JUSTICE SOUTER announced the judgment of the Court and delivered an opinion, in which JUSTICE STEVENS, JUSTICE GINSBURG, and JUSTICE BREYER join.
This case tests a police protocol for custodial interrogation that calls for giving no warnings of the rights to silence and counsel until interrogation has produced a confession. Although such a statement is generally inadmissible, since taken in violation of Miranda v. Arizona, 384 U. S 436 (1966), the interrogating officer follows it with Miranda warnings and then leads the suspect to cover the same ground a second time. The question here is the admissibility of the repeated statement. Because this midstream recitation of warnings after interrogation and unwarned confession could not effectively comply with Miranda’s constitutional requirement, we hold that a statement repeated after a warning in such circumstances is inadmissible.
I
Respondent Patrice Seibert’s 12-year-old son Jonathan had cerebral palsy, and when he died in his sleep she feared charges of neglect because of bedsores on his body. In her presence, two of her teenage sons and two of their friends devised a plan to conceal the facts surrounding Jonathan’s death by incinerating his body in the course of burning the family’s mobile home, in which they planned to leave Donald Rector, a mentally ill teenager living with the family, to avoid any appearance that Jonathan had been unattended. Seibert’s son Darian and a friend set the fire, and Donald died.
Five days later, the police awakened Seibert at 3 a.m. at a hospital where Darian was being treated for burns. In arresting her, Officer Kevin Clinton followed instructions from Rolla, Missouri, officer Richard Hanrahan that he refrain from giving Miranda warnings. After Seibert had been taken to the police station and left alone in an interview room for 15 to 20 minutes, Hanrahan questioned her without Miranda warnings for 30 to 40 minutes, squeezing her arm and repeating “Donald was also to die in his sleep.” App. 59 (internal quotation marks omitted). After Seibert finally admitted she knew Donald was meant to die in the fire, she was given a 20-minute coffee and cigarette break. Officer Hanrahan then turned on a tape recorder, gave Seibert the Miranda warnings, and obtained a signed waiver of rights from her. He resumed the questioning with “Ok, ’trice, we’ve been talking for a little while about what happened on Wednesday the twelfth, haven’t we?,” App. 66, and confronted her with her pre-warning statements:
Hanrahan: “Now, in discussion you told us, you told us
that there was a[n] understanding about Donald.”
Seibert: “Yes.”
Hanrahan: “Did that take place earlier that morning?”
Seibert: “Yes.”
Hanrahan: “And what was the understanding about
Donald?”
Seibert: “If they could get him out of the trailer, to take
him out of the trailer.”
Hanrahan: “And if they couldn’t?”
Seibert: “I, I never even thought about it. I just figured
they would.”
Hanrahan: “’Trice, didn’t you tell me that he was
supposed to die in his sleep?”
Seibert: “If that would happen, ’cause he was on that
new medicine, you know . . . .”
Hanrahan: “The Prozac? And it makes him sleepy. So
he was supposed to die in his sleep?”
Seibert: “Yes.” Id., at 70.
After being charged with first-degree murder for her role in Donald’s death, Seibert sought to exclude both her prewarning and postwarning statements. At the suppression hearing, Officer Hanrahan testified that he made a “conscious decision” to withhold Miranda warnings, thus resorting to an interrogation technique he had been taught: question first, then give the warnings, and then repeat the question “until I get the answer that she’s already provided once.” App. 31–34. He acknowledged that Seibert’s ultimate statement was “largely a repeat of information . . . obtained” prior to the warning. Id., at 30.
The trial court suppressed the prewarning statement but admitted the responses given after the Miranda recitation. A jury convicted Seibert of second-degree murder. On appeal, the Missouri Court of Appeals affirmed, treating this case as indistinguishable from Oregon v. Elstad, 470 U. S. 298 (1985). No. 23729, 2002 WL 114804 (Jan. 30, 2002) (not released for publication).
