Khaimraj Singh, a Guyanan national, petitions for review of a final order of removal grounded upon the determination that he is an aggravated felon on account of his conviction for touching the breast of his cousin, who was under sixteen years of age. The offense of conviction was 11 Del. C. § 767, "Unlawful sexual contact in the third degree," which provides:

A person is guilty of unlawful sexual contact in the third degree when the person has sexual contact with another person or causes the victim to have sexual contact with the person or a third person and the person knows that the contact is either offensive to the victim or occurs without the victim's consent.

The question presented on this petition for review-whether Singh has been convicted of the aggravated felony of "sexual abuse of a minor," 8 U.S.C. § 1101(a)(43)(A)-turns on whether we must apply the so-called "formal categorical approach" announced in Taylor v. United States, 495 U.S. 575, 110 S. Ct. 2143, 109 L. Ed. 2d 607 (1990). Under that approach, an adjudicator "must look only to the statutory definitions of the prior offenses," and may not "consider other evidence concerning the defendant's prior crimes," including, "the particular facts underlying [a] conviction[ ]." Id. at 600, 110 S. Ct. 2143. If we apply the formal categorical approach, Singh has not been convicted of the aggravated felony of sexual abuse of a minor because § 767 does not contain an element specifying the age of the victim. If we do not apply the formal categorical approach, Singh has been convicted of the aggravated felony of sexual abuse of a minor because the victim of his sex offense was, indeed, a minor.

Our jurisprudence in the aggravated felony area-twelve cases in all-is not a seamless web. In order to resolve the appeal we have found it necessary to analyze and synthesize this body of case law, and we do so at length, see infra Part III.B. As will appear, a pattern emerges, causing us to conclude that, while the formal categorical approach of Taylor presumptively applies in assessing whether an alien has been convicted of an aggravated felony, in some cases the language of the particular subsection of 8 U.S.C. § 1101(a)(43) at issue will invite inquiry into the underlying facts of the case, and in some cases the disjunctive phrasing of the statute of conviction will similarly invite inquiry into the specifics of the conviction. But in this case, neither 8 U.S.C. § 1101(a)(43)(A) nor 11 Del. C. § 767 invite inquiry into the facts underlying Singh's conviction. Therefore, because Taylor' s formal categorical approach applies to Singh's case, we will grant the petition for review.

I. Factual Background and Proceedings Before the Immigration Judge and Board of Immigration Appeals

Singh is a native and citizen of Guyana. He was admitted to the United States in June 1988 as an immigrant. About ten years later, he touched the breast of his cousin, who was under the age of sixteen. For this, the State of Delaware charged him under 11 Del. C. § 768, "Unlawful sexual contact in the second degree." Apparently as part of a plea agreement with the state prosecutor, Singh pled guilty to the lesser included offense of 11 Del. C. § 767, "Unlawful sexual contact in the third degree." On May 19, 1998, the Delaware Superior Court imposed a one-year suspended sentence.

This conviction, the government asserts, renders Singh an aggravated felon under 8 U.S.C. § 1101(a)(43)(A), which provides that "sexual abuse of a minor" is an aggravated felony.1 Under 8 U.S.C. § 1227(a)(2)(A)(iii), an alien convicted of an aggravated felony is removable. The Immigration Judge (IJ) agreed with the government's position, noting that "the sentencing order of the [Delaware] Court reflects clearly, under special conditions of probation, `note: victim is under 16 years of age.'" Citing 18 U.S.C. § 3509(2) (which defines the age of majority as 18), the IJ explained that "the victim was under 16 years of age and, consequently, would be classified as a minor." The IJ then looked to 18 U.S.C. § 3509(8) for the definition of "sexual abuse," which includes "sexually explicit conduct":

The term sexually explicit conduct includes touching of one's breast under [18 U.S.C. § 3509(9)(A)]. Consequently, the Court finds that the respondent has engaged in sexually explicit conduct of a child. Likewise, the Court would find that the respondent's conviction, notwithstanding the fact that the age of the victim is not specifically designated in the statute, has indeed ... engaged in sexual abuse of a minor as defined in Title 18.

The Board of Immigration Appeals (BIA) affirmed the IJ's decision without opinion. See 8 C.F.R. § 1003.1(e)(4).2 Under 8 U.S.C. § 1252(a)(2)(C) and our decision in Drakes v. Zimski, 240 F.3d 246, 247 (3d Cir.2001), we have jurisdiction to consider our jurisdiction over this timely petition for review of a final decision of the BIA.

This case turns on a question of statutory interpretation-specifically, the meaning and application of the aggravated felony of "sexual abuse of a minor." As we noted in Patel v. Ashcroft, 294 F.3d 465, 467 (3d Cir.2002), "there is some confusion surrounding the proper standard of review in cases such as this." Patel, which was an aggravated felony case, discusses at length the role of Chevron deference in cases interpreting the Immigration and Nationality Act (INA) generally, and the aggravated felony statute of 8 U.S.C. § 1101(a)(43) in particular. In our most recent aggravated felony case, we described the scope of our Chevron deference thus:

"The first step in interpreting a statute is to determine `whether the language at issue has a plain and unambiguous meaning with regard to the particular dispute in the case.'" [Valansi v. Ashcroft, 278 F.3d 203, 209 (3d Cir.2002)] (quoting Marshak v. Treadwell, 240 F.3d 184, 192 (3d Cir.2001)). If the statutory meaning is clear, our inquiry is at an end. Id. If the statutory meaning is not clear, we must try to discern Congress' intent using the ordinary tools of statutory construction. See INS v. Cardoza-Fonseca, 480 U.S. 421, 447-48, 107 S. Ct. 1207, 94 L. Ed. 2d 434 (1987). "If, by employing traditional tools of statutory construction, we determine that Congress' intent is clear, that is the end of the matter." Valansi, 278 F.3d at 208 (quoting Bell v. Reno, 218 F.3d 86, 90 (2d Cir.2000)). If we are unable to discern Congress' intent using the normal tools of statutory construction, we will generally give deference to the Board's interpretation, so long as it is reasonable. Id.

