Requiring Exhaustion for Cumulative Error Review of Harmlessness Does Not Add Up

Brian J. Levy


In January 2014, the Court of Appeals for the Third Circuit became the fifth circuit court to hold that a habeas corpus procedural rule—exhaustion—bars habeas petitioners from advancing all cumulative error arguments when they failed to raise cumulative error in their state court proceedings. Only one circuit—the Fifth Circuit—allows petitioners to raise cumulative error for the first time in his or her habeas appeal. 1 This Essay argues that the plurality has erred.

First, this Essay briefly describes the cumulative error and exhaustion doctrines. Habeas’ exhaustion requirements only apply to “claims.” 2 There are two types of cumulative error arguments. 3 First, there is a standalone claim that state trial errors taken together violated a defendant’s due process rights. 4 Second, there is a cumulative error argument that functions as a standard of review (cumulative prejudice), in which courts can determine whether claims not separately prejudicial together show sufficient prejudice. 5 Because the second argument is not a claim, exhaustion should not apply to it.

Then, this Essay offers three reasons why exhaustion should not prevent habeas petitioners from raising “unexhausted” cumulative prejudice claims in federal court: (1) cumulative prejudice review—unlike the first type of cumulative error—is not a claim; (2) preventing a litigant from arguing cumulative error with regards to prejudice is inequitable; and (3) barring cumulative error arguments does not support “comity,” the policy basis of exhaustion.



The habeas statute declares that federal courts cannot grant habeas to state prisoners unless “the applicant has exhausted the remedies available in the courts of the State.” 6 In turn, the Supreme Court has explained, exhaustion of state remedies means a petitioner “must ‘fairly present’ his [or her] claim in each appropriate state court (including a state supreme court with powers of discretionary review), thereby alerting that court to the federal nature of the claim.” 7 In this description, as in others, the Supreme Court has held that exhaustion is only required for “claims.” 8 The basis for this exhaustion requirement is comity, or respect for the competence and independence of state courts. 9

Habeas Harmless Error Review

For virtually the same policy reasons, 10 courts apply a more deferential standard of review when determining whether petitioners’ claimed errors are harmless in habeas cases than on direct review. 11 On direct review of criminal cases, the “Chapman standard” applies, under which “a federal constitutional error can be considered harmless only if a court is ‘able to declare a belief that it was harmless beyond a reasonable doubt.’” 12 Meanwhile, on habeas review, the standard announced by Brecht v. Abrahamson 13 applies. Under that standard, no habeas relief will be given for a federal constitutional error unless the error “had substantial and injurious effect or influence in determining the jury’s verdict.” 14 To apply either standard, the court weighs the harm of the alleged error against the probative value of all of the evidence against the defendant. 15 The State has the burden of showing an error is harmless. 16

Cumulative Error

Courts have been confused about cumulative error because there are actually two types of “cumulative error” arguments. 17 The first argument is a claim derived from Chambers v. Mississippi. 18 In Chambers, the Supreme Court held that a defendant’s due process right to a “fair opportunity to defend against the State’s accusations” was violated as a result of two state-law evidentiary errors. 19 Those two errors prevented the defendant from presenting evidence to the jury that a third party, who confessed to the crime and then repudiated his confession, had been the real perpetrator because, under the first error, the defendant “was unable . . . to cross‑examine [the third party]” and, under the second error, was prevented from “present[ing] witnesses in his own behalf who would have discredited [the third party’s] repudiation and demonstrated his complicity.” 20 In Chambers, the two evidentiary errors were not separate constitutional claims; rather, the deprivation of due process rights was based on the interaction between two trial errors that “frustrat[ed] [the defendant’s] efforts to develop an exculpatory defense” and violated the defendant’s due process rights. 21

The second type of cumulative error argument (cumulative prejudice) argues that claims should be reviewed together to overcome the State’s harmless error defense. 22 Because the errors raised are usually unrelated, 23 they do not form a specific due process violation as in Chambers. Rather, this argument is meant to overcome the State’s argument that the record as a whole shows harmlessness 24 and the State’s interest in comity, finality, and federalism. 25

The origins of cumulative prejudice are unclear. 26 While the Supreme Court has hinted at cumulative prejudice in footnotes 27 and some habeas courts claimed to derive cumulative prejudice from the direct review in Chambers, 28 the Supreme Court has never used cumulative prejudice in a habeas case. 29

In January 2014, the Court of Appeals for the Third Circuit held that “cumulative error” is a separate “claim,” and therefore the cumulative error “claim” must be exhausted. 30 The Third Circuit’s holding is troubling because its reasoning was essentially limited to citing the plurality of courts that have also held cumulative error is a claim that must be exhausted. 31 In other words, the Third Circuit’s holding may suggest the sheer numerosity of holdings requiring exhaustion has ended the argument. 32 However, there are at least three reasons why these circuits’ holdings with regard to cumulative prejudice should be rejected: (1) cumulative prejudice review—unlike the first type of cumulative error—is not a claim; (2) preventing a litigant from arguing cumulative error with regards to prejudice is inequitable; and (3) barring cumulative error arguments does not support “comity,” the policy basis of exhaustion.

