U.S. v. Scott

  The United States appeals from an opinion and order of the United States District Court for the Southern District of New York (Laura T. Swain, J.) vacating Gerald Scott’s sentence and resentencing him to time served. The district court held that because New York first-degree manslaughter can be committed by omission, it cannot serve as a predicate felony for the sentencing enhancements prescribed in the Armed Career Criminal Act (“ACCA”) or the Career Offender Sentencing Guideline. We hold that the district court properly concluded that New York first-degree manslaughter is not a predicate crime of violence because it can be committed by omission and therefore without the use of force, as defined in Curtis Johnson v. United States, 559 U.S. 133 (2010). We also hold that New York first-degree manslaughter does not match any of the generic offenses enumerated in the Career Offender Guideline. We therefore affirm the district court’s vacatur of Scott’s sentence under ACCA and its decision to resentence Scott to time served. Affirmed. Judge Leval joins in Judge Pooler’s opinion for the Court and also concurs by a separate opinion, in which Judge Pooler joins. Judge Raggi dissents in a separate opinion. ROSEMARY POOLER, C.J. The United States appeals from an opinion and order of the United States District Court for the Southern District of New York (Laura T. Swain, J.) vacating Gerald Scott’s sentence and resentencing him to time served based on his service of only eleven years of imprisonment. The district court held that because New York first-degree manslaughter can be committed by omission, it cannot serve as a predicate felony for the sentencing enhancements prescribed in the Armed Career Criminal Act (“ACCA”) or the Career Offender Sentencing Guideline. We hold that the district court properly concluded that New York first-degree manslaughter is not a predicate crime of violence because it can be committed by complete inaction and therefore without the use of force, as defined in Curtis Johnson v. United States, 559 U.S. 133 (2010). We also hold that New York first-degree manslaughter does not match any of the generic offenses enumerated in the Career Offender Guideline. We therefore affirm the district court’s vacatur of Scott’s sentence under ACCA and its decision to resentence Scott to time served. BACKGROUND On October 10, 2007, Gerald Scott pled guilty to Hobbs Act robbery in violation of 18 U.S.C. §§1951, 2 (“Count I”); brandishing a weapon during Hobbs Act robbery in violation of 18 U.S.C. §§924(c)(1)(A)(ii), 2 (“Count III”); and being a felon in possession of a firearm in violation of 18 U.S.C. §922(g)(1) and §924(e) (“Count IV”). Judge Swain sentenced Scott to a term of 151 months’ imprisonment on Count I to be served concurrently with a term of 180 months on Count IV, followed by a consecutive term of 84 months on Count III. Scott’s ACCA conviction on Count IV, which is the focus of this appeal, was based on the following predicate offenses: a 1983 conviction for New York robbery in the first degree and two 1988 convictions for New York manslaughter in the first degree. In the course of these crimes, Scott shot and killed one victim and stabbed another to death. These prior convictions subjected him to a mandatory sentence of 180 months on Count IV.¹ On April 26, 2010, Scott filed a habeas petition to vacate his conviction, arguing that he received ineffective assistance of counsel. Judge Swain denied that motion. In November 2016, Scott received this Court’s permission to file a successive habeas petition in light of Samuel Johnson v. United States, 135 S. Ct. 2551 (2015). The district court granted Scott’s successive habeas petition, finding Scott’s convictions for New York manslaughter were not violent felonies so that Scott was ineligible for an ACCA sentence. The district court reasoned that based on decisions of the New York Court of Appeals, “first degree manslaughter can be committed in New York State by omission and thus without using force.” United States v. Scott, No. 06 CR 988-LTS, 2017 WL 2414796, at *2 (S.D.N.Y. June 2, 2017). The court thus held that Scott was “not subject to the career offender 1 enhancement under ACCA” and vacated his sentence for recalculation. Id. at *3. At the resentencing hearing, the district court considered whether Scott could be sentenced under the Career Offender Guideline, U.S.S.G. §4B1.1. Relying on its analysis above, the district court determined that New York first-degree manslaughter was not a “crime of violence” under the force clause of the Career Offender Guideline for the same reasons it was not a crime of violence under ACCA’s identical force clause. The district court also rejected the government’s argument that the Career Offender Guideline applies to Scott under the enumerated-offenses clause because New York first-degree manslaughter criminalizes the same conduct as generic murder, voluntary manslaughter, or aggravated assault in the Guideline’s enumerated-offenses clause. Having decided that the Career Offender Guideline could not apply, the district court calculated Scott’s sentencing range at 121 to 130 months. At the time of the hearing, Scott had served approximately 134 months of his sentence — four months above the Guidelines range’s upper boundary. The district court therefore sentenced Scott to time served and five years of supervised release. Scott is out on supervised release. DISCUSSION On appeal, the government argues that Scott qualifies for ACCA’s 180-month mandatory minimum sentence because New York first-degree manslaughter is a crime of violence that “has as an element the use, attempted use, or threatened use of physical force against the person of another.” 18 U.S.C. §924(e)(2)(B)(i). In the alternative, the government argues that Scott should be sentenced under the Career Offender Guideline because New York first-degree manslaughter criminalizes the same conduct as the generic offenses of murder, voluntary manslaughter, or aggravated assault, which are enumerated crimes of violence in the Guideline. Scott responds that the minimum criminal conduct required to commit New York first-degree manslaughter does not have as an element the “use” of force because the crime can be committed by omission — otherwise put, by doing nothing. He also argues that a majority of states do not criminalize the conduct that New York first-degree murder penalizes as first-degree manslaughter as generic murder, voluntary manslaughter, or aggravated assault. I. New York First-Degree Manslaughter Is Not a Crime of Violence Under the Force Clause of 18 U.S.C. §924(e) ACCA sets a mandatory minimum sentence of “not less than fifteen years” for criminal defendants who have at least three prior convictions for “a violent felony or a serious drug offense.” 18 U.S.C. §924(e)(1). As relevant here, the statute defines a “violent felony” as “any crime punishable by imprisonment for a term exceeding one year” that “has as an element the use, attempted use, or threatened use of physical force against the person of another.” Id. §924(e)(1)(B)(i). The question before us is whether New York first-degree manslaughter satisfies that definition. New York Penal Law lists three criminal acts for which a person may be found “guilty of manslaughter in the first degree.” N.Y. Penal Law §125.20. First, the statute holds a person liable for first-degree manslaughter who “with intent to cause serious physical injury to another person,…causes the death of such person or of a third person.” Id. §125.20(1). Second, the statute criminalizes causing the death of another person “under the influence of extreme emotional disturbance,” id. §125.20(2). Third, the statute makes it illegal to, with intent to cause physical injury, “create[] a grave risk of serious physical injury to a person less than eleven years old” that causes their death. Id. §125.20(4). Where a statute, like New York’s first-degree manslaughter statute, “criminalize[s] multiple acts in the alternative,” we refer to the statute as “divisible” and apply the “modified categorical approach” to determine whether a conviction under the statute can serve as a predicate offense for a federal sentence enhancement. United States v. Jones, 878 F.3d 10, 16 (2d Cir. 2017); accord Descamps v. United States, 570 U.S. 254, 261-62 (2013); see also United States v. Castillo, 896 F.3d 141, 149-50 (2d Cir. 2018) (determining New York Penal Law §125.20 is “a state statute [that] criminalizes multiple acts in the alternative” and thus requires the modified categorical approach (alterations and internal quotations marks omitted)). Under that approach, we first look to “a limited class of documents” to determine of which act the defendant was convicted. Id. “[L]ook[ing] only to the statutory definitions — i.e., the elements — of [that portion of the statute], and not to the particular underlying facts,” Hill, 890 F.3d at 55 (alterations and internal quotation marks omitted), we then ask whether those elements can be satisfied by conduct that might reach beyond the parameters of ACCA. If so, we ask whether there is “a realistic probability, not a theoretical possibility,” that the statute would be applied in that manner. Id. at 56. Unless “the state statute…on its face extends to conduct beyond the definition of the corresponding federal offense,” see Hylton v. Sessions, 897 F.3d 57, 63 (2d Cir. 2018), we require that in either the defendant’s case or another case “the courts in fact did apply the statute” in that manner, Hill, 890 F.3d at 56. The proposition that a state statute can apply to conduct beyond the application of ACCA cannot be based on mere “flights of fancy” or “the application of legal imagination.” Id. at 56. With those realistic readings of the statute in mind, we then determine whether “the state statute sweeps more broadly” than ACCA, in which case a conviction under the state statute cannot serve as an ACCA predicate. Stuckey v. United States, 878 F.3d 62, 67 (2d Cir. 2017) (internal quotation marks omitted). A statute that penalizes activity that does not require the use of force thus “cannot count as a predicate ‘violent felony’ [under the force clause] for the ACCA’s fifteen-year mandatory minimum.” Id. The government has provided two certificates of disposition establishing that both of Scott’s manslaughter convictions were under New York Penal Law §125.20(1), which penalizes a person who “[w]ith intent to cause serious physical injury to another person,…causes the death of such person or of a third person.” We therefore consider only Section of New York Penal Law §125.20, and ask whether the minimum criminal conduct required for conviction under this provision falls within ACCA. A. The Minimum Criminal Conduct that New York First-Degree Manslaughter Requires Is an Omission Despite a Duty to Act The district court determined that the minimum criminal conduct required for a defendant to be held liable under New York Penal Law §125.20(1) is a failure to act in the face of a duty to do so, with the intent to cause serious physical injury, which failure causes death. The government argues that the district court’s conclusion was erroneous because criminal liability for New York first-degree manslaughter attaches only when a person is complicit in the commission of an act that sets harm into motion and then fails to act to stop that harm from killing another. We reject the government’s strained argument because the New York Court of Appeals has explicitly held that New York first-degree manslaughter can be committed by omission. The New York Court of Appeals has clearly and emphatically asserted on two occasions that New York first-degree manslaughter may be committed by a defendant’s failure to act. First, in People v. Steinberg, 29 N.Y.2d 673 (1992), the Court of Appeals considered whether the first-degree manslaughter conviction of a father who had both physically abused his six-year-old daughter and then failed to secure medical care for her resultant serious injuries, which ultimately caused her death, was sustainable on either basis. New York’s highest court “found no defect in the prosecution’s legal theory” that the “defendant performed both acts of commission (striking [the daughter]) and acts of omission (failure to obtain medical care), each with intent to cause serious physical injury, and that such acts caused [the daughter's] death.” Id. at 680-81. In so finding, the Court of Appeals held that Steinberg’s “acts of omission” (the failure to secure medical care) would have been sufficient to sustain the manslaughter conviction, even if he had not