Justice White

delivered the opinion of the Court.

This case requires us to determine the constitutionality of the use of deadly force to prevent the escape of an apparently unarmed suspected felon. We conclude that such force may not be used unless it is necessary to prevent the escape and the officer has probable cause to believe that the suspect poses a significant threat of death or serious physical injury to the officer or others.

I

At about 10:45 p. m. on October 3, 1974, Memphis Police Officers Elton Hymon and Leslie Wright were dispatched to answer a “prowler inside call.” Upon arriving at the scene they saw a woman standing on her porch and gesturing toward the adjacent house.1 She told them she had heard glass breaking and that “they” or “someone” was breaking in next door. While Wright radioed the dispatcher to say that they were on the scene, Hymon went behind the house. He heard a door slam and saw someone run across the backyard. The fleeing suspect, who was appellee-respondent’s decedent, Edward Garner, stopped at a 6-feet-high chain link fence at the edge of the yard. With the aid of a flashlight, Hymon was able to see Garner’s face and hands. He saw no sign of a weapon, and, though not certain, was “reasonably sure” and “figured” that Garner was unarmed. App. 41, 56; Record 219. He thought Garner was 17 or 18 years old and *4about 5' 5" or 5' 7" tall.2 While Garner was crouched at the base of the fence, Hymon called out “police, halt” and took a few steps toward him. Garner then began to climb over the fence. Convinced that if Garner made it over the fence he would elude capture,3 Hymon shot him. The bullet hit Garner in the back of the head. Garner was taken by ambulance to a hospital, where he died on the operating table. Ten dollars and a purse taken from the house were found on his body.4

In using deadly force to prevent the escape, Hymon was acting under the authority of a Tennessee statute and pursuant to Police Department policy. The statute provides that “[i]f, after notice of the intention to arrest the defendant, he either flee or forcibly resist, the officer may use all the necessary means to effect the arrest.” Tenn. Code Ann. *5§40-7-108 (1982).5 The Department policy was slightly more restrictive than the statute, but still allowed the use of deadly force in cases of burglary. App. 140-144. The incident was reviewed by the Memphis Police Firearm’s Review Board and presented to a grand jury. Neither took any action. Id., at 57.

Garner’s father then brought this action in the Federal District Court for the Western District of Tennessee, seeking damages under 42 U. S. C. § 1983 for asserted violations of Garner’s constitutional rights. The complaint alleged that the shooting violated the Fourth, Fifth, Sixth, Eighth, and Fourteenth Amendments of the United States Constitution. It named as defendants Officer Hymon, the Police Department, its Director, and the Mayor and city of Memphis. After a 3-day bench trial, the District Court entered judgment for all defendants. It dismissed the claims against the Mayor and the Director for lack of evidence. It then concluded that Hymon’s actions were authorized by the Tennessee statute, which in turn was constitutional. Hymon had employed the only reasonable and practicable means of preventing Garner’s escape. Garner had “recklessly and heedlessly attempted to vault over the fence to escape, thereby assuming the risk of being fired upon.” App. to Pet. for Cert. A10.

The Court of Appeals for the Sixth Circuit affirmed with regard to Hymon, finding that he had acted in good-faith reliance on the Tennessee statute and was therefore within the scope of his qualified immunity. 600 F. 2d 52 (1979). It remanded for reconsideration of the possible liability of the city, however, in light of Monell v. New York City Dept. of Social Services, 436 U. S. 658 (1978), which had come down after the District Court’s decision. The District Court was *6directed to consider whether a city enjoyed a qualified immunity, whether the use of deadly force and hollow point bullets in these circumstances was constitutional, and whether any unconstitutional municipal conduct flowed from a “policy or custom” as required for liability under Monell. 600 F. 2d, at 54-55.

The District Court concluded that Monell did not affect its decision. While acknowledging some doubt as to the possible immunity of the city, it found that the statute, and Hymon’s actions, were constitutional. Given this conclusion, it declined to consider the “policy or custom” question. App. to Pet. for Cert. A37-A39.

The Court of Appeals reversed and remanded. 710 F. 2d 240 (1983). It reasoned that the killing of a fleeing suspect is a “seizure” under the Fourth Amendment,6 and is therefore constitutional only if “reasonable.” The Tennessee statute failed as applied to this case because it did not adequately limit the use of deadly force by distinguishing between felonies of different magnitudes — “the facts, as found, did not justify the use of deadly force under the Fourth Amendment.” Id., at 246. Officers cannot resort to deadly force unless they “have probable cause ... to believe that the suspect [has committed a felony and] poses a threat to the safety of the officers or a danger to the community if left at large.” Ibid.7

*7The State of Tennessee, which had intervened to defend the statute, see 28 U. S. C. § 2403(b), appealed to this Court. The city filed a petition for certiorari. We noted probable jurisdiction in the appeal and granted the petition. 465 U. S. 1098 (1984).

II

Whenever an officer restrains the freedom of a person to walk away, he has seized that person. United States v. Brignoni-Ponce, 422 U. S. 873, 878 (1975). While it is not always clear just when minimal police interference becomes a seizure, see United States v. Mendenhall, 446 U. S. 544 (1980), there can be no question that apprehension by the use of deadly force is a seizure subject to the reasonableness requirement of the Fourth Amendment.

A

A police officer may arrest a person if he has probable cause to believe that person committed a crime. E. g., United States v. Watson, 423 U. S. 411 (1976). Petitioners and appellant argue that if this requirement is satisfied the Fourth Amendment has nothing to say about how that seizure is made. This submission ignores the many cases in which this Court, by balancing the extent of the intrusion against the need for it, has examined the reasonableness of *8the manner in which a search or seizure is conducted. To determine the constitutionality of a seizure “[w]e must balance the nature and quality of the intrusion on the individual’s Fourth Amendment interests against the importance of the governmental interests alleged to justify the intrusion.” United States v. Place, 462 U. S. 696, 703 (1983); see Delaware v. Prouse, 440 U. S. 648, 654 (1979); United States v. Martinez-Fuerte, 428 U. S. 543, 555 (1976). We have described “the balancing of competing interests” as “the key principle of the Fourth Amendment.” Michigan v. Summers, 452 U. S. 692, 700, n. 12 (1981). See also Camara v. Municipal Court, 387 U. S. 523, 536-537 (1967). Because one of the factors is the extent of the intrusion, it is plain that reasonableness depends on not only when a seizure is made, but also how it is carried out. United States v. Ortiz, 422 U. S. 891, 895 (1975); Terry v. Ohio, 392 U. S. 1, 28-29 (1968).

Applying these principles to particular facts, the Court has held that governmental interests did not support a lengthy detention of luggage, United States v. Place, supra, an airport seizure not “carefully tailored to its underlying justification,” Florida v. Royer, 460 U. S. 491, 500 (1983) (plurality-opinion), surgery under general anesthesia to obtain evidence, Winston v. Lee, 470 U. S. 753 (1985), or detention for fingerprinting without probable cause, Davis v. Mississippi, 394 U. S. 721 (1969); Hayes v. Florida, 470 U. S. 811 (1985). On the other hand, under the same approach it has upheld the taking of fingernail scrapings from a suspect, Cupp v. Murphy, 412 U. S. 291 (1973), an unannounced entry into a home to prevent the destruction of evidence, Ker v. California, 374 U. S. 23 (1963), administrative housing inspections without probable cause to believe that a code violation will be found, Camara v. Municipal Court, supra, and a blood test of a drunken-driving suspect, Schmerber v. California, 384 U. S. 757 (1966). In each of these cases, the question was whether *9the totality of the circumstances justified a particular sort of search or seizure.

B

The same balancing process applied in the cases cited above demonstrates that, notwithstanding probable cause to seize a suspect, an officer may not always do so by killing him. The intrusiveness of a seizure by means of deadly force is unmatched. The suspect’s fundamental interest in his own life need not be elaborated upon. The use of deadly force also frustrates the interest of the individual, and of society, in judicial determination of guilt and punishment. Against these interests are ranged governmental interests in effective law enforcement.8 It is argued that overall violence will be reduced by encouraging the peaceful submission of suspects who know that they may be shot if they flee. Effectiveness in making arrests requires the resort to deadly *10force, or at least the meaningful threat thereof. “Being able to arrest such individuals is a condition precedent to the state’s entire system of law enforcement.” Brief for Petitioners 14.

Without in any way disparaging the importance of these goals, we are not convinced that the use of deadly force is a sufficiently productive means of accomplishing them to justify the killing of nonviolent suspects. Cf. Delaware v. Prouse, supra, at 659. The use of deadly force is a self-defeating way of apprehending a suspect and so setting the criminal justice mechanism in motion. If successful, it guarantees that that mechanism will not be set in motion. And while the meaningful threat of deadly force might be thought to lead to the arrest of more live suspects by discouraging escape attempts,9 the presently available evidence does not support this thesis.10 The fact is that a majority of police de*11partments in this country have forbidden the use of deadly force against nonviolent suspects. See infra, at 18-19. If those charged with the enforcement of the criminal law have abjured the use of deadly force in arresting nondangerous felons, there is a substantial basis for doubting that the use of such force is an essential attribute of the arrest power in all felony cases. See Schumann v. McGinn, 307 Minn. 446, 472, 240 N. W. 2d 525, 540 (1976) (Rogosheske, J., dissenting in part). Petitioners and appellant have not persuaded us that shooting nondangerous fleeing suspects is so vital as to outweigh the suspect’s interest in his own life.

The use of deadly force to prevent the escape of all felony suspects, whatever the circumstances, is constitutionally unreasonable. It is not better that all felony suspects die than that they escape. Where the suspect poses no immediate threat to the officer and no threat to others, the harm resulting from failing to apprehend him does not justify the use of deadly force to do so. It is no doubt unfortunate when a suspect who is in sight escapes, but the fact that the police arrive a little late or are a little slower afoot does not always justify killing the suspect. A police officer may not seize an unarmed, nondangerous suspect by shooting him dead. The Tennessee statute is unconstitutional insofar as it authorizes the use of deadly force against such fleeing suspects.

It is not, however, unconstitutional on its face. Where the officer has probable cause to believe that the suspect poses a threat of serious physical harm, either to the officer or to others, it is not constitutionally unreasonable to prevent escape by using deadly force. Thus, if the suspect threatens the officer with a weapon or there is probable cause to believe that he has committed a crime involving the infliction or threatened infliction of serious physical harm, deadly force may be used if necessary to prevent escape, and if, where *12feasible, some warning has been given. As applied in such circumstances, the Tennessee statute would pass constitutional muster.

Ill

A

It is insisted that the Fourth Amendment must be construed in light of the common-law rule, which allowed the use of whatever force was necessary to effect the arrest of a fleeing felon, though not a misdemeanant. As stated in Hale’s posthumously published Pleas of the Crown:

“[I]f persons that are pursued by these officers for felony or the just suspicion thereof . . . shall not yield themselves to these officers, but shall either resist or fly before they are apprehended or being apprehended shall rescue themselves and resist or fly, so that they cannot be otherwise apprehended, and are upon necessity slain therein, because they cannot be otherwise taken, it is no felony.” 2 M. Hale, Historia Placitorum Coronae 85 (1736).

See also 4 W. Blackstone, Commentaries *289. Most American jurisdictions also imposed a flat prohibition against the use of deadly force to stop a fleeing misdemeanant, coupled with a general privilege to use such force to stop a fleeing felon. E. g., Holloway v. Moser, 193 N. C. 185, 136 S. E. 375 (1927); State v. Smith, 127 Iowa 534, 535, 103 N. W. 944, 945 (1905); Reneau v. State, 70 Tenn. 720 (1879); Brooks v. Commonwealth, 61 Pa. 352 (1869); Roberts v. State, 14 Mo. 138 (1851); see generally R. Perkins & R. Boyce, Criminal Law 1098-1102 (3d ed. 1982); Day, Shooting the Fleeing Felon: State of the Law, 14 Crim. L. Bull. 285, 286-287 (1978); Wilgus, Arrest Without a Warrant, 22 Mich. L. Rev. 798, 807-816 (1924). But see Storey v. State, 71 Ala. 329 (1882); State v. Bryant, 65 N. C. 327, 328 (1871); Caldwell v. State, 41 Tex. 86 (1874).

*13The State and city argue that because this was the prevailing rule at the time of the adoption of the Fourth Amendment and for some time thereafter, and is still in force in some States, use of deadly force against a fleeing felon must be “reasonable.” It is true that this Court has often looked to the common law in evaluating the reasonableness, for Fourth Amendment purposes, of police activity. See, e. g., United States v. Watson, 423 U. S. 411, 418-419 (1976); Gerstein v. Pugh, 420 U. S. 103, 111, 114 (1975); Carroll v. United States, 267 U. S. 132, 149-153 (1925). On the other hand, it “has not simply frozen into constitutional law those law enforcement practices that existed at the time of the Fourth Amendment’s passage.” Payton v. New York, 445 U. S. 573, 591, n. 33 (1980). Because of sweeping change in the legal and technological context, reliance on the common-law rule in this case would be a mistaken literalism that ignores the purposes of a historical inquiry.

B

It has been pointed out many times that the common-law rule is best understood in light of the fact that it arose at a time when virtually all felonies were punishable by death.11 “Though effected without the protections and formalities of an orderly trial and conviction, the killing of a resisting or *14fleeing felon resulted in no greater consequences than those authorized for punishment of the felony of which the individual was charged or suspected.” American Law Institute, Model Penal Code §3.07, Comment 3, p. 56 (Tentative Draft No. 8, 1958) (hereinafter Model Penal Code Comment). Courts have also justified the common-law rule by emphasizing the relative dangerousness of felons. See, e. g., Schumann v. McGinn, 307 Minn., at 458, 240 N. W. 2d, at 533; Holloway v. Moser, supra, at 187, 136 S. E., at 376 (1927).

Neither of these justifications makes sense today. Almost all crimes formerly punishable by death no longer are or can be. See, e. g., Enmund v. Florida, 458 U. S. 782 (1982); Coker v. Georgia, 433 U. S. 584 (1977). And while in earlier times “the gulf between the felonies and the minor offences was broad and deep,” 2 Pollock & Maitland 467, n. 3; Carroll v. United States, supra, at 158, today the distinction is minor and often arbitrary. Many crimes classified as misdemeanors, or nonexistent, at common law are now felonies. Wilgus, 22 Mich. L. Rev., at 572-573. These changes have undermined the concept, which was questionable to begin with, that use of deadly force against a fleeing felon is merely a speedier execution of someone who has already forfeited his life. They have also made the assumption that a “felon” is more dangerous than a misdemeanant untenable. Indeed, numerous misdemeanors involve conduct more dangerous than many felonies.12

There is an additional reason why the common-law rule cannot be directly translated to the present day. The common-law rule developed at a time when weapons were rudimentary. Deadly force could be inflicted almost solely in a hand-to-hand struggle during which, necessarily, the safety *15of the arresting officer was at risk. Handguns were not carried by police officers until the latter half of the last century. L. Kennett & J. Anderson, The Gun in America 150-151 (1975). Only then did it become possible to use deadly force from a distance as a means of apprehension. As a practical matter, the use of deadly force under the standard articulation of the common-law rule has an altogether different meaning — and harsher consequences — now than in past centuries. See Wechsler & Michael, A Rationale for the Law of Homicide: I, 37 Colum. L. Rev. 701, 741 (1937).13

One other aspect of the common-law rule bears emphasis. It forbids the use of deadly force to apprehend a misde-meanant, condemning such action as disproportionately severe. See Holloway v. Moser, 193 N. C., at 187, 136 S. E., at 376; State v. Smith, 127 Iowa, at 535, 103 N. W., at 945. See generally Annot., 83 A. L. R. 3d 238 (1978).

