concurring in part and concurring in the judgment.
I join Parts I and III of the principal opinion and concur in the Court’s judgment affirming in part and reversing in part the judgment of the Court of Appeals.
For me, any error that existed in the admission of the confessions of the codefendants, in violation of Bruton v. United States, 391 U. S. 123 (1968), was, on the facts of this case, clearly harmless beyond a reasonable doubt. I refrain from joining Part II of the principal opinion because, as I read it, it abandons the harmless-error analysis the Court previously has applied in similar circumstances and now adopts a per se rule to the effect that Bruton is inapplicable in an interlocking confession situation.
In Bruton, of course, the Court held that the admission in a joint trial of the confession of a codefendant who did not take the stand violated the Sixth Amendment confrontation right of the other defendant. Because in most cases the impact of admitting a codefendant’s confession is severe, and because the credibility of any such confession “is inevitably suspect,” id., at 136, the Court went on to hold that a limiting jury instruction could not alleviate the resultant substantial threat to a fair trial the Confrontation Clause was designed to protect. Id., at 136-137.
*78In Harrington v. California, 395 U. S. 250 (1969), however, the Court recognized that evidence of guilt could be sufficiently overwhelming so as to render any Bruton error “harmless beyond a reasonable doubt,” under Chapman v. California, 386 U. S. 18 (1967). Reversal of a conviction, then, was not required merely because of the existence of a Bruton error. The Court applied a similar harmless-error analysis in Schneble v. Florida, 405 U. S. 427 (1972), a case concerning the defendant’s own confession and a partially corroborating statement given by a nontestifying codefendant.
In the present case, the principal opinion appears to me to depart from this harmless-error approach and analysis to hold that Bruton simply does not apply in a case involving interlocking confessions. It concludes that in circumstances where one defendant has confessed, the interlocking confession of a codefendant “will seldom, if ever, be of the 'devastating’ character referred to in Bruton.” Ante, at 73. Similarly, it finds that the fact that the confession of a codefendant is “inevitably suspect” is of little weight where interlocking confessions are in evidence. Ibid. Thus, it holds that the right protected by Bruton, i. e., the Confrontation Clause right of cross-examination, “has far less practical value to a defendant who has confessed to the crime than to one who has consistently maintained his innocence.” Ibid. Accordingly, it concludes “that admission of interlocking confessions with proper limiting instructions conforms to the requirements” of the Constitution. Ante, at 75.
The Court has not departed heretofore from a harmless-error approach in Bruton cases. It is unclear where the present analysis will lead in cases where interlocking confessions are not in issue, but where any Bruton error appears harmless under Chapman; for where the Bruton error is harmless, the error in admitting the nontestifying codefend-ant’s confession will be far from devastating. I would be unwilling to depart from the traditional harmless-error anal*79ysis in the straightforward Bruton-error situation. Neither would I depart from the harmless-error approach in interlocking confession cases. The fact that confessions may interlock to some degree does not ensure, as a per se matter, that their admission will not prejudice a defendant so substantially that a limiting instruction will not be curative. The two confessions may interlock in part only. Or they may cover only a portion of the events in issue at the trial. Although two interlocking confessions may not be internally inconsistent, one may go far beyond the other in implicating the confessor’s codefendant. In such circumstances, the admission of the confession of the codefendant who does not take the stand could very well serve to prejudice the defendant who is incriminated by the confession, notwithstanding that the defendant’s own confession is, to an extent, interlocking. I fully recognize that in most interlocking-confession cases, any error in admitting the confession of a nontestifying codefendant will be harmless beyond a reasonable doubt. Even so, I would not adopt a rigid per se rule that forecloses a court from weighing all the circumstances in order to determine whether the defendant in fact was unfairly prejudiced by the admission of even an interlocking confession. Where he was unfairly prejudiced, the mere fact that prejudice was caused by an interlocking confession ought not to override the important interests that the Confrontation Clause protects.
It is possible, of course, that the new approach will result in no more than a shift in analysis. Instead of focusing on whether the error was harmless, defendants and courts will be forced, instead, to inquire whether the confessions were sufficiently interlocking so as to permit a conclusion that Bruton does not apply. And I suppose that after making a determination that the confessions did not interlock to a sufficient degree, the court then would have to make a harmless-*80error determination anyway, thus adding another step to the process.
Unfortunately, it is not clear that the new approach mandates even an inquiry whether the confessions interlock. Respondents have argued that the confessions in this case, in fact, did not interlock. Brief for Respondents 3-U38. The principal opinion, however, simply assumes the interlock. It thus comes close to saying that so long as all the defendants have made some type of confession which is placed in evidence, Bruton is inapplicable without inquiry into whether the confessions actually interlock and the extent thereof. If it is willing to abandon the factual inquiry that accompanies a harmless-error determination, it should be ready, at least, to substitute an inquiry into whether there is genuine interlocking before it casts the application of Bruton, and the underlying Confrontation Clause right, completely aside.
I merely add that in this case, any Bruton error, in my view, clearly was harmless. The principal issue concerning respondents at trial was whether three Negro males identified by a number of witnesses as having been at the murder scene were indeed the respondents. Each confession placed the confessing respondent at the scene of the killing. Each confession implicated the confessor in the Woods’ plan to rob the poker game. Each confession largely overlapped with and was cumulative to the others. Corroborative testimony from witnesses who were in the apartment placed respondent Hamilton at the scene of the murder and tentatively identified respondent Randolph as one of the Negroes who received a share of the proceeds in Hamilton’s apartment immediately after the killing. The testimony of five witnesses to the events outside the apartment strongly corroborated the confessions. In these circumstances, considering the confession of each respondent against him, I cannot believe that “there is a reasonable- possibility that the improperly admitted evidence contributed to the conviction.” *81Schneble v. Florida, 405 U. S., at 432. Reversal on the Bruton issue, therefore, is required.