Concept
Codefendant Confessions
Confrontation and presence
Codefendant Confessions
A joint trial puts two accused people before a single jury, and that arrangement creates a confrontation problem that a separate trial would never raise: one defendant's confession may name the other. In Bruton v. United States, Bruton and Evans were tried together, and a postal inspector testified that Evans had orally confessed in a statement that also implicated Bruton.1 Evans did not take the stand, so Bruton had no means of questioning the man whose words were being repeated to the jury. The trial judge told the jurors to consider the confession only against Evans and to disregard it in deciding Bruton's guilt. The Supreme Court held that the instruction could not stand in for cross-examination, and that Bruton had been denied his Sixth Amendment right to confront the witnesses against him.2
The difficulty has two parts, and each explains why the ordinary remedies fail. The first is that Bruton could not repair the gap himself. A confession is legitimate evidence against the person who made it, so the prosecution was entitled to put Evans's words before the jury, and the only way for Bruton to test them was to question Evans, who kept his constitutional right to decline to testify.3 A defendant cannot compel a codefendant to submit to cross-examination, which leaves the accusation untested. The second part concerns the jury. Once Evans's confession was in evidence for one purpose, the jurors had to hold it in their minds as proof against one man while setting it aside as to the man sitting beside him.
The Court rejected the assumption that jurors can reliably perform that division. An earlier decision, Delli Paoli, had approved reliance on limiting instructions because the jury system makes little sense unless jurors are presumed to follow clear directions.4 Bruton accepted that presumption as sound in most settings, since inadmissible evidence slips in during almost every trial and a defendant is entitled to a fair trial but not a perfect one. It then identified a narrow context in which the risk that jurors will not or cannot comply is so great, and the consequences so vital to the defendant, that the limits of the jury system cannot be ignored: the powerfully incriminating confession of a codefendant who stands accused side by side with the defendant.5 The confession is devastating, and its credibility is inherently suspect because an accomplice has every motive to shift blame onto someone else. That weakness is exactly what cross-examination exposes, so when the confessor stays silent the unreliability is compounded, and the instruction has the same effect as no instruction at all.6
The Court also drew support from Douglas v. Alabama, in which a prosecutor read a purported confession of a witness to the jury while the witness refused to answer on privilege grounds, so that the jury might infer both that the statement was made and that it was true.7 Bruton's situation was more serious than that one, because Evans's confession was actually testified to and was therefore in evidence, which made it more likely that the jury would believe it was made and true, including the portions that implicated Bruton.8
The protection reaches further than the facts of Bruton itself. In Cruz v. New York, the defendant's own confession was admitted at the joint trial and interlocked with the confession of a nontestifying codefendant, yet the codefendant's confession remained inadmissible against the defendant despite the overlap.9 The argument for admitting an interlocking confession is that the defendant has already confessed, so the codefendant's account adds little and, being consistent, looks trustworthy. That argument overlooks what the second confession does at trial. A defendant may contest the confession attributed to him, and a matching confession from a second person makes the contest harder, so the codefendant's statement does real work against the defendant. Whether the two accounts truly agree, and whether the codefendant had reason to shade the story, can be tested only by questioning the codefendant.
The link to Crawford is direct. The Washington Supreme Court there had upheld admission of Sylvia Crawford's station-house statement because it interlocked with her husband's own account, and the United States Supreme Court nonetheless held that the Clause requires unavailability and a prior opportunity for cross-examination before a testimonial statement of an absent witness comes in.10 The two rules are not the same. Crawford governs testimonial hearsay offered against the accused in any setting, while Bruton addresses the joint trial and the failure of the limiting instruction, the usual remedy for evidence admissible against only one party. Both refuse to let reliability substitute for cross-examination. Before Crawford, a custodial statement to police that blamed another person was treated as presumptively unreliable when used against that person, which was a reliability analysis that Crawford later displaced for testimonial statements.11
Severance works as a cure because the confession then reaches only the jury deciding the confessor's own case.12 It does not make the confession usable against the other defendant at that defendant's own trial, since the recording would still be a testimonial statement of an absent witness and would need unavailability and prior cross-examination. Students lose this issue when they treat a limiting instruction as a cure, because the instruction is the very device the Court held inadequate. A related mistake is to reason that the confession is admissible against its maker and is therefore admissible. It is admissible against the maker alone, and the constitutional question is whether the jury can be trusted to confine it to that use, which in this context it cannot.
