concurring in the result.
Chimel v. California, 395 U. S. 752 (1969), applied principles established by a long line of cases1 to determine the permissible scope of a warrantless search sought *661to be justified as the necessary incident of a lawful arrest. But in applying these principles to the circumstances involved in Chimel, we were compelled to overrule Harris v. United States, 331 U. S. 145 (1947), and United States v. Rabinowitz, 339 U. S. 56 (1950). Harris and Rabino-witz were founded on “little more than a subjective view regarding the acceptability of certain sorts of police conduct, and not on considerations relevant to Fourth Amendment interests.” Chimel, supra, at 764-765; see United States v. Rabinowitz, supra, at 83 (Frankfurter, J., dissenting). By the time of Chimel, this view had long since been rejected; but until that day, Harris and Rabinowitz survived as direct authority for the proposition that a lawful arrest would somehow justify a war-rantless search of the premises on which the arrest was made, beyond the immediate reach of the person arrested.2
Accordingly, we are presented in these cases with the question whether Chimel should be applied to require the exclusion at trial of evidence which is the fruit of a search, carried out before our decision in Chimel, and which would be lawful if measured by the standards of Harris and Rabinowitz, but unlawful under the rule of Chimel. The Court today holds that the fruits of searches made prior to our decision in Chimel may be used in criminal trials if the searches may be justified *662under the standards of Harris and Rabinowitz as those standards had previously been applied. See, e. g., Von Cleef v. New Jersey, 395 U. S. 814 (1969). I agree. In Stovall v. Denno, 388 U. S. 293, 297 (1967), we said that
“[t]he criteria guiding resolution of [this] question implicate (a) the purpose to be served by the new standards, (b) the extent of the reliance by law enforcement authorities on the old standards, and (c) the effect on the administration of justice of a retroactive application of the new standards.”
All three factors imply that the rule of Chimel should be applied only to searches carried out after Chimel was decided.
I
Like the Fifth Amendment’s protection against compulsory self-incrimination, the warrant requirement of the Fourth Amendment stakes out boundaries beyond which the government may not tread in forcing evidence or information from its citizens. When coercion, impermissible under the Fifth Amendment, has actually produced an involuntary statement, we have invariably held that the fruits of that unconstitutional coercion may not be used to prosecute the individual involved for crime. E. g., Rochin v. California, 342 U. S. 165, 173 (1952) (Frankfurter, J.); Ashcraft v. Tennessee, 322 U. S. 143 (1944); Boyd v. United States, 116 U. S. 616, 630-635, 638 (1886).3 Exclusion of statements impermissibly coerced is not merely a device to deter government agents from improper conduct in the future. Exclusion of coerced testimony is part and parcel of the privilege *663against self-incrimination. Likewise, when a search impermissible under the Fourth Amendment results in the seizure of evidence, exclusion of the fruits of that unconstitutional invasion is required not merely in hope of deterring unconstitutional searches in the future, but in order to vindicate the right of privacy guaranteed by the Fourth Amendment. See Boyd v. United States, supra; Weeks v. United States, 232 U. S. 383, 390-394, 398 (1914); Mapp v. Ohio, 367 U. S. 643, 656, 660 (1961). Exclusion of evidence in order to vindicate the right of privacy, however, does not improve the reliability of the factfinding process at trial. See Desist v. United States, 394 U. S. 244, 249-250 (1969), and cases cited. Accordingly, this factor does not require that the standards of Chimel be retroactively applied. Desist v. United States, supra; Stovall v. Denno, 388 U. S., at 297-299.
II
The factor of reliance by law-enforcement officials on Harris and Rabinowitz points in the same direction. As we recognized in Chimel itself, Fourth Amendment jurisprudence has often followed a tortuous path. 395 U. S., at 755-762. So long as Harris and Rabinowitz were not visibly overruled, we cannot be surprised that policemen and those who offer them guidance may not have scrutinized their doctrinal underpinnings for signs of erosion. And the extent of reliance, it appears, has been considerable. The Government represents, and petitioners do not seriously dispute, that a very substantial number of searches have been carried out in reliance upon these cases. In many of these, there is no reason to doubt that a warrant could and would have been obtained if the officials involved had been aware that a warrant would be required. This factor as well, therefore, implies that Chimel should have only prospective application.