The Supreme Court of Missouri reversed, holding that “[i]n the circumstances here, where the interrogation was nearly continuous, . . . the second statement, clearly the product of the invalid first statement, should have been suppressed.” 93 S. W. 3d 700, 701 (2002). The court distinguished Elstad on the ground that warnings had not intentionally been withheld there, 93 S. W. 3d, at 704, and reasoned that “Officer Hanrahan’s intentional omission of a Miranda warning was intended to deprive Seibert of the opportunity knowingly and intelligently to waive her Miranda rights,” id., at 706. Since there were “no circumstances that would seem to dispel the effect of the Miranda violation,” the court held that the postwarning confession was involuntary and therefore inadmissible. Ibid. To allow the police to achieve an “end run” around Miranda, the court explained, would encourage Miranda violations and diminish Miranda’s role in protecting the privilege against self-incrimination. 93 S. W. 3d, at 706– 707. One judge dissented, taking the view that Elstad applied even though the police intentionally withheld Miranda warnings before the initial statement, and believing that “Seibert’s unwarned responses to Officer Hanrahan’s questioning did not prevent her from waiving her rights and confessing.” 93 S. W. 3d, at 708 (opinion of Benton, J.).
We granted certiorari, 538 U. S. 1031 (2003), to resolve a split in the Courts of Appeals. Compare United States v. Gale, 952 F. 2d 1412, 1418 (CADC 1992) (while “deliberate ‘end run’ around Miranda” would provide cause for suppression, case involved no conduct of that order); United States v. Carter, 884 F. 2d 368, 373 (CA8 1989) (“Elstad did not go so far as to fashion a rule permitting this sort of end run around Miranda”), with United States v. Orso, 266 F. 3d 1030, 1034–1039 (CA9 2001) (en banc) (rejecting argument that “tainted fruit” analysis applies because deliberate withholding of Miranda warnings constitutes an “improper tactic”); United States v. Esquilin, 208 F. 3d 315, 319–321 (CA1 2000) (similar).
We now affirm.
II
“In criminal trials, in the courts of the United States, wherever a question arises whether a confession is incompetent because not voluntary, the issue is controlled by that portion of the Fifth Amendment . . . commanding that no person ‘shall be compelled in any criminal case to be a witness against himself.’” Bram v. United States, 168 U. S. 532, 542 (1897). A parallel rule governing the admissibility of confessions in state courts emerged from the Due Process Clause of the Fourteenth Amendment, see, e.g., Brown v. Mississippi, 297 U. S. 278 (1936), which governed state cases until we concluded in Malloy v. Hogan, 378 U. S. 1, 8 (1964), that “[t]he Fourteenth Amendment secures against state invasion the same privilege that the Fifth Amendment guarantees against federal infringement—the right of a person to remain silent unless he chooses to speak in the unfettered exercise of his own will, and to suffer no penalty . . . for such silence.” In unifying the Fifth and Fourteenth Amendment voluntariness tests, Malloy “made clear what had already become apparent—that the substantive and procedural safeguards surrounding admissibility of confessions in state cases had become exceedingly exacting, reflecting all the policies embedded in the privilege” against self-incrimination. Miranda, 384 U. S., at 464.
In Miranda, we explained that the “voluntariness doctrine in the state cases . . . encompasses all interrogation practices which are likely to exert such pressure upon an individual as to disable him from making a free and rational choice,” id., at 464–465. We appreciated the difficulty of judicial enquiry post hoc into the circumstances of a police interrogation, Dickerson v. United States, 530 U. S. 428, 444 (2000), and recognized that “the coercion inherent in custodial interrogation blurs the line between voluntary and involuntary statements, and thus heightens the risk” that the privilege against self-incrimination will not be observed, id., at 435. Hence our concern that the “traditional totality-of-the-circumstances” test posed an “unacceptably great” risk that involuntary custodial confessions would escape detection. Id., at 442.
Accordingly, “to reduce the risk of a coerced confession and to implement the Self-Incrimination Clause,” Chavez v. Martinez, 538 U. S. 760, 790 (2003) (KENNEDY, J., concurring in part and dissenting in part), this Court in Miranda concluded that “the accused must be adequately and effectively apprised of his rights and the exercise of those rights must be fully honored,” 384 U. S., at 467. Miranda conditioned the admissibility at trial of any custodial confession on warning a suspect of his rights: failure to give the prescribed warnings and obtain a waiver of rights before custodial questioning generally requires exclusion of any statements obtained.1 Conversely, giving the warnings and getting a waiver has generally produced a virtual ticket of admissibility; maintaining that a statement is involuntary even though given after warnings and voluntary waiver of rights requires unusual stamina, and litigation over voluntariness tends to end with the finding of a valid waiver. See Berkemer v. McCarty, 468 U. S. 420, 433, n. 20 (1984) (“[C]ases in which a defendant can make a colorable argument that a self-incriminating statement was ‘compelled’ despite the fact that the law enforcement authorities adhered to the dictates of Miranda are rare”). To point out the obvious, this common consequence would not be common at all were it not that Miranda warnings are customarily given under circumstances allowing for a real choice between talking and remaining silent.
