Maurice Robinson v. Patrick Arvonio, Superintendent, East Jersey State Prison Robert J. Del Tufo, Attorney General of the State of New Jersey

OPINION OF THE COURT

SCIRICA, Circuit Judge.

Petitioner Maurice J. Robinson, currently confined in a New Jersey state prison, appeals the district court’s denial of his application for a writ of habeas corpus under 28 U.S.C. § 2254 (1988). Robinson asserts the *879writ should have been granted because the New Jersey prosecutor failed to correct a witness’s perjured denial that his cooperation had been secured by the prosecutor’s promise to tell the sentencing judge of his cooperation.1 The district court found the failure of the prosecution to correct the witness’s perjured testimony was harmless error because there was no reasonable likelihood the perjured testimony had affected the judgment of the jury. This finding rested on the court’s judgment that the jury had sufficient evidence before it to assess the credibility of the witness. For reasons that follow, we will affirm the denial of the writ.

BACKGROUND

A jury found Robinson guilty of: (1) murder, under N.J.Stat.Ann. § 2C:ll-3 (West 1982 & 1993 Supp.); (2) armed robbery, under N.J.StatAnn. § 2C:15-lb (West 1982); (3) possession of a handgun without a permit, under N.J.Stat.Ann. § 2C:39-5b (West 1982); and (4) possession of a handgun for unlawful purposes, under N.J.Stat.Ann. § 2C:39-4a (West 1982). Co-defendant Melvin Shark was granted a severance and a jury found him guilty of the same charges. On February 24, 1984, Maurice J. Robinson was sentenced to a thirty-year term of imprisonment for murder, with a minimum of fifteen years without parole. He also received a concurrent fifteen-year term for armed robbery, with a minimum of seven years without parole.

After exhausting state remedies,2 Robinson petitioned for habeas corpus relief in United States District Court. The district court agreed the prosecutor had failed to correct the witness’s testimony, but it found no reasonable likelihood that the error had affected the judgment of the jury.

A.

In his opening statement at trial, the prosecutor informed the jury that Melvin Shark had previously been found guilty and that an agreement had been made with Shark in return for his testimony at Robinson’s trial. The prosecutor stated:

Mr. Shark is going to be a witness in this trial and the State is going to call him as a witness. I would like you to understand at this point that Mr. Shark has had his day in court and has already been adjudicated. And in terms of why he would be coming here at this time, I’d like you to know beforehand that Mr. Shark denied his guilt when he was on trial and subsequently he was found guilty. He has agreed to testify for the State, and what was promised to him was merely he would be kept away from Mr. Robinson, that is it. He will— He was told, and the extent of any promise made to him when his sentence came, his cooperation would be made known to the sentencing judge, but no deal in terms of time, years, anything like that.
I’d also like to note to you that Mr. Shark will be testifying under what’s known as immunity, because he has been recently convicted and he has not yet been sentenced, as I said earlier.

(Emphasis added.)

At trial, however, Shark initially denied that he had been promised anything at all in return for his testimony. On direct examination, Shark testified:

*880Q. Mr. Shark, in reference to your testimony in this courtroom today, was there a deal made with you so you would testify? A. No.
Q. What is your understanding of why you were — what you expect to gain as a result of testifying here?
A. Can you repeat that one more time, please?
Q. What prompted you to testify at this particular trial after your own trial?
A. I was hoping that—
Q. Mr. Shark, do you know what will be done with you as a result of your trial and your conviction?
A. I know I’ll be going for a long time. Q. Did you discuss that with your lawyer?
A. Yes, I did.
Q. And were you promised anything in exchange for testifying here as a result of your trial?
A. No, I wasn’t.
Q. Are you telling the truth today?
A. Yes, I am.

On cross-examination, Shark acknowledged that he was promised immunity for his testimony:

Q. Mr. Shark, is it not a fact that you have been promised immunity from Prosecution based on anything you said in court here today?
A. Yes, it is.

