Harmless Error

Nearly every trial contains a mistake. Evidence law is complex, every trial poses many admissibility questions, and a judge often must rule on an objection in a matter of seconds.¹ A rule that sent every mistake back for retrial would multiply second trials over errors that changed nothing. Federal law therefore adds a question to every appeal in which error is found: did the mistake matter? A mistake that mattered is reversible, and one that did not is harmless.

The earlier sections of this chapter asked whether an appellate court may reach an error at all and how closely it examines the ruling. The appellant must preserve the error and must be aggrieved by the judgment, and the standard of review fixes how much deference the trial court receives. An appellant who clears those hurdles has shown a mistake, and harmless-error doctrine then decides whether the mistake earns a remedy. The answer depends on the error's effect on the case.

Three steps follow: the command of Rule 61 and 28 U.S.C. § 2111, the method courts use to sort harmless errors from reversible ones, and the stricter standard that governs federal constitutional error in criminal cases.

Rule 61 and Section 2111

Two texts give the command. Section 2111 of Title 28 directs a court hearing any appeal to give judgment after examining the record without regard to errors or defects that do not affect the substantial rights of the parties.² Rule 61 states the principle for federal civil proceedings at every stage. Unless justice requires otherwise, no error in admitting or excluding evidence, and no other error by the court or a party, is ground for granting a new trial, setting aside a verdict, or vacating, modifying, or otherwise disturbing a judgment or order, and the court must disregard all errors and defects that do not affect any party's substantial rights.³

Rule 61 reaches widely. It names evidentiary rulings first and then sweeps in any other error by the court or a party, so a flawed jury instruction or an improper closing argument falls within it. It withholds every remedy a losing litigant might seek: a new trial, the setting aside of a verdict, and the vacating, modifying, or disturbing of a judgment. And it measures every mistake by one test, whether the mistake affected substantial rights.

The command binds the trial judge as well as the court of appeals. A judge deciding a motion for a new trial may grant it only for an error serious enough to affect the substantial rights of the parties, and unless the judge believes the error might have made the case come out differently, the motion must be denied.⁴ The court of appeals then asks the same question under § 2111 when it reviews the judgment. The plain-error rules use the same words: an error that no one raised may be considered if it affects substantial rights.⁵ The phrase therefore appears at both ends of review, in deciding whether an unpreserved error may be considered and whether a preserved error warrants reversal.

The modern command reversed an older presumption. Under the so-called Exchequer Rule, any error was presumed harmful and required reversal.⁶ Retrials for seemingly insignificant errors mounted, and appellate courts were criticized as “impregnable citadels of technicality.” Reformers urged harmless-error legislation, a substantial number of states adopted it in the early 1900s, and by the 1960s all fifty states had a harmless-error statute or rule. The federal statute of 1919 provided the model for much of that legislation. It required a federal appellate court to give judgment after examining the entire record, without regard to technical errors, defects, or exceptions that did not affect the substantial rights of the parties.⁷

Because § 2111 was enacted to reverse the Exchequer presumption, finding error no longer ends an appeal.⁸ A court that concludes there was error must also decide whether the error was harmful, and the appellant who proves a ruling wrong has won only the first of two questions.

Courts accept the second question because few trials are error-free, and a litigant is entitled to a fair trial, not a perfect one.⁹ In Chapman v. California, the Supreme Court described harmless-error rules, state and federal, as serving a useful purpose by blocking the setting aside of convictions for small errors or defects that have little, if any, likelihood of having changed the result of the trial.¹⁰ Chapman concerned a criminal conviction, but a civil reversal buys the same thing, a second trial, and a second trial is worth its cost only when the first was compromised. Palmer v. State states the principle for jury verdicts: mere technical errors that do not affect the substantial rights of the parties are not sufficient to set aside a verdict on appeal.¹¹

Rule 61's opening phrase, unless justice requires otherwise, keeps the rule from hardening into a formula.¹² Chapman warned that harmless-error rules can work unfair results when highly important and persuasive evidence or argument, though legally forbidden, finds its way into a trial in which the question of guilt or innocence is close.¹³ The phrase preserves room to correct an error of that kind. Even so, the default runs one way: a court disregards the error unless it mattered.

The judge asked two questions in order: was there error, and did it touch substantial rights? An error by a party falls within Rule 61, and the remark left the tenant's substantial rights where they stood. Having answered yes and then no, the judge had nothing further to decide, and the court of appeals would run the same sequence under § 2111 and affirm.

