This appeal by the United States presents the narrow issue of whether the offense of violating Connecticut's statute punishing first degree assault, Conn. Gen. Stat. § 53a-59(a)(1), qualifies as a "violent felony" for purposes of enhanced sentencing under the federal Armed Career Criminal Act of 1984 ("ACCA"), 18 U.S.C. § 924(e). That issue turns on whether the *125Connecticut statute, analyzed under the so-called "modified categorical approach," Mathis v. United States , --- U.S. ----, 136 S.Ct. 2243, 2249, 195 L.Ed.2d 604 (2016), has as an element what federal law means by defining "violent felony" to require the use of "physical force." 18 U.S.C. § 924(e)(2)(B)(i). This issue arises on an appeal from the June 14, 2016, amended judgment of the District Court for the District of Connecticut (Janet C. Hall, Chief Judge), modifying the sentence of Appellee Richard Villanueva. That judgment brings up for review the District Court's June 10, 2016, ruling that Villanueva's assault offense was not a "violent felony" for purposes of the ACCA.
Because we conclude that Villanueva's assault conviction qualified as an ACCA predicate, we remand for resentencing.
Background
The ACCA authorizes a punishment of up to ten years' imprisonment for any person who possesses a firearm after being convicted of a felony. See 18 U.S.C. §§ 922(g)(1), 924(a)(2). The ACCA also requires a minimum fifteen year term of imprisonment for any person who violates subsection 922(g) and has three previous convictions for a "violent felony" or a "serious drug offense." See id. § 924(e)(1). "Violent felony" is defined, as relevant to this case, as any crime punishable by imprisonment for a term exceeding one year that "has as an element the use ... of physical force against the person of another," id. § 924(e)(2)(B)(i), or "involves conduct that presents a serious potential risk of physical injury to another," id. § 924(e)(2)(B)(ii). Subsection 924(e)(2)(B)(i) is known as the "elements clause," and the quoted portion of subsection 924(e)(2)(B)(ii) is known as the "residual clause." See Welch v. United States , --- U.S. ----, 136 S.Ct. 1257, 1261, 194 L.Ed.2d 387 (2016). The "elements clause" is sometimes called the "force clause." See , e.g. , United States v. Jones , No. 15-1518-cr, 2017 WL 3974269, at *1 (2d Cir. Sept. 11, 2017).
In June 1999, Villanueva, then using the name Richard Zebrowski, was indicted for unlawful possession of a firearm by a convicted felon in violation of 18 U.S.C. § 922(g)(1). The indictment alleged four prior convictions, two for narcotics violations, a third for first degree assault in violation of Conn. Gen. Stat. 53a-59(a), and a fourth for assault on an officer in violation of Conn. Gen. Stat. 53a-167(c). After a jury found Villanueva guilty, a presentence report ("PSR") recommended sentencing under the ACCA's minimum fifteen year sentence provision. The PSR reflected that the first degree assault conviction resulted from Villanueva's firing a gun three or four times and hitting his victim in the shoulder. The PSR calculated a Guidelines range of 262-327 months.
The District Court concluded that each of the two narcotics convictions was a "serious drug offense" within the meaning of subsection 924(e)(2)(A)(ii) and at least one of the assault convictions was a "violent felony" within the meaning of subsection 924(e)(2)(B), without specifying whether the elements clause or the residual clause of that subsection applied. Because the District Court did not specify which of the two assault convictions qualified as a "violent felony" for purposes of the ACCA or whether it was using the elements clause or the residual clause, it left open the possibility that it was implicitly using the residual clause. The Court imposed a sentence pursuant to the ACCA of 262 months. This Court affirmed in part and dismissed in part. United States v. Zebrowski , 229 F.3d 1136 (2d Cir. 2000) (table).
*126In 2007, this Court affirmed a denial of Villanueva's first motion to vacate his sentence under 28 U.S.C. § 2255. That motion raised no issue relating to the ACCA.
On February 13, 2013, the District Court entered an amended judgment to reflect the fact that Villanueva's name had been legally changed from Zebrowski. The sentence of 262 months remained unchanged.
In June 2015, the Supreme Court ruled that the "residual clause" of the ACCA was unconstitutionally vague. See Johnson v. United States , --- U.S. ----, 135 S.Ct. 2551, 2557, 192 L.Ed.2d 569 (2015).
Villanueva then filed in this Court, pursuant to 28 U.S.C. § 2244(b)(3)(A), a motion for leave to file in the District Court a second motion under 28 U.S.C. § 2255, relying on Johnson . This Court granted the motion, concluding that Villanueva had made "a prima facie showing" that his two assault convictions are not violent felonies "under any provision of the ACCA that remains in effect after Johnson ." We instructed the District Court to "determine whether the assault convictions remain proper ACCA predicates after Johnson , and what evidence may be considered in making that determination."1
On June 10, 2016, the District Court granted Villanueva's second section 2255 motion and vacated his sentence. See Villanueva v. United States , 191 F.Supp.3d 178 (D. Conn. 2016). In a carefully considered opinion, the Court first said "that it is more likely than not that [Villanueva] was sentenced under ACCA's Residual Clause," id. at 184, which led to the Court's conclusion that he had "shown by a preponderance of the evidence that the court [had] sentenced him under the Residual Clause, and not the Elements Clause," see id . at 188.
The Court then rejected the Government's contention that the error of using the residual clause was harmless because it was not a structural error.2 See id . at 190. The Court then stated that it need not rest on its structural error determination because, even on harmless error review, the error was not harmless. See id. This conclusion rested on the Court's determination, which is at the heart of the pending appeal, that neither of Villanueva's assault convictions qualified as ACCA predicates because neither was a "violent felony" for lack of a required element of the use of physical force. See id. at 191-98.
On June 14, 2016, the Court resentenced Villanueva. First, the Court recalculated his Sentencing Guidelines range to be 63-78 months.3 Then, recognizing that Villanueva had served more than the ten-year maximum term for a violation of subsection 922(g)(1) without an ACCA enhancement, the Court resentenced him to time served and placed him on supervised release.
