Introduction

The proper allocation of authority between those who execute the law and those who adjudicate individual cases has been the subject of debate for centuries—if not millennia.¹See, e.g., Aristotle, Politics 229, 265 (H. Rackham trans., Harvard University Press, 1944):

[I]t is proper for the laws when rightly laid down to be sovereign, while the ruler or rulers in office should have supreme powers over matters as to which the laws are quite unable to pronounce with precision because of the difficulty of making a general rule to cover all cases. . . .

[I]t is preferable for the law to rule rather than any one of the citizens, and according to this same principle, even if it be better for certain men to govern, they must be appointed as guardians of the laws and in subordination to them; for there must be some government, but it is clearly not just, men say, for one person to be governor when all the citizens are alike. It may be objected that any case which the law appears

to be unable to define, a human being also would be unable to decide. But the law first specially educates the magistrates for the purpose and then commissions them to decide and administer the matters that it leaves over “according to the best of their judgment,” and furthermore it allows them to introduce for themselves any amendment that experience leads them to think better than the established code. He therefore that recommends that the law shall govern seems to recommend that God and reason alone shall govern, but he that would have man govern adds a wild animal also; for appetite is like a wild animal, and also passion warps the rule even of the best men.

I quote this passage from Aristotle not to suggest that his views apply in the context of American administrative law, but rather to show that political philosophers have long grappled with the kinds of questions confronted in this Article. In the context of American administrative law, the debate has revolved around several different considerations. Some have argued that constitutional provisions, such as the clause vesting “judicial Power” in Article III courts²U.S. Const. art. III, § 1. and the Due Process Clause,³U.S. Const. amend. V. establish a standard of review that courts must employ when assessing factual and legal determinations previously made by agencies. Others have relied on policy-inflected judgments about the sound allocation of authority between courts and executive branch actors.⁴Most notably, the Supreme Court itself relied on considerations of sound policy in the allocation of authority to different institutions. See Chevron U.S.A. Inc. v. Nat. Res. Def. Council, Inc., 467 U.S. 837, 865–66 (1984) (stating that judges “have no constituency,” “are not experts in the field” or “part of either political branch” and, hence, are not competent to “resolv[e] the competing interests which Congress itself either inadvertently did not resolve, or intentionally left to be resolved by the agency charged with the administration of the statute in light of everyday realities”). Still others have focused on the meaning of statutory provisions setting forth the standard of review—specifically, those contained in the Administrative Procedure Act (“APA”).⁵5 U.S.C §§ 551-559, 701–706.

This Article addresses this latter question—the APA’s implications for the scope of review that Article III courts employ in interpreting legal text previously construed by administrative agencies. After a prolonged, multi-year debate,⁶See, e.g., George B. Shepherd, Fierce Compromise: The Administrative Procedure Act Emerges from New Deal Politics, 90 Nw. U. L. Rev. 1557, 1560 (1996); Aditya Bamzai, The Origins of Judicial Deference to Executive Interpretation, 126 Yale L.J. 908, 981–90 (2017) [hereinafter Origins of Judicial Deference]. Congress passed the APA in 1946, with a standard-of-review provision currently codified at 5 U.S.C. § 706.⁷5 U.S.C. § 706. In an introductory clause, section 706 authorizes reviewing courts to “decide all relevant questions of law, interpret constitutional and statutory provisions, and determine the meaning or applicability of the terms of an agency action.”⁸Id. Section 706 then proceeds to list a series of standards of review that, seemingly, apply in specific situations. For instance, section 706(2)(A) provides that a court should “hold unlawful and set aside agency action, findings, and conclusions” that a court finds to be “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.”⁹Id. § 706(2)(A). Sections 706(2)(E) and (F) provide that a court should hold unlawful and set aside agency findings “unsupported by substantial evidence” in “formal” proceedings,¹⁰More precisely, section 706(2)(E) applies “in a case subject to sections 556 and 557 of this title or otherwise reviewed on the record of an agency hearing provided by statute.” 5 U.S.C. § 706(2)(E); see id. §§ 556–57. For the canonical case addressing whether proceedings are “formal” for purposes of the APA, see United States v. Florida East Coast Railway Co., 410 U.S. 224, 242-44 (1973). or any findings “unwarranted by the facts to the extent that the facts are subject to trial de novo by the reviewing court.”¹¹5 U.S.C. § 706(2)(F). Section 706’s last clause requires the court to “review the whole record or those parts of it cited by a party” and to take “due account . . . of the rule of prejudicial error.” Id. § 706. And in sections 706(B) through (D), the APA authorizes review of agency decisions “contrary to constitutional right, power, privilege, or immunity”; “in excess of statutory jurisdiction, authority, or limitations, or short of statutory right”; and “without observance of procedure required by law.”¹²Id. § 706(B)–(D).

This Article addresses these provisions’ implications for the proper allocation of authority between those who interpret the law while executing it and those who interpret the law while adjudicating individual cases. As a textual matter, many have viewed the APA’s provisions as requiring some form of de novo review for legal questions. For example, five years after the APA’s passage in 1946, the Court of Appeals for the Ninth Circuit reasoned that, “[i]n enacting the Administrative Procedure Act Congress did not merely express a mood that questions of law are for the courts rather than agencies to decide,—it so enacted with explicit phraseology.”¹³SEC v. Cogan, 201 F.2d 78, 86–87 (9th Cir. 1951). And scholars, spanning from just after the APA’s passage to the recent past, have said much the same.¹⁴See, e.g., Thomas W. Merrill, The Chevron Doctrine: Its Rise and Fall, and the Future of the Administrative State 47 (2022) (noting that the APA appears “unequivocally to instruct courts to apply independent judgment on all questions of law”); John F. Duffy, Administrative Common Law in Judicial Review, 77 Tex. L. Rev. 113, 194 & n.408 (1998); Cynthia R. Farina, Statutory Interpretation and the Balance of Power in the Administrative State, 89 Colum. L. Rev. 452, 473 n.85 (1989) (“That section 706 appears to contemplate de novo judicial determination of questions of statutory meaning is generally acknowledged. This reading is supported by the section’s failure to distinguish in any way between the interpretation of constitutional and statutory provisions, the former of which has always been subject to independent judgment.” (citations omitted)); John Dickinson, Administrative Procedure Act: Scope and Grounds of Broadened Judicial Review, 33 A.B.A. J. 434, 516 (1947). To be sure, not all scholars agree. See infra Section II.D.

In recent years, the 1984 decision in Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc.,¹⁵467 U.S. 837 (1984). has been the relevant touchstone for courts conducting such an analysis. Chevron articulated a standard of review that might be understood to be in significant tension with section 706. In construing statutory language, Chevron reasoned that, if “Congress has not directly addressed the precise question at issue, the court does not simply impose its own construction on the statute, as would be necessary in the absence of an administrative interpretation.”¹⁶Id. at 843 (footnote omitted). Rather, the so-called “second step” of Chevron asks “whether the agency’s answer is based on a permissible construction of the statute,” even if that answer rejects the statute’s “best” interpretation as a reviewing court sees it.¹⁷Id. At the same time, Chevron required courts to employ all the “traditional tools of statutory construction” in addressing ambiguities.¹⁸Id. at 843 n.9. If taken seriously, this requirement would appear to compel courts to resolve all ambiguities in statutory text, to the extent that they can, even where the agency’s interpretation is a “permissible” one.

Precisely how to reconcile Chevron with the APA’s text has been hotly debated. Some claim that the APA’s text says precious little about judicial review of legal questions and, thus, says little about Chevron,¹⁹For a judicial articulation of this position, see Kisor v. Wilkie, 139 S. Ct. 2400, 2407, 2419 (2019) (plurality opinion) (contending that section 706 “does not specify the standard of review a court should use,” particularly given “the practice of judicial review at the time of the APA’s enactment”). For leading articles taking this position, see Ronald M. Levin, The APA and the Assault on Deference, 106 Minn. L. Rev. 125, 190 (2021) (arguing that the “text of § 706 is essentially noncommittal on the issue of what deference, if any, courts should display when they review agency legal interpretations” and that the statute left “the operative law on that subject as it stood”); Cass R. Sunstein, Chevron as Law, 107 Geo. L.J. 1613, 1642 (2019) (“It is true that courts ‘shall decide all relevant questions of law,’ but the right way to decide those questions might be to consult the agency’s view and to accept it so long as it is reasonable. . . . Perhaps the law means what the agency says it means (so long as it is ambiguous).” (emphasis omitted)). Although I part ways with these scholars in some respects, I greatly appreciate their engagement with my own work, and I have learned a great deal from each of their contributions to the literature. For a recent debate on Chevron and stare decisis, compare Cass R. Sunstein, Zombie Chevron: A Celebration, 82 Ohio St. L.J. 565 (2021), with Aditya Bamzai, Judicial Deference and Doctrinal Clarity, 82 Ohio St. L.J. 585 (2021). despite section 706’s directive that courts should “decide all relevant questions of law” and “interpret constitutional and statutory provisions.”²⁰5 U.S.C. § 706. This Article argues that section 706 of the APA is best understood to establish a de novo standard of review for legal questions in the sense that a reviewing court should give statutory text the “best” reading possible, assuming one exists, using the traditional tools of construction. But by its very nature, de novo review incorporates traditional canons of construction, including those that have traditionally given weight to contemporaneous and customary agency understandings of legal text.²¹See infra Section I.A.1.

Part I of this Article addresses the pre-APA backdrop, discussing three doctrines relating to the judicial review of agency statutory interpretation before the APA’s passage. One of those doctrines called for differing standards of review for questions of fact and questions of law—a conceptual division that the Court became suspicious of in the immediate pre-APA era. Part II discusses the APA’s drafting, starting with predecessor bills before turning to the statute’s text, structure, and history. Part II also addresses contemporaneous commentary and other congressional action that casts light on the APA’s meaning. Finally, Part III returns to the Court’s cases, this time addressing the post-APA era and the statute’s echoes over the decades.

I.      The Pre-APA Backdrop

Before turning to the APA’s drafting, it is helpful to get a sense of the various approaches to interpretation that prevailed in the era immediately preceding the statute’s passage. In a nutshell, the preexisting interpretive framework relied on a distinction between law and fact that formed the basis for differing standards of review. This approach began to shift in the 1940s, when the Supreme Court embraced a critique of the distinction.

A.      Three Themes in the Judicial Review of Agency Statutory Interpretation

Three separate strands of jurisprudence helped frame the backdrop against which Congress enacted the Administrative Procedure Act in 1946: (1) the interpretive principle that courts respect customary and contemporaneous interpretations of law; (2) the differing standards of review for questions of law and questions of fact; and (3) judicial review via a writ of mandamus and its accompanying standard. Consider the three in turn.

1.      Customary and Contemporaneous Interpretation

Courts frequently respected executive branch interpretations of statutory text that were contemporaneous with the statute’s enactment or articulated a customary practice that had developed under the statute. In the context of constitutional interpretation, courts gave (and continue to give) similar weight to certain persuasive contemporaneous and customary understandings of legal text.²²See, e.g., NLRB v. Noel Canning, 573 U.S. 513, 525–29 (2014) (relying on a set of United States Attorney General opinions to interpret the Constitution and noting that “this Court has treated practice as an important interpretive factor”). And because the basic approach to interpreting the U.S. Constitution derived from preexisting theories of interpretation generally, it should be unsurprising to find that courts likewise relied on contemporaneous and customary understandings of statutory text.²³See, e.g., Edwards Lessee v. Darby, 25 U.S. (12 Wheat.) 206, 210 (1827) (“In the construction of a doubtful and ambiguous law, the contemporaneous construction of those who were called upon to act under the law, and were appointed to carry its provisions into effect, is entitled to very great respect.” (emphasis added)); United States v. Vowell, 9 U.S. (5 Cranch) 368, 372 (1810) (“If the question had been doubtful, the court would have respected the uniform construction which it is understood has been given by the treasury department . . . .” (emphasis added)). For a recent opinion applying this methodology, see Sackett v. EPA, 143 S. Ct. 1322, 1365 (2023) (Kavanaugh, J., concurring in the judgment) (reasoning that a “longstanding and consistent agency interpretation reflects and reinforces the ordinary meaning of the statute”). Contrast these cases with the interpretive approach of Chevron. See Nat’l Cable & Telecomms. Ass’n v. Brand X Internet Servs., 545 U.S. 967, 981 (2005) (“Agency inconsistency is not a basis for declining to analyze the agency’s interpretation under the Chevron framework.”). This approach to statutory interpretation reflected a broader commitment to weighing contemporaneous and customary practices in assessing the ordinary meaning of legal text.