Lee v. Ashcroft, 368 F.3d 218, 222 (3d Cir.2004).

Canvassing the dozen aggravated felony cases decided by this Court, one indisputable and surprising pattern emerges: We have never affirmatively deferred to an interpretation by the BIA (or an IJ) of 8 U.S.C. § 1101(a)(43), i.e., of whether the crime at issue constitutes an aggravated felony. Many times we have not even discussed Chevron deference to the BIA, irrespective of whether we ultimately agreed or disagreed with the Board. See Munroe v. Ashcroft, 353 F.3d 225 (3d Cir.2003); Wilson v. Ashcroft, 350 F.3d 377 (3d Cir.2003); Bovkun v. Ashcroft, 283 F.3d 166 (3d Cir.2002); United States v. Graham, 169 F.3d 787 (3d Cir.1999). We also have suggested that we conduct de novo review because the question goes to our jurisdiction. See Nugent v. Ashcroft, 367 F.3d 162, 165 (3d Cir.2004); Valansi, 278 F.3d at 207-08 (citing cases). Twice we have declined to reach the question of deference because we concluded that our result would be the same on deferential review as it would on plenary review. See Patel, 294 F.3d at 468; Drakes v. Zimski, 240 F.3d 246, 251 (3d Cir.2001). In two cases, we acknowledged that we must defer to the BIA if the statute's meaning is ambiguous, but both times we held that the BIA's interpretation conflicted with the statute's plain meaning. See Lee, 368 F.3d at 224-25; Valansi, 278 F.3d at 208. In yet another case we affirmatively held that the BIA's interpretation of 18 U.S.C. § 16, which is incorporated by reference in the aggravated felony statute, 8 U.S.C. § 1101(a)(43)(F), was not subject to the general principles of Chevron, but that, even if it was, the specific interpretation at issue was unreasonable and therefore not entitled to deference. See Francis v. Reno, 269 F.3d 162, 168 & n. 8 (3d Cir.2001). Finally, two decisions appear to be more deferential to the BIA. In Gerbier v. Holmes, 280 F.3d 297, 310 (3d Cir.2002), we found the BIA's interpretation "persuasive" (hardly a strong general endorsement), while in Steele v. Blackman, 236 F.3d 130, 133 (3d Cir.2001), we stated that "if a statute administered by the INS is ambiguous, and the BIA has provided a reasonable interpretation of its language, we must simply ask whether the BIA's construction is a permissible one." Even in Steele, however, we looked more closely into the BIA's interpretation, finding it "troublesome," but we "assume [d] its validity" because even the BIA's own interpretation of the statute did not support its disposition of the case. 236 F.3d at 136 & n. 5.

Why then have we never found it necessary and appropriate to defer to the BIA's or IJ's interpretation of 8 U.S.C. § 1101(a)(43)? First, as we explained in Francis, the interpretation and exposition of criminal law is a task outside the BIA's sphere of special competence. See 269 F.3d at 168; see also Drakes, 240 F.3d at 250 ("Chevron [U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837, 104 S. Ct. 2778, 81 L. Ed. 2d 694 (1984)] deference is not required where the interpretation of a particular statute does not `implicate[ ] agency expertise in a meaningful way' but presents instead `a pure question of statutory construction for the courts to decide.'" (quoting Sandoval v. Reno, 166 F.3d 225, 239-40 (3d Cir.1999) (quoting INS v. Cardoza-Fonseca, 480 U.S. 421, 446, 107 S. Ct. 1207, 94 L. Ed. 2d 434 (1987))) (alteration in original)). Second, we have been mindful, as in Nugent, 367 F.3d at 165, and Valansi, 278 F.3d at 207-08, that although the statute is part of Title 8, and not Title 28, of the United States Code, it nonetheless controls our jurisdiction (via 8 U.S.C. § 1252(a)(2)(C)) and we normally consider jurisdictional matters de novo. Moreover, here the IJ offered no reason for his decision not to apply Taylor' s categorical approach; the BIA, by affirming without opinion, gave no considered and authoritative agency-wide interpretation of the statute; and now on petition for review, the government's entire position on deference consists of a single citation to an admittedly vague comment from this Court in Patel, 294 F.3d at 467 ("[S]ome deference is still required under Chevron, even though we are reviewing a purely legal question such as the BIA's interpretation of a criminal statute."). Under all these circumstances, we conclude that the IJ's summary application of § 1101(a)(43)(A)-it can hardly be described as a full-blown reasoned interpretation-is not entitled to deference. As we have done in previous cases, however, we will here expressly reserve decision on whether some BIA interpretations of § 1101(a)(43) are entitled to deference.

