocence in another guise, it is unpersuasive for the reasons we gave in Part II. To the extent Branion asks for a favorable exercise of equitable discretion, he has no equities. Judge Getzendanner wrote (emphasis in original):

The reason that the only evidence now available regarding these events is so old is that Branion himself chose to flee to Africa rather than serve his prison term as was his legal obligation. He flouted the very legal processes that he now seeks to invoke. In these circumstances, the continued incarceration of Branion does not strike me as a “miscarriage of justice.” I therefore decline to suspend the cause-and-prejudice rule for ... Branion’s ineffective assistance claim.

We agree.

Judge Plunkett’s opinion and judgment of November 20, 1987, are vacated because he had no jurisdiction. Appeal No. 87-3053, which challenges that judgment, is dismissed for want of jurisdiction. Segments J through Q of Branion’s appendix, and Part III of his opening brief, are stricken because they incorporate materials not in the record. On Appeal No. 87-3052, from Judge Getzendanner’s final order of September 22, the judgment is

AFFIRMED.

APPENDIX

3. Judge Plunkett’s scenario requires improbable assumptions that do not comport with the standard of guilt beyond a reasonable doubt.

It is one thing to construe record evidence “in the light most favorable to the prosecution.” But it is quite another to

adopt so improbable a view of the evidence that it neither rationally nor fairly supports a finding of guilty “beyond a reasonable doubt.” In the Branion case there were two key pieces of uncontradicted prosecuto-rial evidence given in the form of a range of estimated times: Dr. Belmonte’s evidence that the bruises on Donna Branion’s neck took between 15 to 30 minutes to form, and Detective Boyle’s evidence that the driving time of Dr. Branion’s route took between 6 to 12 minutes. When evi-dentiary ranges are given, the most probable events can be plotted on a standard statistical deviation curve or “bell curve”:

Frequency

[[Image here]]

6 min.DRIVING TIME.12 min.

15 min.GARROTING TIME .... 30 min.

If the probability standard for proof “beyond a reasonable doubt” is 80% to 90% certainty, as it is generally taken to be, then the most reasonable assumption of the two times are: (a) the bruise took between 21 and 24 minutes to form, and (b) the drive took between 8 to 10 minutes. As the times diverge in either direction from these limits, the probabilities grow rapidly smaller. At (a) 15 minutes and (b) 6 minutes, the probabilities are less than 1%. Moreover, since the two time-sequences are independent of each other, the probabilities must be multiplied to find the chances that both were applicable to Dr. Branion. Hence the probability that the time between garroting and shooting was 15 minutes and the probability that the driving time was 6 minutes is .01 x .01 = .0001. To suggest, as Judge Plunkett does, that this improbability of .0001 actually occurred is in fact to concede the irrationality of a finding that Dr. Branion was guilty beyond a reasonable doubt. A reasonable doubt probability of 90%, or 0.9, is 9,000 times larger in magnitude than the joint probability of .0001. In other words, if there were 9,000 cases similar to Dr. Bran-ion’s where the only two facts known were the garroting time and the driving time, then 8,999 defendants would be innocent and only one guilty. In short, the chance that Dr. Branion is guilty, on the basis of the garroting time and the driving time alone, is one in 9,000.

But even that insignificant chance is too high in Dr. Branion’s case, because even if the utterly improbable happened, there still would not be enough time for Dr. Branion to have killed his wife. The minimum driving time (6 minutes) plus the minimum garroting time (15 minutes) subtracted from the window of time of 27 minutes the state contends Dr. Branion had (11:30 to 11:57) is 27 — 21 = 6 minutes. But Dr. Branion could not have done all the things the state concedes he did in those remaining 6 minutes. According to the police detectives’ conclusion based upon interro gation of witnesses, it took 4 to 5 of those 6 minutes just to account for the time between Dr. Branion’s running out on the back porch calling for “Helen!” and when the police were called at 11:57. A. 216. And it took another 5 minutes for Dr. Bran-ion to dress Joby in winter outerclothes, according to prosecution witness Joyce Kelly. A. 99. These two uncontroverted time segments alone add up to a total of 9 to 10 minutes, overwhelming the 6 minutes available. In addition, there is a host of other events that consumed additional time.

1 . Amin, a brutal dictator, was overthrown in 1979, and the country has since undergone several changes of government. See David Lamb, The Africans 77-107 (1987).

2 . The jurisdictional statement in Branion’s brief compounds matters. The statement omits any reference to either the Rule 59 motion before Judge Getzendanner or the notice of appeal filed on October 20, despite the fact that Circuit Rule 28(b) requires all such matters to be spelled out. The jurisdictional statement as drafted initially led us to believe that the motion filed on October 15 was the first post-trial motion, and the notice of appeal on November 25 an appeal from its denial. The appellee’s brief certified that the appellant’s jurisdictional statement is "complete and correct”. It is difficult for this court to carry out the necessary inquiry into its own jurisdiction without the assistance of counsel. Counsel for both sides in this case misled rather than assisted the court.

3 . Branion furnished Judge Getzendanner with a report from the Northwestern University Traffic Institute concluding that the police estimates were too short. Perhaps traffic patterns (and the number of stop signs and stop lights) changed in the intervening 18 years. At all events, disputes of this kind do not assist the accused under Jackson; we look at the evidence in the light most favorable to the verdict.

