with whom
The Chief Justice and Justice Thomas join, and with whom Justice Souter joins as to Part I, dissenting.In Church of the Holy Trinity v. United States, 143 U. S. 457 (1892), this Court conceded that a church’s act of contracting with a prospective rector fell within the plain meaning of a federal labor statute, but nevertheless did not apply the statute to the church: “It is a familiar rule,” the Court pronounced, “that a thing may be within the letter of the statute and yet not within the statute, because not within its spirit, nor within the intention of its makers.” Id., at 459. That is a judge-empowering proposition if there ever was one, and in the century since, the Court has wisely retreated from it, in words if not always in actions. But today Church of the Holy Trinity arises, Phoenix-like, from the ashes. The Court’s contrary assertions aside, today’s decision is nothing other than the elevation of judge-supposed legislative intent over clear statutory text. The plain language of the federal Impact Aid statute clearly and unambiguously forecloses the Secretary of Education’s preferred methodology for determining whether a State’s school-funding system is equalized. Her selection of that methodology is therefore entitled to zero deference under Chevron U. S. A. Inc. v. Natural Resources Defense Council, Inc., 467 U. S. 837 (1984).
I
The very structure of the Court’s opinion provides an obvious clue as to what is afoot. The opinion purports to place a premium on the plain text of the Impact Aid statute, ante, at 93-94, but it first takes us instead on a roundabout tour of “[cjonsiderations other than language,” ante, at 90 (emphasis added) — page after page of unenacted congressional intent and judicially perceived statutory purpose, Part II-A, ante. Only after we are shown “why Zuni concentrates its argument upon language alone,” ante, at 90 (impliedly a shameful practice, or at least indication of a feeble case), are we in*109formed how the statute’s plain text does not unambiguously preclude the interpretation the Court thinks best. Part II-B, ante (beginning “But what of the provision’s literal language? The matter is important . . . ”). This is a most suspicious order of proceeding, since our case law is full of statements such as “We begin, as always, with the language of the statute,” Duncan v. Walker, 533 U. S. 167, 172 (2001), and replete with the affirmation that, when “[g]iven [a] straightforward statutory command, there is no reason to resort to legislative history,” United States v. Gonzales, 520 U. S. 1, 6 (1997). Nor is this cart-before-the-horse approach justified by the Court’s excuse that the statute before us is, after all, a technical one, ante, at 90. This Court, charged with interpreting, among other things, the Internal Revenue Code, the Employee Retirement Income Security Act of 1974, and the Clean Air Act, confronts technical language all the time, but we never see fit to pronounce upon what we think Congress meant a statute to say, and what we think sound policy would counsel it to say, before considering what it does say. As almost a majority of today’s majority worries, “[w]ere the inversion [of inquiry] to become systemic, it would create the impression that agency policy concerns, rather than the traditional tools of statutory construction, are shaping the judicial interpretation of statutes.” Ante, at 107 (Kennedy, J., joined by Alito, J., concurring). True enough — except I see no reason to wait for the distortion to become systemic before concluding that that is precisely what is happening in the present case. For some, policy-driven interpretation is apparently just fine. See ante, at 105 (Stevens, J., concurring). But for everyone else, let us return to Statutory Interpretation 101.
We must begin, as we always do, with the text. See, e. g., Gonzales, supra, at 4. Under the federal Impact Aid program, 20 U. S. C. § 7701 et seq. (2000 ed. and Supp. IV), States distributing state aid to local school districts (referred to in *110the statute as “local educational agencies,” or “LEAs” 1) may not take into account the amount of federal Impact Aid that its LEAs receive. See § 7709(a). But the statute makes an exception if the Secretary of Education certifies that a State “has in effect a program of State aid that equalizes expenditures for free public education among local educational agencies in the State.” § 7709(b)(1) (2000 ed., Supp. IV). Congress has specified a formula for the Secretary to use when making this equalization determination:
“[A] program of State aid equalizes expenditures among local educational agencies if... the amount of per-pupil expenditures made by, or per-pupil revenues available to, the local educational agency in the State with the highest such per-pupil expenditures or revenues did not exceed the amount of such per-pupil expenditures made by, or per-pupil revenues available to, the local educational agency in the State with the lowest such expenditures or revenues by more than 25 percent.” § 7709(b)(2)(A).