But he still denied the agreement included making his cooperation known to the sentencing judge:

Q. Is it not a fact, sir, that the prosecutor promised you that if you would get on that stand and testify today for the State against Mr. Robinson, they would tell the Judge of your cooperation?
A. Well, he didn’t promise me he’d talk to the Judge about that.
Q. He didn’t say that?
A. He didn’t promise me he’d go back in and talk to the Judge and tell him I cooperated. Was nothing promised to me.
Q. And that’s the truth?
A. Yes. Only thing was promised to me was I’d be taken care of.

On redirect by the prosecutor, Shark emphasized he had been promised protective custody in exchange for his testimony:

Q. Mr. Shark, earlier during your cross-examination, in response to one of [defense attorney’s] questions about what you were promised, you said something along the line that you would be taken care of. What does that mean?
A. Protective custody.
Q. And what does that mean to you?
A. Means a lot to me.
Q. Why?
A. Behind the fact what’s going on here, word can get back to the jail, you know, that I did something that, you know, is against rules and regulations inside the jail and I can be seriously hurt for doing something like this.

Shark never acknowledged the prosecutor had agreed to inform the sentencing judge of his cooperation, and the prosecutor failed to correct his perjured testimony. The defense attorney revived the issue in his closing statement by referring to the prosecutor’s promise to speak to the sentencing judge. He stated: .

You recall [Shark] saying from this witness stand, “I just want to get the truth out”? What else did he say? Unconsciously he said, “I want to help myself.” Then I asked him — this is the man who now wants to tell the truth — I said, “Did the Prosecutor promise you anything?” I’m sure you mil recall, ladies and gentlemen, that when the Prosecutor made his opening remarks, he indicated, and stated truthfully to you, that he had made certain promises to this defendant. On this witness stand — a man who’s worried about his survival is not going to forget something like that, if any promises were made to help him, but on this witness stand, after taking the oath on that Bible, I asked him, “Were any promises made to you?” “No.” “Did the Prosecutor promise that he’d talk to the Judge on your behalf?” *881“No. No •promises made.” That’s what he said.
And if you’ll recall, ladies and gentlemen, we took a recess, and when we came back, I asked him, I said, “Did the Prosecutor promise you immunity?” What did he say — “Oh, yes, he did promise me immunity.”
You heard him yesterday say in response to my question that, “I was mentioning 60 years to myself,” ladies and gentlemen, if you face that kind of time and you can buy your way out, do it. He did it. He got up on that witness stand and said to all of you, “I perjured myself, I’m a liar,” Fine. To help himself. Do you think that that young man, streetwise, wouldn’t have certain ideas about what it means if a Prosecutor said, “I will speak to the Judge for you? Ladies and gentlemen, he was sinking. And it was up to his nose and he reached out and he grabbed that straw of immunity to try to save himself. That’s what he did. And I suggest to you that you cannot believe him. You can’t believe anything that he said.
He said on that witness stand that the very first day when he was arrested he was concerned about getting himself out of this. He said on the witness stand yesterday that he was trying to help himself, and in order to do his job, he even said, “Nothing was promised to me,” and either he thought better of it or felt that he would be entrapped and then suddenly he admitted what was promised to him. And the Prosecutor would say to you, “Yes, I did promise him I’d speak to the Judge, ” but, ladies and gentlemen, the Judge is independent, the Judge makes up his own mind and, you know, that’s true. Sentencing is for the Judge. The Prosecutor can’t tell the Judge what to say.
But, ladies and gentlemen, don’t you think some little weasel who’s trying to save himself will take those words and, because he is so desperate, make those words say things that they really did not say, that maybe, maybe, this is a chance for me to get out of this, ...

(Emphasis added.) In his closing statement, the prosecutor referred to Shark’s sentencing and protective custody:

Yes, the State called him as a witness but what did Melvin Shark expect? I opened to you and told you certain things were promised to him, yes. What [was] his deal? His big deal? His lawyer was present when we discussed it. And he explained his answers.... “I know that the Judge can sentence me and I know the Prosecutor doesn’t have any power over the Judge because the jury found me guilty. No plea bargaining. Jury verdict, guilty. So the Prosecutor does not have the power to recommend anything to the Judge anymore in terms of years, months, days. ” So what did he testify for? And this is where it’s coming out now: Well, what did he say was facing him over there in the code, a different code, not the code we live by here, but in the jail here: Subjected to some physical beatings if the word gets out. So what did he want from the State? “I want protective custody. I’m subject to being severely beaten if I testify against another defendant.”