In the pre-APA era, courts followed this approach. For example, in Burnet v. Chicago Portrait Co.,²⁴285 U.S. 1 (1932). a case relied on by Chevron itself,²⁵See Chevron U.S.A. Inc. v. Nat. Res. Def. Council, Inc., 467 U.S. 837, 843 n.9 (1984). the Court, in an opinion by Chief Justice Charles Evans Hughes, addressed the question whether a foreign territory was a “foreign country” within the meaning of a federal revenue statute.²⁶See Burnet, 285 U.S. at 4–5. The Court explained that the “word ‘country,’ in the expression ‘foreign country,’ is ambiguous,” because it could “be taken to mean foreign territory or a foreign government.”²⁷Id. at 5. In its analysis, the Court alluded to the “familiar principle . . . that great weight is attached to the construction consistently given to a statute by the executive department charged with its administration.”²⁸Id. at 16. But the Court identified a “qualification of that principle” that was “as well established as the principle itself”—namely, that the Court was “not bound by an administrative construction, and if that construction is not uniform and consistent, it will be taken into account only to the extent that it is supported by valid reasons.”²⁹Id.

Chief Justice Hughes’s approach was consistent with cases decided before and after 1932. Consider Justice Louis Brandeis’s statement in Iselin v. United States,³⁰270 U.S. 245 (1926). that the agency’s “construction was neither uniform, general, nor long-continued; neither is the statute ambiguous. Such departmental construction cannot be given the force and effect of law.”³¹Id. at 251. Or consider Justice Benjamin Cardozo’s reasoning the year after Burnet in Norwegian Nitrogen Products Co. v. United States.³²288 U.S. 294 (1933). In that case, Justice Cardozo explained the then-current state of the law as follows:

True indeed it is that administrative practice does not avail to overcome a statute so plain in its commands as to leave nothing for construction. True it also is that administrative practice, consistent and generally unchallenged, will not be overturned except for very cogent reasons if the scope of the command is indefinite and doubtful. . . . The practice has peculiar weight when it involves a contemporaneous construction of a statute by the men charged with the responsibility of setting its machinery in motion. . . .³³Id. at 315 (citations omitted); see also United States v. Moore, 95 U.S. 760, 762–63 (1878) (noting a construction of a statute that had “always heretofore obtained in the” agency was “entitled to the most respectful consideration, and ought not to be overruled without cogent reasons”); cf. Skidmore v. Swift & Co., 323 U.S. 134, 139–40 (1944).

Justice Cardozo’s approach thus stressed that administrative practice (“consistent and generally unchallenged” custom) and “contemporaneous construction” are relevant to construing statutes.³⁴Norwegian Nitrogen Prods., 288 U.S. at 315.

Scholars took note of this approach. Then-Professor Erwin Griswold wrote an article in 1941 summarizing what he described as the “regulations problem.”³⁵Erwin N. Griswold, A Summary of the Regulations Problem, 54 Harv. L. Rev. 398 (1941). Griswold explained that then-current law depended on two factors: “contemporaneousness, and long-continuedness.”³⁶Id. at 404. Although Griswold’s article dealt specifically with the effect to “be given to Treasury Regulations in the construction and application of the Federal Revenue Acts,”³⁷Id. at 398. he based his reasoning on the broader principle that “contemporaneousness is a significant factor in evaluating an administrative regulation [that] goes back to the earliest considerations of the problem.”³⁸Id. at 405. As Griswold put it, reference to contemporaneous administrative interpretation was warranted, because “unless the language of the statute is very general, the primary problem is to ascertain the meaning of the statute as of the time it was enacted.”³⁹Id. “Contemporaneous regulations may thus represent actual evidence of the elusive legislative intent.”⁴⁰Id. at 405–06 (listing cases, as well as other reasons justifying the principle); see id. at 408–11 (discussing cases addressing long-continuedness); see also White v. Winchester Country Club, 315 U.S. 32, 41 & n.17 (1942) (relying on Griswold’s article for the proposition that an agency’s “substantially contemporaneous expressions of opinion are highly relevant and material evidence of the probable general understanding of the times . . . .”).

Griswold’s approach paralleled Jabez Sutherland’s in his treatise on statutory interpretation.⁴¹See J.G. Sutherland, Statutes and Statutory Construction (1891). Sutherland devoted a section to “contemporaneous construction,” claiming that its benefit comes when a statute’s language is unclear “and cannot be made plain by the help of any other part of the same statute.”⁴²Id. § 307, at 391. Sutherland devoted several sections to “general usage,”⁴³See §§ 308–12, at 392–97. most pertinently explaining that “[a] practical construction, of long standing, by those for whom the law was enacted, will not be lightly questioned.”⁴⁴Id. § 309, at 392. Other treatise authors echoed these points.⁴⁵See, e.g., Origins of Judicial Deference, supra note 6, at 962–65. To be sure, there were cases that seemed not to fit the pattern of contemporaneousness and long-continuedness. See, e.g., Bates & Guild Co. v. Payne, 194 U.S. 106, 107–10 (1904). Notably, in Bates, Justice Harlan’s dissent pointed out that the government’s then-current position conflicted with its longstanding one and that the Court had “overthrown” the “settled” principle that “the established practice of an Executive Department charged with the execution of a statute will be respected and followed—especially if it has been long continued.” Id. at 111 (Harlan, J., dissenting); see also Origins of Judicial Deference, supra note 6, at 966–69, 968 n.253 (discussing Bates and citing other similar cases). Out of the thousands upon thousands of statutory cases that the Supreme Court decided before the APA was passed, it is unsurprising to find some that break the mold or some that stretch the established principles. But the general pattern identified by Griswold and Sutherland reflects a pre-1940s framework—as Chief Justice Hughes’s opinion in Burnet indicates. In recent years, scholars have acknowledged that “contemporaneity and continuity were important factors in the common law of judicial review,” but have sought to identify other cases that (like Bates) departed from the pattern in an effort to claim that there was no pattern. See Levin, supra note 19, at 167–70. In doing so, these scholars do not acknowledge that contemporaneous authorities (like Griswold, Sutherland, and Hughes, among others) identified the very same pattern I have described. See id. No less importantly, many of the cases that they cite are consistent with the customary-and-contemporaneous approach. See Michael B. Rappaport, Chevron and Originalism: Why Chevron Deference Cannot Be Grounded in the Original Meaning of the Administrative Procedure Act, 57 Wake Forest L. Rev. 1281, 1314–20, 1320 n.243 (2022).

2.      The Law-Fact Distinction

In addition, courts distinguished between questions of fact and questions of law in assessing the scope of review. Unless the Constitution or a statute required de novo review, courts reviewed questions of fact deferentially. The standard of review for questions of law was different.

The canonical case of this era addressing standards of review was Crowell v. Benson.⁴⁶285 U.S. 22 (1932). In Crowell, although Chief Justice Hughes’s majority and Justice Brandeis’s dissent disagreed on the scope of review of factual questions, they agreed on its application to questions of law. Chief Justice Hughes reasoned that the question of scope of review at issue in the case “relate[d] only to determinations of fact,” because authority over “legal questions” was reserved to federal courts.⁴⁷Id. at 49. As Chief Justice Hughes explained: “Rulings of the [agency official] upon questions of law are without finality. So far as the latter [i.e., ‘questions of law’] are concerned, full opportunity is afforded for their determination by the Federal courts through proceedings to suspend or to set aside a compensation order.”⁴⁸Id. at 45–46. “The Congress did not attempt to define questions of law,” Chief Justice Hughes observed, but the statute left “no doubt of the intention to reserve to the Federal court full authority to pass upon all matters which this Court had held to fall within that category.”⁴⁹Id. at 49; see also id. at 54 (“Findings of fact by the deputy commissioner upon such questions [of fact] are closely analogous to the findings of the amount of damages that are made, according to familiar practice, by commissioners or assessors; and the reservation of full authority to the court to deal with matters of law provides for the appropriate exercise of the judicial function in this class of cases.”).

In his dissent, Justice Brandeis agreed that, under the statute, the agency’s “conclusions are, as a matter of right, open to reexamination in the courts on all questions of law.”⁵⁰Id. at 88 (Brandeis, J., dissenting). And Justice Brandeis described “the prevailing practice” under the relevant review provisions as “confin[ing]” judicial review “to questions of law.”⁵¹Id. at 93.

In Crowell, Justice Brandeis’s dissent embraced a de novo standard of review for legal questions as a statutory matter.⁵²I will return to the specific statute that he was addressing below. See infra notes 133–40 and accompanying text. But a few years later, he appeared to suggest that the Constitution compelled de novo review and, thus, barred judicial deference to executive statutory legal interpretation. In a prominent concurrence, he reasoned that “[t]he supremacy of law demands that there shall be opportunity to have some court decide whether an erroneous rule of law was applied.”⁵³St. Joseph Stock Yards Co. v. United States, 298 U.S. 38, 84 (1936) (Brandeis, J., concurring); see id. at 73–74 (reasoning that due process requires that an administrative order “may be set aside for any error of law, substantive or procedural”). Many commentators during this era understood Brandeis’s concurrence as arguing that the Constitution required de novo review. See, e.g., James M. Landis, The Administrative Process 124 (1938) (interpreting Justice Brandeis as meaning that the “supremacy of law” requires the “right of a party to a judicial determination as to the appropriate rule of law applicable to his particular case”); Dickinson, supra note 14, at 516 (arguing that “[a] very broad statement of th[e] principle” that “questions of law are for the determination of the reviewing Court” is “contained in a classic passage of Mr. Justice Brandeis’ concurring opinion in the St. Joseph Stock Yards case”). Setting to one side whether this constitutional claim was correct, at a minimum, the concurrence reflected that prominent jurists of this era characterized the standard of review for legal questions as de novo.

Many scholars recognized that the distinction between law and fact was central to the applicable standard of review. For example, writing in 1927, John Dickinson observed that, under prevailing case law, courts “review[ed] for error[s] of law, but not findings of fact, at least where, on the evidence, the findings are within the bounds of reason.”⁵⁴John Dickinson, Administrative Justice and the Supremacy of Law in the United States 50 (1927). Professor James Landis said much the same about a decade later. He observed that the “scope of judicial review” for factual and legal questions was “wholly different.”⁵⁵Landis, supra note 53, at 145. At the same time, he remarked that “[t]he interesting problem as to the future of judicial review over administrative action is the extent to which judges will withdraw, not from reviewing findings of fact, but conclusions upon law.”⁵⁶Id. at 144 (emphasis added). This statement suggests that Landis recognized that current law did not incorporate a generally deferential standard of review for legal questions, but anticipated that such a standard might emerge in the future.