The Secretary is further instructed, however, that when making this determination, she shall “disregard local educational agencies with per-pupil expenditures or revenues above the 95th percentile or below the 5th percentile of such expenditures or revenues in the State.” § 7709(b)(2)(B)(i). It is this latter subsection which concerns us here.
The casual observer will notice that the Secretary’s implementing regulations do not look much like the statute. The regulations first require the Secretary to rank all of the LEAs in a State (New Mexico has 89) according to their per-pupil expenditures or revenues. 34 CFR pt. 222, subpt. K, App., ¶ (1)(a)(i) (2006). So far so good. But critically here, *111the Secretary must then “[i]dentif[y] those LEAs . . . that fall at the 95th and 5th percentiles of the total number of pupils in attendance in the schools of those LEAs.” Id., ¶ (1)(a)(ii) (emphasis added). Finally, the Secretary compares the per-pupil figures of those two LEAs for the purpose of assessing whether a State exceeds the 25% disparity measure. Id., ¶ (1)(a)(iii). The majority concludes that this method of calculation, with its focus on student population, is a permissible interpretation of the statute.
It most assuredly is not. To understand why, one first must look beyond the smokescreen that the Court lays down with its repeated apologies for inexperience in statistics, and its endless recitation of technical mathematical definitions of the word “percentile.” See, e. g., ante, at 95 (“ ‘The n-th percentile is the value xwioo such that n per cent of the population is less than or equal to Xw/ioo’ ” (quoting C. Clapham & J. Nicholson, The Concise Oxford Dictionary of Mathematics 378 (3d ed. 2005))). This case is not a scary math problem; it is a straightforward matter of statutory interpretation. And we do not need the Court’s hypothetical cadre of number-crunching amici, ante, at 99-100, to guide our way.
There is no dispute that for purposes relevant here “ ‘percentile’ refers to a division of a distribution of some population into 100 parts.’ ” Ante, at 95. And there is further no dispute that the statute concerns the percentile of “per-pupil expenditures or revenues,” for that is what the word “such” refers to. See 20 U. S. C. § 7709(b)(2)(B)(i) (Secretary shall “disregard local educational agencies with per-pupil expenditures or revenues above the 95th percentile or below the 5th percentile of such expenditures or revenues in the State” (emphasis added)). The question is: Whose per-pupil expenditures or revenues? Or, in the Court’s terminology, what “population” is assigned the “characteristic” “per-pupil expenditure” or revenue? Ante, at 95-96. At first blush, second blush, or twenty-second blush, the answer is abundantly clear: local educational agencies. The statute re*112quires the Secretary to “disregard local educational agencies with” certain per-pupil figures above or below specified percentiles of those per-pupil figures. § 7709(b)(2)(B)(i). The attribute “per-pupil expenditure] or revenu[e]” is assigned to LEAs — there is no mention of student population whatsoever. And thus under the statute, “per-pupil expenditures or revenues” are to be arrayed using a population consisting of LEAs, so that percentiles are determined from a list of (in New Mexico) 89 per-pupil expenditures or revenues representing the 89 LEAs in the State. It is just that simple.
The Court makes little effort to defend the regulations as they are written. Instead, relying on a made-for-litigation theory that bears almost no relationship to the regulations themselves, the Court believes it has found a way to shoehorn those regulations into the statute. The Impact Aid statute is ambiguous, the Court says, because it “does not specify precisely what population is to be ‘distributed’ (i. e., ranked according to the population’s corresponding values for the relevant characteristic).” Ante, at 96. Thus the Court finds that it is permissible for the Secretary to attribute the characteristic “per-pupil expenditure or revenue” to pupils, with the result that the Secretary may “us[e]... the State’s students as the relevant population for calculating the specified percentiles.” Ibid. Under that interpretation, as the State manages to explain with a straight face, “[i]n New Mexico, during the time at issue, there were approximately 317,777 pupils in the [S]tate and thus there were 317,777 per-pupil revenues in the [S]tate.” Brief for Respondent New Mexico Public Education Department 37; see also id., at 36 (“Each and every student in an LEA and in a [S]tate may be treated as having his or her own ‘per-pupil’ expenditure or revenue amount”). The Court consequently concludes that “linguistically speaking, one may attribute the characteristic of per-pupil expenditure to each [student].” Ante, at 97.