(Emphasis added.)3

In his final instructions to the jury, the trial judge referred only to the immunity *882part of the agreement and reminded the jury the opening and closing statements were not evidence. As we have noted, however, the prosecutor, in his opening statement, acknowledged he had promised to speak to the sentencing judge on Shark’s behalf, and the defense attorney, in his closing argument, built on that acknowledgement. Therefore, the jury was made aware of the terms of the agreement and the defense attorney made a strong case for Shark’s lack of credibility based on the agreement and Shark’s conflicting testimony.

B.

Shark’s testimony was important for Robinson’s conviction because only one other witness could place Robinson near the scene of the crime and no other witness could testify Robinson had taken part in it. We will recite the relevant facts.

On May 28, 1991, between 1:30 a.m. and 1:45 a.m., the night manager of the New Ridgewood Bar in Newark, N.J., Robert Conaway, unlocked the doors and admitted Melvin Shark, a customer whom he knew by sight, and his male companion. A third man waited outside. In addition to Conaway, barmaid Barbara Evans, her friend Constance Brooks, and two others were in the bar. Shark purchased beer from Evans and asked her for change for the cigarette machine. After conferring with his companion at the machine, Shark asked her for more change, which she refused. Conaway asked Shark and his companion to leave because it was almost closing time. When he opened the door, the third man appeared, asking if they had gotten the beer. Conaway saw the three young men walk away.

Approximately five minutes later, Conaway heard Benjamin Sanders pounding on the bar door shouting to be let in. He said that K.O. Floyd had been shot. Sanders and Floyd, both elderly men, had been sitting outside a social club next door to the bar. Sanders noticed three young men walk past them and saw the first one look at him as they went by. (He later identified Shark as that man.) He saw no other people in the area. The three young men stopped a short distance past the elderly men and had a conversation. Sanders was frightened of the young men; he told Floyd they should leave and started to walk toward Floyd’s car. The young men came back and surrounded the two elderly men. When Sanders saw the second of the three young men pull a gun out of his pocket, he ran to the New Ridgewood Bar and pounded on the door. At that time, he heard Floyd call for help and heard a shot. Just before Conaway opened the door, Sanders turned around and saw Floyd lying in the street and the three men running away. Floyd died of his wounds.

Some two years after the shooting, Sanders was talking to a friend in front of the same social club when he noticed someone staring at him. After seeing the same man on several other occasions, he recognized him as one of the three men involved in Floyd’s murder. He contacted the police and Shark was arrested. Shark admitted being present at the murder, but denied firing the gun. He identified the other two men with him as Maurice Robinson and Robinson’s brother Charles (a minor at the time of the murder).

At Maurice Robinson’s trial, neither Sanders nor the bar manager, Conaway, could identify Maurice Robinson as someone he had seen on the night of the murder. But Constance Brooks, the bar patron, testified she recognized Maurice Robinson as one of the two men who had purchased beer before the shooting.4

*883Shark testified to the following: that night, he met the Robinson brothers near his residence and went with them to the bar; he and Maurice Robinson went in to buy beer while Charles Robinson waited outside; when leaving the bar, they saw two men outside a social club; they walked past the men, with Shark in the lead, and slowed down; Shark heard sounds behind him, scuffling, a shout, and a gun shot; he turned and saw one of the men fall down; Maurice Robinson was standing a few feet away with a gun in his hand; Shark picked up some money lying next to the fallen man and fled with his companions; he gave half the money to Maurice Robinson and went home.

DISCUSSION

A.

In a habeas corpus petition, if the district court held an evidentiary hearing, we review its findings of fact for clear error, Lesko v. Owens, 881 F.2d 44, 50-51 (3d Cir.1989), cert. denied, 493 U.S. 1036, 110 S.Ct. 759, 107 L.Ed.2d 775 (1990), but if it relied on the state court record, our review is plenary. Zettlemoyer v. Fulcomer, 923 F.2d 284, 291 (3d Cir.), cert. denied, — U.S. -, 112 S.Ct. 280, 116 L.Ed.2d 232 (1991); Reese v. Fulcomer, 946 F.2d 247, 253-54 (3d Cir.1991), cert. denied, — U.S. -, 112 S.Ct. 1679, 118 L.Ed.2d 396 (1992). In this case, the district court did not hold an evidentiary hearing, so our review is plenary. We review the trial record de novo, as did the district court.

