[120]*120Justice Stevens
announced the judgment of the Court and delivered the opinion of the Court with respect to Parts I, II, and VI, and an opinion with respect to Parts III, IV, and V, in which Justice Souter, Justice Ginsburg, and Justice Breyer join.
The question presented in this case is whether the accused’s Sixth Amendment right “to be confronted with the witnesses against him” was violated by admitting into evidence at his trial a nontestifying accomplice’s entire confession that contained some statements against the accomplice’s penal interest and others that inculpated the accused.
1 — I
On December 4, 1995, three men — Benjamin Lee Lilly (petitioner), his brother Mark, and Mark’s roommate, Gary Wayne Barker — broke into a home and stole nine bottles of liquor, three loaded guns, and a safe. The next day, the men drank the stolen liquor, robbed a small country store, and shot at geese with their stolen weapons. After their car broke down, they abducted Alex DeFilippis and used his vehicle to drive to a deserted location. One of them shot and killed DeFilippis. The three men then committed two more robberies before they were apprehended by the police late in the evening of December 5.
After taking them into custody, the police questioned each of the three men separately. Petitioner did not mention the murder to the police and stated that the other two men had forced him to participate in the robberies. Petitioner’s brother Mark and Barker told the police somewhat different accounts of the crimes, but both maintained that petitioner [121]*121masterminded the robberies and was the one who had killed DeFilippis.
A tape recording of Mark’s initial oral statement indicates that he was questioned from 1:35 a.m. until 2:12 a.m. on December 6. The police interrogated him again from 2:30 a.m. until 2:53 a.m. During both interviews, Mark continually emphasized how drunk he had been during the entire spree. When asked about his participation in the string of crimes, Mark admitted that he stole liquor during the initial burglary and that he stole a 12-pack of beer during the robbery of the liquor store. Mark also conceded that he had handled a gun earlier that day and that he was present during the more serious thefts and the homicide.
The police told Mark that he would be charged with armed robbery and that, unless he broke “family ties,” petitioner “may be dragging you right in to a life sentence,” App. 257. Mark acknowledged that he would be sent away to the penitentiary. He claimed, however, that while he had primarily been drinking, petitioner and Barker had “got some guns or something” during the initial burglary. Id., at 250. Mark said that Barker had pulled a gun in one of the robberies. He further insisted that petitioner had instigated the carjacking and that he (Mark) “didn’t have nothing to do with the shooting” of DeFilippis. Id., at 256. In a brief portion of one of his statements, Mark stated that petitioner was the one who shot DeFilippis.
The Commonwealth of Virginia charged petitioner with several offenses, including the murder of DeFilippis, and tried him separately. At trial, the Commonwealth called Mark as a witness, but he invoked his Fifth Amendment privilege against self-incrimination. The Commonwealth therefore offered to introduce into evidence the statements Mark made to the police after his arrest, arguing that they were admissible as declarations of an unavailable witness against penal interest. Petitioner objected on the ground that the statements were not actually against Mark’s penal [122]*122interest because they shifted responsibility for the crimes to Barker and to petitioner, and that their admission would violate the Sixth Amendment’s Confrontation Clause. The trial judge overruled the objection and admitted the tape recordings and written transcripts of the statements in their entirety. The jury found petitioner guilty of robbery, abduction, carjacking, possession of a firearm by a felon, and four charges of illegal use of a firearm, for which offenses he received consecutive prison sentences of two life terms plus 27 years. The jury also convicted petitioner of capital murder and recommended a sentence of death, which the court imposed.
The Supreme Court of Virginia affirmed petitioner’s convictions and sentences. As is relevant here, the court first concluded that Mark’s statements were declarations of an unavailable witness against penal interest; that the statements’ reliability was established by other evidence; and, therefore, that they fell within an exception to the Virginia hearsay rule. The court then turned to petitioner’s Confrontation Clause challenge. It began by relying on our opinion in White v. Illinois, 502 U. S. 346 (1992), for the proposition that “ '[w]here proffered hearsay has sufficient guarantees of reliability to come within a firmly rooted exception to the hearsay rule, the Confrontation Clause is satisfied.’” 255 Va. 558, 574, 499 S. E. 2d 522, 534 (1998) (quoting White, 502 U. S., at 356). The Virginia court also remarked:
“[AJdmissiblity into evidence of the statement against penal interest of an unavailable witness is a 'firmly rooted’ exception to the hearsay rule in Virginia. Thus, we hold that the trial court did not err in admitting Mark Lilly’s statements into evidence.” 255 Va., at 575, 499 S. E. 2d, at 534.
