with whom Justice Blackmun joins, and with whom Justice Souter joins in part, dissenting.
Assuming that petitioner’s conviction was in fact tainted by a constitutional violation that, while not harmless beyond a reasonable doubt, did not have “substantial and injurious effect or influence in determining the jury’s verdict,” Kot-teakos v. United States, 328 U. S. 750, 776 (1946), it is undisputed that he would be entitled to reversal in the state courts on appeal or in this Court on certiorari review. If, however, the state courts erroneously concluded that no violation had occurred or (as is the case here) that it was harmless beyond a reasonable doubt, and supposing further that certiorari was either not sought or not granted, the majority would foreclose relief on federal habeas review. As a result of today’s decision, in short, the fate of one in state custody turns on whether the state courts properly applied the Federal Constitution as then interpreted by decisions of this Court, and on whether we choose to review his claim on cer-tiorari. Because neither the federal habeas corpus statute nor our own precedents can support such illogically disparate treatment, I dissent.
*645J — H
A
Chapman v. California, 386 U. S. 18 (1967), established the federal nature of the harmless-error standard to be applied when constitutional rights are at stake. Such rights, we stated, are “rooted in the Bill of Rights, offered and championed in the Congress by James Madison, who told the Congress that the ‘independent’ federal courts would be the ‘guardians of those rights.’” Id., at 21 (footnote omitted). Thus,
“[wjhether a conviction for crime should stand when a State has failed to accord federal constitutionally guaranteed rights is every bit as much of a federal question as what particular federal constitutional provisions themselves mean, what they guarantee, and whether they have been denied. With faithfulness to the constitutional union of the States, we cannot leave to the States the formulation of the authoritative laws, rules, and remedies designed to protect people from infractions by the States of federally guaranteed rights.” Ibid, (emphasis added).
Chapman, it is true, never expressly identified the source of this harmless-error standard. But, whether the standard be characterized as a “necessary rule” of federal law, ibid,., or criticized as a quasi-constitutional doctrine, see id., at 46, 51 (Harlan, J., dissenting), the Court clearly viewed it as essential to the safeguard of federal constitutional rights. Otherwise, there would have been no justification for imposing the rule on state courts. Cf. id., at 48-51 (Harlan, J., dissenting). As far as I can tell, the majority does not question Chapman’s vitality on direct review and, therefore, the federal and constitutional underpinnings on which it rests. •
That being so, the majority’s conclusion is untenable. Under Chapman, federal law requires reversal of a state *646conviction involving a constitutional violation that is not harmless beyond a reasonable doubt. A defendant whose conviction has been upheld despite the occurrence of such a violation certainly is “in custody in violation of the Constitution or laws ... of the United States,” 28 U. S. C. § 2254(a), and therefore is entitled to habeas relief. Although we have never explicitly held that this was the case, our practice before this day plainly supports this view, as the majority itself acknowledges. See, e. g., Rose v. Clark, 478 U. S. 570, 584 (1986); see also ante, at 630.
B
The Court justifies its decision by asserting that “collateral review is different from direct review,” ante, at 633, and that “we have applied different standards on habeas than would be applied on direct review with respect to matters other than harmless-error analysis,” ante, at 634. All told, however, it can only uncover a single example of a constitutional violation that would entitle a state prisoner to relief on direct, but not on collateral, review. Thus, federal habeas review is not available to a defendant claiming that the conviction rests on evidence seized in violation of the Fourth Amendment, even though such claims remain cognizable in state courts. Stone v. Powell, 428 U. S. 465 (1976). I have elsewhere stated my reasons for disagreeing with that holding, id,., at 536-537 (White, J., dissenting), but today’s decision cannot be supported even under Stone’s own terms.
Stone was premised on the view that the exclusionary rule is not a “personal constitutional right,” id., at 486, and that it “does not exist to remedy any wrong committed against the defendant, but rather to deter violations of the Fourth Amendment by law enforcement personnel,” Kimmelman v. Morrison, 477 U. S. 365, 392 (1986) (Powell, J., concurring in judgment). In other words, one whose conviction rests on evidence obtained in a search or seizure that violated the Fourth Amendment is deemed not to be unconstitutionally *647detained. It is no surprise, then, that the Court of Appeals in this case rested its decision on an analogy between the rights guaranteed in Doyle v. Ohio, 426 U. S. 610 (1976), and those at issue in Stone. See 944 F. 2d 1363, 1371-1372 (CA7 1991). Doyle, it concluded, “is ... a prophylactic rule designed to protect another prophylactic rule from erosion or misuse.” 944 F. 2d, at 1370.