In short, though the common-law pedigree of Tennessee’s rule is pure on its face, changes in the legal and technological context mean the rule is distorted almost beyond recognition when literally applied.

C

In evaluating the reasonableness of police procedures under the Fourth Amendment, we have also looked to pre*16vailing rules in individual jurisdictions. See, e. g., United States v. Watson, 423 U. S., at 421-422. The rules in the States are varied. See generally Comment, 18 Ga. L. Rev. 137, 140-144 (1983). Some 19 States have codified the common-law rule,14 though in two of these the courts have significantly limited the statute.15 Four States, though without a relevant statute, apparently retain the common-law rule.16 Two States have adopted the Model Penal Code’s *17provision verbatim.17 Eighteen others allow, in slightly varying language, the use of deadly force only if the suspect has committed a felony involving the use or threat of physical or deadly force, or is escaping with a deadly weapon, or is likely to endanger life or inflict serious physical injury if not arrested.18 Louisiana and Vermont, though without statutes or case law on point, do forbid the use of deadly force to prevent any but violent felonies.19 The remaining States either have no relevant statute or case law, or have positions that are unclear.20

*18It cannot be said that there is a constant or overwhelming trend away from the common-law rule. In recent years, some States have reviewed their laws and expressly rejected abandonment of the common-law rule.21 Nonetheless, the long-term movement has been away from the rule that deadly force may be used against any fleeing felon, and that remains the rule in less than half the States.

This trend is more evident and impressive when viewed in light of the policies adopted by the police departments themselves. Overwhelmingly, these are more restrictive than the common-law rule. C. Milton, J. Halleck, J. Lardner, & G. Abrecht, Police Use of Deadly Force 45-46 (1977). The Federal Bureau of Investigation and the New York City Police Department, for example, both forbid the use of firearms except when necessary to prevent death or grievous bodily harm. Id., at 40-41; App. 88. For accreditation by the Commission on Accreditation for Law Enforcement Agencies, a department must restrict the use of deadly force to situations where “the officer reasonably believes that the action is in defense of human life ... or in defense of any person in immediate danger of serious physical injury.” Commission on Accreditation for Law Enforcement Agencies, Inc., Standards for Law Enforcement Agencies 1-2 (1983) (italics deleted). A 1974 study reported that the police department regulations in a majority of the large cities of the United States allowed the firing of a weapon only when a *19felon presented a threat of death or serious bodily harm. Boston Police Department, Planning & Research Division, The Use of Deadly Force by Boston Police Personnel (1974), cited in Mattis v. Schnarr, 547 F. 2d 1007, 1016, n. 19 (CA8 1976), vacated as moot sub nom. Ashcroft v. Mattis, 431 U. S. 171 (1977). Overall, only 7.5% of departmental and municipal policies explicitly permit the use of deadly force against any felon; 86.8% explicitly do not. K. Matulia, A Balance of Forces: A Report of the International Association of Chiefs of Police 161 (1982) (table). See also Record 1108-1368 (written policies of 44 departments). See generally W. Geller & K. Karales, Split-Second Decisions 33-42 (1981); Brief for Police Foundation et al. as Amici Curiae. In light of the rules adopted by those who must actually administer them, the older and fading common-law view is a dubious indicium of the constitutionality of the Tennessee statute now before us.

D

Actual departmental policies are important for an additional reason. We would hesitate to declare a police practice of long standing “unreasonable” if doing so would severely hamper effective law enforcement. But the indications are to the contrary. There has been no suggestion that crime has worsened in any way in jurisdictions that have adopted, by legislation or departmental policy, rules similar to that announced today. Amici note that “[a]fter extensive research and consideration, [they] have concluded that laws permitting police officers to use deadly force to apprehend unarmed, non-violent fleeing felony suspects actually do not protect citizens or law enforcement officers, do not deter crime or alleviate problems caused by crime, and do not improve the crime-fighting ability of law enforcement agencies.” Id., at 11. The submission is that the obvious state interests in apprehension are not sufficiently served to warrant the use of lethal weapons against all fleeing felons. See supra, at 10-11, and n. 10.

*20Nor do we agree with petitioners and appellant that the rule we have adopted requires the police to make impossible, split-second evaluations of unknowable facts. See Brief for Petitioners 25; Brief for Appellant 11. We do not deny the practical difficulties of attempting to assess the suspect’s dangerousness. However, similarly difficult judgments must be made by the police in equally uncertain circumstances. See, e. g., Terry v. Ohio, 392 U. S., at 20, 27. Nor is there any indication that in States that allow the use of deadly force only against dangerous suspects, see nn. 15, 17-19, supra, the standard has been difficult to apply or has led to a rash of litigation involving inappropriate second-guessing of police officers’ split-second decisions. Moreover, the highly technical felony/misdemeanor distinction is equally, if not more, difficult to apply in the field. An officer is in no position to know, for example, the precise value of property stolen, or whether the crime was a first or second offense. Finally, as noted above, this claim must be viewed with suspicion in light of the similar self-imposed limitations of so many police departments.

IV

The District Court concluded that Hymon was justified in shooting Garner because state law allows, and the Federal Constitution does not forbid, the use of deadly force to prevent the escape of a fleeing felony suspect if no alternative means of apprehension is available. See App. to Pet. for Cert. A9-A11, A38. This conclusion made a determination of Garner’s apparent dangerousness unnecessary. The court did find, however, that Garner appeared to be unarmed, though Hymon could not be certain that was the case. Id., at A4, A23. See also App. 41, 56; Record 219. Restated in Fourth Amendment terms, this means Hymon had no articu-lable basis to think Garner was armed.

In reversing, the Court of Appeals accepted the District Court’s factual conclusions and held that “the facts, as found, did not justify the use of deadly force.” 710 F. 2d, at 246. *21We agree. Officer Hymon could not reasonably have believed that Garner — young, slight, and unarmed — posed any threat. Indeed, Hymon never attempted to justify his actions on any basis other than the need to prevent an escape. The District Court stated in passing that “[t]he facts of this case did not indicate to Officer Hymon that Garner was ‘non-danger ous.’ ” App. to Pet. for Cert. A34. This conclusion is not explained, and seems to be based solely on the fact that Garner had broken into a house at night. However, the fact that Garner was a suspected burglar could not, without regard to the other circumstances, automatically justify the use of deadly force. Hymon did not have probable cause to believe that Garner, whom he correctly believed to be unarmed, posed any physical danger to himself or others.

The dissent argues that the shooting was justified by the fact that Officer Hymon had probable cause to believe that Garner had committed a nighttime burglary. Post, at 29, 32. While we agree that burglary is a serious crime, we cannot agree that it is so dangerous as automatically to justify the use of deadly force. The FBI classifies burglary as a “property” rather than a “violent” crime. See Federal Bureau of Investigation, Uniform Crime Reports, Crime in the United States 1 (1984).22 Although the armed burglar would present a different situation, the fact that an unarmed suspect has broken into a dwelling at night does not automatically mean he is physically dangerous. This case demonstrates as much. See also Solem v. Helm, 463 U. S. 277, 296-297, and nn. 22-23 (1983). In fact, the available statistics demonstrate that burglaries only rarely involve physical violence. During the 10-year period from 1973-1982, only 3.8% of all burglaries involved violent crime. Bureau of Justice Statistics, House*22hold Burglary 4 (1985).23 See also T. Reppetto, Residential Crime 17, 105 (1974); Conklin & Bittner, Burglary in a Suburb, 11 Criminology 208, 214 (1973).

V

We wish to make clear what our holding means in the context of this case. The complaint has been dismissed as to all the individual defendants. The State is a party only by virtue of 28 U. S. C. § 2403(b) and is not subject to liability. The possible liability of the remaining defendants — the Police Department and the city of Memphis — hinges on Monell v. New York City Dept. of Social Services, 436 U. S. 658 (1978), and is left for remand. We hold that the statute is invalid insofar as it purported to give Hymon the authority to act as he did. As for the policy of the Police Department, the absence of any discussion of this issue by the courts below, and the uncertain state of the record, preclude any consideration of its validity.

The judgment of the Court of Appeals is affirmed, and the case is remanded for further proceedings consistent with this opinion.

So ordered.

Justice O’Connor,

with whom The Chief Justice and Justice Rehnquist join, dissenting.

The Court today holds that the Fourth Amendment prohibits a police officer from using deadly force as a last resort to *23apprehend a criminal suspect who refuses to halt when fleeing the scene of a nighttime burglary. This conclusion rests on the majority’s balancing of the interests of the suspect and the public interest in effective law enforcement. Ante, at 8. Notwithstanding the venerable common-law rule authorizing the use of deadly force if necessary to apprehend a fleeing felon, and continued acceptance of this rule by nearly half the States, ante, at 14, 16-17, the majority concludes that Tennessee’s statute is unconstitutional inasmuch as it allows the use of such force to apprehend a burglary suspect who is not obviously armed or otherwise dangerous. Although the circumstances of this case are unquestionably tragic and unfortunate, our constitutional holdings must be sensitive both to the history of the Fourth Amendment and to the general implications of the Court’s reasoning. By disregarding the serious and dangerous nature of residential burglaries and the longstanding practice of many States, the Court effectively creates a Fourth Amendment right allowing a burglary suspect to flee unimpeded from a police officer who has probable cause to arrest, who has ordered the suspect to halt, and who has no means short of firing his weapon to prevent escape. I do not believe that the Fourth Amendment supports such a right, and I accordingly dissent.

I-H

The facts below warrant brief review because they highlight the difficult, split-second decisions police officers must make in these circumstances. Memphis Police Officers Elton Hymon and Leslie Wright responded to a late-night call that a burglary was in progress at a private residence. When the officers arrived at the scene, the callér said that “they” were breaking into the house next door. App. in No. 81-5605 (CA6), p. 207. The officers found the residence had been forcibly entered through a window and saw lights *24on inside the house. Officer Hymon testified that when he saw the broken window he realized “that something was wrong inside,” id., at 656, but that he could not determine whether anyone — either a burglar or a member of the household — was within the residence. Id., at 209. As Officer Hymon walked behind the house, he heard a door slam. He saw Edward Eugene Garner run away from the house through the dark and cluttered backyard. Garner crouched next to a 6-foot-high fence. Officer Hymon thought Garner was an adult and was unsure whether Garner was armed because Hymon “had no idea what was in the hand [that he could not see] or what he might have had on his person.” Id., at 658-659. In fact, Garner was 15 years old and unarmed. Hymon also did not know whether accomplices remained inside the house. Id., at 657. The officer identified himself as a police officer and ordered Garner to halt. Garner paused briefly and then sprang to the top of the fence. Believing that Garner would escape if he climbed over the fence, Hymon fired his revolver and mortally wounded the suspected burglar.

Appellee-respondent, the deceased’s father, filed a 42 U. S. C. § 1983 action in federal court against Hymon, the city of Memphis, and other defendants, for asserted violations of Garner’s constitutional rights. The District Court for the Western District of Tennessee held that Officer Hymon’s actions were justified by a Tennessee statute that authorizes a police officer to “use all the necessary means to effect the arrest,” if “after notice of the intention to arrest the defendant, he either flee or forcibly resist.” Tenn. Code Ann. §40-7-108 (1982). As construed by the Tennessee courts, this statute allows the use of deadly force only if a police officer has probable cause to believe that a person has committed a felony, the officer warns the person that he intends to arrest him, and the officer reasonably believes that no means less than such force will prevent the escape. See, e. g., Johnson v. State, 173 Tenn. 134, 114 S. W. 2d 819 *25(1938). The District Court held that the Tennessee statute is constitutional and that Hymon’s actions as authorized by that statute did not violate Garner’s constitutional rights. The Court of Appeals for the Sixth Circuit reversed on the grounds that the Tennessee statute “authorizing the killing of an unarmed, nonviolent fleeing felon by police in order to prevent escape” violates the Fourth Amendment and the Due Process Clause of the Fourteenth Amendment. 710 F. 2d 240, 244 (1983).

The Court affirms on the ground that application of the Tennessee statute to authorize Officer Hymon’s use of deadly force constituted an unreasonable seizure in violation of the Fourth Amendment. The precise issue before the Court deserves emphasis, because both the decision below and the majority obscure what must be decided in this case. The issue is not the constitutional validity of the Tennessee statute on its face or as applied to some hypothetical set of facts. Instead, the issue is whether the use of deadly force by Officer Hymon under the circumstances of this case violated Garner’s constitutional rights. Thus, the majority’s assertion that a police officer who has probable cause to seize a suspect “may not always do so by killing him,” ante, at 9, is unexceptionable but also of little relevance to the question presented here. The same is true of the rhetorically stirring statement that “[t]he use of deadly force to prevent the escape of all felony suspects, whatever the circumstances, is constitutionally unreasonable.” Ante, at 11. The question we must address is whether the Constitution allows the use of such force to apprehend a suspect who resists arrest by attempting to flee the scene of a nighttime burglary of a residence.

II

For purposes of Fourth Amendment analysis, I agree with the Court that Officer Hymon “seized” Garner by shooting him. Whether that seizure was reasonable and therefore permitted by the Fourth Amendment requires a careful bal*26ancing of the important public interest in crime prevention and detection and the nature and quality of the intrusion upon legitimate interests of the individual. United States v. Place, 462 U. S. 696, 703 (1983). In striking this balance here, it is crucial to acknowledge that police use of deadly force to apprehend a fleeing criminal suspect falls within the “rubric of police conduct . . . necessarily [involving] swift action predicated upon the on-the-spot observations of the officer on the beat.” Terry v. Ohio, 392 U. S. 1, 20 (1968). The clarity of hindsight cannot provide the standard for judging the reasonableness of police decisions made in uncertain and often dangerous circumstances. Moreover, I am far more reluctant than is the Court to conclude that the Fourth Amendment proscribes a police practice that was accepted at the time of the adoption of the Bill of Rights and has continued to receive the support of many state legislatures. Although the Court has recognized that the requirements of the Fourth Amendment must respond to the reality of social and technological change, fidelity to the notion of constitutional— as opposed to purely judicial — limits on governmental action requires us to impose a heavy burden on those who claim that practices accepted when the Fourth Amendment was adopted are now constitutionally impermissible. See, e. g., United States v. Watson, 423 U. S. 411, 416-421 (1976); Carroll v. United States, 267 U. S. 132, 149-153 (1925). Cf. United States v. Villamonte-Marquez, 462 U. S. 579, 585 (1983) (noting “impressive historical pedigree” of statute challenged under Fourth Amendment).