B.

A defendant’s right to due process is implicated when the state obtains a conviction based upon testimony the state knows is perjured. Napue v. Illinois, 360 U.S. 264, 269, 79 S.Ct. 1173, 1177, 3 L.Ed.2d 1217 (1959); Giglio v. United States, 405 U.S. 150, 153-55, 92 S.Ct. 763, 765-66, 31 L.Ed.2d 104 (1972). In Napue and Giglio, the prosecution made agreements with witnesses in exchange for their testimony. Both witnesses falsely denied the existence of the agreements, and the prosecutors failed to correct their perjured testimony.

In Napue, the Court held a conviction is obtained through the use of false evidence, and therefore violates the Fourteenth Amendment, when the state, “although not soliciting false evidence, allows it to go uncorrected when it appears.” 360 U.S. at 269, 79 S.Ct. at 1177. The Court elaborated:

The principle that a State may not knowingly use false evidence, including false testimony, to obtain a tainted conviction, implicit in any concept of ordered liberty, does not cease to apply merely because the false testimony goes only to the credibility of the witness. The jury’s estimate of the truthfulness and reliability of a given witness may well be determinative of guilt or innocence, and it is upon such subtle factors as the possible interest of the witness in testifying falsely that a defendant’s life or liberty may depend.

Id.

In Giglio, the government’s ease depended almost entirely on the testimony of a witness whom the government promised it would not prosecute if he testified. The trial prosecutor had not himself made the agreement and was unaware of it, but the Court charged him with knowledge of the agreement made by his predecessor. The Court held that because the evidence was relevant to the jury’s assessment of the credibility of the witness, a new trial would be required if “ ‘the false testimony could ... in any reasonable likelihood have affected the judgment of the jury ... ’ ” Giglio, 405 U.S. at 154, 92 S.Ct. at 766, (quoting Napue, 360 U.S. at 271, 79 S.Ct. at 1178) (alteration in original). Both Giglio and Napue embody the rule that the state’s knowing use of perjured testimony to obtain a conviction is constitutional error, but that does not automatically entitle petitioner to relief. The court must also decide the error is not harmless.

The Supreme Court has distinguished two kinds of constitutional error at trial: structural error and trial error. A structural error compromises the entire trial; it requires reversal because it involves a *884deprivation of a constitutional protection so basic that in its absence, ‘“a criminal trial cannot reliably serve its function as a vehicle for determination of guilt or innocence, and no criminal punishment may be regarded as fundamentally fair.’ ” Arizona v. Fulminante, 499 U.S. 279, 310, 111 S.Ct. 1246, 1265, 113 L.Ed.2d 302 (1991) (citation omitted). Examples of structural error are a biased judge, Tumey v. Ohio, 273 U.S. 510, 47 S.Ct. 437, 71 L.Ed. 749 (1927), or the denial of trial counsel, Gideon v. Wainwright, 372 U.S. 335, 83 S.Ct. 792, 9 L.Ed.2d 799 (1963). A trial error is one which “occurred during the presentation of the case to the jury, and which may therefore be quantitatively assessed in the context of other evidence presented in order to determine whether its admission was harmless-” Fulminante, 499 U.S. at 307-08, 111 S.Ct. at 1263. Examples of constitutional trial error are jury instructions misstating an element of the offense, Rose v. Clark, 478 U.S. 570, 106 S.Ct. 3101, 92 L.Ed.2d 460 (1986), or improper comment on defendant’s silence at trial, U.S. v. Hasting, 461 U.S. 499, 103 S.Ct. 1974, 76 L.Ed.2d 96 (1983). Structural error cannot be harmless; trial error can be. Fulminante, 499 U.S. at 310, 111 S.Ct. at 1265. The error in this case was trial error. The testimony must therefore be assessed in the context of the other evidence to determine whether its admission was harmless.

