Harvard Law Review Harvard Law Review Harvard Law Review

Harvard Law Review Forum

Patterns of Error

Responding to Justin Murray, A Contextual Approach to Harmless Error Review

Chief Justice Traynor introduced his famous book on the problem of harmless error from the perspective of a state supreme court judge as a harried exterminator. Faced with swarms of appeals, the judge called errors “the insects in the world of law,” with many “plainly harmless,” but others “ominously harmful” that may “mark the way for a plague of followers.”1×1. Roger J. Traynor, The Riddle of Harmless Error ix (1970). To be properly equipped to swat away the stray harmless bug but also smother with pesticides any incoming plague, perhaps a judge must have varied tools to remedy both solitary errors and entire patterns of error.

Harmless error rules, which in theory decide whether trial errors are of sufficient gravity that they can result in relief on appeal or postconviction, have become the “nebulous test of reasonableness” that Chief Justice Traynor feared would be “unlikely to foster uniformity either in the application of standards, should there be any, or in the pragmatic exercise of discretion.”2×2. Id. at 34. In criminal cases, the tests have evolved into a family of doctrines so ornate that, as Justice Scalia has commented, the doctrine creates “ineffable gradations of probability . . . beyond the ability of the judicial mind (or any mind) to grasp.”3×3. United States v. Dominguez Benitez, 542 U.S. 74, 86 (2004) (Scalia, J., concurring in the judgment). Or as Justice O’Connor put it, discussing two standards of harmless error, “each requires an exercise of judicial judgment that cannot be captured by the naked words of verbal formulae.”4×4. Brecht v. Abrahamson, 507 U.S. 619, 656 (1993) (O’Connor, J., dissenting). Has the house, our system of fair trials in the entomological metaphor, become so infested that we need to build a new one?

I write here to comment on a wonderful new Article by Justin Murray, who has ably described a persistent problem: judges have wide discretion when finding error harmless, they do so frequently, and they do so with guidance ill-suited to the task.5×5. Justin Murray, A Contextual Approach to Harmless Error Review, 130 Harv. L. Rev. 1791, 1794–95 (2017). Murray’s Article contributes to a decades-long literature bemoaning the state of harmless error review6×6. See, e.g., Traynor, supra note 1; Brent M. Craig, “What Were They Thinking?” — A Proposed Approach to Harmless Error Analysis, 8 Fla. Coastal L. Rev. 1, 6–12 (2006); Anne Bowen Poulin, Tests for Harm in Criminal Cases: A Fix for Blurred Lines, 17 U. Pa. J. Const. L. 991, 994 (2015); see generally Roger A. Fairfax, Jr., A Fair Trial, Not a Perfect One: The Early Twentieth-Century Campaign for the Harmless Error Rule, 93 Marq. L. Rev. 433, 433–34 (2009) (providing a historical overview of the evolution of the modern harmless error rule). with a proposal to change the tools of the trade. Murray recommends that judges conduct what he calls a “contextual approach to harmless error review.”7×7. Murray, supra note 5, at 1791. Using such an approach, a judge would not just ask whether an error contributed to the outcome, but also the implications of an error for the broader “constellation of interests” served by a particular rule.8×8. Id. at 1811. Murray suggests that these broader considerations may explain at least some of what courts currently do, and that this approach would be less result-based and more sensitive to underlying constitutional values.

In this Response, I will first ask: why harmless error? Today, judges use many more tools aside from harmless error rules to fend off plagues of errors. I ask whether this proposal comes several decades too late. Harmless error is not as relevant a doctrine as it used to be, now that appellate and postconviction law has become more ornate. Indeed, much of the result-based doctrine has been incorporated into the substance of key constitutional rights. Being more sensitive to underlying constitutional values would require changing constitutional doctrine and not just harmless error rules.

Second, I ask whether the proposed cure will actually help judges to differentiate between the harmless and the harmful. The proposal risks magnifying cognitive bias. Encouraging judges to confirm results, as current doctrine does, is troubling, but a contextual approach might not reduce that problem and it could separately liberate judges to further rely on their own value judgments. Murray is right to be cautious about the implications of the proposal.

Third, I suggest that we redirect our attention to the bulk of the swarm. Today, almost all criminal cases are plea bargained. I ask whether this proposal might instead prove more useful in an area Murray does not address: reviewing errors in cases that are plea bargained.