The public interest involved in the use of deadly force as a last resort to apprehend a fleeing burglary suspect relates primarily to the serious nature of the crime. Household burglaries not only represent the illegal entry into a person’s home, but also “pos[e] real risk of serious harm to others.” Solem v. Helm, 463 U. S. 277, 315-316 (1983) (Burger, C. J., dissenting). According to recent Department of Justice statistics, “[t]hree-fifths of all rapes in the home, *27three-fifths of all home robberies, and about a third of home aggravated and simple assaults are committed by burglars.” Bureau of Justice Statistics Bulletin, Household Burglary 1 (January 1985). During the period 1973-1982, 2.8 million such violent crimes were committed in the course of burglaries. Ibid. Victims of a forcible intrusion into their home by a nighttime prowler will find little consolation in the majority’s confident assertion that “burglaries only rarely involve physical violence.” Ante, at 21. Moreover, even if a particular burglary, when viewed in retrospect, does not involve physical harm to others, the “harsh potentialities for violence” inherent in the forced entry into a home preclude characterization of the crime as “innocuous, inconsequential, minor, or ‘nonviolent.’” Solem v. Helm, supra, at 316 (Burger, C. J., dissenting). See also Restatement of Torts §131, Comment g (1934) (burglary is among felonies that normally cause or threaten death or serious bodily harm); R. Perkins & R. Boyce, Criminal Law 1110 (3d ed. 1982) (burglary is dangerous felony that creates unreasonable risk of great personal harm).

Because burglary is a serious and dangerous felony, the public interest in the prevention and detection of the crime is of compelling importance. Where a police officer has probable cause to arrest a suspected burglar, the use of deadly force as a last resort might well be the only means of apprehending the suspect. With respect to a particular burglary, subsequent investigation simply cannot represent a substitute for immediate apprehension of the criminal suspect at the scene. See President’s Commission on Law Enforcement and Administration of Justice, Task Force Report: The Challenge of Crime in a Free Society 97 (1967). Indeed, the Captain of the Memphis Police Department testified that in his city, if apprehension is not immediate, it is likely that the suspect will not be caught. App. in No. 81-5605 (CA6), p. 334. Although some law enforcement agencies may choose to assume the risk that a criminal will remain at large, the *28Tennessee statute reflects a legislative determination that the use of deadly force in prescribed circumstances will serve generally to protect the public. Such statutes assist the police in apprehending suspected perpetrators of serious crimes and provide notice that a lawful police order to stop and submit to arrest may not be ignored with impunity. See, e. g., Wiley v. Memphis Police Department, 548 F. 2d 1247, 1252-1253 (CA6), cert. denied, 434 U. S. 822 (1977); Jones v. Marshall, 528 F. 2d 132, 142 (CA2 1975).

The Court unconvincingly dismisses the general deterrence effects by stating that “the presently available evidence does not support [the] thesis” that the threat of force discourages escape and that “there is a substantial basis for doubting that the use of such force is an essential attribute to the arrest power in all felony cases.” Ante, at 10, 11. There is no question that the effectiveness of police use of deadly force is arguable and that many States or individual police departments have decided not to authorize it in circumstances similar to those presented here. But it should go without saying that the effectiveness or popularity of a particular police practice does not determine its constitutionality. Cf. Spaziano v. Florida, 468 U. S. 447, 464 (1984) (“The Eighth Amendment is not violated every time a State reaches a conclusion different from a majority of its sisters over how best to administer its criminal laws”). Moreover, the fact that police conduct pursuant to a state statute is challenged on constitutional grounds does not impose a burden on the State to produce social science statistics or to dispel any possible doubts about the necessity of the conduct. This observation, I believe, has particular force where the challenged practice both predates enactment of the Bill of Rights and continues to be accepted by a substantial number of the States.

Against the strong public interests justifying the conduct at issue here must be weighed the individual interests implicated in the use of deadly force by police officers. The *29majority declares that “[t]he suspect’s fundamental interest in his own life need not be elaborated upon.” Ante, at 9. This blithe assertion hardly provides an adequate substitute for the majority’s failure to acknowledge the distinctive manner in which the suspect’s interest in his life is even exposed to risk. For purposes of this case, we must recall that the police officer, in the course of investigating a nighttime burglary, had reasonable cause to arrest the suspect and ordered him to halt. The officer’s use of force resulted because the suspected burglar refused to heed this command and the officer reasonably believed that there was no means short of firing his weapon to apprehend the suspect. Without questioning the importance of a person’s interest in his life, I do not think this interest encompasses a right to flee unimpeded from the scene of a burglary. Cf. Payton v. New York, 445 U. S. 573, 617, n. 14 (1980) (White, J., dissenting) (“[T]he policeman’s hands should not be tied merely because of the possibility that the suspect will fail to cooperate with legitimate actions by law enforcement personnel”). The legitimate interests of the suspect in these circumstances are adequately accommodated by the Tennessee statute: to avoid the use of deadly force and the consequent risk to his life, the suspect need merely obey the valid order to halt.

A proper balancing of the interests involved suggests that use of deadly force as a last resort to apprehend a criminal suspect fleeing from the scene of a nighttime burglary is not unreasonable within the meaning of the Fourth Amendment. Admittedly, the events giving rise to this case are in retrospect deeply regrettable. No one can view the death of an unarmed and apparently nonviolent 15-year-old without sorrow, much less disapproval. Nonetheless, the reasonableness of Officer Hymon’s conduct for purposes of the Fourth Amendment cannot be evaluated by what later appears to have been a preferable course of police action. The officer pursued a suspect in the darkened backyard of a house that from all indications had just been burglarized. The *30police officer was not certain whether the suspect was alone or unarmed; nor did he know what had transpired inside the house. He ordered the suspect to halt, and when the suspect refused to obey and attempted to flee into the night, the officer fired his weapon to prevent escape. The reasonableness of this action for purposes of the Fourth Amendment is not determined by the unfortunate nature of this particular case; instead, the question is whether it is constitutionally impermissible for police officers, as a last resort, to shoot a burglary suspect fleeing the scene of the crime.

Because I reject the Fourth Amendment reasoning of the majority and the Court of Appeals, I briefly note that no other constitutional provision supports the decision below. In addition to his Fourth Amendment claim, appellee-respondent also alleged violations of due process, the Sixth Amendment right to trial by jury, and the Eighth Amendment proscription of cruel and unusual punishment. These arguments were rejected by the District Court and, except for the due process claim, not addressed by the Court of Appeals. With respect to due process, the Court of Appeals reasoned that statutes affecting the fundamental interest in life must be “narrowly drawn to express only the legitimate state interests at stake.” 710 F. 2d, at 245. The Court of Appeals concluded that a statute allowing police use of deadly force is narrowly drawn and therefore constitutional only if the use of such force is limited to situations in which the suspect poses an immediate threat to others. Id., at 246-247. Whatever the validity of Tennessee’s statute in other contexts, I cannot agree that its application in this case resulted in a deprivation “without due process of law.” Cf. Baker v. McCollan, 443 U. S. 137, 144-145 (1979). Nor do I believe that a criminal suspect who is shot while trying to avoid apprehension has a cognizable claim of a deprivation of his Sixth Amendment right to trial by jury. See Cunningham v. Ellington, 323 F. Supp. 1072, 1075-1076 (WD Tenn. 1971) (three-judge court). Finally, because there is no indication that the use *31of deadly force was intended to punish rather than to capture the suspect, there is no valid claim under the Eighth Amendment. See Bell v. Wolfish, 441 U. S. 520, 538-539 (1979). Accordingly, I conclude that the District Court properly entered judgment against appellee-respondent, and I would reverse the decision of the Court of Appeals.

III

Even if I agreed that the Fourth Amendment was violated under the circumstances of this case, I would be unable to join the Court’s opinion. The Court holds that deadly force may be used only if the suspect “threatens the officer with a weapon or there is probable cause to believe that he has committed a crime involving the infliction or threatened infliction of serious physical harm.” Ante, at 11. The Court ignores the more general implications of its reasoning. Relying on the Fourth Amendment, the majority asserts that it is constitutionally unreasonable to use deadly force against fleeing criminal suspects who do not appear to pose a threat of serious physical harm to others. Ibid. By declining to limit its holding to the use of firearms, the Court unnecessarily implies that the Fourth Amendment constrains the use of any police practice that is potentially lethal, no matter how remote the risk. Cf. Los Angeles v. Lyons, 461 U. S. 95 (1983).

Although it is unclear from the language of the opinion, I assume that the majority intends the word “use” to include only those circumstances in which the suspect is actually apprehended. Absent apprehension of the suspect, there is no “seizure” for Fourth Amendment purposes. I doubt that the Court intends to allow criminal suspects who successfully escape to return later with §1983 claims against officers who used, albeit unsuccessfully, deadly force in their futile attempt to capture the fleeing suspect. The Court’s opinion, despite its broad language, actually decides only that the *32shooting of a fleeing burglary suspect who was in fact neither armed nor dangerous can support a § 1983 action.

The Court’s silence on critical factors in the decision to use deadly force simply invites second-guessing of difficult police decisions that must be made quickly in the most trying of circumstances. Cf. Payton v. New York, 445 U. S., at 619 (White, J., dissenting). Police are given no guidance for determining which objects, among an array of potentially lethal weapons ranging from guns to knives to baseball bats to rope, will justify the use of deadly force. The Court also declines to outline the additional factors necessary to provide “probable cause” for believing that a suspect “poses a significant threat of death or serious physical injury,” ante, at 3, when the officer has probable cause to arrest and the suspect refuses to obey an order to halt. But even if it were appropriate in this case to limit the use of deadly force to that ambiguous class of suspects, I believe the class should include nighttime residential burglars who resist arrest by attempting to flee the scene of the crime. We can expect an escalating volume of litigation as the lower courts struggle to determine if a police officer’s split-second decision to shoot was justified by the danger posed by a particular object and other facts related to the crime. Thus, the majority opinion portends a burgeoning area of Fourth Amendment doctrine concerning the circumstances in which police officers can reasonably employ deadly force.

IV

The Court’s opinion sweeps broadly to adopt an entirely new standard for the constitutionality of the use of deadly force to apprehend fleeing felons. Thus, the Court “lightly brushe[s] aside,” Payton v. New York, supra, at 600, a longstanding police practice that predates the Fourth Amendment and continues to receive the approval of nearly half of the state legislatures. I cannot accept the majority’s creation of a constitutional right to flight for burglary sus*33pects seeking to avoid capture at the scene of the crime. Whatever the constitutional limits on police use of deadly force in order to apprehend a fleeing felon, I do not believe they are exceeded in a case in which a police officer has probable cause to arrest a suspect at the scene of a residential burglary, orders the suspect to halt, and then fires his weapon as a last resort to prevent the suspect’s escape into the night. I respectfully dissent.

Justice Stevens

delivered the opinion of the Court.

In January 2003, petitioner Keshia Dixon purchased multiple firearms at two gun shows, during the course of which she provided an incorrect address and falsely stated that she was not under indictment for a felony. As a result of these illegal acts, petitioner was indicted and convicted on one count of receiving a firearm while under indictment in violation of 18 U. S. C. § 922(n) and eight counts of making false statements in connection with the acquisition of a firearm in violation of § 922(a)(6). At trial, petitioner admitted that *4she knew she was under indictment when she made the purchases and that she knew doing so was a crime; her defense was that she acted under duress because her boyfriend threatened to kill her or hurt her daughters if she did not buy the guns for him.

Petitioner contends that the trial judge’s instructions to the jury erroneously required her to prove duress by a preponderance of the evidence instead of requiring the Government to prove beyond a reasonable doubt that she did not act under duress. The Court of Appeals rejected petitioner’s contention, 413 F. 3d 520 (CA5 2005); given contrary treatment of the issue by other federal courts,1 we granted certiorari, 546 U. S. 1135 (2006).

I

At trial, in her request for jury instructions on her defense of duress, petitioner contended that she “should have the burden of production, and then that the Government should be required to disprove beyond a reasonable doubt the duress.” App. 300. Petitioner admitted that this request was contrary to Fifth Circuit precedent, and the trial court, correctly finding itself bound by Circuit precedent, denied petitioner’s request. Ibid. Instead, the judge’s instructions to the jury defined the elements of the duress defense2 and *5stated that petitioner has “the burden of proof to establish the defense of duress by a preponderance of the evidence.” Id., at 312.

Petitioner argues here, as she did in the District Court and the Court of Appeals, that federal law requires the Government to bear the burden of disproving her defense beyond a reasonable doubt and that the trial court’s erroneous instruction on this point entitles her to a new trial. There are two aspects to petitioner’s argument in support of her proposed instruction that merit separate discussion. First, petitioner contends that her defense “controverted the mens rea required for conviction” and therefore that the Due Process Clause requires the Government to retain the burden of persuasion on that element. Brief for Petitioner 41. Second, petitioner argues that the Fifth Circuit’s rule is “contrary to modern common law.” Id., at 14.

II

The crimes for which petitioner was convicted require that she have acted “knowingly,” § 922(a)(6), or “willfully,” § 924(a)(1)(D).3 As we have explained, “unless the text of the statute dictates a different result, the term ‘knowingly’ merely requires proof of knowledge of the facts that constitute the offense.” Bryan v. United States, 524 U. S. 184, 193 (1998) (footnote omitted). And the term “willfully” in § 924(a)(1)(D) requires a defendant to have “acted with knowledge that his conduct was unlawful.” Ibid. In this case, then, the Government bore the burden of proving beyond a reasonable doubt that petitioner knew she was mak*6ing false statements in connection with the acquisition of firearms and that she knew she was breaking the law when she acquired a firearm while under indictment. See In re Winship, 397 U. S. 358, 364 (1970). Although the Government may have proved these elements in other ways, it clearly met its burden when petitioner testified that she knowingly committed certain acts—she put a false address on the forms she completed to purchase the firearms, falsely claimed that she was the actual buyer of the firearms, and falsely stated that she was not under indictment at the time of the purchase—and when she testified that she knew she was breaking the law when, as an individual under indictment at the time, she purchased a firearm. App. 221-222.

