dissenting.

The Court decides that the voluntariness of a confession is not an issue of fact presumed to be correct under 28 U. S. C. § 2254(d). I think it is difficult to sensibly distinguish the determination that a particular confession was voluntary from the determinations which we have held to be entitled to a presumption of correctness under § 2254(d). See Sumner v. Mata, 449 U. S. 539 (1981); Sumner v. Mata, 455 U. S. 591 (1982) (per curiam); Marshall v. Lonberger, 459 U. S. 422, 431-437 (1983); Maggio v. Fulford, 462 U. S. 111 (1983) (per curiam); Rushen v. Spain, 464 U. S. 114 (1983) (per curiam); Patton v. Yount, 467 U. S. 1025, 1036-1038 (1984); and Wainwright v. Witt, 469 U. S. 412, 426-430 (1985). While the Court relies principally on stare decisis for the result it reaches today, almost all the cases upon which it relies entailed direct review by this Court of state-court decisions *119rather than federal habeas review. But even if that difference were deemed immaterial, it seems to me that stare deci-sis is not a sufficient reason for excluding a finding as to the voluntariness of a confession from the presumption embodied in § 2254(d). All of the recent cases cited evince a more reasoned approach to this issue than the interesting but somewhat mystical exegesis in cases such as Culombe v. Connecticut, 367 U. S. 568, 603-605 (1961) (opinion of Frankfurter, J.).

I also disagree with the Court’s decision to remand this case to the Court of Appeals for a second run at the voluntariness issue. I think the majority of that court made it clear that it had evaluated the admissibility of the confession under the correct standard as defined by this Court today. It is unfortunate that petitioner’s challenge to his conviction for a murder which occurred 12 years ago should be the subject of additional and unnecessary litigation and delay.

I respectfully dissent.

delivered the opinion of the Court.

Under 28 U. S. C. § 2254(d), state-court findings of fact “shall be presumed to be correct” in a federal habeas corpus proceeding unless one of eight enumerated exceptions applies.1 The question presented is whether the voluntariness *106of a confession is an issue of fact entitled to the § 2254(d) presumption.

I

On the morning of August 13, 1973, a stranger approached the rural New Jersey home of 17-year-old Deborah Margolin and told her that a heifer was loose at the foot of her driveway. She set out alone to investigate and never returned. Later that day, her mutilated body was found in a nearby stream.

The victim’s brothers were able to provide a description of the stranger’s car and clothing. Based on this information, officers of the New Jersey State Police tentatively identified petitioner and, later that evening, found him at his place of employment. Petitioner responded to the officers’ preliminary inquiries and agreed to return to the police barracks for further questioning. Approximately two hours later, Detective Charles Boyce led petitioner to an interrogation room and informed him of his Mimnda rights. Petitioner inquired about the scope of his privilege to remain silent and then executed a written waiver, the validity of which is not at issue.

A 58 minute long interrogation session ensued. During the course of the interview, Detective Boyce told petitioner that Ms. Margolin had just died. That statement, which Boyce knew to be untrue, supported another officer’s earlier, and equally false, suggestion that the victim was still alive and could identify her attacker. App. 16-17; Record 109 and 305. Detective Boyce also told petitioner that he had been identified at the Margolin home earlier in the day. In fact, Ms. Margolin’s brothers had only provided a general description of the stranger’s car and clothing. Finally, Detective Boyce indicated that blood stains had been found on petitioner’s front stoop. No such evidence was introduced at trial, and respondents do not now contend that it ever in fact existed.

Throughout the interview, Detective Boyce presented himself as sympathetic to petitioner’s plight. On several *107occasions, he stated that he did not consider petitioner to be a criminal because the perpetrator of the deed had a “mental problem” and needed medical help rather than punishment. App. 19.2 Eventually, petitioner fully confessed to the crime. After doing so, he lapsed into what Detective Boyce described as a “state of shock.” Record 84-85. Repeated *108efforts to rouse him from his stupor failed, and the police summoned an ambulance to transport him to the hospital.

The trial court rejected petitioner’s motion to suppress the confession, and the jury found petitioner guilty of murder in the first degree. The Superior Court Appellate Division reversed, finding as a matter of law that the confession was the result of “intense and mind bending psychological compulsion” and therefore was impermissible under the Fourteenth Amendment’s guarantee of due process. App. 53. Over three dissents, the Supreme Court of New Jersey reversed again. State v. Miller, 76 N. J. 392, 388 A. 2d 218 (1978). After examining the “totality of all the surrounding circumstances,” including petitioner’s educational level, age, and awareness of his Miranda rights, the court found that the interrogation “did not exceed proper bounds,” and that the resulting confession, being voluntary, had been properly admitted into evidence. Id., at 402-405, 388 A. 2d, at 223-224.