inflicted the injuries that required medical attention. Id. Steinberg had argued in his defense that “failure to obtain medical care for a child cannot, as a matter of law, support the mens rea element of first degree manslaughter…unless defendant has medical expertise, and would thereby know that serious injury will result from a lack of medical attention.” Id. at 680. The court rejected that argument. It explained: “The Penal Law provides that criminal liability may be based on an omission, which is defined as the failure to perform a legally imposed duty. Parents have a nondelegable affirmative duty to provide their children with adequate medical care. Thus, a parent’s failure to fulfill that duty can form the basis of a homicide charge.” Id. (emphasis added) (citations omitted). While acknowledging that it relied upon cases involving prosecutions for reckless manslaughter and criminally negligent homicide in reaching this conclusion, the court repeated, “the failure to obtain medical care can also support a first degree manslaughter charge, so long as there is sufficient proof of the requisite mens rea — intent to cause serious physical injury.” Id. The Court of Appeals reaffirmed this conclusion in People v. Wong, 81 N.Y.2d 600 (1993). The prosecution in Wong charged an infant’s two caretakers, Eugene and Mary Wong, with first-degree and second-degree manslaughter. Id. at 606. There, the infant died from shaken baby syndrome after one of the Wongs shook the infant and both of the Wongs subsequently failed to seek medical care for the infant. Id. The prosecution contended that even though only one of the defendants had shaken the child, both defendants were liable for first-degree manslaughter. Id. The prosecution could not prove which of the two defendants had shaken the infant and, for want of this proof, argued that a defendant who had no role in the harmful act could be held liable for first-degree manslaughter based on a failure to act. The defendants were convicted on all counts, and the Appellate Division reversed the convictions for first-degree manslaughter as against the weight of the evidence on the issue of intent to cause serious bodily injury. As to the second-degree manslaughter convictions, the Court of Appeals made clear “that the People’s theory against the ‘passive’ defendant is legally sound” because in Steinberg, it had “held that parents have an affirmative duty to provide their children with adequate medical care and that, under certain circumstances, the failure to perform that duty can form the basis of a homicide charge.” Id. at 607. That is, “a person in the position of the ‘passive’ defendant here may be held criminally liable for failing to seek emergency medical aid for a seriously injured child.” Id. at 608. The court ultimately reversed the defendants’ convictions because of “the absence of proof from which the jury could infer that the ‘passive’ defendant was even aware that the infant had been violently shaken,” and therefore needed medical care. Id. at 609. The Government argues that, notwithstanding the explicit statement of the Court of Appeals in Steinberg that “the failure to obtain medical care can…support a first degree manslaughter charge,” 79 N.Y.2d at 680, that case “does not stand for the proposition that a pure omission without any active use of force would suffice to ground a conviction under Section 125.20(1).” Appellant’s Br. at 26. It imports great significance to the court’s phrase: “so long as there is sufficient proof of the requisite mens rea.” The Government contends that an affirmative act of violence — such as defendant Steinberg’s striking the victim prior to his withholding of medical care — is necessary to establish the mens rea of intent to cause serious physical harm. It thus argues that, notwithstanding the court’s statement that failure to obtain medical care is sufficient for a conviction of first-degree manslaughter, the need to establish mens rea effectively requires a preceding act of violence on the part of the defendant. Judge Raggi argues for similar reasons that the Steinberg defendant’s acts of violence and acts of omission were “inextricably linked,” so that the case cannot establish a realistic probability that New York courts would apply the manslaughter statute to acts of omission alone. We disagree. A defendant’s intent to cause serious physical harm to the victim can be established in many ways other than that defendant’s preceding act of violence, including most obviously the defendant’s having observed (or having been told of) the infliction of violence on the child by another, or the defendant’s statement of intention to inflict harm. Although the facts in Steinberg happened to include a preceding act of violence by the defendant, the court’s independent analysis of inaction makes clear that a failure to act would be alone sufficient to sustain a conviction where the required mens rea was shown. Nothing in that discussion suggests the defendant’s inaction must be linked to that defendant’s affirmative act of violence; indeed everything in that discussion supports the contrary. The Government and Judge Raggi also argue that, because the defendants’ convictions in Wong were reversed due to insufficiency of evidence of either defendant’s awareness of the fact that the child had been shaken, the court’s discussion of the sufficiency of inaction was dictum, and therefore the case cannot show a realistic probability that a New York conviction for first-degree manslaughter based on omission alone would be sustained. The argument that Wong’s dictum is inadequate to show “realistic probability” is both irrelevant (as the proposition was a holding in Steinberg) and exaggerates the meaning of our precedents requiring a “realistic probability” of application of a statute beyond the scope of ACCA. We have stated that a “realistic probability” requires that “the courts did in fact apply” the statute in the manner considered, as opposed to a mere “flight of fancy” or explanation of “the legal imagination.” Hill, 890 F.3d at 56. For the proposition that New York courts would apply the first-degree manslaughter statute to inaction in the face of a duty to act, we do not rely on a flight of fancy or of legal imagination. We rely on the carefully considered and subsequently repeated assertion of that proposition by the highest court of New York. In Steinberg, the Court so held, explicitly sustaining the defendant’s conviction on the basis of inaction (notwithstanding the court’s recognition that the conviction would also be sustained on another basis). In Wong, furthermore, the Court of Appeals described its prior statement in Steinberg as a holding. In view of the Steinberg holding by the state’s highest court, coupled with the Court’s emphatic reaffirmance of the same proposition in Wong, it appears to us that a New York trial court would be compelled to conclude that first-degree manslaughter can be satisfied by inaction, and we are compelled in this case to construe New York law following what New York’s highest court has explicitly held. Wong additionally contradicts Judge Raggi’s assertion that no defendant has been “actually prosecuted” under an omission theory. As we noted earlier, the government in Wong “argued…that the People’s case rested on the theory that each defendant was independently liable for [the child's] death because one of them had shaken the baby while the other had stood by and failed to intervene.” 81 N.Y.S.2d at 606. Wong thus establishes not only that a defendant may be convicted for first-degree manslaughter based on omission, but also that New York has actually prosecuted defendants under this theory. In short, these two decisions from New York’s highest court are more than sufficient to establish a realistic probability that New York first-degree manslaughter could be applied to a defendant who intentionally causes death by an act of omission in the face of a duty to protect the victim from a perceived harm. New York’s highest court “did in fact apply” the first-degree manslaughter statute to the facts of a particular case. See Hill, 890 F.3d at 56. Judge Raggi acknowledges that the standard is whether there is a “realistic probability” that the courts would accept omission as the basis of a first-degree manslaughter conviction. Hill, 890 F.3d at 56. That standard requires that the courts “in fact did [so] apply the statute.” Id. Nonetheless, in her reasoning, she converts that standard into whether the courts of New York have in fact affirmed a manslaughter conviction based on omission. Notwithstanding that in both Steinberg and Wong, cases in which the manslaughter charge was (as an alternative) predicated on omission, New York’s highest court expressly stated that omission alone would support liability, Judge Raggi asserts that there is no such realistic probability because in neither case did the Court of Appeals affirm a conviction predicated on inaction without complicity in any action. By doing so, Judge Raggi has effectively converted the requirement of “realistic probability” to a requirement of actual affirmance of a manslaughter conviction based on omission alone, a reading that has no basis in our case law.² Accordingly, we consider whether a crime committed by omission is a violent felony under ACCA’s force clause.³ B. New York First-Degree Manslaughter Sweeps More Broadly than 18 U.S.C. §924(e) We must decide whether a crime that can be committed by omission “has as an element the use, attempted use, or threatened use of physical force against the person of another.” 18 U.S.C. §924(e)(2)(B)(i). The government, relying on United States v. Castleman, 572 U.S. 157 (2014), argues that where a defendant’s omission causes a serious injury, force is implicit in the inaction and thus the omission constitutes a use of force. Scott counters that ACCA’s text requires an active deployment of force that an omission, by nature, cannot effectuate and that Castleman cannot be read to hold that a defendant who commits nothing more than a culpable omission (without any affirmative act) uses force. The Supreme Court articulated the standard for what constitutes the use of “physical force” sufficient to satisfy Section 924(e) in Curtis Johnson v. United States, 559 U.S. 133. Contrary to the position stated by Justice Alito in dissent that “ACCA uses ‘violent felony’ as a term of art with a wider meaning than the phrase may convey in ordinary usage,” see id. at 148 (Alito, J., dissenting), the majority of the Supreme Court concluded that the term “physical force” is to be given its “ordinary meaning,” id. at 138 (majority opinion). The Court explained that “force” means “strength or energy; active power; vigor; often an unusual degree of strength or energy[;] power to affect strongly in physical relations;…power, violence, compulsion, or constraint exerted upon a person” and that “physical force” means “power, violence, or pressure directed against a person or thing[;]…force consisting in a physical act, esp. a violent act directed against a robbery victim.” Id. at 139 (emphasis added) (alterations and internal quotation marks omitted). The Court distilled the term’s ordinary meaning into a now-indelible phrase: “violent force — that is, force capable of causing physical pain or injury to another person.” Id. at 140. The Court’s reliance on the ordinary meaning to shape its conception of “physical force” communicates that only an active crime constitutes a “violent felony” under ACCA. See also Stokeling v. United States, 139 S. Ct. 544, 553 (2019) (noting that a conviction must be “within the ‘category of violent, active crimes’ that Congress included in ACCA”). The Government and Judge Raggi argue that the Supreme Court’s subsequent decision in Castleman, 572 U.S. 157, as well as our later decision in Villanueva, 893 F.3d 123 (2d Cir. 2018), compel the conclusion that a crime of manslaughter that may be committed by a failure to act necessarily involves the “use of force” as defined under the ACCA force clause. They rely principally on Castleman’s statement that “the knowing or intentional causation of bodily injury necessarily involves the use of physical force,” 572 U.S. at 169, from which they conclude that a defendant who causes death by inaction necessarily uses force. What was said in Castleman and Villanueva must be considered in the light of the facts of those cases. In Castleman, the Supreme Court considered whether a Tennessee statute, which made it a crime to “intentionally, knowingly or recklessly cause bodily injury to another” qualified as a misdemeanor crime of domestic violence under a different statute, 18 U.S.C. §921(a)(33)(A)(ii), because it involved “the use or attempted The defendant argued, as relevant here, that the offense did not categorically involve “the use or attempted use of physical force” because it could be accomplished through the indirect application of force, such as by sprinkling poison into a beverage. The Supreme Court rejected that argument. First, it concluded that, in contrast to the ACCA definition of force established in Curtis Johnson, the “use of physical force” under §921(a)(33)(A) is satisfied by “the degree of force that supports a common-law battery conviction.” 