Some of our cases interpreting 8 U.S.C. § 1101(a)(43) have employed the rule of Taylor, described there as the "formal categorical approach," 495 U.S. at 600, 110 S. Ct. 2143. See, e.g., Francis, 269 F.3d at 171-72. Taylor addressed the meaning and application of the term "burglary" in 18 U.S.C. § 924(e), which imposes enhanced sentences on defendants convicted under the unlawful-possession-of-a-firearm statute, 18 U.S.C. § 922(g), when the defendant has three prior convictions for specified offenses, including "burglary." Taylor' s analysis can be readily imported here, because 8 U.S.C. § 1101(a)(43) is similar to 18 U.S.C. § 924(e) in that it too enumerates offenses, conviction of which places an alien in the category of "aggravated felon." Taylor put the "general issue" in interpreting this sort of statute as follows:

whether the sentencing court in applying § 924(e) must look only to the statutory definitions of the prior offenses, or whether the court may consider other evidence concerning the defendant's prior crimes. The Courts of Appeals uniformly have held that § 924(e) mandates a formal categorical approach, looking only to the statutory definitions of the prior offenses, and not to the particular facts underlying those convictions.

495 U.S. at 600, 110 S. Ct. 2143. Citing "the practical difficulties and potential unfairness [to a defendant] of a factual approach," id. at 601, 110 S. Ct. 2143, the Taylor Court adopted the "formal categorical approach."

The facts of Taylor provide an apt illustration of the principle at work: Taylor had been twice convicted of second degree burglary in Missouri. Under Missouri law, second degree burglary encompassed several discrete sets of statutory elements. As the Court explained, "All seven offenses required entry into a structure, but they varied as to the type of structure and the means of entry involved." Id. at 578 n. 1, 110 S. Ct. 2143. Compared with the definition of "generic burglary" adopted by the Court-"convict[ion] of any crime, regardless of its exact definition or label, having the basic elements of unlawful or unprivileged entry into, or remaining in, a building or structure, with intent to commit a crime," id. at 599, 110 S. Ct. 2143-not all variants of Missouri second degree burglary qualified under the federal sentencing enhancement statute. Since the formal categorical approach does not permit looking beyond the literal elements of the statute (i.e., to the facts supporting the convictions), the Court could not say that Taylor had been previously convicted of crimes encompassing the elements of generic burglary. See id. at 602, 110 S. Ct. 2143. Accordingly, the Court remanded for further development on the question of "which of [the Missouri second degree burglary] statutes were the bases for Taylor's prior convictions." Id. The inquiry was limited to the statute of conviction, however, as the formal categorical approach "generally requires the trial court to look only to the fact of conviction and the statutory definition of the prior offense." Id. The Taylor Court also acknowledged that, under limited circumstances, resort to the charging instrument may be permissible: "[The formal categorical approach] may permit the sentencing court to go beyond the mere fact of conviction in a narrow range of cases where a jury was actually required to find all the [necessary] elements." Id.

The questions presented in this case are readily apparent: Does Taylor' s formal categorical approach apply to "sexual abuse of a minor" under 8 U.S.C. § 1101(a)(43)(A), and if it does, does a conviction under 11 Del. C. § 767 nonetheless qualify as a conviction for sexual abuse of a minor? We find it more efficient to address the second question first.

A. Under the Formal Categorical Approach of Taylor, Does a Conviction Under 11 Del. C. § 767 Qualify as a Conviction for "Sexual Abuse of a Minor"?

The government first argues that, even under the formal categorical approach, a conviction under 11 Del. C. § 767 is a conviction for sexual abuse of a minor. To evaluate this argument, we follow the Supreme Court's two-step approach in Taylor, where it first construed the term "burglary" in the federal statute, 495 U.S. at 590-99, 110 S. Ct. 2143, and then compared the elements of the Missouri statutes to the federal definition, id. at 602, 110 S. Ct. 2143. The IJ looked to 18 U.S.C. § 3509(2), (8), and (9) to define "sexual abuse of a minor." While we have no quarrel with this approach, it is unnecessary for our purposes to give a comprehensive definition of "sexual abuse of a minor," and we reserve decision on that question. It is sufficient to say that "sexual abuse of a minor" entails some conduct involving a minor, i.e., someone under the age of eighteen.

This is enough to defeat the government's argument, for § 767 says nothing whatsoever about the age of the victim. In full, the statute under which Singh was convicted reads:

A person is guilty of unlawful sexual contact in the third degree when the person has sexual contact with another person or causes the victim to have sexual contact with the person or a third person and the person knows that the contact is either offensive to the victim or occurs without the victim's consent.

Since a finding of the age of the victim is not required for conviction, § 767 does not appear to be an aggravated felony (or at least not the aggravated felony of sexual abuse of a minor).

In the face of the literal and unambiguous text of § 767, the government argues that the statute nonetheless criminalizes sexual abuse of a minor. Of course it is irrelevant that sexually abusing a minor may be sufficient for conviction under the statute; what matters is whether such conduct is necessary for such a conviction. The government seems to argue that the overall statutory scheme in Delaware establishes that sexual abuse of a minor is necessary for a conviction under § 767. Even assuming that appeals to statutes other than the statute of conviction are within the bounds of the formal categorical approach, we still cannot agree with the government's position.

10 Del. C. § 922(a)(19) provides the Delaware Family Court with exclusive original criminal jurisdiction over "unlawful sexual conduct in the third degree against a child under 11 Del. C. § 767." Similarly, 11 Del. C. § 1112(b)(4)(A) defines "sexual offender" as a person who has been convicted of "any sexual offense upon a child under 16 years of age under § 767, § 768, § 769, § 770, § 771, § 772, § 773...." The government would have us read these provisions as definitions of § 767. We do not agree. There are at least two problems with the government's approach. First, "against a child" and "upon a child," in 10 Del. C. § 922(a)(19) and 11 Del. C. § 1112(b)(4)(A), respectively, are better read not as definitions of Del.Code Ann. tit. 11, § 767, but rather as qualifications or limitations on the § 767 convictions that are intended to be within the scope of the jurisdictional and sex offender statutes. Unless these phrases are read as qualifiers, they are superfluous-why not, in § 1112(b)(4)(a), simply omit "upon a child under 16 years of age" if the statutes referred to already incorporate such a notion?