*113The sheer applesauce of this statutory interpretation should be obvious. It is of course true that every student in New Mexico causes an expenditure or produces a revenue that his LEA either enjoys (in the case of revenues) or is responsible for (in the case of expenditures). But it simply defies any semblance of normal English usage to say that every pupil has a “per-pupil expenditure or revenue.” The word “per” connotes that the expenditure or revenue is a single average figure assigned to a unit the composite members of which are individual pupils. And the only such unit mentioned in the statute is the local educational agency.2 See 20 U. S. C. § 7709(b)(2)(B)(i). It is simply irrelevant that “[n]o dictionary definition ... suggests that there is any single logical, mathematical, or statistical link between [per-pupil expenditures or revenues] and . . . the nature of the relevant population.” Ante, at 96. Of course there is not. It is the text at issue which must identify the relevant population, and it does so here quite unambiguously: “local educational agencies with per-pupil expenditures or revenues.” § 7709(b)(2)(B)(i) (emphasis added). That same phrase shows the utter irrelevance of the Court’s excursus upon the meaning of the word “per.” See ante, at 97. It does indeed mean “ ‘for each, or ‘for every’ ” — and when it is contained in a clause that reads “local educational agencies with per-pupil expenditures or revenues” it refers to (and can only refer to) the average expenditure or revenue “for each” or “for every” student out of the total expenditures or revenues of the LEA.
*114The violence done to this statute would be severe enough if the Secretary used the actual expenditure or revenue for each individual pupil. But in fact the Secretary determines the per-pupil expenditure or revenue for each individual student by (guess what) computing the per-pupil expenditure or revenue of each LEA! As the New Mexico brief explains:
“[A] per-pupil expenditure or revenue is an average number. It is not the amount actually spent on any given pupil, an amount which would be impossible to calculate in any meaningful way. It is roughly the total amount expended by an LEA divided by the number of pupils in that LEA.” Brief for Respondent New Mexico Public Education Department 36.
The Secretary thus assigns an artificial number to each student that corresponds exactly to his LEA’s per-pupil expenditure or revenue. In other words, at the end of the day the Secretary herself acknowledges that “per-pupil expenditures or revenues” pertains to LEAs, and not students. And she is interpreting “per-pupil expenditure or revenue” not as the Court suggests (an amount attributable to each student), but rather as I suggest (an average amount for the pupils in a particular LEA). But she then proceeds to take a step not at all permitted by the statutory formula — in effect applying “per-pupil expenditure or revenue” a second time (this time according to the Court’s fanciful interpretation of “per-pupil”) in order to reach the result she desires. Of course, if the Secretary did apply the “per-pupil expenditure or revenue” only once, arraying students by their actual expenditures or revenues, her entire system would collapse. Students from the same LEA, rather than appearing on the list with the same per-pupil figure, would be located at various points on the spectrum. And so long as an LEA had at least one student above the 95th or below the 5th percentile of pupil “per-pupil expenditures or revenues,” that LEA would *115have to be excluded from the disparity analysis. The result would be a serious distortion of the disparity determination, excluding many more LEAs (in fact, perhaps all of them) from the disparity calculation. This would render the 25% disparity measure in § 7709(b)(2)(A) all but meaningless.
The Court makes one final attempt to rescue the Secretary’s interpretation, appealing to “statutory context.” “Context here tells us,” it says, “that the instruction to identify school districts with ‘per-pupil expenditures’ above the 95th percentile ‘of such expenditures’ is . . . ambiguous, because both students and school districts are of concern to the statute.” Ante, at 99. This is a complete non sequitur. Of course students are a concern to a statute dealing with school funding. But that does not create any ambiguity with respect to whether, under this statute, pupils can reasonably be said to have their own “per-pupil expenditures or revenues.” It is simply irrational to say that the clear dispositions of a statute with regard to the entities that it regulates (here LEAs) are rendered ambiguous when those entities contain subunits that are the ultimate beneficiaries of the regulation (here students). Such a principle of interpretation — if it could be called that — would inject ambiguity into many statutes indeed.
The Court’s reliance on statutory context is all the more puzzling since the context obviously favors petitioners. “The focus [of the Impact Aid statute] is upon LEAs, not upon the number of pupils.” 393 F. 3d 1158, 1172 (CA10 2004) (O’Brien, J., dissenting), opinion vacated, 437 F. 3d 1289, 1290 (2006) (en banc) (per curiam). In fact, the provisions at issue here make not the slightest mention of students. That is both sensible and predictable, since the Impact Aid program’s equalization formula is designed to address funding disparities between LEAs, not between students. See 20 U. S. C. § 7709(b)(2)(A) (referring to “a program of State aid [that] equalizes expenditures among local educational agencies”); see also § 7709(d)(1). Indeed, *116the whole point of the equalization determination is to figure out whether States may reduce state aid to LEAs. See § 7709(a).