572 U.S. at 168.⁴ Second, the Court concluded that this definition of force “encompasses even its indirect application.” Id. at 170. The Court explained that the “use of force” in the poisoning example “is not the act of sprinkling the poison; it is the act of employing poison knowingly as a device to cause physical harm. That the harm occurs indirectly, rather than directly (as with a kick or punch), does not matter.” Id. at 171. Similarly, shooting a victim with a gun is no less a use of force “because it is the bullet, not the trigger, that actually strikes the victim.” Id. Later, in Villanueva, 893 F.3d 123, this Circuit considered whether an offense of first-degree assault that could be accomplished by poisoning qualified as a violent felony under the ACCA force clause. Extending the Supreme Court’s reasoning in Castleman to the context of ACCA, we concluded the statute did qualify as a violent felony. We found persuasive Castleman’s “explanation of why sprinkling poison constitutes use of physical force even if the offender does not hit or even touch the victim,” noting the example of pulling the trigger of a gun. Id. at 129; cf. Hill, 832 F.3d at 144 (concluding in light of Castleman that Hobbs Act robbery is a crime of violence under §924(c)(3)(A), notwithstanding that it may be committed by indirect application of force). Accordingly, both Castleman and Villanueva were addressed to the question whether a low level of force that produces a violent result can qualify as use of force under ACCA. Neither addressed, much less resolved, the question before us, namely, whether the failure to act — by one who has a duty to act to protect the victim — constitutes a “use of physical force” under ACCA.⁵ See 18 U.S.C. §924(e)(2)(B)(i); cf. United States v. Mayo, 901 F.3d 218, 228-29 (3d Cir. 2018) (recognizing that neither Castleman, nor subsequent Third Circuit case law applying Castleman to the ACCA context, addressed the issue of crime by omission). Unlike New York first-degree manslaughter, the acts discussed in Castleman and Villanueva require some action that initiates a harmful consequence. By contrast, a defendant who commits a crime by omission definitionally takes no action and thus initiates nothing. This might occur, for example, when the mandatory caretaker defendant observes that the victim has, entirely through natural causes, entered into a condition that urgently requires administration of medical help and, with intention to harm the victim, the defendant fails to summon such help. As we read Castleman and Johnson, this scenario would not constitute ACCA’s required ” use of physical force.” Judge Raggi also argues that we conflate “physical force” itself with the “use” of such force. It is true, as Judge Raggi notes, that the Supreme Court said in Castleman that “use” “conveys the idea that the thing being used (here, ‘physical force’) has been made the user’s instrument.” Castleman, 572 U.S. at 170- 71. She further notes that, in the example of a person who kills his victim by poison, the Supreme Court explained that the “use of force” is not the “act of sprinkling the poison; it is the act of employing poison knowingly as a device to cause physical harm.” Id. at 171 (internal quotation marks and brackets omitted) (emphasis added). Under Castleman’s reading of the term “use,” Judge Raggi argues, because someone guilty of New York first-degree manslaughter must have intended to cause at least serious physical injury, he “must be said to have intentionally used the force causing death as his instrument without regard to whether he did so by commission or omission.” To the extent that Judge Raggi argues that Castleman compels her reading, she overstates the holding of that case, which explicitly did not reach the question whether “causation of bodily injury necessarily entails violent force [for ACCA purposes].” Id. at 167. As to the poison example, the Supreme Court’s analysis does not compel Judge Raggi’s conclusion. It is at most ambiguous. There, the Court explained that the use of force was “the act of employing poison.” Id. at 171 (emphasis added). It did not contemplate a situation in which a person stands by, taking no action whatsoever, and allows a physical harm to occur with the intent to cause serious physical injury, and thus it cannot be said to have answered the question whether such a situation constitutes “use” of the injury-causing force. We conclude that a crime that may be committed by complete inaction does not have “as an element the use…of physical force against the person of another,” under the meaning of ACCA. See 18 U.S.C. §924(e)(B)(i). We consider the example of a person having a legal duty to provide care to another who, with intent to cause serious physical harm, fails to seek medical help when his ward requires it. Notwithstanding the Government’s and Judge Raggi’s respectable argument that, by availing himself or taking advantage of the preexisting force that affected the ward, that the person did “use…physical force,” that conclusion does not, as Judge Raggi suggests, follow from Castleman. Judge Raggi further argues that it “makes no sense” to distinguish between a defendant who, “intent on killing or seriously injuring another person…himself sprinkled [poison] in a victim’s drink,” and one who, with the same intentions and a “legal duty to intervene,…stood by and let the person imbibe what defendant knew was a poisoned drink.” If Judge Raggi’s argument depends on the notion that the two cases involve the same degree of moral culpability, we do not disagree that both defendants are equally morally culpable. Perhaps a perfectly constructed criminal law would treat them identically. But what determines whether two defendants who have acted differently but with similar degrees of moral culpability will be treated the same or differently is the language of the statutes that cover their cases. Notwithstanding Judge Raggi’s reasonable argument that a few words in Castleman could be read broadly to mean that “use of violent force” includes doing nothing, neither the statute nor the Supreme Court’s interpretation of it clearly communicates such a meaning. In addition, to the extent that it may be seen as a close question whether “use of physical force” includes crimes committed by omission, the burden is on the government to show that a prior conviction counts as a predicate offense for the purpose of an ACCA sentence enhancement. See United States v. Savage, 542 F.3d 959, 964 (2d Cir. 2008). Moreover, the rule of lenity, which “requires ambiguous criminal laws to be interpreted in favor of the defendants subject to them,” United States v. Santos, 553 U.S. 507, 514 (2008), requires construing that ambiguity in the defendant’s favor. See also Johnson v. United States, 529 U.S. 695, 713 n.13 (2000) (noting that lenity applies “when the equipoise of competing reasons cannot otherwise be resolved”). We therefore hold that New York first-degree manslaughter is not a crime of violence under the force clause of ACCA because it can be committed by inaction, while the ordinary meaning of the terms of ACCA are not satisfied by inaction. The district court’s vacatur of Scott’s ACCA sentence under 18 U.S.C. §§922(g)(1) and 924(e) is thus affirmed. II. A Prior Conviction for New York First-Degree Manslaughter Is Not a Crime of Violence Pursuant to the Career Offender Guideline The government also appeals the district court’s interpretation of the Career Offender Guideline as not applicable to New York first-degree manslaughter. The Career Offender Guideline considers as a “career offender” a defendant who at eighteen years of age or more committed an offense of conviction “that is a felony [and is] either a crime of violence or a controlled substance offense,” and “ha[s] at least two prior felony convictions of either a crime of violence or a controlled substance offense.” U.S.S.G. §4B1.1(a). Classification as a career offender typically results in a significantly higher Guidelines sentencing range than would otherwise apply. The Career Offender Guideline defines a “crime of violence” as: any offense under federal or state law, punishable by imprisonment for a term exceeding one year, that — (1) has as an element the use, attempted use, or threatened use of physical force against the person of another, or (2) is murder, voluntary manslaughter, kidnapping, aggravated assault, a forcible sex offense, robbery, arson, extortion, or the use or unlawful possession of a firearm described in 26 U.S.C. §5845(a) or explosive material as defined in 18 U.S.C. §841(c). Id. §4B1.2(a). The first clause of the Guideline, the “force clause,” is identical to the force clause in ACCA. Compare id. (defining “crime of violence” as involving “use, attempted use, or threatened use of physical force against the person of another”), with 18 U.S.C. §924(e)(2)(B)(i) (defining “violent felony” as involving “use, attempted use, or threatened use of physical force against the person of another”). Accordingly, our analysis of whether New York first-degree manslaughter comes within the definition of a crime of violence under ACCA’s force clause is pertinent in answering whether the offense is a crime of violence under the force clause of the Career Offender Guideline. The Supreme Court in Curtis Johnson specified that in interpreting ACCA, the phrase “physical force” carries its “ordinary meaning.” See 559 U.S. at 138. We see no reason why the same phrase in the Career Offender Guideline should not also be interpreted in accordance with its ordinary meaning. We conclude that, pursuant to its ordinary meaning, the phrase “has as an element the use of physical force” similarly does not apply to an offense that may be committed by taking no action whatsoever. The Government argues that, even if the force clause does not apply, New York first-degree manslaughter is a crime of violence under the Career Offender Guideline’s “enumerated-offenses clause.” It argues that New York first-degree manslaughter matches the generic offenses of murder, voluntary manslaughter, or aggravated assault. “Where the basis for categorizing a prior conviction as a crime of violence is that the offense is specifically enumerated as such in the Career Offender Guideline or its commentary, we undertake the categorical approach by comparing the state statute to the generic definition of the offense.” Jones, 878 F.3d at 18. We determine the “generic” definition of the offense by looking to the “sense in which the term is now used in the criminal codes of most States.” Taylor v. United States, 495 U.S. 575, 598 (1990). To carry out this obligation, we may “consult other sources, including federal criminal statutes, the Model Penal Code, scholarly treatises, and legal dictionaries,” in addition to the state codes themselves. Castillo, 896 F.3d at 150 (footnotes omitted). If New York first-degree manslaughter “has the same elements” or is narrower than the generic Career Offender Guideline analog, “the prior conviction can serve as a [Career Offender] predicate.” Descamps, 570 U.S. at 261. If, however, the state statute “sweeps more broadly than the generic crime,” the crime of conviction cannot serve as a Career Offender predicate. Id. We apply this exhaustive approach to determine whether New York first-degree manslaughter is the equivalent of or narrower than the generic crimes of murder, voluntary manslaughter, or aggravated assault. A. New York First-Degree Manslaughter Is Not Generic Voluntary Manslaughter As the government concedes, only a minority of states penalize the conduct of New York first-degree manslaughter as “murder” or “manslaughter.” The government nonetheless raises the novel argument that the offense qualifies as an enumerated crime of violence because the conduct it describes “is penalized in a majority of states as either ‘murder’ or ‘voluntary manslaughter,’” and those states can be aggregated to form a majority. Appellant’s Br. at 41 (emphasis added). It clarifies in its reply brief that its claim is that those states that penalize this conduct as “murder” should be counted as states in which the New York statute fits the “lesser crime” of manslaughter. The government identifies no authority that has interpreted the enumerated offenses clause in this manner. In support, it cites to Taylor for the proposition that “courts should disregard…’labels’ in applying the categorical approach.” Id. at 47 (citing Taylor, 495 U.S. at 590). It further notes that in Taylor, the Court held that the generic definition of burglary encompasses both “aggravated burglaries” and “run-of-the-mill burglaries,” and argues that here, murder should be treated as an aggravated form of “voluntary manslaughter.” Appellant’s Reply Br. at 17-18. We find this “aggregation” argument unpersuasive. It may be true that the conduct in the New York statute “is penalized in a majority of states as either ‘murder’ or ‘voluntary manslaughter.’” Appellant’s Br. at 41. Nonetheless, according to the government’s own account, a majority of States (30) do not penalize that conduct as “murder,” and a majority of States (42) do not penalize that conduct as “voluntary manslaughter.” Under our case law, the statute does not fit any of the generic crimes. See Castillo, 896 F.3d at 152. To the extent that the government attempts to find support in Taylor for its novel approach, that reliance is misplaced. Far from holding that “courts should disregard…’labels’ in applying the categorical approach,” Appellant’s Br. at 47, Taylor held that an enumerated offense in Section 924(e) refers to “the generic sense in which the term is now used in the criminal codes of most States.” 