Fourth, I turn to the source of the swarm itself. I examine the potential use of judicial review to examine patterns of errors in the aggregate using mechanisms like consolidation of cases. Our system not only poorly identifies errors in individual cases, but may also systematically deny relief to errors that do affect results, as well as fairness, dignity, and other interests. I applaud the effort to begin a conversation about how to rethink our system of appeals and postconviction review. Unfortunately, harmless error is just one of a host of related problems and other doctrines have eclipsed it in importance. The infestation may be so serious that we need to rebuild from the ground up.

I. Error in a Changing Appellate and Postconviction System

When Chief Justice Traynor sounded his call for better tools to police errors in criminal appeals, the number of prisoners and number of claims they raised in the courts were a small fraction of what they are today. Appellate and postconviction doctrine, as well as constitutional criminal procedure, were also simpler then. Back then, rethinking harmless error, as Chief Justice Traynor tried to do and as Murray recommends, was a sensible response to the challenge of appellate review. Today, practical considerations and changes in constitutional doctrine and statutory restrictions make harmless error a far less relevant tool. The swarm is too large. Judges have had their hands tied in too many new ways.

We are now a mass incarceration nation. Despite the millions of convicts in our country, or perhaps because we now have so many convicts, any review at all of a conviction is an unusual event in the United States. Each year, over a million people are convicted of felonies and many more of misdemeanors.9×9. Bureau of Justice Statistics, U.S. Dep’t of Justice, Felony Sentences in State Courts, 2006 — Statistical Tables 1 (2009), https://www.bjs.gov/content/pub/pdf/fssc06st.pdf [https://perma.cc/K95Q-36GT] (estimating over 1.1 million individuals convicted of felonies in state courts in 2006). Perhaps 70,000 appeals are filed each year, chiefly in felonies with lengthy sentences.10×10. Nicole L. Waters et al., Bureau of Justice Statistics, U.S. Dep’t of Justice, Criminal Appeals in State Courts 1 (2015), https://www.bjs.gov/content/pub/pdf/casc.pdf [https://perma.cc/65TA-23ZY]; Michael Heise, Nancy J. King & Nicole A. Heise, State Criminal Appeals Revealed, 70 Vand. L. Rev. (forthcoming 2017). The vast majority of offenders plead guilty and waive or lose the right to an appeal or to postconviction review;11×11. See 5 Wayne R. LaFave et al., Criminal Procedure § 21.6(a) (4th ed. 2015); see also id. § 21.5(b). in addition, any relief could result in a worse bargain and sentence following the reversal.12×12. Padilla v. Kentucky, 559 U.S. 356, 373 (2010) (“Those who . . . attack their guilty pleas lose the benefit of the bargain obtained as a result of the plea. . . . [U]ltimately, the challenge may result in a less favorable outcome for the defendant . . . .”). For cases, typically more serious ones with long sentences, that do receive an appeal, few receive any relief, with a success rate of 2.8% according to one study of state supreme court discretionary review of direct appeals.13×13. Heise, King & Heise, supra note 10. Reversals during state post-conviction review and federal habeas corpus are far less frequent.14×14. Nancy J. King, Fred L. Cheesman II & Brian J. Ostrom, Final Technical Report: Habeas Litigation in U.S. District Courts: An Empirical Study of Habeas Corpus Cases Filed by State Prisoners Under the Antiterrorism and Effective Death Penalty Act of 1996, at 19–20 (2007), https://www.ncjrs.gov/pdffiles1/nij/grants/219558.pdf [https://perma.cc/5CJ5-L9GQ]. In assessing whether a claim, particularly a federal constitutional claim, deserves relief, judges must mentally replay a trial they did not attend. It is no easy task.