While recognizing that the law-fact distinction reflected current law on the standard of review, some of the same scholars challenged this distinction altogether. In the same 1927 book, Dickinson both acknowledged that the standard of review under prevailing cases depended on a distinction between law and fact and challenged that distinction.⁵⁷Dickinson, supra note 54, at 312–13. In his view, it was impossible “to establish a clear line between so-called ‘questions of law’ and ‘questions of fact,’” because standards inevitably “bridged [the gap] between the special subsidiary facts . . . and the ultimate [legal] conclusion.”⁵⁸Id. at 312, 315. Put differently, Dickinson claimed that “[m]atters of law grow downward into roots of fact, and matters of fact reach upward, without a break, into matters of law,” with “[t]he knife of policy alone effect[ing] an artificial cleavage at the point where the court chooses to draw the line between public interest and private right.”⁵⁹Id. at 55. The alleged fuzziness of the line between questions of law and questions of fact was a theme picked up by scholars, executive branch actors, and courts in the years to come.⁶⁰See Landis, supra note 53, at 145, 152–53.

3.      The Mandamus Standard

For much of the nineteenth century, judicial review occurred in certain areas using a writ of mandamus, in which an Article III court would not resolve questions of law de novo.⁶¹See, e.g., Origins of Judicial Deference, supra note 6, at 947–58; United States v. Mead Corp., 533 U.S. 218, 242–43 (2001) (Scalia, J., dissenting) (observing that under mandamus practice some “[s]tatutory ambiguities . . . were left to reasonable resolution by the Executive”); Decatur v. Paulding, 39 U.S. (14 Pet.) 497, 514–15 (1840) (holding that mandamus could not guide or control an executive officer’s judgment where the law authorized him to exercise discretion). But by the APA’s passage in 1946, actions in equity—rather than actions in mandamus—had long been understood as the catchall remedy for judicial review of administrative action.⁶²See, e.g., Duffy, supra note 14, at 118–19 (noting that “statutes conferring equity jurisdiction” were interpreted to “vest the federal courts with a power to fashion and administer a judge-made law of equity” and that, consequently, “[j]udicial review in the early administrative era grew up in the federal equity jurisdiction”); 3 Kenneth Culp Davis, Administrative Law Treatise § 23.04, at 307 (1st ed. 1958) (describing action in equity as “the mainstay for review of federal administrative action”); Louis L. Jaffe, Judicial Control of Administrative Action 193 (1965) (referring to action for injunction in equity as a “catchall”); see also Am. Sch. of Magnetic Healing v. McAnnulty, 187 U.S. 94, 110–11 (1902). That understanding meant that Congress enacted the APA’s provisions against a backdrop where the mandamus standard was not central to judicial review of agency action. And that backdrop meant that the APA’s standard-of-review provision was not understood to incorporate the mandamus standard.

B.      The Judicial Breakdown of the Law-Fact Distinction

In the era immediately surrounding the APA’s passage, several Supreme Court cases undercut a sharp distinction between “questions of law” and “questions of fact” for purposes of the standard of review.⁶³See Origins of Judicial Deference, supra note 6, at 977–81. Three cases exemplified this trend.

In Gray v. Powell,⁶⁴314 U.S. 402 (1941). the Court reasoned that Congress had “delegate[d] th[e] function” of interpreting the statutory term “producer” “to those whose experience in a particular field gave promise of a better informed, more equitable” judgment.⁶⁵Id. at 411–12, 412 n.7. Applying the general term “producer” to a “particular instance,” according to the Court, “call[ed] for the expert, experienced judgment of those familiar with the industry.”⁶⁶Id. at 413. Although this language might have suggested a broader approach to statutory interpretation, Gray could also have been understood as limited to the meaning of the term “producer.”⁶⁷See id. at 411. Two later cases of the trilogy, however, suggested that comparable interpretive principles applied to other statutory terms.

In Dobson v. Commissioner,⁶⁸320 U.S. 489 (1943). the Court addressed the proper interpretation of statutory language in the Revenue Act authorizing review of Tax Court determinations “not in accordance with law.”⁶⁹See id. at 494. The Court read into the statutory review provision a deferential standard, reasoning that “when the court cannot separate the elements of a decision so as to identify a clear-cut mistake of law, the decision of the Tax Court must stand.”⁷⁰Id. at 502. The Court also reasoned that when deciding questions of law, “courts may properly attach weight to the decision of points of law by an administrative body having special competence to deal with the subject matter.”⁷¹Id. In doing so, Dobson echoed the criticisms lodged by scholars against the law-fact distinction, remarking that “[p]erhaps the chief difficulty in consistent and uniform compliance with the congressional limitation upon court review lies in the want of a certain standard for distinguishing ‘questions of law’ from ‘questions of fact.’”⁷²Id. at 500–01; see, e.g., John Kelley Co. v. Comm’r, 326 U.S. 521, 527 (1946) (relying on Dobson and remarking on the difficulty “in drawing a line between questions of fact and questions of law”). In this fashion, Dobson appeared to generalize the approach that some might have understood Gray to have taken.

Finally, in NLRB v. Hearst Publications, Inc.,⁷³322 U.S. 111 (1944). the Court reasoned that “where the question is one of specific application of a broad statutory term in a proceeding in which the agency administering the statute must determine it initially, the reviewing court’s function is limited.”⁷⁴Id. at 131. The Court rejected the argument that it could “import wholesale the traditional common-law conceptions” into the statutory term “employee,” instead reasoning that the agency construction “is to be accepted if it has ‘warrant in the record’ and a reasonable basis in law.”⁷⁵Id. at 125, 130–31. For a discussion of how Hearst seems out of step with the Court’s ordinary treatment of a common-law term (“employee”) incorporated into a statute, see Caleb Nelson, Statutory Interpretation 621–22 (2011).

These were not the only cases, and the Court was by no means consistent in its approach to statutory interpretation.⁷⁶See Origins of Judicial Deference, supra note 6, at 977–81. For example, the very same year Hearst was decided, the Court seemed to articulate a more forthrightly multifactor and contextual approach to judicial deference in Skidmore v. Swift & Co.⁷⁷323 U.S. 134 (1944). There, the Court reasoned that the “weight” given to an agency’s legal interpretation “depend[ed] upon the thoroughness evident in its consideration, the validity of its reasoning, its consistency with earlier and later pronouncements, and all those factors which give it power to persuade, if lacking power to control.”⁷⁸Id. at 140. And several cases appeared to ignore the approach of Gray, Dobson, and Hearst altogether.⁷⁹See, e.g., Kenneth Culp Davis, Administrative Law § 247, at 887–93 (1951) (first citing, as cases contrary to Gray, Davies Warehouse Co. v. Bowles, 321 U.S. 144, 156 (1944); then citing Otis Co. v. SEC, 323 U.S. 624, 637–38 (1945); and then citing SEC v. Central-Illinois Sec. Corp., 338 U.S. 96, 127–29 (1949)); Packard Motor Car Co. v. NLRB, 330 U.S. 485, 491 (1947). Thus, writing in 1951, Professor Kenneth Culp Davis remarked that “[t]he one statement that can be made with confidence about applicability of the doctrine of Gray v. Powell is that sometimes the Supreme Court applies it and sometimes it does not.”⁸⁰Davis, supra note 79, § 248, at 893.

II.      Drafting the APA

When Congress enacted the APA in 1946, it included a section entitled “Scope of review” setting forth a series of standards that courts should employ when reviewing agency action.⁸¹5 U.S.C. § 706. This Part traces the drafting of this scope-of-review provision, from a vetoed bill in 1939 to its final form in the enacted statute.

A.      Antecedents

The genesis of a modern code governing administrative procedure and review can be found seven years before the APA’s enactment, when Representative Francis Walter and Senator Mills Logan introduced such a bill in the House and Senate.⁸²H.R. 6324, 76th Cong. (1939); S. 915, 76th Cong. (1939). Although President Franklin Delano Roosevelt ultimately vetoed what was known as the “Walter-Logan bill,”⁸³86 Cong. Rec. 13942–43, 13945 (1940). its provisions reflect how Congress approached the question of scope of review during the era.⁸⁴Although this Article focuses on the Walter-Logan bill’s scope-of-review provisions, those provisions were not the most significant aspects of the bill. The Walter-Logan bill created a seemingly cumbersome internal three-person board review framework, but it tempered that apparatus by carving out of its jurisdiction several agencies. H.R. 6324 § 4(a), at 6. As with the later APA, the Walter-Logan bill did not apply to military or naval operations or courts martial. Id. § 7(b), at 15–16. Moreover, as introduced by Representative Francis Walter, the proposal also did not apply to the Federal Reserve Board, the Office of the Comptroller of the Currency, the Federal Deposit Insurance Corporation, the Interstate Commerce Commission, the Department of State, the Department of Justice (including U.S. attorneys), or any matters “concerning or relating to the internal revenue, customs, patent, trade-mark, copyright, personnel, or longshoremen and harbor workers’ laws . . . or any case where the aggrieved party may be dissatisfied with a grading service in connection with the purchase or sale of agricultural products, or has failed to receive appointment or employment by any agency or independent agency.” Id.

First, the Walter-Logan bill provided that courts would have the authority to determine whether any “administrative rule” “is in conflict with the Constitution of the United States or the statute under which [it was] issued.”⁸⁵Id. § 3, at 5. Notably, unlike the APA, see 5 U.S.C. § 553, the Walter-Logan bill did not differentiate between legislative rules, interpretative rules, and statements of policy. Instead, the bill referred generally to “administrative rules,” which included “rules, regulations, orders, and amendments thereto of general application issued by officers in the executive branch of the United States Government interpreting the terms of statutes they are respectively charged with administering.” H.R. 6324 § 1(1), at 1. Accordingly, the Walter-Logan bill expressly linked “administrative rules” with generality, as well as with the task of statutory interpretation. To emphasize this aspect of judicial review, the bill reiterated that federal courts could invalidate rules “for violation of the Constitution or for conflict with a statute or for lack of authority conferred upon the agency issuing it by the statute or statutes pursuant to which it was issued or for failure to comply with [the bill’s procedural provisions].”⁸⁶H.R. 6324 § 3, at 5. The standard of review for rules, thus, included judicial review for “conflict” with statutory authority.

Second, the bill contained a separate standard-of-review provision for administrative agency “decisions or orders.”⁸⁷Id. § 5, at 11 (capitalization altered). As to such “decisions or orders,” the bill provided that they should be set aside if a court determined:

(1) that the findings of fact are clearly erroneous; or (2) that the findings of fact are not supported by substantial evidence; or (3) that the decision is not supported by the findings of fact; or (4) that the decision was issued without due notice and a reasonable opportunity having been afforded the aggrieved party for a full and fair hearing; or (5) that the decision is beyond the jurisdiction of the agency or independent agency, as the case may be; or (6) that the decision infringes the Constitution or statutes of the United States; or (7) that the decision is otherwise contrary to law.⁸⁸Id. § 5(a), at 13–14.

Most pertinently, the provision authorized a court to set aside an order if the agency’s decision “infringe[d] . . . statutes” or was “contrary to law.”

Thus, the Walter-Logan bill provided separate standards for judicial review of “administrative rules” and agency “decisions or orders.”⁸⁹Id. §§ 3–5(a), at 5, 11. But as to legal questions, both standards appeared to depend on a conflict with statute.

