Plain Versus Harmless Error

Plain Versus Harmless Error

Rule 52 handles an error that may not have changed the result in two ways, and the defendant's conduct at trial decides which way applies. Subsection (a) directs that any error, defect, irregularity, or variance that does not affect substantial rights must be disregarded, and it governs errors the party preserved with a timely objection.¹ Subsection (b) allows a court to notice a plain error that affects substantial rights even though it was never brought to the trial court's attention, and it governs errors the party forfeited.² The four steps taught above are therefore the machinery of subsection (b) alone. A preserved error goes to the harmless-error inquiry instead, where a different standard applies and the burden of persuasion sits with a different party.

The two subsections share a phrase, which is why the comparison has to be drawn carefully. In United States v. Olano the Court observed that the third limitation on plain-error review uses the same language as Rule 52(a), and that in most cases the language means the error was prejudicial, in the sense that it affected the outcome of the proceedings.³ When the defendant objected, the court of appeals examines the record in a specific harmless-error analysis to see whether the error was prejudicial. Rule 52(b) calls for the same kind of inquiry with one important difference: the defendant, not the government, bears the burden of persuasion on prejudice.⁴ The Court found the reason in the wording, because Rule 52(a) precludes correction only if the error “does not” affect substantial rights, while Rule 52(b) authorizes a remedy only if the error “does” affect them.⁵ Read that way, the party who wants an error disregarded under subsection (a) must show that it had no effect, and the party who wants a remedy under subsection (b) must show that it did.

The allocation also tracks who is in the best position to speak to the record. A timely objection tells the trial judge and the opposing party that a problem exists while it can still be fixed, and it leaves a record on which the party who benefited from the error can explain why the error did not matter. A defendant who stayed silent let the error stand, and a rule that placed the same burden on the government whether or not the defendant spoke would remove any reason to object at trial. Evidence Rule 103 builds the same two tracks: a timely objection stating the specific ground preserves the claim for the more forgiving harmless-error standard, and the absence of one leaves only plain-error review.⁶

The vocabulary of appellate review follows the same line. A preserved error that probably affected the judgment is reversible error, a preserved error that probably did not is harmless error, and a forfeited error clear and serious enough to be corrected despite the silence is plain error.⁷ The labels do not move together, because a forfeited error that clears the prejudice step must still pass the fourth step, so in theory an error that would require reversal if preserved will not always require it under plain-error review.⁸ The three subsections below take the comparison in order: the Kotteakos standard for ordinary preserved errors, the stricter Chapman standard for preserved federal constitutional errors, and the reasons a silent defendant carries the burden of showing prejudice.

The two versions differ in one fact, the objection, and the result flips only because the record is evenly balanced, which is the situation in which the allocation of the burden decides the case. Where the record shows clearly that the error made no difference, the outcome is the same under either subsection. Where it shows clearly that the error mattered, the defendant prevails under either, subject in the forfeited case to the remaining plain-error steps. Students go wrong when they assume the burden is the same because both subsections speak of substantial rights, since Olano located the difference in the wording of the two subsections.⁹

Harmless Error Under 52(a)

Rule 52(a) is the baseline against which plain-error review is measured, and its command is short: any error, defect, irregularity, or variance that does not affect substantial rights must be disregarded.¹¹ The word must leaves a reviewing court no discretion to reverse for a harmless mistake, because Rule 52 binds federal courts as firmly as a statute does.¹² In Bank of Nova Scotia v. United States, a grand jury case, the Court accordingly held that a federal court may not use its supervisory power to circumvent the harmless-error inquiry that the Rule prescribes.¹³ Subsection (a) governs errors the party preserved with a timely objection, so the live question is whether substantial rights were affected, and in most cases that phrase is read to ask what the error did to the outcome.¹⁴

The leading account of how to ask that question comes from Kotteakos v. United States, which construed the 1919 harmless-error statute that Rule 52(a) later absorbed.¹⁵ If, after pondering all that happened, the reviewing court is sure the error did not influence the jury or had but very slight effect, the judgment stands.¹⁶ If the court cannot say with fair assurance, without stripping the erroneous action from the whole, that the judgment was not substantially swayed by the error, then it cannot conclude that substantial rights were unaffected, and a court left in grave doubt must let the conviction fall.¹⁷ The inquiry cannot be merely whether enough evidence supported the result apart from the phase the error touched, because the question is whether the error itself had substantial influence.¹⁸ That framing protects the jury's role. A reviewing court that asked only whether the untainted evidence sufficed would be deciding guilt for itself, and the Court directed instead that the error be judged by its effect on the jurors, in the total setting of the trial, and not by the reviewing judge's own reaction.¹⁹ The federal courts have accordingly rejected the so-called correct-result approach in favor of an evaluation of the error's probable effect on the actual decision.²⁰