495 U.S. at 598. Moreover, an interpretation of “voluntary manslaughter” that includes States in which the State offense would qualify as murder, on the basis that “voluntary manslaughter” is a lesser-included crime of murder, would render entirely superfluous Congress’s inclusion of the separate crime of “murder,” because such states would necessarily count as states in which the offense in question qualifies as “voluntary manslaughter.” See Duncan v. Walker, 533 U.S. 167, 174 (2001) (“It is our duty to give effect, if possible, to every clause and word of a statute.” (internal quotation marks omitted)); see also Marx v. Gen. Revenue Corp., 578 U.S. 371, 386 (2013) (“[T]he canon against surplusage is strongest when an interpretation would render superfluous another part of the same statutory scheme.”). The government’s analogy of murder to the inclusion of “aggravated burglary” under “burglary” is thus inapt, among other potential reasons, because Congress did not include a separate crime of “aggravated burglary” in addition to “burglary,” whereas it did include the crime “murder” in addition to “voluntary manslaughter.” For these reasons, we reject the government’s proposed “aggregation” approach. In any event, it is certainly not the clear meaning of the Guideline to allow aggregation of states that treat manslaughter as one of the enumerated crimes with those that treat manslaughter as another. The rule of lenity therefore bars adding punitive consequences based on an interpretation that is not clearly communicated. The government makes two additional arguments as to why the New York statute fits the generic crime of voluntary manslaughter, notwithstanding the fact that it identifies only eight states in which its mens rea requirement of “intent to do serious injury” is sufficient. First, the government points to a treatise stating, “The usual view of voluntary manslaughter…presupposes an intent to kill (or perhaps an intent to do serious injury or to engage in very reckless conduct).” LaFave, Substantive Crim. L. §15.2(a)) (emphasis added). We do not find that an equivocal (“perhaps”) statement in a law treatise, supported by a citation to only one state court case addressing the law of only that state, to be sufficient to carry the government’s burden in demonstrating that the New York statute fits the generic crime of voluntary manslaughter. Moreover, although there is no requirement that the government expressly identify a majority of states in order to establish the meaning of a generic crime, and although courts may consult sources other than state codes including “scholarly treatises” in doing so, see Castillo, 896 F.3d at 150, we find the government’s reliance on this treatise particularly unavailing in light of the fact that it has undertaken an apparently exhaustive attempt to identify state criminal codes under which the New York statute would qualify as “voluntary manslaughter,” and yet came up only with a small minority. Second, the government points to our decision in Castillo, in which we held that “the generic definition of ‘manslaughter’ includes the unlawful killing of another human being recklessly.” Id. at 152. We explained: [C]ommon law manslaughter…served as “a sort of catch-all category…[for] homicides which are not bad enough to be murder but which are too bad to be no crime whatever. Manslaughter was later subdivided into voluntary and involuntary varieties. Voluntary manslaughter was the intentional killing “in a heat of passion upon adequate provocation,” and involuntary manslaughter was an unintentional killing caused by “criminal negligence” or “recklessness,” or during the commission of an unlawful act not amounting to a felony. Id. at 151 (citing LaFave, Substantive Crim. L. §§15.1, 15.4). Relying on this passage, the government argues that the New York statute qualifies as voluntary manslaughter because, “Plainly, a violation of Section 125.20(1)…is not ‘unintentional killing caused by criminal negligence or recklessness.’” In making this statement, the government seems to assume that because the New York statute is not involuntary manslaughter, it must be voluntary manslaughter. We are not persuaded. It could equally be said that, plainly, Section 125.20(1) — which criminalizes causing death with the intent to cause serious physical injury — includes conduct beyond “intentional killing ‘in a heat of passion upon adequate provocation.’” See Castillo, 896 F.3d at 151. It is clear that the New York statute covers circumstances beyond provocation. Reliance on this passage in Castillo is thus unavailing. We conclude that the government has failed to carry its burden in showing that Scott’s conviction fits the generic crime of voluntary manslaughter. B. New York First-Degree Manslaughter Is Not Generic Aggravated Assault The government also argues that New York first-degree manslaughter penalizes conduct that would be generic aggravated assault in “[v]irtually every state.” Appellant’s Br. at 49. We reject this argument because the government has not demonstrated that most states would penalize an omission as assault. “The government bears the burden of showing that a prior conviction counts as a predicate offense for the purpose of a sentencing enhancement.” United States v. Savage, 542 F.3d 959, 964 (2d Cir. 2008). Here, however, the government has not identified a single state that has actually punished a defendant’s failure to act as aggravated assault. Instead, the government relies on three sources to establish that a failure to act could be so punished: the treatise American Jurisprudence, the Model Penal Code, and LaFave’s treatise. These three sources do not carry the government’s burden. The government’s citations to American Jurisprudence and the Model Penal Code are inapposite because they do not address the question of whether an omission may be punished as assault. The portion of American Jurisprudence upon which the government relies concerns only the sufficiency of an indirect application of force to qualify as battery and has no bearing on whether battery can be committed by omission. 6 Am. Jur. 2d Assault & Battery §6 (“The force used in a common law battery need not be applied directly to the body of the victim.”). The government then points this Court to the Model Penal Code for the general proposition that a person can be held liable for a culpable omission, but it fails to explain how these provisions constitute a statement that battery can be committed by omission. See Model Penal Code §1.13 (defining “conduct,” in relevant part, as “an action or omission and its accompanying state of mind”); id. §2.01 (requiring liability to be “based on conduct that includes a voluntary act or the omission to perform an act of which [a person] is physically capable”).⁶ These provisions provide little guidance on whether a majority of states would criminalize a failure to act as aggravated assault. Lastly, the government relies on LaFave to support its argument that a majority of states would criminalize New York first-degree manslaughter as aggravated assault. Judge Raggi, in support of this argument, points to a few cases cited by LaFave that involve assaults by omission. Our directive under Taylor v. United States is to look to “the generic sense in which the term is now used in the criminal codes of most States.” 495 U.S. at 598 (emphasis added). The few cases cited by Judge Raggi fall far short of showing that assault may be committed by omission in most states. The treatise therefore has limited persuasive value both because it cites only a few examples and because it is the sole source of support for the government’s contention.⁷ Given the government’s failure to cite any state code that criminalizes an omission as aggravated assault, we decline to rest our decision today on a treatise that cites only a few cases of assault by omission. We hold that New York first-degree manslaughter does not fit the generic definitions of murder, voluntary manslaughter, or aggravated assault, and we therefore affirm the district court’s conclusion that Scott cannot be sentenced under the Career Offender Sentencing Guideline. CONCLUSION The State of New York has prosecuted New York first-degree manslaughter based on a defendant’s failure to act. We hold that because a defendant can commit New York first-degree manslaughter without taking any action, a defendant so convicted cannot be said to have used force sufficient to trigger ACCA’s sentencing enhancements or the force clause of the Career Offender Guideline. We also hold that New York first-degree manslaughter is not a match to the generic offenses of murder, voluntary manslaughter, or aggravated assault and therefore does not satisfy the en umerated-offenses clause of the Career Offender Guideline. The order and judgment of the district court is therefore AFFIRMED. LEVAL, Circuit Judge, Concurring: I join in Judge Pooler’s persuasive opinion for the Court holding that New York first-degree manslaughter does not fall within ACCA because it extends more broadly than ACCA’s definition of violent felony. In this opinion, in which Judge Pooler joins, I offer an additional reason, relying on application of the rule of lenity by reason of lack of clarity whether ACCA’s requirement of use of physical force can be satisfied — and ACCA’s mandatory 15-year sentence imposed — where the defendant has taken no action whatsoever. The rule of lenity “requires ambiguous criminal laws to be interpreted in favor of the defendants subjected to them.” United States v. Santos, 553 U.S. 507, 514 (2008); see also United States v. Universal C.I.T. Credit Corp., 344 U.S. 218, 221-22 (1952) (“[W]hen choice has to be made between two readings of what conduct Congress has made a crime, it is appropriate, before we choose the harsher alternative, to require that Congress should have spoken in language that is clear and definite.”). That rule, which is “‘perhaps not much less old than’ the task of statutory ‘construction itself,’” United States v. Davis, 139 S. Ct. 2319, 2333 (2019) (quoting United States v. Wiltberger, 18 U.S. (5 Wheat.) 