A second and independent problem with the government's proposed reading is that it leads to absurdities elsewhere in the statutory scheme. For example, 11 Del. C. §§ 767-773 are the gamut of traditional sex offenses under Delaware law, from misdemeanor sexual assault to first degree rape. Adopting the government's argument would imply not only that § 767 is an offense against a child, but also that §§ 768-773 are as well. This would have the astonishing result that all sex offenses in Delaware (as Delaware does not define any general sex offenses outside this list) require that the victim be a child. These problems only reinforce our conclusion from the plain meaning of the statute: 11 Del. C. § 767 does not include as an element that the victim be a minor, and accordingly, under the formal categorical approach of Taylor, Singh does not stand convicted of sexual abuse of a minor.

B. Does the Formal Categorical Approach of Taylor Apply to the Aggravated Felony of "Sexual Abuse of a Minor"?

Because the IJ's decision cannot stand if we apply Taylor' s formal categorical approach, we must turn to the government's fallback argument that the formal categorical approach does not apply to "sexual abuse of a minor" under 8 U.S.C. § 1101(a)(43)(A). If the government is free from the strictures of the formal categorical approach, we would simply review the IJ's decision under the deferential substantial evidence standard, evaluating whether the factual record before the IJ could fairly support the conclusion that Singh was convicted of sexual abuse of a minor. See Dia, 353 F.3d at 247-49 (describing the substantial evidence standard). As the record supports this conclusion-indeed, Singh concedes that, as a factual matter, he was convicted for touching the breast of his minor cousin-we would dismiss the petition if the formal categorical approach did not apply here. But for the reasons that follow, we conclude that the formal categorical approach does apply.

1. Our prior aggravated felony jurisprudence

We have decided a dozen cases implicating 8 U.S.C. § 1101(a)(43), and at times we have applied the formal categorical approach of Taylor, and at other times we have not. (In the latter instances, though we have never explicitly considered and rejected Taylor' s approach, our decisions cannot be fairly read as employing the formal categorical approach.) The question here, then, is whether sexual abuse of a minor under 8 U.S.C. § 1101(a)(43)(A) is more akin to those provisions of 8 U.S.C. § 1101(a)(43) to which we have applied the formal categorical approach, or to those to which we have not. We begin with a comprehensive survey of this Court's aggravated felony jurisprudence. The United States Supreme Court has not decided any case involving 8 U.S.C. § 1101(a)(43), so we consider only our own cases in this survey. Three of the twelve cases we have decided, see supra Part II, do not even implicitly involve the question whether to use Taylor' s formal categorical approach.3 Thus, in the sections that follow, we address the other nine cases, which all (explicitly or implicitly) take a position on Taylor' s applicability.

We expressly invoked and applied Taylor' s formal categorical approach in Francis. There, the question presented was whether "a state misdemeanor conviction for vehicular homicide is a `crime of violence' within the meaning of 18 U.S.C. § 16," 269 F.3d at 164, which is incorporated by reference in the definition of "aggravated felony," 8 U.S.C. § 1101(a)(43)(F). We concluded that the petitioner's conviction was not a crime of violence under 18 U.S.C. § 16, and therefore that he was not an aggravated felon. Although we were able to reach this result on an analysis of 18 U.S.C. § 16 that did not implicate Taylor, Francis, 269 F.3d at 168-71, we also held in the alternative that, under the formal categorical approach, Francis's conviction did not satisfy 18 U.S.C. § 16(b)'s requirement that a crime of violence be one that, "by its nature, involves a substantial risk that physical force against the person or property of another may be used in the course of committing the offense."

We explained that under the formal categorical approach,

we must look to Pennsylvania's definition of homicide by vehicle.... 75 Pa.C.S.A. § 3732 provides:

Any person who unintentionally causes the death of another person while engaged in the violation of any law of this Commonwealth or municipal ordinance applying to the operation or use of a vehicle or to the regulation of traffic except section 3731 (relating to driving under influence of alcohol or controlled substance) is guilty of homicide by vehicle, a misdemeanor of the first degree, when the violation is the cause of death.

75 Pa.C.S.A. § 3732.

On its face, homicide by vehicle is certainly not an offense that "by its nature, involves a substantial risk that physical force against the person or property of another may be used in the course of committing the offense." 18 U.S.C. § 16(b). The BIA acknowledged that § 3732 involves a range of behavior that "may or may not" fall under § 16(b).

The categorical approach does "permit the sentencing court to go beyond the mere fact of conviction in a narrow range of cases where a jury was actually required to find all the elements of [the relevant] generic [offense]." Taylor, 495 U.S. at 602, 110 S. Ct. 2143. Here, the criminal complaint stated:

Southbound on Route 95 in the vicinity of Comly Street the defendant unintentionally caused the death of the decedent # 1 Harry B. Rutter, Driver of vehicle # 1, by operating a 1985 Chevrolet Caprice, Pa. License ADB 7268, while his operating privilege was suspended, and in such a manner as to cause a eight vehicle accident between four cars, one van, and three tractor trailers and a near miss by a tanker truck carrying 8000 gallons of gasoline, causing the deaths of two people and injuring a third.

(emphasis added).

Francis was therefore charged with the "unintentional" conduct, of operating an automobile in such a manner as to cause a car accident resulting in two deaths.