The facts of Kotteakos show the standard working in the defendant's favor, and the Court's own comparison to an earlier case shows which fact moved the answer. The indictment charged one conspiracy, but the proof showed at least eight separate ones, with thirty-two defendants at the outset and thirteen by the time the case reached the jury.²¹ The court of appeals had examined the evidence, found guilt manifest, and called reversal a miscarriage of justice, yet the Supreme Court disagreed, because the right at stake was the right not to be tried en masse for a conglomeration of distinct offenses committed by others.²² In Berger v. United States an error of the same abstract character, a variance between one charged conspiracy and the proof of two, had been held harmless, since only four people were involved in all.²³ Scale was the difference: with so many defendants and transactions, a defendant tied to one or a few of them could not realistically guard against evidence about the others spilling over.²⁴

The opposite result appears when the record shows the error was inert. In Hamling v. United States the district court had refused to let counsel object to the jury instructions outside the jury's presence, as Rule 30 requires, and the Court declined to adopt a rule that any such violation requires reversal.²⁵ Reading Rule 30 together with Rule 52(a), the Court examined the record and found no prejudice: the bench conference was one of many, nothing indicated that the jury heard it, and it concerned purely legal questions.²⁶ Both cases show that the same label, procedural or evidentiary error, does not settle the result, because Rule 52(a) turns on the effect shown by the record as a whole.

The statute that Kotteakos construed answered a specific complaint about appellate practice. Congress adopted it in 1919 after long agitation over criminal appeals, when courts of review were criticized for letting technicality control outcomes and the threat of reversal was so great in many jurisdictions that a trial could become a game of planting reversible error in the record in order to win a second one.²⁷ The Court described the aim as substituting judgment for the automatic application of rules, so that review would remain a check on arbitrary action and essential unfairness without handing fairly convicted defendants the loopholes that a rigid scheme of errors creates.²⁸ That aim explains both halves of the standard. An error that left the judgment unswayed is disregarded, which closes the loophole, and the test is a judgment about the whole record rather than a catalogue of forbidden mistakes, which keeps review meaningful. Because judgment replaced formula, the Court also declined to announce a generalized presumption for every case and tied the burden instead to whether the error is technical or naturally prejudicial.²⁹

Since Kotteakos disclaims a blanket presumption, the working allocation for ordinary preserved errors comes from United States v. Olano. When the defendant made a timely objection, the court of appeals normally analyzes the district court record in a harmless-error inquiry to decide whether the error was prejudicial, and Rule 52(b) calls for the same inquiry with the burden of persuasion reversed, so that it falls on the defendant instead of the government.³⁰ The Court tied that reversal to the wording of the two subsections: Rule 52(a) precludes correction only if the error does not affect substantial rights, so the party who wants the error disregarded must show that it had no such effect.³¹ Hamling left open whether a court of appeals could instead require the defendant to demonstrate prejudice from a Rule 30 violation, so the baseline allocation is taken from Olano and not from Hamling.³² Kotteakos itself left room for an exception when the departure is from a constitutional norm or a specific command of Congress, which is why the next subsection turns to preserved constitutional errors.³³

Students lose this issue when they ask only whether the other evidence was enough to convict, because the Rule 52(a) inquiry looks to the error's own influence on the judgment and an affirmance resting on sufficiency alone skips that question.³⁴

Constitutional Harmless Error

Rule 52(a) tells a reviewing court to disregard preserved error that does not affect substantial rights, but how hard the court must look before calling an error harmless depends on the kind of error the defendant objected to. When the preserved error violates the federal Constitution, the Court set a demanding test in Chapman v. California: before a federal constitutional error can be held harmless, the court must be able to declare a belief that it was harmless beyond a reasonable doubt.³⁶ The burden of reaching that belief does not rest on the defendant. It falls on the beneficiary of the error, ordinarily the government, who must show that the error did not contribute to the verdict.³⁷ The standard governs errors the defendant preserved and then raised on direct appeal, so it is the constitutional counterpart of the Kotteakos inquiry just discussed, with a more demanding threshold for the party defending the judgment.³⁸ It says nothing yet about a defendant who stayed silent, a gap this section marks at its close.