76, 95 (1820) (Marshall, C.J.)), is founded, at least in part, on the notion that “a fair warning should be given to the world in language that the common world will understand, of what the law intends to do if a certain line is passed. To make the warning fair, so far as possible the line should be clear.” United States v. Bass, 404 U.S. 336, 348 (1971) (quoting McBoyle v. United States, 283 U.S. 25, 27 (1931) (Holmes, J.)). Mandatory sentencing statutes such as ACCA, when their applicability is not clearly commanded by the words of the statute, present a particularly compelling ground for the application of the rule of lenity. This is because mandatory sentencing statutes (for reasons explained below) rarely serve a useful purpose, and frequently serve as arbitrary instruments of injustice. Given the overriding objectives of sentencing to provide appropriate protection to the public with fairness to the defendant and the victim, sentences must be imposed on the basis of the relevant facts of the case. The “purposes” of sentencing are specified in the basic federal sentencing statute, 18 U.S.C. §3553(a). It requires the court to consider “the nature and circumstances of the offense and the history and characteristics of the defendant,” id. §3553(a)(1), as well as “the need for the sentence imposed…to reflect the seriousness of the offense[;] to promote respect for the law[;] to provide just punishment for the offense;…[t]o afford adequate deterrence to criminal conduct;…[and] to protect the public from further crimes of the defendant,” id. §3553(a)(2). The statute goes on, in what is called the “parsimony clause,” to command the courts to impose a sentence that is “sufficient, but not greater than necessary, to comply with the purposes” of sentencing. All of this requires that the sentencing court, and the reviewing court, give careful attention to the facts that characterize both the commission of the offense and the particular defendant. The task of assuring that the sentencing court imposes a sufficient sentence is assigned to the reviewing court. If the sentencing judge imposes a sentence that is not “sufficient…to comply with” the purposes of sentencing, that sentence would violate the statute and should be set aside on appeal. When a legislature imposes a mandatory sentence, all consideration of the relevant facts disappears. Necessarily knowing nothing of the facts — neither of the crime nor the victim nor the defendant (as the crime has not yet occurred) — the legislature, by requiring mandatory sentences, arbitrarily commands that all crimes falling within the broadly described category be sentenced as specified, regardless of whether the particular facts can reasonably justify such a sentence. No doubt, mandatory sentences are most often legislated for circumstances in which the majority of occurrences will appropriately call for a sentence of such severity. Because the need for a severe sentence in most of the covered cases will be as evident to courts as it is to the legislature, in most such cases the mandatory requirement serves no purpose. The courts would impose a sentence of comparable severity as the just resolution of the case, regardless of whether the legislature had imposed a mandatory sentence, precisely because such a sentence would be well justified by the facts. And in the relatively rare case where the sentencing judge failed to impose an appropriately severe sentence that complied with the purposes of sentencing, the prosecutor could appeal, and it would be the duty of the appeals court to vacate an unreasonably lenient sentence, requiring a sentence that is sufficiently severe. On the other hand, notwithstanding that the sentence required by mandatory sentencing laws is appropriate in most circumstances, it does not follow that the mandatory sentence will be just or even reasonable in all circumstances. Almost invariably, the standards that govern the imposition of mandatory sentences are crude and generalized, and lack the fine distinctions that are crucial to fair sentencing. Inevitably mandatory sentences will govern cases where sentences of such severity cannot be reasonably justified. Thus, when the sentence mandated is reasonably called for in the majority of cases it covers, the main consequence of the legislature’s making that sentence mandatory is to require substantial injustice in the minority of cases for which a sentence of such severity cannot be justified.⁸ A 15-year sentence, as mandated by ACCA, for one who possesses a firearm after three convictions for either a violent felony or a felony drug offense does not on its face sound unreasonable. The court would reliably impose such a sentence on most persons who fall into that category. That is not to say, however, that such a sentence would be reasonable for everyone who is ACCA-eligible. A foolish 16-year-old, who yielded to the temptation of serving as a runner or a lookout for drug sellers in an urban drug marketplace, would have committed numerous felony drug offenses in the course of a single day, and could easily have accumulated three separate felony drug convictions in a period of a few days. Should it matter if those convictions were 40 years in the past, and if the foolish teenager had thereafter rehabilitated himself and lived an exemplary, productive, and useful life for 40 years before being in possession of a gun? Should it matter whether the subsequent illegal gun possession occurred in relatively innocuous circumstances? Suppose for example that, 40 years after his drug convictions, he acquired a rifle because he lived on a farm and his family or livestock were menaced by coyotes. Should it matter if, on his father’s death many years after his early disposition toward crime, he simply inherited his father’s property, which he knew included a hunting rifle? Probably a violation. But warranting 15 years of imprisonment? Mandatory minimum sentencing statutes do not make allowance for the inevitable circumstances of an individual case that would make the commanded sentences outrageously unjust. Notwithstanding that they rarely serve a useful purpose and almost inevitably cause appalling injustices in some cases, mandatory sentences are tolerated almost without comment because there is no voice or constituency in our society for convicted felons who are the victims of unjust sentencing laws. No member of Congress ever lost votes for supporting unfairly harsh sentences. Nonetheless, such sentences represent a serious injustice — one that we ought to consider intolerable. For these reasons, mandatory sentences are inevitably at times in tension with the parsimony clause of 18 U.S.C. §3553(a), which requires that a district judge impose a sentence that is “not greater than necessary” to satisfy the goals of sentencing. See United States v. Dorvee, 616 F.3d 174, 182-83 (2d Cir. 2010); cf. United States v. Ministro-Tapia, 470 F.3d 137, 141 (2d Cir. 2006) (“Plainly, if a district court were explicitly to conclude that two sentences equally served the statutory purpose of §3553, it could not, consistent with the parsimony clause, impose the higher.”). I do not suggest that Congress lacks the power to impose mandatory sentences that will inevitably inflict injustice. Nonetheless, under the rule of lenity, statutes imposing criminal sanctions are not enforced unless they give clear notice of their applicability. When the rule in question is a harsh mandatory sentence, which, notwithstanding reasonable application to many circumstances, will cause serious injustices in others, courts have a particularly important duty to see to it that the rule is not enforced beyond the circumstances to which it unambiguously applies. Even if one is persuaded by the government’s and Judge Raggi’s imaginative arguments that Congress probably intended that ACCA’s requirement of a “violent felony” involving “use of physical force” be satisfied in cases where the defendant has done nothing at all, these arguments fall far short of showing that the reach of the statute unambiguously commands this. I conclude for the reasons persuasively stated by Judge Pooler and those expressed in this opinion that ACCA’s mandatory 15-year sentence does not apply to cases where the defendant has taken no action whatever. Although the case is less clear as to the inapplicability of the Career Offender Guideline, which, since United States v. Booker, 543 U.S. 220 (2005), Gall v. United States, 552 U.S. 38 (2007), and Kimbrough v. United States, 552 U.S. 85 (2007), has become entirely discretionary, our court has applied the rule of lenity’s requirement of clarity to the Sentencing Guidelines. United States v. Parkins, 935 F.3d 63, 66 (2d Cir. 2019). I join in Judge Pooler’s opinion on this issue as well. Even if Congress intended that the Guideline cover such statutes as New York manslaughter, the applicability of the phrase “crime of violence” to a crime that can be committed by complete inaction is at best ambiguous. REENA RAGGI, Circuit Judge, dissenting: Gerald Scott is a violent criminal, who has repeatedly threatened, and on two occasions taken, human life. The killings were undoubtedly brutal: Scott shot one of his victims in the head at point-blank range; he stabbed the other to death. For these homicide crimes, Scott stands twice convicted in New York State of first-degree manslaughter under N.Y. Penal Law §125.20(1). At issue on this appeal is whether Scott’s manslaughter convictions are for crimes of violence. An affirmative answer might appear obvious to a man on the street aware of Scott’s conduct. But precedent, as the panel majority notes, requires judges to view that conduct as “irrelevant.” Majority Op., ante at 4, n.1. Instead, the violent-crime question must be determined categorically by asking whether an element of New York first-degree manslaughter requires the defendant to have used “force capable of causing physical pain or injury to another person.” Curtis Johnson v. United States, 559 U.S. 133, 140 (2010). Following that principle here, the majority concludes that New York first-degree manslaughter is not a categorical crime of violence because — Scott’s own homicidal conduct notwithstanding — it is possible to violate §125.20(1) without a defendant himself actively deploying physical force. For example, the crime can be committed by “omission,” as when a parent, intent on seriously injuring his child, fails to seek necessary medical care for the child, thereby causing the child’s death.⁹ See Majority Op., ante at 3. Carried to its logical — or illogical — conclusion, the majority’s reasoning would preclude even intentional murder from being recognized as a categorical violent felony or crime of violence because, presumably, a person can cause death through omission whether his specific intent is to kill, see N.Y. Penal Law §125.25 (second-degree murder), or to cause serious physical injury, see id. §125.20(1) (first-degree manslaughter). Rather than start down a path leading so far from the violent reality of homicide crimes, I respectfully dissent from the majority’s conclusion that New York first-degree manslaughter is not a categorical crime of violence. Two reasons inform my decision. First, Scott has not carried his threshold burden to identify a case in which a defendant has been “actually prosecute[d]” for New York first-degree manslaughter solely on a theory of omission. Moncrieff v. Holder, 569 U.S. 184, 206 (2013). In People v. Steinberg, 79 N.Y.2d 673, 680, 584 N.Y.S.2d 770, 772 (1992), the New York Court of Appeals recognized the legal possibility of a first-degree manslaughter conviction being based solely on a theory of omission. But Steinberg itself was not such a case. It upheld the first-degree manslaughter conviction of a parent who struck his six-year old child senseless and then failed to secure lifesaving medical assistance. See id. at 678–80, 584 N.Y.S.2d at 771–72. Because Scott has not pointed to a single case in the nearly three decades following Steinberg in which a defendant has “ in fact” been prosecuted for New York first-degree manslaughter based solely on omission, I cannot conclude that he has adequately demonstrated a “realistic probability” of New York law being so applied. United States v. Hill, 890 F.3d 51, 56 (2d Cir. 2018) (internal quotation marks omitted), cert. denied 139 S. Ct. 844 (2019). Second — and more important — even assuming a probability of §125.20(1) crimes being prosecuted based only on omissions, I would hold New York first-degree manslaughter a categorical crime of violence. New York law states that to be guilty of first-degree manslaughter a person “[w]ith intent to cause serious physical injury to another person, [must] cause[] the death of such person or of a third person.” N.Y. Penal Law §125.20(1). As the Supreme Court has recognized, “the knowing or intentional causation of bodily injury necessarily involves the use of physical force.” United States v. Castleman, 572 U.S. 157, 169 (2014) (emphasis added). Thus, to satisfy the causation element of §125.20(1), a defendant who causes death while intent on causing serious physical injury must use physical force. See Villanueva v. United States, 893 F.3d 123, 128 (2d Cir. 2018) (construing Castleman to focus “force” inquiry for federal purposes “on the causation of a consequence”). It makes no difference whether a defendant does so through commission, by himself initiating that force, or through omission, by making force already in motion the “instrument” of his own injurious intent by failing to act on a duty to remediate the force. United States v. Castleman, 572 U.S. at 170–71 (holding that “word ‘use’ conveys the idea that the thing used (here, ‘physical force’) has been made the user’s instrument” (internal quotation marks omitted)); see Villanueva v. United States, 893 F.3d at 129 (explaining that “relevant force is the impact of [the injurious agent] on the victim, not the impact of the user on [that agent]“). In either circumstance, the defendant has necessarily used physical force in causing death. Thus, because New York first-degree manslaughter can be committed only by a defendant’s use of force, I would identify it as a categorical crime of violence. I. Background Whether New York first-degree manslaughter is a crime of violence is a question that here arises in the context of determining the appropriate sentence for Scott’s latest conviction, his fourth for a crime using life-threatening violence.