Petitioner then sought a writ of habeas corpus in the United States District Court for the District of New Jersey. That court dismissed the application without an evidentiary hearing. A divided panel of the Court of Appeals for the Third Circuit affirmed. 741 F. 2d 1456 (1984). Relying on Circuit precedent,3 the court held that the voluntariness of a confession is a “factual issue” within the meaning of 28 U. S. C. § 2254(d). Accordingly, federal review of the New Jersey Supreme Court’s determination that petitioner’s confession was voluntary was “limited to whether the state court applied the proper legal test, and whether [its] factual conclusions . . . [were] supported on the record as a whole.” 741 F. 2d, at 1462. Under this standard, the court concluded, *109the District Court’s denial of the petition for habeas relief was proper.

Because the Courts of Appeals have reached differing conclusions on whether state-court voluntariness determinations are entitled to the § 2254(d) presumption of correctness, and because of the issue’s importance to the administration of criminal justice, we granted certiorari. 471 U. S. 1003 (1985). Compare Brantley v. McKaskle, 122 F. 2d 187, 188 (CA5 1984) “([V]oluntariness of a confession is a mixed question of law and fact”), with Alexander v. Smith, 582 F. 2d 212, 217 (CA2) (state-court voluntariness determination entitled to § 2254(d) presumption), cert. denied, 439 U. S. 990 (1978). We now reverse and remand.

I — i HH

This Court has long held that certain interrogation techniques, either in isolation or as applied to the unique characteristics of a particular suspect, are so offensive to a civilized system of justice that they must be condemned under the Due Process Clause of the Fourteenth Amendment. Brown v. Mississippi, 297 U. S. 278 (1936), was the wellspring of this notion, now deeply embedded in our criminal law. Faced with statements extracted by beatings and other forms of physical and psychological torture, the Court held that confessions procured by means “revolting to the sense of justice” could not be used to secure a conviction. Id., at 286. On numerous subsequent occasions the Court has set aside convictions secured through the admission of an improperly obtained confession. See, e. g., Mincey v. Arizona, 437 U. S. 385 (1978); Haynes v. Washington, 373 U. S. 503 (1963); Ashcraft v. Tennessee, 322 U. S. 143 (1944); Chambers v. Florida, 309 U. S. 227, 235-238 (1940). Although these decisions framed the legal inquiry in a variety of different ways, usually through the “convenient shorthand” of asking whether the confession was “involuntary,” Blackburn v. Alabama, 361 U. S. 199, 207 (1960), the Court’s analysis has *110consistently been animated by the view that “ours is an accu-satorial and not an inquisitorial system,” Rogers v. Richmond, 365 U. S. 534, 541 (1961), and that, accordingly, tactics for eliciting inculpatory statements must fall within the broad constitutional boundaries imposed by the Fourteenth Amendment’s guarantee of fundamental fairness. Indeed, even after holding that the Fifth Amendment privilege against compulsory self-incrimination applies in the context of custodial interrogations, Miranda v. Arizona, 384 U. S. 436, 478 (1966), and is binding on the States, Malloy v. Hogan, 378 U. S. 1, 6 (1964), the Court has continued to measure confessions against the requirements of due process. See, e. g., Mincey v. Arizona, supra, at 402; Beecher v. Alabama, 389 U. S. 35, 38 (1967) (per curiam).

Without exception, the Court’s confession cases hold that the ultimate issue of “voluntariness” is a legal question requiring independent federal determination. See, e. g., Haynes v. Washington, supra, at 515-516; Ashcraft v. Tennessee, supra, at 147-148. As recently as 1978, the Court reaffirmed that it was “not bound by” a state-court volun-tariness finding and reiterated its historic “duty to make an independent evaluation of the record.” Mincey v. Arizona, supra, at 398. That duty, as Mincey makes explicit, is not limited to instances in which the claim is that the police conduct was “inherently coercive.” Ashcraft v. Tennessee, supra, at 154. It applies equally when the interrogation techniques were improper only because, in the particular circumstances of the case, the confession is unlikely to have been the product of a free and rational will. See Mincey v. Arizona, supra, at 401. Because the ultimate issue in both categories of cases is the same — whether the State has obtained the confession in a manner that comports with due process — the decisions leave no doubt that our independent obligation to decide the constitutional question is identical.