Two features of the rule answer two different problems, so each deserves its own account. The first is that harmlessness is available at all. Chapman declined to hold that every federal constitutional error must be deemed harmful, because that would require automatic reversal, and it reasoned that some constitutional errors are, in the setting of a particular case, "so unimportant and insignificant" that they may be deemed harmless.³⁹ The Court restated that reasoning in Estelle v. Williams, a habeas case that is useful here only because it quotes Chapman at length. Refusing automatic reversal keeps a sound conviction from being discarded over a violation that plainly changed nothing.

The second feature is the severity of the standard and the placement of the burden. Forbidden evidence or argument does its worst damage in a trial where guilt is a close question, and a lenient test would let a reviewing court affirm in exactly the cases where the error may have tipped the balance.⁴⁰ Proof beyond a reasonable doubt, which Chapman treated as little different from asking whether there is a reasonable possibility that the complained-of evidence might have contributed to the conviction, resolves real doubt in favor of the right.⁴¹ The burden sits with the beneficiary because that party gained the advantage from the error and can point to the rest of the record to explain why it did not matter, while the defendant has already established a constitutional violation and should not also carry the risk of an unclear record. Chapman added a historical reason: the original common-law harmless-error rule required the beneficiary of an error either to prove that there was no injury or to suffer reversal of the judgment.⁴²

The facts of Chapman show the standard working for the defendant. The prosecutor repeatedly commented on the defendants' decision not to testify, and the trial judge's charge told the jurors they could treat a failure to deny or explain evidence as tending to indicate its truth.⁴³ After looking at the comments in the record, the Court had no doubt that the error was not harmless, which means the government could not carry its burden.⁴⁴

The opposite pole is the large group of constitutional errors that the analysis lets a court forgive. Where the defendant had counsel and an impartial adjudicator, there is a strong presumption that other constitutional errors are subject to harmless-error analysis, and the Court has recognized that most constitutional errors can be harmless.⁴⁵ A concurring opinion in United States v. Gaudin collects the instruction cases: jury instructions containing an unconstitutional burden-shifting presumption, a mandatory presumption, a misstated element, or an erroneous rebuttable presumption have all been put through harmless-error analysis, and the only listed exception is an erroneous burden-of-proof instruction.⁴⁶ An instruction flaw of constitutional dimension therefore does not end the inquiry. It starts the Chapman question of whether the government can show beyond a reasonable doubt that the flaw did not contribute to the verdict.

The exception is the structural error, a defect in the framework within which the trial proceeds, as opposed to an error within the trial process itself. Such defects infect the entire trial, so reversal is automatic and no harmlessness inquiry is made.⁴⁷ The Court has found structural error only in a very limited class of cases: the complete denial of counsel (Gideon), a biased trial judge (Tumey), racial discrimination in the selection of a grand jury (Vasquez), the denial of self-representation (McKaskle), the denial of a public trial (Waller), and a defective reasonable-doubt instruction (Sullivan).⁴⁸ Chapman itself acknowledged that some rights are so basic to a fair trial that their infraction can never be treated as harmless, and it pointed to coerced confessions, the right to counsel, and an impartial judge.⁴⁹ The reason is practical. When the defect corrupts the whole process, the effect on the verdict is impossible to measure, and a harmlessness inquiry would require unguided speculation, as the Court observed when it presumed prejudice from discriminatory grand jury selection in Bank of Nova Scotia v. United States.⁵⁰

The line between the two groups produced a split in Neder v. United States. There the district court, over the defendant's objection, took the materiality element of the tax counts from the jury.⁵¹ The majority held that the omission was not structural, because an instruction omitting an element does not necessarily make a trial fundamentally unfair or an unreliable way to decide guilt: the defendant had counsel, an impartial judge, the correct standard of proof, and a fairly selected jury instructed to weigh all the evidence.⁵² The error therefore went to Rule 52(a), under which reviewing courts disregard constitutional errors that are harmless beyond a reasonable doubt.⁵³ The dissenting opinion would have treated the omission as structural in effect. It reasoned that a directed verdict against the defendant is reversible no matter how overwhelming the evidence, because structural review exists to protect a basic right even when the result looks right.⁵⁴ The majority view controls, and it shows what pushes an error into the automatic-reversal class: damage to the framework of the trial, not the presence of a constitutional label.