¹⁰ Specifically, on September 26, 2006, Scott entered a Bronx jewelry store, pointed a gun at the store owner, and ordered him to surrender the contents of his cash register. The robbery, and any possible ensuing injury, were thwarted by the fortuitous intervention of a police officer. For his actions, Scott now stands federally convicted, based on his guilty plea, for attempted Hobbs Act robbery, see 18 U.S.C. §§1951, 1952; brandishing a firearm during that robbery, see id. §924(c)(1)(A)(ii); and being a previously convicted felon in possession of a firearm, see id. §§922(g)(1), 924(e). In sentencing Scott for these crimes, the district court had to consider his criminal history, particularly his history for violent crimes. Such consideration is relevant to assessing, at a minimum, the seriousness of Scott’s most recent crimes, the likelihood of his committing future crimes, and the risk of danger — life-threatening danger — to innocent persons from such crimes. See id. §3553(a)(2)(A)–(C). While the law generally proscribes any limits on the facts and circumstances that a district court may consider in determining a defendant’s sentence, see id. §3661, it sometimes requires specific findings to trigger particular sentencing consequences. Thus, to apply a “Career Offender” enhancement to a defendant’s Sentencing Guidelines calculation, a court must find that a defendant has two or more prior felony convictions for “a crime of violence or a controlled substance offense.” U.S.S.G. §4B1.1(a).¹¹ And if the court finds that a defendant found guilty of being a felon in possession of a firearm has three or more prior violent felony or serious drug convictions, the Armed Career Criminal Act (“ACCA”) mandates a minimum prison term of fifteen years for the unlawful possession. See 18 U.S.C. §924(e)(1).¹² To decide whether a crime of conviction is a violent felony or a crime of violence under the “force” clauses of ACCA and the Career Offender Guideline, a court must make a categorical assessment, asking whether an element of the crime requires the use of physical “force capable of causing physical pain or injury to another person.” Curtis Johnson v. United States, 559 U.S. at 140 (construing ACCA’s “physical force” requirement); accord United States v. Hill, 890 F.3d at 58. To answer that question, a court looks to the minimum criminal conduct necessary for conviction under the statute, without regard to whether the defendant himself engaged in more egregious conduct. See United States v. Hill, 890 F.3d at 55.¹³ At Scott’s initial sentencing, the district court concluded — with no objection and, therefore, little discussion — that Scott warranted enhancements under the force clauses of both ACCA and the Career Offender Guideline. This was based on his two 1988 New York manslaughter convictions, as well as his 1983 conviction for firstdegree robbery. See N.Y. Penal Law §160.15(3); Stuckey v. United States, 878 F.3d 62, 71–72 (2d Cir. 2017) (holding New York firstdegree robbery categorical violent felony under ACCA). Accordingly, the district court sentenced Scott to a total 22 years’ imprisonment. This sentence, toward the low end of Scott’s 262-to- 327-month Guidelines range, reflected the ACCA mandated 15-year sentence for being a felon in possession, a concurrent 15-year term for attempted Hobbs Act robbery, and a mandated consecutive 7-year sentence for brandishing a firearm. See 18 U.S.C. §924(c)(1)(A)(ii). In 2017, on habeas review of Scott’s sentence, see 28 U.S.C. §2255, the district court reversed course, holding that neither the original ACCA nor Career Offender Guideline enhancements were warranted because New York first-degree manslaughter is not a categorical violent felony or crime of violence as it can be committed passively, by omission, without a defendant himself using, attempting to use, or threatening to use physical force. The district court further concluded that New York first-degree manslaughter committed by omission did not fit within the generic definitions of “voluntary manslaughter,” “murder,” or “aggravated assault,” crimes specifically enumerated in the second clause of the Guidelines’ definition of a “crime of violence.” See supra note 3. Accordingly, it vacated Scott’s original sentence; recalculated his Guidelines range without the career offender enhancement as 121 to 130 months; and sentenced Scott to time served, which by that time was approximately 135 months. The panel majority now affirms that judgment, reasoning that New York first-degree manslaughter is not a categorical violent felony or crime of violence because the minimum conduct required to commit the crime does not include a defendant’s active use of force. Rather, the crime may be committed through omission, as occurs when a defendant, intent on causing serious physical injury, causes death by failing to prevent deadly harm despite the duty and ability to do so. I think precedent forecloses this reasoning. II. The Probability of a First-Degree Manslaughter Conviction by Omission Alone As the majority acknowledges, see Majority Op., ante at 9–10, to hold that the minimum conduct for a predicate crime of conviction does not require the use of violence “there must be a realistic probability, not a theoretical possibility, that the statute at issue could be applied to conduct that does not constitute a [violent felony],” United States v. Hill, 890 F.3d at 56 (quoting Gonzales v. Duenas-Alvarez, 549 U.S. 183, 193 (2007)). To demonstrate that probability, a defendant “must at least point to his own case or other cases in which the…courts in fact did apply the statute in the…manner for which he argues.” Id. (emphasis added) (omissions in original) (internal quotation marks omitted); see Moncrieffe v. Holder, 569 U.S. at 206 (holding that “realistic probability” requires defendant to show that “State actually prosecutes the relevant offense” in the non-violent circumstances hypothesized).¹⁴ This requirement ensures that patently violent crimes do not escape being denominated as such on theories that, even if legally possible, are not reflected in actual prosecutions. Scott cannot rely on his own case to carry his realisticprobability burden. As noted at the outset, Scott actively deployed violence to kill his manslaughter victims, shooting one point-blank in the head and stabbing the other. Thus, he must point to other cases in which New York actually prosecuted first-degree manslaughter based only on a theory of omission. He cannot do so. In concluding otherwise, the majority relies on two cases: People v. Steinberg, 79 N.Y.2d 673, 584 N.Y.S.2d 770, and People v. Wong, 81 N.Y.2d 600, 601 N.Y.S.2d 440 (1993). In neither, however, was a defendant actually prosecuted, much less successfully convicted, of first-degree manslaughter based only on omission. In Steinberg the defendant was convicted for inextricably linked acts of commission and omission. In Wong, the defendants’ manslaughter convictions, whether by commission or omission, were reversed. At issue in Steinberg was the notorious killing of a six-year-old child, Lisa Steinberg, at the hands of her “adoptive” father, attorney Joel Steinberg. Steinberg was convicted of first-degree manslaughter on a theory that he engaged in both “acts of commission (striking Lisa) and acts of omission (failing to obtain medical care), each with intent to cause serious physical injury.” People v. Steinberg, 79 N.Y.2d at 680, 584 N.Y.S.2d at 772.¹⁵ The New York Court of Appeals held that the law allowed Steinberg to be prosecuted on either theory. See id. at 680–81, 584 N.Y.S.2d at 772–73. While this may admit the possibility that, in some future case, even a parent who did not himself fatally injure his child — as Steinberg undoubtedly did — could be convicted under §125.20(1) for failing to secure necessary life-saving care for the child, Steinberg itself sheds little light on the realistic probability of such prosecutions resulting in successful convictions. When the Court of Appeals, in Steinberg, stated that “the failure to obtain medical care can…support a first degree manslaughter charge,” it observed that conviction nevertheless requires “sufficient proof of the requisite mens rea — intent to cause serious physical injury.” Id. at 680, 584 N.Y.S.2d at 772. In Steinberg, the fact that defendant acted with such injurious intent when he failed to secure medical care for his child was powerfully demonstrated by evidence that he himself had committed the brutal assault triggering the child’s need for care. In short, in the Steinberg prosecution, the defendant’s act of omission was inextricably linked to his acts of commission, and the Court of Appeals discussed them in conjunction in rejecting Steinberg’s sufficiency challenge to his first-degree manslaughter conviction. See id. at 683, 584 N.Y.S.2d at 774 (holding that “jury could have inferred from the evidence that defendant’s objective in assaulting Lisa and failing to summon medical assistance was to cause serious physical injury”). For this reason, Steinberg is not itself a case in which a defendant was “actually prosecute[d]” for New York first-degree manslaughter for omission alone. Moncrieffe v. Holder, 569 U.S. at 206; accord People v. Hill, 890 F.3d at 56.¹⁶ Nor is People v. Wong, 81 N.Y.2d 600, 601 N.Y.S.2d 440, such a case. There, a couple entrusted with the care of a three-month-old baby was prosecuted for manslaughter in both the first and second degrees, see N.Y. Penal Law §§125.15(1), 125.20(1),¹⁷ when, after only four days in their care, the child died from “internal brain injuries…that could only be attributed to ‘shaken baby syndrome,’” People v. Wong, 81 N.Y.2d at 605, 601 N.Y.S.2d at 443. Unable to prove which defendant had actually shaken the baby, the prosecution maintained that both were guilty “either by commission for having actually shaken the child, or by omission for failing to protect the child.” People v. Wong, 182 A.D.2d 98, 108, 588 N.Y.S.2d 119, 124 (1st Dep’t 1992). Thus, commission and omission were also closely linked in Wong, even if not as inextricably intertwined as in Steinberg. In any event, the New York courts ruled that neither Wong could stand convicted of manslaughter on either theory. The Appellate Division, First Department, reversed the Wongs’ first-degree manslaughter convictions, reasoning that record evidence indicating that the shaker’s “probable intention” was “to quiet [the child]…and not necessarily to harm him,” raised a “reasonable doubt” as to either defendant’s possession of the required specific intent to cause serious physical injury. Id. at 107, 588 N.Y.S.2d at 123. The Court of Appeals then reversed the Wongs’ second-degree manslaughter convictions. It was as to that charge, which required proof only of recklessness, and not of a specific intent to cause physical injury, see People v. Rivera, 23 N.Y.3d 112, 122, 989 N.Y.S.2d 446, 453 (2014), that the Court observed that the prosecution’s theory — that one defendant “shook the baby violently while the other, aware of the harm being done, failed to seek medical assistance” — “could theoretically support [second-degree manslaughter] convictions in a proper case,” People v. Wong, 81 N.Y.2d at 607, 601 N.Y.S.2d at 443 (emphasis added). But it held that Wong was not that case, a conclusion that means Wong does not demonstrate a realistic probability of applying New York’s first-degree manslaughter statute only on a theory of omission. See United States v. Hill, 890 F.3d at 56 (requiring more than “theoretical possibility” of statute being applied to non-violent conduct). Highlighting “the absence of proof from which the jury could infer that the ‘passive’ defendant was even aware that the infant had been violently shaken,” the Court concluded that the record would not admit a finding that such defendant recklessly failed to procure needed medical care. People v. Wong, 81 N.Y.2d at 609, 601 N.Y.S.2d at 445. Indeed, the Court suggested that, given the difficulty of recognizing shaken baby syndrome, the proof would have to admit an inference that the “passive” defendant “actually witnessed the shaking” to support guilt on a theory of omission. Id. at 610, 601 N.Y.S.2d at 445. Such evidence might imply more than failure to seek medical care; it might imply complicity in the violence necessitating such care. But even if it would not, the reversals of all manslaughter convictions in Wong preclude us from identifying that case as one in which a defendant has in fact successfully been prosecuted for New York first-degree manslaughter based only on an omission.