with whom Mr. Justice Brennan and Mr. Justice Marshall join, dissenting.
In the name of preserving legislative flexibility, the Court today drains In re Winship, 397 U. S. 358 (1970), of much of its vitality. Legislatures do require broad discretion in the drafting of criminal laws, but the Court surrenders to the legislative branch a significant part of its responsibility to protect the presumption of innocence.
I
An understanding of the import of today’s decision requires a comparison of the statutes at issue here with the statutes and practices of Maine struck down by a unanimous Court just two years ago in Mullaney v. Wilbur, 421 U. S. 684 (1975).
A
Maine’s homicide laws embodied the common-law distinctions along with the colorful common-law language. Murder *217was defined in the statute as the unlawful killing of a human being “with malice aforethought, either express or implied.” Manslaughter was a killing “in the heat of passion, on sudden provocation, without express or implied malice aforethought.” Id., at 686, and n. 3. Although “express malice” at one point may have had its own significant independent meaning, see Perkins, A Re-Examination of Malice Aforethought, 43 Yale L. J. 537, 546-552 (1934), in practice a finding that the killing was committed with malice aforethought had come to mean simply that heat of passion was absent. Indeed, the trial court in Mullaney expressly charged the jury that “malice aforethought and heat of passion on sudden provocation are two inconsistent things.” 421 U. S., at 686-687. And the Maine Supreme Judicial Court had held that instructions concerning express malice (in the sense of premeditation) were unnecessary. The only inquiry for the jury in deciding whether a homicide amounted to murder or manslaughter was the inquiry into heat of passion on sudden provocation. State v. Lafferty, 309 A. 2d 647, 664-665 (Me. 1973). See 421 U. S., at 686 n. 4.
Our holding in Mullaney found no constitutional defect in these statutory provisions. Rather, the defect in Maine practice lay in its allocation of the burden of persuasion with respect to the crucial factor distinguishing murder from manslaughter. In Maine, juries were instructed that if the prosecution proved that the homicide was both intentional and unlawful, the crime was to be considered murder unless the defendant proved by a preponderance of the evidence that he acted in the heat of passion on sudden provocation. Only if the defendant carried this burden would the offense be reduced to manslaughter.
New York’s present homicide laws had their genesis in lingering dissatisfaction with certain aspects of the common-law framework that this Court confronted in Mullaney. Critics charged that the archaic language tended to obscure the fac*218tors of real importance in the jury’s decision. Also, only a limited range of aggravations would lead to mitigation under the common-law formula, usually only those resulting from direct provocation by the victim himself. It was thought that actors whose emotions were stirred by other forms of outrageous conduct, even conduct by someone other than the ultimate victim, also should be punished as manslaughterers rather than murderers. Moreover, the common-law formula was generally applied with rather strict objectivity. Only provocations that might cause the hypothetical reasonable man to lose control could be considered. And even provocations of that sort were inadequate to reduce the crime to manslaughter if enough time had passed for the reasonable man’s passions to cool, regardless of whether the actor’s own thermometer had registered any decline. See generally W. LaFave & A. Scott, Criminal Law 528-530, 539-540, 571-582 (1972); Wechsler, Codification of Criminal Law in the United States: The Model Penal Code, 68 Colum. L. Rev. 1425, 1446 (1968); ALI, Model Penal Code § 201.3, Comment (Tent. Draft No. 9, 1959); Perkins, supra. Cf. B. Cardozo, Law and Literature and Other Essays 99-101 (1931).
The American Law Institute took the lead in moving to remedy these difficulties. As part of its commendable undertaking to prepare a Model Penal Code, it endeavored to bring modern insights to bear on the law of homicide. The result was a proposal to replace “heat of passion” with the moderately broader concept of “extreme mental or emotional disturbance.” The proposal first appeared in a tentative draft published in 1959, and it was accepted by the Institute and included as § 210.3 of the 1962 Proposed Official Draft.
