The Substantial Rights Standard

The Substantial Rights Standard

The earlier sections left one question open. Rule 52(a) tells a court to disregard errors that do not affect substantial rights, and the reasons behind it point toward the error's effect on the result, yet the words themselves do not say how a court measures that effect.¹ For ordinary non-constitutional error, such as a misstep in admitting evidence or a gap between the indictment and the proof, the measure comes from Kotteakos v. United States, which asked what the error did to the jury that decided the case.² As United States v. Olano later explained, the phrase “affecting substantial rights” means in most cases that the error was prejudicial, that is, that it affected the outcome of the proceeding.³

This section follows the standard in the order a reviewing court applies it. It begins with the test itself, then takes up two things the test rules out, an inquiry into whether enough other evidence supported the verdict and a blanket presumption for or against the appellant, and it ends with a matched pair of conspiracy cases in which the same kind of error produced opposite results.

The Fair-Assurance Test

Kotteakos construed the federal harmless-error statute of its day, which directed courts to give judgment after examining the entire record, without regard to technical errors or defects that do not affect the substantial rights of the parties.⁴ Rule 52(a) uses the same operative words, and the Court has since applied Kotteakos when deciding whether an error is harmless under the Rule.⁵ The Court's statement of the test has two halves. If the court is sure that the error did not influence the jury, or had but “very slight effect,” the verdict and judgment stand. If the court cannot say with “fair assurance,” after pondering all that happened 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 set the conviction aside.⁶

The method has two features worth holding onto. The reviewing court reads the entire record with the erroneous ruling left in the picture, and it asks what the error meant to the jurors in relation to everything else that happened at trial.⁷ It must also judge the jurors' reactions by how jurors might reasonably react, and not by the judge's own reaction, a discipline that is easy to forget when a cold transcript makes guilt look obvious.⁸

The fourth step does the real work, because it settles who bears the risk of uncertainty. Reviewing courts often cannot tell for sure whether even a serious error changed a verdict, since the deliberations are closed to them.¹¹ If a court affirmed whenever it was unsure, an error that quietly decided a close case would stand, and the rule would shelter the very mistakes that matter most. Grave doubt therefore counts against the judgment, which is how the Court resolved a close case in which it could not say that an erroneously admitted statement may not have tipped the scales.¹² The standard also resists reduction to a formula, because the opinion that announced it acknowledged that what counts as substantial varies with the judge and with the circumstances.¹³

Courts do not always state the standard in the Supreme Court's words. Some ask whether the error “probably affected” the result, and others ask whether it is “highly probable” that the error did not contribute to the verdict.¹⁴ The Supreme Court has also described the reversible nonconstitutional error as one with a “substantial and injurious effect or influence in determining the jury's verdict.”¹⁵ In the decision under review in Dowling v. United States, the Third Circuit found an erroneous admission of other-act testimony harmless because it was highly probable that the error had not prejudiced the defendant, and it declined to apply the stricter constitutional standard because the mistake was evidentiary only.¹⁶ These formulations differ in how much doubt they tolerate, with the range running from error that “may have” affected the verdict to error that “probably did,” but each asks about the error's influence on the verdict.¹⁷ The wise course is to state the Kotteakos wording and treat the others as variants. The same passage in Kotteakos set aside departures from a constitutional norm, which receive a stricter test taken up in the next major section.¹⁸

Effect, Not Sufficiency

The most common mistake in applying the test is to ask whether the record contains enough untainted evidence to convict. Kotteakos closes off that shortcut: the inquiry cannot be merely whether there was enough to support the result apart from the phase affected by the error, and the question is whether the error itself had substantial influence.²⁰ The shortcut resembles a “correct result” test, which asks whether, on all the admissible evidence, the finding of guilt is clearly correct, and the Court rejected that test for two reasons.²¹ A correct-result test turns the appellate court into the trier of fact, and it overlooks that a defendant is entitled to a fair trial even when guilt looks clear.²² The question is instead what effect the error had, or reasonably may be taken to have had, on the jury's decision, so the impact that counts is the one on the minds of the jurors rather than on the minds of the judges.²³

The confusion is understandable, because sufficiency review also looks at the record. It asks whether, viewing the evidence in the light most favorable to the prosecution, any rational trier of fact could have found the elements of the crime beyond a reasonable doubt.²⁴ That question concerns whether the record could support a verdict, and a record can pass it while an error still left its mark on the particular verdict this jury returned. Harmless-error review asks the second question, about the actual influence of the mistake, so a conviction that survives a sufficiency challenge can still fail the Kotteakos test. Students lose this issue when they treat enough other evidence to convict as if it answered the harmlessness question.