¹⁸ In sum, while Steinberg admits the legal “possibility” of a first-degree manslaughter conviction being obtained solely on a theory of omission, because that was not “in fact” the case in Steinberg — and even less so in Wong — these cases are insufficient to demonstrate the “realistic probability” of application that is a necessary prerequisite for the majority’s conclusion that New York first-degree manslaughter is not a categorical violent felony or crime of violence. United States v. Hill, 890 F.3d at 56 (internal quotation marks omitted).¹⁹ III. A Defendant Who Commits New York First-Degree Manslaughter Necessarily “Uses” Physical Force in Causing Death In United States v. Castleman, 572 U.S. 157, the Supreme Court stated that the “knowing or intentional causation of bodily injury necessarily involves the use of physical force.” Id. at 169 (emphasis added). Indeed, the Court went on to state that “it is impossible to cause bodily injury without applying force.” Id. at 170 (emphasis added); accord id. at 174 (Scalia, J., concurring) (“[I]t is impossible to cause bodily injury without using force capable of producing that result.” (internal quotation marks omitted)).²⁰ Thus, Castleman compels the conclusion that to be guilty of causing death while intending to cause at least serious physical injury — the required elements of N.Y. Penal Law §125.20(1) — a defendant must have used physical force. The conclusion obtains, moreover, regardless of whether the use was direct or indirect, by commission or omission. Castleman makes that clear in rejecting an argument that a poisoner can kill his victim without using physical force by the non-violent act of surreptitiously sprinkling poison in his drink. The Court explained that the “use of force” in that circumstance is “not the act of sprinkling the poison; it is the act of employing poison knowingly as a device to cause physical harm.” United States v. Castleman, 572 U.S. at 171 (internal quotation marks and brackets omitted). In holding otherwise, the panel majority attempts to narrow Castleman, insisting that the Court there addressed only “whether a low level of force that produces a violent result” — sprinkling poison — “can qualify as use of force under ACCA.” Majority Op., ante at 25. It maintains that a defendant can only be said to have used injurious physical force if he took “some action that initiates a harmful consequence.” Id. at 26 (emphasis in original). This reading of Castleman, however, is belied by the just-quoted text, which makes clear that not only is the degree of force used in sprinkling poison irrelevant to the Court’s use-of-force conclusion, so too is the sprinkling itself. That is because the “use of force” in that circumstance is “not the act of sprinkling the poison” at all; rather, “it is the act of employing poison knowingly as a device to cause physical harm.” United States v. Castleman, 572 U.S. at 171 (internal quotation marks and brackets omitted). Applying Castleman’s reasoning in United States v. Hill, this court effectively recognized that a defendant can employ force without affirmatively initiating it when it rejected the argument that threatening to withhold vital medicine from a victim (as well as threatening to poison him) would not threaten the use of physical force. See 890 F.3d at 59. By the majority’s reasoning, however, a defendant intent on killing or seriously injuring another person could be said to have used the deadly force of a poison if he himself sprinkled that substance in a victim’s drink, but not to have used such force if, possessed of the same injurious intent, and with the legal duty to intervene, he stood by and let the person imbibe what defendant knew was a poisoned drink. The distinction makes no sense, and not simply because the two circumstances involve comparable moral culpability. See Majority Op., ante at 28–29. Rather, the reason no distinction is warranted is because of how the Supreme Court has construed statutory text penalizing crimes committed by the use of force. In both hypothesized circumstances, the defendant is “employing poison knowingly as a device to cause [intended] physical harm” — the use-of-force standard articulated in Castleman, 572 U.S. at 171. In the second circumstance, the person employs poison as the instrument for his injurious intent by deliberately failing to act on a duty to remediate the poison. Indeed, the law equates such inaction in the face of a duty to act to voluntary action. See N.Y. Penal Law §15.10 (equating “omission to perform an act which [person] is physically capable of performing” with “voluntary act” in identifying minimal conduct required for criminal culpability). For that reason, the majority cannot fairly characterize a culpable omission as taking “no action whatsoever,” Majority Op., ante at 28, or as doing “nothing at all,” Concurring Op., ante at 8. The culpable “action” taken is deliberate inaction, while under a duty to act, with the specific intent thereby to cause serious physical harm. The majority nevertheless insists that the distinction it draws is compelled by Curtis Johnson v. United States, which, in holding that the term “physical force” should be construed according to its “ordinary meaning,” observed that dictionary definitions describe “force” as “active power; vigor,” and “physical force” as “force consisting in a physical act.” 559 U.S. at 139 (internal quotation marks and brackets omitted); see Majority Op., ante at 21–22. From this, the majority concludes that “only an active crime constitutes a ‘violent felony’ under ACCA.” Majority Op., ante at 22. I respectfully suggest that, in reaching this conclusion, my colleagues conflate “physical force” itself with the “use” of such force. Curtis Johnson holds that the former connotes active physical power and vigor. But it says nothing about the latter; certainly not that the “use” of physical force must be affirmative rather than passive. It was only subsequently, in United States v. Castleman, that the Supreme Court discussed what constitutes a “use” of “physical force.” There, the Court explained that the word “use” “conveys the idea that the thing being used (here, ‘physical force’) has been made the user’s instrument.” 572 U.S. at 170–71 (internal quotation marks omitted). I respectfully submit that when a person, (1) who has the legal duty and the ability to protect another from physical injury, (2) fails to act on that duty, (3) because he specifically intends to cause serious physical injury, the person must be said, by his omission, to have adopted the physical force causing harm as the “instrument” of his own injurious intent and, thus, to have used that physical force as much as if he had actively deployed it in the first instance. The majority suggests that this “overstates the holding” in Castleman because the Supreme Court did not there consider the physical force question “for ACCA purposes.” Majority Op., ante at 28. Such criticism is unwarranted. This court has already recognized that the determinative significance Castleman affords causation in identifying crimes committed by the use of force applies in the ACCA context. In Villanueva v. United States, an ACCA case, we observed that, following Castleman, a federal court’s “use of force” inquiry properly “focuses on the causation of a consequence, rather than the physical act of initiating an action that leads to a consequence.” 893 F.3d at 128. Put another way, in deciding whether a crime categorically requires the use of force, a court properly considers the necessary impact of an instrument of force on the crime’s victim, not the impact of a defendant’s actions — or inactions — on that instrument. See id. at 129 (“[R]elevant force is the impact of the substance on the victim, not the impact of the user on the substance.”). Because the only consequence for a victim of New York first-degree manslaughter is death — the ultimate physical injury — and because that consequence can only be caused by the application of physical force to the victim, see United States v. Castleman, 572 U.S. at 169–70, a defendant convicted of that homicide crime because he intended to cause at least serious physical injury, must be said to have intentionally used the force causing death as his instrument without regard to whether he did so by commission or omission.²¹ Five of our sister circuits have already sensibly refused to draw a distinction between crimes causing physical injury by commission and those causing such injury by omission in identifying crimes of violence under various provisions of federal law. See United States v. Rumley, — F.3d —, 2020 WL 1222681 (4th Cir. Mar. 13, 2020) (holding Virginia unlawful wounding crime of violence under ACCA’s force clause because, following Castleman, “there is just as much ‘use of force’ when a murderous parent uses the body’s need for food to intentionally cause his child’s death as when that parent uses the forceful physical properties of poison to achieve the same result”); United States v. Sanchez, 940 F.3d 526, 536 (11th Cir. 2019) (holding New York second-degree murder crime of violence under ACCA’s force clause because, under “the logic of Castleman, we see no reason to draw a distinction between administering a poisonous substance with the intent to cause death and withholding a life-saving substance with the intent to cause death”); United States v. Peeples, 879 F.3d 282, 287 (8th Cir. 2018) (holding Iowa attempted murder categorical crime of violence under U.S.S.G. §2K2.1(a)(4)(A), because, even when committed by omission, such as by withholding food with intent to cause dependent to starve, it is impossible to cause death without force); United States v. Ontiveros, 875 F.3d 533, 538 (10th Cir. 2017) (holding Colorado second-degree assault crime of violence under ACCA’s force clause because, following Castleman’s logic, “[i]f it is impossible to commit a battery without applying force, and a battery can be committed by an omission to act, then second-degree assault must also require physical force” (emphasis omitted)); United States v. Waters, 823 F.3d 1062, 1066 (7th Cir. 2016) (adhering to precedent recognizing domestic battery under Illinois law as crime of violence for purposes of Career Offender Guideline because, assuming crime could be committed by withholding medicine, that omission “causes physical harm, albeit indirectly, and thus qualifies as the use of force under Castleman”). But see United States v. Mayo, 901 F.3d 218, 227, 230 n.15 (3d Cir. 2018) (declining to extend Castleman to ACCA context, although recognizing this court to have done so in Villanueva based on Castleman’s focus on causation). I would join them, and hold New York first-degree manslaughter a categorical violent felony and crime of violence under the force clauses of both ACCA and the Career Offender Guideline. IV. The Career Offender Guideline’s Enumerated Offenses Clause Because I think Scott’s manslaughter convictions are categorical violent felonies and crimes of violence under the force clauses of both ACCA and the Career Offender Guideline, I would not reach the government’s alternative argument urging a crime-of-violence conclusion under the Guideline’s enumerated offenses clause. See supra note 3. Nevertheless, because the panel majority rejects this alternative argument, I here briefly explain why I am not persuaded by its reasoning. A court identifying a crime of violence by reference to an enumerated offense properly determines the “generic” definition of the enumerated offense by looking to the “sense in which the term is now used in the criminal codes of most States.” Taylor v. United States, 495 U.S. 575, 598 (1990). In the process, a court may also “consult other sources, including federal criminal statutes, the Model Penal Code, scholarly treatises,…legal dictionaries,” and “the common law…to frame the analysis.” United States v. Castillo, 896 F.3d 141, 150 (2d Cir. 2018). For a crime of conviction to qualify as an enumerated violent offense, its elements must be “the same as, or narrower than, those of the generic offense.” Descamps v. United States, 570 U.S. 254, 257 (2013); accord Quarles v. United States, 139 S. Ct. 1872, 1877 (2019); United States v. Castillo, 896 F.3d at 150. The government asserts that New York first-degree manslaughter is an enumerated crime of violence under the Career Offender Guideline because a majority of the states penalize conduct with the elements of N.Y. Penal Law §125.20(1) — causing death while intending serious bodily injury — as either murder (20 states)²² or voluntary manslaughter (eight states),²³ while almost all states penalize such conduct, at a minimum, as aggravated assault.