President Roosevelt vetoed the Walter-Logan bill, but he responded to the urge to enact a generalized administrative code by forming a Committee on Administrative Procedure, which produced a “Final Report” shortly thereafter.⁹⁰See Final Report of the Attorney General’s Committee on Administrative Procedure, S. Doc. No. 77-8, app. A, at 252–53 (1941) [hereinafter Final Report]. The Committee’s majority analyzed the preexisting standards of review that Congress had included in different statutory schemes, which varied from statute to statute.⁹¹Id. at 89. Some, such as certain provisions governing review of orders of the Interstate Commerce Commission, were silent on the “scope of review.”⁹²Id. The majority cited, in this respect, the provision authorizing courts “to enjoin, set aside, annul, or suspend” orders of the Interstate Commerce Commission under the Urgent Deficiencies Act of 1913, 38 Stat. 208 (1913), and suits “for the enforcement . . . of any order of the Interstate Commerce Commission other than for the payment of money,” 36 Stat. 555 (1910). Id. For discussion of these provisions, see Aditya Bamzai, The Path of Administrative Law Remedies, 98 Notre Dame L. Rev. 2037, 2047–51 (2023). Others, by contrast, established a standard of review for factual findings, while saying nothing about the standard of review for legal questions.⁹³See Federal Trade Commission Act, 15 U.S.C. § 45(c) (establishing that the “findings of the Commission as to the facts, if supported by evidence, shall be conclusive”); Securities Exchange Act, 15 U.S.C. § 78y(a)(4) (establishing that findings of fact shall be conclusive if they are “supported by substantial evidence”); Federal Alcohol Administration Act, 27 U.S.C. § 204(h) (same); Federal Power Commission Act, 16 U.S.C. § 825l(b) (same). For the series of cases interpreting, and the statutory amendment inserting the word “substantial” in, the National Labor Relations Act, see infra notes 96, 148–49, 151. Still others, such as the Communications Act of 1934, addressed both the standard of review for legal and factual questions, using the following language (which the majority quoted in full): “review by the court shall be limited to questions of law and that findings of fact by the Commission, if supported by substantial evidence, shall be conclusive unless it shall clearly appear that the findings of the Commission are arbitrary or capricious.”⁹⁴Communications Act of 1934, ch. 652, 48 Stat. 1063, 1094 (codified as amended at 47 U.S.C. § 402). Notably, the Communications Act of 1934 contained both a standard of review for “questions of law” as well as a reference to the “arbitrary or capricious” standard, both of which would later be incorporated into the APA. See id.; 5 U.S.C. § 706. The majority noted that the Urgent Deficiencies Act of 1913 set the standard of review for other FCC orders without “any [further] provision as to the scope of review.” Final Report, supra note 90, at 89 (citing 50 Stat. 197 (1937), codified as amended at 47 U.S.C. § 402).

The majority recognized that some of the “differences in language” in the various statutory review provisions “seem to involve no difference in meaning.”⁹⁵Final Report, supra note 90, at 89. For example, the majority noted that the Supreme Court had interpreted the requirement in the National Labor Relations Act that courts treat findings of fact as conclusive “if supported by evidence” as identical to the “substantial evidence” standard expressly used in other statutes.⁹⁶See id. at 89 & n.50 (citing Consol. Edison Co. v. NLRB, 305 U.S. 197, 229 (1938)) (reasoning that the “statute in providing that ‘the findings of the Board as to the facts, if supported by evidence, shall be conclusive’ . . . means supported by substantial evidence . . . such relevant evidence as a reasonable mind might accept as adequate to support a conclusion”); see also id. (noting that “[s]imilar conclusiveness is attached to findings of the Interstate Commerce Commission” despite a lack of a provision establishing the standard of review); Carl McFarland, Judicial Control of the Federal Trade Commission and the Interstate Commerce Commission 135–37 (1933). Other “variations in statutory language,” however, “seem to make a substantial difference.”⁹⁷Final Report, supra note 90, at 90.

Somewhat more speculatively, the majority reasoned that:

Even on questions of law [independent] judgment [by the court] seems not to be compelled. The question of statutory interpretation might be approached by the court de novo and given the answer which the court thinks to be the “right interpretation.” Or the court might approach it, somewhat as a question of fact, to ascertain, not the “right interpretation,” but only whether the administrative interpretation has substantial support. Certain standards of interpretation guide in that direction. Thus, where the statute is reasonably susceptible of more than one interpretation, the court may accept that of the administrative body. Again, the administrative interpretation is to be given weight—not merely as the opinion of some men or even of a lower tribunal, but as the opinion of the body especially familiar with the problems dealt with by the statute and burdened with the duty of enforcing it. This may be particularly significant when the legislation deals with complex matters calling for expert knowledge and judgment.⁹⁸Id. at 90–91 (footnote omitted).

This reasoning mirrored John Dickinson’s argument on the difficulty of drawing the line between questions of fact and questions of law.⁹⁹Id. at 88 n.37 (citing Dickinson, supra note 54, at 55). And the reasoning foreshadowed the approach embraced in cases like Gray, Hearst, and Dobson—all decided a few years following the Final Report.¹⁰⁰Gray v. Powell, 314 U.S. 402, 411–12 (1941); Dobson v. Comm’r, 320 U.S. 489, 501–02 (1943); NLRB v. Hearst Publ’ns, Inc., 322 U.S. 111, 131 (1944). The sole case that the Final Report cited on this point was SEC v. Associated Gas & Electric Co., 99 F.2d 795 (2d Cir. 1938), which could be understood as a case applying the customary and contemporaneous canons of construction. See Final Report, supra note 90, at 90; Associated Gas & Electric Co., 99 F.2d at 798 (noting “uniform[] treat[ment]” and “long settled practice,” as well as the “benefit of [the agency’s] special knowledge acquired through continuous experience in a difficult and complicated field”).

At the end of the day, the Committee’s majority recommended that Congress not change the scope of review “so long as the courts continue to discharge conscientiously the functions of review.”¹⁰¹Final Report, supra note 90, at 209; see also Dean Acheson, Summary of Attorney General’s Committee Report, reprinted in 27 A.B.A. J. 143, 145 (1941) (“The Committee believes that the judicial review which now exists is wise and should be maintained . . . [T]he Committee believes that [changes] may not wisely be effected by general legislation.”). The majority’s draft bill did not codify a standard for the scope of review of agency decisions.¹⁰²See Final Report, supra note 90, at 191–202.

A “minority” of the Committee—composed of Carl McFarland, E. Blythe Stason, and Arthur Vanderbilt—claimed that “Congress should prescribe the scope of judicial review rather than leave it to the courts” without any guidance.¹⁰³Id. at 209. The minority pointed in part to the “haphazard, uncertain, and variable results of the present system or lack of system of judicial review.”¹⁰⁴Id. at 210. Like the majority, the minority relied on the previously cited “general statutory phrases now in use, purporting to express [ ] the congressional intent as to the scope of judicial review of administrative determinations of facts.” Id. at 210 & n.3 (citing the statutory review provisions listed supra note 93, as well as those from the Interstate Commerce Act, Walsh-Healey Act, and Commodities Exchange Act). These statutes had, in the minority’s words, led to “[w]ide variations in results in specific cases [that] defy explanation.” Id. at 210. They also pointed to “a fundamental change” that, according to them, revolutionized the scope of judicial review, previously based on “due process, separation of powers, and the nature of judicial power under Article III of the Constitution.”¹⁰⁵Id. at 210. For this proposition, the minority cited the Supreme Court’s then-recent opinion in Railroad Commission of Texas v. Rowan & Nichols Oil Co., 310 U.S. 573 (1940). In that (now somewhat obscure) case, Justice Frankfurter, writing for the Court, rejected a Fourteenth Amendment challenge to an oil proration order from a state agency (the Railroad Commission of Texas), see id. at 577, remarking in that context that “the evolution of these formulas belongs to the Commission and not to the judiciary” and that “courts must not substitute their notions of expediency and fairness for those which have guided the agencies to whom the formulation and execution of policy have been entrusted,” id. at 580–81. Justice Roberts, in dissent, contended that the Court’s opinion “announce[d] principles with respect to the review of administrative action challenged under the due process clause directly contrary to those which have been established.” Id. at 585 (Roberts, J., dissenting) (citing Thompson v. Consol. Gas Utilities Corp., 300 U.S. 55 (1937)). As the minority of both the Attorney General’s Report and the Rowan & Nichols Court saw it, the Court’s holding suggested that “fact issues involving due process, equal protection, and doubtless also other constitutional guarantees will in all probability no longer be subject to court review as a matter of constitutional right.” Final Report, supra note 90, at 210; see also Barry Cushman, Court-Packing and Compromise, 29 Const. Comment. 1, 28–29 (2013).

All that said, the minority focused, not on judicial review of legal questions, but rather on factual issues. According to the minority, courts had struggled to interpret the “substantial evidence” rule included in many statute’s judicial review provisions.¹⁰⁶Final Report, supra note 90, at 210. Specifically, the minority pointed to decisions that held the standard was satisfied if “‘substantial evidence’ is found anywhere in the record to support conclusions of fact . . . without reference to how heavily the countervailing evidence may preponderate—unless indeed the stage of arbitrary decision is reached.” Id. at 210–11. In contrast, the minority advocated that review of findings of fact should occur “upon the whole record.” Id. at 211. Those provisions, moreover, “fail[ed] to take account of differences between the various types of fact determinations,” some of which involved “highly technical matters and require special experience and training,” whereas others “impinge heavily upon private rights.”¹⁰⁷Id. at 211. More generally, the minority contended that the then-existing standards of review were “unsatisfactory because of the very manner of their establishment”¾namely “the usual case-to-case procedure of the courts.”¹⁰⁸Id.

Like the Walter-Logan bill, the minority proposed two different standards of review¾one for rules and one for adjudications, with both authorizing judicial review for lack of statutory authority. The minority’s proposed bill authorized judicial review of administrative adjudication for all questions of (1) constitutional right, power, privilege, or immunity; (2) the statutory authority or jurisdiction of the agency; (3) the lawfulness and adequacy of procedure; (4) findings, inferences, or conclusions of fact unsupported, upon the whole record, by substantial evidence; and (5) administrative action otherwise arbitrary or capricious.¹⁰⁹See id. at 246; id. at 230 (authorizing courts to review rules for “statutory authority or discretion of the agency, including the propriety of interpretative rules”); see also Additional Views and Recommendations of Messrs. McFarland, Stason and Vanderbilt, reprinted in 27 A.B.A. J. 146, 146–47 (1941) (outlining the minority’s proposal for a comprehensive code of administrative procedure). The minority draft also included a judicial review proviso stating that “upon such review due weight shall be accorded the experience, technical competence, specialized knowledge, and legislative policy of the agency involved as well as the discretionary authority conferred upon it.” Final Report, supra note 90, at 246–47. Contemporaneous scholars pointed out that the final version of the APA omitted this language. See Dickinson, supra note 14, at 518 n.40 (noting that, although the APA “adopts most of the judicial review provisions of the minority bill,” the proviso “seems never to have been seriously considered by Congress or its committees”).

B.      The APA: Text, Structure, and Legislative History

When Congress returned to the business of enacting an administrative code after a hiatus prompted by World War II, the authors of the legislation began with the Final Report’s minority proposal as a starting point.¹¹⁰See Nathaniel L. Nathanson, Some Comments on the Administrative Procedure Act, 41 Ill. L. Rev. 368, 414 (1946) (“In general Section 10 of the Act adopts the proposal of the minority of the Attorney General’s Committee.”); Shepherd, supra note 6, at 1649 (observing that bills introduced in 1943 “mirrored the Attorney General’s Committee’s minority bill” and became the APA after two years of negotiation). But they altered the standard-of-review provision to produce the final language of section 706. As originally introduced, the bill that became the APA did not include language authorizing courts to “interpret constitutional and statutory provisions” or to set aside agency action “otherwise not in accordance with law.”¹¹¹See § 9(e), S. 7, 79th Cong. (1945).