Id. at 172 (some citations omitted).

As Francis's predicate "violation of [the] law" for purposes of 75 Pa. Cons.Stat. § 3732 was driving with a suspended license, and not something that "by its nature, involves a substantial risk [of] physical force," 18 U.S.C. § 16(b), we could not agree with the BIA that his offense was a crime of violence. Significantly, we disagreed with the BIA's conclusion that Francis's conduct involved a substantial risk of physical force because it was reckless, stating: "It may well have been [reckless]. However, recklessness was not charged, and he was not convicted of an offense requiring that mens rea." Francis, 269 F.3d at 173. In sum, the BIA reasoned from the facts supporting Francis's conviction, but we felt constrained to grant the petition for review by applying Taylor' s formal categorical approach and confining our inquiry to the statute of conviction, illuminated by the charging instrument.

Three other cases — Steele, Gerbier, and Wilson — apply Taylor through their use of the "hypothetical federal felony" (or "hypothetical federal conviction") approach.4 This method was developed by the BIA in applying 8 U.S.C. § 1101(a)(43)(B), which categorizes as an aggravated felony "illicit trafficking in a controlled substance (as defined in section 802 of Title 21), including a drug trafficking crime (as defined in section 924(c) of Title 18)." In Steele, our first case to examine the hypothetical federal felony approach, we explained it this way:

Section 924(c)(2) of Title 18 defines "drug trafficking crime" as meaning "any felony punishable under the Controlled Substance Act[,] ... the Controlled Substance Import and Export Act[s]... or the Maritime Drug Law Enforcement Act...." Accordingly, the BIA finds within [this] category of aggravated felony convictions any federal conviction for a violation of one of the specified statutes that is a felony conviction under federal law, i.e., a conviction for an offense punishable by imprisonment for over one year. See 18 U.S.C. § 3559. More relevant for present purposes, the BIA understands this ... category to encompass convictions for state offenses, however characterized by the state, if those offenses would be "punishable" under one of the three specified federal statutes if federally prosecuted, so long as the hypothetical federal conviction would be a felony under federal law, i.e., would be punishable by a term of imprisonment of over one year.

This hypothetical federal conviction approach "require[s] a comparison between the elements of the [state] drug offense and [the elements of] a federal drug provision referenced in 18 U.S.C. § 924(c)(2)...." Matter of Davis, 20 I. & N. Dec. 536, 544, 1992 WL 443920 (BIA 1992). Since the basis for the incapacities under the Immigration Act is "convict[ion] of an aggravated felony," 8 U.S.C. § 1229b(a), the Board looks to what the convicting court must necessarily have found to support the conviction and not to other conduct in which the defendant may have engaged in connection with the offense. Thus where, as here, the Service is relying on a state misdemeanor conviction, the requirements of this ... category of "aggravated felony convictions" are "satisfied [only] by proving a conviction that includes all the elements of [a felony] offense for which an alien `could be convicted and punished' under the cited federal laws." Matter of Barrett, 20 I. & N. Dec. 171, 174, 1990 WL 385754 (BIA 1990).

Steele, 236 F.3d at 135-36 (some alterations in original). The hypothetical federal felony approach is essentially the formal categorical approach of Taylor, as applied to a specific federal statute.

Though we did not actually approve the hypothetical federal felony approach in Steele, we accepted it arguendo, because even it did not support the BIA's disposition-the proposed hypothetical federal felony required a finding of an additional, prior drug conviction, a prior conviction that, though existing in fact, had not been proven in the course of Steele's state criminal proceedings. Id. at 137. We thus granted Steele's petition for review. A little over a year later, in Gerbier, we did adopt the BIA's hypothetical federal felony approach to 8 U.S.C. § 1101(a)(43)(B). 280 F.3d at 308-11. But as in Steele, Gerbier's status as a recidivist had not been litigated or otherwise decided in his state criminal proceeding, and we therefore granted his petition for review. Id. at 317.

In Wilson, 350 F.3d 377, the third case in our hypothetical-federal-felony trilogy, we again granted the petition for review. There, the proposed hypothetical federal felony-21 U.S.C. § 841(a)(1), which is the general federal felony criminal prohibition on unauthorized manufacture, distribution, and possession with intent to distribute of controlled substances-included an escape clause making distribution of "a small amount of marihuana for no remuneration" a misdemeanor. See 21 U.S.C. § 841(b)(1)(D) and (b)(4). Wilson had possessed with the intent to distribute a small amount of marijuana, a misdemeanor under N.J. Stat. Ann. § 2C:35-5(b)(11).5 Wilson, 350 F.3d at 381. He argued to this Court that, because his state conviction did not necessarily imply that he sought remuneration for his distribution, it could not therefore be shown that his conduct would have been a felony under federal controlled substance law-he might have been able to invoke the escape clause. Id.

The government countered that Wilson's conviction was for possession, not distribution (though both distribution and possession-with-intent were included in New Jersey's law, see supra note 5), and that he therefore would not have been eligible for the escape clause. We rejected the government's invitation to look beyond the New Jersey statute itself, citing Steele and Gerbier for the proposition that "in evaluating whether a state violation is analogous to a federal felony, we look to the elements of the statutory state offense, not to the specific facts." Wilson, 350 F.3d at 381. "Since the state statutory elements would be satisfied by proof of either distribution or possession with intent to distribute, we cannot draw the federal analogy by presuming that the statute only covers possession." Id. at 382. Wilson may thus represent the zenith of our faithfulness to Taylor.