At about this time the New York Legislature undertook the preparation of a new criminal code, and the Revised Penal Law of 1967 was the ultimate result. The new code adopted virtually word for word the ALI formula for distinguishing murder from manslaughter. N. Y. Penal Law §§ 125.20 (2), *219125.25 (l)(a) (McKinney 1975).1 Under current New York law,2 those who kill intentionally are guilty of murder. But there is an affirmative defense left open to a defendant: If his act was committed “under the influence of extreme emotional disturbance for which there was a reasonable explanation or excuse,” the crime is reduced to manslaughter. The supposed defects of a formulation like Maine’s have been removed. Some of the rigid objectivity of the common law is relieved, since reasonableness is to be determined “from the viewpoint of a person in the defendant’s situation under the circum*220stances as the defendant believed them to be.” § 125.25 (1) (a). The New York law also permits mitigation when emotional disturbance results from situations other than direct provocation by the victim. And the last traces of confusing archaic language have been removed. There is no mention of malice aforethought, no attempt to give a name to the state of mind that exists when extreme emotional disturbance is not present. The statute is framed in lean prose modeled after the ALI approach, giving operative descriptions of the crucial factors rather than attempting to attach the classical labels.
Despite these changes, the major factor that distinguishes murder from manslaughter in New York — “extreme emotional disturbance” — is undeniably the modern equivalent of “heat of passion.” The ALI drafters made this abundantly clear. They were not rejecting the notion that some of those who kill in an emotional outburst deserve lesser punishment; they were merely refining the concept to relieve some of the problems with the classical formulation. See ALI, Model Penal Code, § 201.3, Comment, pp. 46-48 (Tent. Draft No. 9, 1959). The New York drafters left no doubt about their reliance on the ALI work. See 39 N. Y. 2d 288, 300-301, 347 N. E. 2d 898, 906 (1976). Both the majority and the dissenters in the New York Court of Appeals agreed that extreme emotional disturbance is simply “a new formulation” for the traditional language of heat of passion. Id., at 301, 347 N. E. 2d, at 906; id., at 312, 347 N. E. 2d, at 913-914 (Cooke, J., dissenting).
But in one important respect the New York drafters chose to parallel Maine’s practice precisely, departing markedly from the ALI recommendation. Under the Model Penal Code the prosecution must prove the. absence of emotional disturbance beyond a reasonable doubt once the issue is properly raised. See ALI, Model Penal Code §§ 1.12, 210.3 (Proposed Official Draft 1962); id., § 1.13, Comment, pp. 108-118 (Tent'. Draft No. 4, 1955). In New York, however, extreme emotional disturbance constitutes an affirmative defense rather *221than a simple defense. Consequently the defendant bears not only the burden of production on this issue; he has the burden of persuasion as well. N.' Y. Penal Law § 25.00 (McKinney 1975).
B
Mullaney held invalid Maine’s requirement that the defendant prove heat of passion. The Court today, without disavowing the unanimous holding of Mullaney, approves New York’s requirement that the defendant prove extreme emotional disturbance. The Court manages to run a constitutional boundary line through the barely visible space that separates Maine’s law from New York’s. It does so on the basis of distinctions in language that are formalistic rather than substantive.
This result is achieved by a narrowly literal parsing of the holding in Winship: “[T]he Due Process Clause protects the accused against conviction except upon proof beyond a reasonable doubt of every fact necessary to constitute the crime with which he is charged.” 397 U. S., at 364. The only “facts” necessary to constitute a crime are said to be those that appear on the face of the statute as a part of the definition of the crime.3 Maine’s statute was invalid, the Court reasons, because it “defined [murder] as the unlawful killing of a human being 'with malice aforethought, either express or implied.’ ” Ante, at 212. “[MJalice,” the Court reiterates, “in the sense of the absence of provocation, was part of the definition of that crime.” Ante, at 216. Winship was violated only because this “fact” — malice—was “presumed” unless the defendant persuaded the jury otherwise by showing that he acted in the heat of passion.4 New York, in form presuming *222no affirmative “fact” against Patterson,5 and blessed with a statute drafted in the leaner language of the 20th century, escapes constitutional scrutiny unscathed even though the effect on the defendant of New York’s placement of the burden of persuasion is exactly the same as Maine’s. See 39 N. Y. 2d, at 312-313, 347 N. E. 2d, at 913-914 (Cooke, J., dissenting).