*664Ill
Finally, we must evaluate the probable impact of retroactive application on the administration of justice. Persons convicted through the use of evidence inadmissible under Chimel have been found to have- engaged in conduct that the government involved may legitimately punish. Chimel casts no doubt upon the propriety of the government’s interest in punishing those who have engaged in such conduct. Accordingly, it may fairly be assumed that retroactive application of its standards would result in a substantial number of retrials. Yet Chimel likewise casts no doubt upon the reliability of the initial determination of guilt at the previous trial. Moreover, the legitimate reliance of law-enforcement officials on Harris and Rabinowitz, as already noted, may well have led them to conduct a warrantless search merely because the warrant requirement, although easily satisfied, was understandably not understood. The consequence of this is that retroactive application of the standards applied in Chimel would impose a substantial burden upon the federal and state judicial systems, while serving neither to redress knowing violations of individual privacy nor to protect a class of persons the government has no legitimate interest in punishing.
IY
This is not to say, however, that petitioners are to be denied relief because they are probably guilty. “ [T] here is always in litigation a margin of error, representing error in factfinding.” Speiser v. Randall, 357 U. S. 513, 525 (1958). The constitutional requirement that guilt in criminal cases be proved beyond a reasonable doubt serves to limit, but cannot eliminate, the number of criminal defendants found guilty who are in fact innocent. See In re Winship, 397 U. S. 358, 370-372 (1970) (con*665curring opinion). In the present cases, both petitioners asserted their innocence by pleading not guilty and going to trial; and petitioner in No. 81, whose case is here on direct review, raised in his petition for certiorari the question whether the evidence presented at trial was sufficient to support a finding of guilt. But this Court does not sit to review such questions. In denying retroactive application to the rule of Chimel, we neither do nor could determine that every person convicted by the use of evidence obtained contrary to that rule is in fact guilty of the crime of which he was convicted. The question we face is not the legitimacy or sincerity of petitioners’ claims of innocence, or indeed whether any such claims are expressly made at all. It is, instead, whether Chimel v. California compels us to conclude that the invasion of petitioners’ privacy, conducted in justifiable but mistaken reliance upon the continuing validity of Harris and Rabinowitz, requires the exclusion of the fruits of that invasion from the factfinding process at trial. I agree with the Court that it does not, and that the standards of Chimel should apply only to searches carried out after June 23, 1969.
Our cases have settled the proposition that the Fourth Amendment requires agents of the Government to obtain prior judicial approval of all searches and seizures, see, e. g., Davis v. Mississippi, 394 U. S. 721, 728 (1969); Katz v. United States, 389 U. S. 347, 356-357 (1967); James v. Louisiana, 382 U. S. 36 (1965); Preston v. United States, 376 U. S. 364, 368 (1964); McDonald v. United States, 335 U. S. 451, 455-456 (1948); Agnello v. United States, 269 U. S. 20, 33 (1925), subject only to a few narrow and well-delineated exceptions grounded upon urgent necessity. Terry v. Ohio, 392 U. S. 1, 16-27 (1968); see Katz v. United States, supra, *661at 357 n. 19 and cases cited; cf. Chambers v. Maroney, 399 U. S. 42 (1970). And, in all events, “[t]he scope of [a] search must be 'strictly tied to and justified by’ the circumstances which rendered its initiation permissible.” Terry v. Ohio, supra, at 19, quoting Warden v. Hayden, 387 U. S. 294, 310 (1967) (concurring opinion).
Long before Chimel, of course, we had made clear that Harris and Rabinowitz were not themselves without limit. James v. Louisiana, 382 U. S. 36 (1965); Kremen v. United States, 353 U. S. 346 (1957); see Von Cleef v. New Jersey, 395 U. S. 814 (1969); Stanley v. Georgia, 394 U. S. 557, 569-572 (1969) (Stewart, J., concurring in result).
Under what circumstances the Fifth Amendment requires that the individual concerned be granted immunity from prosecution for the matters revealed in his statements is a question not pertinent here. See Piccirillo v. New York, 400 U. S. 548, 561-573 (1971) (Brennan, J., dissenting).