Petitioner contends, however, that she cannot have formed the necessary mens rea for these crimes because she did not freely choose to commit the acts in question. But even if we assume that petitioner’s will was overborne by the threats made against her and her daughters, she still knew that she was making false statements and knew that she was breaking the law by buying a firearm. The duress defense, like the defense of necessity that we considered in United States v. Bailey, 444 U. S. 394, 409-410 (1980), may excuse conduct that would otherwise be punishable, but the existence of duress normally does not controvert any of the elements of the offense itself.4 As we explained in Bailey, “[c]riminal liability is normally based upon the concurrence of two factors, 'an evil-meaning mind [and] and evil-doing hand____’ ” Id., at 402 (quoting Morissette v. United States, *7342 U. S. 246, 251 (1952)). Like the defense of necessity, the defense of duress does not negate a defendant’s criminal state of mind when the applicable offense requires a defendant to have acted knowingly or willfully; instead, it allows the defendant to “avoid liability . . . because coercive conditions or necessity negates a conclusion of guilt even though the necessary mens rea was present.” Bailey, 444 U. S., at 402.5

The fact that petitioner’s crimes are statutory offenses that have no counterpart in the common law also supports our conclusion that her duress defense in no way disproves an element of those crimes. We have observed that “[t]he definition of the elements of a criminal offense is entrusted to the legislature, particularly in the case of federal crimes, which are solely creatures of statute.” Liparota v. United States, 471 U. S. 419, 424 (1985). Here, consistent with the movement away from the traditional dichotomy of general versus specific intent and toward a more specifically defined hierarchy of culpable mental states, see Bailey, 444 U. S., at 403-404, Congress defined the crimes at issue to punish defendants who act “knowingly,” § 922(a)(6), or “willfully,” § 924(a)(1)(D). It is these specific mental states, rather than some vague “evil mind,” Brief for Petitioner 42, or “ ‘criminal’ intent,” Martin v. Ohio, 480 U. S. 228, 235 (1987), that the Government is required to prove beyond a reasonable doubt, see Patterson v. New York, 432 U. S. 197, 211, n. 12 (1977) (“The applicability of the reasonable-doubt standard, *8however, has always been dependent on how a State defines the offense that is charged in any given case”). The jury instructions in this case were consistent with this requirement and, as such, did not run afoul of the Due Process Clause when they placed the burden on petitioner to establish the existence of duress by a preponderance of the evidence.

Ill

Having found no constitutional basis for placing upon the Government the burden of disproving petitioner’s duress de-. fense beyond a reasonable doubt, we next address petitioner’s argument that the modern common law requires the Government to bear that burden. In making this argument, petitioner recognizes that, until the end of the 19th century, common-law courts generally adhered to the rule that “the proponent of an issue bears the burden of persuasion on the factual premises for applying the rule.” Fletcher, Two Kinds of Legal Rules: A Comparative Study of Burden-of-Persuasion Practices in Criminal Cases, 77 Yale L. J. 880, 898 (1967-1968). In petitioner’s view, however, two important developments have established a contrary common-law rule that now prevails in federal courts: this Court’s decision in Davis v. United States, 160 U. S. 469 (1895), which placed the burden on the Government to prove a defendant’s sanity, and the publication of the Model Penal Code in 1962.

Although undisputed in this case, it bears repeating that, at common law, the burden of proving “affirmative defenses—indeed, ‘all... circumstances of justification, excuse or alleviation’—rested on the defendant.” Patterson, 432 U. S., at 202 (quoting 4 W. Blackstone, Commentaries *201); see also Martin v. Ohio, 480 U. S., at 235; Mullaney v. Wilbur, 421 U. S. 684, 693 (1975). This common-law rule accords with the general evidentiary rule that “the burdens of producing evidence and of persuasion with regard to any given issue are both generally allocated to the same party.” 2 J. Strong, McCormick on Evidence §337, p. 415 (5th ed. *91999). And, in the context of the defense of duress, it accords with the doctrine that “where the facts with regard to an issue lie peculiarly in the knowledge of a party, that party has the burden of proving the issue.” Id., at 418. Although she claims that the common-law rule placing the burden on a defendant to prove the existence of duress “was the product of flawed reasoning,” petitioner accepts that this was the general rule, at least until this Court’s decision in Davis. Brief for Petitioner 18. According to petitioner, however, Davis initiated a revolution that overthrew the old common-law rule and established her proposed rule in its place.

Davis itself, however, does not support petitioner’s position. In that case, we reviewed a defendant’s conviction for having committed murder “feloniously, wilfully, and of his malice aforethought.” 160 U. S., at 474. It was undisputed that the prosecution’s evidence, “if alone considered, made it the duty of the jury to return a verdict of guilty of the crime charged”; the defendant, however, adduced evidence at trial tending to show that he did not have the mental capacity to form the requisite intent. Id., at 475. At issue before the Court was the correctness of the trial judge’s instruction to the jury that the law “ ‘presumes every man is sane, and the burden of showing it is not true is upon the party who asserts it.’” Id., at 476. Under this instruction, “if the evidence was in equilibrio as to the accused being sane, that is, capable of comprehending the nature and effect of his acts, he was to be treated just as he would be if there were no defence of insanity or if there were an entire absence of proof that he was insane.” Id., at 479.

In reversing the defendant’s conviction, we found ourselves “unable to assent to the doctrine that in a prosecution for murder... it is the duty of the jury to convict where the evidence is equally balanced on the issue as to the sanity of the accused at the time of the killing.” Id., at 484 (emphasis added). Instead, we concluded that this defendant was “entitled to an acquittal of the specific crime charged if upon *10all the evidence there is reasonable doubt whether he was capable in law of committing [the] crime.” Ibid, (emphasis added). Our opinion focused on the “definition of murder,” explaining that “it is of the very essence of that heinous crime that it be committed by a person of ‘sound memory and discretion,’ and with ‘malice aforethought.’” Ibid. Reviewing “the adjudged cases” and “elementary treatises upon criminal law,” we found that “[a]ll admit that the crime of murder necessarily involves the possession by the accused of such mental capacity as will render him criminally responsible for his acts.” Id., at 485. Thus, when we ultimately found that the burden of proving the accused’s sanity rested on the Government, our holding rested on the conclusion that

“[Davis’] guilt cannot be said to have been proved beyond a reasonable doubt—his will and his acts cannot be held to have joined in perpetrating the murder charged—if the jury, upon all the evidence, have a reasonable doubt whether he was legally capable of committing crime, or (which is the same thing) whether he wilfully, deliberately, unlawfully, and of malice aforethought took the life of the deceased. As the crime of murder involves sufficient capacity to distinguish between right and wrong, the legal interpretation of every verdict of guilty as charged is that the jury believed from all the evidence beyond a reasonable doubt that the accused was guilty, and was therefore responsible, criminally, for his acts. How then upon principle or consistently with humanity can a verdict of guilty be properly returned, if the jury entertain a reasonable doubt as to the existence of a fact which is essential to guilt, namely, the capacity in law of the accused to commit that crime?” Id., at 488.

Our opinion in Davis, then, interpreted a defendant’s sanity to controvert the necessary mens rea for the crime of *11murder committed “feloniously, wilfully, and of his malice aforethought,” id., at 474, as “[o]ne who takes human life cannot be said to be actuated by malice aforethought, or to have deliberately intended to take life, or to have ‘a wicked, depraved, and malignant heart/ . . . unless at the time he had sufficient mind to comprehend the criminality or the right and wrong of such an act,” id., at 485. We required the Government to prove the defendant’s sanity beyond a reasonable doubt because the evidence that tended to prove insanity also tended to disprove an essential element of the offense charged. See Davis v. United States, 165 U. S. 373, 378 (1897) (“[T]he fact of sanity, as any other essential fact in the case, must be established to the satisfaction of the jury beyond a reasonable doubt” (emphasis added)). Whether or not this reasoning correctly treated insanity as negating the mens rea for murder as defined in the statute at issue, cf. n. 4, supra, it does not help petitioner: The evidence of duress she adduced at trial does not contradict or tend to disprove any element of the statutory offenses that she committed.

Nor does the proposition for which Davis has come to stand help petitioner’s cause. Although written more narrowly in the context of a prosecution for the crime of murder, Davis was later interpreted to establish a general “rule for federal prosecutions . . . that an accused is ‘entitled to an acquittal of the specific crime charged if upon all the evidence there is reasonable doubt whether he was capable in law of committing crime.’ ” Leland v. Oregon, 343 U. S. 790, 797 (1952) (quoting Davis, 160 U. S., at 484); see also Lynch v. Overholser, 369 U. S. 705, 713 (1962) (explaining that the Davis rule applied in all federal courts). After Davis, if a federal defendant introduced sufficient evidence to raise a reasonable doubt as to his sanity, it was sufficient to create a question for the jury on which the Government bore the ultimate burden of persuasion beyond a reasonable doubt. *12See, e. g., Hall v. United States, 295 F. 2d 26, 28 (CA4 1961); Holloway v. United States, 148 F. 2d 665, 666 (CADC 1945); Post v. United States, 135 F. 1, 10 (CA5 1905).

In apparent recognition of the fact that Davis relied on the heightened mens rea applicable to the particular statute at issue, we held in Leland that this rule was not constitutionally mandated, 343 U. S., at 797, and Congress overruled it by statute in 1984, requiring a defendant to prove his insanity by clear and convincing evidence, 98 Stat. 2057, codified at 18 U. S. C. § 17(b). Moreover, Congress has treated the defense of insanity differently from that of duress not only by codifying it but by requiring defendants who intend to rely on an insanity defense to provide advance notice to the Government. See Fed. Rule Crim. Proc. 12.2(a). Thus, even if the rule arising from Davis may have once been relevant to an evaluation of other affirmative defenses, Congress’ differential treatment of the insanity defense and its rejection of the Davis rule are inconsistent with petitioner’s invitation to follow Davis’ lead in this case.

Indeed, petitioner’s reliance on Davis ignores the fact that federal crimes “are solely creatures of statute,” Liparota, 471 U. S., at 424, and therefore that we are required to effectuate the duress defense as Congress “may have contemplated” it in the context of these specific offenses, United States v. Oakland Cannabis Buyers’ Cooperative, 532 U. S. 483, 491, n. 3 (2001) (internal quotation marks omitted); see also id., at 499 (Stevens, J., concurring in judgment) (explaining that Court was addressing whether the statute at issue foreclosed a necessity defense to specific charges brought under the statute); Bailey, 444 U. S., at 410 (“We need not speculate now, however, on the precise contours of whatever defenses of duress or necessity are available against charges brought under [18 U. S. C.J § 751(a)”). The offenses at issue in this case were created by statute in 1968, when Congress enacted the Omnibus Crime Control and Safe Streets Act (hereinafter Safe Streets Act or Act). See 82 *13Stat. 197. There is no evidence in the Act’s structure or history that Congress actually considered the question of how the duress defense should work in this context, and there is no suggestion that the offenses at issue are incompatible with a defense of duress.6 Cf. Oakland Cannabis Buyers’ Cooperative, 532 U. S., at 491. Assuming that a defense of duress is available to the statutory crimes at issue,7 then, we must determine what that defense would look like as Congress “may have contemplated” it.

As discussed above, the common law long required the defendant to bear the burden of proving the existence of duress. Similarly, even where Congress has enacted an affirmative defense in the proviso of a statute, the “settled rule in this jurisdiction [is] that an indictment or other pleading . . . need not negative the matter of an exception made by a proviso or other distinct clause . . . and that it is incumbent on one who relies on such an exception to set it up and establish it.” McKelvey v. United States, 260 U. S. 353, 357 (1922); see also United States v. Dickson, 15 Pet. 141, 165 (1841) (calling this “the general rule of law which has always prevailed, and become consecrated almost as a maxim in the interpretation of statutes”). Even though the Safe Streets Act does not mention the defense of duress, we can safely assume that the 1968 Congress was familiar with *14both the long-established common-law rule8 and the rule applied in McKelvey and that it would have expected federal courts to apply a similar approach to any affirmative defense that might be asserted as a justification or excuse for violating the new law.9

This conclusion is surely more reasonable than petitioner’s hypothesis that Davis dramatically upset a well-settled rule of law. Petitioner cites only one federal case decided before 1968 for the proposition that it has been well established in federal law that the Government bears the burden of disproving duress beyond a reasonable doubt. But that case involved a defendant’s claim that he “lacked the specific intent to defraud required by the statute for the reason that he committed the offense under duress and coercion.” Johnson v. United States, 291 F. 2d 150, 152 (CA8 1961). Thus, when the Court of Appeals explained that “there is no burden upon the defendant to prove his defense of coercion,” id., at 155, that statement is best understood in context as a corollary to the by-then-unremarkable proposition that “the burden of proof rests upon the Government to prove the defendant’s guilt beyond a reasonable doubt,” ibid. Properly understood, Johnson provides petitioner little help in her uphill struggle to prove that a dramatic shift in the federal common-law rule occurred between Davis and the enactment of the Safe Streets Act in 1968.

Indeed, for us to be able to accept petitioner’s proposition, we would need to find an overwhelming consensus among *15federal courts that it is the Government’s burden to disprove the existence of duress beyond a reasonable doubt. The existence today of disagreement among the Federal Courts of Appeals on this issue, however—the very disagreement that caused us to grant certiorari in this case, see n. 1, supra— demonstrates that no such consensus has ever existed. See also post, at 25-27 (Breyer, J., dissenting) (discussing differences in treatment of the duress defense by the various Courts of Appeals). Also undermining petitioner’s argument is the fact that, in 1970, the National Commission on Reform of Federal Criminal Laws proposed that a defendant prove the existence of duress by a preponderance of the evidence. See 1 Working Papers 278. Moreover, while there seem to be few, if any, post-Davis, pre-1968 cases placing the burden on a defendant to prove the existence of duress,10 or even discussing the issue in any way, this lack of evidence does not help petitioner. The long-established common-law rule is that the burden of proving duress rests on the defendant. Petitioner hypothesizes that Davis fomented a revolution upsetting this rule. If this were true, one would expect to find cases discussing the matter. But no such cases exist.

It is for a similar reason that we give no weight to the publication of the Model Penal Code in 1962. As petitioner notes, the Code would place the burden on the government to disprove the existence of duress beyond a reasonable doubt. See ALI, Model Penal Code § 1.12, p. 88 (2001) (hereinafter Model Penal Code or Code) (stating that each element *16of an offense must be proved beyond a reasonable doubt); § 1.13(9)(c), at 91 (defining as an element anything that negatives an excuse for the conduct at issue); §2.09, at 131-132 (establishing affirmative defense of duress). Petitioner argues that the Code reflects “well established” federal law as it existed at the time. Brief for Petitioner 25. But, as discussed above, no such consensus existed when Congress passed the Safe Streets Act in 1968. And even if we assume Congress’ familiarity with the Code and the rule it would establish, there is no evidence that Congress endorsed the Code’s views or incorporated them into the Safe Streets Act.

In fact, the Act itself provides evidence to the contrary. Despite the Code’s careful delineation of mental states, see Model Penal Code §2.02, at 94-95, the Safe Streets Act attached no explicit mens rea requirement to the crime of receiving a firearm while under indictment, § 924(a), 82 Stat. 233 (“Whoever violates any provision of this chapter ... shall be fined not more than $5,000 or imprisoned not more than five years, or both”). And when Congress amended the Act to impose a mens rea requirement, it punished people who “willfully” violate the statute, see § 104(a), 100 Stat. 456, a mental state that has not been embraced by the Code, see Model Penal Code '§ 2.02(2), at 94-95 (defining “purposely,” “knowingly,” “recklessly,” and “negligently”); id., Explanatory Note, at 97 (“Though the term ‘wilfully’ is not used in the definitions of crimes contained in the Code, its currency and its existence in offenses outside the criminal code suggest the desirability of clarification”). Had Congress intended to adopt the Code’s structure when it enacted or amended the Safe Streets Act, one would expect the Act’s form and language to adhere much more closely to that used by the Code. It does not, and, for that reason, we cannot rely on the Model Penal Code to provide evidence as to how Congress would have wanted us to effectuate the duress defense in this context.