A brief timeline-sketch of the changes to harmless error doctrine suggests the direction of change has been consistently against broadening the scope of judicial review, as Murray proposes, but instead has trended toward restricting judicial review. Judges and scholars have been grappling with the problem of harmless error since well before Chief Justice Traynor’s time. In Kotteakos v. United States,15×15. 328 U.S. 750 (1946). the Supreme Court held, in the context of federal criminal appeals, that any presumptions regarding how to apply harmless error rules should arise “from the nature of the error and ‘its natural effect’ for or against prejudice in the particular setting.”16×16. Id. at 765–66. That sounds like a context-specific harmless error test, but more specific language stated that only an error substantially affecting the right should result in a reversal.17×17. Id. at 765. Decades after Kotteakos, the Court in 1967 adopted a more protective test for constitutional rights in Chapman v. California;18×18. 386 U.S. 18 (1967). but the Court has since retrenched, separating trial errors and “structural defects,”19×19. Arizona v. Fulminante, 499 U.S. 279, 309 (1991); see also id. at 309–11. and adopting specific and less inmate-friendly tests incorporated into the substance of constitutional rights. In 1993, the Supreme Court adopted in Brecht v. Abrahamson20×20. 507 U.S. 619 (1993). a standard more forgiving of constitutional errors for federal habeas actions.21×21. See id. at 623. Scholars debated whether Brecht would have a negative effect on federal habeas corpus.22×22. See, e.g., John H. Blume & Stephen P. Garvey, Harmless Error in Federal Habeas Corpus After Brecht v. Abrahamson, 35 Wm. & Mary L. Rev. 163 (1993). However, the rules soon tightened further. Congress enacted the Antiterrorism and Effective Death Penalty Act of 199623×23. Pub. L. No. 104-132, 110 Stat. 1214 (codified as amended in scattered sections of the U.S. Code). (AEDPA), which contained limitations on granting relief for constitutional violations in state criminal trials. In 2007, the Court in Fry v. Pliler24×24. 551 U.S. 112 (2007). ruled that the AEDPA requires a new level of deference unless a state court’s denial of relief under Chapman is itself “unreasonable” under 28 U.S.C. § 2254(d)(1).25×25. Id. at 119–20.

That trend toward more restrictive harmless error review might give the impression that narrowing harmless error was an important mechanism for regulating the review of criminal convictions. That impression would be wrong. The restrictions in harmless error review are merely symptomatic of the larger judicial and statutory approach toward stripping judicial review in criminal cases. Today, a wide range of procedural obstacles typically make it unnecessary for a judge to even reach the question whether a claim has any merit. Harmless error rules lie below many layers of procedural sediment, statutes of limitations, procedural default rules, nonretroactivity rules, standards limiting the evidentiary record relevant to a claim, and standards limiting access to relief.26×26. See, e.g., Strickland v. Washington, 466 U.S. 668, 688, 694 (1984). During appellate review, courts often deny relief summarily, without issuing any rulings with reasons. During postconviction review, federal courts now apply the AEDPA, which more broadly restricts relief on the merits.27×27. 28 U.S.C. § 2254(d) (2012). See generally Brandon L. Garrett & Lee Kovarsky, Federal Habeas Corpus: Executive Detention and Post-Conviction Litigation ch. 5 (2013). Murray describes some of these rules, setting out different standards for granting relief, but not doing so in detail, as the Article generally discusses “result-based” standards.28×28. Murray, supra note 5, at 1795.

Second, improving harmless error review would not affect rulings on many of the key constitutional criminal procedure rights that inmates rely on. That is because the Supreme Court has interpreted a range of constitutional criminal procedure rules to incorporate harmless error–type reasoning or tests. This creates what I have called a second layer of harmless error rules.29×29. For a discussion of this general phenomenon, see Brandon L. Garrett, Innocence, Harmless Error, and Federal Wrongful Conviction Law, 2005 Wis. L. Rev. 35, 62. They place the burden of showing harm on the convict. Such rules include those for ineffective assistance of counsel,30×30. Brady v. Maryland, 373 U.S. 83 (1963). and the right to be free from suggestive identification procedures.31×31. See Garrett, supra note 29, at 69–88. For such commonly litigated claims, harmless error analysis is not conducted. Murray’s proposal would not affect such claims. Moreover, such rulings provide a cautionary tale. It is no accident that the Supreme Court used constitutional rulings to heighten the standards for finding error harmless in precisely the areas in which post-trial claims are most heavily litigated.32×32. See Joseph L. Hoffmann & Nancy J. King, Essay, Rethinking the Federal Role in State Criminal Justice, 84 N.Y.U. L. Rev. 791, 815–17 (2009) (discussing the debilitating effect of habeas petitions on judicial resources). If judges are free to analyze harm interests, they may do so reflecting their own institutional interests, such as those based on which claims are more burdensome or more frequently brought.