——————
1. “[T]he burden of showing admissibility rests, of course, on the prosecution.” Brown v. Illinois, 422 U. S. 590, 604 (1975). The prosecution bears the burden of proving, at least by a preponderance of the evidence, the Miranda waiver, Colorado v. Connelly, 479 U. S. 157, 169 (1986), and the voluntariness of the confession, Lego v. Twomey, 404 U. S. 477, 489 (1972).
——————
III
There are those, of course, who preferred the old way of doing things, giving no warnings and litigating the voluntariness of any statement in nearly every instance. In the aftermath of Miranda, Congress even passed a statute seeking to restore that old regime, 18 U. S. C. §3501, although the Act lay dormant for years until finally invoked and challenged in Dickerson v. United States, supra. Dickerson reaffirmed Miranda and held that its constitutional character prevailed against the statute. The technique of interrogating in successive, unwarned and warned phases raises a new challenge to Miranda. Although we have no statistics on the frequency of this practice, it is not confined to Rolla, Missouri. An officer of that police department testified that the strategy of withholding Miranda warnings until after interrogating and drawing out a confession was promoted not only by his own department, but by a national police training organization and other departments in which he had worked. App. 31–32. Consistently with the officer’s testimony, the Police Law Institute, for example, instructs that “officers may conduct a two-stage interrogation. . . . At any point during the pre-Miranda interrogation, usually after arrestees have confessed, officers may then read the Miranda warnings and ask for a waiver. If the arrestees waive their Miranda rights, officers will be able to repeat any subsequent incriminating statements later in court.” Police Law Institute, Illinois Police Law Manual 83 (Jan. 2001–Dec. 2003), http://www.illinoispolicelaw.org/training/lessons/ ILPLMIR.pdf (as visited Dec. 31, 2003, and available in the Clerk of Court’s case file) (hereinafter Police Law Manual) (emphasis in original).2 The upshot of all this advice is a question-first practice of some popularity, as one can see from the reported cases describing its use, sometimes in obedience to departmental policy.3
——————
2. Emphasizing the impeachment exception to the Miranda rule approved by this Court, Harris v. New York, 401 U. S. 222 (1971), some training programs advise officers to omit Miranda warnings altogether or to continue questioning after the suspect invokes his rights. See, e.g., Police Law Manual 83 (“There is no need to give a Miranda warning before asking questions if . . . the answers given . . . will not be required by the prosecutor during the prosecution’s case-in-chief”); California Commission on Peace Officer Standards and Training, Video Training Programs for California Law Enforcement, Miranda: Post-Invocation Questioning (broadcast July 11, 1996) (“We . . . have been encouraging you to continue to question a suspect after they’ve invoked their Miranda rights”); D. Zulawski & D. Wicklander, Practical Aspects of Interview and Interrogation 50–51 (2d ed. 2002) (describing the practice of “[b]eachheading” as useful for impeachment purpose (emphasis deleted)); see also Weisselberg, Saving Miranda, 84 Cornell L. Rev. 109, 110, 132–139 (1998) (collecting California training materials encouraging questioning “outside Miranda”). This training is reflected in the reported cases involving deliberate questioning after invocation of Miranda rights. See, e.g., California Attorneys for Criminal Justice v. Butts, 195 F. 3d 1039, 1042–1044 (CA9 2000); Henry v. Kernan, 197 F. 3d 1021, 1026 (CA9 1999); People v. Neal, 31 Cal. 4th 63, 68, 72 P. 3d 280, 282 (2003); People v. Peevy, 17 Cal. 4th 1184, 1189, 953 P. 2d 1212, 