The final Act included such language. As enacted, the prefatory clause to the scope-of-review provision authorized a reviewing court to “decide all relevant questions of law, interpret constitutional and statutory provisions, and determine the meaning or applicability of the terms of an agency action.”¹¹²5 U.S.C. § 706. Section 706 then lists a series of standards of review that, seemingly, apply in specific situations. Of relevance, section 706(2)(A) provides that a court should “hold unlawful and set aside agency action, findings, and conclusions” that are “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.”¹¹³Id. § 706(2)(A). And sections 706(2)(E) and (F) authorize a court to hold unlawful and set aside agency findings “unsupported by substantial evidence” in “formal” proceedings and any findings “unwarranted by the facts to the extent that the facts are subject to trial de novo by the reviewing court.”¹¹⁴Id. § 706(2)(E)–(F). The final clause of section 706 requires the court to “review the whole record or those parts of it cited by a party” and to take “due account . . . of the rule of prejudicial error.” Id. § 706. And in sections 706(B) through (D), the APA authorizes review of agency decisions “contrary to constitutional right, power, privilege, or immunity”; “in excess of statutory jurisdiction, authority, or limitations, or short of statutory right”; and “without observance of procedure required by law.” Id. § 706(B)–(D).

To begin, consider that the prefatory clause instructs courts to “decide all relevant questions of law,” which seems to contemplate a form of de novo review for legal questions.¹¹⁵Id. § 706. It also authorizes courts to “interpret constitutional and statutory provisions,” which appears to contemplate that statutory and constitutional interpretation will be conducted using the same standard of review.¹¹⁶Id. In addition, although the “not in accordance with law” language in section 706(2)(A) requires more context and interpretation, Congress had used such language in statutory standards of review before the APA’s enactment in a manner tending to suggest de novo review.¹¹⁷A matter that I will discuss below. See infra Section II.C.1. Finally, the provision contemplates deferential “substantial evidence” review of factfinding in certain agency proceedings. When taken together with the prefatory clause, the provision reestablishes a distinction between the standard of review for questions of “law” and “fact” that would have been familiar in the mid-1940s. Questions of law would be reviewed using a court’s independent judgment (subject to the relevant canons of construction), while questions of fact would be reviewed deferentially.¹¹⁸The primary textual argument to the contrary is Justice Kagan’s claim in Kisor v. Wilkie, 139 S. Ct. 2400 (2019), that section 706 “does not specify the standard of review a court should use,” particularly given “the practice of judicial review at the time of the APA’s enactment.” Id. at 2419 (plurality opinion). But if taken seriously, that claim would mean that the courts could adopt any standard of review whatsoever, including complete abdication of review. Perhaps for that reason, academic supporters of Justice Kagan’s general position have retreated from this particular claim. For example, although Professor Levin sometimes argues that section 706 “is essentially noncommittal on the issue of” deference, Levin, supra note 19, at 190, he elsewhere acknowledges that “the APA presupposes a common law background” and that a court’s “elaborations on its judicial review requirements must bear at least a reasonable relationship to prevailing principles as of 1946.” Id. at 143. Thus, according to Professor Levin, “[a] regime in which courts may not review agency legal interpretations at all would be fundamentally incompatible with administrative practice as of the time of the APA.” Id. If that is correct, then section 706 is not entirely “noncommittal” on the standard of review, but rather rules out certain possible interpretive approaches based on the intentions of the APA’s drafters and the relevant “common law background.” See id. at 143. If, as I have argued, the relevant traditional principles gave respect to customary and contemporaneous interpretation, see supra Section I.A.1, then those principles provided the relevant backdrop to which our current interpretive regime must bear a reasonable relationship.

To the extent that legislative history might be relevant to understanding section 706, it tends to cut in favor of this understanding. Before passage of the bill that became the APA, Representative Francis Walter, the chairman of the House Subcommittee on Administrative Law, described the scope-of-review provision as “requir[ing] courts to determine independently all relevant questions of law, including the interpretation of constitutional or statutory provisions. . . .”¹¹⁹92 Cong. Rec. 5654 (1946) (statement of Rep. Walter). Professor Levin contends that the term “independently” in this statement is ambiguous because, shortly thereafter, Representative Walter remarked that “[t]he term ‘substantial evidence’ as used in this bill means evidence which on the whole record as reviewed by the court and in the exercise of the independent judgment of the reviewing court is material to the issues, clearly substantial, and plainly sufficient to support a finding or conclusion.” Levin, supra note 19, at 156 (emphasis omitted) (citing 92 Cong. Rec. 5654 (1946) (statement of Rep. Walter)). Professor Levin argues that Walter “must have meant something more modest” when he earlier used the term “independently” because “courts do not find facts de novo when they conduct substantial evidence review,” id. at 157, and, hence, Walter’s second use of the term “independent” suggests a form of reasonableness review, see id. But note that Walter refers to a court’s “exercise of . . . independent judgment” to determine the “materiality” of the issues, which could be understood as a form of de novo review. Id. at 156 (emphasis omitted) (citing 92 Cong. Rec. 5654 (1946) (statement of Rep. Walter)). The most natural reading is that Representative Walter had the ordinary meaning of “independent” in mind both times he used the term. Moreover, in the context of a discussion of why “interpretative rules” were exempted from the APA’s notice-and-comment requirements,¹²⁰See 5 U.S.C. § 553(b)(3)(A). a Senate Judiciary Committee print indicated that agency statutory interpretations “are subject to plenary judicial review.”¹²¹S. Doc. No. 79-248, at 18 (1946). The House and Senate reports also provided that “questions of law are for the courts rather than agencies to decide in the last analysis.”¹²²H.R. Rep. No. 79-1980, at 278 (1946); S. Rep. No. 79-752, at 214 (1945). The House report included a “diagram synopsis” of the APA, which said that “reviewing courts . . . are to determine all questions of law . . . and hold unlawful action found . . . in violation of any statute.”¹²³H.R. Rep. No. 79-1980, at 267; see generally Origins of Judicial Deference, supra note 6, at 988–90 (summarizing other portions of the legislative history).

C.      Adjacent Statutes: The Congressional Repudiation of Dobson and Hearst

In the era surrounding the APA’s passage, Congress repudiated two of the leading 1940s cases addressing judicial deference—Dobson and Hearst.¹²⁴See supra Section I.B. Congress’s actions cast light on the APA. Consider the two actions in turn.

1.      Dobson v. Commissioner

When Congress repudiated Dobson, it did so by construing language—”not in accordance with law”—that it had used in the APA.¹²⁵Dobson v. Comm’r, 320 U.S. 489, 494 (1943) (quoting Revenue Act of 1926 § 1003(b), 44 Stat. 9, 110). Congress had previously used the phrase “not in accordance with law” in a statutory review provision in 1926 to describe the federal courts’ authority over the newly constituted Board of Tax Appeals.¹²⁶See Revenue Act of 1926, § 1003(b), 44 Stat. 9, 105–06, 109–10. Before 1924, a taxpayer could contest the Commissioner of Internal Revenue’s determination of the amount of tax only after payment.¹²⁷See Old Colony Trust Co. v. Comm’r, 279 U.S. 716, 721 (1929). In the Revenue Act of 1924, Congress altered this system by creating the Board—the predecessor of the current U.S. Tax Court—to give taxpayers an avenue to review the Commissioner’s determinations of taxes owed,¹²⁸Revenue Act of 1924, § 900, 43 Stat. 253, 336–38. though initially without any further avenue for direct federal court review.¹²⁹Old Colony Trust, 279 U.S. at 721 (“There was under the Act of 1924 no direct judicial review of the proceedings before the Board of Tax Appeals. But each party had the unhindered right to seek separate action by a court of competent jurisdiction to test the correctness of the Board’s action.”). Two years later, in the 1926 statute, Congress “continued” the Board “as an independent agency in the Executive Branch of the Government.”¹³⁰§ 900, 44 Stat. 9, 105-06. The Board was established with sixteen members “appointed by the President, by and with the advice and consent of the Senate” who could “[b]e removed by the President, after notice and opportunity for public hearing, for inefficiency, neglect of duty, or malfeasance in office, but for no other cause.” § 901(a), 44 Stat. at 106. The statute was enacted on February 26, 1926, see id. at 9, just over ten months after the reargument, and eight months before the decision, in Myers v. United States, 272 U.S. 52, 52 (1926).

The Revenue Act of 1926 authorized courts of appeals to exercise direct judicial review of the Board of Tax Appeals through “exclusive jurisdiction to review the decisions of the Board.”¹³¹§ 1003(a), 44 Stat. at 110. In conducting such review, federal courts were given the “power to affirm or, if the decision of the Board is not in accordance with law, to modify or to reverse the decision of the Board.”¹³²§ 1003(b), 44 Stat. at 110 (emphasis added). Three years after the Board’s creation, writing for the Court in Old Colony Trust, Chief Justice Taft reasoned that the review provisions established by the Revenue Act of 1926 were constitutional. In doing so, Taft noted that “[t]he Board of Tax Appeals is not a court,” but rather “an executive or administrative board, upon the decision of which the parties are given an opportunity to base a petition for review to the courts after the administrative inquiry of the Board has been had and decided.” Old Colony Trust, 279 U.S. at 725. During the same congressional session that Congress enacted the Revenue Act of 1926, it used the same language to authorize federal courts to review orders issued under the Longshoremen’s and Harbor Workers’ Compensation Act (“LHWCA”).¹³³Longshoremen’s and Harbor Workers’ Compensation Act, § 21(b), 44 Stat. 1424, 1436 (1927) (codified as amended at 33 U.S.C. §§ 901–50).

The Supreme Court first addressed the meaning of this phrase while construing the LHWCA in the landmark opinion in Crowell v. Benson.¹³⁴285 U.S. 22, 44 (1932). Recall that, in Crowell, both Chief Justice Hughes’s majority and Justice Brandeis’s dissent agreed that courts would review legal questions de novo. Chief Justice Hughes explained that “[r]ulings of the [agency official] upon questions of law are without finality,” with “full opportunity . . . afforded for their determination by the Federal courts through proceedings to suspend or to set aside a compensation order.”¹³⁵Id. at 45–46; id. at 49 (reasoning that the statute left “no doubt of the intention to reserve to the Federal court full authority to pass upon all matters which this Court had held to fall within” the category of questions of law). In his dissent, Justice Brandeis observed that “[t]he initial question” that he would address was “one of construction of the” LHWCA.¹³⁶Id. at 66 (Brandeis, J., dissenting). He noted that the LHWCA provided that “‘if not in accordance with law, a compensation order may be suspended or set aside, in whole or in part, through injunction proceedings . . . instituted in the Federal district court.’”¹³⁷Id. at 67 (emphasis added); see also id. (observing that “[t]he phrase in [the LHWCA] providing that the order may be set aside ‘if not in accordance with law’ was adopted from the statutory provision, enacted by the same Congress, for review by the Circuit Courts of Appeals of decisions of the Board of Tax Appeals”). On this matter, Justice Brandeis agreed with the Crowell majority that, under the LHWCA, the agency’s “conclusions are, as a matter of right, open to reexamination in the courts on all questions of law.”¹³⁸Id. at 88. As Brandeis put it, the LHWCA “created fact-finding and fact-gathering tribunals.” Id. Under the statutory standard of review, these tribunals did not displace the federal courts’ decisionmaking “on all questions of law” and, as Brandeis noted, “may be open even on all questions of the weight of the evidence.” Id. As he put it, “the prevailing practice” under the LHWCA’s review provisions “confin[es]” judicial review “to questions of law.”¹³⁹Id. at 93. Thus, as a statutory matter, Justice Brandeis agreed that federal courts could freely reexamine “questions of law” under the “not in accordance with law” standard.¹⁴⁰Id. at 73, 88.