Third, federal habeas corpus rules add an additional layer of deference, consisting of the Brecht rule and the AEDPA restrictions. Regarding the AEDPA, the Supreme Court held in Cullen v. Pinholster33×33. 563 U.S. 170 (2011). that the Act requires “doubly deferential” review of the “prejudice” prong of an ineffective assistance of counsel claim.34×34. Id. at 189–90. Thus, at the federal level, harmless error inquiry is already so deferential that it is not clear that Murray’s more context-sensitive approach would be compatible or make much difference. To be sure, during state direct appeals, such rules do not apply. But for all of these reasons, harmless error doctrine does not do the work that it did decades ago when harmless error was a more common subject for scholarship.

Finally, exceptions to result-based harmless error review presently exist in the doctrine, as Murray notes, but they, too, provide a cautionary tale. For example, the Supreme Court has noted that Brecht left open “an exception” for cases in which “a deliberate and especially egregious error of the trial type, or one that is combined with a pattern of prosecutorial misconduct, . . . infect[s] the integrity of the proceeding.”35×35. Fry v. Pliler, 551 U.S. 112, 117 (2007) (quoting Brecht v. Abrahamson, 507 U.S. 619, 638 n.9 (1993)). Perhaps that exception could provide a doctrinal foothold for the approach Murray recommends. Lower courts have tended to view the prejudice exception as a hybrid category designed for situations in which unusual prosecutorial misconduct, for example, makes it difficult to conduct a prejudice inquiry.36×36. For examples of lower courts employing the Brecht prejudice exception, see, for example, United States v. Bowen, 799 F.3d 336, 339 (5th Cir. 2015) (finding that “novel and extraordinary” reasons, including prosecutorial misconduct in the form of online statements about the case, supported trial reversal of conviction); Burgess v. Dretke, 350 F.3d 461, 471 (5th Cir. 2003) (describing such errors as hybrid errors, sharing attributes of structural and trial errors); United States v. Harbin, 250 F.3d 532, 545 (7th Cir. 2001) (stating that such error requires automatic reversal); and Hardnett v. Marshall, 25 F.3d 875, 879–80 (9th Cir. 1994) (suggesting that such error would not be subject to harmless error review). The Supreme Court has also adopted tests for harmless error geared specifically for death sentencing cases.37×37. Sawyer v. Whitley, 505 U.S. 333 (1992). That test does not suggest that it is a good idea to create context-specific harmless error tests; the test heavily weighs the state interest in nonreversal of death sentences by selecting a very difficult to meet “clear and convincing evidence” standard. Courts have also adopted tests for when and whether new evidence of innocence can justify relaxing otherwise applicable procedural bars in a range of habeas contexts; those tests have been criticized as insufficiently results-protective.38×38. For an overview, see generally Brandon L. Garrett, Claiming Innocence, 92 Minn. L. Rev. 1629 (2008). Similarly, the Supreme Court’s extra-deferential Brecht harmless error test for federal habeas corpus was premised on a larger institutional interest argument, reasoning that federalism values support deference to state courts that err.39×39. Brecht, 507 U.S. at 638–39.

However, Murray would not have courts focus on institutional or state interests. Murray concludes the proposal is “tentative” and more would have to be done to explore whether and how it could realistically be implemented.40×40. Murray, supra note 5, at 1826. But these examples suggest there are pitfalls to substituting neutral harmless error tests for standards that the courts may prefer to adopt to fit their institutional priorities.

II. Context Bias and Contextual Approaches to Error

There is a broader concern that the cure might be worse than the disease. Sometimes simpler rules function better than more nuanced rules. What is the evidence that a more all-things-considered approach would produce better results? By what standard should an approach be judged? As just discussed, importing substantive concerns into remedial questions can itself blend the distinction between right and remedy. Such interest analysis may magnify the discretion that so troubled Chief Justice Traynor.

