Structural Error Analysis

Finding a violation does not end an appeal, because the reviewing court must still decide what follows from it. The last section settled that the defendant need not prove prejudice to establish the violation. A separate question concerns the verdict: whether the conviction can stand if the prosecution argues that the error did not change the outcome. In United States v. Gonzalez-Lopez the Court answered that erroneous deprivation or disqualification of a defendant's chosen retained counsel is structural error, which means harmless-error review does not apply and reversal is automatic.¹ The defendant gets a new trial without any inquiry into whether the denied lawyer would have changed the result.

The reason for that second holding lies in what harmless-error review requires. The review asks whether an error affected the result, and it answers by weighing the error against the rest of the trial record. That weighing is workable only when the error's effect can be measured, as when an improperly admitted statement is set against the other proof of guilt. When the effect cannot be measured, the reviewing court would have to guess what a different lawyer might have done, and the guess would stand in for the constitutional guarantee. The three parts below take the doctrine in order: first the framework that separates trial errors from structural defects, then its application to the choice of counsel, and finally a comparison with other errors that escape harmless-error review.

Trial Error Versus Structural Defects

Whether a verdict survives a constitutional violation depends on which of two classes the violation falls into, a division the Court drew in Arizona v. Fulminante.² The Court applied that framework in United States v. Gonzalez-Lopez to decide whether the erroneous loss of chosen counsel could be reviewed for harmlessness.³ The first class is trial error, meaning mistakes that occur during presentation of the case to the jury, whose effect can be assessed against the other evidence to decide whether they were harmless beyond a reasonable doubt. This class includes most constitutional errors. The second class is structural defects, which defy harmless-error analysis because they affect the framework within which the trial proceeds and are not simply an error in the trial process itself.⁴ Keep the polarities straight: a structural label means no harmless-error review and automatic reversal, while a trial-error label means the conviction survives if the government shows harmlessness.

The framework was needed because the earlier rule supplied no principle for sorting errors. In Chapman v. California the prosecutor commented in closing argument on the defendants' failure to testify, and the Court recognized that some constitutional violations require automatic reversal while others are unimportant enough in a particular case to be deemed harmless, but it separated the two groups only by illustration.⁵ Fulminante supplied the missing principle by asking whether the error's effect can be measured. A trial error can be weighed, so on direct review the conviction stands if the government shows the error harmless beyond a reasonable doubt.⁶ A structural defect cannot be weighed, so the harmlessness question is never asked.

The line falls at measurability because of how harmless-error review works. The reviewing court must estimate what the verdict would have been without the error, and in practice the strength of the remaining evidence is the most important consideration.⁷ That estimate is possible when the error is a discrete event, such as one improperly admitted exhibit, because the court can remove it and look at what is left. It is impossible when the error runs through the whole proceeding. Fulminante illustrated the point with the total absence of counsel and a judge who is not impartial, since either one affects the entire conduct of the trial from beginning to end and leaves no untainted trial to compare.⁸ Where no comparison exists, harmless-error review would substitute speculation for the constitutional guarantee, so the error requires reversal without any inquiry into harm. Structural defects are therefore the exception, because most constitutional errors can be harmless.⁹

The Court has not claimed that its list of structural defects is complete, but the recognized examples are these:¹⁰

Total denial of the right to counsel at trial (Gideon v. Wainwright).
Denial of the right of self-representation (McKaskle v. Wiggins).
Denial of the right to a public trial (Waller v. Georgia).
A defective reasonable-doubt instruction (Sullivan v. Louisiana).
Erroneous denial of the defendant's chosen retained counsel (Gonzalez-Lopez).
Trial before a biased judge (Tumey v. Ohio).

What these share is harm so pervasive or so hard to quantify that asking whether the violation was harmless makes no sense.¹¹ The counsel-of-choice entry is the one this chapter turns on, and the next part explains why it belongs on the list.