Two other cases also follow Taylor' s formal categorical approach-though only silently (in the case of Bovkun) or weakly (in the case of Drakes). Although we did not cite Taylor in Bovkun, we plainly followed the formal categorical approach. There, the petitioner had been convicted of making terroristic threats under Pennsylvania law,6 and the government sought to classify him as an aggravated felon under 8 U.S.C. § 1101(a)(43)(F) (incorporating 18 U.S.C. § 16 by reference), for committing "a crime of violence."7 Bovkun argued that "mere public inconvenience" would not qualify as a crime of violence, but we rejected his argument because it confused the actus reus of the offense ("threat[ ] to commit a crime of violence") and the mens rea ("with intent to ... or reckless disregard of ..."). Bovkun, 283 F.3d at 170. We held that it was the actus reus of the state offense that had to be aligned with the federal statute, and on that basis we concluded that a Pennsylvania conviction for making terroristic threats was a crime of violence as defined in 18 U.S.C. § 16.

Finally, in Drakes, we considered 8 U.S.C. § 1101(a)(43)(R), which classifies as an aggravated felony "an offense relating to ... forgery." Though the facts were not entirely clear, the petitioner had been convicted of second-degree forgery under Delaware law in connection with providing a false name to the Delaware State Police during a traffic stop.8 As noted above, see supra Part III.A (discussing Taylor' s two-step approach), Taylor requires both interpretation of the federal statute describing the offense, and a comparison with the statute of criminal conviction. Though most of our aggravated felony cases have turned on the second step, Drakes concentrated more on the first. In Drakes, we discussed at length the meaning of "forgery" in 8 U.S.C. § 1101(a)(43)(R). See 240 F.3d at 248-50. Upon determining that "Congress evidenced an intent to define forgery in its broadest sense," id. at 249, it became easy to conclude that 11 Del. C. § 861 came within the wide sweep of the offenses described in 8 U.S.C. § 1101(a)(43)(R), see Drakes, 240 F.3d at 250. Thus Drakes only briefly touched on Taylor' s second step-the formal categorical approach as we have been discussing it.

We turn now to the cases in which we did not confine ourselves to the formal categorical approach of Taylor. All three such cases — Nugent, Munroe, and Valansi — concerned 8 U.S.C. § 1101(a)(43)(M)(i), which defines as an aggravated felony an offense that "involves fraud or deceit in which the loss to the victim or victims exceeds $10,000." In all three cases, the relevant criminal statute did not include a "loss greater than $10,000" element. See Nugent, 367 F.3d at 168 n. 2 (quoting 18 Pa. Cons.Stat. § 3922(a)9); Munroe, 353 F.3d at 226 (citing N.J. Stat. Ann. 2C:20-410); Valansi, 278 F.3d at 210 (quoting 18 U.S.C. § 65611).12 Yet in these cases we expressly rested our holding on the underlying facts about the amount of loss involved: In Nugent, 367 F.3d at 169, the bad check at issue was in the amount of $4831.26 (and thus insufficient to support the aggravated felony classification); in Munroe, 353 F.3d at 226, several bad checks written by the petitioner totaled in excess of $10,000 (and thus were sufficient to support the aggravated felony classification).

Valansi, in which the petitioner had embezzled over $400,000 in cash and checks in her capacity as a bank teller, 278 F.3d at 205, bears further discussion. The monetary threshold was clearly reached in Valansi. See 278 F.3d at 209 ("Valansi does not dispute that her conviction satisfies the $10,000 monetary requirement."). The case turned instead on whether Valansi's crime "involve[d] fraud or deceit." In an extensive discussion of 18 U.S.C. § 656, the Court focused on the mens rea requirement that had been judicially imposed on the statute: The embezzler must do so "with the intent to injure or defraud the bank." Valansi, 278 F.3d at 210 (citing United States v. Schoenhut, 576 F.2d 1010, 1024 (3d Cir.1978) (citing United States v. Schmidt, 471 F.2d 385 (3d Cir.1972))). We held that this disjunctive mens rea requirement-either intent to defraud or intent to injure suffices-put the statute with one foot in 8 U.S.C. § 1101(a)(43)(M)(i) and one foot out:

The plain meaning of § 1101(a)(43)(M)(i) defines an aggravated felony as an offense that has fraud or deceit as at least one required element. Some but not all convictions under 18 U.S.C. § 656 qualify as an aggravated felony under that definition: a conviction for embezzlement with specific intent to defraud qualifies as an offense involving fraud or deceit, and thus an aggravated felony; a conviction with only the specific intent to injure does not.

Valansi, 278 F.3d at 217.

Under a strict reading of Taylor' s formal categorical approach, this would be the end of the story, because a conviction under 18 U.S.C. § 656 does not necessarily establish fraudulent intent any more than Singh's conviction under 11 Del. C. § 767 necessarily establishes sexual abuse of a minor (because, after all, some but not all convictions under 11 Del. C. § 767 involve a minor victim). The Valansi panel went on, however:

We have cautioned that where "a criminal statute on its face fits the INA's deportability classification ... [,][t]o go beyond the offense as charged and scrutinize the underlying facts would change our inquiry from a jurisdictional one into a full consideration of the merits. Such an approach would fly in the face of the jurisdiction limiting language of IIRIRA." Drakes, 240 F.3d at 247-48. However, in this case we have determined that the criminal statute does not fit squarely within the INA's deportability classification because some, but not all, of the convictions under 18 U.S.C. § 656 qualify as offenses involving fraud or deceit. Because we are unable to determine from the face of the statute whether Valansi's conviction is among those that qualify as an aggravated felony, we must take the additional step of examining the underlying facts to determine whether Valansi pled guilty to an offense involving fraud or deceit.