This explanation of the Mullaney holding bears little re*223semblance to the basic rationale of that decision.6 But this is not the cause of greatest concern. The test the Court today establishes allows a legislature to shift, virtually at will, the-burden of persuasion with respect to any factor in a criminal case, so long as it is careful not to mention the nonexistence of that factor in the statutory language that defines the crime. The sole requirement is that any references to the factor be confined to those sections that provide for an affirmative defense.7
Perhaps the Court’s interpretation of Winship is consistent with the letter of the holding in that case. But little of the spirit survives. Indeed, the Court scarcely could distinguish this case from Mullaney without closing its eyes to the constitutional values for which Winship stands. As Mr. Justice Harlan observed in Winship, “& standard of proof represents an attempt to instruct the factfinder concerning the degree of *224confidence our society thinks he should have in the correctness of factual conclusions for a particular type of adjudication.” 397 U. S., at 370 (concurring opinion). See Speiser v. Randall, 357 U. S. 513, 525-526 (1958). Explaining Mul-laney, the Court says today, in effect, that society demands full confidence before a Maine factfinder determines that heat of passion is missing — a demand so insistent that this Court invoked the Constitution to enforce it over the contrary decision by the State. But we are told that society is willing to tolerate far less confidence in New York’s factual determination of precisely the same functional issue. One must ask what possibly could explain this difference in societal demands. According to the Court, it is because Maine happened to attach a name — “malice aforethought” — to the absence of heat of passion, whereas New York refrained from giving a name to the absence of extreme emotional disturbance. See 39 N. Y. 2d, at 313, 347 N. E. 2d, at 914 (Cooke, J., dissenting).
With all respect, this type of constitutional adjudication is indefensibly formalistic. A limited but significant check on possible abuses in the criminal law now becomes an exercise in arid formalities. What Winship and Mullaney had sought to teach about the limits a free society places on its procedures to safeguard the liberty of its citizens becomes a rather simplistic lesson in statutory draftsmanship. Nothing in the Court’s opinion prevents a legislature from applying this new learning to many of the classical elements of the crimes it punishes.8 It would be preferable, if the Court has found *225reason to reject the rationale of Winship and Mullaney, simply and straightforwardly to overrule those precedents.
The Court understandably manifests some uneasiness that its formalistic approach will give legislatures too much latitude in shifting the burden of persuasion. And so it issues a warning-that “there are obviously constitutional limits beyond which the States may not go in this regard.” Ante, at 210; The Court thereby concedes that legislative abuses may occur and that they must be curbed by the judicial branch. But if the State is careful to conform to the drafting formulas articulated today, the constitutional limits are anything but “obvious.” This decision simply leaves us without a conceptual framework for distinguishing abuses from legitimate legislative adjustments of the burden of persuasion in criminal cases.9
II
It is unnecessary for the Court to retreat to a formalistic test for applying Winship. Careful attention to the Mullaney decision reveals the principles that should control in this and like cases. Winship held that the prosecution must bear the burden of proving beyond a reasonable doubt “ ‘the existence of every fact necessary to constitute the crime charged.’ ” 397 U. S., at 363, quoting Davis v. United States, 160 U. S. 469, 493 (1895). In Mullaney we concluded that heat of passion was one of the “facts” described in Winship — that is, a *226factor as to which the prosecution must bear the burden of persuasion beyond a reasonable doubt. 421 U. S., at 704. We reached that result only after making two careful inquiries. First, we noted that the presence or absence of heat of passion made a substantial difference in punishment of the offender and in the stigma associated with the conviction. Id., at 697-701. Second, we reviewed the history, in England and this country, of the factor at issue. Id., at 692-696. Central to the holding in Mullaney was our conclusion that heat of passion “has been, almost from the inception of the common law of homicide, the single most important factor in determining the degree of culpability attaching to an unlawful homicide.” Id., at 696.
Implicit in these two inquiries are the principles that should govern this case. The Due Process Clause requires that the prosecutor bear the burden of persuasion beyond a reasonable doubt only if the factor at issue makes a substantial difference in punishment and stigma. The requirement of course applies a fortiori if the factor makes the difference between guilt and innocence. But a substantial difference in punishment alone is not enough. It also must be shown that in the Anglo-American legal tradition 10 the factor in question historically has held that level of importance.11 If either branch *227of the test is not met, then the legislature retains its traditional authority over matters of proof. But to permit a shift in the burden of persuasion when both branches of this test are satisfied would invite the undermining of the presumption of innocence, “that bedrock 'axiomatic and elementary’ principle whose ‘enforcement lies at the foundation of the administration of our criminal law.’ ” In re Winship, 397 U. S., at 363, quoting from Coffin v. United States, 156 U. S. 432, 453 (1895). See Cool v. United States, 409 U. S. 100, 104 (1972); Ivan V. v. City of New York, 407 U. S. 203, 204 (1972); Lego v. Twomey, 404 U. S. 477, 486-487 (1972); Morissette v. United States, 342 U. S. 246, 275 (1952); Bailey v. Alabama, 219 U. S. 219, 236 (1911); Davis v. United States, supra. This is not a test that rests on empty form, for “Winship is concerned with substance rather than . . . formalism.” Mullaney v. Wilbur, 421 U. S., at 699.