The error and the objection are the same in both trials, and the result flips because of what the record shows about the element the instruction removed from the jury. Students lose points here when they treat every constitutional violation as automatically reversible, because the structural class is short and the presumption runs toward harmless-error analysis.⁵⁵ They lose points in the other direction when they treat the constitutional label as proof of harmlessness, since the government must still carry its burden on the actual record.

One boundary keeps this standard in its place. Chapman governs direct review. On collateral review, such as a federal habeas petition, Brecht v. Abrahamson applies the less stringent Kotteakos formulation, under which the petitioner must prove actual prejudice, meaning a substantial and injurious effect or influence in determining the jury's verdict.⁵⁶ Brecht is a habeas decision and not a Rule 52 case, so it matters here only to keep the two review settings apart.

Everything so far concerns error the defendant preserved, and the constitutional label does not carry the Chapman allocation into a record where the defendant stayed silent. Neder itself draws the contrast by describing Johnson v. United States, where the same kind of omitted-materiality error went unobjected-to and was reviewed only for plain error. The Court there declined to correct it, because the evidence on materiality was overwhelming and uncontroverted, so the error did not seriously affect the fairness, integrity, or public reputation of the proceedings.⁵⁷ Under Rule 52(b) the defendant who forfeited a constitutional error ordinarily must prove a reasonable probability that it affected the outcome, and the deprivation of a substantial right does not by itself warrant reversal without the further showing about fairness and integrity.⁵⁸ The next subsection takes up why the silent defendant carries that burden.

The Silent Defendant's Burden

The preceding discussion ended on a gap: it has not yet said who must persuade the court of appeals about prejudice when the defendant never objected. For preserved errors under Rule 52(a), the party defending the judgment carries that burden, as the harmless-error discussion above showed. Rule 52(b) reverses the allocation. The court of appeals makes much the same record-based inquiry into whether the error was prejudicial, but the burden of persuasion falls on the defendant who stayed silent rather than on the government.⁶⁰ In most cases, then, a court of appeals cannot correct a forfeited error unless the defendant shows that it affected the outcome.⁶¹ The reasons for that allocation lie in the wording of the two subsections and in the incentives that silence would otherwise create, and each deserves its own account.

The textual reason turns on a single word. Rule 52(a) precludes correction of an error only if it "does not affect substantial rights," so the party who wants the error disregarded must show that it had no such effect.⁶² Rule 52(b) authorizes a remedy only if the error "does" affect substantial rights, so the party who wants the remedy must show that it did.⁶³ Each subsection therefore places the burden on whoever is asking for the result that the subsection makes conditional, and that placement decides every case in which the record leaves the effect of the error uncertain. Under Rule 52(a) such doubt works against the judgment, because the government has not shown that the error made no difference. Under Rule 52(b) the same doubt works in favor of the judgment, because the defendant has not shown that it did.

The wording explains the allocation, and the incentive behind it explains why the allocation makes sense. If silence cost a defendant nothing, a lawyer who spotted an obvious error would have every reason to hold it back, take the chance of an acquittal, and raise the point on appeal after a conviction, with the government then obliged to prove harmlessness. An objection is what lets the trial judge repair the problem while repair is still possible, and a defendant who withholds it forfeits the claim and also deprives the court of that opportunity. Putting the burden of showing prejudice on the silent defendant holds the defendant accountable for that failure and deters strategic withholding of objections.⁶⁴ Trial practice on evidence follows the same two-track design: a timely, specific objection keeps the claim on the more forgiving harmless-error track, and silence sends it to plain-error review.⁶⁵

The allocation is the ordinary rule, and the Court did not make it absolute. In United States v. Olano the Court said the defendant must normally, though perhaps not in every case, make a specific showing of prejudice. It expressly declined to decide whether the phrase "affecting substantial rights" always means prejudicial, whether some forfeited errors may be corrected regardless of their effect on the outcome, or whether some errors should be presumed prejudicial when the defendant cannot make a specific showing.⁶⁶ Where the ordinary rule applies, the showing required is a reasonable probability that the error affected the outcome.⁶⁷ That is a much higher bar than the harmless-error review given to preserved errors, because the defendant who objected never has to make the showing at all.⁶⁸