In sum, the plain language of the Impact Aid statute compels the conclusion that the Secretary’s method of calculation is ultra vires. Employing the formula that the statute requires, New Mexico is not equalized. Ante, at 89.
II
How then, if the text is so clear, are respondents managing to win this case? The answer can only be the return of that miraculous redeemer of lost causes, Church of the Holy Trinity. In order to contort the statute’s language beyond recognition, the Court must believe Congress’s intent so crystalline, the spirit of its legislation so glowingly bright, that the statutory text should simply not be read to say what it says. See Part II-A, ante. Justice Stevens is quite candid on the point: He is willing to contradict the text. See ante, at 106-107 (concurring opinion).3 But Justice Stevens’ candor should not make his philosophy seem unassuming. He maintains that it is “a correct performance of the judicial function” to “override a strict interpretation of the text” so long as policy-driven interpretation “is faithful to the intent of Congress.” Ante, at 105. But once one departs from “strict interpretation of the text” (by which Jus*117tice Stevens means the actual meaning of the text) fidelity to the intent of Congress is a chancy thing. The only thing we know for certain both Houses of Congress (and the President, if he signed the legislation) agreed upon is the text. Legislative history can never produce a “pellucidly clear” picture, ante, at 106 (Stevens, J., concurring), of what a law was “intended” to mean, for the simple reason that it is never voted upon — or ordinarily even seen or heard — by the “intending” lawgiving entity, which consists of both Houses of Congress and the President (if he did not veto the bill). See U. S. Const., Art. I, §§ 1, 7. Thus, what judges believe Congress “meant” (apart from the text) has a disturbing but entirely unsurprising tendency to be whatever judges think Congress must have meant, i. e., should have meant. In Church of the Holy Trinity, every Justice on this Court disregarded the plain language of a statute that forbade the hiring of a clergyman from abroad because, after all (they thought), “this is a Christian nation,” 143 U. S., at 471, so Congress could not have meant what it said. Is there any reason to believe that those Justices were lacking that “intellectual!] honestly]” that Justice Stevens “presume[s]” all our judges possess, ante, at 105? Intellectual honesty does not exclude a blinding intellectual bias. And even if it did, the system of judicial amendatory veto over texts duly adopted by Congress bears no resemblance to the system of lawmaking set forth in our Constitution.
Justice Stevens takes comfort in the fact that this is a case in which he “cannot imagine anyone accusing any Member of the Court of voting one way or the other because of that Justice’s own policy preferences.” Ante, at 106. I can readily imagine it, given that the Court’s opinion begins with a lengthy description of why the system its judgment approves is the better one. But even assuming that, in this rare case, the Justices’ departure from the enacted law has nothing to do with their policy view that it is a bad law, nothing in Justice Stevens’ separate opinion limits his ap*118proach to such rarities. Why should we suppose that in matters more likely to arouse the judicial libido — voting rights, antidiscrimination laws, or environmental protection, to name only a few — a judge in the School of Textual Subversion would not find it convenient (yea, righteous!) to assume that Congress must have meant, not what it said, but what he knows to be best?
Lest there be any confusion on the point, I must discuss briefly the two cases Justice Stevens puts forward, ante, at 104-105, as demonstrating this Court’s recent endorsement of his unorthodox views. They demonstrate just the opposite. Griffin v. Oceanic Contractors, Inc., 458 U. S. 564 (1982), involved a maritime statute that required the master of a vessel to furnish unpaid wages to a seaman within a specified period after the seaman’s discharge, and further provided that a master who failed to do so without sufficient cause “ ‘shall pay to the seaman a sum equal to two days’ pay for each and every day during which payment is delayed.’ ” Id., at 570 (quoting 46 U. S. C. §596 (1976 ed.)). We explained that “Congress intended the statute to mean exactly what its plain language says,” 458 U. S., at 574, and held that the seaman was entitled to double wages for every day during which payment was delayed, even for the period in which he had obtained alternative employment. The result was that the seaman would receive approximately $300,000 for his master’s improper withholding of $412.50, id., at 575, even though “[i]t [was] probably true that Congress did not precisely envision the grossness of the difference . . . between the actual wages withheld and the amount of the award required by the statute,” id., at 576. We suggested in dicta that there might be a “rare cas[e]” in which the Court could relax its steadfastness to statutory text, id., at 571, but if Griffin itself did not qualify, it is hard to imagine what would. The principle Justice Stevens would ascribe to Griffin is in fact the one he advocated in dissent. “[T]his is one of the cases in which the exercise of judgment dictates *119a departure from the literal text in order to be faithful to the legislative will.” Id., at 586 (Stevens, J., dissenting).