“That Mark Lilly’s statements were self-serving, in that they tended to shift principal responsibility to others or to offer claims of mitigating circumstances, goes to the [123]*123weight the jury eould assign to them and not to their admissibility.” Id., at 574,499 S. E. 2d, at 534.
Our concern that this decision represented a significant departure from our Confrontation Clause jurisprudence prompted us to grant certiorari. 525 U. S. 981 (1998).
II
As an initial matter, the Commonwealth asserts that we should decline to exercise jurisdiction over petitioner’s claim because he did not fairly present his Confrontation Clause challenge to the Supreme Court of Virginia. We disagree. Although petitioner focused on state hearsay law in his challenge to the admission of Mark’s statements, petitioner expressly argued in his opening brief to that court that the admission of the statements violated his Sixth Amendment right to confrontation. He expanded his Sixth Amendment argument in his reply brief and cited Lee v. Illinois, 476 U. S. 530 (1986), and Williamson v. United States, 512 U. S. 594 (1994), in response to the Commonwealth’s contention that the admission of the statements was constitutional. These arguments, particularly the reliance on our Confrontation Clause opinion in Lee, sufficed to raise in the Supreme Court of Virginia the constitutionality of admitting Mark’s statements. See Taylor v. Illinois, 484 U. S. 400, 406, n. 9 (1988). Indeed, the court addressed petitioner’s Confrontation Clause claim without mentioning any waiver problems.
1 — i HH ► — i
In all criminal prosecutions, state as well as federal, the accused has a right, guaranteed by the Sixth and Fourteenth Amendments to the United States Constitution, “to be confronted with the witnesses against him.” U. S. Const., Amdt. 6; Pointer v. Texas, 380 U. S. 400 (1965) (applying Sixth Amendment to the States). “The central concern of the Confrontation Clause is to ensure the reliability of the evidence against a criminal defendant by subjecting it to rig[124]*124orous testing in the context of an adversary proceeding before the trier of fact.” Maryland v. Craig, 497 U. S. 836, 845 (1990). When the government seeks to offer a declarant’s out-of-court statements against the accused, and, as in this case, the declarant is unavailable,1 courts must decide whether the Clause permits the government to deny the accused his usual right to force the declarant “to submit to cross-examination, the 'greatest legal engine ever invented for the discovery of truth.’ ” California v. Green, 399 U. S. 149, 158 (1970) (footnote and citation omitted).
In our most recent case interpreting the Confrontation Clause, White v. Illinois, 502 U. S. 346 (1992), we rejected the suggestion that the Clause should be narrowly construed to apply only to practices comparable to “a particular abuse common in 16th- and 17th-century England: prosecuting a defendant through the presentation of ex parte affidavits, without the affiants ever being produced at trial.” Id., at 352. This abuse included using out-of-court depositions and “ 'confessions of accomplices.’ ” Green, 399 U. S., at 157. Accord, White, 502 U. S., at 361, 363 (Thomas, J., concurring in part and concurring in judgment) (noting that this rule applies even if the confession is “found to be reliable”). Because that restrictive reading of the Clause’s term “witnesses” would have virtually eliminated the Clause’s role in restricting the admission of hearsay testimony, we considered it foreclosed by our prior cases. Instead, we adhered to our general framework, summarized in Ohio v. Roberts, 448 U. S. 56 (1980), that the veracity of hearsay statements is sufficiently dependable to allow the untested admission of such statements against an accused when (1) “the evidence [125]*125falls within a firmly rooted hearsay exception” or (2) it contains “particularized guarantees of trustworthiness” such that adversarial testing would be expected to add little, if anything, to the statements’ reliability. Id., at 66.
Before turning to the dual Roberts inquiries, however, we note that the statements taken from petitioner’s brother in the early, morning of December 6 were obviously obtained for the purpose-of creating evidence that would be useful at a future trial. The analogy to the presentation of ex parte affidavits in the early English proceedings thus brings the Confrontation Clause into play no matter how narrowly its gateway might be read.