*17IV

Congress can, if it chooses, enact a duress defense that places the burden on the Government to disprove duress beyond a reasonable doubt. In light of Congress’ silence on the issue, however, it is up to the federal courts to effectuate the affirmative defense of duress as Congress “may have contemplated” it in an offense-specific context. Oakland Cannabis Buyers’ Cooperative, 532 U. S., at 491, n. 3 (internal quotation marks omitted). In the context of the firearms offenses at issue—as will usually be the case, given the long-established common-law rule—we presume that Congress intended the petitioner to bear the burden of proving the defense of duress by a preponderance of the evidence. Accordingly, the judgment of the Court of Appeals is affirmed.

It is so ordered.

Justice Kennedy,

concurring.

No one disputes that, subject to constitutional constraints, Congress has the authority to determine the content of a duress defense with respect to federal crimes and to direct whether the burden of proof rests with the defense or the prosecution. The question here is how to proceed when Congress has enacted a criminal statute, the Omnibus Crime Control and Safe Streets Act of 1968, 82 Stat. 197 et seq. (hereinafter Safe Streets Act), without explicit instructions regarding the duress defense or its burden of proof. See ante, at 12-13.

When issues of congressional intent with respect to the nature, extent, and definition of federal crimes arise, we assume Congress acted against certain background understandings set forth in judicial decisions in the Anglo-American legal tradition. See United States v. Bailey, 444 U. S. 394, 415, n. 11 (1980). Those decisions, in turn, consult sources such as legal treatises and the American Legal Insti*18tute’s Model Penal Code. See, e. g., United States v. Jimenez Recio, 537 U. S. 270, 275-276 (2003); Salinas v. United States, 522 U. S. 52, 64-65 (1997). All of these sources rely upon the insight gained over time as the legal process continues. Absent some contrary indication in the statute, we can assume that Congress would not want to foreclose the courts from consulting these newer sources and considering innovative arguments in resolving issues not confronted in the statute and not within the likely purview of Congress when it enacted the criminal prohibition applicable in the particular case.

While the Court looks to the state of the law at the time the statute was enacted, see ante, at 14, the better reading of the Court’s opinion is that isolated authorities or writings do not control unless they were indicative of guiding principles upon which Congress likely would have relied. Otherwise, it seems altogether a fiction to attribute to Congress any intent one way or the other in assigning the burden of proof. It seems unlikely, moreover, that Congress would have wanted the burden of proof for duress to vary from statute to statute depending upon the date of enactment. Consistent with these propositions, the Court looks not only to our precedents and common-law traditions, but also to the treatment of the insanity defense in a 1984 statute and a proposal of the National Commission on Reform of Federal Criminal Laws, even though they both postdated the passage of the Safe Streets Act. See ante, at 12, 15.

As there is no reason to suppose that Congress wanted to depart from the traditional principles for allocating the burden of proof, the proper approach is simply to apply these principles to the context of duress. See, e. g., Schaffer v. Weast, 546 U. S. 49, 56 (2005) (where the plain text of the statute is “silent on the allocation of the burden of persuasion,” we proceed to consider the “ordinary default rule” and its exceptions). The facts needed to prove or disprove the defense “lie peculiarly in the knowledge of” the defendant. *192 K. Broun, McCormick on Evidence §337, p. 475 (6th ed. 2006); see ante, at 8-9. The claim of duress in most instances depends upon conduct that takes place before the criminal act; and, as the person who allegedly coerced the defendant is often unwilling to come forward and testify, the prosecution may be without any practical means of disproving the defendant’s allegations. There is good reason, then, to maintain the usual rule of placing the burden of production and persuasion together on the party raising the issue. See 2 Broun, supra, §337; ante, at 8. The analysis may come to a different result, of course, for other defenses.

With these observations, I join the Court’s opinion.

Justice Alito, with whom Justice Scalia joins, concurring.

I join the opinion of the Court with the understanding that it does not hold that the allocation of the burden of persuasion on the defense of duress may vary from one federal criminal statute to another.

Duress was an established defense at common law. See 4 W. Blackstone, Commentaries on the Laws of England 30 (1769). When Congress began to enact federal criminal statutes, it presumptively intended for those offenses to be subject to this defense. Moreover, Congress presumptively intended for the burdens of production and persuasion to be placed, as they were at common law, on the defendant. Although Congress is certainly free to alter this pattern and place one or both burdens on the prosecution, either for all or selected federal crimes, Congress has not done so but instead has continued to revise the federal criminal laws and to create new federal crimes without addressing the issue of duress. Under these circumstances, I believe that the burdens remain where they were when Congress began enacting federal criminal statutes.

I do not assume that Congress makes a new, implicit judgment about the allocation of these burdens whenever it ere*20ates a new federal crime or, for that matter, whenever it substantially revises an existing criminal statute. ' It is unrealistic to assume that on every such occasion Congress surveys the allocation of the burdens of proof on duress under the existing federal case law and under the law of the States and tacitly adopts whatever the predominant position happens to be at the time. Such a methodology would create serious problems for the district courts and the courts of appeals when they are required to decide where the burden of persuasion should be allocated for federal crimes enacted on different dates. If the allocation differed for different offenses, there might be federal criminal cases in which the trial judge would be forced to instruct the jury that the defendant bears the burden of persuasion on this defense for some of the offenses charged in the indictment and that the prosecution bears the burden on others.

I would also not assume, as Justice Breyer does, see post, at 22 (dissenting opinion), that Congress has implicitly delegated to the federal courts the task of deciding in the manner of a common-law court where the burden of persuasion should be allocated. The allocation of this burden is a debatable policy question with an important empirical component. In the absence of specific direction from Congress, cf. Fed. Rule Evid. 501, I would not assume that Congress has conferred this authority on the Judiciary.

Justice Breyer, with whom Justice Souter joins, dissenting.

Courts have long recognized that “duress” constitutes a defense to a criminal charge. Historically, that defense “excuse[d] criminal conduct” if (1) a “threat of imminent death or serious bodily injury” led the defendant to commit the crime, (2) the defendant had no reasonable, legal alternative to breaking the law, and (3) the defendant was not responsible for creating the threat. United States v. Bailey, 444 U. S. 394, 409-410 (1980); see also 2 W. LaFave, Substantive *21Criminal Law § 9.7(b), pp. 74-82 (2003) (hereinafter LaFave); ante, at 4, n. 1 (opinion of the Court). The Court decides today in respect to federal crimes that the defense must bear the burden of both producing evidence of duress and persuading the jury. I agree with the majority that the burden of production lies on the defendant, that here the burden of persuasion issue is not constitutional, and that Congress may allocate that burden as it sees fit. But I also believe that, in the absence of any indication of a different congressional intent, the burden of persuading the jury beyond a reasonable doubt should lie where such burdens normally lie in criminal cases, upon the prosecution.

I

My disagreement with the majority in part reflects my different view about how we should determine the relevant congressional intent. Where Congress speaks about burdens of proof, we must, of course, follow what it says. But suppose, as is normally the case, that the relevant federal statute is silent. The majority proceeds on the assumption that Congress wished courts to fill the gap by examining judicial practice at the time that Congress enacted the particular criminal statute in question. Ante, at 12-16. I would not follow that approach.

To believe Congress intended the placement of such burdens to vary from statute to statute and time to time is both unrealistic and risks unnecessary complexity, jury confusion, and unfairness. It is unrealistic because the silence could well mean only that Congress did not specifically consider the “burden of persuasion” in respect to a duress defense. It simply did not think about that secondary matter. Had it done so, would Congress have wanted courts to freeze current practice statute by statute? Would it have wanted to impose different burden-of-proof requirements where claims of duress are identical, where statutes are similar, where the only relevant difference is the time of enactment? Why? *22Indeed, individual instances of criminal conduct often violate several statutes. In a trial for those violations, is the judge to instruct the jury to apply different standards of proof to a duress defense depending upon when Congress enacted the particular statute in question? What if in this very case the defendant’s boyfriend had given her drug money and insisted (under threat of death) not only that she use some of the money to buy him a gun, but that she launder the rest? See 18 U. S. C. § 1956 (2000 ed. and Supp. II); see infra, at 25-27.

I would assume instead that Congress’ silence typically means that Congress expected the courts to develop burden rules governing affirmative defenses as they have done in the past, by beginning with the common law and taking full account of the subsequent need for that law to evolve through judicial practice informed by reason and experience. See Davis v. United States, 160 U. S. 469 (1895); McNabb v. United States, 318 U. S. 332, 341 (1943); ante, at 14, n. 8 (opinion of the Court) (proposed general revision of the federal criminal code would have instructed courts to determine the contours of affirmative defenses “ ‘according to the principles of the common law as they may be interpreted in the light of reason and experience’ ”); 9 J. Wigmore, Evidence § 2486, p. 291 (J. Chadbourn rev. ed. 1981) (allocation of the burdens of proof present courts with questions “of policy and fairness based on experience in the different situations”). That approach would produce uniform federal practice across different affirmative defenses, as well as across statutes passed at different points in time.

II

My approach leads me to conclude that in federal criminal cases, the prosecution should bear the duress defense burden of persuasion. The issue is a close one. In Blackstone’s time the accused bore the burden of proof for all affirmative defenses. See 4 W. Blackstone, Commentaries *201; Patterson v. New York, 432 U. S. 197, 201-202 (1977). And 20th-century experts have taken different positions on the matter. *23The Model Penal Code, for example, recommends placing the burden of persuasion on the prosecution. ALI, Model Penal Code §1.12, p. 16, § 1.13(9)(c), p. 18, §2.09, pp. 37-38 (1985). The Brown Commission recommends placing it upon the defendant. 1 National Commission on Reform of Federal Criminal Laws, Working Papers 278 (1970). And the proposed revision of the federal criminal code, agnostically, would have turned the matter over to the courts for decision. S. 1722, 96th Cong., 1st Sess., §501 (1979). Moreover, there is a practical argument that favors the Government’s position here, namely, that defendants should bear the burden of persuasion because defendants often have superior access to the relevant proof.

Nonetheless, several factors favor placing the burden on the prosecution. For one thing, in certain respects the question of duress resembles that of mens rea, an issue that is always for the prosecution to prove beyond a reasonable doubt. See In re Winship, 397 U. S. 358, 364 (1970); Martin v. Ohio, 480 U. S. 228, 234 (1987). The questions are not the same. The defendant’s criminal activity here was voluntary; no external principle, such as the wind, propelled her when she acted. The Nicomachean Ethics of Aristotle, p. 54 (R. Browne transí. 1865). Moreover, her actions were intentional. Whether she wanted to buy the guns or not, and whether she wanted to lie while doing so or not, she decided to do these things and knew that she was doing them. Indeed, her action was willful in the sense that she knew that to do them was to break the law. Ante, at 5-7 (opinion of the Court); see also Ratzlaf v. United States, 510 U. S. 135, 136-137 (1994).

Nonetheless, where a defendant acts under duress, she lacks any semblance of a meaningful choice. In that sense her choice is not free. As Blackstone wrote, the criminal law punishes “abuse[s] of th[e] free will”; hence “it is highly just and equitable that a man should be excused for those acts, which are done through unavoidable force and compul*24sion.” 4 Commentaries *27. And it is in this “force and compulsion,” acting upon the will, that the resemblance to lack of mens rea lies. Cf. Austin, Ifs and Cans, in Proceedings of the British Academy 123-124 (1956) (noting difference between choosing to do something where one has the opportunity and ability to do otherwise and choosing to do something where one lacks any such opportunity or ability). Davis v. United States, supra, allocated the federal insanity defense burden to the Government partly for these reasons. That case, read in light of Leland v. Oregon, 343 U. S. 790, 797 (1952), suggests that, even if insanity does not always show the absence of mens rea, it does show the absence of a “ ‘vicious will.’ ” Davis, supra, at 484 (citing Blackstone; emphasis added).

For another thing, federal courts (as a matter of statutory construction or supervisory power) have imposed the federal-crime burden of persuasion upon the prosecution in respect to self-defense, insanity, and entrapment, which resemble the duress defense in certain relevant ways. In respect to both duress and self-defense, for example, the defendant’s illegal act is voluntary, indeed, intentional; but the circumstances deprive the defendant of any meaningful ability or opportunity to act otherwise, depriving the defendant of a choice that is free. Insanity, as I said, may involve circumstances that resemble, but are not identical to, a lack of mens rea. And entrapment requires the prosecution to prove that the defendant was “predisposed” to commit the crime—a matter sometimes best known to the defendant.

As to self-defense, see First Circuit Pattern Criminal Jury Instructions § 5.04 (1998); United States v. Thomas, 34 F. 3d 44, 47 (CA2 1994); Government of Virgin Islands v. Smith, 949 F. 2d 677, 680 (CA3 1991); United States v. Harris, Nos. 95-5637, 95-5638, 1996 U. S. App. LEXIS 22040, *4-*5 (CA4, Aug. 27, 1996); United States v. Branch, 91 F. 3d 699, 714, n. 1 (CA5 1996); Sixth Circuit Pattern Criminal Jury Instructions § 6.06 (2005); United States v. Jackson, 569 F. 2d *251003, 1008, n. 12 (CA7 1978); United States v. Pierre, 254 F. 3d 872, 876 (CA9 2001); United States v. Corrigan, 548 F. 2d 879, 883 (CA10 1977); United States v. Alvarez, 755 F. 2d 830, 842 (CA11 1985); Bynum v. United States, 408 F. 2d 1207 (CADC 1968); see also Mullaney v. Wilbur, 421 U. S. 684, 702, n. 30 (1975) (noting this as the “‘majority rule’”).

As to insanity, see Davis, 160 U. S., at 486; Leland, supra, at 797 (making clear that Davis determined burden allocations as a matter of federal, but not constitutional, law); but see 18 U. S. C. § 17(b) (overruling this default rule to place the burden on the defendant by clear and convincing evidence). As to entrapment, see Jacobson v. United States, 503 U. S. 540, 554 (1992) (reversing the judgment affirming the conviction because “the prosecution failed, as a matter of law, to adduce evidence to support the jury verdict that petitioner was predisposed, independent of the Government’s acts and beyond a reasonable doubt,” to commit the crime). See also Patterson, 432 U. S., at 202 (noting that Davis “had wide impact on the practice in the federal courts with respect to the burden of proving various affirmative defenses”); Patterson, supra, at 231 (Powell, J., dissenting) (“[SJince this Court’s decision in Davis . . . 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”).