A violation does not automatically undo the conviction. Confrontation violations, Bruton violations included, are trial errors and not structural errors, so an appellate court asks whether the error was harmless.13 The government bears the burden of proving harmlessness beyond a reasonable doubt, and the court weighs the importance of the evidence to the prosecution's case, whether it was cumulative, and the overall strength of the case.14 In Harrington v. California the Court found a Bruton violation but affirmed, because the defendant's own statements placed him at the scene and the other evidence of guilt was overwhelming, which made the codefendants' confessions cumulative.15 That reasoning carries less force when the contested element is the defendants' state of mind and the defense has built its case around that point, because the question there is whether a reasonable juror could reach more than one conclusion.16
Three recognized circumstances allow a codefendant's confession to be used at a joint trial without violating this rule, and each is taken up in the sections that follow.
Redacted Confessions
Separate trials are not the only way to keep a codefendant's confession from damaging the person it names. The statement can be edited so that it no longer refers to that person, and a confession edited this way may be read at the joint trial against the one who made it.20 The reason this works follows from what went wrong in Bruton. The Court treated the confession as dangerous because it was a devastating accusation by a codefendant whose credibility was inherently suspect and who could not be cross-examined, and it concluded that jurors could not be expected to wall that accusation off in their minds.21 Redaction removes the accusation itself. Once nothing in the statement points at the defendant, nothing is left for the defendant to test through cross-examination, and the situation returns to the ordinary one in which jurors are told to use evidence against one party only, a direction that Bruton accepted jurors can usually follow.22
The difficulty lies in making sure the edit really leaves nothing behind. Deleting a name does not solve the problem if the jury will plainly infer who was meant, and a redaction is constitutionally defective when the jurors are certain to supply the omitted name.23 The inquiry is therefore practical, and it asks what the jury will understand from the statement as it is delivered. The more powerfully the edited statement implicates the defendant who did not make it, the less likely the redaction is to cure the problem.24 The decided cases mark both ends of that spectrum.
Two decisions show the redaction succeeding. In Richardson v. Marsh the confession was edited to remove every reference to the defendant and referred instead to a third party who was not a codefendant, and the Court approved its use at the joint trial.25 The more instructive example is Samia v. United States, where the confession was edited to say that the confessor committed the crime with other persons, a reference that gave no sign the codefendant took part, and the statement could be admitted with a limiting instruction.26 Courts have accepted this style of edit, a neutral term in place of the defendant's name, because the edited statement is no longer powerfully incriminating to the defendant.27
Gray v. Maryland shows the failure. The confession there was read to the jury with conspicuous gaps where the names had been, so that it described the confessor, two blanks, and a few other men killing the victim. The detective then confirmed that the police arrested Gray right after hearing the statement, and the Court, dividing five to four, held that this directly implicated Gray and that using the statement at the joint trial violated his confrontation rights.28 Both Samia and Gray replaced a name with something else, so the act of substitution cannot be the test. What separates them is whether the edited statement signals that something was taken out. A neutral reference reads as a complete account with no hole in it, so jurors have no reason to wonder who was removed. A visible blank advertises the removal, and at a joint trial the defendant is sitting in the courtroom, so jurors need little imagination to fill the gap. That obviousness restores the accusation the redaction was meant to eliminate, and the limiting instruction, which Bruton found too weak a substitute for cross-examination, again has to carry a load it cannot bear.
Students lose this issue when they treat removing the name as the whole test, because the question is whether the edited statement still points at the defendant after the name is gone. A successful redaction also leaves the confession where it started: it is evidence against the confessor only, and the jury is instructed to use it that way, since the redaction works by obscuring the implication against the other defendant and does not turn the confession into evidence against that defendant.29 The redaction route and the severance route therefore solve the same problem differently, one by changing what the jury hears and the other by changing which jury hears it.
Confessor Takes the Stand
Separate trials and careful redaction both keep a codefendant's accusation away from the jury that decides the other defendant's case. A third cure leaves the confession in the record and supplies the missing safeguard instead: the confessing defendant takes the stand and submits to cross-examination. When that happens, the codefendant can question the accuser in front of the jury, so the confrontation the Clause promises has been provided and no Bruton violation arises.31 What made the confession so damaging in Bruton was that it reached the jury in a form no one could test, because Evans never took the stand.32 Once the confessor testifies, that defect is gone.