As previously discussed,¹⁴¹See supra notes 68–72 and accompanying text. just a decade later in Dobson, the Court read into the Revenue Act’s identical language a seemingly deferential standard. The Court reasoned that courts should not disturb a Tax Court decision “when the court cannot separate the elements of a decision so as to identify a clear-cut mistake of law,” and that “[i]n deciding law questions courts may properly attach weight to the decision of points of law by an administrative body having special competence to deal with the subject matter.”¹⁴²Dobson v. Comm’r, 320 U.S. 489, 502 (1943). Fascinating insight on the drafting of Dobson—that tends to support the proposition that its author knowingly changed the law in this area—can be found in explorations of Justice Jackson’s papers. See John Q. Barrett, Justice Jackson on “What the Law’s Going to Be”—At Least Until Its “Gelding,” 6 Green Bag 2d 125, 127 (2003) (reporting that Justice Jackson told a law clerk who wrote a memorandum supporting traditional review of the Board that “you may be right about what the law is, but that’s not what it’s going to be”); Kirk J. Stark, The Unfulfilled Tax Legacy of Justice Robert H. Jackson, 54 Tax. L. Rev. 171, 221–44 (2001); cf. Arrowsmith v. Comm’r, 344 U.S. 6, 12 (1952) (Jackson, J., dissenting) (remarking on Dobson’s statutory demise). For a modern assessment of judicial review of the Tax Court, see Leandra Lederman, (Un)Appealing Deference to the Tax Court, 63 Duke L.J. 1835 (2014).

But Dobson was not long for this world. Two years after enacting the APA (with the “not in accordance with law” language contained in section 706(2)(A)), Congress overturned Dobson by statute.¹⁴³Act of June 25, 1948, ch. 646, § 36, 62 Stat. 869, 991 (codified as amended at 26 U.S.C. § 7482(a)(1)) (providing that the courts of appeals should review Tax Court decisions “in the same manner and to the same extent as decisions of the district courts in civil actions tried without a jury”). Supporters of the bill, such as Representative Sam Hobbs—the author of the Administrative Orders Review Act (otherwise known as the Hobbs Act)¹⁴⁴28 U.S.C. § 2342; see Jason N. Sigalos, The Other Hobbs Act: An Old Leviathan in the Modern Administrative State, 54 Ga. L. Rev. 1095, 1111–14 (2020).—contended that “[p]rior to the Dobson decision it was assumed by all the courts, including the Supreme Court, that on appeal from the Tax Court all questions of law were fully reviewable.”¹⁴⁵93 Cong. Rec. A3281 (1947) (statement of Rep. Sam Hobbs); see also 94 Cong. Rec. 8500–01 (1948) (statement of Rep. Reed) (similar).

Thus, although the 1948 legislation amended just the Revenue Act (and not the APA), it did so in a manner that suggested Congress’s understanding of the meaning of the statutory review phrase “not in accordance with law.”¹⁴⁶See 26 U.S.C. § 7482(a)(1). According to Representative Hobbs, all had agreed that, prior to Dobson, such language rendered questions of law “fully reviewable.”¹⁴⁷93 Cong. Rec. A3281.

2.      NLRB v. Hearst Publications, Inc.

Congress also appeared to repudiate NLRB v. Hearst Publications, Inc., shortly after the passage of the APA. Here, too, a brief overview of the steps leading to Congress’s actions is revealing.

In the Wagner Act of 1935, Congress specified that “[t]he findings of the Board as to the facts, if supported by evidence, shall be conclusive.”¹⁴⁸National Labor Relations Board Act, § 10(e), 49 Stat. 449, 454 (1935) (codified as amended at 29 U.S.C. § 160(e). In the following years, the Court read the word “evidence” in the Wagner Act to mean “substantial evidence.”¹⁴⁹Wash., Va. & Md. Coach Co. v. NLRB, 301 U.S. 142, 146–47 (1937); Consol. Edison Co. v. NLRB, 305 U.S. 197, 229 (1938) (reasoning that “[s]ubstantial evidence is more than a mere scintilla,” but rather “such relevant evidence as a reasonable mind might accept as adequate to support a conclusion”); see also NLRB v. Columbian Enameling & Stamping Co., 306 U.S. 292, 300 (1939). As previously discussed, interpreting the “substantial evidence” test in Hearst, the Court reasoned that “where the question is one of specific application of a broad statutory term in a proceeding in which the agency administering the statute must determine it initially, the reviewing court’s function is limited.”¹⁵⁰NLRB v. Hearst Publ’ns, Inc., 322 U.S. 111, 131 (1944).

When Congress passed the Taft-Hartley Act in 1947,¹⁵¹See Labor Management Relations Act, ch. 120, 61 Stat. 136 (1947) (codified as amended at 29 U.S.C. § 141). an accompanying House Report explained that the statute’s judicial review provisions “will be adequate to preclude” the holdings in several cases.¹⁵²H.R. Rep. No. 80-510, at 56 (1947) (Conf. Rep.). Among the listed cases was Hearst.¹⁵³See id. A few years later, the Court addressed the implications of the Taft-Hartley Act for the scope of review of factual issues in the landmark opinion of Universal Camera Corporation v. National Labor Relations Board.¹⁵⁴340 U.S. 474 (1951). In doing so, the Court made two remarks relevant to the scope of review of legal questions. First, the Court noted that “[i]t would be mischievous word-playing to find that the scope of review under the Taft-Hartley Act is any different from that under the [APA].”¹⁵⁵Id. at 487. Second, the Court recognized Congress’s repudiation of Hearst, quoting the House Report language.¹⁵⁶See id. at 485–86; see H.R. Rep. No. 80-510, supra note 152, at 56.

As a result, in this context, too, members of Congress were alert to cases like Dobson and Hearst and sought to repudiate them.

D.      Contemporaneous Commentary

Reactions to the APA’s standard-of-review provision came from the executive branch, congressional supporters, and the scholarly community. The executive branch’s interpretation of the APA came in the form of the Attorney General’s Manual on the Administrative Procedure Act, which was published by the Department of Justice a year after the APA’s enactment.¹⁵⁷U.S. Dep’t of Just., Attorney General’s Manual on the Administrative Procedure Act 5–7 (1947). The Department characterized section 706 as “restat[ing] the present law as to the scope of judicial review” and as a “general restatement of the principles of judicial review embodied in many statutes and judicial decisions.”¹⁵⁸Id. at 93, 108. Even assuming, however, that the “restatement” characterization was accurate,¹⁵⁹But see Kisor v. Wilkie, 139 S. Ct. 2400, 2436 (2019) (Gorsuch, J., concurring in the judgment) (noting that the Manual “reflected the interests of the executive branch” and expressed views “far from universally shared”). the manual did not explain what exactly was being “restated.” Recall that the Court did not consistently apply Gray during this era.¹⁶⁰See supra notes 64–67 and accompanying text. As a result, terming the scope-of-review provision a “restatement” did not resolve which line of inconsistent cases the section restated. The “restatement,” for instance, might well have been of the traditional independent-judgment rule.¹⁶¹See, e.g., Sunstein, supra note 19, at 1652–53 (agreeing that, although the Manual described the APA as restating principles of judicial review, “frustratingly, the Manual did not say what those principles were”).

By contrast, Senator Pat McCarran, the chair of the Senate Committee on the Judiciary when the APA was enacted, wrote an article contending that the APA “simply and expressly provides that Courts ‘shall decide all relevant questions of law,’” a provision that would cut down “the ‘cult’ of discretion” that had “gained considerable currency in the last decade or so.”¹⁶²Pat McCarran, Improving “Administrative Justice”: Hearings and Evidence; Scope of Judicial Review, 32 A.B.A. J. 827, 828, 831 (1946).

On the scholarly front, John Dickinson echoed Senator McCarran’s perspective.¹⁶³Dickinson, supra note 14, at 516. He contended that it had “been generally understood that in a review proceeding[,] questions of law are for the determination of the reviewing Court.”¹⁶⁴Id. But Dickinson claimed that “[i]ncreasingly, in recent years the Supreme Court has tended to treat many issues, which, when subjected to adequate analysis, would be seen to be issues of law, as lying within the discretion of an administrative agency, and, therefore, non-reviewable.”¹⁶⁵Id.; see id. at 516–17 (arguing that courts had “begun” to distinguish “between two kinds of questions of law: Those which involve what are sometimes spoken of as general law or legal principles, and others which involve the construction of technical terms and the application of knowledge thought to be expert and specialized,” which left to “the Court’s discretion the determination of whether it would or would not review a legal question”). According to Dickinson, section 706 was a “clear mandate” requiring a court to decide legal questions “for itself, and in the exercise of its own independent judgment.”¹⁶⁶Id. at 516.

Dickinson was not alone in making these two points.¹⁶⁷In this regard, I part ways with the way others have characterized scholarship written roughly contemporaneously with the APA’s passage. See Levin, supra note 19, at 181 (describing Dickinson’s views as “almost completely isolated,” with only a single other article endorsing his analysis, and claiming that “[a]part from this one exception . . . the verdict of contemporary scholarship regarding Dickinson’s position appears to have been entirely negative”); Sunstein, supra note 19, at 1653 n.206 (describing Dickinson as “the only prominent contemporaneous voice on behalf of the specific view that section 706 had changed the law with respect to judicial review of agency judgments of law”). As explained in the text, there was a fair deal of support for Dickinson’s understanding of section 706, such that his position was not at all “completely isolated.. For instance, in a 1951 article, Professor Louis Jaffe characterized the Court’s opinions in Gray, Hearst, and Dobson as “rather unfortunate.”¹⁶⁸Louis L. Jaffe, Judicial Review: “Substantial Evidence on the Whole Record,” 64 Harv. L. Rev. 1233, 1258 (1951). To Jaffe, the trilogy’s holdings were “heresy” to the extent that they said that “if the [agency’s] judgment is reasonable the courts are powerless to interfere, though independently they would have arrived at a different conclusion.”¹⁶⁹Id. According to Jaffe, “[t]he question whether the action of the administrative body is within the limits of relevance is always a question for the courts, regardless of how reasonable the agency’s conception of relevance may be.”¹⁷⁰Id. at 1258–59.

Equally importantly, Jaffe claimed that, as of the time he was writing, “[i]t [wa]s thought to be open for decision whether the so-called doctrine of Gray v. Powell has been repealed by the Administrative Procedure Act.”¹⁷¹Id. at 1260 (citing specifically the clause in section 706 providing “that the court shall decide ‘all relevant questions of law’” (quoting 5 U.S.C. § 706)). Jaffe observed that “[i]n the view of some distinguished authorities[,] this provision was intended to overcome the doctrine of Gray v. Powell.”¹⁷²Id. (citing John Dickinson, The Judicial Review Provisions of the Federal Administrative Procedure Act (Section 10) Background and Effect, in Federal Administrative Procedure Act and the Administrative Agencies 546, 587 n.57 (Warren ed. 1947)). And he thought it “significant” that “the Taft-Hartley legislative committees” cited Hearst “with disapproval.”¹⁷³Id. He concluded that “[t]his question”¾namely, whether section 706 rejected Gray, Hearst, and other similar cases¾“must ultimately be faced” and “cannot be avoided by labeling these questions as questions of fact.”¹⁷⁴Jaffe, supra note 168, at 1260. Were the Court to face the question, Jaffe argued that, to the extent that the “opinions suggest what [he had] called heresy (and one quite unnecessary to the decisions in those cases)[,] they should be regarded as disapproved.”¹⁷⁵Id. Jaffe’s perspective was nuanced and perhaps even changed subtly over time. For example, Jaffe sought to maintain “valid areas of administrative discretion” where agency judgment would “be set aside only if unreasonable or arbitrary.” Id. Later, Jaffe observed that “[t]he device of characterizing a question as one of fact or as ‘mixed’ permit[ted] a court to pretend that it must affirm the administrative action if it is ‘supported by evidence’ or is ‘reasonable.’” Jaffe, supra note 62, at 547; see also id. at 569–70, 570 n.79.