But the Court clearly and, in my view, properly rejects that view. Indeed, it repeatedly emphasizes that Doyle “is rooted in fundamental fairness and due process concerns,” that “due process is violated whenever the prosecution uses for impeachment purposes a defendant’s post-Miranda silence,” and that it “does not bear the hallmarks of a prophylactic rule.” Ante, at 629. Because the Court likewise leaves undisturbed the notion that Chapman’s harmless-error standard is required to protect constitutional rights, see supra, at 645, its conclusion that a Doyle violation that fails to meet that standard will not trigger federal habeas relief is inexplicable.
II
The majority’s decision to adopt this novel approach is far from inconsequential. Under Chapman, the State must prove beyond a reasonable doubt that the constitutional error “did not contribute to the verdict obtained.” 386 U. S., at 24. In contrast, the Court now invokes Kotteakos v. United States, 328 U. S. 750 (1946) — a case involving a non-constitutional error of trial procedure — to impose on the defendant the burden of establishing that the error “resulted in ‘actual prejudice.’ ” Ante, at 637. Moreover, although the Court of Appeals limited its holding to Doyle and other so-called “prophylactic” rules, 944 F. 2d, at 1375, and although the parties’ arguments were similarly focused, see Brief for Respondent 36-37; Brief for United States as Amicus Curiae 16, 19, n. 11, the Court extends its holding to all “constitutional error[s] of the trial type,” ante, at 638. Given that *648all such “trial errors” are now subject to harmless-error analysis, see Arizona v. Fulminante, 499 U. S. 279, 307-308 (1991), and that “most constitutional errors” are of this variety, id., at 306, the Court effectively has ousted Chapman from habeas review of state convictions.* In other words, a state court determination that a constitutional error — even one as fundamental as the admission of a coerced confession, see Fulminante, supra, at 308 — is harmless beyond a reasonable doubt has in effect become unreviewable by lower federal courts by way of habeas corpus.
I believe this result to be at odds with the role Congress has ascribed to habeas review, which is, at least in part, to deter both prosecutors and courts from disregarding their constitutional responsibilities. “[T]he threat of habeas serves as a necessary additional incentive for trial and appellate courts throughout the land to conduct their proceedings in a manner consistent with established constitutional standards.” Desist v. United States, 394 U. S. 244, 262-263 (1969) (Harlan, J., dissenting); see also Teague v. Lane, 489 U. S. 288, 306 (1989) (plurality opinion). In response, the majority characterizes review of the Chapman determination by a federal habeas court as “scarcely .. . logical,” ante, at 636, and, in any event, sees no evidence that deterrence is needed. Ibid. Yet the logic of such practice is not ours to assess for, as Justice Frankfurter explained:
“Congress could have left the enforcement of federal constitutional rights governing the administration of criminal justice in the States exclusively to the State courts. These tribunals are under the same duty as the federal courts to respect rights under the United States Constitution. . . . But the wisdom of such a modification in the law is for Congress to consider . . . .” Brown v. *649Allen, 344 U. S. 443, 499-500 (1953) (opinion of Frankfurter, J.).
“[T]he prior State determination of a claim under the United States Constitution cannot foreclose consideration of such a claim, else the State court would have the final say which the Congress ... provided it should not have.” Id., at 500.
See also Reed v. Ross, 468 U. S. 1, 10 (1984). As for the “empirical evidence” the majority apparently seeks, I cannot understand its import. Either state courts are faithful to federal law, in which case there is no cost in applying the Chapman as opposed to the Kotteakos standard on collateral review; or they are not, and it is precisely the role of habeas corpus to rectify that situation.
Ultimately, the central question is whether States may detain someone whose conviction was tarnished by a constitutional violation that is not harmless beyond a reasonable doubt. Chapman dictates that they may not; the majority suggests that, so long as direct review has not corrected this error in time, they may. If state courts remain obliged to apply Chapman, and in light of the infrequency with which we grant certiorari, I fail to see how this decision can be reconciled with Congress’ intent.
HH
Our -habeas jurisprudence is taking on the appearance of a confused patchwork in which different constitutional rights are treated according to their status, and in which the same constitutional right is treated differently depending on whether its vindication is sought on direct or collateral review. I believe this picture bears scant resemblance either to Congress’ design or to our own precedents. The Court of Appeals having yet to apply Chapman to the facts of this case, I would remand to that court for determination of whether the Doyle violation was harmless beyond a reasonable doubt. I dissent.
As I explained in Fulminante, I have serious doubt regarding the effort to classify in systematic fashion constitutional violations as either “trial errors” — that are subject to harmlessness analysis — or “structural defects” — that are not. See 499 U. S., at 290 (White, J., dissenting).