1215 (1998). Scholars have noted the growing trend of such practices. See, e.g., Leo, Questioning the Relevance of Miranda in the Twenty-First Century, 99 Mich. L. Rev. 1000, 1010 (2001); Weisselberg, In the Stationhouse After Dickerson, 99 Mich. L. Rev. 1121, 1123–1154 (2001). It is not the case, of course, that law enforcement educators en masse are urging that Miranda be honored only in the breach. See, e.g., C. O’Hara & G. O’Hara, Fundamentals of Criminal Investigation 133 (7th ed. 2003) (instructing police to give Miranda warnings before conducting custodial interrogation); F. Inbau, J. Reid, & J. Buckley, Criminal Interrogation and Confessions 221 (3d ed. 1986) (hereinafter Inbau, Reid, & Buckley) (same); John Reid & Associates, Interviewing & Interrogation: The Reid Technique 61 (1991) (same). Most police manuals do not advocate the question-first tactic, because they understand that Oregon v. Elstad, 470 U. S. 298 (1985), involved an officer’s good-faith failure to warn. See, e.g., Inbau, Reid, & Buckley 241 (Elstad’s “facts as well as [its] specific holding” instruct that “where an interrogator has failed to administer the Miranda warnings in the mistaken belief that, under the circumstances of the particular case, the warnings were not required, . . . corrective measures . . . salvage an interrogation opportunity”).
——————
IV
When a confession so obtained is offered and challenged, attention must be paid to the conflicting objects of Miranda and question-first. Miranda addressed “interrogation practices . . . likely . . . to disable [an individual] from making a free and rational choice” about speaking, 384 U. S., at 464–465, and held that a suspect must be “adequately and effectively” advised of the choice the Constitution guarantees, id., at 467. The object of question-first is to render Miranda warnings ineffective by waiting for a particularly opportune time to give them, after the suspect has already confessed. Just as “no talismanic incantation [is] required to satisfy [Miranda’s] strictures,” California v. Prysock, 453 U. S. 355, 359 (1981) (per curiam), it would be absurd to think that mere recitation of the litany suffices to satisfy Miranda in every conceivable circumstance. “The inquiry is simply whether the warnings reasonably ‘conve[y] to [a suspect] his rights as required by Miranda.’” Duckworth v. Eagan, 492 U. S. 195, 203 (1989) (quoting Prysock, supra, at 361). The threshold issue when interrogators question first and warn later is thus whether it would be reasonable to find that in these circumstances the warnings could function “effectively” as Miranda requires. Could the warnings effectively advise the suspect that he had a real choice about giving an admissible statement at that juncture? Could they reasonably convey that he could choose to stop talking even if he had talked earlier? For unless the warnings could place a suspect who has just been interrogated in a position to make such an informed choice, there is no practical justification for accepting the formal warnings as compliance with Miranda, or for treating the second stage of interrogation as distinct from the first, unwarned and inadmissible segment.4
——————
3. See, e.g., United States v. Orso, 266 F. 3d 1030, 1032–1033 (CA9 2001) (en banc); Pope v. Zenon, 69 F. 3d 1018, 1023–1024 (CA9 1995), overruled by Orso, supra; Cooper v. Dupnik, 963 F. 2d 1220, 1224–1227, 1249 (CA9 1992) (en banc); United States v. Carter, 884 F. 2d 368, 373 (CA9 1989); United States v. Esquilin, 208 F. 3d 315, 317 (CA1 2000); Davis v. United States, 724 A. 2d 1163, 1165–1166 (D. C. App. 1998).
——————
There is no doubt about the answer that proponents of question-first give to this question about the effectiveness of warnings given only after successful interrogation, and we think their answer is correct.