Mincey, Ashcraft, and many of the early decisions applying the independent-determination rule in confession cases came *111to the Court on direct appeal from state-court judgments. The rule, however, is no less firmly established in cases coming to the federal system on application for a writ of habeas corpus. Davis v. North Carolina, 384 U. S. 737 (1966), resolved the issue with unmistakable clarity. There, the State had admitted into evidence a confession elicited from an impoverished, mentally deficient suspect who had been held incommunicado for 16 days with barely adequate nourishment. Expressly relying on the direct-appeal cases, the Court stated unequivocally that state-court determinations concerning the ultimate question of the voluntariness of a confession are not binding in a federal habeas corpus proceeding. Id., at 741-742.

Davis was decided four months before 28 U. S. C. § 2254(d) was signed into law. Act of Nov. 2, 1966, Pub. L. 89-711, 80 Stat. 1105. Respondent contends that, whatever may have been the ease prior to 1966, the enactment of § 2254(d) in that year fundamentally altered the nature of federal habeas review of state voluntariness findings. That suggestion finds no support in this Court’s decisions. See, e. g., Boulden v. Holman, 394 U. S. 478, 480 (1969) (finding confession voluntary after making “an independent study of the entire record”); Frazier v. Cupp, 394 U. S. 731, 739 (1969) (examining “totality of the circumstances” to assess admissibility of confession). More importantly, the history of § 2254(d) undermines any argument that Congress intended that the ultimate question of the admissibility of a confession be treated a “factual issue” within the meaning of that provision. The 1966 amendment was an almost verbatim codification of the standards delineated in Townsend v. Sain, 372 U. S. 293 (1963), for determining when a district court must hold an evidentiary hearing before acting on a ha-beas petition. When a hearing is not obligatory, Townsend held, the federal court “ordinarily should . . . accept the facts as found” in the state proceeding. Id., at 318. Congress elevated that exhortation into a mandatory presumption of *112correctness. But there is absolutely no indication that it intended to alter Townsend’s understanding that the “ultimate constitutional question” of the admissibility of a confession was a “mixed questio[n] of fact and law” subject to plenary federal review. Id., at 309, and n. 6.

In short, an unbroken line of cases, coming to this Court both on direct appeal and on review of applications to lower federal courts for a writ of habeas corpus, forecloses the Court of Appeals’ conclusion that the “voluntariness” of a confession merits something less than independent federal consideration. To be sure, subsidiary factual questions, such as whether a drug has the properties of a truth serum, id., at 306, or whether in fact the police engaged in the intimidation tactics alleged by the defendant, LaVallee v. Delle Rose, 410 U. S. 690, 693-695 (1973) (per curiam), are entitled to the § 2254(d) presumption. And the federal habeas court, should, of course, give great weight to the considered conclusions of a coequal state judiciary. Culombe v. Connecticut, 367 U. S. 568, 605 (1961) (opinion of Frankfurter, J.). But, as we now reaffirm, the ultimate question whether, under the totality of the circumstances, the challenged confession was obtained in a manner compatible with the requirements of the Constitution is a matter for independent federal determination.

h-H HH

The Court of Appeals recognized that treating the volun-tariness of a confession as an issue of fact was difficult to square with “fifty years of caselaw” in this Court. 741F. 2d, at 1462. It believed, however, that this substantial body of contrary precedent was not controlling in light of our more recent decisions addressing the scope of the § 2254(d) presumption of correctness. See Wainwright v. Witt, 469 U. S. 412, 429 (1985) (trial court’s determination that a prospective juror in a capital case was properly excluded for cause entitled to presumption); Patton v. Yount, 467 U. S. 1025 (1984) *113(impartiality of an individual juror); Rushen v. Spain, 464 U. S. 114 (1983) (per curiam) (effect of ex parte communication on impartiality of individual juror); Maggio v. Fulford, 462 U. S. 111 (1983) (per curiam) (competency to stand trial); Marshall v. Lonberger, 459 U. S. 422, 431-437 (1983) (determination that defendant received and understood sufficient notice of charges against him to render guilty plea voluntary). We acknowledge that the Court has not charted an entirely clear course in this area. We reject, however, the Court of Appeals’ conclusion that these case-specific holdings tacitly overturned the longstanding rule that the voluntariness of a confession is a matter for independent federal determination.