*127A second amended judgment was entered the same day.
The Government seeks review of the District Court's June 10 ruling, made reviewable by entry of the second amended judgment on June 14.
Discussion
The Government contends that each of Villanueva's two assault convictions was a "violent felony" within the meaning of the elements clause of ACCA, 18 U.S.C. § 924(e)(2)(B)(i). Villanueva contends that neither assault conviction was a "violent felony."
The elements clause defines a "violent felony," required for enhanced punishment, to include a crime that "has as an element the use, attempted use, or threatened use of physical force against the person of another." Id. The Supreme Court has stated that, for purposes of the elements clause, "the phrase 'physical force' means violent force-that is, force capable of causing physical pain or injury to another person." (Curtis) Johnson v. United States , 559 U.S. 133, 140, 130 S.Ct. 1265, 176 L.Ed.2d 1 (2010) (emphasis in original).4 So the issue on this appeal is whether either of the statutes defining Villanueva's two assault offenses includes, as an element, the use of violent force. We confine our consideration to the first degree assault conviction under Conn. Gen. Stat. § 53a-59(a)(1).
The Connecticut first degree assault provision has two subdivisions. The first states:
A person is guilty of assault in the first degree when: (1) With intent to cause serious physical injury to another person, he causes such injury to such person or to a third person by means of a deadly weapon or a dangerous instrument.
Conn. Gen. Stat. § 53a-59(a)(1).
Two definition provisions are relevant to our inquiry:
"Serious physical injury" means physical injury which creates a substantial risk of death, or which causes serious disfigurement, serious impairment of health or serious loss or impairment of the function of any bodily organ.
Conn. Gen. Stat. § 53a-3(4).
"Dangerous instrument" means any instrument, article or substance which, under the circumstances in which it is used or attempted or threatened to be used, is capable of causing death or serious physical injury ...."
Conn. Gen. Stat. § 53a-3(7).
The Supreme Court has instructed that in determining whether a state criminal statute qualifies as an ACCA predicate offense, courts are to use the so-called "categorical approach," and, when the state statute has subdivisions, courts are to use the so-called "modified categorical approach." See *128Descamps v. United States , 570 U.S. 254, 133 S.Ct. 2276, 2281, 186 L.Ed.2d 438 (2013). Under these approaches, "courts identify the minimum criminal conduct necessary for conviction under a particular statute" and "look only to the statutory definitions-i.e ., the elements-of [the] ... offense[ ], and not to the particular [underlying] facts." United States v. Hill , 890 F.3d 51, 55 (2d Cir. 2018) (internal quotation marks omitted) (emphasis, brackets, and ellipsis in original).
Under either the general or the modified categorical approach, courts ascertain the elements of the offense from such materials as the indictment, a plea agreement, or a plea colloquy, but courts are not permitted to consider the facts of the offense conduct, as might be revealed by a trial transcript or a PSR. "ACCA ... treats such facts [the means by which the crime was committed] as irrelevant: Find them or not, by examining the record or anything else, a court still may not use them to enhance a[n ACCA] sentence." Mathis v. United States, --- U.S. ----, 136 S.Ct. 2243, 2253, 195 L.Ed.2d 604 (2016). Thus, in this case, under the rigidly structured regime of categorical analysis, we may not consider the fact, disclosed by Villanueva's PSR, that he committed his assault by shooting his victim in the shoulder.
In the pending case, the Connecticut first degree assault statute is divisible into subsections, and the modified categorical approach is therefore applicable. We are concerned with subsection (1).
Villanueva contends, and the District Court agreed, that the first degree assault statute does not require the use of physical force and is therefore not a "violent felony" for purposes of the ACCA because, under the definition of "dangerous instrument," the offense could be committed by use of a "substance" and, for example, the use of a poisonous substance to kill or injure would not constitute the use of physical force. We disagree.
The Supreme Court rejected Villanueva's reasoning in United States v. Castleman , 571 U.S. 1045, 134 S.Ct. 405, 187 L.Ed.2d 449 (2013). Castleman was indicted for violating 18 U.S.C. § 922(g)(9), which prohibits possession of a firearm by a person previously convicted of a "misdemeanor crime of domestic violence." He argued that his previous conviction for causing bodily injury to the mother of his child did not involve the use of "physical force" as required by the federal statute's definition of a "misdemeanor crime of domestic violence," see 18 U.S.C. § 921(a)(33)(A)(ii). The District Court dismissed the charge, reasoning "that one can cause bodily injury without the use of physical force-for example, by deceiving [the victim] into drinking a poisoned beverage, without making contact of any kind." 134 S.Ct. at 1414 (internal quotation marks omitted).
The Supreme Court's rejection of the District Court's reasoning is instructive:
The "use of force" in Castleman's example is not the act of "sprinkl[ing]" 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. Under Castleman's logic, after all, one could say that pulling the trigger on a gun is not a "use of force" because it is the bullet, not the trigger, that actually strikes the victim.
Id . at 1415. Making clear its view that the inquiry as to "force," for federal law purposes, focuses on the causation of a consequence, rather than the physical act of initiating an action that leads to a consequence, the Court stated, "[T]he knowing or intentional causation of bodily injury *129necessarily involves the use of physical force." Id. at 1414.
Villanueva discounts the relevance of what the Supreme Court said in Castleman because that case concerned the element of "physical force" for purposes of a misdemeanor offense, one involving a domestic relations context. We recognize that language in one judicial opinion, even when written by the Supreme Court, cannot be uncritically applied to other cases. But the Supreme Court's explanation of why sprinkling poison constitutes use of physical force even if the offender does not hit or even touch the victim is precisely relevant to our case. The explanation that initiating, however gently, a consequence that inflicts injury constitutes the use of physical force was independent of both the domestic relations context and the fact that the offense at issue was a misdemeanor.