Further, most federal courts, in respect to most federal crimes, have imposed the burden of persuasion in respect to the duress defense upon the Government, following Johnson v. United States, 291 F. 2d 150, 155 (CA8 1961), and authorities such as E. Devitt & C. Blackmar, Federal Jury Practice and Instructions § 13.14, p. 293 (2d ed. 1970), and Sthe Federal Judicial Center Pattern Criminal Jury Instructions §56 (1988). By the mid-1990’s, seven Circuits had squarely placed the burden of persuasion upon the prosecution; one Circuit (the Fifth) placed the burden on the defendant; and four (the Third, Fourth, Eleventh, and District of Columbia) *26did not, as far as I can tell, have a definitive practice. Compare United States v. Arthurs, 73 F. 3d 444, 448 (CA1 1996); United States v. Mitchell, 725 F. 2d 832, 836 (CA2 1983); United States v. Campbell, 675 F. 2d 815, 821 (CA6 1982); United States v. Talbott, 78 F. 3d 1183, 1186 (CA7 1996) (per curiam); United States v. Campbell, 609 F. 2d 922, 925 (CA8 1979); United States v. Hearst, 563 F. 2d 1331, 1336, and n. 2 (CA9 1977) (per curiam); and United States v. Falcon, 766 F. 2d 1469, 1477 (CA10 1985), with United States v. Willis, 38 F. 3d 170, 179 (CA5 1994) (putting the burden on the defendant by a preponderance). Compare also First Circuit Pattern Criminal Jury Instructions § 5.05 (1998); Sixth Circuit Pattern Criminal Jury Instructions §6.05 (1991); Seventh Circuit Pattern Criminal Federal Jury Instructions § 6.08 (1998); and Eighth Circuit Pattern Criminal Jury Instructions §§3.09, 9.02 (2000), with Fifth Circuit Pattern Criminal Jury Instructions §1.36 (2001). Petitioner adds, without contradiction, that the States allocate the burden similarly by a ratio of 2 to 1. Brief for Petitioner 32-34; Brief for United States 38, n. 30.

Beginning in 1991, the matter became more complicated because the Ninth Circuit began to require the defendant to bear the burden of proving duress in certain circumstances. United States v. Dominguez-Mestas, 929 F. 2d 1379, 1382, 1384 (per curiam). And a few years later the Third, Sixth, and Eleventh Circuits followed suit in cases concerning a closely related justification defense. See United States v. Dodd, 225 F. 3d 340, 347-350 (CA3 2000); United States v. Brown, 367 F. 3d 549, 555-556 (CA6 2004); United States v. Deleveaux, 205 F. 3d 1292, 1298-1300 (CA11 2000); Eleventh Circuit Pattern Criminal Jury Instructions § 16 (2003). But see Sixth Circuit Pattern Criminal Jury Instructions §6.05 (2005) (stating that the burden-of-proof issue for duress is undecided in that Circuit).

These latter cases, however, put the burden on the defendant only where the criminal statute narrows its mens rea *27requirement, i. e., the burden is the defendant’s where the statute requires that the defendant act with “knowledge” but not, suggest these courts, where the statute requires that the defendant act “willfully,” “intentionally,” or “voluntarily.” See, e. g., Dominguez-Mestas, supra, at 1382, 1384; United States v. Meraz-Solomon, 3 F. 3d 298, 300 (CA9 1993) (per curiam); Ninth Circuit Pattern Criminal Jury Instructions §§6.5, 6.6 (2003); but see United States v. Fei Lin, 139 F. 3d 1303, 1307-1308 (CA9 1998). See also Eleventh Circuit Pattern Criminal Jury Instructions § 16 (2003); United States v. Diaz, 285 F. 3d 92, 97 (CA1 2002) (indicating that this bifurcated rule might be appropriate, but noting Circuit precedent to the contrary). Similarly, the Tenth Circuit placed the burden of proving duress upon the defendant in “strict liability” cases where mens rea is not an element of the crime at all. United States v. Unser, 165 F. 3d 755, 763-765 (1999).

The apparent upshot is that four Circuits now place the burden of persuasion on the prosecution across the board; one places the burden on the prosecution if the statute requires mens rea but not otherwise; and four have held or suggested that the burden should be on the prosecution if the statute requires an intentional or willful state of mind, but not if the statute requires only knowledge. While the Circuits are divided, apparently only one (the Fifth) agrees with the position taken by the Court today.

Further, while I concede the logic of the Government’s practical argument—that defendants have superior access to the evidence—I remain uncertain of the argument’s strength. After all, “[i]n every criminal case the defendant has at least an equal familiarity with the facts and in most a greater familiarity with them than the prosecution.” Tot v. United States, 319 U. S. 463, 469 (1943). And the strict contours of the duress defense, as well as the defendant’s burden of production, already substantially narrow the circumstances under which the defense may be used. A defendant may find it difficult, for example, to show duress where the *28relevant conduct took place too long before the criminal act. Cf. ante, at 18-19 (Kennedy, J., concurring). That is because the defendant must show that he had no alternative to breaking the law. Supra, at 20-21. And that will be the more difficult to show the more remote the threat. See also LaFave § 9.7, at 77-79 (duress generally requires an “immediate” or “imminent” threat, that the defendant “take advantage of a reasonable opportunity to escape,” and that the defendant “terminate his conduct ‘as soon as the claimed duress . . . had lost its coercive force’”). More important, the need to prove mens rea can easily present precisely the same practical difficulties of proof for the prosecutor. Suppose for example the defendant claims that an old lady told him that the white powder he transported across the border was medicine for her dying son. Cf. United States v. Mares, 441 F. 3d 1152 (CA10 2006). See also Mullaney v. Wilbur, 421 U. S., at 702 (requiring the government to prove an absence of passion in a murder conviction imposes “no unique hardship on the prosecution”).

It is particularly difficult to see a practical distinction between this affirmative defense and, say, self-defense. The Government says that the prosecution may “be unable to call the witness most likely to have information bearing on the point,” namely, the defendant. Brief for United States 21. But what is the difference in this respect between the defendant here, who says her boyfriend threatened to kill her, and a battered woman who says that she killed her husband in self-defense, where the husband’s evidence is certainly unavailable? See also Jacobson, 503 U. S. 540 (entrapment; need to prove “propensity”). Regardless, unless the defendant testifies, it could prove difficult to satisfy the defendant’s burden of production; and, of course, once the defendant testifies, cross-examination is possible.

In a word, I cannot evaluate the claim of practicality without somewhat more systematic evidence of the existence of a problem, say, in those Circuits that for many years have *29imposed the burden on the prosecutor. And, of course, if I am wrong about the Government’s practical need (and were my views to prevail), the Government would remain free to ask Congress to reallocate the burden.

Finally, there is a virtue in uniformity, in treating the federal statutory burden of persuasion similarly in respect to actus reus, mens rea, mistake, self-defense, entrapment, and duress. The Second Circuit, when imposing the burden of persuasion for duress on the prosecution, wrote that differences in this respect create “a grave possibility of juror confusion.” United States v. Mitchell, 725 F. 2d 832, 836 (1983) (Newman, J., joined by Feinberg, C. J., and Friendly, J.). They risk unfairness as well.

For these reasons I believe that, in the absence of an indication of congressional intent to the contrary, federal criminal law should place the burden of persuasion in respect to the duress defense upon the prosecution, which, as is now common in respect to many affirmative defenses, it must prove beyond a reasonable doubt. With respect, I dissent.

LYNCH, Circuit Judge.

Defendant Robert C. Luisi, Jr., an admitted member of the “La Cosa Nostra” (LCN) crime family, appeals his convictions on three cocaine-related charges. These convictions stemmed from an FBI investigation that employed a paid cooperating witness and LCN member, Ronald Previte.

At trial, Luisi testified and admitted his involvement in the cocaine transactions. His defense was entrapment, on intertwined theories. He claimed that Previte, acting for the government along with undercover FBI agent Michael McGowan, had improperly tried to induce him to commit drug crimes. He further claimed that when he resisted, Previte persuaded Philadelphia LCN boss Joseph Merlino to order Luisi to engage in the charged drug transactions. Merlino was Luisi’s superior in the LCN, and the government was aware of the serious consequences Luisi would face if he refused to follow Merlino’s order.

The district court instructed the jury on the entrapment defense. However, the court’s supplemental instructions — given in response to a jury question — foreclosed the jury from considering Merlino’s role in the asserted government entrapment of Luisi. We conclude that those instructions were erroneous, and we vacate the convictions and remand the case.

I.

In July 1999, a grand jury in the District of Massachusetts indicted Luisi and three co-defendants on three charges: one count of conspiracy to possess cocaine with intent to distribute, see 21 U.S.C. §§ 841(a)(1), 846, and two counts of possession of cocaine with intent to distribute, see id. § 841(a). The conspiracy was alleged to have run from February 1999 through June 28, 1999. The two possession counts stemmed from transactions on April 30,1999, and June 3,1999.

Pursuant to plea agreements with the government, two of Luisi’s co-defendants pled guilty to the possession counts. The indictment against the third was dismissed following that defendant’s death, leaving Luisi as the sole defendant at a trial that commenced on September 9, 2002. We recount the key testimony.

In the late 1990s, the FBI conducted a major investigation into the operations of the Philadelphia LCN. Previte, a captain or “capo regime” in the LCN, assisted the FBI investigation by working as a cooperating witness under a personal services contract with the FBI. He was paid a substantial sum of money in return.

The FBI came to learn that Luisi was working for the Philadelphia LCN as a captain, and that he was supervising the criminal activity that the organization undertook in Boston. Eager to get evidence against Luisi, the FBI had Previte introduce Luisi to McGowan, who posed as a source of illegal money-making opportunities.

The introduction took place over January 11th and 12th, 1999. McGowan operated under the pseudonym “Michael O’Sullivan” and purported to be in the import/export business. He told Luisi he had previously worked with Capo Previte in Philadelphia, and said he had now relocated to Boston. McGowan explained to Luisi that as part of his business he was sometimes presented with “opportunities” *46and that sometimes he needed help taking advantage of these “deals.” Luisi agreed that he would look at future deals with McGowan. Unbeknownst to Luisi, this conversation was being recorded by the FBI — as were the vast majority of the future conversations Luisi would have with McGowan.

The first “opportunity” occurred on February 10, 1999, when McGowan presented Luisi with several “stolen” furs. Luisi was not sure if he would be able to sell them, but he stated he would look into it. He also inquired whether McGowan had other items; when McGowan mentioned the possibility of obtaining jewelry, Luisi expressed more interest in that, and particularly in diamonds. There was no mention of any drugs at this point.

Several days later, Luisi and Previte spoke to each other at a party in Philadelphia. Merlino was also present at the party. Luisi testified that Previte proposed a “swap” of cocaine for diamonds, and that Luisi’s response was that he would “try” to get the deal done. He testified that he gave this response because “at the party [Merlino] ... made it very clear to me that he wanted these drugs,” although Luisi later clarified that he did not at that time understand Merlino to be giving an “order” to do the deal, but merely “permission.” That would change. In any event, Luisi testified that at or shortly after the party he chose not to do the deal.

On March 8, 1999, McGowan again met with Luisi, along with two of Luisi’s associates.1 McGowan referred to Previte’s proposed swap and stated that he (McGowan) knew a guy with diamonds, and that the guy was looking to exchange them for “three bricks.”2 Luisi’s immediate response was: “I want to get them, I want to bring them to [a jeweler friend of mine], if he likes it, boom. We’ll do the deal and I’ll do it that way, whatever [Previte] wants.” McGowan interpreted this to mean that Luisi wanted to see the diamonds, and that he would be willing to exchange cocaine for them.

Several minutes later, however, Luisi took McGowan aside privately. This part of the conversation was not recorded. According to McGowan, Luisi told him that because Previte had referred McGowan to him, Luisi would make “every effort” to get the cocaine, but it would be difficult and it would take time. Luisi testified that he did not actually agree to do a drugs-for-diamonds swap.

Also during the March 8th conversation, McGowan asked Luisi what items, other than the diamonds, he would be interested in. Luisi responded that he would be interested in jewelry, watches, and cigarettes, and some of his associates mentioned film and razor blades.

McGowan’s next meeting with Luisi and his associates came on March 11, 1999. McGowan had some “stolen” Polaroid film, and the participants discussed how it was to be sold. Luisi reported on his only partially successful attempts to sell fur coats, and the participants also discussed diamonds and jewelry. Later during the meeting, McGowan mentioned that Previte was coming up to Boston in a few days, and Luisi agreed to meet with both Previte *47and McGowan then. Luisi and his associates left with the film.

Previte came to Boston, and on March 16, 1999, he met with McGowan, Luisi, and some of Luisi’s associates. The participants had a cryptic conversation during which, according to McGowan, Luisi confirmed that he would get the eocaine-for-diamonds deal done. The following day, McGowan talked to Luisi over the phone, and again inquired into the status of the cocaine deal with Previte. Luisi replied that he would work on it, but indicated that the deal would not happen immediately-

During this time, Luisi had also been trying to sell the film that McGowan had given him on March 11th. He was unable to do so at a price that McGowan was willing to accept, and so on March 19th Luisi returned the film. After Luisi again expressed his preference for jewelry, and after McGowan again reaffirmed his ability to get jewelry, the conversation turned back to the proposed diamonds-for-cocaine deal. Luisi made comments that, if taken at face value, expressed a reluctance to go ahead with the deal and indicated that Luisi had “nothing to do with” the cocaine business. Luisi also explained to McGowan that “in the last ... three years I lost over a dozen and a half guys to that.... And I have to make a stern, a firm stand here.... I don’t wanna have nothing to do with it.”

Luisi then said that he would send a guy named Danny White, not affiliated with Luisi, to do the deal. Luisi told McGowan that once White made contact with McGowan, McGowan would have exactly seven days to complete the transaction. Luisi and McGowan also discussed cash terms for the deal (even though the deal had originally been conceived of as a barter for diamonds).

Luisi testified that Danny White actually is a fictitious person whom Luisi made up in order to pretend that he was cooperating with McGowan.3 McGowan testified that he never met White at any point, that he never had any conversation with him, and that he did not know whether White was a fictitious person or not.

Luisi and McGowan had no contact with each other for the next three weeks.4 On April 19, 1999, McGowan initiated a phone conversation with Luisi. McGowan turned the discussion to dealings with Previte, and Luisi responded that “[ejverything’s gonna be okay soon.” McGowan understood this to be a reference to the cocaine transaction.

McGowan initiated another phone conversation with Luisi on April 23, 1999. McGowan told him that Previte would be coming to Boston on April 28th, and he asked Luisi to be available then. Luisi said he would probably be available, and he also said “Pm ah gonna be calling you ah with my other friend any day.” McGowan interpreted the “other friend” to be a reference to Danny White. In the same phone conversation, McGowan mentioned that he might soon have more stolen property coming in.