There are good reasons to fear that more substance-oriented review would be more error prone, or that at minimum, it would not address the problem. If the current problem is that judges unduly focus on results, rather than on the underlying values in constitutional law, the question is how does one help judges focus on information apart from the result, that is, the conviction? Providing more contextual information about a case may not help. In fact, contextual information may itself be a driver of the problematic outcomes one observes in harmless error rulings. Professors Keith A. Findley and Michael S. Scott have described how “[u]nder [the harmless error] doctrine, cognitive biases can contribute in powerful ways to a conclusion that the defendant was indeed guilty, and that the error was therefore harmless.”41×41. Keith A. Findley & Michael S. Scott, The Multiple Dimensions of Tunnel Vision in Criminal Cases, 2006 Wis. L. Rev. 291, 350. Scholars in law and psychology have described how hindsight bias, confirmation bias, and outcome bias may all encourage judges to affirm what the trial factfinder already concluded.42×42. See Stephanos Bibas, The Psychology of Hindsight Bias and After-the-Fact Review of Ineffective Assistance of Counsel, 2004 Utah L. Rev. 1, 1–5; Gregory Mitchell, Comment, Against “Overwhelming” Appellate Activism: Constraining Harmless Error Review, 82 Calif. L. Rev. 1335 (1994); Dan Simon, A Third View of the Black Box: Cognitive Coherence in Legal Decision Making, 71 U. Chi. L. Rev. 511, 578–80 (2004). It is difficult for judges to ignore the fact that the defendant was found guilty by the factfinder.

Moreover, certain types of evidence, like confession evidence, can in theory be harmless but strongly impact appellate judges, who may be likely to disregard constitutional error and admit the evidence. Studies have shown that judges will find highly coerced confessions, improperly admitted into evidence, as nevertheless highly probative of guilt.43×43. D. Brian Wallace & Saul M. Kassin, Harmless Error Analysis: How Do Judges Respond to Confession Errors, 36 Law & Hum. Behav. 151, 156 (2012). Certain types of statistical and forensic evidence, which jurors may incorrectly interpret depending on how that evidence is presented to them, may similarly lead to errors in appellate judging.44×44. Jonathan J. Koehler, When Are People Persuaded by DNA Match Statistics?, 25 Law & Hum. Behav. 493, 493–95 (2001). Furthermore, one sees in rulings how judges frequently discount defense theories, view evidence that would be contrary to the prosecution’s theory as merely cumulative, and view the prosecution’s case as “overwhelming.”45×45. Poulin, supra note 6, at 1024–30; see also Mitchell, supra note 42, at 1343.

Indeed, not only may judges unduly focus on results at the expense of procedural fairness, but they may also deny relief to actually innocent individuals for whom the wrong result was in fact reached. I have described how judges found “overwhelming” evidence of guilt in cases where the convict was later shown by DNA testing to be actually innocent.46×46. Brandon L. Garrett, Convicting the Innocent: Where Criminal Prosecutions Go Wrong 202 (2011); Brandon L. Garrett, Judging Innocence, 108 Colum. L. Rev. 55, 95 (2008).

Providing judges with a standard that is ostensibly less result-based would not obviously help to counteract the tendency to focus on case-specific contextual information. Indeed, as Murray notes, statutes providing for harmless error review and the Supreme Court have long been clear that the question is whether the error in question contributed to the outcome. The question is not the guilt-based inquiry whether, pretending the tainted evidence simply did not exist, there was nevertheless sufficient evidence for a jury to convict. If the latter were the question, any error could disappear, even if it did contribute reasonably or even substantially to the outcome. Chief Judge Edwards has criticized the slippage into that latter guilt-based reasoning in his classic article on the harmless error standard.47×47. Harry T. Edwards, To Err Is Human, but Not Always Harmless: When Should Legal Error Be Tolerated?, 70 N.Y.U. L. Rev. 1167 (1995). The Court has remained clear that such guilt-based reasoning is improper, stating, for example: “The inquiry . . . [is] whether the guilty verdict actually rendered in this trial was surely unattributable to the error.”48×48. Sullivan v. Louisiana, 508 U.S. 275, 279 (1993) (emphasis omitted).

How to combat that slippage from result-based to guilt-based harmless error review? Murray agrees commentators are right to call the guilt-based approach improper, but also describes how courts continue to slip into such analysis.49×49. Murray, supra note 5, at 1801–04. Telling judges to broaden their focus is unlikely to help. It does not address the underlying structural problems, including the forms of cognitive bias that would make higher-court judges likely to affirm convictions. Instead, as I discuss in the final Part, we need to give judges better information about errors. We need to help them to better identify the harmful pests rather than just instruct them to do a better job.