——————
4. Respondent Seibert argues that her second confession should be excluded from evidence under the doctrine known by the metaphor of the “fruit of the poisonous tree,” developed in the Fourth Amendment context in Wong Sun v. United States, 371 U. S. 471 (1963): evidence otherwise admissible but discovered as a result of an earlier violation is excluded as tainted, lest the law encourage future violations. But the Court in Elstad rejected the Wong Sun fruits doctrine for analyzing the admissibility of a subsequent warned confession following “an initial failure . . . to administer the warnings required by Miranda.” Elstad, 470 U. S., at 300. In Elstad, “a simple failure to administer the warnings, unaccompanied by any actual coercion or other circumstances calculated to undermine the suspect’s ability to exercise his free will” did not “so tain[t] the investigatory process that a subsequent voluntary and informed waiver is ineffective for some indeterminate period. Though Miranda requires that the unwarned admission must be suppressed, the admissibility of any subsequent statement should turn in these circumstances solely on whether it is knowingly and voluntarily made.” Id., at 309. Elstad held that “a suspect who has once responded to unwarned yet uncoercive questioning is not thereby disabled from waiving his rights and confessing after he has been given the requisite Miranda warnings.” Id., at 318. In a sequential confession case, clarity is served if the later confession is approached by asking whether in the circumstances the Miranda warnings given could reasonably be found effective. If yes, a court can take up the standard issues of voluntary waiver and voluntary statement; if no, the subsequent statement is inadmissible for want of adequate Miranda warnings, because the earlier and later statements are realistically seen as parts of a single, unwarned sequence of questioning.
——————
By any objective measure, applied to circumstances exemplified here, it is likely that if the interrogators employ the technique of withholding warnings until after interrogation succeeds in eliciting a confession, the warnings will be ineffective in preparing the suspect for successive interrogation, close in time and similar in content. After all, the reason that question-first is catching on is as obvious as its manifest purpose, which is to get a confession the suspect would not make if he understood his rights at the outset; the sensible underlying assumption is that with one confession in hand before the warnings, the interrogator can count on getting its duplicate, with trifling additional trouble. Upon hearing warnings only in the aftermath of interrogation and just after making a confession, a suspect would hardly think he had a genuine right to remain silent, let alone persist in so believing once the police began to lead him over the same ground again.5 A more likely reaction on a suspect’s part would be perplexity about the reason for discussing rights at that point, bewilderment being an unpromising frame of mind for knowledgeable decision. What is worse, telling a suspect that “anything you say can and will be used against you,” without expressly excepting the statement just given, could lead to an entirely reasonable inference that what he has just said will be used, with subsequent silence being of no avail. Thus, when Miranda warnings are inserted in the midst of coordinated and continuing interrogation, they are likely to mislead and “depriv[e] a defendant of knowledge essential to his ability to understand the nature of his rights and the consequences of abandoning them.” Moran v. Burbine, 475 U. S. 412, 424 (1986). By the same token, it would ordinarily be unrealistic to treat two spates of integrated and proximately conducted questioning as independent interrogations subject to independent evaluation simply because Miranda warnings formally punctuate them in the middle.
——————
5. It bears emphasizing that the effectiveness Miranda assumes the warnings can have must potentially extend through the repeated interrogation, since a suspect has a right to stop at any time. It seems highly unlikely that a suspect could retain any such understanding when the interrogator leads him a second time through a line of questioning the suspect has already answered fully. The point is not that a later unknowing or involuntary confession cancels out an earlier, adequate warning; the point is that the warning is unlikely to be effective in the question-first sequence we have described.
——————
V
Missouri argues that a confession repeated at the end of an interrogation sequence envisioned in a question-first strategy is admissible on the authority of Oregon v. Elstad, 470 U. S. 298 (1985), but the argument disfigures that case. In Elstad, the police went to the young suspect’s house to take him into custody on a charge of burglary. Before the arrest, one officer spoke with the suspect’s mother, while the other one joined the suspect in a “brief stop in the living room,” id., at 315, where the officer said he “felt” the young man was involved in a burglary, id., at 301 (internal quotation marks omitted). The suspect acknowledged he had been at the scene. Ibid. This Court noted that the pause in the living room “was not to interrogate the suspect but to notify his mother of the reason for his arrest,” id., at 315, and described the incident as having “none of the earmarks of coercion,” id., at 316. The Court, indeed, took care to mention that the officer’s initial failure to warn was an “oversight” that “may have been the result of confusion as to whether the brief exchange qualified as ‘custodial interrogation’ or . . . may simply have reflected . . . reluctance to initiate an alarming police procedure before [an officer] had spoken with respondent’s mother.” Id., at 315–316. At the outset of a later and systematic station house interrogation going well beyond the scope of the laconic prior admission, the suspect was given Miranda warnings and made a full confession. Elstad, supra, at 301, 314–315. In holding the second statement admissible and voluntary, Elstad rejected the “cat out of the bag” theory that any short, earlier admission, obtained in arguably innocent neglect of Miranda, determined the character of the later, warned confession, Elstad, 470 U. S., at 311–314; on the facts of that case, the Court thought any causal connection between the first and second responses to the police was “speculative and attenuated,” id., at 313. Although the Elstad Court expressed no explicit conclusion about either officer’s state of mind, it is fair to read Elstad as treating the living room conversation as a good-faith Miranda mistake, not only open to correction by careful warnings before systematic questioning in that particular case, but posing no threat to warn-first practice generally. See Elstad, supra, at 309 (characterizing the officers’ omission of Miranda warnings as “a simple failure to administer the warnings, unaccompanied by any actual coercion or other circumstances calculated to undermine the suspect’s ability to exercise his free will”); 470 U. S., at 318, n. 5 (Justice Brennan’s concern in dissent that Elstad would invite question-first practice “distorts the reasoning and holding of our decision, but, worse, invites trial courts and prosecutors to do the same”).