Valansi, 278 F.3d at 214 (alterations in original). Ultimately, we concluded that

[i]n Valansi's case, the specific intent to defraud was not established. It appears that Valansi was counseled to avoid admitting to that intent, and the plea colloquy fails to pin down the mens rea element sufficiently for us to conclude that Valansi acted with the intent to defraud rather than to injure her employer.

Id. at 217. Accordingly, we granted the petition for review.

In dissent, Judge Scirica concluded that embezzlement under 18 U.S.C. § 656 is always a crime of fraud or deceit, and therefore had no need to analyze the intent-to-defraud/intent-to-injure distinction. While this seems more consistent with the formal categorical approach of Taylor, Judge Scirica's dissent still relies on (1) the plea colloquy (to establish that Valansi committed embezzlement, and not "abstract[ion], purloin[ing], or ... misappli [cation]"), and (2) the factual record (to establish the fiduciary relationship between Valansi and her employing bank necessary to satisfy his definition of embezzlement). Both steps are beyond the strict scope of Taylor' s categorical approach. Thus, whatever disagreements there were within the panel in Valansi, all agreed that the case required a look beyond the literal elements of 18 U.S.C. § 656.

c. A governing principle?

Our survey complete, the question we now pose is whether these decisions-some applying Taylor, some not-can be reconciled under a governing principle. We believe that they can. As Taylor itself demonstrates, there are two facets to these cases: the federal statute enumerating categories of crimes on the one hand (the "enumerating statute"), and the criminal statute of conviction, whether federal or state, on the other (the "statute of conviction"). While Taylor' s formal categorical approach presumptively applies in comparing the two, under certain conditions, both the enumerating statute and the statute of conviction can require a departure from the formal categorical approach.

In the case of the enumerating statute, a departure from the formal categorical approach seems warranted when the terms of the statute invite inquiry into the facts underlying the conviction at issue. The qualifier "in which the loss to the victim or victims exceeds $10,000" in 8 U.S.C. § 1101(a)(43)(M)(i) is the prototypical example-it expresses such a specificity of fact that it almost begs an adjudicator to examine the facts at issue. This principle explains our holdings in Nugent and Munroe. Another example would be an enumerating statute specifying crimes "committed within the last two years." Such a statute could not be read to cover only crimes which have "within the last two years" as an element; instead, a court would read "within the last two years" as a limiting provision on crimes that would otherwise qualify.

In contrast, cases interpreting relatively unitary categorical concepts-like "forgery" (Drakes), "burglary" (Taylor itself) or "crime of violence" (Francis and Bovkun) — do not look to underlying facts because the enumerating statute does not invite any such inquiry. Likewise, the hypothetical federal felony trilogy (Steele, Gerbier, and Wilson) asks only whether the elements of a federal criminal statute can by satisfied by reference to the actual statute of conviction; this presents no invitation to depart from Taylor' s formal categorical approach and examine the underlying facts.

Though we have little case law on point, the contrast we have described appears to be mirrored in the references in 8 U.S.C. § 1101(a)(43) to the duration of sentences. Correcting for the scrivener's error in 8 U.S.C. § 1101(a)(43)(G), we held in Graham, 169 F.3d at 791, that that section specifies theft and burglary offenses "for which the term of imprisonment [imposed is] at least one year." This obviously invites an inquiry into the sentence actually imposed on the alien, rather than a categorical inquiry into the statutory punishment for the offense. Similarly, provisions like 8 U.S.C. § 1101(a)(43)(S), which classifies as an aggravated felony certain obstruction of justice offenses "for which the term of imprisonment is at least one year," invites inquiry into the alien's actual sentence. See Graham, 169 F.3d at 790-91. In contrast, provisions like 8 U.S.C. § 1101(a)(43)(J), which classifies as an aggravated felony certain racketeering and gambling offenses "for which a sentence of one year imprisonment or more may be imposed," seem to direct inquiry toward the statutory sentencing scheme, not the alien's actual sentence. See Graham, 169 F.3d at 790-91.

Turning to the statute of conviction, there are also cases where a look into the underlying facts-or at least the charging instrument-is called for. Valansi is a good example of such a case: There, the statute of conviction was phrased in the disjunctive-a mens rea of either intent to defraud or intent to injure would suffice for conviction-which, in our view, called for an exploration of which of the alternative elements was the actual basis for conviction. Statutes phrased in the disjunctive are akin to, and can be readily converted to, statutes structured in outline form, with a series of numbered or letter elements. See, e.g., statutes cited supra notes 8-10. Such statutes may sometimes more clearly invite further inquiry into exactly which subsection the defendant violated. The exercise of analyzing disjunctive statutes for an invitation to further inquiry is much more difficult than that described in the preceding paragraphs, for it poses the vexing question of how far below the judgment or plea colloquy the court may look. The cases are few and the jurisprudence is not clear. However, in the hope that it may shed some light on this troublesome area, we will do our best to analyze the problem.

We have used a numbered subsection of such a statute as a statute of conviction for purposes of the Taylor inquiry. See Wilson, 350 F.3d at 381 (citing N.J. Stat. Ann. § 2C:35-5(b)(11), which specifically criminalizes marijuana). In Wilson, however, the numbered paragraphs of § 2C:35-5(b) were distinct offenses carrying separate penalties, not alternate types of conduct that constituted the same offense. The lesson there is that sometimes disjunctive parts of statutes of conviction represent distinct offenses, with distinct punishments. Where different crimes with different penalties are involved, under the categorical approach, further inquiry is clearly invited into which particular crime the petitioner was actually convicted of.