The second case Justice Stevens relies upon, United States v. Ron Pair Enterprises, Inc., 489 U. S. 285 (1989), is equally inapt. The Court’s opinion there (unlike the one here) explained that our analysis “must begin . . . with the language of the statute itself,” and concluded that that was “also where the inquiry should end, for where . . . the statute’s language is plain, ‘the sole function of the courts is to enforce it according to its terms.’” Id., at 241 (quoting Caminetti v. United States, 242 U. S. 470, 485 (1917)). My “fifth vote” in Ron Pair was thus only “decisive,” ante, at 105 (Stevens, J., concurring), in reaffirming this Court’s adherence to statutory text, decisively preventing it from falling off the precipice it plunges over today.
Contrary to the Court and Justice Stevens, I do not believe that what we are sure the Legislature meant to say can trump what it did say. Citizens arrange their affairs not on the basis of their legislators’ unexpressed intent, but on the basis of the law as it is written and promulgated. I think it terribly unfair to expect that the two rural school districts that are petitioners here should have pored over some 30 years of regulatory history to divine Congress’s “real” objective (and with it the “real” intent that a majority of Justices would find honest and true). To be governed by legislated text rather than legislators’ intentions is what it means to be “a Government of laws, not of men.” And in the last analysis the opposite approach is no more beneficial to the governors than it is to the governed. By “depriving legislators of the assurance that ordinary terms, used in an ordinary context, will be given a predictable meaning,” we deprive Congress of “a sure means by which it may work the people’s will.” Chisom v. Roemer, 501 U. S. 380, 417 (1991) (Scalia, J., dissenting).
I do not purport to know what Congress thought it was doing when it amended the Impact Aid program in 1994. *120But even indulging Justice Stevens’ erroneous premise that there exists a “legislative intent” separate and apart from the statutory text, ante, at 105 (concurring opinion), I do not see how the Court can possibly say, with any measure of confidence, that Congress wished one thing rather than another. There is ample evidence, for example, that at the time it amended the Impact Aid statute, Congress knew exactly how to incorporate student population into a disparity calculation. Most prominently, in the very same Act that added § 7709(b)(2)(B)(i) to the Impact Aid program, Congress established the Education Finance Incentive Program, known as EFIG. See Improving America’s Schools Act of 1994,108 Stat. 3575. That statute allocates grants to States based in part on an “equity factor” which requires a disparity calculation similar to that in the Impact Aid statute. See 20 U.S.C. § 6337(b)(1)(A) (2000 ed„ Supp. IV). In EFIG, however, Congress specifically required the Secretary to take student population into account: “[TJhe Secretary shall weigh the variation between per-pupil expenditures in each local educational agency . . . according to the number of pupils served by the local educational agency.” § 6337(b)(3)(A)(ii)(II) (emphasis added); see also Brief for Federal Respondent 28-29. And there is more. In EFIG, Congress expressly provided that a State would be accorded a favorable “equity factor” rating if it was considered equalized under the Secretary’s Impact Aid regulations. See § 6337(b)(3)(B) (2000 ed., Supp. IV). Congress thus explicitly incorporated the Impact Aid regulations into EFIG, but did no such thing with respect to the Impact Aid statute itself. All this on the very same day.
Nor do I see any significance in the fact that no legislator in 1994 expressed the view that § 7709(b)(2)(B)(i) was designed to upend the Secretary’s equalization formula. Ante, at 91 (majority opinion). It is quite plausible — indeed, eminently plausible — that the Members of Congress took the *121plain meaning of the language whick the Secretary himself had proposed to be what the Secretary himself had previously been doing. It is bad enough for this Court to consider legislative materials beyond the statutory text in aid of resolving ambiguity, but it is truly unreasonable to require such extratextual evidence as a precondition for enforcing an unambiguous congressional mandate. See Koons Buick Pontiac GMC, Inc. v. Nigh, 543 U.S. 50, 73-74 (2004) (Scalia, J., dissenting). The Court points to the fact that “no Member of Congress has ever criticized the method the [Secretary’s] regulation^] sets forth.” Ante, at 90. But can it really be that this case turns, in the Court’s view, on whether a freshman Congressman from New Mexico gave a floor speech that only late-night C-SPAN junkies would witness? The only fair inference from Congress’s silence is that Congress had nothing further to say, its statutory text doing all of the talking.