A prosecutor’s agreement to speak to the sentencing judge on a witness’s behalf in return for the witness’s testimony is especially likely to create an issue of credibility. As the United States Court of Appeals for the Fourth Circuit noted: Boone v. Paderick, 541 F.2d 447, 451 (4th Cir.1976), cert. denied, 430 U.S. 959, 97 S.Ct. 1610, 51 L.Ed.2d 811 (1977).

[Rjather than weakening the significance for credibility purposes of an agreement of favorable treatment, tentativeness may increase its relevancy. This is because a promise to recommend leniency (without assurance of it) may be interpreted by the promisee as contingent upon the quality of the evidence produced — the more uncertain the agreement, the greater the incentive to make the testimony pleasing to the promisor.

C.

Until recently, the standard for assessing harmless error on both direct and collateral review was whether it appeared “beyond a reasonable doubt that the error complained of did not contribute to the verdict obtained.” Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 828 (1967). Under Chapman, the burden of proving an error harmless rested with the state. Id.

In Brecht v. Abrahamson, — U.S. -, 113 S.Ct. 1710, 123 L.Ed.2d 353 (1993), the Court announced a new standard for harmless error on collateral review: whether, in light of the record as a whole, the error resulted in “actual prejudice” to the defendant. Brecht, -U.S. at -, 113 S.Ct. at 1722. Actual prejudice occurs when the constitutional error “ ‘had substantial and injurious effect or influence in determining the jury’s verdict.’ ” — U.S. at -, 113 S.Ct. at 1722 (quoting Kotteakos v. United States, 328 U.S. 750, 776, 66 S.Ct. 1239, 1253, 90 L.Ed. 1557 (1946)); Kontakis v. Beyer, 19 F.3d 110, 116 (3d Cir.1994).

By introducing a more deferential standard on collateral review, the Court emphasized that historically the writ of habeas corpus has been regarded as an extraordinary remedy to afford relief for grievous wrongs. Brecht, — U.S. at -, 113 S.Ct. at 1719. It reasoned that “‘[fjederal intrusions into state criminal trials frustrate both the States’ sovereign power to punish offenders and their good-faith attempts to honor constitutional rights,’ ” id. — U.S. at -, 113 S.Ct. at 1720 (quoting Engle v. Issac, 456 U.S. 107, 128, 102 S.Ct. 1558, 1572, 71 L.Ed.2d 783 (1982)), and stated that liberal allowance of the writ “‘degrades the prominence of the trial itself,’ ” id. — U.S. at -, 113 S.Ct. at 1720-21 (quoting Engle, 456 U.S. at 127, 102 S.Ct. at 1571), and encourages habe-as petitioners to relitigate their claims on collateral review, id., — U.S. at -, 113 S.Ct. at 1721. Thus, considerations of federalism and comity, the state’s interest in finality of convictions that have survived direct *885review, and the social costs of retrying defendants years after the original conviction outweigh any additional effect the Chapman standard would have in deterring state courts from failing to fully enforce constitutional rights. Id. — U.S. at -, 113 S.Ct. at 1721-22.

The new standard, announced in Brecht and derived from Kotteakos, does not require a showing that “but for” the error the jury would have rendered a verdict in favor of defendant. Duest v. Singletary, 997 F.2d 1336, 1338 (11th Cir.1993) (citing Kotteakos, 328 U.S. at 763, 66 S.Ct. at 1247). Nor is it relevant whether the reviewing court is persuaded the defendant is guilty. The court must stand in the shoes of the jury.5 Id., — U.S. at -, 113 S.Ct. at 1724. The question it must consider is: Did the constitutional error “substantially [and injuriously] influence” the verdict? Id. at 765, 66 S.Ct. at 1248. If so, the petitioner is entitled to habeas relief.