The reasoning of Castleman has been deemed relevant to the proper interpretation of various criminal statutes by several courts, including ours. See Hill, 890 F.3d at 58 (construing 18 U.S.C. § 924(c)(3)(A) ); Hernandez v. Lynch , 831 F.3d 1127, 1131 (9th Cir. 2016) (construing Cal. Penal Code §§ 422, 664 ); United States v. Waters , 823 F.3d 1062, 1066 (7th Cir. 2016) (construing 720 Ill. Comp. Stat. § 5/12-3.2 ); United States v. Rice , 813 F.3d 704, 705-06 (8th Cir. 2016) (construing Ark. Code Ann. § 5-13-202(a)(4) ).
Under the reasoning of Castleman , the use of a "substance" (the term in the Connecticut definition of "dangerous instrument") constitutes use of physical force, for federal law purposes, because the relevant force is the impact of the substance on the victim, not the impact of the user on the substance. Furthermore, the state statute's requirement that the use of a "substance" must be "capable of causing death or serious physical injury" to qualify as a "dangerous instrument" for purposes of first degree assault, see Conn. Gen. Stat. § 53(a)-3(7), satisfies the ACCA requirement that the predicate offense has as an element the use of physical force that is violent, as (Curtis) Johnson requires. See (Curtis) Johnson , 559 U.S. at 14, 130 S.Ct. 983.
Villanueva also endeavors to discount the reasoning of Castleman by pointing out that the Connecticut courts do not construe the first degree assault statute to require force,5 and reminding us that the Supreme Court has instructed that while federal law determines the meaning of "violent felony" in subsection 924(e)(1), state law determines the elements of a state offense, see (Curtis) Johnson , 559 U.S. at 138, 130 S.Ct. 1265. Castleman itself refutes this aspect of Villanueva's argument. The Supreme Court pointed out that "Castleman pleaded guilty to having 'intentionally or knowingly cause[d] bodily injury,' " 134 S.Ct. at 1414 (quoting the words of the indictment that tracked Tenn. Code Ann. § 39-13-101(a)(1) ), and then ruled that causing such a resulting injury was, as a matter of federal law, the use of physical force for purposes of determining whether "physical force," required by 18 U.S.C. § 921(a)(33)(A)(ii), was satisfied. State law specified the elements, but federal law determined whether the consequences of the conduct that those elements required for a state law offense rendered conviction for that conduct a "violent felony" under federal law.
*130Villanueva seeks to support the District Court's ruling by citing this Court's decision in Chrzanoski v. Ashcroft , 327 F.3d 188 (2d Cir. 2003). That decision ruled that a Connecticut conviction for third degree assault, see Conn. Gen. Stat. § 53a-61(a), was not a "crime of violence" for purposes of 18 U.S.C. § 16(a), which makes conviction for such a crime a basis for removal of an alien. We accept Villanueva's premise that "crime of violence" in subsection 16(a) is the equivalent of "violent felony" in subsection 924(e). See (Curtis) Johnson , 559 U.S. at 140, 130 S.Ct. 1265 (equating "crime of violence" in 18 U.S.C. § 16 with "violent felony" in subsection 924(e)(2)(B)(i) ). Nevertheless, we do not agree that Chrzanoski aids Villanueva.
Chrzanoski explained an understanding of the use of force that has been abrogated by the Supreme Court's decision in Castleman .6 We explained the effect of Castleman on Chrzanoski in Hill , 890 F.3d at 59-60. The issue in Hill was whether Hobbs Act robbery is a "crime of violence" within the meaning of 18 U.S.C. § 924(c)(3).7 See Hill , 890 F.3d at 53. Hill had argued that his offense could not qualify as a crime of violence under the elements clause of subsection 924(c)(3)(A) because it can be committed without physical force. See id .
In rejecting Hill's argument, we said, "In Castleman , the Supreme Court, construing 'physical force' as it is employed in connection with § 922(g)(9), made clear that physical force 'encompasses even its indirect application,' as when a battery is committed by administering a poison." Id. at 59 (quoting Castleman , 134 S.Ct. at 1414-15 ). Then, explicitly reckoning with Hill's reliance on Chrzanoski , we added, "[T]he Chrzanoski panel did not have the benefit of the Supreme Court's reasoning in Castleman to the effect that a use of physical force can encompass acts undertaken to cause physical harm, even when the harm occurs indirectly (as with poisoning)." Id. at 60 (quoting Castleman , 134 S.Ct. at 1415 ). Chrzanoski is as unavailing for Villanueva as it was for Hill.8
We deem it worth noting that we do not agree with the Government effort to enlist several other decisions of this Court in support of its argument that Villanueva's first degree assault conviction has as an element the use of physical force. See Brief for Appellant at 17 (citing United States v. Walker , 442 F.3d 787, 788 (2d Cir. 2006) ); id . at 19-20 (citing Harris v.
*131United States , No. 15-2679 (2d Cir. Nov. 17, 2015), United States v. Williams , 526 F. App'x 29, 37 (2d Cir. June 4, 2013), and United States v. Bogle , 522 F. App'x 15, 19-20 (2d Cir. May 23, 2013) ).
Walker , Williams , and Bogle ruled that a New York conviction for second degree assault, see N.Y. Penal L. § 120.05(2), qualified as an ACCA predicate offense. Although the New York second degree assault statute is very similar to the Connecticut first degree assault statute, and the New York definition of "dangerous weapon" is very similar to Connecticut's definition, there is no indication in any of these cases that the appellants contended that the "substance" language of the "dangerous instrument" definition precluded the New York convictions from qualifying as ACCA predicates. In view of the persuasive force of the reasoning in Castleman , we need not decide whether to rely on what would be at most an implicit holding of Walker , Williams , and Bogle . See Webster v. Fall , 266 U.S. 507, 511, 45 S.Ct. 148, 69 L.Ed. 411 (1925) ("Questions which merely lurk in the record, neither brought to the attention of the court nor ruled upon, are not to be considered as having been so decided as to constitute precedents.").
Harris , the fourth decision cited by the Government, concerned a Connecticut conviction for first degree assault "by means of the discharge of a firearm" in violation of subdivision (5) of Conn. Gen. Stat. 53a-59(a), not subdivision (1) as in the pending case, and did not require consideration of the definition of "dangerous instrument." See Government's Memorandum in Opposition to Petitioner's Motion for Leave to File a Successive Motion under 28 U.S.C. § 2255, at 12-14, and Harris's Traverse in Response, at 2-5, Harris v. United States , No. 15-2679 (2d Cir. Nov. 17, 2015).