IV
The Supreme Court of Virginia held that the admission of Mark Lilly’s confession was constitutional primarily because, in its view, it was against Mark’s penal interest and because “the statement against penal interest of an unavailable witness is a ‘firmly rooted’ exception to the hearsay rule in Virginia.” 255 Va., at 575, 449 S. E. 2d, at 534. We assume, as we must, that Mark’s statements were against his penal interest as a matter of state law, but the question whether the statements fall within a firmly rooted hearsay exception for Confrontation Clause purposes is a question of federal law. Accordingly, it is appropriate to begin our analysis by examining the “firmly rooted” doctrine and the roots of the “against penal interest” exception.
We have allowed the admission of statements falling within a firmly rooted hearsay exception since the Court’s recognition in Mattox v. United States, 156 U. S. 237 (1895), that the Framers of the Sixth Amendment “obviously intended to ... respec[t]” certain unquestionable rules of evidence in drafting the Confrontation Clause. Id., at 243. Justice Brown, writing for the Court in that case, did not question the wisdom of excluding deposition testimony, ex parte affidavits and their equivalents. But he reasoned that an unduly strict and “technical” reading of the Clause would [126]*126have the effect of excluding other hearsay evidence, such as dying declarations, whose admissibility neither the Framers nor anyone else 100 years later “would have [had] the hardihood ... to question.” Ibid.
We now describe a hearsay exception as “firmly rooted” if, in light of “longstanding judicial and legislative experience,” Idaho v. Wright, 497 U. S. 805, 817 (1990), it “rest[s] [on] such [a] solid foundatio[n] that admission of virtually any evidence within [it] comports with the ‘substance of the constitutional protection.5 " Roberts, 448 U. S., at 66 (quoting Mattox, 156 U. S., at 244). This standard is designed to allow the introduction of statements falling within a category of hearsay whose conditions have proved over time “to remove all temptation to falsehood, and to enforce as strict an adherence to the truth as would the obligation of an oath" and cross-examination at a trial. Ibid. In White, for instance, we held that the hearsay exception for spontaneous declarations is firmly rooted because it “is at least two centuries old,” currently “widely accepted among the States,” and carries “substantial guarantees of... trustworthiness ... [that] cannot be recaptured even by later in-court testimony.” 502 U. S., at 355-856, and n. 8. Established practice, in short, must confirm that statements falling within a category of hearsay inherently “earr[y] special guarantees of credibility" essentially equivalent to, or greater than, those produced by the Constitution’s preference for cross-examined trial testimony. Id., at 356.
The “against penal interest” exception to the hearsay rule — unlike other previously recognized firmly rooted exceptions — is not generally based on the maxim that statements made without a motive to reflect on the legal consequences of one’s statement, and in situations that are exceptionally conducive to veracity, lack the dangers of inaccuracy that typically accompany hearsay. The exception, rather, is founded on the broad assumption “that a person is unlikely to fabricate a statement against his own interest at [127]*127the time it is made.” Chambers v. Mississippi, 410 U. S. 284, 299 (1973).
We have previously noted that, due to the sweeping scope of the label, the simple categorization of a statement as a “ ‘declaration against penal interest’ . . . defines too large a class for meaningful Confrontation Clause analysis.” Lee v. Illinois, 476 U. S., at 544, n. 5. In criminal trials, statements against penal interest are offered into evidence in three principal situations: (1) as voluntary admissions against the declarant; (2) as exculpatory evidence offered by a defendant who claims that the declarant committed, or was involved in, the offense; and (3) as evidence offered by the prosecution to establish the guilt of an alleged accomplice of the declarant. It is useful to consider the three categories and their roots separately.
Statements in the first category — voluntary admissions of the declarant — are routinely offered into evidence against the maker of the statement and carry a distinguished heritage confirming their admissibility when so used. See G. Gilbert, Evidence 139-140 (1756); Lambe’s Case, 2 Leach 552, 168 Eng. Rep. 379 (1791); State v. Kirby, 1 Strob. 155, 156 (1846); State v. Cowan, 29 N. C. 239, 246 (1847). Thus, assuming that Mark Lilly’s statements were taken in conformance with constitutional prerequisites, they would unquestionably be admissible against him if he were on trial for stealing alcoholic beverages.
If Mark were a codefendant in a joint trial, however, even the use of his confession to prove his guilt might have an adverse impact on the rights of his accomplices. When dealing with admissions against penal interest, we have taken great care to separate using admissions against the declarant (the first category above) from using them against other criminal defendants (the third category).