Harmless Versus Reversible Error

Rule 61 and § 2111 tell a court to disregard errors that leave substantial rights untouched, but neither says how to recognize them.¹⁵ The cases supply a prediction: an error is reversible when it probably affected the result and harmless when it probably did not.¹⁶ Two connected ideas drive the prediction. A substantial right means essentially the outcome, which is the verdict and resulting judgment in a jury case and the judgment alone in a judge-tried case. And a reviewing court often cannot tell for sure that even a glaring error changed that outcome, so it needs a standard for uncertainty. The usual standard directs reversal only for error that “probably affected” the result, a formula that tells litigants little by itself.¹⁷

Error affects a substantial right when it has a material effect on the jury's deliberations or substantially sways them, and a merely technical violation of an evidence rule is no basis for reversal.¹⁸ The inquiry therefore looks at the decision the fact-finder actually made. Evidence wrongly admitted may have led the jury to find against the appellant on a critical point, and evidence wrongly excluded may have tipped an issue the other way, so the court asks whether the mistake probably moved that decision.¹⁹

That question differs from asking whether the rest of the record could support the same verdict. Whether other evidence suffices is a sufficiency question, like the one Rule 50(a) poses when it asks whether a reasonable jury would have a legally sufficient evidentiary basis to find for a party.²⁰ A contested trial usually contains enough to sustain the verdict it produced, so treating sufficiency as the test would excuse nearly every error. An error therefore calls for corrective action when it probably affected the outcome, even if enough other evidence exists to support the conclusion the trier reached.²¹ Some opinions nonetheless treat an erroneous admission as harmless whenever sufficient other evidence could support the verdict, even if the item probably was decisive in the jury's mind.²² That formulation departs from the effect-on-decision test, and you should apply the effect-on-decision test unless your jurisdiction says otherwise.

Hypothetical Illustration

A homeowner sues a roofing contractor for installing a cheaper shingle than the contract specified. Which shingle went on the roof is the only contested question. The contractor's foreman is the only witness who can describe it, and over a timely objection the judge excludes his account on a ground the evidence rules do not support. The jury, which saw delivery invoices listing the cheaper product, finds for the homeowner. The invoices would sustain the verdict, yet the foreman was the contractor's only direct answer on the one disputed point, so the exclusion probably affected the result and the court of appeals orders a new trial.

Change one fact. The contractor first called two other installers who gave the same account of the specified brand, and the judge excludes the foreman as a third repetition. Now the excluded testimony only repeats what the jury heard, the jury probably would not have decided differently, and the court of appeals affirms.

Both scenes involve the same erroneous ruling and a verdict the record could support. Only what else the jury heard on the contested point differs: the first exclusion removed the contractor's sole answer, and the second removed a repetition.

Courts answer the probability question no in a few recurring circumstances, and three doctrines describe them. Each turns an apparently reversible error into a harmless one:²³

The cumulative-evidence doctrine supports affirmance despite errors in admitting and in excluding. When so much other proper evidence supported the same point that the jury would likely have found against the appellant anyway, an erroneous admission is harmless, and when so much other proof on the point was admitted that the jury would not likely have changed its mind, an erroneous exclusion is harmless too.
The curative-instruction doctrine applies when the judge instructs the jury to disregard improperly admitted evidence or to consider it only for a limited purpose. Courts usually view such an instruction as effective, so it “cures” the error and makes it harmless.²⁴
The overwhelming-evidence doctrine applies when the properly admitted evidence supports the judgment so strongly that, as the opinions suggest, it invited a directed verdict. An appellate court generally affirms in that situation even when the errors might otherwise be serious.

The overwhelming-evidence cases echo Rule 50(a), under which a court that finds a reasonable jury would lack a legally sufficient evidentiary basis to find for a party on an issue may resolve that issue against the party.²⁵ Where a defendant offered so much unrebutted evidence of non-defectiveness that a rational jury could not have found for the plaintiff, an appeals court will treat the admission of an inadmissible affidavit as harmless error and affirm, because the plaintiff was extremely unlikely to win even without the affidavit.²⁶

The three doctrines share one logic and one set of limits. Each points to something in the record that makes it improbable that the error changed the decision reached, so a reversal would buy a second trial likely to end the same way. Cumulative evidence helps only when the contested item merely repeated what the jury already had. A treatise problem shows the contrast: three excluded records that were merely cumulative of a larger volume of consistent evidence were likely harmless, but excluding a plaintiff's attending psychiatrist, whose diagnosis would be powerful rebuttal to two defense doctors who described a delusional disorder, almost certainly affected substantial rights.²⁷ A verdict can cure an error as well, as when the jury finds for the appellant on the only issues the error could have affected.²⁸ Instructions have limits that the constitutional-error discussion below takes up.