The District Court, apparently unaware that Harris did not concern the same subdivision of the first degree assault statute under which Villanueva was convicted, see Villanueva , 191 F.Supp.3d at 194,9 declined to consider Harris persuasive "where, as here, the Second Circuit has issued a mandate to this court directing the court to determine 'whether the assault conviction[ ] remain[s] a proper ACCA predicate after Johnson .' " Id . (brackets in original, italics added) (quoting Zebrowski v. United States , No. 15-3676 (2d Cir. Feb. 16, 2016) (order granting leave to file successive section 2255 motion) ).
The language of that direction is included in all orders of this Court granting leave to file a successive section 2255 motion in the aftermath of the Supreme Court's decision in Johnson . The language simply calls upon a district court to assess a prior conviction, without intimating any view of this Court as to what the outcome of that assessment should be. Such orders also customarily state, as the "leave to file" order did in this case, that the applicant has made a "prima facie showing that he has satisfied the successive motion requirements with respect to his proposed claim based on *132Johnson v. United States, --- U.S. ----, 135 S.Ct. 2551, 192 L.Ed.2d 569 (2015)." "Prima facie" has a different meaning in different contexts.10 In this context, it has the weakest possible meaning, indicating no more than that the applicant's claim should be considered by a district court.
Since Villanueva's first degree assault conviction qualified as an ACCA predicate, the District Court's order of June 10, 2016, must be vacated, the June 14, 2016, second amended judgment must also be vacated, and we will remand with directions to resentence. The District Court's options on resentencing remain to be determined.
The Government contends that the District Court should re-impose the original sentence of 262 months (21 years and 10 months).11 Villanueva, arguing for affirmance, does not consider the scope of resentencing if such should be ordered.
"{A} court's duty is always to sentence the defendant as he stands before the court on the day of sentencing," United States v. Bryson , 229 F.3d 425, 426 (2000), and a court is "required to resentence [the defendant] in light of the circumstances as they stood at the time of resentencing," Werber v. United States , 149 F.3d 172, 178 (2d Cir. 1998). At resentencing, Villanueva will have been out of prison for more than two years (after having served more than the ACCA mandatory fifteen year sentence). That is a circumstance the District Court is entitled to consider in deciding whether to impose a sentence that requires him to serve all, part, or none of the unexpired term of the original sentence. See Pepper v. United States, 562 U.S. 476, 490, 131 S.Ct. 1229, 179 L.Ed.2d 196 (2011) (rehabilitation since prior sentencing may be considered on resentencing); United States v. Hernandez , 604 F.3d 48, 55 (2d Cir. 2010) (defendant's rehabilitation during prolonged freedom pending resentencing warranted consideration).
Conclusion
The case is remanded with directions to vacate the ruling of June 10, 2016, vacate the sentence imposed on June 14, 2014, and resentence.12
POOLER, Circuit Judge:
I respectfully dissent.
This case calls on us to decide whether Villanueva's prior crimes qualify as "violent felonies" that render him eligible for an ACCA sentencing enhancement. See (Curtis) Johnson v. United States , 559 U.S. 133, 140, 130 S.Ct. 1265, 176 L.Ed.2d 1 (2010). My colleagues conclude that United States v. Castleman , 572 U.S. 157, 134 S.Ct. 1405, 188 L.Ed.2d 426 (2014), a *133case which analyzed the level of force applicable to a "misdemeanor crime of domestic violence," effectively decides the question. But this reasoning overlooks the plain and decisive fact that the Castleman Court repeatedly distinguished the ACCA context we confront here, and carefully cabined its discussion to exclude ACCA elements clause determinations. Indeed, the Supreme Court has now twice noted that the definition of force applicable to the Castleman inquiry was "a comical misfit" to ACCA. (Curtis) Johnson , 559 U.S. at 145, 130 S.Ct. 1265 ; Castleman , 134 S.Ct. at 1410 (quoting (Curtis) Johnson ).
As a result, my colleagues' reliance on Castleman improperly extends that decision to the very statutory context that the Castleman Court specifically and repeatedly differentiated. See United States v. Rico-Mejia , 859 F.3d 318, 322 (5th Cir. 2017) (declining to rely on Castleman in Sentencing Guidelines crime of violence determination because, "[b]y its express terms, Castleman 's analysis is not applicable to the physical force requirement for a crime of violence, which 'suggests a category of violent, active crimes' that have as an element a heightened form of physical force that is narrower in scope than that applicable in the domestic violence context"); see also Williams v. United States , 712 F. App'x 50, 53 (2d Cir. 2017) (summary order) (" Castleman ... only addressed the meaning of a 'misdemeanor crime of domestic violence' as defined by 18 U.S.C. § 922(g)(9). Thus, the decision does not squarely address the 'violent felony' definition of ACCA...").
Their application of Castleman here is all the more unwarranted since binding Circuit authority stands for the contrary proposition. In Chrzanoski v. Ashcroft , we reasoned that statutes such as the one we analyze today do not have, as an element, physical force. 327 F.3d 188, 195 (2d Cir. 2003). Given that Castleman explicitly confined its application to a specific setting of misdemeanor crimes of domestic violence, I would follow Chrzanoski and affirm the judgment of the district court following Circuit precedent-rather than taking the path Castleman seemed to expressly forbid, of applying a lower level of force to an ACCA case.
But as my colleagues note, they are not alone in their views. Our Circuit has previously interpreted Castleman in similar fashion, see United States v. Hill , 890 F.3d 51, 58-60 (2d Cir. 2018),1 and so have most circuit courts. For that reason, I recognize that the decision issued today is reasonable. But I cannot agree that it is correct.
