PER CURIAM:

Petitioner Allan Ross, a federal prisoner, appeals the district court's denial of his 28 U.S.C. § 2255 motion for habeas relief. We conclude that the district court correctly denied Petitioner's section 2255 motion, and we affirm that ruling.

In 1992, Ross was tried under a three-count indictment for engaging in a continuing criminal enterprise ("CCE") in violation of 21 U.S.C. § 848 (Count I), conspiracy to import cocaine and marijuana in violation of 21 U.S.C. §§ 952, 963 (Count II), and conspiracy to possess with intent to distribute cocaine and marijuana in violation of 21 U.S.C. §§ 841, 846 (Count III). At trial, he requested a jury instruction requiring the jury to agree unanimously on which offenses constituted the continuing series of violations of the CCE charge. The district court, pursuant to Circuit precedent at that time, denied Ross's request. The jury convicted Ross on all three counts, and he was sentenced to three concurrent life sentences and fined $2 million pursuant to Count I and $4 million each pursuant to Counts II and III.

On direct appeal, Ross argued, among other things, that the district court's failure to give his requested unanimity instruction constituted reversible error. We affirmed Ross's conviction and sentencing. See United States v. Ross, 33 F.3d 1507 (11th Cir. 1994). Ross filed a petition for habeas relief under 28 U.S.C. § 2255 and a motion for a new trial. While his section 2255 motion was pending, the Supreme Court decided Richardson v. United States, 526 U.S. 813, 119 S.Ct. 1707, 143 L.Ed.2d 985 (1999), in which the Court concluded that "a jury has to agree unanimously about which specific violations make up the `continuing series of violations'" of a CCE charge. See Richardson, 526 U.S. at 815, 119 S.Ct. at 1709. The district court denied all claims raised in Ross's original section 2255 motion but granted Ross leave to amend his petition to add a claim based on Richardson.

The district court ultimately denied Ross habeas relief on his Richardson claim. The district court concluded that, while Richardson does apply retroactively to cases on collateral review, Richardson error was subject to review for harmless error. And the district court determined that the failure to give a unanimity instruction at Ross's trial constituted harmless error. We granted a certificate of appealability on whether Ross was due relief in the light of Richardson.

The district court's determination that the rule articulated in Richardson should be applied retroactively to cases on collateral review is an issue of law to be examined de novo. Our sister circuits which have addressed this issue have concluded that Richardson does apply retroactively to cases on collateral review. See Santana-Madera v. United States, 260 F.3d 133, 139 (2d Cir. 2001); United States v. Lopez, 248 F.3d 427, 432 (5th Cir. 2001); Lanier v. United States, 220 F.3d 833, 838 (7th Cir. 2000); Murr v. United States, 200 F.3d 895, 906 (6th Cir. 2000).

The government contends that Teague v. Lane, 489 U.S. 288, 310, 109 S.Ct. 1060, 1075, 103 L.Ed.2d 334 (1989), under which new constitutional rules of criminal procedure are generally inapplicable to cases on collateral review, bars application of the rule in Richardson in this case. Teague's bar, however, does not apply when the Supreme Court decides a new substantive rule of criminal law as opposed to a new procedural rule. And a new rule is substantive when it interprets "the meaning of a criminal statute enacted by Congress" so that the conduct for which a defendant was convicted may no longer be illegal. Bousley v. United States, 523 U.S. 614, 620, 118 S.Ct. 1604, 1610, 140 L.Ed.2d 828 (1998).

In Richardson, the Supreme Court concluded that 21 U.S.C. § 848's "series of violations" refers to several "violations," each of which the jury must agree on unanimously. Richardson, 526 U.S. at 824, 119 S.Ct. at 1713. Therefore, the Court changed the elements of a CCE offense and, thus, changed the meaning of that statute.