I hardly need add that New York’s provisions allocating the burden of persuasion as to “extreme emotional disturbance” are unconstitutional when judged by these standards. “Extreme emotional disturbance” is, as the Court of Appeals recognized, the direct descendant of the “heat of passion” factor considered at length in Mullaney. I recognize, of course, that the differences between Maine and New York law are not unimportant to the defendant; there is a somewhat broader opportunity for mitigation. But none of those distinctions is relevant here. The presence or absence of extreme emotional disturbance makes a critical difference in punishment and stigma, and throughout our history the resolution of this issue of fact, although expressed in somewhat different terms, has distinguished manslaughter from murder. See 4 W. Blackstone, Commentaries *190-193, 198-201.
Ill
The Court beats its retreat from Winship apparently because of a concern that otherwise the federal judiciary will in-*228trade too far into substantive choices concerning the content of a State’s criminal law.12 The concern is legitimate, see generally Powell v. Texas, 392 U. S. 514, 533-534 (1968) (plurality opinion); Leland v. Oregon, 343 U. S. 790, 803 (1952)' (Frankfurter, J., dissenting), but misplaced. Winship and Mullaney are no more than what they purport to be: decisions addressing the procedural requirements that States must meet to comply with due process. They are not outposts for policing the substantive boundaries of the criminal law.
The Winship/Mullaney test identifies those factors of such importance, historically, in determining punishment and stigma that the Constitution forbids shifting to the defendant the burden of persuasion when such a factor is at issue. Win-ship and Mullaney specify only the procedure that is required when a State elects to use such a factor as part of its substantive criminal law. They do not say that the State must elect to use it. For example, where a State has chosen to retain the traditional distinction between murder and manslaughter, as have New York and Maine, the burden of persuasion must remain on the prosecution with respect to the distinguishing factor, in view of its decisive historical importance. But nothing in Mullaney or Winship precludes a State from'abolishing the distinction between murder and manslaughter and treating all unjustifiable homicide as murder.13 In this sig*229nificant respect, neither Winship nor Mullaney eliminates the substantive flexibility that should remain in legislative hands.
Moreover, it is unlikely that more than a few factors — although important ones — for which a shift in the burden of persuasion seriously would be considered will come within the Mullaney holding. With some exceptions, then, the State has the authority “to recognize a factor that mitigates the degree of criminality or punishment” without having “to prove its nonexistence in each case in which the fact is put in issue.” Ante, at 209. New ameliorative affirmative defenses,14 about *230which the Court expresses concern, generally remain undisturbed by the holdings in Winship and Mullaney — and need not be disturbed by a sound holding reversing Patterson’s conviction.15
Furthermore, as we indicated in Mullaney, 421 U. S., at 701-702, n. 28, even as to those factors upon which the prosecution must bear the burden of persuasion, the State retains an important procedural device to avoid jury confusion and prevent the prosecution from being unduly hampered. The State normally may shift to the defendant the burden of production,16 that is, the burden of going forward with sufficient *231evidence “to justify [a reasonable] doubt upon the issue.” 17 ALI, Model Penal Code § 1.13, Comment, p. 110 (Tent. Draft No. 4, 1955). If the defendant’s evidence does not cross this threshold, the issue — be it malice, extreme emotional disturbance, self-defense, or whatever — will not be submitted to the jury.18 See Sansone v. United States, 380 U. S. 343, 349 (1965); Stevenson v. United States, 162 U. S. 313, 314-316 (1896). Ever since this Court’s decision in Davis v. United States, 160 U. S. 469 (1895), federal prosecutors have borne the burden of persuasion with respect to factors like insanity, self-defense, and malice or provocation, once the defendant has carried this burden of production. See, e. g., Blake v. United States, 407 F. 2d 908, 910-911 (CA5 1969) (en banc) (insanity); Frank v. United States, 42 F. 2d 623, 629 (CA9 1930) (self-defense); United States v. Alexander, 152 U. S. App. D. C. 371, 389-395, 471 F. 2d 923, 941-947, cert. denied sub nom. Murdock v. United States, 409 U. S. 1044 (1972) (provocation). I know of no indication that this *232practice has proven a noticeable handicap to effective law enforcement.19
To be sure, there will be many instances when the Winship/ Mullaney test as I perceive it will be more difficult to apply than the Court’s formula. Where I see the need for a careful and discriminating review of history, the Court finds a bright-line standard that can be applied with a quick glance at the face of the statute. But this facile test invites tinkering with the procedural safeguards of the presumption of innocence, an invitation to disregard the principles of Winship that I would not extend.