At bottom, much of the difficulty here is exacerbated by the unhappy marriage between the elements clause analysis we must perform in the wake of Johnson v. United States , --- U.S. ----, 135 S.Ct. 2551, 192 L.Ed.2d 569 (2015) (" Johnson 2015 "), and legislative draftsmanship that did not account for this altered landscape. But, given the stakes-and the many years of imprisonment that will be spent behind bars, not only by Villanueva, but also by a class of individuals with a similar conviction record-our meticulous attention to this categorical approach analysis is vital to the integrity of our justice system. I fear that all too frequently we fall prey to *134the impulse to decide these cases instinctively, in a way that renders elements clauses just as capacious and chameleon-like as the residual clauses that are now broadly invalid.
I. Castleman Cabined Itself to Its Statutory Setting
In 2010, the Supreme Court in (Curtis) Johnson ruled that ACCA's elements clause-the same clause at issue here-required qualifying crimes have "violent force" as an element. 559 U.S. at 140, 130 S.Ct. 1265 (emphasis in original). Importantly, the (Curtis) Johnson Court specifically rejected the notion that the definition of "physical force" found in the common law should govern the ACCA elements clause inquiry. Id. at 139-42, 130 S.Ct. 1265. The (Curtis) Johnson Court reasoned that a common-law conception of force was ill-suited to ACCA, because the common-law notion of force was derived from the misdemeanor crime of battery, and required only "the slightest offensive touching." Id. at 139, 130 S.Ct. 1265. Thus common-law force was "a comical misfit" for ACCA's elements clause, which defined a category of "violent felonies." Id. at 145, 130 S.Ct. 1265.
In Castleman , the Supreme Court ruled that common-law force-"namely, offensive touching"-did apply to the determination of whether a crime qualifies as "a misdemeanor crime of domestic violence" under 18 U.S.C. § 922(g)(9). 134 S.Ct. at 1410. In so deciding, the Castleman Court was at pains to distinguish (Curtis) Johnson and the ACCA setting. The Castleman Court reasoned that common-law force, with its roots in misdemeanor crimes, was a poor fit for the "violent felonies " discussed in ACCA, but, in the context of a "misdemeanor crime of domestic violence," it was "likely that Congress meant to incorporate that misdemeanor-specific meaning of 'force.' " 134 S.Ct. at 1411 (emphasis added). The Castleman Court also noted that the two contexts were different because " '[d]omestic violence' is not merely a type of 'violence'; it is a term of art encompassing acts that one might not characterize as 'violent' in a nondomestic context." Id. As a result, "[m]inor uses of force may not constitute 'violence' in the generic sense," but may well be "easy to describe as 'domestic violence.' " Id. ; see also Stuckey v. United States , 878 F.3d 62, 69 (2d Cir. 2017) (observing that in Castleman , "the Court confirmed that certain 'minor uses of force' do not rise to the level of violence that the ACCA requires") (brackets omitted).
In light of the different statutory context at issue in Castleman , and the common-law force applicable in that statutory setting, the Castleman Court found that a Tennessee domestic abuse statute constituted a "misdemeanor crime of domestic violence," even though a conviction could lie for acts such as "deceiving the victim into drinking a poisoned beverage, without making contact of any kind." Id. at 1414 (brackets omitted). Such a poison-based conviction would appropriately qualify as a misdemeanor crime of domestic violence because "the common-law concept of 'force' encompasses even its indirect application," and "[i]t is impossible to cause bodily injury without applying force in the common-law sense." Id. at 1414-15 (emphasis added).
Given this carefully cabined discussion, I cannot agree with my colleagues' efforts to graft Castleman back onto the ACCA inquiry. Contrary to their reasoning, Castleman did not create a regime where causation of an injury is the dispositive question for force inquiries under federal law (and such an approach is ill advised for myriad reasons, as I explain below). Nor does Castleman command that we reject the *135poison-related argument pressed by Villanueva here. Tellingly, my colleagues quote Hill 's recounting that "[i]n Castleman , the Supreme Court ... made clear that physical force 'encompasses even its indirect application.' " Op. at 130 (quoting Hill , 890 F.3d at 59 (itself quoting Castleman , 134 S.Ct. at 1414-15 ) ). But of course, the Castleman Court in fact made a much narrower pronouncement, asserting only that "the common-law concept of 'force' encompasses even its indirect application." Castleman , 134 S.Ct. at 1414 (emphasis added). The partial quote employed by my colleagues reflects the difficulties that inhere in the general approach of reading Castleman to extend far beyond its explicit bounds.
My colleagues, and other courts, have found support for this broad extension by looking to a portion of Castleman that discusses what it means to use force. Op. at 128-29. The Castleman Court in this passage commented that under the defendant's specious argument, "one could say that pulling the trigger on a gun is not a 'use of force' because it is the bullet, not the trigger, that actually strikes the victim." 134 S.Ct. at 1415. However, once again, context reveals that this passage does not militate in favor of the broad application my colleagues assert. Rather, that section addresses only how force is used in the relevant sense. And as the Supreme Court itself noted, it is simply not the case that "the word 'use' somehow alters the meaning of 'force.' " Id. Thus this passage is of little independent value in analyzing the meaning of "force."
On my read of Castleman , it is quite apparent that the opinion teaches that (i) there is significant daylight between common-law force and the violent force required here, and (ii) causation of a consequence will conclusively demonstrate the applicable degree of force only where the common-law definition of force applies.
For these same reasons, a poisoning could easily be found to constitute an offensive touching-but not violent force. Castleman , 134 S.Ct. at 1414-15. An offensive touching may occur through poisoning, since the poison is a "concrete bod[y]" that comes into contact with the victim. Id. at 1410 n.2. But poisoning does not necessitate violent force, in the sense that it is not "moving, acting, or characterized, by physical force, esp. by extreme and sudden or by unjust or improper force; furious; severe; vehement." (Curtis) Johnson , 559 U.S. at 140, 130 S.Ct. 1265 (quoting Webster's New International Dictionary 986) (internal punctuation omitted). And poisoning does not comport with the conception of violent force evinced by the Castleman Court's opinion. In an illustrative passage, the Castleman majority explained that a lower force threshold was appropriate in the domestic violence context because it is "hard to describe as 'violence' a squeeze of the arm that causes a bruise... But an act of this nature is easy to describe as domestic violence." 134 S.Ct. at 1412 (quoting Flores v. Ashcroft , 350 F.3d 666, 670 (2003) ) (internal punctuation omitted). It would seem that poisoning requires even less violence than such an abusive squeeze. And as we have recognized, poisoning simply does not require physical force, as it is generally understood. See Chrzanoski , 327 F.3d at 195.