Still more troublingly, in decades of constitutional criminal procedure rulings, the courts have generally been less willing to use constitutional remedies to provide relief to defendants who do not personally “deserve” relief, because their cases were not causally harmed by the errors. Some might also argue that if harmless error doctrines did not create a safety valve, courts might separately interpret substantive rights in a manner still more deferential to law enforcement or government. Moreover, non-results-oriented constitutional interests have available nonexclusionary remedies. To be sure, courts have eroded access to damages and injunctive remedies as well. But those remedies may more directly address underlying violations by law enforcement than by tinkering with harmless error.

III. Plea Bargaining and Harmlessness

We have a criminal justice system of plea bargaining. Thus, to tackle the bulk of the swarm of errors in criminal justice, today our judges must engage with errors in plea bargaining. What is the importance of harmless error doctrine in the cases that do not go to trial? Typically, plea deals include a waiver that largely bars any appeal. Murray briefly notes rules forbidding prosecutors from breaking promises in plea deals50×50. Id. at 1813. but otherwise does not discuss plea bargaining in his Article.

Nevertheless, the question of how to review errors that occur in plea bargaining is still highly relevant and pressing. Courts have begun to ask whether plea bargains should be totally immune from review. Claims such as ineffective assistance of counsel can now be brought to challenge certain plea bargains.51×51. E.g., Padilla v. Kentucky, 559 U.S. 356, 374 (2010). Scholars have written careful work grappling with how to assess the effect that ineffective assistance can have on a plea.52×52. See, e.g., Heidi Altman, Prosecuting Post-Padilla: State Interests and the Pursuit of Justice for Noncitizen Defendants, 101 Geo L.J. 1, 38–40 (2012); Jenny Roberts, Proving Prejudice, Post-Padilla, 54 How. L.J. 693, 733 (2011). The scholarship also addresses whether some errors should not be waivable, such as effective assistance of counsel and innocence.53×53. Samuel R. Wiseman, Waiving Innocence, 96 Minn. L. Rev. 952 (2012).

The types of conceptual or interest-based distinctions that Murray focuses on may be quite helpful in the context of plea bargaining. After the fact, it can be difficult to assess what might have reasonably affected a plea negotiation. Moreover, the relevant harms cannot be thought of purely in terms of likely sentences, but must also include collateral consequences. In general, courts have tended to identify stark violations, such as failure to communicate a plea offer or a substantial collateral consequence to a client.54×54. E.g., Missouri v. Frye, 566 U.S. 134, 143 (2012); Lafler v. Cooper, 566 U.S. 156, 160 (2012). Murray’s Article could suggest tools for addressing such difficult questions. How larger values of fairness or noncoercion intersect with the interests that prosecutors may assert raises difficult questions, but the ability of an interest analysis to untangle such problems would be its test. We must adopt a contextual approach to error in plea bargaining.

IV. Identifying Patterns of Error

To tackle the growing swarms of error that our criminal justice system produces, we need to rethink the fundamental tools that judges have been given. We need to give clearer direction to lower courts so that they can avoid making errors in the first instance. Those detection and deterrence goals of trial and appellate review have been neglected.

To remedy those defects, some scholars have made the opposite suggestion from Murray’s complication of harmless error doctrine: simplifying harmless error standards. Professor John Greabe has recently argued for a “simplified, unitary, and transcontextual harmless-error test,” like what Chief Justice Traynor advanced.55×55. John M. Graebe, The Riddle of Harmless Error Revisited, 54 Hous. L. Rev. 59, 60 (2016). However, as I have described in this response, I worry that the differences between such doctrinal interventions miss the underlying problem. I find far more attractive proposed alternatives to the current system.

One source of solutions has been to improve information judges collect at the trial stage, both to incentivize better decisionmaking at that stage and to improve the quality and ease of making decisions on appeal. Trial judges could give jurors more detailed specialized verdict forms.56×56. Craig, supra note 6, at 16. For guilty pleas, producing a more detailed factual basis or sentencing analysis might similarly permit more meaningful appellate review.57×57. Cf. McCarthy v. United States, 394 U.S. 459, 467 (1969) (stating that a district judge’s interrogation of the defendant at a plea colloquy “not only facilitates his own determination of a guilty plea’s voluntariness, but . . . also facilitates that determination in any subsequent post-conviction proceeding”).