The far larger class is trial error. Fulminante itself involved an involuntary confession, which the Court had once treated as never harmless. In 1991 it held that such a confession is a trial error similar in kind to other improperly admitted evidence, so the reviewing court examines the remaining evidence to decide whether admission was harmless beyond a reasonable doubt.¹² The same treatment applies to improper comment on a defendant's silence, evidence obtained in violation of the Fourth Amendment, an instruction that omits an element of the offense, and testimony admitted in violation of the Sixth Amendment right to counsel.¹³ The last example shows that the right at stake does not decide the category. Total denial of counsel at trial is structural and requires automatic reversal,¹⁴ while one piece of testimony taken in violation of the same right is trial error, because testimony can be weighed against the rest of the record and the absence of a lawyer cannot.

The two named poles make the contrast concrete. A denial of the right to a public trial, as in Waller v. Georgia, appears on the Court's structural list, so a conviction obtained that way is reversed without a search for harm.¹⁵ The prosecutor's comment on silence in Chapman is a trial error that a reviewing court can weigh against the other evidence.¹⁶ The comment is one event inside a trial that otherwise ran normally, whereas a trial closed to the public changes the setting for every witness, lawyer, and juror, and no record can show what an open trial would have looked like.

Hypothetical Illustration

Two defendants are convicted of the same armed robbery on identical proof: surveillance video, a fingerprint, and two eyewitnesses. In the first trial, the court erroneously admits an involuntary confession. In the second, the judge presiding is not impartial. On the first appeal, the reviewing court sets the confession aside, looks at the video, the fingerprint, and the eyewitnesses, and affirms if the government shows beyond a reasonable doubt that the confession did not contribute to the verdict. On the second appeal, the same overwhelming proof does not help the government, and the conviction is reversed because a biased judge is a structural defect and the court never weighs the evidence. The proof was identical in both cases. The fact that moved the answer is whether the error was an event that can be measured against the evidence or a defect in the structure through which the evidence was heard.

Students lose this issue when they let the strength of the prosecution's other evidence drive an answer about a structural defect. That evidence is the entire inquiry for a trial error, and it has no role once the error is structural. The classification therefore has to come first.

Exam Analysis Steps

Because the category decides the remedy, it should be fixed before any inquiry into how strong the prosecution's case was. The steps are these:

Identify the constitutional error the defendant has established, and confirm that it is an error at all before asking about its effect.
Ask whether the error is one of the recognized structural defects, such as total denial of counsel, denial of self-representation, denial of a public trial, a defective reasonable-doubt instruction, or erroneous denial of counsel of choice.
If the error is on that list, or if it affects the framework of the whole trial so that its effect cannot be measured, classify it as structural and conclude that reversal is automatic without any look at the other evidence.
If the error occurred during presentation of the case to the jury and its effect can be weighed against the other evidence, classify it as trial error and ask whether the government can show it harmless beyond a reasonable doubt.

Why Counsel Choice Is Structural

The framework from the last part sorts errors by whether a reviewing court can measure their effect, so applying it to the loss of a chosen lawyer comes down to one question: what would that loss have changed, and could anyone say so with confidence? In United States v. Gonzalez-Lopez a defendant charged with conspiracy to distribute marijuana hired a California lawyer, and the district court denied that lawyer's applications for admission to the case. At trial the court then barred the lawyer from sitting at counsel table or communicating with the local lawyer who tried the case, and even stationed a marshal between them.¹⁸ The government conceded that the deprivation was erroneous, which left only the question whether the error could be reviewed for harmlessness.¹⁹ The Court answered that an erroneous deprivation of counsel of choice has consequences that are "necessarily unquantifiable and indeterminate," so that it "unquestionably qualifies as 'structural error.'"²⁰ Reversal follows automatically, with no showing of specific prejudice.²¹

The conclusion rests on how much of a case a lawyer controls. Different attorneys pursue different strategies in investigation and discovery, in developing the theory of the defense, in selecting the jury, in presenting witnesses, and in the style of examining witnesses and arguing to the jury.²² The choice of lawyer also affects whether and on what terms the defendant cooperates with the prosecution, bargains for a plea, or decides instead to go to trial.²³ That second group of decisions deserves attention, because many of them never concern the conduct of the trial and so leave no trace in the transcript that a reviewing court reads. The error therefore reaches the framework within which the trial proceeds, and in some cases whether a trial happens at all.²⁴