The reason follows from what cross-examination does. A confession that blames someone else comes from an accomplice with every motive to shift responsibility, and questioning is the tool that exposes that motive. Counsel for the codefendant can ask about the circumstances in which the statement was given, about the confessor's reasons for naming someone else, and about whether the account is true, and the jury watches the answers and weighs them. The Clause therefore secures an opportunity to cross-examine and says nothing about what the confessor must answer. That is why it makes no difference whether the confessor admits the earlier statement, denies making it, or qualifies it: the codefendant can cross-examine on both the earlier statement and the present testimony, and the Clause is satisfied.33
In Nelson v. O'Neil the confessing codefendant testified at the joint trial and was cross-examined about the truth of what his statement said, yet he denied ever making the confession, and the confession could still be used against the other defendant.34 A denial is an answer, and an answer given under questioning is exactly what jurors need in order to choose between the denial and the statement. How much the denial undermines the confession is a question of weight for the jury, and it raises no confrontation issue.
The contrast case is Douglas v. Alabama. Loyd, already convicted at an earlier trial, was called by the State against Douglas, invoked the privilege against self-incrimination, and refused to answer any question. The prosecutor then read from a document purported to be Loyd's confession and asked him to confirm or deny each statement, and the Court held that Douglas's inability to cross-examine Loyd denied him confrontation.35 Loyd was physically in the courtroom, so presence alone gave Douglas nothing. Because Loyd answered nothing, the reading may have been the equivalent, in the jury's mind, of testimony that Loyd made the statement, and his reliance on the privilege let the jury infer both that it was made and that it was true.36 Douglas observed that effective confrontation of Loyd would have required him to affirm the statement as his, but that observation described a witness who answered no questions at all, and where a confessor does testify the governing question is whether the other defendant could cross-examine him, whatever he says.37 Neither Loyd nor the Nelson codefendant embraced the statement, so agreement with it cannot be what separates the two cases. The difference is that one witness answered questions and the other did not.
The hypothetical shows who controls the outcome. The confessor holds the right not to testify, and the codefendant cannot force the confessor onto the stand, so this cure operates only when the confessor in fact testifies.38 The condition also belongs to jury trials, since a judge sitting alone raises no Bruton concern in the first place.39 Students lose this issue when they treat a denial on the stand as though it defeated the cure, because a confessor who denies the statement is still a witness the codefendant can cross-examine, and that opportunity is all the Clause requires.
Rebutting Coercion Claims
The Bruton rule bars a nontestifying codefendant's confession when the prosecution uses it as an accusation against the defendant. The same document can sometimes do a different job, and that is where this limit on the rule lies. When a defendant claims that his own confession was coerced, the prosecution may answer by offering the codefendant's confession. The purpose is not to prove what the codefendant said about the defendant. It is to show something about how the defendant's own statement came to be.
In Tennessee v. Street the defendant claimed that the sheriff had coerced him into adopting his accomplice's confession, so that his own statement was a copy fed to him by the police. The prosecution read the accomplice's confession to the jury so that jurors could compare the two accounts and see whether the defendant's statement really tracked it. The trial judge told the jury to consider the accomplice's confession only on that question, and the Court found no confrontation violation. The jury was not being asked whether the accomplice told the truth. It was being asked whether the defendant's statement was a copy, and the document was the only available yardstick for answering that.
The reason the result differs from Bruton lies in the dangers the Bruton Court identified. It described a codefendant's confession as devastating to the defendant, as inherently suspect because accomplices have every motive to shift blame, and as intolerably unreliable when the accomplice cannot be cross-examined.41 All of those dangers come from using the confession as evidence of guilt. A confession offered only as a comparison against the defendant's own statement asks jurors for a narrower and more manageable act, which is to treat the document as a measuring stick and not as a witness. Bruton itself did not abandon limiting instructions. It recognized that jurors can and will follow them in many settings and refused to rely on them only where the risk of failure is so great, and the stakes so high, that the limits of the jury system cannot be ignored.42 An instruction that confines a document to a rebuttal use falls on the safe side of that line.
There is also an incentive reason for the exception. A defendant who claims coercion puts the circumstances of his own statement in issue and asks the jury to believe that the police manufactured it. If the prosecution could not show the jury the document the defendant says he was made to copy, the jury would hear a one-sided story that could not be checked, and a coercion claim would cost the defendant nothing to make. Allowing the rebuttal restores the balance, and the instruction keeps the prosecution from converting the rebuttal into proof of guilt.
The contrast with Bruton therefore depends on use and not on the document. In Bruton v. United States the codefendant's confession reached the jury as an accusation of the petitioner at a joint trial, the confessor did not testify, and the Court held that clear instructions to disregard it were no substitute for cross-examination.43
Students lose this issue when they treat the document as the test, because what decides the question is the purpose for which the confession is offered. The rebuttal use is available only while a coercion or copying claim is actually in play and the jury has been told to use the confession for that narrow purpose alone. If the confession is instead used to prove the defendant's guilt, the limiting instruction fails as it did in Bruton.