There are also other forms of manslaughter set forth in the New York statute, not all of which conform to the ALI recommendations. Those provisions are not implicated in this case.
The 1967 provisions marked a considerable departure from the prior New York statutes defining manslaughter. As we noted in Mullaney v. Wilbur, 421 U. S. 684, 694 (1975), the grounds for distinguishing murder from manslaughter developed along two distinct paths in this country. Prior to the 1967 change New York, with a handful of other jurisdictions, see ALI, Model Penal Code § 201.3, Comment, p. 43 (Tent. Draft No. 9, 1959), pursued the first path: to establish malice (and hence to convict of murder) the prosecution bore the burden of persuasion, being required to establish a substantive element of intent- — that the defendant possessed “a design to effect death.” See 39 N. Y. 2d 288, 299, 347 N. E. 2d 898, 905 (1976) (case below); Stokes v. People, 53 N. Y. 164 (1873). Maine, in contrast, followed the second path, marked out most prominently by Chief Justice Shaw’s opinion in Commonwealth v. York, 50 Mass. 93 (1845): malice was presumed unless the defendant established that he acted in the heat of passion.
This difference between the old New York practice and the York approach was substantial — as noted by the Court of Appeals below. But that court placed entirely too much weight on this distinction as a basis for concluding that Mullaney’s holding was inapplicable. The statute at issue here is the 1967 Revised Penal Law, not the earlier formulation. In 1967, New York broke from the first branch and aligned itself with York, although casting its statute in more modern language. No matter how extensive the differences between the pre-1967 practice and the Maine statutes found deficient in Mullaney, this case must be decided on the basis of current New York law.
The Court holds that the prosecution must prove beyond a reasonable doubt “all of the elements included in the definition of the offense of which the defendant is charged.” Ante, at 210 (emphasis added).
The Court explains: “Such shifting of the burden of persuasion with respect to a fact which the State deems so important that it must be *222either proved or presumed is impermissible under the Due Process Clause.” Ante, at 215. I must point out, however, that this is a less than faithful reading of Maine law. The Maine Supreme Judicial Court, rejecting a recent holding to the contrarjr by the Court of Appeals for the First Circuit, emphatically insisted that the words “malice aforethought” appearing in the Maine statute did not connote a “fact” to be “presumed” in the sense the latter terms are customarily used:
“As we read the [First Circuit] case, the Federal Court was of the impression that [murder] includes, in addition to an intentional and unlawful killing, the independent element of ‘malice aforethought.’ Such is not, and never has been, the law in Maine. As we said in [State v. Rollins, 295 A. 2d 914, 920 (1972)]:
‘“[T]be “malice” (said to be “presumed”) is not a designation of any subjective state of mind existing as a fact. Similarly, the “presumption” (of “malice”) arising from the fact of an intentional killing is not a designation of any probative relationship between the fact of “intention” relating to the killing and any further facts State v. Lafferty, 309 A. 2d 647, 664 (1973) (emphasis in original).
See id., at 672 (concurring opinion); Mullaney v. Wilbur, 421 U. S., at 689, 699.
“The crime of murder is defined by the [New York] statute ... as causing the death of another person with intent to do so. The death, the intent to kill, and causation are the facts that the State is required to prove beyond a reasonable doubt if a person is to be convicted of murder. No further facts are either presumed or inferred in order to constitute the crime. . . .
"... [The] affirmative defense [of extreme emotional disturbance] . . . does not serve to negative any facts of the crime which the State is to prove in order to convict of murder.” Ante, at 205-206, 206-207.
In Mullaney we made it clear that Winship is not “limited to a State’s definition of the elements of a crime.” 421 U. S., at 699 n. 24.