In other statutes, disjunctive wording or outline formatting simply describes variations of the same offense, with no difference in punishment and no distinction on the judgment of conviction. Even here, though, further inquiry might be warranted, as we found in Valansi, because the face of the statute might not make clear whether the conviction qualifies as an aggravated felony. It is not clear that only those disjunctions reflected in the penalty or the judgment of conviction are relevant for the purposes of § 1101(a)(43); where some variations of the crime of conviction meet the aggravated-felony requisites and others do not, we have thus allowed further inquiry to see which variation was actually committed.

Taylor itself, in some ways, provides an example. Taylor was convicted of second-degree burglary under some section of Mo. Rev.Stat. § 560 (1969); the Court remanded for consideration of whether he was convicted under §§ 560.045, .050, .055, .060, .070, .075, or .080. Taylor, 495 U.S. at 578 n. 1, 110 S. Ct. 2143. These seven separate statutory sections, each of which contained different factual predicates for the same crime with the same penalty, are in practice very similar to a single statute defining a crime, with seven numbered subsections containing alternate factual predicates. And in Taylor, the Supreme Court found that this division invited further inquiry.

Since any statute that is phrased in the disjunctive can be readily converted to outline form, it would be strange to think that Congress intended the application of the categorical approach to turn on the typography used by the statute's drafters. Commonly, the best way to resolve the question raised by a conviction under a statute phrased in the disjunctive, or structured in outline form, will be to look to the charging instrument or to a formal guilty plea (as we did in Valansi, for example). But even in such cases, we have not taken the further step of looking to facts outside the charging instrument or further plea; we leave for another day the question whether statutes phrased in the disjunctive invite such inquiry beyond a charging instrument or a formal plea.

As suggested above, our jurisprudence is not a seamless web. In Wilson, we did not conduct any further inquiry into the underlying facts of the conviction, even though the statute at issue was phrased in the disjunctive. As discussed above, the petitioner had been convicted under N.J. Stat. Ann. § 2C:35-5(a)(1), which made it unlawful "[t]o manufacture, distribute or dispense, or to possess or have under his control with intent to manufacture, distribute or dispense, [certain substances]." Despite this disjunctive phrasing, we declined to examine whether Wilson had engaged in distribution or possession with the intent to distribute. Subject to our discussion in the margin,13 this may be in tension with our earlier decision in Valansi. But there may also be other cases in which a statute phrased in the disjunctive would not invite further inquiry; we cannot with great confidence draw any general rule on this point from our cases.

At all events, for purposes of deciding Singh's petition, we need not resolve the matter, because Singh's statute of conviction is not phrased in the disjunctive in a relevant way. The statute is phrased in the disjunctive, both with respect to its actus reus (which can be either (1) sexual contact or (2) causing sexual contact) and its mens rea (which can be either (a) knowing that the contact is offensive to the victim, or (b) knowing that the contact occurs without the victim's consent). But none of this gives insight into the question whether 11 Del. C. § 767 constitutes "sexual abuse of a minor" because, though any combination of actus reus and mens rea seem to suffice as the actus reus and mens rea of "sexual abuse," the statute is silent on the critical matter of the age of the victim.

2. Does the formal categorical approach apply here?

At long last, we come to the operative question in this case: Does Taylor's formal categorical approach apply to 8 U.S.C. § 1101(a)(43)(A) and 11 Del. C. § 767? We have already indicated that Singh's statute of conviction does not invite us to go beyond the formal categorical approach, because it is not phrased in the disjunctive in a relevant way. Thus we must ask whether 8 U.S.C. § 1101(a)(43)(A) (and specifically "sexual abuse of a minor") invites inquiry-that is, whether it is more like the amount-of-loss provision of 8 U.S.C. § 1101(a)(43)(M)(i), or more like provisions for "burglary," 8 U.S.C. § 1101(a)(43)(G), or "crime of violence," 8 U.S.C. § 1101(a)(43)(F).

We think it clear that "sexual abuse of a minor" belongs in the latter category. First, it is listed in the same subsection as "murder" and "rape," two terms that share the common law pedigree of "burglary," which was examined in Taylor itself. Applying the maxim noscitur a sociis,14 we would place "sexual abuse of a minor" in a similar mold. Second, nothing in the phrase "sexual abuse of a minor" signals that a factual investigation is called for. Congress could have enacted, for example, the language "any sex offense, where the victim of such offense was a minor"; such language, parallel to provisions like 8 U.S.C. § 1101(a)(43)(G) & (S), might direct our inquiry into the facts of the crime rather than its definition. Third, in an area that so routinely implicates state laws, Congress is presumed to legislate against the backdrop of existing state statutes. The widespread existence of state statutes specifically criminalizing sexual abuse of a minor, see, e.g., Ala.Code § 13A-6-67 ("Sexual abuse in the second degree"); Cal.Penal Code § 288.5 ("Continuous sexual abuse of a child"); 11 Del. C. § 778 ("Continuous sexual abuse of a child"); 18 Pa. Cons.Stat. § 6312 ("Sexual abuse of children"), supports the conclusion that Congress intended Taylor' s formal categorical approach to be applied in this case.

In view of our conclusion in Part III.B that Taylor' s formal categorical approach applies to "sexual abuse of a minor" under 8 U.S.C. § 1101(a)(43)(A) and 11 Del. C. § 767, our analysis in Part III.A stands: Because § 767 does not establish the age of the victim, a conviction under that statute is not the aggravated felony of sexual abuse of a minor. Because Singh has not been convicted of an aggravated felony, we have jurisdiction and will grant the petition for review.