In addition to announcing a new standard, the Brecht Court shifted the burden of proof to the petitioner. The Court noted that, under Chapman, the state bore the burden of proving harmless error beyond a reasonable doubt. Brecht, — U.S. at -, 113 S.Ct. at 1717. Under the new standard for collateral review, “[petitioners] are not entitled to habeas relief based on trial error unless they can establish that it resulted in ‘actual prejudice.’ ” Id., — U.S. at -, 113 S.Ct. at 1722 (citing United States v. Lane, 474 U.S. 438, 449, 106 S.Ct. 725 (1986)).6

Robinson notes an exception to the Brecht rule, set forth in a footnote in that case. Declaring that the Kotteakos harmless-error standard applies with constitutional error “of the trial type,” the Court noted:

Our holding does not foreclose the possibility that in an unusual case, a deliberate and especially egregious error of the trial type, or one that is combined with a pattern of prosecutorial misconduct, might so infect the integrity of the proceeding as to warrant the grant of habeas relief, even if it did not substantially influence the jury’s verdict.

Brecht, — U.S. at - n. 9, 113 S.Ct. at 1722 n. 9. Robinson contends the error in this case is of such magnitude that it constitutes an exception to the new rule. He argues that in these circumstances, we should resort to the standard of United States v. Agurs, 427 U.S. 97, 96 S.Ct. 2392, 49 L.Ed.2d 342 (1976), where the Court stated, *886“[A] conviction obtained by the knowing use of perjured testimony is fundamentally unfair, and must be set aside if there is any reasonable likelihood that the false testimony could have affected the judgment of the jury.” Id. at 103, 96 S.Ct. at 2397 (footnotes omitted).

Robinson would have us apply the “reasonable likelihood” standard of Agurs rather than the “substantial and injurious effect” standard of Brecht, and he cites pre-Brecht cases evaluating “materiality” in the context of prosecutorial misconduct and applying the pre-Brecht “reasonable likelihood” standard of harmless error. Landano v. Rafferty, 856 F.2d 569 (3d Cir.1988), cert. denied, 489 U.S. 1014, 109 S.Ct. 1127, 103 L.Ed.2d 189 (1989) (habeas appeal); Brown v. Wainwright, 785 F.2d 1457 (11th Cir.1986) (habeas appeal); United States v. Wallach, 935 F.2d 445 (2d Cir.1991) (direct appeal). Regardless whether the Agurs standard still retains any vitality in habeas cases after Brecht, we do not find the error in this case warrants a departure from the Brecht standard.

There is little doubt the prosecutor should have corrected the perjured testimony of his witness. This was the duty of the prosecutor, not the defense attorney. Giglio, 405 U.S. at 154, 92 S.Ct. at 766; United States v. Harris, 498 F.2d 1164, 1169 (3d Cir.), cert. denied, 419 U.S. 1069, 95 S.Ct. 655, 42 L.Ed.2d 665 (1974). Nonetheless, when it became clear that the prosecutor had not corrected the perjured testimony, the defense attorney could have alerted the judge and sought a remedy that would have eliminated any possibility of prejudice to his client, such as a stipulation or an instruction on the details of the agreement. Instead, the defense attorney sought to counter the misleading impression through cross-examination and closing argument. Although we agree with Robinson that his attorney did not waive the error by failing to call it to the attention of the court, an error which the defense attorney could have corrected at trial is not likely “to infect the integrity of the proceeding_” Brecht, — U.S. at - n. 9, 113 S.Ct. at 1722 n. 9.

The prosecutor must be charged with the error because he failed to take advantage of opportunities on direct and re-direct examination to correct the perjured testimony. But the error was neither deliberate nor planned; in his opening statement, the prosecutor referred to all aspects of the agreement: immunity, protective custody, and the promise to inform the sentencing judge of Shark’s cooperation. Presumably, neither counsel knew the perjured testimony would occur, and both knew as soon as it did occur, as did the trial judge.7 We conclude the error was neither a “deliberate and especially egregious error of the trial type, [n]or one that is combined with a pattern of prosecuto-rial misconduct” that “might so infect the integrity of the proceeding as to warrant the grant of habeas relief, even if it did not substantially influence the jury’s verdict.” Brecht, — U.S. at - n. 9, 113 S.Ct. at 1722 n. 9. Therefore, we will apply the Brecht standard to determine whether petitioner has shown the error “ ‘had substantial and injurious effect or influence in determining the jury’s verdict.’”8 Id. — U.S. at -, 113 S.Ct. at 1722 (quoting Kotteakos, 328 U.S. at 776, 66 S.Ct. at 1253).