The mismatch with harmless-error review explains why the Court refused to apply it. Harmless-error review estimates what the verdict would have been without the error, and in practice the strength of the remaining evidence carries most of the weight.²⁵ That method works when the error is an event that can be subtracted from a record that otherwise stands, such as one improperly admitted exhibit. A wrongly barred lawyer leaves nothing to subtract. The court would have to learn what different choices the rejected lawyer would have made and then quantify their effect on the outcome, and since neither step can be done, the inquiry would be speculation about "an alternate universe."²⁶

Two situations that look alike on the surface fall on opposite sides of this line. In Gonzalez-Lopez itself a chosen lawyer was wrongly barred and the defendant was convicted with replacement counsel, so whatever the strength of the government's proof, the court does not weigh it and the conviction is reversed.²⁷ A defendant who instead claims that his own lawyer made an identifiable mistake at trial is in a different position. Under Strickland v. Washington that claim requires both deficient performance and prejudice, so the harm is built into the definition of the error and courts do not run a separate harmless-error analysis once a violation is shown.²⁸ The difference lies in what the court is asked to evaluate: a specific mistake by a lawyer who was present can be set against the record, while the choices of a lawyer who was excluded cannot.

The dissenting Justices in Gonzalez-Lopez would have kept harmless-error review available. Their position treated the choice right as a means of securing a fair trial, so a defendant who loses a chosen lawyer should have to show some difference between the representation received and the representation lost.²⁹ The concern has the most force where substitute counsel performed well and the evidence of guilt was strong, because automatic reversal then sets aside a conviction that looks reliable. The majority did not adopt that view, for reasons that follow from the earlier steps. The choice right has never been derived from the purpose of ensuring a fair trial and is regarded as the root meaning of the guarantee, so the quality of the resulting trial is not the interest the right protects.³⁰ Harmless-error review would also require comparing the trial actually held with a trial that a different lawyer would have conducted, and the majority held that comparison cannot be made. Finally, letting the government save the conviction by pointing to a capable substitute would bring back, through the harmlessness inquiry, the prejudice requirement that the Court had already refused to attach to the violation itself.

The structural label applies only after an erroneous deprivation has been established. A refusal that rests on a valid ground, such as a lawyer's lack of qualification to practice or a conflict of interest, is no error at all, so the remedy question is never reached.³¹ Once the deprivation is erroneous, students lose points when they let the skill of the substitute lawyer bear on the answer, because it plays no role in either the violation or the remedy. The next part places this error beside other errors that escape ordinary harmless-error review.

Structural Error Compared

Erroneous deprivation of counsel of choice is one of several constitutional violations that a reviewing court will not test for harmlessness, and the reason becomes easier to hold once the error is set beside the others. Three groups of violations are exempt from the ordinary harmless-error inquiry. The first is structural error in the sense Fulminante gave the term, where the harm cannot be measured against the rest of the record. The second is a violation whose remedy is a bar on reprosecution. The third is a violation that exists only if the defendant was prejudiced, because prejudice is part of the right itself.³³ All three skip the harmless-error step, but they reach that result by different paths, and a reader who runs the paths together will misclassify a new error.

The first group is the familiar one. Erroneous denial of a chosen lawyer, the holding of United States v. Gonzalez-Lopez, belongs there beside total denial of counsel at trial and denial of self-representation, because each leaves no untainted trial to serve as a comparison.³⁴ The label attaches to the error and the stage at which it occurs, so counsel-related violations do not all land in this group. In Gilbert v. California a post-indictment lineup was held without the defendant's lawyer, and the Court did not order automatic reversal. It vacated the conviction and left the State free to show that the in-court identifications had an independent source or that their admission was harmless.³⁵ The same pattern appears in Coleman v. Alabama, where the denial of counsel at a preliminary hearing led to vacatur and a remand for a harmless-error determination.³⁶ The difference tracks measurability. A tainted identification or a flawed pretrial hearing is a discrete event whose effect on the verdict a court can examine, while a trial conducted without the chosen lawyer offers no record to examine.