Although the Court never says so explicitly, its new standards appear to be designed for application to the language of a criminal statute on its face, regardless of how the state court construes the statute. The Court, in explaining Mullaney, persistently states that in Maine malice “was part of the definition of that crime [murder],” ante, at 216, even though the Maine Supreme Judicial Court, construing its own statute, had ruled squarely to the contrary. See n. 4, supra. In the usual ease it is well established that an authoritative construction by the State’s highest court “puts [appropriate] words in the statute as definitely as if it had been so amended by the legislature.” Winters v. New York, 333 U. S. 507, 514 (1948). See Mullaney, supra, at 690-691; Hebert v. Louisiana, 272 U. S. 312, 316-317 (1926); Murdock v. Memphis, 20 Wall. 590, 635 (1875). Why an apparent exception should be engrafted on that doctrine today goes unexplained.
The result, under the Court’s holding, is that only the legislature can remedy any defects that come to light as a result of the Court’s decision. No matter how clear the legislative intent that defendants bear the burden of persuasion on an issue — an ultimate result the Court approves — state courts may not effectuate that intent until the right verbal formula appears in the statute book.
For example, a state statute could pass muster under the only solid standard that appears in the Court’s opinion if it defined murder as mere physical contact between the defendant and the victim leading to the victim’s death, but then set up an affirmative defense leaving it to the defendant to prove that he acted without culpable mens rea. The State, in other words, could be relieved altogether of responsibility for proving anything regarding the defendant’s state of mind, provided only that the face of the statute meets the Court’s drafting formulas.
To be sure, it is unlikely that legislatures will rewrite their criminal laws *225in this extreme form. The Court seems to think this likelihood, of restraint is an added reason for limiting review largely to formalistic examination. Ante, at 211. But it is completely foreign to this Court's responsibility for constitutional adjudication to limit the scope of judicial review because of the expectation — however reasonable — that legislative bodies will exercise appropriate restraint.
1 have no doubt that the Court would find some way to strike down a formalistically correct statute as egregious as the one hypothesized in n. 8, supra. Cf. Morissette v. United States, 342 U. S. 246, 250-263 (1952). But today’s ruling suggests no principled basis for concluding that such a statute falls outside the “obvious” constitutional limits the Court invokes.
Cf. Brinegar v. United States, 338 U. S. 160, 174 (1949):
“Guilt in a criminal case must be proved beyond a reasonable doubt and by evidence confined to that which long experience in the common-law tradition, to some extent embodied in the Constitution, has crystallized into rules of evidence consistent with that standard. These rules are historically grounded rights of our system, developed to safeguard men from dubious and unjust convictions, with resulting forfeitures of life, liberty and property.”
As the Court acknowledges, ante, at 207-208, n. 10, the clear trend over the years has been to require the prosecutor to carry the burden of persuasion with respect to all important factors in a criminal case, including traditional affirmative defenses. See W. LaFave & A. Scott, Criminal Law 50 (1972); C. McCormick, Evidence § 341, pp. 800-802 (1972).
See Low & Jeffries, DICTA: Constitutionalizing the Criminal Law?, 29 Va. Law Weekly, No. 18, p. 1 (1977); Tushnet, Constitutional Limitation of Substantive Criminal Law: An Examination of the Meaning of Mullaney v. Wilbur, 55 B. U. L. Rev. 775 (1975).
Perhaps under other principles of due process jurisprudence, certain factors are so fundamental that a State could not, as a substantive matter, refrain from recognizing them so 'long as it chooses to punish given conduct as a crime. Cf. Bailey v. Alabama, 219 U. S. 219 (1911) (holding a criminal-law presumption invalid proeedurally and also finding a substantive defect under the Thirteenth Amendment and the Anti-Peonage Act). But substantive limits were not at issue in Winship or Mullaney, and they are not at issue here.
Even if there are no constitutional limits preventing the State, for *229example, from treating all homicides as murders punishable equally regardless of mitigating factors like heat of passion or extreme emotional disturbance, the Winship/Mullaney rule still plays an important role. The State is then obliged to make its choices concerning the substantive content of its criminal laws with full awareness of the conso-quences, unable to mask substantive policy choices by shifts in the burden of persuasion. See Fletcher, Two Kinds of Legal Rules: A Comparative Study of Burden-of-Persuasion Practices in Criminal Cases, 77 Yale L. J. 880, 894 (1968) (“The burden of persuasion has proved to be a subtle, low-visibility tool for adjusting the interests of competing classes of litigants”). The political check on potentially harsh legislative action is then more likely to operate. Cf. Tot v. United States, 319 U. S. 463, 472 (1943); United States v. Romano, 382 U. S. 136 (1965).