The district court emphasized the jury knew Shark testified under an agreement with the state by which he would benefit and that there was ample evidence to assess Shark’s credibility. It cited the following evidence:

The jury was aware that Shark was a convicted felon who had perjured himself at his own trial. Furthermore, it is apparent from the record that the jury was *887apprised of Shark’s conflicting testimony at Robinson’s trial. Shark at one point denied that he had been promised immunity and protective custody in exchange for his [testimony], yet he later admitted to these facts.
In bringing out this agreement under direct and cross-examination, both the prosecution and the defense conveyed to the jury that Shark was testifying subject to an agreement with the State from which he would benefit. The defense attorney questioned Shark regarding this agreement in an effort to impeach his credibility. With all of these factors relating to Shark’s credibility before the jury, this Court finds that there is no reasonable likelihood that Shark’s perjured testimony regarding the agreement with the State could have affected the judgment of the jury.

Robinson contended at oral argument that the evidence that Shark was a convicted felon who had perjured himself at his own trial is qualitatively different from the evidence the prosecutor improperly withheld. We agree a jury’s knowledge that a witness has just lied about what he stands to gain from testifying is generally more damaging to his credibility than the knowledge that he has lied in the past.

In this ease, however, the jury knew Shark had lied to it about what he stood to gain from the agreement as a whole. First, he denied he had been promised anything at all; then, on cross-examination, he admitted he had been promised immunity; finally, on redirect, he admitted he had been promised protective custody. Significantly, unlike the prosecutors in Napue and Giglio, the prosecutor here, in his opening statement, acknowledged his promise to tell the sentencing judge of Shark’s cooperation. The defense attorney built on that acknowledgment in arguing Shark was not a credible witness. We think it unlikely that further evidence Shark was misrepresenting the agreement would have changed the jury’s evaluation of his testimony.

The crucial inquiry is whether, under Brecht, Robinson has shown the error “had substantial and injurious effect or influence in determining the jury’s verdict.” As we have noted, in determining the effect of this trial error, which “occurred during the presentation of the case to the jury,” we must assess it “in the context of other evidence presented in order to determine whether its admission was harmless.” Fulminante, 499 U.S. at 308, 111 S.Ct. at 1264. We conclude there was sufficient evidence to undermine Shark’s credibility. The evidence of the prosecutor’s promise to tell the sentencing judge of Shark’s cooperation would have added only incrementally to the evidence presented on the agreement and Shark’s credibility.

The prosecutor had a duty to correct the perjured testimony of his witness. In these circumstances, however, we agree with the district court that the error was harmless. We conclude the absence of the evidence of the prosecutor’s promise to speak to the sentencing judge did not result in “actual prejudice” because it did not have “substantial and injurious effect or influence in determining the jury’s verdict.”

CONCLUSION

For the foregoing reasons, we will affirm the district court’s denial of the writ of habe-as corpus.

. Robinson contends the prosecutor’s conduct deprived him of due process of law and a fair trial under the Fifth, Sixth and Fourteenth Amendments of the United States Constitution.

. Robinson appealed to the Superior Court of New Jersey, Appellate Division, which affirmed his conviction in an unreported per curiam opinion filed on February 28, 1986. The New Jersey Supreme Court denied Robinson's request for certification. He then filed a pro se petition for post-conviction relief, which was denied after a hearing in his absence. He appealed, and in an unreported per curiam opinion, filed August 1, 1988, the Appellate Division found the hearing should not have been conducted in Robinson's absence and remanded for another hearing. After the second hearing, at which Robinson was present and represented by counsel, the trial court again denied his petition. Robinson appealed, and on October 1, 1990, in an unreported per curiam opinion, the Appellate Division affirmed the denial of post-conviction relief. The New Jersey Supreme Court denied Robinson's petition for certification on December 24, 1990.