²⁴ The majority rejects the argument insofar as it requires aggregating states holding what the majority perceives as minority views of murder and voluntary manslaughter. Indeed, it maintains that a majority of the states limit murder and voluntary manslaughter to homicides committed with the specific intent to kill. As for aggravated assault, the majority insists that it cannot be committed by omission. I am not convinced. The government’s urged aggregation approach finds support in the Supreme Court’s instruction that courts identifying categorical crimes of violence by reference to enumerated offenses should not be controlled by the “label” states attach to their crimes but, rather, by the elements of those crimes. Quarles v. United States, 139 S. Ct. at 1880. Mindful of that instruction, I conclude that New York first-degree manslaughter is an enumerated crime of violence because a clear majority of the states penalize the conduct proscribed by the two elements of N.Y. Penal Law §125.20(1) — (1) causing death while (2) intending to cause serious physical injury — as a Guidelines-enumerated crime of violence, whether labeled murder or voluntary manslaughter. While the panel majority characterizes these states as holding minority views of the mens rea element of murder and voluntary manslaughter, see Majority Op., ante at 35–36, the fact remains that a majority of the states treat the conduct proscribed by the elements of N.Y. Penal Law §125.20(1) as one or the other of these enumerated crimes of violence. Indeed, it would be more than curious to conclude that New York first-degree manslaughter is not an enumerated crime of violence because only eight states would label it voluntary manslaughter when another 20 would label it as the more serious crime of murder. To explain, as Professor Wayne LaFave details, at common law, “one who intended to do serious bodily injury short of death, but who actually succeeded in killing, was guilty of murder in spite of his lack of an intent to kill, in the absence of circumstances which mitigated the offense to voluntary manslaughter or which justified or excused it.” 2 Wayne R. LaFave, SUBSTANTIVE CRIMINAL LAW §14.3, at 590 (3d ed. 2018). “This type of common-law murder became a part of the law of murder in America.” Id. It was only over time that “[m]ost” — but far from all — “modern [state penal] codes” came to “define murder as not including the intent-to-do-serious-bodily injury type.” Id. at 591 & n.5 (comparing codes that do not treat death caused intending serious bodily injury as murder with those that do, either expressly or as a lesser degree of murder). Instead, death caused with the intent to cause serious bodily harm was punished as the lesser homicide of voluntary manslaughter. See id. §15.2(a), at 670 (observing that “[a]lthough the killing of another person — when accompanied by an intent to kill, or by an intent to do serious bodily harm short of death, or when resulting from such unreasonable and highly reckless conduct as to evince a depraved heart — often amounts to murder, yet it may under certain circumstances amount only to voluntary manslaughter” (emphasis added) (internal quotation marks omitted)). Nevertheless, as the New York Court of Appeals has observed, a homicide caused with intent to cause serious physical injury rather than death, i.e., first-degree manslaughter, is a lesser included offense of second-degree murder. See People v. Hull, 27 N.Y.3d 1056, 1058, 35 N.Y.S.3d 284, 286 (2016). This background explains why homicides caused while intending to cause serious physical injury, which New York punishes as first-degree manslaughter, are properly viewed as categorical crimes of violence under the enumerated offenses clause of the Career Offender Guideline. A clear majority of states punish crimes having these same elements as either murder or voluntary manslaughter. Only a minority of states fails to punish this conduct as one of these enumerated crimes of violence. Our recent decision in United States v. Castillo, 896 F.3d 141, further supports this conclusion. There, we held New York first-degree manslaughter a categorical crime of violence under the pre-2016 Guidelines by analogy to the enumerated crime of “generic manslaughter,” which we construed to include both voluntary and involuntary manslaughter. See id. at 151 n.36, 153–54.²⁵ We explained that, historically, “murder was the unlawful killing of a human being with malice aforethought,” while “manslaughter was the unlawful killing of a human being without malice aforethought.” See id. at 151 (internal quotation marks omitted). Over time, the term “malice aforethought came to encompass a variety of mental states, including intent to kill in the absence of extenuating circumstances, intent to do serious bodily injury, depraved heart, and intent to commit a felony.” Id. (emphasis added) (internal quotation marks omitted). Meanwhile, manslaughter was subdivided into “voluntary and involuntary varieties,” the former being “intentional killing in a heat of passion upon adequate provocation,” and the latter being “unintentional killing caused by criminal negligence or recklessness, or during the commission of an unlawful act not amounting to a felony.” Id. (internal quotation marks omitted). While some state codes, and the federal criminal code, preserve a voluntary/involuntary distinction as to manslaughter, Castillo acknowledges that “the modern trend, reflected in a majority of recent codifications, is for there to be but one single manslaughter crime.” Id. (quoting 2 LaFave §15.1, at 668-69). Thus, Castillo concluded that while the “mens rea of recklessness…serves as the proverbial floor for the generic definition of ‘manslaughter,’” “the states of mind that historically fell under the category of ‘malice aforethought’ associated with murder,” including the intent to kill and to do serious bodily injury, also satisfy that generic definition. Id. at 152. From this, I conclude that where manslaughter remains divided into voluntary and involuntary homicides, the mens rea element for generic involuntary manslaughter is recklessness. Higher mental state requirements, indicative of the malice aforethought originally associated with murder — such as intent to kill or to do serious bodily injury — denominate homicides now generically recognized as voluntary manslaughter. Cf. id. at 153 (recognizing “recklessness (a mens rea of [New York] second-degree manslaughter) [as] a lower mental state than intent to cause serious injury,” as required for New York first-degree manslaughter). Thus, contrary to my colleagues in the majority, I identify New York first-degree manslaughter as the equivalent of generic voluntary manslaughter under the Career Offender Guideline’s enumerated offenses clause. Further, I think the conduct satisfying the elements of New York first-degree manslaughter — causing death intending to cause serious physical injury — would be punished in virtually all states, at a minimum, as aggravated assault, a term that also includes aggravated battery. See supra note 16. Certainly, the Model Penal Code §211.1(2)(a) defines “aggravated assault” to include causing “serious bodily injury…purposely [or] knowingly.” The treatises are also in accord. See 6 AM. JUR. 2D Assault & Battery §32, Westlaw (database updated Nov. 2019) (“Aggravated assault usually consists of intentionally or recklessly causing great or serious bodily harm to another.”); 2 LaFave §16.2(d), at 764-65 (observing that aggravated battery includes “intentionally or knowingly” inflicting “serious bodily injury”). To avoid this conclusion, the panel majority faults the government for failing to show that most states would prosecute serious bodily injury purposely caused by omission as aggravated assault. See Majority Op., ante at 40. I disagree. First, such a showing is unnecessary given Scott’s failure to cite a case in which New York actually prosecuted first-degree manslaughter based only on omission. Second, even assuming such a showing were necessary, it is satisfied here. It is well established in law that injury can be proximately caused by either commission or omission, particularly when the latter pertains to a legal duty. See W. Page Keeton et al., PROSSER & KEETON ON TORTS §41, at 263 (5th ed. 1984) (stating that “essential element” of negligence claim is “proximate cause,” i.e., “reasonable connection between the act or omission of the defendant and the damage which the plaintiff has suffered” (emphasis added)). The same conclusion obtains for criminal liability generally, which the Model Penal Code instructs can “be based on conduct that includes a voluntary act or the omission to perform an act of which [a defendant] is physically capable.” MODEL PENAL CODE §2.01(1) (emphasis added); see id. §1.13 (defining “conduct” of crime as “an action or omission and its accompanying state of mind” (emphasis added)); see also 2 WHARTON’S CRIMINAL LAW §25, at 139 (15th ed. 1993) (observing that to be criminally liable, defendant must engage in “conduct,” which “means an act or omission” (internal quotation marks omitted)). Citing the Model Penal Code, Professor LaFave observes that “modern recodifications typically state that crimes may be committed either by an act or an omission,” when there is a “legal duty” to act. 1 LaFave §6.2, at 588 & n.1. Indeed, LaFave identifies as the first element of criminal battery “the defendant’s conduct (act or omission).”²⁶ In sum, because a clear majority of the states penalize the conduct proscribed by the elements of N.Y. Penal Law §125.20(1) as either generic voluntary manslaughter or the more serious homicide crime of murder, and because an even larger majority of the states penalize such conduct, at a minimum, as generic aggravated assault, I conclude that the Career Offender Guideline applies to Scott’s case. V. Conclusion For the reasons stated in this opinion, I conclude that Scott should not have been granted habeas relief from his original sentence because he has three prior convictions for violent felonies under ACCA and for crimes of violence under the Career Offender Guideline. To the extent Scott has two prior convictions for first-degree manslaughter in violation of N.Y. Penal Law §125.20(1), I conclude that those convictions are for categorical violent felonies and crimes of violence under the force clauses of both ACCA and the Career Offender Guideline. To commit New York first-degree manslaughter, a defendant must (1) cause death while (2) specifically intending to cause at least serious physical injury. Death can only be caused by intentionally using physical force, that is, by a person making fatal force his own instrument for causing intended injury. See United States v. Castleman, 572 U.S. at 169. Whether a defendant himself deploys that force in the first instance or avails himself of such force already in motion (by failing to act on a duty to remediate that force because he specifically intends serious injury) does not alter the conclusion that he has intentionally used physical force to commit the crime. Thus, New York first-degree manslaughter is a categorical crime of violence. While I do not think it necessary to decide whether New York first-degree manslaughter is also a categorical crime of violence under the enumerated offenses clause of the Career Offender Guideline, to the extent the panel majority concludes that it is not, I cannot agree with its reasoning. I, therefore, respectfully dissent from the court’s decision today affirming the district court judgment in favor of Scott.