In Krulewitch v. United States, a conspiracy prosecution, the trial court erroneously admitted an out-of-court statement attributed to a co-conspirator, made after the conspiracy's objectives had ended, which plainly implied the defendant's guilt.²⁵ It was urged that the statement was cumulative and that the rest of the case was strong enough to make the error harmless. The Court declined, because the defendant had been tried four times with two mistrials and two convictions, a Florida grand jury had earlier failed to indict, and on every vital phase the jury had to choose between the defendant and the complaining witness, so the Court could not say the statement may not have been the weight that tipped the scales.²⁶ Justice Burton, dissenting, applied the same Kotteakos language and reached the opposite conclusion, relying on an 800-page record of cumulative and corroborated evidence and on the negligible dependability of the two people whose conversation was reported.²⁷ The split shows where the difficulty lies: both opinions asked about influence, and they parted over how close the case was and what the statement would have meant to these jurors.

Presumptions and Burdens

Kotteakos also declined to settle the matter with a presumption. The Court read the legislative statements behind the harmless-error statute as an injunction against generalizing broadly, by presumption or otherwise, because the statute's sponsors distinguished between two kinds of error.²⁹ For technical errors, the party seeking a new trial bears the burden of showing that substantial rights were affected, but when an error is of a kind whose natural effect is to prejudice a litigant's substantial rights, the burden of sustaining the verdict rests on the party who claims under it.³⁰ The only permissible presumption is therefore a particular one, arising from the nature of the error and its natural effect for or against prejudice in the particular setting.³¹

The refusal to adopt a uniform presumption follows from how much errors differ. A presumption of harm for every error would send cases back for retrial over slips like the misstated exhibit number discussed earlier, while a presumption of harmlessness would require the appellant to prove what happened inside a jury room that no appellate court can observe. Tying the presumption to the error's natural effect lets a court infer prejudice where the character of the error makes harm likely, as with a mass of evidence against others that tends to spill onto the accused, and decline to infer it where the error is trivial. In the ordinary appeal from a preserved error, the party asserting that the error was harmless, which in a criminal case is the government, carries the burden of showing it.³² The allocation reverses when the objection was never made, because under Rule 52(b) it is the defendant who must show prejudice, a contrast that closes this chapter.³³

One Variance, Two Outcomes

Rule 52(a) names variance among the departures it reaches, meaning a gap between what the indictment charges and what the evidence at trial proves.³⁵ Two conspiracy cases show the standard working on that one kind of error at different scales. In Berger v. United States the indictment charged a single conspiracy, and the proof showed two, tied to contemporaneous counterfeit-money transactions, one with two participants and the other with three, with a single defendant common to both.³⁶ The Court treated the case as no different in substance from one in which the same four people had been charged in two counts and the proof had failed to connect one defendant to one of them.³⁷ Nothing in the facts showed prejudice or surprise, and proof of two conspiracies could not hurt a later defense of former acquittal or conviction, so the variance did not affect substantial rights.³⁸ Berger therefore holds that the existence of a variance does not decide the case, and what matters is whether the variance affected the accused's substantial rights.³⁹

Kotteakos presented the same error on a far larger scale. The indictment again charged one conspiracy, but the proof showed at least eight separate and independent groups, none connected to any other except through Brown, the one person through whom each obtained National Housing Act loans, a pattern the Government itself described as separate spokes meeting in a common center without a rim to enclose them.⁴⁰ Thirty-two defendants were charged at the outset, nineteen were on trial when the trial began, and thirteen went to the jury.⁴¹ The Government conceded the variance and argued only that it was not prejudicial because guilt was manifest, an argument the Court rejected under the effect-on-the-verdict inquiry described above.⁴²

The Court said that, abstractly considered, the errors in the two cases were identical in character, and Berger itself had warned that a variance might be material in a different case.⁴³ What separated them was scale and the risk of spillover. With four people and two schemes, the accused could readily prepare a defense and evidence of the other scheme added little. With thirty-two defendants and eight schemes, a defendant tied to one or a few transactions bore a far heavier burden, both in preparing for trial and in guarding against evidence about others passing to the defendant by psychological effect despite instructions to keep separate transactions separate.⁴⁴ A later court read Kotteakos as shifting attention from surprise and double jeopardy toward the transference of prejudicial evidence among defendants, and as refusing to apply the harmless-error rule mechanically.⁴⁵ That reading shows the particular presumption described in the previous section at work, since the natural effect of a mass trial is to spread evidence across defendants, an inference of prejudice that two schemes and four people do not support.