Romano involved a challenge to a federal statute that authorized the jury to infer possession, custody, and control of an illegal still from mere presence at the site. The Government contended that the statute should be sustained since it was merely Congress’ way of broadening the substantive provisions in order to make a crime of mere presence. The Court rejected this argument, serving notice that Congress could not work a substantive change of that magnitude in such a disguised form. Id., at 144. See Ashford & Risinger, Presumptions, Assumptions, and Due Process in Criminal Cases: A Theoretical Overview, 79 Yale L. J. 165, 177-178 (1969); Osenbaugh, The Constitutionality of Affirmative Defenses to Criminal Changes, 29 Ark. L. Rev. 429, 461 (1976).
Numerous examples of such defenses are available: New York subjects an armed robber to lesser punishment than he would otherwise receive if he proves by a preponderance of the evidence that the gun he used was unloaded or inoperative. N. Y. Penal Law §160.15 (McKinney 1975). A number of States have ameliorated the usual operation of statutes *230punishing statutory rape, recognizing a defense if the defendant shows that he reasonably believed his partner was of age. E. g., Ky. Rev. Stat. Ann. §§ 500.070, 510.030 (1975); Wash. Rev. Code Ann. § 9.79.160 (2) (Supp. 1975). Formerly the age of the minor was a strict-liability element of the crime. The Model Penal Code also employs such a shift in the burden of persuasion for a limited number of defenses. For example, a corporation can escape conviction of an offense if it proves by a preponderance of the evidence that the responsible supervising officer exercised due diligence to prevent the commission of the offense. § 2.07 (5) (Proposed Official Draft 1962).
A number of commentators have suggested that the Constitution permits the States some latitude in adjusting the burden of persuasion with respect to new ameliorative affirmative defenses that result from legislative compromise, but not with respect to other factors. See, e. g., W. LaFave & A. Scott, supra, n. 11, at 49; 1 National Commission on Reform of Federal Criminal Laws, Working Papers 18-19 (1970); ALI, Model Penal Code § 1.13, Comment, p. 113 (Tent. Draft No. 4, 1955) (quoted, ante, at 209 n. 11); Note, 51 Wash. L. Rev. 953, 964 (1976); Osenbaugh, supra, n. 13, at 459-467. Cf. Fletcher, supra, n. 13, at 928-929.
There are outer limits on shifting the burden of production to a defendant, limits articulated in a long line of cases in this Court passing on the validity of presumptions. Most important are the “rational connection” requirement of Mobile, J. & K. C. R. Co. v. Turnipseed, 219 U. S. 35, 43 (1910), and Bailey v. Alabama, supra, at 238-239, and also the “comparative convenience” criterion of Morrison v. California, *231291 U. S. 82 (1934). See also, e. g., Tot v. United States, supra, at 467-468; Speiser v. Randall, 357 U. S. 513, 523-524 (1958); Leary v. United States, 395 U. S. 6, 33-34 (1969); Barnes v. United States, 412 U. S. 837, 843 (1973). Caution is appropriate, however, in generalizing about the application of any of these cases to a given procedural device, since the term “presumption” covers a broad range of procedural mechanisms having significantly different consequences for the defendant. See McCormick, n. 11, supra, at 802-806; Evans v. State, 28 Md. App. 640, 675-678, 349 A. 2d 300, 324-325 (1975).
This does not mean that the defendant must introduce evidence in every case. In some instances the prosecution’s case may contain sufficient evidence in support of the defendant’s position to generate a jury issue.
On many occasions this Court has sustained a trial court’s refusal to submit an issue to the jury in a criminal case when the defendant failed to meet his burden of production. See, e. g., Sparf v. United States, 156 U. S. 51, 63-64 (1895); Andersen v. United States, 170 U. S. 481, 510-511 (1898); Battle v. United States, 209 U. S. 36, 38 (1908). Cf. Galloway v. United States, 319 U. S. 372, 395 (1943).
Dean McCormick 'emphasized that the burden of production is “a critical and important mechanism in a jury trial.” In his view, “this mechanism has far more influence upon the final outcome of cases than does the burden of persuasion, which has become very largely a matter of the technique of the wording of instructions to juries.” C. McCormick, Evidence § 307, pp. 638-639, and n. 2 (1st ed. 1954). Cf. Fletcher, supra, n. 13, at 930,