On April 27, 1999, one day before Prev-ite’s planned trip to Boston, Previte had a conversation with Merlino, his superior in the LCN. Previte was wearing a wire, and the conversation was recorded. As re*48vealed by the tape, Previte complained to Merlino that Luisi had not yet done the cocaine transaction, despite Luisi’s representations. Previte explained to Merlino that he had “big money sittin[g] on the line,” and that Merlino would also make money from the transaction. He directly asked Merlino if there was “any way you could just tell [Luisi] to do what he gotta do.” Merlino agreed to do so. Previte and Merlino then agreed that when Prev-ite went to Boston the next day, Previte would put Luisi on the phone with Merlino, at which point Merlino would tell Luisi to do the cocaine deal. As Merlino put it on the tape: “I’ll say [to Luisi:] whatever [Previte] says to do[,] just do it.”

Previte was still cooperating with the FBI at the time he had the conversation with Merlino, and a jury could conclude that the FBI had in fact directed Previte’s request of Merlino. Previte did not testify, nor did any of the Philadelphia FBI agents who had worked with him. But McGowan was asked if the FBI had arranged the meeting between Previte and Merlino, and his response was that while he did not know, he “assume[d] so because Previte was cooperating.”

Previte flew to Boston on the morning of April 28th, the day after his conversation with Merlino. He went to McGowan’s office, and McGowan arranged for Previte to make a three-way call with LCN boss Merlino and Capo Luisi, Merlino’s underling. The call was recorded in its entirety. It began with Previte calling Luisi. Once Luisi was on the line, Previte brought Merlino into the conversation. After an exchange of preliminaries, Merlino (somewhat cryptically) got down to business.5 Luisi testified that he understood Merlino to be ordering him to get the cocaine deal done, and that he agreed to do the deal as a result of this.

McGowan, who had been listening in to the conversation, testified to having a similar understanding. As he put it, “[a]fter this phone conversation, I expected to receive cocaine.” His hopes were soon realized.

Within an hour after Merlino spoke to Luisi, Luisi met with Previte and McGowan to confirm details of the drug transaction. Initially, Previte and Luisi had a private conversation to work out certain points, and Previte explained that McGowan wanted to do multiple cocaine deals. Luisi agreed, and the private conversation ended shortly thereafter. With Luisi looking on, Previte then informed McGowan that the mysterious Danny White would not be playing a role in the deal, and that his place would be taken by Shawn Vetere, one of Luisi’s associates. Vetere promptly put McGowan in touch with Bobby Carroz-za.6 McGowan worked out more details with Carrozza. On April 30, 1999 — two days after the call with Merlino — Carrozza sold two kilograms of cocaine to McGowan. Carrozza told McGowan that the cocaine came “right from [Luisi and Vetere]. I wouldn’t be able to do it any other way.”

Two weeks later, McGowan gave Luisi a $1,000 “tribute” payment for arranging the *49transaction. McGowan and Luisi also engaged in preliminary discussions about a future cocaine deal, and Luisi told McGowan to work the rest out with Carrozza. McGowan did so.

The next cocaine delivery was not immediately forthcoming, however. On May 24, 1999, McGowan complained to Luisi, who told him to be patient. On June 1, 1999, Luisi proposed certain changes to the impending cocaine transaction; McGowan agreed to the revised cocaine deal the next day, and he paid Luisi the $24,000 cash price. On June 3, 1999, Carrozza and Tommy Wilson (another of Luisi’s associates) came to McGowan’s office to deliver one kilogram of cocaine. McGowan later gave Luisi a $500 “tribute” payment for arranging the deal.7

Luisi was the sole witness to testify for the defense. He testified to several additional pertinent points. He admitted that he had been a captain in the LCN since the fall of 1998, and he agreed that the LCN was properly described as “the Mafia” and as “the mob.” As a captain, one of his jobs had been to make “tribute” payments to Merlino, and these payments had come out of Luisi’s earnings from the criminal enterprises conducted by his “crew” in Boston. Luisi explained that the LCN was extremely hierarchical, and he stated that when the head of the LCN ordered him to do the cocaine transaction, he felt that he had no alternative other than to fulfill the order. He was also asked why he had engaged in the second cocaine transaction; while he did not specifically reference Merlino’s order, he responded that he had not wanted to do the drug deal, and he did so because he “had to bring money to Philadelphia” and so he “was desperate.”

Luisi admitted that at one point in his life he had been involved in drug distribution. However, he testified that he stopped his involvement in that business when he joined the LCN in mid-1998, and he gave several reasons for this. At the time Luisi joined, Georgie Borghesi (another LCN captain) and Merlino both told Luisi that he was not to deal drugs. Additionally, Luisi testified that around that time he had a “spiritual encounter,” and he realized that his “whole lifestyle was wrong.” Because he was an LCN captain, he found it difficult to realize fully his spiritual aspirations, but he testified that he was at least partially able to implement them by ending his involvement with drugs as of late 1998.8

Luisi also offered an explanation for why he had seemed receptive to the cocaine deal, even before receiving the order from Merlino on April 28th. He claimed he had been trying to “pal off’ McGowan and Previte; that is, he politely pretended to be cooperating with them on the drug deal, while in fact he had no intention of ever delivering drugs to them.

Before the jury was instructed, Luisi’s attorney asked the district court to dismiss the case on the ground that the govern*50ment had engaged in allegedly outrageous conduct, thereby violating Luisi’s due process rights. The court never ruled on that motion; such a motion is an issue for the judge and not the jury. See United States v. Bradley, 820 F.2d 3, 7 n. 5 (1st Cir.1987). However, over the government’s objection, the district court did agree that Luisi was entitled to an entrapment instruction.

The district court’s entrapment instructions correctly informed the jury that the government had the burden to prove, beyond a reasonable doubt, that Luisi had not been entrapped. See United States v. Walter, 434 F.3d 30, 37 (1st Cir.2006). The district court further explained that the government had to prove, beyond a reasonable doubt, that at least one of two things was true: either (1) “no government agent9 or person acting on behalf [of] or ... under [the] auspices of the government persuaded or induced the defendant to commit” the charged crimes; or (2) “the defendant was ready and willing to commit the [charged] crimefs] without persuasion from the government.” This was also a correct statement of the law. See United States v. Gamache, 156 F.3d 1, 9 (1st Cir.1998) (explaining that the two prongs of an entrapment defense are improper government inducement and lack of predisposition).

Luisi specifically asked for an instruction indicating that if the jury found that Previte had induced Merlino, that meant that Merlino’s order could be considered government action. The district court refused, stating that the instruction was improper because there was a factual dispute over whether the government’s responsibility ended due to the presence of an intermediary.

Shortly after commencing deliberations, the jury sent the court a question that revealed it was considering Merli-no’s role and how it related to the entrapment defense. The jury asked: “Is Merlino’s request of Luisi, if determined to be excessive pressure, considered to be government persuasion or inducement because the contact between Merlino and Luisi resulted from the government agent Previte and Merlino?”

The court and the parties researched the issue overnight, and returned in the morning to discuss the proper response. Luisi contended that because Previte had spoken to Merlino about the cocaine transaction, and because Merlino’s order had been facilitated by Previte and McGowan (the latter of whom had actually placed the three-way call to Merlino), the actions of Merlino, Previte, and McGowan together could be attributed to the government. The government disagreed, and based its argument on a case from outside this circuit, United States v. Washington, 106 F.3d 983 (D.C.Cir.1997). The government described Luisi’s asserted defense as “derivative entrapment,” and it claimed that Washington was the only case it could find recognizing the defense. As the government read Washington, Luisi’s claim was foreclosed on the facts presented.

The court took a different route, and it ultimately concluded that a case from this circuit, United States v. Bradley, was controlling. In Bradley a prison inmate named Constanza, who could have been deemed to have been acting as a government agent, directly threatened an intermediary named Brenner to do a drug deal on pain of physical harm. 820 F.2d at 5-6. The intermediary was unable to do the deal on his own, and he in turn pressured his friend — the defendant Bradley — to assist him. See id. at 7-8. We found the *51evidence sufficient to support an entrapment instruction for the intermediary Brenner, as the government agent had directly threatened him. Id. But we held that Brenner’s friend, defendant Bradley, was not entitled to an entrapment instruction: while Brenner could claim duress, Bradley had “only an appeal to sympathy, which he was free to reject.” Id. at 7. We stated that we “would not extend the [entrapment] defense to a remote defendant without, at least, a showing that pressure had been put upon him by the intermediary at the instruction of the government agent.” Id. at 8.

Thus in Bradley this court rejected defendant Bradley’s argument that the government’s improper inducement of Brenner could be an indirect entrapment of Bradley, as the agent (Constanza) had neither ordered nor expected Brenner to entrap other persons. See id. at 7. We said that a “quite different case would be presented if Constanza had targeted a putative seller and had instructed Brenner to put the arm on him.” Id. We then added a footnote to “suggest that such a case, though argued to be a third-party case, is not really a third-party case at all. The intermediary in such instance is really acting under instructions, as a government subagent — a quite different situation.” Id. at 7 n. 6.

The district court focused upon Bradley’s use of the word “instructions,” and it read that case as concerned with whether the government agent (Previte) had “instructed” the intermediary (Merlino) to pressure Luisi. The court concluded that because Previte ranked below Merlino in the LCN hierarchy, Previte was in no position to “instruct” Merlino to do anything.

Although Luisi argued for a broader reading of the word “instruct,” the district court rejected such a reading of Bradley. In the alternative, Luisi asked that the jury be allowed to determine whether Merlino had been “instructed,” but the court rejected that option as well.

The district court then called the jury back into the courtroom and answered the jury’s question as follows:

[I]n your consideration of the entrapment question, you should focus your attention on the relationship — the direct relationship between Mr. Previte ... and Mr. McGowan on the one hand, and Mr. Luisi on the other hand. You should consider evidence as [it] relates to the direct contact between and among those people.

Because the district court omitted Merlino from this statement, while simultaneously mentioning other individuals, the jury likely concluded that it could not consider Merlino’s role in any inducement of Luisi. This was particularly so since the jury’s question specifically asked about Merlino, and the court told the jury to focus on individuals other than Merlino. As a result, and as the government has not disputed, the jury was precluded from finding that Merlino’s order to Luisi could be deemed improper governmental inducement.

Later that day, the jury convicted Luisi on all three counts of the indictment.

II.

Luisi’s primary contention on appeal is that the district court’s jury instructions, as supplemented by its answer to the jury’s question, were incorrect as a matter of law. We review that issue de novo here. See United States v. Buttrick, 432 F.3d 373, 376 (1st Cir.2005).

We agree with Luisi that the district court’s instructions were erroneous. We begin by explaining the nature and policies behind the entrapment defense generally. *52We address Bradley, and conclude that it supports Luisi’s claim of error. We also reject the government’s alternative argument, which we review de novo, that there was no reversible error on the facts presented because Luisi was not entitled to any entrapment instruction whatsoever. See United States v. Nishnianidze, 342 F.3d 6, 17 (1st Cir.2003) (stating that a defendant is entitled to a jury instruction so long as his legal theory is valid and there is evidence in the record to support it).

A. Understanding the Entrapment Defense

In federal criminal cases, the entrapment defense is neither a doctrine of constitutional dimension, nor a defense specifically granted by statute. See United States v. Russell, 411 U.S. 423, 432-33, 93 S.Ct. 1637, 36 L.Ed.2d 366 (1973). Rather, the defense has its origins in an inference about congressional intent. Sherman v. United States, 356 U.S. 369, 372, 78 S.Ct. 819, 2 L.Ed.2d 848 (1958); see also United States v. Gendron, 18 F.3d 955, 961 (1st Cir.1994). The Supreme Court has explained that the “function of law enforcement is the prevention of crime and the apprehension of criminals. Manifestly, that function does not include the manufacturing of crime.... Congress could not have intended that its statutes were to be enforced by tempting innocent persons into violations.” Sherman, 356 U.S. at 372, 78 S.Ct. 819. A successful entrapment defense requires that there be a reasonable doubt on both prongs of a two-pronged test.

The first prong necessitates a showing of improper government inducement. See Gamache, 156 F.3d at 9. This aspect of the defense plainly seeks to deter improper government conduct. Gendron, 18 F.3d at 961. Indeed, a defendant cannot claim entrapment when government conduct played no causal role in the defendant’s inducement. See Sherman, 356 U.S. at 372, 78 S.Ct. 819. Nevertheless, the entrapment defense only deters government misbehavior in cases where the defendant would otherwise be law-abiding. Gendron, 18 F.3d at 962. That is because the second prong requires that the defendant have a lack of predisposition to commit the charged offense. See id.; see also Sorrells v. United States, 287 U.S. 435, 448, 53 S.Ct. 210, 77 L.Ed. 413 (1932).

These two prongs, and the policy concerns behind them, play important roles in delimiting the scope of the entrapment defense. But they are not the only considerations that matter. This court and the Supreme Court have taken account of the practical problems facing law enforcement, particularly in the prosecution of “victimless” crimes where “significant governmental involvement in illegal activities” is often required. Bradley, 820 F.2d at 6-8; see also Hampton v. United States, 425 U.S. 484, 494-95 & n. 7, 96 S.Ct. 1646, 48 L.Ed.2d 113 (1976) (Powell, J., concurring in the judgment); Russell, 411 U.S. at 432, 93 S.Ct. 1637; Gendron, 18 F.3d at 961. We must be mindful that “the defense of entrapment ... [does] not ... give the federal judiciary a ‘chancellor’s foot’ veto over law enforcement practices of which it [does] not approve.” Russell, 411 U.S. at 435, 93 S.Ct. 1637.

These various considerations are sometimes in tension with one another, and we have treated the resolution of questions about the scope of the entrapment defense as essentially exercises in balancing. Indeed, in Bradley we weighed the competing factors and realized that we ultimately faced “a question of social policy.” 820 F.2d at 8; cf. United States v. Hollingsworth, 27 F.3d 1196, 1198 (7th Cir.1994) *53(en banc) (characterizing entrapment as a “common law doctrine”).

We synthesize the key facts on which we must balance competing concerns in this case. It is beyond dispute that an individual like Previte, hired by the government as an informant, is a “government agent” for entrapment purposes. See Sherman, 356 U.S. at 373-75, 78 S.Ct. 819. Nor can there be any dispute that Merlino’s order to Luisi, with its implied threat of physical harm or other serious retribution, could be found by a jury to be improper inducement if attributed to the government. See Gendron, 18 F.3d at 961; Bradley, 820 F.2d at 7.

It is also clear that Luisi’s case does not fit the pattern of what has come to be known as “vicarious entrapment.” In “vicarious entrapment” an unknowing middleman merely tells the defendant about an inducement that the government had used to target the middleman. See United States v. Valencia, 645 F.2d 1158, 1168-69 (2d Cir.1980) (recognizing the vicarious entrapment defense). Here, the target was not the middleman Merlino, but the defendant Luisi. Further, a jury could find that Merlino had himself threatened Luisi. This was not a case where Merlino repeated to Luisi a threat that Previte had made against Merlino. Indeed, Previte did not threaten Merlino at all.

Instead, this case is much closer to what has been called “derivative entrapment,” a situation in which a government agent “uses the unsuspecting middleman as a means of passing on an inducement” to the defendant. 2 W. LaFave, Substantive Criminal Law, § 9.8(a), at 93 (2d ed.2003).10 Yet even within this category, further refinement is required.