Cumulative Prejudice Is Not a Claim

Because cumulative prejudice is not a claim, it need not be exhausted. Cumulative prejudice is not a claim because, if it were,
(1) there would have to be a Supreme Court holding establishing it, and (2) it would be a claim solely composed of other claims.

Cumulative Error Cannot Be a Claim Because the Claim Is Not “Clearly Established”

Under the habeas statute, a petitioner can only be granted habeas for a legal error if the state court’s decision “was contrary to, or involved an unreasonable application of, clearly established Federal law, as determined by the Supreme Court of the United States.” 33 As discussed above, the Supreme Court has never applied cumulative prejudice in a habeas case. 34 As a result, cumulative prejudice is not “clearly established” under the habeas statute. 35 Therefore, were cumulative prejudice actually a claim, it would be nonactionable. Indeed, commentators and courts have so argued. 36

However, the circuit courts’ opinions assume that cumulative prejudice would be viable in every circuit if exhausted. If cumulative prejudice was a not-clearly established claim, courts could dart around the procedural thicket of exhaustion and dismiss the claim on the merits. Indeed, courts have shown no compunction in dismissing unexhausted claims on the merits. 37

However, cumulative prejudice is clearly established when understood as a necessary corollary to harmless error analysis. Under harmless error analysis, a court is required to determine harmlessness “in light of the record as a whole.” 38 No further Supreme Court guidance is needed to determine that the “record as whole” includes items on both sides of the prejudice scale: showing harmlessness or prejudice. 39 Thus, cumulative prejudice is only viable when a part of harmless error analysis—not as a claim. If it is not a claim, it does not need to be exhausted.

Cumulative Error Cannot Be a Claim Because It Comprises Claims That Are Already Constitutional Errors

Generally, things are not composed of themselves. A combination of items either forms a group or a new whole. 40 Geese form a flock; cells form a person. Many circuit courts have explicitly limited their cumulative prejudice analysis only to alleged federal constitutional violations. 41 Common sense indicates that cumulative prejudice cannot be a claim just like its component errors. Those component errors are the claims. The Fifth Circuit’s outlying holding explained that cumulative error arguments can always be raised so long as the component claims are themselves exhausted, showing that cumulative prejudice merely overlays other claims. 42

Cumulative prejudice does not raise a novel due process claim about other claims; it addresses the harmless error defense to those claims.

Cumulative Error and Harmless Error Are Treated Inequitably

Unreasonably preventing a petitioner from arguing cumulative prejudice inequitably favors the State and allows constitutional errors to justify the denial of relief for other constitutional errors.

First, when courts deny cumulative prejudice review, they inequitably allow the State—but not the petitioner—to cumulate without justification. The State almost always cumulates its evidence. For instance, no rational juror would consider the prosecution’s presentation of an eyewitness, ballistics evidence, and DNA evidence separately to determine whether each proved that a defendant was guilty beyond a reasonable doubt. Indeed, in Brecht itself, the Court had license to rove through the record and cumulate evidence against the petitioner to create a conclusive case and render errors harmless. 43 Justice is a “big picture” concept. 44 It is inequitable for the State to be able to discuss the “big picture” while limiting petitioners to isolated pixels.

Second, when cumulative prejudice is ignored, the State can show a constitutional error is harmless because other constitutional errors contributed to a petitioner’s conviction as the following hypothetical shows. 45 Assume that a harmful error under Brecht—one that “had substantial and injurious effect or influence in determining the jury’s verdict” 46—is an error that is 10% of the cause of the verdict. Because the State has the burden to show harmlessness, 47 the State would then have to show any verdict was more than 90% caused by something other than the alleged error. If there were four unrelated errors that each were causes of 6% of the verdict (e.g., ineffective assistance of counsel, prosecutorial misconduct during the closing argument), the State could show that the verdict was more than 90% likely with respect to each claimed error in isolation even though the verdict would be only 74% caused by non-errors. 48

This potentially fractious relationship between a petitioner’s competing habeas claims—in which the prejudice from other claims prevents any one claim from being singularly, sufficiently prejudicial 49 when cumulative prejudice is denied and the Brecht standard is applicable—is inequitable. When multiple constitutional errors together are prejudicial, a petitioner should not be kept in custody because no single error was the but-for cause of the flawed verdict. Such reasoning has rightly been repudiated in torts, 50 and, a fortiori, should be equally repudiated where an individual’s liberty is at stake.