The contrast between Elstad and this case reveals a series of relevant facts that bear on whether Miranda warnings delivered midstream could be effective enough to accomplish their object: the completeness and detail of the questions and answers in the first round of interrogation, the overlapping content of the two statements, the timing and setting of the first and the second, the continuity of police personnel, and the degree to which the interrogator’s questions treated the second round as continuous with the first. In Elstad, it was not unreasonable to see the occasion for questioning at the station house as presenting a markedly different experience from the short conversation at home; since a reasonable person in the suspect’s shoes could have seen the station house questioning as a new and distinct experience, the Miranda warnings could have made sense as presenting a genuine choice whether to follow up on the earlier admission.
At the opposite extreme are the facts here, which by any objective measure reveal a police strategy adapted to undermine the Miranda warnings.6 The unwarned interrogation was conducted in the station house, and the questioning was systematic, exhaustive, and managed with psychological skill. When the police were finished there was little, if anything, of incriminating potential left unsaid. The warned phase of questioning proceeded after a pause of only 15 to 20 minutes, in the same place as the unwarned segment. When the same officer who had conducted the first phase recited the Miranda warnings, he said nothing to counter the probable misimpression that the advice that anything Seibert said could be used against her also applied to the details of the inculpatory statement previously elicited. In particular, the police did not advise that her prior statement could not be used.7 Nothing was said or done to dispel the oddity of warning about legal rights to silence and counsel right after the police had led her through a systematic interrogation, and any uncertainty on her part about a right to stop talking about matters previously discussed would only have been aggravated by the way Officer Hanrahan set the scene by saying “we’ve been talking for a little while about what happened on Wednesday the twelfth, haven’t we?” App. 66.
——————
6Because the intent of the officer will rarely be as candidly admitted as it was here (even as it is likely to determine the conduct of the interrogation), the focus is on facts apart from intent that show the question-first tactic at work.
7We do not hold that a formal addendum warning that a previous statement could not be used would be sufficient to change the character of the question-first procedure to the point of rendering an ensuing statement admissible, but its absence is clearly a factor that blunts the efficacy of the warnings and points to a continuing, not a new, interrogation.
——————
The impression that the further questioning was a mere continuation of the earlier questions and responses was fostered by references back to the confession already given. It would have been reasonable to regard the two sessions as parts of a continuum, in which it would have been unnatural to refuse to repeat at the second stage what had been said before. These circumstances must be seen as challenging the comprehensibility and efficacy of the Miranda warnings to the point that a reasonable person in the suspect’s shoes would not have understood them to convey a message that she retained a choice about continuing to talk.8
VI
Strategists dedicated to draining the substance out of Miranda cannot accomplish by training instructions what Dickerson held Congress could not do by statute. Because the question-first tactic effectively threatens to thwart Miranda’s purpose of reducing the risk that a coerced confession would be admitted, and because the facts here do not reasonably support a conclusion that the warnings given could have served their purpose, Seibert’s post-warning statements are inadmissible. The judgment of the Supreme Court of Missouri is affirmed.
It is so ordered.
——————
8. Because we find that the warnings were inadequate, there is no need to assess the actual voluntariness of the statement.