Finally, the Court expresses its belief that Congress must have intended to adopt the Secretary’s pre-1994 disparity test because that test is the more reasonable one, better able to account for States with small numbers of large LEAs, or large numbers of small ones. See ante, at 91-93. This, to tell the truth, is the core of the opinion. As I have suggested, it is no accident that the countertextual legislative intent judges perceive invariably accords with what judges think best. It seems to me, however, that this Court is no more capable of saying with certainty what is best in this area than it is of saying with certainty (apart from the text) what Congress intended. There is good reason to be concerned — in the implementation of a statute that makes a limited exception for States that have “in effect a program of State aid that equalizes expenditures for free public education among local educational agencies,” 20 U. S. C. § 7709(b)(1) (2000 ed., Supp. IV) (emphasis added) — that the Secretary’s methodology eliminates from the disparity calcu*122lation too many LEAs. In the certification at issue in this very case, the Secretary excluded 23 of New Mexico’s 89 LEAs, approximately 26%. Is this Court such an expert in school finance that it can affirm the desirability of excluding one in four of New Mexico’s LEAs from consideration?
As for the Secretary’s concerns about the discrepancy between large and small LEAs, does the Court have any basis for its apparent confidence that other parts of the Impact Aid statute do not adequately address the problem? Immediately after setting forth the 95th and 5th percentile cutoffs, §7709(b)(2)(B)(i), the statute instructs the Secretary to “take into account the extent to which a program of State aid reflects the additional cost of providing free public education in particular types of local educational agencies, such as those that are geographically isolated, or to particular types of students, such as children with disabilities.” § 7709(b)(2)(B)(ii). Respondents do not explain why the Secretary could not use § 7709(b)(2)(B)(ii) to temper any unintended effects of § 7709(b)(2)(B)(i). Respondents further maintain that States could take advantage of the statute’s plain meaning by subdividing their LEAs. But again, the statute itself contains a remedy. Under § 7713(9)(B)(ii), “[t]he term ‘local educational agency’ does not include any agency or school authority that the Secretary determines on a case-by-case basis ... is not constituted or reconstituted for legitimate educational purposes.”
* * *
The only sure indication of what Congress intended is what Congress enacted; and even if there is a difference between the two, the rule of law demands that the latter prevail. This case will live with Church of the Holy Trinity as an exemplar of judicial disregard of crystal-clear text. We must interpret the law as Congress has written it, not as we would wish it to be. I would reverse the judgment of the Court of Appeals.
The Court’s opinion has replaced the phrase “‘local educational agencies’ ” with “ ‘local school districts.’ ” See ante, at 100. While I have no objection to that terminology, I will instead use “local educational agencies” and “LEAs.”
The Court maintains that the phrase “per-pupil expenditures” or revenues may also be attributed to schools or grade levels. Ante, at 97. Standing alone and abstracted from the rest of the statute, indeed it may. But not when it appears in the phrase “local educational agencies with per-pupil expenditures or revenues.” (Emphasis added.) In any case, the fact that “per-pupil expenditures or revenues” could be applied to composite entities other than LEAs does not establish that speaking of the “per-pupil expenditure or revenue” of an individual student makes any sense (it does not).
Like Justice Stevens, respondents themselves were aboveboard when they litigated this case at the administrative level. After hearing argument from the Department of Education, the Administrative Law Judge (ALJ) protested: “The problem is I don’t see the ambiguity of the statute.” App. 29. To this the Department’s counsel responded: “The only way I can do that is by reference to the statutory purpose.” Ibid. Later in the hearing, the ALJ similarly asked the State of New Mexico how its interpretation was consistent with the statute. The State answered: “Literally, on the face of the words, perhaps not, probably not." Id., at 53. Despite his misgivings, the ALJ ultimately decided that he did not possess the authority to invalidate the regulations. App. to Pet. for Cert. 38a, 51a.