In the § 2254(d) context, as elsewhere, the appropriate methodology for distinguishing questions of fact from questions of law has been, to say the least, elusive. See Bose Corp. v. Consumers Union of United States, Inc., 466 U. S. 485 (1984); Baumgartner v. United States, 322 U. S. 665, 671 (1944). A few principles, however, are by now well established. For example, that an issue involves an inquiry into state of mind is not at all inconsistent with treating it as a question of fact. See, e. g., Maggio v. Fulford, supra. Equally clearly, an issue does not lose its factual character merely because its resolution is dispositive of the ultimate constitutional question. See Dayton Board of Education v. Brinkman, 443 U. S. 526, 534 (1979) (finding of intent to discriminate subject to “clearly erroneous” standard of review). But beyond these elemental propositions, negative in form, the Court has yet to arrive at “a rule or principle that will unerringly distinguish a factual finding from a legal conclusion.” Pullman-Standard v. Swint, 456 U. S. 273, 288 (1982).

Perhaps much of the difficulty in this area stems from the practical truth that the decision to label an issue a “question of law,” a “question of fact,” or a “mixed question of law and fact” is sometimes as much a matter of allocation as it is of *114analysis. See Monaghan, Constitutional Fact Review, 85 Colum. L. Rev. 229, 237 (1985). At least in those instances in which Congress has not spoken and in which the issue falls somewhere between a pristine legal standard and a simple historical fact, the fact/law distinction at times has turned on a determination that, as a matter of the sound administration of justice, one judicial actor is better positioned than another to decide the issue in question. Where, for example, as with proof of actual malice in First Amendment libel cases, the relevant legal principle can be given meaning only through its application to the particular circumstances of a case, the Court has been reluctant to give the trier of fact’s conclusions presumptive force and, in so doing, strip a federal appellate court of its primary function as an expositor of law. See Bose Corp. v. Consumers Union of United States, Inc., 466 U. S., at 503. Similarly, on rare occasions in years past the Court has justified independent federal or appellate review as a means of compensating for “perceived shortcomings of the trier of fact by way of bias or some other factor. ...” Id., at 518 (Rehnquist, J., dissenting). See, e. g., Haynes v. Washington, 373 U. S., at 516; Watts v. Indiana, 338 U. S. 49, 52 (1949) (opinion of Frankfurter, J.). Cf. Norris v. Alabama, 294 U. S. 587 (1935).

In contrast, other considerations often suggest the appropriateness of resolving close questions concerning the status of an issue as one of “law” or “fact” in favor of extending deference to the trial court. When, for example, the issue involves the credibility of witnesses and therefore turns largely on an evaluation of demeanor, there are compelling and familiar justifications for leaving the process of applying law to fact to the trial court and according its determinations presumptive weight. Patton v. Yount, supra, and Wainwright v. Witt, supra, are illustrative. There the Court stressed that the state trial judge is in a position to assess juror bias that is far superior to that of federal judges reviewing an application for a writ of habeas corpus. Principally *115for that reason, the decisions held, juror bias merits treatment as a “factual issue” within the meaning of § 2254(d) notwithstanding the intimate connection between such determinations and the constitutional guarantee of an impartial jury.

For several reasons we think that it would be inappropriate to abandon the Court’s longstanding position that the ultimate question of the admissibility of a confession merits treatment as a legal inquiry requiring plenary federal review. We note at the outset that we do not write on a clean slate. “Very weighty considerations underlie the principle that courts should not lightly overrule past decisions.” Moragne v. States Marine Lines, Inc., 398 U. S. 375, 403 (1970). Thus, even assuming that contemporary considerations supported respondent’s construction of the statute, nearly a half century of unwavering precedent weighs heavily against any suggestion that we now discard the settled rule in this area. Moreover, as previously noted, Congress patterned § 2254(d) after Townsend v. Sain, 372 U. S. 293 (1963), a case that clearly assumed that the voluntariness of a confession was an issue for independent federal determination. Thus, not only are stare decisis concerns compelling, but, unlike in Marshall v. Lonberger, 459 U. S. 422 (1983), Rushen v. Spain, 464 U. S. 114 (1983), or any of our other recent § 2254(d) cases, in the confession context we have the benefit of some congressional guidance in resolving whether the disputed issue falls outside of the scope of the § 2254(d) presumption. Although the history of that provision is not without its ambiguities, it is certainly clear enough to tip the scales in favor of treating the voluntariness of a confession as beyond the reach of § 2254(d).