Courts do not agree on how probable the effect must be. The standard varies by jurisdiction and by whether the case is civil, criminal, or criminal-constitutional, and the required probability runs from the error having “may have” affected the verdict to its having “probably did.”²⁹ Courts have asked whether the party claiming error has shown it “highly probable” that the error affected the right to a fair trial, whether the error affected that right “to a substantial degree,” and whether the verdict was “more probably than not” tainted.³⁰ In Kotteakos v. United States, the Supreme Court warned against technicality, urged courts to consider the proceedings in their entirety, and held reversal proper unless the court can say with “fair assurance” that the error did not affect the result.³¹ An answer should reason from the error's effect on the actual decision, name the probability standard the jurisdiction uses, and fall back on “probably affected” when none is given.

The same reasoning explains why an error on an undisputed issue rarely warrants reversal. A verdict rests on the questions the parties actually contested, so a ruling about a fact no one disputed cannot have moved the decision.³²

Here the ruling sat beside the contested issue; in the roofing case it sat on it. An exam answer therefore begins by finding the question the fact-finder actually had to decide, asks whether the mistake probably moved the answer, and then checks the record for cumulative proof, a curative instruction, or overwhelming evidence that would make that unlikely.

Constitutional Error Contrast

A statute that speaks only of substantial rights does not say how to treat a violation of the Federal Constitution. Chapman v. California answered that question for a criminal conviction. The Court declined the defendants' request to treat every federal constitutional error as automatically reversible, noting that all fifty states have harmless-error statutes or rules, that Congress long ago set the same rule for federal courts, and that some constitutional errors, in the setting of a particular case, are so unimportant and insignificant that they may be deemed harmless.³⁴

But the Court did not carry the ordinary inquiry over unchanged. 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, and the burden of showing harmlessness falls on the beneficiary of the error, as it did under the original common-law rule.³⁵ Applying that test to the prosecutor's comments, the Court had no doubt that the error was not harmless to the petitioners.³⁶

The two features of the test answer two different worries. The demanding standard answers the danger Chapman named: harmless-error rules can work unfair results when highly important and persuasive material, though forbidden, reaches a trial in which the question of guilt or innocence is close.³⁷ The burden rule answers the question of who pays when no one can tell whether the error mattered. The Court placed that cost on the party who gained from the violation and not on the person it prejudiced.³⁸ An ordinary error asks only for a probability judgment about its effect on the result, as the preceding subsection showed.

Chapman did not make every constitutional error subject to the harmless-error test. The Court recognized earlier decisions holding that some constitutional rights are so basic to a fair trial that their infraction can never be treated as harmless error.³⁹ Appellate courts also developed a related line for rights concerned with the structure of the proceeding, such as jury selection and venue, and some states retain it. A technical violation of such a right does not require reversal, but if the defendant was deprived of the substantive protection of the right, a new trial is required automatically and the strength of the evidence is irrelevant.⁴⁰ Curative instructions have limits too in criminal cases of constitutional error. The Supreme Court has held, at least on particular facts, that an instruction to disregard an involuntary confession does not cure the harm of admitting it, and that a limiting instruction ordinarily cannot make a co-defendant's confession safe to introduce against the other defendant in a joint trial.⁴¹

Hypothetical Illustration

A jury convicts a defendant after the trial judge admits an involuntary confession and then tells the jury to disregard it. On appeal the State points to strong other evidence of guilt. The State cannot rest on the instruction, which the Supreme Court has held, at least on particular facts, does not cure the harm of admitting an involuntary confession, and it must persuade the court beyond a reasonable doubt that the error was harmless.

In a civil case, a shipper loses a damages suit after the judge admits a second repair estimate that repeats figures already in evidence through a properly admitted invoice. Nothing suggests the jury relied on the estimate. The court of appeals affirms because the cumulative item did not probably affect the result, and no one must prove beyond a reasonable doubt that it did not.

The two scenes share a mistaken evidentiary ruling and a verdict the record could support, and they differ in the source of the mistake and therefore in the standard. A federal constitutional error in a criminal trial meets the reasonable-doubt test, and an ordinary civil error meets the substantial-rights inquiry.

An appellate court sorting a claimed error asks whether the issue was preserved, whether there was error, and, if so, whether the error is subject to the harmless-error doctrine at all.⁴² On an exam, ask first whether the error violated a federal constitutional right of a criminal defendant. If it did, and the right is not one whose infraction is never harmless, the beneficiary of the error must show harmlessness beyond a reasonable doubt. If the error is nonconstitutional, or the case is civil, the ordinary rule governs and the court disregards the error unless it affected substantial rights, the same words that govern plain-error review of unpreserved instruction errors.⁴³