A second structural error comes from a different interest, the defendant's authority over the objectives of the defense. Certain decisions are so personal that counsel must follow the client's wishes even when the decision has a strategic side, including whether to plead guilty, waive a jury, testify, or forgo an appeal.³⁷ In McCoy v. Louisiana the defendant insisted that he had been framed and objected to his lawyer's plan to concede that he had killed three members of his estranged wife's family. The lawyer told the jury so anyway, and the Court held that overriding the client's choice was structural error requiring reversal.³⁸ Two features drove that classification. The effect of a lawyer's concession of guilt on a jury cannot be measured, and the right at stake protects the defendant's autonomy, an interest unrelated to avoiding an erroneous conviction.³⁹

Counsel of choice fits with this group for a related reason. The right concerns who controls the defense, and a right defined by control is violated at the moment of the intrusion, whatever the verdict turns out to be. That is why the strength of the prosecution's case never enters the analysis for either error.

The second group has a different logic. When the remedy for a violation is a bar on reprosecution, as with denial of the right to a speedy trial or a double jeopardy violation, reversal is automatic.⁴⁰ Harmless-error review asks whether the verdict would have come out the same without the error, and that question has no use when the remedy ends the case, because the defendant is entitled to be free of further prosecution however strong the evidence at the flawed trial was. Counsel-of-choice error does not belong in this group. Reversal for the loss of a chosen lawyer returns the case for a new trial, and the defendant can now have that trial with the lawyer of choice, so the error draws automatic reversal because it is structural.

The third group differs again, because here harmless-error review is absent as a result of how the right is defined. A claim of ineffective assistance requires both deficient performance by the lawyer and prejudice to the defendant, and a claim that the prosecutor withheld evidence under Brady is viable only if the information was material, meaning there is a reasonable probability the outcome would have differed.⁴¹ The harm is built into the definition of the error, so once a violation is shown courts do not add a separate harmless-error step.⁴² In Strickland v. Washington a defendant who had pleaded guilty to capital murder attacked his lawyer's choices at the sentencing stage, including the decisions not to seek a psychiatric examination, not to present character witnesses, and not to request a presentence report.⁴³ He could not prevail without showing prejudice, meaning a reasonable probability that the result would have been different.⁴⁴

Students most often go wrong by treating that prejudice showing as a harmless-error step. Both inquiries look at the effect of the lawyer's conduct on the outcome, but they sit on opposite sides of the violation. Harmless-error review begins after a violation has been established and puts the burden on the government to show the error harmless.⁴⁵ Strickland prejudice is an element the defendant must prove before any violation exists, so the defendant carries the burden and the government does not. Dissenting in Strickland, one Justice objected to the prejudice standard the Court adopted, reasoning that ineffective assistance amounts to a denial of counsel, which can never be harmless, and that estimating prejudice from a cold record is often impossible.⁴⁶ The majority's standard controls. A right defined by harm builds the prejudice inquiry into the right, while a right that protects the defendant's control over the defense, such as choosing a lawyer or deciding whether to testify, is violated by the intrusion itself.

Exam Analysis Steps

Because the three groups answer different questions, a counsel-related error is best sorted by asking what the right protects before asking what the error did. The steps are these:

Ask whether the right is violated only when the defendant was prejudiced, as with ineffective assistance, and if so require the defendant to prove prejudice as part of the claim.
Ask whether the remedy for the violation is a bar on reprosecution, and if so reverse without any harmless-error inquiry.
Ask whether the error is a structural defect, such as erroneous denial of chosen counsel or the override of a decision reserved to the client, and if so reverse automatically.
If none of those applies, treat the error as trial error and ask whether the government can show it harmless beyond a reasonable doubt.