In addition to considerations of stare decisis and congressional intent, the nature of the inquiry itself lends support to the conclusion that “voluntariness” is a legal question meriting independent consideration in a federal habeas corpus proceeding. Although sometimes framed as an issue of *116“psychological fact,” Culombe v. Connecticut, 367 U. S., at 603, the dispositive question of the voluntariness of a confession has always had a uniquely legal dimension. It is telling that in confession cases coming from the States, this Court has consistently looked to the Due Process Clause of the Fourteenth Amendment to test admissibility. See, e. g., Mincey v. Arizona, 437 U. S., at 402. The locus of the right is significant because it reflects the Court’s consistently held view that the admissibility of a confession turns as much on whether the techniques for extracting the statements, as applied to this suspect, are compatible with a system that presumes innocence and assures that a conviction will not be secured by inquisitorial means as on whether the defendant’s will was in fact overborne. See, e. g., Gallegos v. Colorado, 370 U. S. 49, 51 (1962) (suggesting that “a compound of two influences” requires that some confessions be condemned); Culombe v. Connecticut, supra, at 605 (describing voluntariness as an “amphibian”). This hybrid quality of the volun-tariness inquiry,4 subsuming, as it does, a “complex of values,” Blackburn v. Alabama, 361 U. S., at 207, itself militates against treating the question as one of simple historical fact.

Putting to one side whether “voluntariness” is analytically more akin to a fact or a legal conclusion, the practical considerations that have led us to find other issues within the scope of the § 2254(d) presumption are absent in the confession context. First, unlike the impartiality of a given juror, Patton v. Yount, 467 U. S., at 1036, or competency to stand trial, Maggio v. Fulford, 462 U. S., at 117, assessments of credibil*117ity and demeanor are not crucial to the proper resolution of the ultimate issue of “voluntariness.” Of course, subsidiary-questions, such as the length and circumstances of the interrogation, the defendant’s prior experience with the legal process, and familiarity with the Miranda warnings, often require the resolution of conflicting testimony of police and defendant. The law is therefore clear that state-court findings on such matters are conclusive on the habeas court if fairly supported in the record and if the other circumstances enumerated in § 2254(d) are inapplicable. But once such underlying factual issues have been resolved, and the moment comes for determining whether, under the totality of the circumstances, the confession was obtained in a manner consistent with the Constitution, the state-court judge is not in an appreciably better position than the federal habeas court to make that determination.

Second, the allocution of a guilty plea, Marshall v. Lonberger, supra, the adjudication of competency to stand trial, Maggio v. Fulford, supra, and the determination of juror bias, Wainwright v. Witt, 469 U. S. 412 (1985), take place in open court on a full record. In marked contrast, the critical events surrounding the taking of a confession almost invariably occur in a secret and inherently more coercive environment. Miranda v. Arizona, 384 U. S., at 458. These circumstances, standing alone, cannot be dispositive of the question whether a particular issue falls within the reach of § 2254(d). However, together with the inevitable and understandable reluctance to exclude an otherwise reliable admission of guilt, Jackson v. Denno, 378 U. S. 368, 381 (1964), they elevate the risk that erroneous resolution of the voluntariness question might inadvertently frustrate the protection of the federal right. See Haynes v. Washington, 373 U. S., at 516; Ward v. Texas, 316 U. S. 547 (1942). We reiterate our confidence that state judges, no less than their federal counterparts, -will properly discharge their duty to protect the constitutional rights of criminal defendants. We *118note only that in the confession context, independent federal review has traditionally played an important parallel role in protecting the rights at stake when the prosecution secures a conviction through the defendant’s own admissions.

r-H <1

After defending at length its conclusion that the voluntariness of a confession was entitled to the § 2254(d) presumption, and after carefully analyzing the petitioner’s confession under that standard, the Court of Appeals suggested in a brief footnote that it “would reach the same result” even were it to give the issue plenary consideration. 741 F. 2d, at 1467, n. 21. Inasmuch as it is not clear from this language that the court did in fact independently evaluate the admissibility of the confession, and because, in any event, we think that the case warrants fuller analysis under the appropriate standard, we reverse the decision below and remand for further proceedings consistent with this opinion.

It is so ordered.