Justice Souter, with whom Justice Stevens and Justice Ginsburg join, dissenting.

The majority repeatedly says that the Fifth Amendment does not address the admissibility of nontestimonial evidence, an overstatement that is beside the point. The issue actually presented today is whether courts should apply the fruit of the poisonous tree doctrine lest we create an incentive for the police to omit Miranda warnings, see Miranda v. Arizona, 384 U. S. 436 (1966), before custodial interrogation.1 In closing their eyes to the consequences of giving an evidentiary advantage to those who ignore Miranda, the majority adds an important inducement for interrogators to ignore the rule in that case.

Miranda rested on insight into the inherently coercive character of custodial interrogation and the inherently difficult exercise of assessing the voluntariness of any confession resulting from it. Unless the police give the prescribed warnings meant to counter the coercive atmosphere, a custodial confession is inadmissible, there being no need for the previous time-consuming and difficult enquiry into voluntariness. That inducement to forestall involuntary statements and troublesome issues of fact can only atrophy if we turn around and recognize an evidentiary benefit when an unwarned statement leads investigators to tangible evidence. There is, of course, a price for excluding evidence, but the Fifth Amendment is worth a price, and in the absence of a very good reason, the logic of Miranda should be followed: a Miranda violation raises a presumption of coercion, Oregon v. Elstad, 470 U. S. 298, 306-307, and n. 1 (1985), and the Fifth Amendment privilege against compelled self-incrimination extends to the exclusion of derivative evidence, see United States v. Hubbell, 530 U. S. 27, 37-38 (2000) (recognizing "the Fifth Amendment's protection against the prosecutor's use of incriminating information derived directly or indirectly from ... [actually] compelled testimony"); Kastigar v. United States, 406 U. S. 441, 453 (1972). That should be the end of this case.

The fact that the books contain some exceptions to the Miranda exclusionary rule carries no weight here. In Harris v. New York, 401 U. S. 222 (1971), it was respect for the integrity of the judicial process that justified the admission of unwarned statements as impeachment evidence. But Patane's suppression motion can hardly be described as seeking to "pervert" Miranda "into a license to use perjury" or otherwise handicap the "traditional truth-testing devices of the adversary process." 401 U. S., at 225-226. Nor is there any suggestion that the officers' failure to warn Patane was justified or mitigated by a public emergency or other exigent circumstance, as in New York v. Quarles, 467 U. S. 649 (1984). And of course the premise of Oregon v. Elstad, supra, is not on point; although a failure to give Miranda warnings before one individual statement does not necessarily bar the admission of a subsequent statement given after adequate warnings, 470 U. S. 298; cf. Missouri v. Seibert, ante, at ___ (slip op., at 12-13) (plurality opinion), that rule obviously does not apply to physical evidence seized once and for all.2

There is no way to read this case except as an unjustifiable invitation to law enforcement officers to flout Miranda when there may be physical evidence to be gained. The incentive is an odd one, coming from the Court on the same day it decides Missouri v. Seibert, ante. I respectfully dissent.