Other solutions focus on improving avenues to raise nonrecord claims and evidence.58×58. See, e.g., Keith A. Findley, Innocence Protection in the Appellate Process, 93 Marq. L. Rev. 591, 633 (2009). Trial records could be opened on appeal as Professor Eve Brensike Primus has suggested.59×59. Eve Brensike Primus, Structural Reform in Criminal Defense: Relocating Ineffective Assistance of Counsel Claims, 92 Cornell L. Rev. 679, 682–83 (2007). Some courts have created sentencing-only review procedures.60×60. E.g., Conn. Gen. Stat. § 51–195 (2017). Improving the quality of assistance on appeal and postconviction may similarly have far more of an impact on the quality of decisionmaking. Others have suggested changes to the test to permit more independent judicial thinking. Innocence Commissions, for example, like those that exist in North Carolina and the United Kingdom, systematically investigate convictions to assess whether new evidence of innocence supports reversing a conviction.61×61. For a comparative discussion, see Brandon L. Garrett, Towards an International Right to Claim Innocence, 105 Calif. L. Rev. (forthcoming 2017).

Where patterns of error affect not just one but dozens or even hundreds of cases, the question whether error requires a do-over in any one case may be less pressing than how to remedy the ongoing source of error. A common sentencing error may be corrected across the board without the resource concerns that might cause a judge to fear the costs of a retrial. When forensic errors have come to light, affecting hundreds or even many thousands of cases, courts have developed systems for not only reopening cases tainted by lab errors, but also for resolving harmless error issues.62×62. For an evaluation and comparison of such mechanisms, see Brandon L. Garrett, Bad Hair: The Legal Response to Mass Forensic Errors, 42 Litig. 32 (2016). They have, for example, stated that there should be a presumption that any case affected by forensic error should be reversed, with the burden on prosecutors to show why it should not.63×63. Commonwealth v. Scott, 5 N.E.3d 530, 545 (Mass. 2014) (holding that the defendant is entitled to “a conclusive presumption that egregious government misconduct occurred in the defendant’s case” based on revelations of misconduct at drug laboratory). Further, for plea bargained cases, courts have set out presumptions that new sentences should be lower so that prosecutors do not respond to a reversal by insisting on a harsher plea deal.64×64. Bridgeman v. Dist. Attorney, 30 N.E.3d 806, 830 (Mass. 2015) (adopting rule that a defendant granted a new trial due to misconduct at the crime lab “cannot be charged with a more serious offense” than that to which the defendant previously pleaded guilty, and if convicted again, “cannot be given a more severe sentence than that which originally was imposed”).

A further response to the inefficiency of individual case resolution is to develop better mechanisms to aggregate criminal cases. Just as class actions or judicial consolidation tools can permit more efficient resolution of large numbers of tort claims, aggregation in criminal law can remedy patterns of error that might otherwise go unremedied.65×65. For an extended exploration of that topic, see Brandon L. Garrett, Aggregation in Criminal Law, 95 Calif. L. Rev. 383 (2007). Perhaps Chief Justice Traynor’s image of a solitary appellate judge reviewing errors on a case-by-case basis is itself problematic. Such individualized review cannot easily cope with a swarm. Using aggregate mechanisms to address patterns of error can prevent a plague from developing in the first instance.


If Chief Justice Traynor felt somewhat daunted by the swarms of errors confronting higher-court judges, today the swarms have exploded in size while the tools wielded by judges have been tied down or discarded. The result has been a wave of wrongful convictions. The judges themselves have been stung, although nowhere nearly as badly as the convicts who suffered the errors. Entire court systems have had to cope with not just routine errors, but plagues of systemic errors of all stripes, from sentencing errors to forensic errors. Still more troubling is the manner in which judicial review, including harmless error rulings, can disguise the existence and extent of such problems. Unfortunately, far more elaborate rules now restrict judicial remedies. An entire system of plea bargaining makes relief a remote possibility for most convicts. Work like Murray’s is important because we need out-of-the-box solutions to make review of criminal convictions more meaningful. Our current approach is complex, burdensome, and ineffectual. The time for rethinking trial rules, criminal appeals, and postconviction review has long come due.

* Justice Thurgood Marshall Distinguished Professor of Law, University of Virginia School of Law.