. At Shark's sentencing hearing, the prosecutor asked for leniency for Shark and stressed the critical importance of Shark's testimony in obtaining Robinson’s conviction:

Mr. Shark was told by me in all candor that I would speak for him in this respect at his sentence, that your Honor would note his cooperation in testifying against his codefendant and codefendants.
[W]ithout his testimony, proving the case against his codefendant [Robinson] would have been difficult, if not impossible ...
He should be punished and he will be punished ..., but his cooperation in both the trials [of co-defendants] has to be noted because we would have not had the opportunity to convict
*882the person who actually pulled the trigger ... Maurice Robinson; without the help of Melvin Shark, albeit to help himself, but in exchange for nothing really concrete, Mr. Shark did testify, and I think that should be noted to the court.

As we have noted, Robinson and Shark were found guilty of the same offenses; however, Robinson was given a longer sentence. For the murder, he received a 30 year term with a 15 year mandatory minimum, whereas Shark received a 20 year term with a 10 year mandatory minimum. Robinson's concurrent term for armed robbery was also longer than Shark's.

. Conaway identified Shark whom he knew by sight as a patron of the bar, but he could not identify Robinson. Constance Brooks identified both Shark and Robinson, although she had not known either of them before; her visual memory may have been better than Conaway's because she was an artist. Barbara Evans, the barmaid, *883was not called to identify the men; she died at about the time they were apprehended.

. As the Court in Kotteakos explained, the issue is not whether the jurors were

right in their judgment, regardless of the error or its effect upon the verdict. It is rather what effect the error had or reasonably may be taken to have had upon the jury's decision. The crucial thing is the impact of the thing done wrong on the minds of other men, not on one’s own, in the total setting.
This must take account of what the error meant to them, not singled out and standing alone, but in relation to all else that happened. And one must judge others' reactions not by his own, but with allowance for how others might react and not be regarded generally as acting without reason. This is the important difference, but one easy to ignore when the sense of guilt comes strongly from the record.

Brecht,-U.S. at-, 113 S.Ct. at 1724 (Stevens, J., concurring) (quoting Kotteakos, 328 U.S. at 764, 66 S.Ct. at 1247-48 (citation omitted)).

. Justice Stevens, concurring, wrote separately to set forth his understanding that the Court, in adopting the Kotteakos standard, also adopted the Kotteakos Court's placing the burden on prosecutors to prove harmless error (- U.S. at -, 113 S.Ct. at 1723-24). The language of the opinion, however, clearly puts the burden on the habeas petitioner, not on the state. Most courts of appeals that have applied the new Brecht standard have assumed or stated that the burden of proof is on the petitioner. Kyles v. Whitley, 5 F.3d 806, 818 (5th Cir.1993); O’Neal v. Morris, 3 F.3d 143, 145 (6th Cir.1993); Tague v. Richards, 3 F.3d 1133, 1140 (7th Cir.1993); Henry v. Estelle, 993 F.2d 1423, 1427 (9th Cir. 1993); Jeffries v. Blodgett, 5 F.3d 1180, 1190 (9th Cir.1993); Castillo v. Stainer, 997 F.2d 669, 669 (9th Cir.1993). Some do not discuss who has the burden. Vanderbilt v. Collins, 994 F.2d 189 (5th Cir.1993); Duest, 997 F.2d 1336 (11th Cir.1993). A few quote a concurring or dissenting opinion in Brecht saying the state bears the burden, but do not explicitly adopt that view. Lowery v. Collins, 996 F.2d 770, 773 (5th Cir.1993) (quoting Stevens, J., concurring); Smith v. Dixon, 996 F.2d 667, 677 n. 13 (4th Cir.1993) (quoting White, J., dissenting). Only one opinion states outright that the state has the burden, Stoner v. Sawders, 997 F.2d 209, 213 (6th Cir.1993), and a later opinion by the same court stated that the burden is on the petitioner without referring to *886the earlier opinion. O’Neal v. Morris, 3 F.3d at 145.

. At oral argument, Robinson's counsel acknowledged that at trial, both attorneys and the trial judge knew what was in the agreement between Robinson and the state.

. The standard we apply is not the same standard the district court used. The district court’s opinion, which came out before Brecht, presumably used the Chapman standard when it denied the petition. If it had applied the new Brecht standard, which is less favorable to the habeas petitioner, it certainly would have denied the petition also. That does not matter to our review, however, because we look at the record de novo in reviewing the district court’s denial of petitioner’s habeas petition.