Consider, next, Professor Bernard Schwartz’s views in 1955, acknowledging that cases like Gray had “enabled the highest Court to all but nullify language in the Administrative Procedure Act of 1946,” which Schwartz would have interpreted “to eliminate the Gray v. Powell doctrine.”¹⁷⁶Bernard Schwartz, Gray vs. Powell and the Scope of Review, 54 Mich. L. Rev. 1, 67–68 (1955) (footnote omitted) (citing 5 U.S.C. § 706’s language that “the reviewing court shall decide all relevant questions of law” and “interpret . . . statutory provisions”). Schwartz contended that, although section 706 appeared to “eliminate[] the doctrine of narrow review in the Gray v. Powell situation,” the Court had “avoided this result by its refusal to concede that an agency finding of the kind under discussion involves statutory interpretation.”¹⁷⁷Id. at 68. “By its use of its power to classify challenged agency findings,” Schwartz explained, “the Court has been able to maintain the Gray v. Powell doctrine unaltered, despite the seemingly contrary language of the Administrative Procedure Act.”¹⁷⁸Id. Schwartz “doubt[ed] the soundness of the doctrine and hope[d] for its ultimate repudiation.”¹⁷⁹Id. At the same time, he noted that he “still desire[d], so long as it is not overruled, to see it applied in a logically consistent fashion,” which was why he was concerned by “the embarrassingly large number of Supreme Court decisions that do not adhere to the doctrine of Gray v. Powell.”¹⁸⁰Id. For a further discussion by the same author, see Bernard Schwartz, Mixed Questions of Law and Fact and the Administrative Procedure Act, 19 Fordham L. Rev. 73, 82–87 (1950). For Schwartz’s harsh judgment of Chevron, see Bernard Schwartz, “Apotheosis of Mediocrity”? The Rehnquist Court and Administrative Law, 46 Admin. L. Rev. 141, 178 (1994) (“Because of this, one of our most acute administrative law commentators, Louis L. Jaffe, rejected an earlier version of the Chevron doctrine as ‘heresy.’ With Chevron and its progeny, however, the heresy has become accepted doctrine.”). In a nutshell, Schwartz’s article reflected a disagreement with Gray’s approach; a belief that section 706 repudiated Gray’s interpretive methodology; and a concern that the Court’s precedents, circa 1955, were in disarray on this issue.

Others echoed Dickinson, Jaffe, and Schwartz’s perspective.¹⁸¹See, e.g., Thayer D. Moss, The Administrative Interpretation of Statutes, 39 Geo. L.J. 244, 259–60 (1951); Frank Hinman Jr., Effect of the Administrative Procedure Act on Judicial Review of Administrative Action, 20 Rocky Mtn. L. Rev. 267, 276–77 (1948). Another commentator observed that the Act “preserve[d] the customary dichotomy of law and fact, in spite of argument by distinguished commentators that these categories cannot in application be distinguished, and in spite of recent Supreme Court decisions giving support to such argument.” Ray A. Brown, The Federal “Administrative Procedure Act,” 1947 Wis. L. Rev. 66, 86; Alfred Long Scanlan, Judicial Review Under the Administrative Procedure Act—In Which Judicial Offspring Receive a Congressional Confirmation, 23 Notre Dame Law. 501, 528–29, 531 (1948) (reasoning that the “decide all relevant questions of law” language was “simply a restatement of the present powers which reviewing courts possess, and frequently exercise, of reviewing relevant questions of constitutional and statutory law,” but noting that “mixed” questions are “merely another ramification of the substantial evidence rule”); Frederick F. Blachly & Miriam E. Oatman, The Federal Administrative Procedure Act, 34 Geo. L.J. 407, 427–30 (1946) (arguing that the Act “greatly widens the scope of judicial review” and appearing to assume that both before and after the Act, courts used independent judgment for legal questions); Julius Cohen, Legislative Injustice and the Supremacy “of Law,” 26 Neb. L. Rev. 323, 339 (1947) (claiming that “the language of the section leaves no doubt that it was the major purpose of the drafters to tighten substantially the judicial grip on administrative action,” but appearing to focus on factual questions). But some scholarly commentators took the contrary position. For example, Professor Kenneth Culp Davis argued that section 706 did not establish a general standard of review for legal questions.¹⁸²See, e.g., Davis, supra note 79, § 244, at 871; see also id. § 245, at 877 (asking, rhetorically, whether it is “not high time, now that the APA has reasserted [the law-fact] distinction, that the courts should conform to the statutory formula”); Kenneth Culp Davis, Scope of Review of Federal Administrative Action, 50 Colum. L. Rev. 559, 567–69 (1950). And Professor Nathaniel Nathanson summarized, without objection, the Attorney General Manual’s claim that section 706 had restated then-current law on judicial review.¹⁸³Nathanson, supra note 110, at 414; see also Reginald Parker, The Administrative Procedure Act: A Study in Overestimation, 60 Yale L.J. 581, 587 (1951); Herbert Kaufman, The Federal Administrative Procedure Act, 26 B.U. L. Rev. 479, 500–01 (1946).

In the final analysis, contemporaneous scholars embraced a variety of perspectives on section 706’s implications for judicial review of legal questions. Prominent scholars like Dickinson, Jaffe, and Schwartz believed that Congress had enacted the APA to repudiate cases like Gray. To be sure, others disagreed. But such disagreement returns us to the text and structure of section 706 itself and whether (and how) that text can be made to cohere with the concept of judicial deference as it developed in the post-APA era.

III.      Echoes of the APA

This Part addresses the APA’s aftermath before returning to a discussion of section 706’s meaning.

A.      Aftermath

In the decades after the APA’s passage in 1946, the same confusion that Professor Kenneth Culp Davis had identified about the applicability of “the doctrine of Gray v. Powell”¹⁸⁴Davis, supra note 79, § 248, at 893 (“The one statement that can be made with confidence about applicability of the doctrine of Gray v. Powell is that sometimes the Supreme Court applies it and sometimes it does not.”). To be sure, despite acknowledging this inconsistency, Davis went on to claim that “the doctrine of Gray v. Powell has survived the APA.” Id. § 246, at 885. remained a part of the case law. Thus, writing in 1965, Professor Jaffe observed that the Supreme Court’s cases “sometimes assert[ed] the correctness of the agency rule, at other times, going no further than to hold that the administrator can but is not required to adopt such a rule,” giving “rise to profound difficulties of description and analysis, and to intense controversy.”¹⁸⁵Jaffe, supra note 62, at 557–58 (footnotes omitted). In this 1965 work, Jaffe noted that Gray and Hearst had “recognized perhaps more openly than had been customary in the recent past the law- or policy-making function of the agencies” and could be construed as “an abdication of the customary power and responsibility of the judiciary.” Id. at 575. He nevertheless claimed that the cases were “traditional” and “sound.” Id.; compare supra notes 168–175 and accompanying text (describing Jaffe’s views in 1951).

Cases acknowledged this confusion. In 1976, eight years before Chevron, Judge Henry Friendly addressed “the ever troubling question” whether the interpretation of a statute “is the kind of question which justifies or requires judicial deference.”¹⁸⁶Pittston Stevedoring Corp. v. Dellaventura, 544 F.2d 35, 49 (2d Cir. 1976), aff’d sub nom. Ne. Marine Terminal Co. v. Caputo, 432 U.S. 249 (1977). Judge Friendly recognized that, as of 1976, “there are two lines of Supreme Court decisions on this subject which are analytically in conflict, with the result that a court of appeals must choose the one it deems more appropriate for the case at hand.”¹⁸⁷Id. He then listed “[l]eading cases supporting the view that great deference must be given to the decisions of an administrative agency applying a statute to the facts and that such decisions can be reversed only if without rational basis.”¹⁸⁸Id. He contrasted these cases with “an impressive body of law sanctioning free substitution of judicial for administrative judgment when the question involves the meaning of a statutory term.”¹⁸⁹Id. (first citing Off. Emps. Int’l Union, Local No. 11 v. NLRB, 353 U.S. 313, 318–20 (1957); and then citing Davies Warehouse Co. v. Bowles, 321 U.S. 144, 150 (1944)); see id. (noting that Morton v. Ruiz, 415 U.S. 199, 237 (1974), observed that deference was warranted where the agency interpretation is “‘consistent with the congressional purpose,’” and contending that “this very nearly eliminates the ‘deference’ principle as regards statutory construction altogether since if the agency’s determination is found by a court to be consistent with the congressional purpose, it presumably would be affirmed on that ground without any need for deference”). As Judge Friendly’s opinion indicates, the immediate pre-Chevron caselaw was not entirely consistent on the question of judicial deference. Courts sometimes deferred to agency legal determinations and sometimes did not. When courts deferred, they identified several factors for doing so.¹⁹⁰For example, an article by Professor Colin Diver remarked that “whether to grant deference [to an agency’s legal interpretation] depends on various attributes of the agency’s legal authority and functions and of the administrative interpretation at issue.” Colin S. Diver, Statutory Interpretation in the Administrative State, 133 U. Pa. L. Rev. 549, 562 (1985). The article identified ten factors as relevant to this inquiry: contemporaneousness, long-standing duration, consistency, reliance, importance of the issue, complexity, presence of rulemaking authority, the need for agency action to implement the statute, congressional ratification, and the quality of agency explanation. See id. at 562 n.95; see also Pittston Stevedoring, 544 F.2d at 49–50 (denying deference based on such factors).

In Chevron, the Court appeared to introduce a simplified two-step process for deferring to an agency’s statutory construction. Chevron said that the first step was “whether Congress has directly spoken to the precise question at issue.”¹⁹¹Chevron U.S.A. Inc. v. Nat. Res. Def. Council, Inc., 467 U.S. 837, 842 (1984). Where Congress’s intent is clear, the court must give that intent effect.¹⁹²Id. A court should ascertain that congressional intent “employing traditional tools of statutory construction.”¹⁹³Id. at 843 n.9. But Congress could “explicitly [leave] a gap for the agency to fill” either through an express or implied “delegation of authority.”¹⁹⁴Id. at 843–44. If “Congress has not directly addressed the precise question at issue,” Chevron instructed that “the court does not simply impose its own construction on the statute, as would be necessary in the absence of an administrative interpretation.”¹⁹⁵Id. at 843 (footnote omitted). Rather, the second step of Chevron asked “whether the agency’s answer is based on a permissible construction of the statute.”¹⁹⁶Id. In this fashion, Chevron indicated that “[t]he court need not conclude that the agency construction was . . . even the reading the court would have reached if the question initially had arisen in a judicial proceeding.”¹⁹⁷Chevron, 467 U.S. at 843 n.11. Thus, on the face of the Chevron opinion, there was a tension between the approach suggested in footnote 9 and the one suggested in footnote 11. In footnote 9, Chevron indicated that a court would approach interpretation “employing traditional tools of statutory construction.” Id. at 843 n.9. But in footnote 11, the Court suggested that a reviewing court might abandon “the reading the court would have reached if the question initially had arisen in a judicial proceeding” in the face of an agency construction. Id. at 843 n.11.; cf. Kisor v. Wilkie, 139 S. Ct. 2400, 2415 (2019) (addressing the meaning of ambiguity in the closely related context of an agency’s interpretation of its own regulations and reasoning that “sometimes the law runs out, and policy-laden choice is what is left over” and that “if the law gives an answer . . . then a court has no business deferring to any other reading, no matter how much the agency insists it would make more sense”). For a discussion of Kisor, see Aditya Bamzai, Delegation and Interpretive Discretion: Gundy, Kisor, and the Formation and Future of Administrative Law, 133 Harv. L. Rev. 164, 186–98 (2019).