We reject the government's contention that Richardson is akin to the decision in United States v. Gaudin, 515 U.S. 506, 522-23, 115 S.Ct. 2310, 2319-20, 132 L.Ed.2d 444 (1995), in which the Supreme Court concluded that materiality was an element to be decided by the jury (and not the district court) and which we have held is inapplicable to cases on collateral review. In Gaudin, neither party disputed that materiality was an element of the charged offense; instead, they disputed who decided that element. In contrast, pre- Richardson, a jury was required to find that the accused committed a series of offenses; the jurors, however, were not required to agree on what specific predicate offenses comprised this series. Post- Richardson, if jurors cannot agree on what specific offenses make up the continuing series, the jury may not convict the accused of the CCE offense. Therefore, we join our sister circuits in concluding that Richardson set out a new substantive rule of criminal law which applies retroactively to cases on collateral review.

Having determined that Richardson applies retroactively to cases on collateral review, we must next decide whether harmless error analysis applies to Richardson error. The courts of appeals which have addressed this issue have concluded that Richardson error is subject to harmless error analysis. See Santana-Madera v. United States, 260 F.3d 133, 139-40 (2d Cir. 2001); Lanier v. United States, 220 F.3d 833, 838-39 (7th Cir. 2000); United States v. Brown, 202 F.3d 691, 699 (4th Cir. 2000); United States v. Long, 190 F.3d 471, 476 n. 3 (6th Cir. 1999); United States v. Escobar-de Jesus, 187 F.3d 148, 161-62 (1st Cir. 1999). We agree. A jury instruction which omits an element of the charged offense is generally subject to harmless error analysis. See Neder v. United States, 527 U.S. 1, 9, 119 S.Ct. 1827, 1833, 144 L.Ed.2d 35 (1999) ("[A]n instruction that omits an element of the offense does not necessarily render a criminal trial fundamentally unfair or an unreliable vehicle for determining guilt or innocence."). Structural error, to which harmless error analysis does not apply, occurs only with "extreme deprivations of constitutional rights, such as denial of counsel, denial of self representation at trial, and denial of a public trial." United States v. Nealy, 232 F.3d 825, 830 n. 4 (11th Cir. 2000). And we conclude that instructional error under Richardson is not structural error, but instead is subject to review for harmless error.

The parties dispute precisely what harmless-error standard is applicable in this case. In Brecht v. Abrahamson, 507 U.S. 619, 636-38, 113 S.Ct. 1710, 1721-22, 123 L.Ed.2d 353 (1993), the Supreme Court concluded that the appropriate standard for reviewing the harmlessness of a non-structural constitutional error on collateral review of a state court judgment is the standard enunciated in Kotteakos v. United States, 328 U.S. 750, 66 S.Ct. 1239, 90 L.Ed. 1557 (1946). Under that standard, a defendant is entitled to habeas relief when an error results in actual prejudice because it "had substantial and injurious effect or influence in determining the jury's verdict." Brecht, 507 U.S. at 638, 113 S.Ct. at 1722. Ross, however, contends that the instructional error in this case should be evaluated under the harmless-beyond-a-reasonable-doubt standard of Chapman v. California, 386 U.S. 18, 24, 87 S.Ct. 824, 828, 17 L.Ed.2d 705 (1967). He offers two reasons for this assertion.

Ross notes that, because Brecht involved the habeas review of a state court conviction, the Supreme Court relied on principles of comity and federalism, as well as the State's interest in the finality of convictions and the extraordinary nature of habeas relief, in concluding that the more relaxed Kotteakos standard should apply in reviewing harmlessness of error during collateral review. See Brecht, 507 U.S. at 637, 113 S.Ct. at 1721-22. And, in Brecht, the state court had already evaluated the error under the more stringent Chapman standard on direct appeal. See id. at 636, 113 S.Ct. at 1721. Therefore, Ross contends, Brecht's relaxed harmless error standard should apply only when a federal appellate court is reviewing a state conviction in which the alleged error has been evaluated under the more stringent Chapman standard on direct appeal.

Only the Sixth and Seventh Circuits have addressed which harmless error standard is appropriate when reviewing Richardson error on habeas review. The Sixth Circuit has concluded that the Brecht standard applied, see Murr, 200 F.3d at 906 while the Seventh Circuit has applied the harmless-beyond-a-reasonable-doubt standard of Chapman. See Lanier, 220 F.3d at 839.