In Bruton v. United States, 391 U. S. 123 (1968), two co-defendants, Evans and Bruton, were tried jointly and convicted of armed postal robbery. A postal inspector testified [128]*128that Evans had orally confessed that he and Bruton had committed the crime. The jury was instructed that Evans’ confession was admissible against him, but could not be considered in assessing Bruton’s guilt. Despite that instruction, this Court concluded that the introduction of Evans’ confession posed such a serious threat to Bruton’s right to confront and cross-examine the witnesses against him that he was entitled to a new trial. The case is relevant to the issue before us today, not because of its principal holding concerning the ability or inability of the jury to follow the judge’s instruction, but rather because it was common ground among all of the Justices that the fact that the confession was a statement against the penal interest of Evans did not justify its use against Bruton. As Justice White noted at the outset of his dissent, “nothing in that confession which was relevant and material to Bruton’s case was admissible against Bruton.” Id., at 138.
In the years since Bruton was decided, we have reviewed a number of cases in which one defendant’s confession has been introduced into evidence in a joint trial pursuant to instructions that it could be used against him but not against his codefendant. Despite frequent disagreement over matters such as the adequacy of the trial judge’s instructions, or the sufficiency of the redaction of ambiguous references to the declarant’s accomplice, we have consistently either stated or assumed that the mere fact that one accomplice’s confession qualified as a statement against his penal interest did not justify its use as evidence against another person. See Gray v. Maryland, 523 U. S. 185, 194-195 (1998) (stating that because the use of an accomplice’s confession “creates a special, and vital, need for cross-examination,” a prosecutor desiring to offer such evidence must comply with Bruton, hold separate trials, use separate juries, or abandon the use of the confession); 523 U. S., at 200 (Scalia, J., dissenting) (stating that codefendant’s confessions “may not be considered for the purpose of determining [the defendant’s] guilt”); Richardson [129]*129v. Marsh, 481 U. S. 200, 206 (1987) (“[W]here two defendants are tried jointly, the pretrial confession of one cannot be admitted against the other unless the confessing defendant takes the stand”); Cruz v. New York, 481 U. S. 186, 189-190, 193 (1987) (same).
The second category of statements against penal interest encompasses those offered as exculpatory evidence by a defendant who claims that it was the maker of the statement, rather than he, who committed (or was involved in) the crime in question. In this context, our Court, over the dissent of Justice Holmes, originally followed the 19th-century English rule that categorically refused to recognize any “against penal interest” exception to the hearsay rule, holding instead that under federal law only hearsay statements against pecuniary (and perhaps proprietary) interest were sufficiently reliable to warrant their admission at the trial of someone other than the declarant. See Donnelly v. United States, 228 U. S. 243, 272-277 (1913). Indeed, most States adhered to this approach well into the latter half of the 20th century. See Chambers, 410 U. S., at 299 (collecting citations).
As time passed, however, the precise Donnelly rule, which barred the admission of other persons’ confessions that exculpated the accused, became the subject of increasing criticism. Professor Wigmore, for example, remarked years after Donnelly:
“The only practical consequences of this unreasoning limitation are shocking to the sense of justice; for, in its commonest application, it requires, in a criminal trial, the rejection of a confession, however well authenticated, of a person deceased or insane or fled from the jurisdiction (and therefore quite unavailable) who has avowed himself to be the true culprit.... It is therefore not too late to retrace our steps, and to discard this barbarous doctrine, which would refuse to let an innocent accused vindicate himself even by producing to the tribunal a perfectly authenticated written confession, made [130]*130on the very gallows, by the true culprit now beyond the reach of justice.” 5 J. Wigmore, Evidence § 1477, pp. 289-290 (3d ed. 1940).
See also Scolari v. United States, 406 F. 2d 563, 564 (CA9 1969) (criticizing Donnelly); United States v. Annunziato, 293 F. 2d 373, 378 (CA2 1961) (Friendly, J.) (same); Hines v. Commonwealth, 136 Va. 728, 117 S. E. 843 (1923) (criticizing Donnelly and refusing to incorporate it into state law); Wright, Uniform Rules and Hearsay, 26 U. Cin. L. Rev. 575 (1957).