Requiring Exhaustion Is Contemptuous of State Courts

Requiring cumulative prejudice to be exhausted is not supported by the policy rationale for exhaustion—comity. The Supreme Court has held that exhaustion is a “rule of comity [that] reduces friction between the state and federal court systems by avoiding the ‘unseem[liness]’ of a federal district court’s overturning a state court conviction without the state courts having had an opportunity to correct the constitutional violation in the first instance.” 51 Because cumulative prejudice is not a “constitutional violation,” that goal does not apply literally. More importantly, requiring cumulative prejudice to be exhausted is patronizing rather than respectful. A federal judge who requires cumulative error to be exhausted holds essentially that when a petitioner says that he or she sat down and ate “steak” and “potatoes” the state judge would not understand that the petitioner sat down and ate “steak and potatoes.” Insisting that state judges are limited machines who require “magic words” 52 to process obvious arguments is contemptuous. Because this application of exhaustion runs directly counter to the policy behind exhaustion, there is no policy argument for requiring the exhaustion of cumulative prejudice.


Habeas petitions are often filed by poor, legally ignorant prisoners. While the vast majority of petitions are wholly meritless and deserve quick dismissal, courts should not invent complicated doctrinal traps for the unwary to avoid their Article III duties to consider even meritless petitions. This Commentary has presented three arguments why cumulative prejudice does not need to be exhausted. Hopefully, the reader does not require the author to indicate that those arguments should be considered cumulatively.



  1. See Derden v. McNeel, 978 F.2d 1453, 1454 (5th Cir. 1992) (en banc) (holding cumulative error is available when the constituted “errors were not procedurally defaulted for habeas purposes”).

  2. See infra Part I.A.

  3. See infra Part I.C.

  4. See infra Part I.C.

  5. See infra Part I.C.

  6. 28 U.S.C. § 2254(b)(1)(A) (2012).

  7. Baldwin v. Reese, 541 U.S. 27, 29 (2004) (quoting Duncan v. Henry, 513 U.S. 364, 365–66 (1995) (per curiam)).

  8. Artuz v. Bennett, 531 U.S. 4, 9–10 (2000) (emphasizing that procedural default and exhaustion apply to “claims”).

  9. See, e.g., O’Sullivan v. Boerckel, 526 U.S. 838, 844–45 (1999).

  10. See Fry v. Pliler, 551 U.S. 112, 116 (2007) (relying on “finality, comity, and federalism”).

  11. In addition to general harmless error analysis, most habeas claims will have their own prejudice analyses. See, e.g., Premo v. Moore, 131 S. Ct. 733, 743–45 (2011) (discussing the difficulty in showing prejudice on an ineffective assistance of counsel claim on habeas).

  12. Fry, 551 U.S. at 116 (quoting Chapman v. California, 386 U.S. 18, 24 (1967)).

  13. 507 U.S. 619 (1993).

  14. Id. at 631 (quoting Kotteakos v. United States, 328 U.S. 750, 776 (1946)) (internal quotation marks omitted).

  15. See Marshall v. Hendricks, 307 F.3d 36, 78 (3d Cir. 2002) (“Indeed, in Chapman itself, the cumulative effect of the error was weighed together.”); see also Brecht, 507 U.S. at 629, 638 (applying the harmless error test “in light of the record as a whole”); Chapman, 386 U.S. at 22 (holding that errors may be “so unimportant and insignificant that they may . . . be deemed harmless” “in the setting of a particular case”) (emphasis added).