My colleagues attempt to solve this problem by positing that the severity of the consequence is what demonstrates that the force is sufficiently violent. Op. at 130 n.6. In their view, the Connecticut statute's requirement that the "substance" be "capable of causing death or serious physical injury" suffices to show violent force. Id. And of course they are correct that poison may cause severe consequences, including *136death. However, this view, of force as defined by consequence, overlooks the fact that the Castleman majority-not once, but twice-expressly declined to find that causation of injury could demonstrate violent force. 134 S.Ct. at 1413, 1414. And the Castleman Court further reaffirmed that it viewed violent force as "suggest[ing] a category of violent, active crimes," id. at 1411 n.4, which decidedly does not describe poisoning.
Accordingly, given the distinct statutory setting of Castleman , and the Court's repeated assertions that common law force is a distinct, lesser type of force than the violent force we must find here, I cannot agree with my colleagues that Castleman can be read to decide this case. Where the Supreme Court repeatedly insists that it is analyzing a specific type of force, we should heed their instruction and take care to observe that limitation. We certainly should not take their carefully cabined decision and tack it on to the specific statutory setting the Court distinguished, as my colleagues have done here.
II. Heeding Castleman 's Express Limitations, Its Rationale Should be Read to Accord with Circuit Precedent
There is another, larger problem with extending Castleman in the manner employed by my colleagues. Castleman 's willingness to infer common-law force from the causation of bodily injury sits in tension with Circuit precedent. Where a statute requires only causation of injury, there is no "element" of force, as we have previously understood that term. Again, I readily concede that my colleagues are not unreasonable in their methodology- Castleman 's holdings (if stripped of the careful limitations to statutory context I discuss above) may lend some weight to the notion that we may find an "element" of force where an injury has been caused. But this approach is out of step with the plain command of the elements clause, and our prior case law. In light of Castleman 's persistent cabining of its inquiry, we should not read that case to abrogate our authority in other contexts. Accordingly, my colleagues' interpretation-that Castleman mandates that "the inquiry as to 'force,' for federal law purposes, focuses on the causation of a consequence," 128-represents a fundamental shift in our Circuit's practice, which I believe is in error.
Our Circuit's prior case law reflects the specific inquiry we must perform here. ACCA's elements clause, like many similarly phrased elements clauses, requires that a qualifying statute have "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). Typically, for offenses created by statute, rather than common law, we ascertain the elements from the text of the statute itself. See Descamps v. United States , 570 U.S. 254, 261, 133 S.Ct. 2276, 186 L.Ed.2d 438 (2013) (under categorical approach, "courts may look only to the statutory definitions-i.e., the elements-of a defendant's prior offenses...") (internal quotation marks omitted); see also Mathis v. United States , --- U.S. ----, 136 S.Ct. 2243, 2248-49, 195 L.Ed.2d 604 (2016) ("Elements are the constituent parts of a crime's legal definition-the things the prosecution must prove to sustain a conviction.") (internal quotation marks omitted).
The portion of the Connecticut first degree assault statute that underlies Villanueva's conviction requires that an individual, "[w]ith intent to cause serious physical injury to another person ... causes such injury to such person or to a third person by means of a deadly weapon or a dangerous *137instrument." Conn. Gen. Stat. Ann. § 53a-59(a)(1). The statute makes no reference to force of any kind. It speaks solely in terms of causation. A prosecutor need not prove "force" to a jury to secure a conviction-much less violent force.
Accordingly, if we look only to the text of the statute, it would seem that the crime does not have, "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 some instances, the Supreme Court has indicated that this type of textual inquiry, of looking to the explicit elements of the offense alone, is indeed how elements clause determinations should be performed. See, e.g. , Torres v. Lynch , --- U.S. ----, 136 S.Ct. 1619, 1629, 194 L.Ed.2d 737 (2016) (noting that elements clause of another "crime of violence provision would not pick up demanding a ransom for kidnapping [because] 18 U.S.C. § 875(a)... defin[es] the crime without any reference to physical force") (internal punctuation omitted).
However, in applying the categorical approach, we often do not use a strictly textual method, but also analyze whether force inheres in the crime by looking to the minimum conduct covered by the statute. See, e.g. , Martinez v. Mukasey , 551 F.3d 113, 120 (2d Cir. 2008). This focus on the potential conduct of conviction perhaps has the salutary effect of avoiding an overly formal inquiry, where the word "force" must literally appear in the statute. But it must be noted that by looking to minimum conduct, we are already moving a step beyond what the plain text of the elements clause would seem to envision-namely, locating the elements in the text of the statute, where they have been traditionally found, to see whether those elements include physical force.
Further, defining minimum conduct is often a trying task. While the "focus on the minimum conduct criminalized by the ... statute is not an invitation to apply legal imagination to the ... offense," Moncrieffe v. Holder , 569 U.S. 184, 191, 133 S.Ct. 1678, 185 L.Ed.2d 727 (2013) (internal quotation marks omitted), some degree of abstraction is inherent in this exercise, since we are looking beyond the text of the statute to the real world conduct it might sweep in. In this specific case, all agree that the minimum conduct required for conviction would be poisoning-but not all statutes are so obvious.