Finally, in 1973, this Court endorsed the more enlightened view in Chambers, holding that the Due Process Clause affords criminal defendants the right to introduce into evidence third parties’ declarations against penal interest— their confessions — when the circumstances surrounding the statements “provid[e] considerable assurance of their reliability.” 410 U. S., at 300. Not surprisingly, most States have now amended their hearsay rules to allow the admission of such statements under against-penal-interest exceptions. See 5 J. Wigmore, Evidence § 1476, p. 352, and n. 9 (J. Chadbourn rev. 1974); id., §1477, at 360, and n. 7; J. Wigmore, Evidence §§ 1476 and 1477, pp. 618-626 (A. Best ed. Supp. 1998). But because hearsay statements of this sort are, by definition, offered by the accused, the admission of such statements does not implicate Confrontation Clause concerns. Thus, there is no need to decide whether the reliability of such statements is so inherently dependable that they would constitute a firmly rooted hearsay exception.
The third category includes cases, like the one before us today, in which the government seeks to introduce “a confession by an accomplice which incriminates a criminal defendant.” Lee, 476 U. S., at 544, n. 5. The practice of admitting statements in this category under an exception to the hearsay rule — to the extent that such a practice exists in certain jurisdictions — is, unlike the first category or even the second, of quite recent vintage. This category also typically [131]*131includes statements that, when offered in the absence of the declarant, function similarly to those used in the ancient ex parte affidavit system.
Most important, this third category of hearsay encompasses statements that are inherently unreliable. Typical of the ground swell of scholarly and judicial criticism that culminated in the Chambers decision, Wigmore’s treatise still expressly distinguishes accomplices’ confessions that inculpate themselves and the accused as beyond a proper understanding of the against-penal-interest exception because an accomplice often has a considerable interest in “confessing and betraying his eoeriminals.” 5 Wigmore, Evidence § 1477, at 358, n. 1 (J. Chadbourn rev. 1974). Consistent with this scholarship and the assumption that underlies the analysis in our Bruton line of cases, we have over the years “spoken with one voice in declaring presumptively unreliable accomplices’ confessions that incriminate defendants.” Lee, 476 U. S., at 541. See also Cruz, 481 U. S., at 195 (White, J., dissenting) (such statements “have traditionally been viewed with special suspicion”); Bruton, 391 U. S., at 136 (such statements are “inevitably suspect”).
In Crawford v. United States, 212 U. S. 183 (1909), this Court stated that even when an alleged accomplice testifies, his confession that “incriminate[s] himself together with defendant... ought to be received with suspicion, and with the very greatest care and caution, and ought not to be passed upon by the jury under the same rules governing other and apparently credible witnesses.” Id., at 204. Over 30 years ago, we applied this principle to the Sixth Amendment. We held in Douglas v. Alabama, 380 U. S. 415 (1965), that the admission of a nontestifying accomplice’s confession, which shifted responsibility and implicated the defendant as the triggerman, “plainly denied [the defendant] the right of cross-examination secured by the Confrontation Clause.” Id., at 419.
[132]*132In Lee, we reaffirmed Douglas and explained that its holding “was premised on the basic understanding that when one person accuses another of a crime under circumstances in which the declarant stands to gain by inculpating another, the accusation is presumptively suspect and must be subjected to the scrutiny of cross-examination.” 476 U. S., at 541. This is so because
“th[e] truthfinding function of the Confrontation Clause is uniquely threatened when an accomplice’s confession is sought to be introduced against a criminal defendant without the benefit of cross-examination. . . . 'Due to his strong motivation to implicate the defendant and to exonerate himself, a codefendant’s statements about what the defendant said or did are less credible than ordinary hearsay evidence.’” Ibid. (quoting Bruton, 391 U. S., at 141 (White, J., dissenting)).
Indeed, even the dissenting Justices in Lee agreed that “accomplice confessions ordinarily are untrustworthy precisely because they are not unambiguously adverse to the penal interest of the declarant,” but instead are likely to be attempts to minimize the declarant’s culpability. 476 U. S., at 552-553 (Blackmun, J., dissenting).2
We have adhered to this approach in construing the Federal Rules of Evidence. Thus, in Williamson v. United [133]*133States, 512 U. S. 594 (1994), without reaching the Confrontation Clause issue, we held that an accomplice’s statement against his own penal interest was not admissible against the defendant.3 We once again noted the presumptive unreliability of the “non-self-inculpatory” portions of the statement: “One of the most effective ways to lie is to mix falsehood with truth, especially truth that seems particularly persuasive because of its self-inculpatory nature.” Id., at 599-601.