Although parts of Chevron suggest an attempt to establish a simplified two-step process to deference, it is not clear that the Chevron Court itself intended to change the underlying multifactor approach.¹⁹⁸See Thomas W. Merrill, The Story of Chevron: The Making of an Accidental Landmark, 66 Admin. L. Rev. 253, 275–76 (2014). Indeed, a few years later, the Court appeared to step back from a broad reading of Chevron, describing the issue before it as “a pure question of statutory construction for the courts to decide.”¹⁹⁹INS v. Cardoza-Fonseca, 480 U.S. 421, 446–48 (1987). And a decade and a half later, in United States v. Mead Corp.,²⁰⁰533 U.S. 218 (2001). the Court qualified Chevron by holding that the measure of deference that a court gives to an agency interpretation depends in part on the formality of the agency’s procedures.²⁰¹See id. at 230–31. In doing so, the Court reasoned that “[t]he fair measure of deference to an agency administering its own statute has been understood to vary with circumstances, and courts have looked to the degree of the agency’s care, its consistency, formality, and relative expertness, and to the persuasiveness of the agency’s position.”²⁰²Id. at 228 (footnotes omitted). Mead explained that this approach “has produced a spectrum of judicial responses, from great respect at one end, to near indifference at the other.”²⁰³Id. (citations omitted). Describing the Court’s precedents, Mead contended that the Court had “tailor[ed] deference to variety,” with a recognition of “more than one variety of judicial deference, just as the Court has recognized a variety of indicators that Congress would expect Chevron deference.”²⁰⁴Id. at 236–37. Mead’s approach thus eschewed the simplicity of a two-step process.²⁰⁵See, e.g., id. at 239 (Scalia, J., dissenting) (characterizing Mead as “an avulsive change” and claiming that “[w]hat was previously a general presumption of authority in agencies to resolve ambiguity in the statutes they have been authorized to enforce has been changed to a presumption of no such authority”). Indeed, throughout these developments, some Justices adhered to a multifactor understanding of Chevron.²⁰⁶See, e.g., City of Arlington v. FCC, 569 U.S. 290, 308–09 (2013) (Breyer, J., concurring in part and concurring in the judgment) (“[T]he existence of statutory ambiguity is sometimes not enough to warrant the conclusion that Congress has left a deference-warranting gap for the agency to fill because our cases make clear that other, sometimes context-specific, factors will on occasion prove relevant.”).

Some later cases relied heavily on Chevron’s footnote instructing courts to defer even when the court would not have adopted the same construction as the agency “if the question initially had arisen in a judicial proceeding.”²⁰⁷Chevron U.S.A. Inc. v. Nat. Res. Def. Council, Inc., 467 U.S. 837, 843 n.11 (1984). The most prominent such case was National Cable & Telecommunications Ass’n v. Brand X Internet Services,²⁰⁸545 U.S. 967 (2005). where the Court reasoned that, “[i]f a statute is ambiguous, and if the implementing agency’s construction is reasonable, Chevron requires a federal court to accept the agency’s construction of the statute, even if the agency’s reading differs from what the court believes is the best statutory interpretation.”²⁰⁹Id. at 980 (citing Chevron, 467 U.S. at 843–44, 843 n.11); see id. at 982 (“A court’s prior judicial construction of a statute trumps an agency construction otherwise entitled to Chevron deference only if the prior court decision holds that its construction follows from the unambiguous terms of the statute and thus leaves no room for agency discretion.”); see id. at 1016–17 (Scalia, J., dissenting) (agreeing that the logical consequence of the Court’s Brand X opinion was that an agency could reject the best interpretation of a statute); Christensen v. Harris Cnty., 529 U.S. 576, 589 (2000) (Souter, J., concurring) (similar); see also Brett M. Kavanaugh, Fixing Statutory Interpretation, 129 Harv. L. Rev. 2118, 2153 (2016) (reviewing Robert A. Katzmann, Judging Statutes (2014)) (noting that under Chevron a judge might “uphold the agency’s interpretation even though it is not the best interpretation”). Brand X thus contemplates that an agency’s permissible construction can displace the best construction of a statute under some circumstances. In the views of the Brand X Court, “[t]his principle follows from Chevron itself.”²¹⁰Brand X, 545 U.S. at 982; see id. at 983 (“Chevron teaches that a court’s opinion as to the best reading of an ambiguous statute an agency is charged with administering is not authoritative . . . .”). Several lower courts have embraced and reiterated Brand X’s understanding of ambiguity under Chevron. Consider the following examples from the Courts of Appeals for the Fifth, Ninth, and D.C. Circuits—three courts of appeals that adjudicate a significant portion of the Nation’s administrative law docket. In Acosta v. Hensel Phelps Construction Co., 909 F.3d 723 (5th Cir. 2018), the Fifth Circuit reasoned that “Chevron requires a federal court to accept the agency’s construction of the statute, even if the agency’s reading differs from what the court believes is the best statutory interpretation.” Id. at 730 (quoting Elgin Nursing & Rehab. Ctr. v. HHS, 718 F.3d 488, 492 n.3 (5th Cir. 2013)). In Henriquez-Rivas v. Holder, 707 F.3d 1081 (9th Cir. 2013), the Ninth Circuit reasoned that “[i]f the [agency’s] construction is reasonable, we must accept that construction under Chevron, even if we believe the agency’s reading is not the best statutory interpretation.” Id. at 1087. And in American Council on Education v. FCC, 451 F.3d 226 (D.C. Cir. 2006), the D.C. Circuit reasoned that it could not “set aside the Commission’s reasonable interpretation of the Act in favor of an alternatively plausible (or an even better) one.” Id. at 234; see also Citizens Coal Council v. Norton, 330 F.3d 478, 482 (D.C. Cir. 2003) (“Even assuming the correctness of [an alternative interpretation], the ambiguity of the statute in combination with the Chevron doctrine eclipses the ability of the courts to substitute their preferred interpretation for an agency’s reasonable interpretation . . . .”). At the same time, it appears that other judges do not treat ambiguity under Chevron in this fashion. For example, Judge Kethledge has remarked that he has “personally . . . never had occasion to reach Chevron’s step two in any of [his] cases.” Raymond M. Kethledge, Ambiguities and Agency Cases: Reflections After (Almost) Ten Years on the Bench, 70 Vand. L. Rev. En Banc 315, 323 (2017); cf. Antonin Scalia, Judicial Deference to Administrative Interpretation of Law, 1989 Duke L.J. 511, 521 (1989) (discussing the relationship between statutory interpretation and the triggering requirements for Chevron deference). That remark suggests a robust perspective on how to “employ[] traditional tools of statutory construction” that leaves little room for permissible alternatives. Chevron, 467 U.S. at 843 n.9.

Leaving their precise holdings to one side, these cases show how the “Chevron doctrine” ebbed and flowed in the decades following the Court’s decision. Rather than one consistent approach, the Court adopted several different methods for parceling out deference to agency legal interpretations. Over the years—both before and after the Chevron decision—the standard of review for legal questions varied. In this sense, it is hard to point to a single “Chevron doctrine,” rather than shifting approaches to parceling out deference changing over time.

B.      The APA Revisited

Returning to section 706’s text, the most plausible interpretation is that, much like with its statutory repudiation of Dobson, Congress sought to establish the traditional scope of review for legal questions when it enacted the APA.²¹¹Cf. Davis, supra note 79, § 244, at 868 n.1 (“Before its statutory abolition in 1948, the opinion was widespread that whether or when the Dobson doctrine would be applied was utterly unpredictable. What is not generally recognized is that outside the tax field the use or non-use of what is essentially the Dobson doctrine is equally difficult to predict.”). In doing so, Congress sought to repudiate then-recent innovations regarding the standard of review. Thus, where statutory text has a “best” interpretation, section 706 requires a court to give it that interpretation, subject to canons of construction that require courts to give weight to agency interpretations that are contemporaneous or customary. That was the traditional approach before the APA.²¹²See Burnet v. Chi. Portrait Co., 285 U.S. 1, 16 (1932) (reasoning that “great weight” will be given to an agency’s construction of a statute as long as the interpretation is consistent); see also Norwegian Nitrogen Prods. Co. v. United States, 288 U.S. 294, 315 (1933) (discussing the “weight” that an agency interpretation is given when it employs “contemporaneous construction of a statute”). And although certain cases in the 1940s seemed to depart from the approach’s logic,²¹³As did the Attorney General’s Final Report, which operated on a vision of the law-fact distinction that, like Gray, blurred the line between the two categories. See supra notes 90–92 and accompanying text. But the Final Report’s discussion on this point is not compelling evidence of the meaning of section 706. First, if the APA sought to recrystallize a distinction between law and fact, then the Final Report suffered from the same flaws as precedents like Gray, Dobson, and Hearst. Second, recall that the majority of the Attorney General’s Committee did not believe that Congress should enact a standard of review at all. Final Report, supra note 90, at 209. But Congress did enact such a standard, borrowing a framework from a proposal by the minority of the Attorney General’s Committee. See Shepherd, supra note 6, at 1649. the APA is best understood to repudiate those later cases and embrace the traditional framework. Read as a whole, section 706 reestablishes a form of de novo review, tempered by the customary and contemporaneous approaches to interpretation.

That understanding of the APA’s structure creates a method by which legal text can settle over time, but policy determinations need not. Under that approach, courts must follow a statute’s best interpretation in the face of an alternative permissible interpretation embraced by an agency, consistent with the part of Chevron requiring that courts employ the traditional tools of construction.²¹⁴As Justice Kavanaugh explained in Kisor, “[i]f a reviewing court employs all of the traditional tools of construction”—as Chevron’s footnote 9 directs—then “the court will almost always reach a conclusion about the best interpretation of the [legal text] at issue.” Kisor v. Wilkie, 139 S. Ct. 2400, 2448 (2019) (Kavanaugh, J., concurring in the judgment) (citing Chevron, 467 U.S. at 843 n.9). At that point, a court “will have no need to adopt or defer to an agency’s contrary interpretation.” Id. At the same time, “some cases involve [legal text] that employ broad and open-ended terms like ‘reasonable,’ ‘appropriate,’ ‘feasible,’ or ‘practicable.’”²¹⁵Id. In general, absent contrary indications in the organic statute, “[t]hose kinds of terms afford agencies broad policy discretion,”²¹⁶Id. at 2448–49. subject to the APA’s arbitrary-and-capricious standard.²¹⁷See 5 U.S.C. § 706(2)(A); Motor Vehicle Mfrs. Ass’n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 52 (1983). Put slightly differently, the two approaches to Chevron point in different directions in those cases where a statutory provision is amenable to a “best” interpretation using all the ordinary tools of statutory construction, but there nevertheless exist “permissible” alternative interpretations. Under Justice Kavanaugh’s concurrence in Kisor, such a statute would be given its “best” interpretation. See 139 S. Ct. at 2448. If there were no “best” interpretation, the agency decision would be subject to arbitrary-and-capricious review under State Farm. See id. at 2449. Under the alternative approach, the agency’s “permissible” alternative interpretation would govern. Under these circumstances, it would make sense to say that a legal text does not speak to an issue, which would imply that any questions left unresolved by the text were questions of policy, not legal interpretation.

Conclusion

This Article has explored the interpretive backdrop against which Congress enacted the APA. To be sure, that backdrop is complex, with several strands of precedents—regarding contemporaneous and customary interpretation, the law-fact distinction, and the mandamus standard—leading to the APA’s passage in 1946. But complexity alone does not prevent the modern interpreter from arriving at sound conclusions about section 706’s best interpretation. Against the relevant backdrop, section 706 is best understood to require a form of de novo review, tempered by respect for agency interpretations that are contemporaneous and customary. At the same time, the “arbitrary and capricious” standard leaves room for agency flexibility where policy judgment, not statutory interpretation, is at issue.