  16. See Arizona v. Fulminante, 499 U.S. 279, 296−97 (1991).

  17. See Rachel A. Van Cleave, When Is an Error Not an “Error”? Habeas Corpus and Cumulative Error Analysis, 46 Baylor L. Rev. 59, 60–61 (1994) (comparing a situation in which “a court might find . . . several ‘constitutional errors,’ otherwise individually harmless, were collectively harmful” under “cumulation of harmlessness” and a situation in which “numerous errors of state law may operate together to deprive a petitioner of a fair trial under the Due Process Clause”). Most courts apply some form of cumulative error and apply it the same way on direct and habeas review. See John H. Blume & Christopher Seeds, Reliability Matters: Reassociating Bagley Materiality, Strickland Prejudice, and Cumulative Harmless Error, 95 J. Crim. L. & Criminology 1153, 1185 n.117 (2005) (“Except for the Fifth, Sixth, and Eighth Circuits, the federal courts of appeals apply cumulative error analysis, without explicit distinction from the direct-appeal context, in the review of habeas corpus petitions.”).

  18. 410 U.S. 284 (1973).

  19. Id. at 294.

  20. Id.

  21. Id. at 290 n.3.

  22. See, e.g., United States v. Rivera, 900 F.2d 1462, 1469–70 (10th Cir. 1990) (en banc) (“Such an analysis is an extension of the harmless-error rule, which is used to determine whether an individual error requires reversal.”).

  23. See, e.g., Derden v. McNeel, 978 F.2d 1453, 1459–61 (5th Cir. 1992) (en banc) (attempting to cumulate the judge’s “allegedly offensive” remarks, the court’s sustaining four prosecution objections, prosecution’s misconduct during voir dire, prosecution’s eliciting impermissible evidence, and a Brady error).

  24. See Collins v. Sec’y of Pa. Dep’t of Corr., 742 F.3d 528, 540–43 (3d Cir. 2014) (noting petitioner’s argument “that ‘the cumulative error doctrine is a required method of conducting prejudice analysis,’ not a standalone constitutional claim”).

  25. See Brecht v. Abrahamson, 507 U.S. 619, 635 (1993) (identifying comity, finality, and federalism as reasons to find errors harmless on habeas review); Derden, 978 F.2d at 1462 (Higginbotham, J., concurring) (disputing that cumulative error “is an open‑ended threat to comity, finality, and federalism”).

  26. See, e.g., United States v. Smith, 776 F.2d 892, 899 (10th Cir. 1985) (offering no support for its cumulative prejudice analysis); United States v. Berry, 627 F.2d 193, 200–01 (9th Cir. 1980) (similar); cf. Blume & Seeds, supra note 18, at 1185 (“When one sets out to investigate the origins of cumulative harmless error, there is not a lot to find.”).

  27. See Brecht, 507 U.S. at 638 n.9; Taylor v. Kentucky, 436 U.S. 478, 487 n.15 (1978).

  28. See Parle v. Runnels, 505 F.3d 922, 927 & n.5 (9th Cir. 2007) (applying Supreme Court direct review cases to argue cumulative error is clearly established).

  29. See Ford v. Schofield, 488 F. Supp. 2d 1258, 1368 (N.D. Ga. 2007) (“There is no clearly established Supreme Court precedent requiring states to consider the cumulative effect of alleged constitutional errors in order to determine whether a criminal defendant has received due process of law.”); Van Cleave, supra note 18, at 60 (“The United States Supreme Court has not considered the use of cumulative error analysis in habeas corpus petitions and recently denied certiorari in a case which squarely presented these issues.”); cf. Ruth A. Moyer, To Err Is Human, To Cumulate, Judicious: The Need for U.S. Supreme Court Guidance on Whether Federal Habeas Corpus Reviewing State Convictions May Cumulatively Assess Strickland Errors, 61 Drake L. Rev. 447, 452 (2013) (“[T]he cumulation of Strickland errors is not currently clearly established law . . . .”).

  30. See Collins v. Sec’y of Pa. Dep’t of Corr., 742 F.3d 528, 540–43 (3d Cir. 2014); see also Blume & Seeds, supra note 18, at 1192 n.131 (arguing cumulative error must be exhausted).

  31. See Collins, 742 F.3d at 540–43 (noting its own agreement with the Second, Sixth, Ninth, and Tenth Circuits).

  32. Cf. Brian J. Levy, 20 U.S.C. § 1406(b), 62 Buff. L. Rev. 377, 380 (2014) (“[D]octrines . . . reverberate from court to court, judge to judge, and case to case, with each voice giving the doctrines more force and wider application as they resonate.”).

  33. 28 U.S.C. § 2254(d)(1) (2012).

  34. See supra note 30 and accompanying text.

  35. See Moyer, supra note 30, at 474–75 (arguing cumulative error review of Strickland claims is impermissible on habeas because it has not been “clearly established”); Moore v. Parker, 425 F.3d 250, 256 (6th Cir. 2005) (similar).