It is clear that our cases consistently have viewed an accomplice’s statements that shift or spread the blame to a criminal defendant as falling outside the realm of those “hearsay exeeption[s] [that are] so trustworthy that adversarial testing can be expected to add little to [the statements’] reliability.” White, 502 U. S., at 357. This view is also reflected in several States’ hearsay law.4 Indeed, prior [134]*134to 1995, it appears that even Virginia rarely allowed statements against the penal interest of the declarant to be used at criminal trials. See, e. g., Ellison v. Commonwealth, 219 Va. 404, 247 S. E. 2d 685 (1978). That Virginia relaxed that portion of its hearsay law when it decided Chandler v. Commonwealth, 249 Va. 270, 455 S. E. 2d 219 (1995), and that it later apparently concluded that all statements against penal interest fall within “a ‘firmly rooted’ exception to the hearsay rule in Virginia,” 255 Va., at 575, 499 S. E. 2d, at 534, is of no consequence. The decisive fact, which we make explicit today, is that accomplices’ confessions that inculpate a criminal defendant are not within a firmly rooted exception to the hearsay rule as that concept has been defined in our Confrontation Clause jurisprudence.5
[135]*135V
Aside from its conclusion that Mark’s statements were admissible under a firmly rooted hearsay exception, the Supreme Court of Virginia also affirmed the trial court’s holding that the statements were “reliable].., in the context of the facts and circumstances under which [they were] given” because (i) “Mark Lilly was cognizant of the import of his statements and that he was implicating himself as a participant in numerous crimes” and (ii) “[elements of [his] statements were independently corroborated” by other evidence offered at trial Id., at 574, 499 S. E. 2d, at 534. See also App. 18 (trial court’s decision). The Commonwealth contends that we should defer to this “fact-intensive” determination. It further argues that these two indicia of reliability, coupled with the facts that the police read Mark his Miranda rights and did not promise him leniency in exchange for his statements, demonstrate that the circumstances surrounding his statements bore “particularized guarantees of trustworthiness,” Roberts, 448 U. S., at 66, sufficient to satisfy the Confrontation Clause’s residual admissibility test.6
[136]*136The residual “trustworthiness” test credits the axiom that a rigid application of the Clause’s standard for admissibility might in an exceptional ease exclude a statement of an unavailable witness that is incontestably probative, competent, and reliable, yet nonetheless outside of any firmly rooted hearsay exception. Cf. id., at 68; Mattox, 156 U. S., at 243-244. When a court can be confident — as in the context of hearsay falling within a firmly rooted exception — that “the declarant’s truthfulness is so clear from the surrounding circumstances that the test of cross-examination would be of marginal utility,” the Sixth Amendment’s residual “trustworthiness” test allows the admission of the declarant’s statements. Wright, 497 U. S., at 820.
Nothing in our prior opinions, however, suggests that appellate courts should defer to lower courts’ determinations regarding whether a hearsay statement has particularized guarantees of trustworthiness. To the contrary, those opinions indicate that we have assumed, as with other fact-intensive, mixed questions of constitutional law, that “[^Independent review is ... necessary ... to maintain control of, and to clarify, the legal principles” governing the factual circumstances necessary to satisfy the protections of the Bill of Rights. Ornelas v. United States, 517 U. S. 690, 697 (1996) (holding that appellate courts should review reasonable suspicion and probable-cause determinations de novo). We, of course, accept the Virginia courts’ determination that Mark’s statements were reliable for purposes of state hearsay law, and, as should any appellate court, we review the [137]*137presence or absence of historical facts for clear error. But the surrounding circumstances relevant to a Sixth Amendment admissibility determination do not include the declar-ant’s in-eourt demeanor (otherwise the declarant would be testifying) or any other factor uniquely suited to the province of trial courts. For these reasons, when deciding whether the admission of a declarant’s out-of-court statements violates the Confrontation Clause, courts should independently review whether the government’s proffered guarantees of trustworthiness satisfy the demands of the Clause.