United States v. Vonn shows how firmly the allocation holds. At Vonn's guilty plea hearing the judge failed to mention that Vonn would have a right to counsel if he went to trial, and Vonn said nothing. He raised the omission only on appeal, after the trial court denied his motion to withdraw the plea.⁶⁹ Rule 11(h) directs courts to disregard a variance from the plea procedures that does not affect substantial rights, which is the classic harmless-error formulation, and it contains no exception for defendants who did not object.⁷⁰ Vonn therefore argued that Congress, by writing the harmless-error standard into Rule 11 without its companion plain-error rule, meant to excuse a silent defendant from the plain-error burdens and to make the government prove harmlessness.⁷¹ The Court rejected that reading. The omission of the companion rule did not speak clearly in Vonn's favor, so the defendant who sat silent still had to show that his substantial rights were affected, and because relief is discretionary he carried the further burden of showing that the error seriously affected the fairness, integrity, or public reputation of judicial proceedings.⁷²

The Court's reasoning returns to the incentive problem described above: had Vonn been right, a defendant would lose nothing by failing to object to an obvious Rule 11 error when it occurred.⁷³ The outcome on the record followed from the allocation. Transcripts of other proceedings in the case showed that Vonn had been told of his right to counsel at his initial appearance and at his arraignment, and that he or his counsel had repeatedly confirmed that he understood his rights, so the omission at the plea hearing did not undermine the plea and he could not carry his burden.⁷⁴ A silent defendant thus invites a search of the whole record and not only the plea transcript, and is expected to speak while the problem can still be fixed. Collateral attack is a separate setting with a stricter standard of its own, which Vonn conceded Rule 11(h) left untouched, and it is not the question here.⁷⁵

The same allocation operates outside the plea setting. In United States v. Griffith, the prosecutor broke off her rebuttal argument to ask the defendant, who had been muttering at the defense table, whether he wanted to take the stand and testify.⁷⁶ Defense counsel objected and moved for a mistrial, and the judge denied the motion but instructed the jury on the presumption of innocence and the burden of proof. Counsel did not object to that instruction and never asked for an instruction on the Fifth Amendment privilege.⁷⁷ The Fifth Circuit found the error plain, since a Fifth Amendment violation calls for a Fifth Amendment instruction, but held that the defendant had not shown an effect on his substantial rights and so it had no authority to correct the error.⁷⁸ The comment was isolated, the jury was told at once to disregard it, and the evidence of guilt was strong, which left the burden unmet.⁷⁹ The burden is demanding but it can be carried. In Rosales-Mireles v. United States the defendant first raised a Guidelines miscalculation on appeal and so proceeded under Rule 52(b), and the Court held that a Guidelines error found to be plain and to affect substantial rights will in the ordinary case seriously affect the fairness, integrity, or public reputation of judicial proceedings and thus warrant relief.⁸⁰

The error and the record are identical in the two appeals, and the only variable is the objection, so the allocation of the burden alone decides the result. That also shows when the allocation matters most. A record that plainly shows prejudice, or plainly shows none, produces the same result under either subsection, while a balanced record is lost by whichever party carries the burden. The most common mistake is to treat harmless-error wording as shifting the burden to the government in a case where the defendant never objected, because that wording protects the party who preserved the issue and does not reach one who did not.⁸¹

One court of appeals has adopted a limited exception that confirms the rationale. In United States v. Viola, the Second Circuit reasoned that when an instruction was clearly wrong under the law at the time of trial, the defendant is on notice of the duty to object and rightly bears the burden of proving prejudice. When a supervening decision alters settled circuit law, however, the defendant had no duty to object to an instruction resting on firmly established authority, was not derelict in staying silent, and so the government must show that the error did not affect substantial rights, as under Rule 52(a).⁸² That exception is one circuit's approach, and it goes beyond what Olano decided, since Olano left the question of errors presumed prejudicial open.⁸³ It also depends on the same logic as the ordinary rule: the burden moves to the government only where the defendant's silence cannot fairly be called a failure to object.