  36. See supra note 36.

  37. See, e.g., Ballard v. McNeil, 785 F. Supp. 2d 1299, 1335 (N.D. Fla. 2011) (“Notwithstanding Petitioner’s failure to exhaust this claim, the [cumulative error] claim is without merit.”). Were it otherwise, courts create more traps for the unwary. See Diane E. Courselle, AEDPA Statute of Limitations: Is It Tolled when the United States Supreme Court Is Asked To Review a Judgment from a State Post-Conviction Proceeding?, 53 Clev. St. L. Rev. 585, 585 (2005–06) (“[AEDPA’s] procedural rules have created numerous traps for the unwary.”).

  38. See supra note 16.

  39. See Albrecht v. Horn, 485 F.3d 103, 139 (3d Cir. 2007) (holding that Brecht “is the standard applicable here [to analyze cumulative error], because ‘a cumulative‑error analysis merely aggregates all the errors that individually have been found to be harmless, and therefore not reversible, and it analyzes whether their cumulative effect on the outcome of the trial is such that collectively they can no longer be determined to be harmless’”) (quoting Darks v. Mullin, 327 F.3d 1001, 1018 (10th Cir. 2003)).

  40. Cf. Citizens United v. FEC, 558 U.S. 310, 465–67 (2010) (Stevens, J., dissenting) (describing the difference between a corporation and its stakeholders); Theodore Sturgeon, More than Human (1953).

  41. See, e.g., Hayes v. Ayers, 632 F.3d 500, 524 (9th Cir. 2011) (allowing cumulative error only where an “error of constitutional magnitude occurred”); Turner v. Quarterman, 481 F.3d 292, 301 (5th Cir. 2007) (requiring the cumulated errors to be constitutional errors).

  42. See Derden v. McNeel, 978 F.2d 1453, 1458 (5th Cir. 1992) (en banc) (holding that the component errors “must not have been procedurally barred from habeas corpus review.”).

  43. See Brecht v. Abrahamson, 507 U.S. 619, 638–39 (1993).

  44. See Blume & Seeds, supra note 18, at 1154–55 (“A verdict’s reliability cannot sensibly be measured by assessing deficiencies of counsel, prosecutorial misconduct, and any other errors affecting reliability in isolation from one another.”).

  45. Cf. Arizona v. Fulminante, 499 U.S. 279, 307–08 (1991) (holding trial error “may . . . be quantitatively assessed in the context of other evidence” to determine harmlessness).

  46. Brecht, 507 U.S. at 631 (quoting Kotteakos v. United States, 328 U.S. 750, 776 (1946)) (internal quotation marks omitted).

  47. See Fulminante, 499 U.S. at 296−97.

  48. Cf. Blume & Seeds, supra note 18, at 1183–84 (arguing that courts’ categorization of errors as separately prejudicial cause the fractional prejudice of each error to be rounded to zero before adding them together). For more on categorization, see generally Levy, supra note 33, at 422−37.

  49. In other contexts, this has been called the “three stooges effect.” See Gregory Cochran, John Hawks & Henry Harpending, Overdominance and Rapid Adaptation (July 30, 2011),‌20140422133052/http://‌harpending.‌humanevo.‌‌ (retrieved from Internet Archive) (describing “the ‘stooge effect,’” which was named “after the Three Stooges all trying to get through a doorway at the same time,” causing their mutual failure); The Simpsons: The Mansion Family (FOX television broadcast Jan. 23, 2000) (similar).

  50. See, e.g., Summers v. Tice, 199 P.2d 1 (Cal. 1948).

  51. See, e.g., O’Sullivan v. Boerckel, 526 U.S. 838, 844–45 (1999) (quoting Darr v. Buford, 339 U.S. 200, 204 (1950)) (“Comity thus dictates that when a prisoner alleges that his continued confinement for a state court conviction violates federal law, the state courts should have the first opportunity to review this claim and provide any necessary relief.”) (citing Rose v. Lundy, 455 U.S. 509, 515–16 (1982)); Darr, 339 U.S. at 204).

  52. See, e.g., Jimenez v. Walker, 458 F.3d 130, 149 (2d Cir. 2006) (holding that a petitioner did “not give the state court fair notice of a distinct cumulative-error claim” when he asserted that one error “was exacerbated by” another).

  53. ∗  J.D., New York University School of Law. Thanks to Tommy Bennett for his always insightful suggestions and edits. Thanks to Elissa, John, Janet, Andrew, and Debby for their support.