4 . Branion contends that he had a 5-minute conversation with a neighbor before entering his house. This evidence was not presented to the state courts or to Judge Getzendanner and is among the submissions we have stricken from the appendix.

5 . We suggested in Brown v. Bowen, 847 F.2d 342, 345-46 (7th Cir.1988), that reasonable doubt means 0.9 or so, with adjustments depending on the gravity of the offense. That figure is the median of estimates from a survey of federal judges. See C.M.A. McCauliff, Burdens of Proof: Degrees of Belief, Quanta of Evidence, or Constitutional Guarantees, 35 Vand.L. Rev. 1293, 1325-32 (1982). Judges were not of one mind about this, however. Some thought that reasonable doubt means 100% certainty (as if that were available for any endeavor after Werner Heisenberg's work), while one thought reasonable doubt meant 50% or more. At least no federal judge thought that probabilities of less than 50% could satisfy the reasonable-doubt standard! (Defining "beyond a reasonable doubt” as a probability of 0.9 or more does not imply that 10% of all those convicted are innocent, any more than the civil burden of p > 0.5 implies that half of all civil cases are decided incorrectly. The limit is a cutoff, a minimum degree of confidence after the rules of evidence and other devices narrow the considerations before the court. A range of 0.9 < p < 1.0 could imply a median confidence somewhere around 0.95. Almost all cases are decided at higher levels of confidence still, because constraints on resources lead the prosecutor to filter out borderline cases.)

6 . Compare Michael O. Finkelstein & William B. Fairley, A Bayesian Approach to Identification Evidence, 83 Harv.L.Rev. 489 (1970), with Laurence Tribe, Trial by Mathematics: Precision and Ritual in the Legal Process, 84 Harv.L.Rev. 1329 (1971), and the continuation in Finkel-stein’s Quantitative Methods in Law 288-310 (1978). The more useful recent contributions include Lea Brilmayer & Lewis Kornhauser, Quantitative Methods and Legal Decisions, 46 U.Chi.L.Rev. 116, 135-48 (1978), and David Kaye, Apples and Oranges: Confidence Coeffi cients and the Burden of Persuasion, 73 Cornell L.Rev. 54 (1987). Several of the principals in this debate summarize and extend their arguments in a symposium, Probability and Inference in the Law of Evidence, 66 B.U.L.Rev. 377-952 (1986).

7 . Straight multiplication of probabilities is not the right way to find the joint probability. We need to compute, from the means and variances of two or more independent events, the joint mean and variance. See Hugh D. Young, Statistical Treatment of Experimental Data 96-99 (1962), for a description of the method. Only with these data may we determine the probability that a given observation will fall a given distance from the joint mean. Consider a quantity Q (here, time) to be calculated from several observed quantities a, b, c, and so on. The first step is to calculate the mean and variance of each of the constituents, where:

o 1 N aa = Jf X <Aai)2 /= 1

The complete formula, derived by a process that need not be repeated, is:

2 amQ -m 'ma 2 °mb +

where amQ is the variance of the mean of Q, oma the variance of the

mean of a, and so forth. In this case Q = a + b, so (£) = I and dQ (37 )

= 1. Thus a Q 2 ab-

Branion seems to assume, for reasons he does not explain, that the times given in the testimony correspond to two or three standard deviations on either side of the mean. (We infer this from the scale on the graph in the brief.) this leads to the conclusion that a = 9 minutes and aa = 1, and that

= + 2.5^ = 7.25, implying sr <T> 3 Q O10 3 CL Q 3 0 to to

that Q = 31.5 minutes and oq = 2.73 minutes. It is then straightforward to determine that the probability that Q < 21 ~ 0.001. If we assume instead that the ranges in the evidence correspond to one standard deviation, then the standard deviation of driving times = 3 and the standard deviation of bruise-forming times is 7.5 minutes, and we quickly compute oq = 8.1 minutes, so that the probability that Q < 21 ~ 0.10.

8 . An affidavit signed by William L. Hooks in July 1986 states:

"After Dr. Branion was convicted, and. while Judge Holzer was deciding what to do about the defense motion for a judgment notwithstanding the verdict, Nelson Brown [Donna Brown Bran-ion's brother] came to me and said that Judge Holzer was looking for a bribe.... A couple days later, Nelson Brown told me that he was able to get $20,000 from other friends of John Branion’s, but they had imposed a condition that $10,000 be paid to Judge Holzer in advance and the remaining $10,000 would be paid to him as soon as John Branion was freed.... A day or two later, I inquired of Nelson Brown what was going on, and he informed me of a meeting that was held at the Holiday Inn on Lake Shore Drive, at which Judge Holzer and his bailiff, and Nelson Brown, were present. At that meeting, $10,000 in cash was passed to Judge Holzer. A dáy or two later ... Nelson Brown informed me that Judge Holzer said that he could not accept the remaining $10,000 in addition to the $10,000 he had already received, because the state’s attorney had somehow gotten wind of the meeting in the Holiday Inn and threatened to arrest everyone concerned and to lock up the judge too if the judge reversed the jury’s verdict. Nelson Brown added that Judge Holzer said he would, however, release John Branion on bail in return for the $10,000 that was paid.” Holzer indeed released Branion, allowing him to flee the country. If this story is true, more than one extortion plot was afoot.