We have before us a situation in which a jury could find that Previte specifically targeted Luisi, and then “induced” Merlino to give an order to Luisi when Merlino might not have otherwise done so. But Previte’s inducement of Merlino does not appear to have itself been improper. Previte simply helped set up a drug transaction, explained to Merlino that Merlino would profit from the transaction’s execution, and then encouraged Merlino to order Luisi’s assistance. Cf. Gendron, 18 F.3d at 961-62 (explaining that improper inducement “goes beyond providing an ordinary ‘opportunity to commit a crime,’ ” and providing examples (quoting Jacobson v. United States, 503 U.S. 540, 550, 112 S.Ct. 1535, 118 L.Ed.2d 174 (1992))). A jury would presumably find that Previte merely provided an “ordinary” inducement to Merlino; it was Merlino’s inducement of Luisi that a jury could find improper.

This is an unusual entrapment situation. Under the original, correct instructions given, it is evident that the jury was considering the possibility that Merlino had put excessive pressure on Luisi, and that the jurors were unsure whether Merlino’s order could be considered “government persuasion or inducement because the contact between Merlino and Luisi resulted from the government agent Previte.” The effect of the court’s response was the same as if it had instructed the jury, as a matter of law, that Merlino’s order could not be considered government inducement or persuasion. We must decide whether the is*54sue was correctly removed from the jury’s consideration.

B. Thirdr-Party Entrapment After Bradley

In this circuit, Bradley is the leading case on third-party entrapment. Like the district court, the government now believes that Bradley controls this case. We examine Bradley’s facts and reasoning in greater detail.

The government agent in Bradley was a prison inmate named Constanza. In exchange for a reduced sentence, Constanza agreed to identify drug dealers for the government to assist in undercover operations and prosecutions. See Bradley, 820 F.2d at 5. Constanza told the government about an individual named Brenner, and then he threatened Brenner with physical harm if Brenner did not engage in a cocaine deal with an undercover agent. Id. at 5. Brenner in turn appealed to his friend, defendant Bradley, for assistance in obtaining the cocaine. Id. at 6. Though Bradley initially refused, Brenner explained that his physical safety was in jeopardy. Id. Upon hearing this, Bradley decided to assist Brenner. Id.

It should now be clear that Bradley presented a narrow fact pattern: it was a case of vicarious entrapment. That is, Brenner merely informed Bradley of a threat that had been made against Brenner, not one that had been made against Bradley. Although the Second Circuit’s Valencia opinion had recognized that such a vicarious entrapment defense could be viable, we considered and rejected that position. See id. at 8. At the same time, however, we rejected the government’s argument that a defendant can never be entrapped by a third party. See id. at 6-8.

Bradley reached these conclusions by weighing a variety of policy considerations as applied to the facts of the case. On the one hand was the fact that the crime would not have happened but for the government’s involvement. See id. at 6. On the other hand were several factors. First, the government’s role in ensnaring Bradley was “attenuated” because the agent’s threat neither ordered Brenner to seek assistance nor expected it. Id. at 7. Second, Bradley acknowledged the practical difficulties for the prosecution when it is forced to refute a defendant’s claim of entrapment in a scenario where the only two witnesses — the defendant and the intermediary — are likely to be hostile to the government.11 See id. Third, Bradley observed that the defendant did commit a crime, exhibiting socially inappropriate behavior. See id. at 6. Finally, Bradley noted the fact that undercover investigations are often needed to prosecute drug crimes. See id.

The Bradley court ultimately held that it “would not extend the [entrapment] defense to a remote defendant without, at least, a showing that pressure had been put upon him by the intermediary at the instruction of the government agent.” Id. at 8. The government reads “instruct” to mean “command,” such that Previte is not responsible for Merlino’s threats against Luisi unless Previte had commanded Mer-lino to order Luisi into the cocaine deal. Luisi reads “instruct” to mean “convince” or “inform.” But the dispute about the use of particular language in Bradley is largely beside the point. Bradley was a *55case in which the government agent neither “commanded,” “convinced,” nor “informed” the middleman (Brenner) to target Bradley. Indeed, it was the lack of any government targeting of Bradley whatsoever on which Bradley relied. See id. at 7. Here, unlike in Bradley, the government’s actions were specifically designed to pressure Luisi, and the government in fact expected that Luisi would be pressured. Bradley thus had no occasion to consider the fact pattern at hand.12

In United States v. Rogers, 102 F.3d 641 (1st Cir.1996), we considered a fact pattern that was more on point. In Rogers, the defendant did not move quickly to complete a drug transaction that an undercover agent had attempted to facilitate. Id. at 645. In an isolated statement, the government agent “told” the middleman to “put some heat on [the defendant].” Id. The agent had thus specifically targeted the defendant to receive an inducement, and Rogers relied on Bradley to conclude that the government would have been responsible if the agent had “told” the middleman “to apply the pressure or inducement later deemed improper.” Id. (emphasis added) (italics removed). However, Rogers factually did not involve an agent’s suggestion that improper pressure be applied to the defendant because “putting on some heat” was not improper. Rogers distinguished the more sinister suggestion of putting “the arm” on someone, a phrase that implied threatened force. Id. at 645-46.

Thus after Rogers, the law in this circuit permits an entrapment instruction involving a middleman when there is evidence that (1) a government agent specifically targeted the defendant in order to induce him to commit illegal conduct; (2) the agent acted through the middleman after other government attempts at inducing the defendant had failed; (3) the government agent requested, encouraged, or instructed the middleman to employ a specified inducement, which could be found improper, against the targeted defendant; (4) the agent’s actions led the middleman to do what the government sought, even if the government did not use improper means to influence the middleman; and (5) as a result of the middleman’s inducement, the targeted defendant in fact engaged in the illegal conduct.13

*56The government reads Rogers differently. It contends that the word “told” in Rogers must mean an imperative equivalent to the government’s interpretation of “instruct” in Bradley. It also reads Rogers to state a requirement that the agent have improperly induced the middleman.

We disagree. First, we think that the government’s argument is contrary to the plain meaning of “told,” and we see nothing in the context of Rogers that overrides this. Indeed, in concluding that the agent had not “told” the middleman to improperly induce the defendant, Rogers pointed out that “nothing in the record show[ed] that [the agent] urged, suggested or was even aware of’ the middleman’s improper inducement. See 102 F.3d at 645 (emphases added). Urging and suggesting are hardly equivalent to ordering, nor do they by themselves necessarily constitute improper inducement. Second, if the Rogers court had in fact read Bradley the same way as the government does, we do not see how the Rogers court could have considered it an open question whether the government can be responsible for an agent’s mere knowing tolerance of improper inducement by a middleman.14 See id.

Moreover, the policy concerns discussed in Bradley support our reading of Rogers. Bradley thought it important that the government’s role was “attenuated” in that case because the agent had not attempted to ensnare the defendant. Bradley, 820 F.2d at 7. But in a case where the government agent specifically targets the defendant, and then causes the middleman to take a specifically contemplated action (that is arguably improper pressure) with the goal of ensnaring the defendant, the government’s role is hardly attenuated. Additionally, Bradley worried about the fact that the government had turned a potentially innocent person into a criminal, see id. at 6, though it ultimately decided that this concern was outweighed by other issues. But in Bradley the government had never attempted to induce defendant Bradley directly. Bradley’s concern for potential innocents weighs heavier when the government has tried — and failed — to induce the defendant without the use of a middleman.

In light of our understanding of the law, we think a properly instructed jury could conclude that the government was responsible for Merlino’s order to Luisi. Indeed, such a jury could decide that: (1) Previte specifically requested that Merlino order Luisi to engage in the cocaine deal; (2) Previte’s request came after earlier government efforts to ensnare Luisi, without a middleman, had not been fruitful; (3) Previte, as an LCN captain, understood that the order he requested from Merlino would contain an implied threat of death, physical harm, or serious retribution if Luisi failed to comply; (4) Merlino’s order to Luisi was exactly what Previte had requested; and (5) Merlino would not have given the order if Previte had not encouraged him to do so.15 As a result, we think *57that the district court incorrectly answered the jury’s question.

This result is supported by a Fifth Circuit case. In United States v. Anderton, 629 F.2d 1044 (5th Cir.1980), the Fifth Circuit allowed the entrapment defense in a case where law enforcement officers specifically targeted the defendant, and then put unspecified “pressure” on the unwitting middleman to bring the defendant into a pre-designed criminal scheme. Id. at 1045, 1047. Without discussing whether the middleman had been improperly induced, the court found it important that the criminal design originated from the government itself. See id. at 1046-47.

Our conclusion is also supported by the D.C. Circuit’s analysis in Washington. In Washington, an undercover FBI agent posed as a drug lord and recruited a corrupt police officer to help him. See 106 F.3d at 990-91. After paying the officer for his services, the agent “asked [the officer] ... to recruit as many new officers as he could.” Id. at 991. The corrupt officer agreed, see id., but then recruited the defendants with an inducement not contemplated by the undercover agent. See id. at 992, 994-95. The Washington court rejected the derivative entrapment instruction on the facts of the case. Id. at 995. Importantly, however, it held that the entrapment defense is available where the government does contemplate the improper inducement given to the defendant, and causes the middleman to give that inducement as a result of a government agent’s “ ‘instruction or direction.’ ” Id. at 993 (quoting United States v. Layeni, 90 F.3d 514, 520 (D.C.Cir.1996)); see also id. at 995.16

The government cites to other cases, but none presented the fact pattern at issue here. See, e.g., Hollingsworth, 27 F.3d at 1200-02, 1204-05; United States v. Hodges, 936 F.2d 371, 371-72 (8th Cir.1991); United States v. Pilarinos, 864 F.2d 253, 254-56 (2d Cir.1988).17 The Seventh Circuit has even intimated that the defense might be available when a government agent specifically targets the defendant, and then encourages the middleman to induce that Zdefendant. See Hollingsworth, 27 F.3d at 1204 (citing, inter alia, Bradley).

*58Thus we conclude that the district court erred in answering the jury’s question in a way that excluded Merlino’s order from the jury’s consideration. We do not suggest that a jury would necessarily have concluded that Luisi was entrapped through Merlino. We hold only that the defendant was entitled to have a properly instructed jury consider the issue.

C. Predisposition

As a fallback argument, the government contends that Luisi failed to present sufficient evidence of his lack of predisposition, and thus was not entitled to any entrapment instruction at all. We readily dispose of this argument.

A judge may only instruct the jury on entrapment if the defendant meets his entry-level burden of production. United States v. Gifford, 17 F.3d 462, 467 (1st Cir.1994). That is, a defendant is entitled to an entrapment instruction only if there is “some evidence,” on both elements of the entrapment defense, sufficient to raise a reasonable doubt that the defendant was merely an unwary innocent. See United States v. Joost, 92 F.3d 7, 12 (1st Cir.1996). The defendant must show “more than a scintilla of evidence, more than mere creation of an opportunity for criminal activity.” Id. Nonetheless, even a defendant’s self-serving testimony can suffice, especially when it has some circumstantial corroboration. See id.

Luisi made the requisite entry-level showing. We have already explained why a jury could find improper governmental inducement. Additionally, Luisi introduced “some evidence” of his lack of predisposition. He testified that sometime before he met McGowan, he experienced a “spiritual encounter,” and that as a result of this spiritual encounter he resolved to stop dealing drugs. This story was consistent with evidence that Luisi had deliberately stalled the drug transaction for several months to “pal off’ McGowan, thereby indefinitely delaying a drug deal to which Luisi was opposed. Indeed, a jury could find that this two-month delay stood in stark contrast to the two-day lag between when Merlino gave Luisi the order, and when McGowan received drugs from Lui-si’s colleagues.

The government suggests that Luisi’s “spiritual encounter” was not credible, and that there was “overwhelming” evidence of Luisi’s predisposition, including the fact that he previously had been a cocaine dealer, and the fact that Luisi continued to receive profits from small drug deals undertaken by one of his associates.

In this procedural posture, however, our job is not “to weigh the evidence, make credibility determinations, or resolve conflicts in the proof.” Gamache, 156 F.3d at 9. Accordingly, we find that Luisi introduced sufficient evidence of his lack of predisposition to entitle him to an entrapment instruction.

III.

Luisi further argues that the charges against him in fact should have been dismissed because the government engaged in outrageous conduct that violated his due process rights. Outrageous government conduct is an issue of law, and it is the province of the district court—and not the jury—to rule on a defendant’s motion to dismiss on that ground. See Bradley, 820 F.2d at 7 n. 5. When a district court rules on such a motion, its ultimate conclusion is subject to de novo review, while its factual findings are reviewed for clear error. See United States v. Guzman, 282 F.3d 56, 58 (1st Cir.2002).

In this case, the district court never ruled on Luisi’s motion, and thus we have no factual findings in the record to assist *59us.18 Nonetheless, we think that the basic facts needed to rule on this matter are clear enough, and that we are able to reach the necessary legal conclusions. Luisi has asked us to find that the government’s conduct was outrageous, and the government has also asked us to resolve the issue.

The outrageousness doctrine permits dismissal of criminal charges only in those very rare instances when the government’s misconduct is so appalling and egregious as to violate due process by “shocking ... the universal sense of justice.” Russell, 411 U.S. at 432, 93 S.Ct. 1637 (quoting Kinsella v. United States ex rel. Singleton, 361 U.S. 234, 246, 80 S.Ct. 297, 4 L.Ed.2d 268 (1960))(internal quotation marks omitted); see also United States v. Nunez, 146 F.3d 36, 38 (1st Cir.1998). While the doctrine is often invoked by criminal defendants, it has never yet been successful in this circuit. See United States v. Santana, 6 F.3d 1, 4 (1st Cir.1993) (collecting First Circuit cases rejecting the argument); United States v. Panitz, 907 F.2d 1267, 1272-73 (1st Cir.1990) (collecting additional First Circuit cases rejecting the argument); see also, e.g., United States v. Capelton, 350 F.3d 231, 243 n. 5 (1st Cir.2003) (rejecting the argument on the facts presented); Nunez, 146 F.3d at 38-39 (same); United States v. Matiz, 14 F.3d 79, 82-83 (1st Cir.1994) (same). This case is no exception.19

Luisi’s argument for dismissal of the charges relies heavily on dicta in Bradley that “outrageous conduct ... might well be found in a threat of serious physical harm.” 820 F.2d at 7. But at the same time, Bradley acknowledged that an outrageousness claim might be defeated if a defendant has been “too active himself.” Id. Moreover, defendant Bradley himself had not been threatened, and we declined to consider the matter further. See id.

Whatever fact situation Bradley had in mind, it was not this one. Here, even though the government’s actions have risked giving the defendant a viable entrapment claim, it is another thing entirely to say that the conduct was “outrageous.” After considering the totality of the circumstances in this case, we think the government’s actions fell well short of shocking the “universal sense of justice.”

IV.

The judgments of conviction are vacated and the case is remanded for further proceedings consistent with this opinion.20