Indeed, in some instances, the quest to define minimum conduct has led courts to all but abandon the statutory text, and consider only applications of the statute that arise in the case law. Under this approach, courts will refuse to consider a potential type of minimum conduct permitted by the statutory text, unless a defendant finds a case illustrating that the statute has been used by prosecutors in that manner. See, e.g. , Hill , 890 F.3d at 57 n.9, 58-59 (declining to consider potential nonviolent means of putting an individual in "fear of injury" as permitted by text of statute of conviction, citing lack of case law). This approach has the virtue of simplicity for courts: absent a case on point, a particular application of a statute will be discarded. But in a system where more than ninety percent of all convictions are the product of plea agreements, we can hardly expect a judicial opinion to have issued on each available fact pattern. See Lafler v. Cooper , 566 U.S. 156, 170, 132 S.Ct. 1376, 182 L.Ed.2d 398 (2012) ("Ninety-seven percent of federal convictions and ninety-four percent of state convictions are the result of guilty pleas."). A defendant's inability to find a case illustrating a particular type of conviction does not necessarily indicate that such cases do not exist-rather, it may well reflect the fact that finding such a case would require onerous *138and unwieldy research into the filings of individual cases, to ascertain whether the particular facts might fit the mold obviously encompassed by the statutory language.2 See Ramos v. U.S. Atty. Gen. , 709 F.3d 1066, 1072 (11th Cir. 2013) (explaining that defendant need not point to case law illustrating specific types of prosecution "when the statutory language itself, rather than the application of legal imagination to that language, creates the realistic probability that a state would apply the statute to conduct beyond the generic definition") (internal quotation marks omitted).3
This is all to say that elements clause inquiries are complex, and have led courts to embrace methodologies that are far afield from what the elements clause itself would seem to command. It should be good news for us then, that with specific regard to statutes phrased in terms of causation of bodily injury, such as the one we confront here, we have previously considered the fair inferences that may be drawn from the statutory text. In Chrzanoski , we recognized that "intentional causation of injury does not necessarily involve the use of force." 327 F.3d at 195. This is because a doctor that withholds medicine may cause bodily injury without applying force of any kind, see id. at 196, as may a neglectful parent that fails to provide nourishment to her child.
However, in Castleman , the Court found that a statute, phrased similarly to the one we consider here, has an element of force-although as discussed, the Court found only common-law force, rather than the violent force we must locate today. 134 S.Ct. at 1413-14. The statute considered there, much like the statute here, defined an offense in terms of "caus[ing] bodily injury." Id. at 1409. Despite the fact that the statute did not have an explicit, textual element of force, the Court concluded that the statute did have in fact "as an element," common-law force, because "[i]t is impossible to cause bodily injury without applying force in the common-law sense." Id. at 1415.
The Castleman court did not discuss the types of minimum conduct we identified in Chrzanoski -despite the fact, which we recognized in Chrzanoski and which nearly every first year law student knows, that a person may cause injury without resorting even to offensive touching. Instead of acknowledging these potential applications, the minimum conduct that was located by the Castleman Court was poisoning. See id. at 1414-15. Perhaps the neglectful parent example was simply the other side of the line separating "legal imagination," Moncrieffe , 569 U.S. at 191, 133 S.Ct. 1678, *139from conduct that is properly considered in a categorical analysis. Is it hard to see why, exactly, that would be the case: a neglectful parent starving a child would cause bodily injury, and clearly fulfill all the textual requirements of the statute at issue in Castleman .
Most likely, the Castleman Court was deterred from exploring such minimum conduct by its desire to construe misdemeanor crimes of domestic violence under Section 922(g)(9) to encompass statutes phrased in terms of causation. As the Castleman Court noted, "a contrary reading would have rendered § 922(g)(9) inoperative in many States," since the relevant statutes defining misdemeanor crimes of domestic violence are largely phrased in terms of causation. 134 S.Ct. at 1413.
In my view, Castleman 's willingness to infer an element of common-law force, while discarding other minimum conduct plainly encompassed by the statute at hand, underscores the unique nature of that decision-and the caution we should employ in applying it beyond its explicit bounds. Our trepidation should apply a fortiori here, where our Circuit precedent in Chrzanoski stands for the contrary proposition: that, as a general matter, force cannot be inferred from causation of bodily injury. The Castleman Court emphasized that it was dealing with a specific, lesser type of force than the violence force we must find here. And it must be said again: the Castleman Court twice declined to find that causation of bodily injury could show violence force. 134 S.Ct. at 1413, 1414. Accordingly, Castleman 's analytic approach, by its own express instruction, should be read narrowly-and should not be interpreted, as my colleagues would have it, to cabin otherwise binding Circuit precedent.
* * *
Accordingly, I would affirm the judgment of the district court. I readily acknowledge that my colleagues' opinion accords with the prevailing view across the federal courts of how best to perform elements clause determinations in the wake of Castleman . However, in my assessment, this approach ignores the explicit limitations of Castleman , as well as binding Circuit authority. Happily, I note that the Supreme Court has recently granted certiorari in a case implicating some of these same concerns, and we may soon be given more definitive guidance in how best to perform violent force inquiries. See Stokeling v. United States , --- U.S. ----, 138 S.Ct. 1438, 200 L.Ed.2d 716 (2018).
Of course, the categorical approach itself is a rather bizarre exercise, requiring intricate legal inquiry to determine whether one of Villanueva's convictions, for shooting another person in the shoulder, should indeed be considered a crime of violence. However, in my view, the correct application of precedent requires that we affirm the judgment of the district court here.
For these reasons, I respectfully dissent.
Our ruling also instructed the District Court to determine whether Johnson announced a new rule of constitutional law made retroactive to cases on collateral review, a requirement for a successive motion under section 2255, see 28 U.S.C. § 2244(b)(2)(A). By the time the District Court ruled on the second section 2255 motion, the Supreme Court had ruled that Johnson was retroactive. See Welch v. United States , --- U.S. ----, 136 S.Ct. 1257, 1265, 194 L.Ed.2d 387 (2016).
"Taking the Residual Clause out of the equation so substantially alters the framework of ACCA that a Johnson error can hardly be said to be anything other than 'structural error,' which is unamenable to harmless error review." Villanueva , 191 F.Supp.3d at 190.
The District Court recalculated the Guidelines sentencing range because it did not consider any of Villanueva's four prior convictions to be either a "crime of violence" or a "controlled substance offense" for purposes of U.S.S.G. § 2K2.1.