The Commonwealth correctly notes that “the presumption of unreliability that attaches to codefendants’ confessions . . . may be rebutted.” Lee, 476 U. S., at 543. We have held, in fact, that any inherent unreliability that accompanies co-conspirator statements made during the course and in furtherance of the conspiracy is per se rebutted by the circumstances giving rise to the long history of admitting such statements. See Bourjaily v. United States, 483 U. S. 171, 182-184 (1987). Nonetheless, the historical underpinnings of the Confrontation Clause and the sweep of our prior confrontation cases offer one cogent reminder: It is highly unlikely that the presumptive unreliability that attaches to accomplices’ confessions that shift or spread blame can be effectively rebutted when the statements are given under conditions that implicate the core concerns of the old ex parte affidavit practice — that is, when the government is involved in the statements’ production, and when the statements describe past events and have not been subjected to adversarial testing.
Applying these principles, the Commonwealth’s asserted guarantees of trustworthiness fail to convince us that Mark’s confession was sufficiently reliable as to be admissible without allowing petitioner to cross-examine him. That other evidence at trial corroborated portions of Mark’s statements is irrelevant. We have squarely rejected the notion that “evidence corroborating the truth of a hearsay statement [138]*138may properly support a finding that the statement bears 'particularized guarantees of trustworthiness.’ ” Wright, 497 U. S., at 822. In Wright, we concluded that the admission of hearsay statements by a child declarant violated the Confrontation Clause even though the statements were admissible under an exception to the hearsay rule recognized in Idaho, and even though they were corroborated by other evidence. We recognized that it was theoretically possible for such statements to possess “'particularized guarantees of trustworthiness’” that would justify their admissibility, but we refused to allow the State to “bootstrap on” the trustworthiness of other evidence. “To be admissible under the Confrontation Clause,” we held, “hearsay evidence used to convict a defendant must possess indicia of reliability by virtue of its inherent trustworthiness, not by reference to other evidence at trial.” Ibid.
Nor did the police’s informing Mark of his Miranda rights render the circumstances surrounding his statements significantly more trustworthy. We noted in rejecting a similar argument in Lee that a finding that a confession was “voluntary for Fifth Amendment purposes ... does not bear on the question of whether the confession was also free from any desire, motive, or impulse [the declarant] may have had either to mitigate the appearance of his own culpability by spreading the blame or to overstate [the defendant’s] involvement” in the crimes at issue. 476 U. S., at 544. By the same token, we believe that a suspect’s consciousness of his Miranda rights has little, if any, bearing on the likelihood of truthfulness of his statements. When a suspect is in custody for his obvious involvement in serious crimes, his knowledge that anything he says may be used against him militates against depending on his veracity.
The Commonwealth’s next proffered basis for reliability— that Mark knew he was exposing himself to criminal liability — merely restates the fact that portions of his statements were technically against penal interest. And as we have ex[139]*139plained, such statements are suspect insofar as they inculpate other persons. “[T]hat a person is making a broadly self-inculpatory confession does not make more credible the confession’s non-self-inculpatory parts.” Williamson, 512 U. S., at 599. Accord, Lee, 476 U. S., at 545. Similarly, the absence of an express promise of leniency to Mark does not enhance his statements’ reliability to the level necessary for their untested admission. The police need not tell a person who is in custody that his statements may gain him leniency in order for the suspect to surmise that speaking up, and particularly placing blame on his cohorts, may inure to his advantage.
It is abundantly clear that neither the words that Mark spoke nor the setting in which he was questioned provides any basis for concluding that his comments regarding petitioner’s guilt were so reliable that there was no need to subject them to adversarial testing in a trial setting. Mark was in custody for his involvement in, and knowledge of, serious crimes and made his statements under the supervision of governmental authorities. He was primarily responding to the officers’ leading questions, which were asked without any contemporaneous cross-examination by adverse parties. Thus, Mark had a natural motive to attempt to exculpate himself as much as possible. See id., at 544-545; Dutton v. Evans, 400 U. S. 74, 98 (1970) (Harlan, J., concurring in result). Mark also was obviously still under the influence of alcohol. Each of these factors militates against finding that his statements were so inherently reliable that cross-examination would have been superfluous.
VI
The admission of the untested confession of Mark Lilly violated petitioner’s Confrontation Clause rights. Adhering to our general custom of allowing state courts initially to assess the effect of erroneously admitted evidence in light of substantive state criminal law, we leave it to the Virginia courts [140]*140to consider in the first instance whether this Sixth Amendment error was "harmless beyond a reasonable doubt.” Chapman v. California, 386 U. S. 18, 24 (1967). See also Lee, 476 U. S., at 547. Accordingly, the judgment of the Supreme Court of Virginia is reversed, and the case is remanded for further proceedings.
It is so ordered.