We conclude that the Sixth Circuit is correct and that application of the Brecht standard to Richardson error on collateral appeal is the appropriate approach. In this Circuit, we apply the Brecht harmless error standard to the habeas review of federal court convictions, as well as state court convictions. See Vines v. United States, 28 F.3d 1123, 1130 (11th Cir. 1994). And we, when looking at state decisions, have rejected the theory that the Brecht standard only applies to convictions which have been previously reviewed under the Chapman standard. See Horsley v. Alabama, 45 F.3d 1486, 1492 n. 11 (11th Cir. 1995). Ross offers no compelling argument for why we should treat federal convictions differently. We, therefore, conclude that the Brecht standard for harmless error applies to our habeas review of the Richardson error at Ross's trial.

We must now consider whether the district court's failure to give a unanimity instruction as required by Richardson constitutes harmless error under Brecht in this case. When reviewing the harmlessness of an error under the Brecht standard, "[i]f, when all is said and done, the [court's] conviction is sure that the error did not influence, or had but very slight effect, the verdict and the judgment should stand." O'Neal v. McAninch, 513 U.S. 432, 437, 115 S.Ct. 992, 995, 130 L.Ed.2d 947 (1995) (quoting Kotteakos v. United States, 328 U.S. 750, 764, 66 S.Ct. 1239, 90 L.Ed. 1557 (1946)). But if a federal court is "in grave doubt about whether a trial error of federal law had substantial and injurious effect or influence in determining the jury's verdict, that error is not harmless." Smith v. Singletary, 61 F.3d 815, 818-19 (11th Cir. 1995) (quoting O'Neal, 513 U.S. at 436, 115 S.Ct. 992) (applying Brecht standard).

The "continuing series" of a section 848 offense is made up of at least three connected violations of the controlled substances laws. See United States v. Alvarez-Moreno, 874 F.2d 1402, 1408 (11th Cir. 1989), cert. denied, 494 U.S. 1032, 110 S.Ct. 1484, 108 L.Ed.2d 620 (1990). And in this Circuit, a controlled substance conspiracy conviction may count as a predicate offense of a CCE charge. See United States v. Rosenthal, 793 F.2d 1214, 1227 (11th Cir. 1986), cert. denied, 480 U.S. 919, 107 S.Ct. 1377, 94 L.Ed.2d 692 (1987). Here, Ross's two conspiracy convictions make up two of the three controlled substances offenses necessary to sustain his CCE conviction. Still, no third, connected conviction exists which would complete the continuing series of the CCE. Thus, we must look to the record to determine whether we have at least "grave doubt" about whether the absence of a Richardson instruction in this case had "substantial and injurious effect or influence in determining the jury's verdict." Smith, 61 F.3d at 819. In other words, we must determine whether grave doubt exists on whether the jury would have unanimously found that Ross committed a third connected violation of the controlled substances laws if a Richardson instruction had been given.

Having conducted a close review of the record, we cannot conclude that the absence of a Richardson instruction had substantial and injurious effect on the jury's verdict in the CCE charge. We have no doubt that the jury would have unanimously concluded that Ross committed a third connected controlled substances offense had a Richardson unanimity instruction been given. At trial, more than fifteen witnesses for the Government testified about Ross's participation in hundreds of controlled substance offenses. That the jury found Ross guilty of the two conspiracy charges evidences that the jury believed the testimony of at least some subset of the Government's witnesses, in spite of Ross's attacks on the witnesses' credibility. We agree with the district court that it is virtually inconceivable that the jury would have unanimously agreed that Ross committed the conspiracy offenses and yet would not have unanimously agreed that Ross committed at least one specific offense of the hundreds of uncharged controlled substance offenses in evidence. Thus, because we cannot discern that, in this case, the Richardson error had a substantial and injurious effect on the jury's verdict in the CCE offense, we affirm the district court's order denying Ross habeas relief.

AFFIRMED.