We cite this Johnson case as "(Curtis) Johnson " to distinguish it from the somewhat related later case of Johnson v. United States , --- U.S. ----, 135 S.Ct. 2551, 192 L.Ed.2d 569 (2015), the case invalidating the residual clause. Some courts distinguish the cases by citing the later Johnson case as "Johnson II ." See, e.g., United States v. Herr , 2016 WL 6090714, No. 16-cr-10038-IT, at *1 (D. Mass. Oct. 18, 2016). The use of "II " in case names customarily identifies the second phase of the same litigation. We prefer to distinguish the earlier and less frequently cited Johnson case by using the first name of the petitioner in parentheses. Another way to distinguish the earlier case is to label it (Curtis) Johnson v. United States . See United States v. Vail-Bailon , 868 F.3d 1293, 1297 (11th Cir. 2017) (in banc).
Villanueva cites State v. Gooch , 186 Conn. 17, 438 A.2d 867, 871 (1982), State v. Carter , 141 Conn.App. 377, 61 A.3d 1103 (2013), aff'd, 317 Conn. 845, 120 A.3d 1229 (2013), State v. Holmes, 75 Conn.App. 721, 817 A.2d 689, 702 (2003), and Connecticut's pattern jury instructions for subsection 53a-59(a)(1), 5A Conn. Prac. Jury Instructions § 9.1 (4th ed.).
The holding of Chrzanoski has not been disturbed. First, the defendant's prior offense, which our Court assumed to be third degree assault in violation of Conn. Gen. Stat. 53a-61(a)(1), see 327 F.3d at 192, was a misdemeanor and could not for that reason have qualified as a "violent felony" under ACCA. Second, the misdemeanor offense required intent to cause only "physical injury," not "serious physical injury," as required for Villanueva's first degree assault conviction. That latter element is necessary to make his offense a "violent felony."
Hill was indicted for violating 18 U.S.C. § 924(j)(1) by committing a firearms-related murder in the course of a "crime of violence," as defined in 18 U.S.C. § 924(c)(3). The "crime of violence" was Hobbs Act robbery as defined in 18 U.S.C. § 1951(b)(1). See Hill , 890 F.3d at 54.
We note that last year, this Court issued a summary order that relied on Chrzanoski in determining that a New York conviction for second degree burglary, see N.Y. Penal Law § 140.25 was not a "crime of violence" for purposes of subsection 4B1.2(a) of the Sentencing Guidelines. See United States v. Welch , 641 F. App'x 37, 42 (2d Cir. 2016). Without comparing the Guidelines provision to the ACCA or the New York second degree burglary offense to the Connecticut first degree assault offense, we point out that the summary order did not reckon with Castleman or Hill , nor was either decision cited by the parties in Welch .
The District Court's unawareness of this fact is entirely understandable. This Court's order denying Harris leave to file a successive section 2255 motion did not identify the subdivision of the first degree assault statute under which he had been convicted. See Harris v. United States , No. 15-2679 (2d Cir. Nov. 17, 2015) (order denying leave to file). The Government's brief in the pending appeal also reveals unawareness that Harris had been convicted of assault by means of a firearm under Conn. Gen. Stat. § 53a-59(a)(5). This fact is revealed in the parties' papers on Harris' motion to file a successive section 2255 motion.
See , e.g. , Fisher v. Vassar College , 114 F.3d 1332, 1336 (2d Cir. 1997) (in banc) ("[T]he term prima facie case, as used in Title VII and ADEA actions, has a meaning that is quite different from and more limited than that ascribed to the term in many other actions."); id. at 1365 (Newman, C.J., with whom Kearse, Winter, and Cabranes, JJ., join, dissenting) ("[T]he majority nonetheless asserts that 'prima facie case' in Title VII has a third meaning.
The record does not disclose what the actual consequences of such a sentence would be for Villanueva. From the date of the original sentencing, February 1, 2000, he had served at least 16 years, 4 months, and 13 days by June 14, 2016 (the date of resentencing and placement on supervised release), a term of imprisonment beyond the ACCA's 15 year minimum. If the original sentence were to be re-imposed, his release date would depend on the length of any presentencing confinement and any good time credits earned during imprisonment.
Because we have remanded for resentencing, we need not determine whether the District Court's pre-Johnson error of using the residual clause in imposing the original sentence was a structural or a harmless error.
The Hill panel's discussion of Castleman was not necessary to its judgment, since the panel addressed Castleman only by assuming arguendo that Hill's argument related to Castleman was relevant, even though the Hill panel explicitly found it was not. 890 F.3d at 59. Thus this discussion is dicta, and not binding on this panel. See Jimenez v. Walker , 458 F.3d 130, 142 (2d Cir. 2006) ("[D]icta are not and cannot be binding. Holdings-what is necessary to a decision-are binding. Dicta-no matter how strong or how characterized-are not.") (internal punctuation omitted).
Indeed, in this very case, the minimum conduct criminalized by the statute-poisoning-does not appear to have generated case law. Nonetheless, it seems from public reports that prosecutors have used the first degree assault statute in poisoning cases. See Suspects in poisoning fight charges in court , NewsTimes (Danbury, CT) (Apr. 19, 2007), http://www.newstimes.com/news/article/Suspects-in-poisoning-fight-charges-in-court-236686.php; Mother Faces Charges Of Poisoning Her Child , Hartford Courant (Dec. 10, 1993), http://articles.courant.com/1993-12-10/news/0000000848_1_methanol-poison-child-affidavit; Woman accused of trying to poison husband 'snapped,' lawyer says , Darien News (Apr. 2, 2007) http://www.dariennewsonline.com/news/article/Woman-accused-of-trying-to-poison-husband-46164.php.
Even if there truly were no cases to have ever been charged under such a fact pattern, this could be attributable more to the preferences of prosecutors than a lack of a legal element. Of course, it is fair to observe that the dearth of published opinions regarding a particular version of offense conduct indicates that such cases are rarely brought. But where the text of a statute is clear, we cannot rely on the forbearance of prosecutors to prevent an offense from qualifying as a crime of violence.