Severability First Principles

University of Chicago Law School
Chicago Unbound
Articles Scholarship
2023
Severability First Principles

William Baude
Follow this and additional works at: https://chicagounbound.uchicago.edu/journal_articles
Part of the Law Commons

COPYRIGHT © 2023 VIRGINIA LAW REVIEW ASSOCIATION
VIRGINIA LAW REVIEW

VOLUME 109 MARCH 2023 NUMBER 1
ARTICLES
SEVERABILITY FIRST PRINCIPLES
William Baude*
The United States Supreme Court has decided a number of cases
involving severability in the last decade, from NFIB v. Sebelius and
Murphy v. NCAA to Seila Law v. CFPB, Barr v. AAPC, United States
v. Arthrex, California v. Texas, and Collins v. Yellen. The analysis has
not been consistent, the Justices have not been able to agree, and the
results have not been intuitive. Some of the Justices have proposed a
revisionist approach, but they too have been unable to agree on what it
requires.
This Article proposes a return to first principles. Severability is a
question of what the law is. Severability also includes two principles of
constitutional law: that judges should enforce the law, and that the
Constitution displaces ordinary law that is repugnant to it. And it also
includes principles of non-constitutional law: that validly enacted
statutes are law if they are not repugnant to the Constitution, that
* Professor of Law, Faculty Director of the Constitutional Law Institute, University of

Chicago Law School. This Article benefitted from preliminary workshops at the University of
Michigan Law School and University of Wisconsin Law School, and the subsequent draft
benefitted from further comments by Josh Blackman, Curt Bradley, Jud Campbell, Tara
Grove, John Harrison, Robert Leider, Mike Rappaport, Richard Re, Steve Sachs, Peter Salib,
Mark Shawhan, Kevin Walsh, Lael Weinberger, Ilan Wurman, Sam Bray, Kian Hudson, Brian
Lea, and other participants in workshops at the University of Chicago Law School and DePaul
College of Law, and the Originalism Works in Progress Conference at the University of San
Diego. Simon Carvalho and Sarah Leitner provided excellent research assistance. Any defects
in this Article should be severed.
1
2 Virginia Law Review [Vol. 109:1
unenacted hopes and dreams are not, and that Congress may legislate
for contingencies.
Much of the time, these principles lead to a simple bottom line:
effectively complete severability, rebutted only by an inseverability
clause or something else with the force of law. There are also harder
cases where the bottom line is not so simple, but where the first
principles of severability will nonetheless lead the way—the relevance
of unconstitutional removal restrictions, the non-constitutional law that
resolves unconstitutional combinations, and the relevance of
severability to standing and other procedural questions.
INTRODUCTION .................................................................................. 3
I. PRINCIPLES ..................................................................................... 5
A. Severability Is a Question of What the Law Is ...................... 5
B. Disregarding Repugnant Law ............................................... 8
C. Applying Fallback Law ....................................................... 14
II. IMPLICATIONS ............................................................................. 17
A. Severability Doctrine .......................................................... 17
1. The Presumption of Severability ................................... 17
2. Severability Clauses ...................................................... 18
3. Fully Operative Law and Counterfactual Intent ........... 22
4. State Law ....................................................................... 24
B. Justices Gorsuch and Thomas ............................................. 26
C. Other Doctrines .................................................................. 32
1. Facial Challenges.......................................................... 32
2. National Injunctions ...................................................... 34
3. Contract Law ................................................................. 35
III. DIFFICULTIES ............................................................................. 37
A. Ultra Vires ........................................................................... 37
B. Combinations ...................................................................... 41
1. Fallback Law Solutions ................................................. 44
a. Hypothetical congressional intent ......................... 45
b. Special-purpose canons ......................................... 46
c. General-purpose canons ........................................ 48
d. Final thought ......................................................... 50
2. Non-Fallback Approaches ............................................. 50
a. . . . that do not work ............................................... 51
i. Standing bootstrapping .................................... 51
ii. Avoiding “editorial freedom” ......................... 52
2023] Severability First Principles 3
b. . . . that work better ................................................ 53

i. Stipulation ........................................................ 53
ii. Factual particularity ....................................... 54
C. Severability Procedures ...................................................... 56
CONCLUSION ................................................................................... 59
INTRODUCTION
When part of a statute is unconstitutional, the courts engage in
severability analysis. According to the cases, this analysis couples a
presumption with a possible rebuttal. The presumption is one of
severability: “[T]he invalid part may be dropped.”1 The presumption is
rebutted based on either an objective analysis, asking whether “what is
left is fully operative as a law,”2 or a subjective analysis, asking whether
“it is evident that the Legislature would not have enacted those provisions
which are within its power, independently of that which is not.”3 Slightly
more controversially, the same seems to be true for a single provision with
constitutional and unconstitutional applications.4
There have been many calls to abandon or reform severability
doctrine.5 But there is no consensus about what the problem is or what to
do instead. At least one problem, though, is methodological: the modern
approach to statutory interpretation is heavily influenced by formalism
generally and textualism specifically. Such judges have extra reason to be
skeptical of current doctrine. They doubt the coherence or the relevance
of counterfactual inquiries into legislative intent and also tend to resist the
normative analysis that sometimes lies behind particular severability
arguments. And severability can look uncomfortably like “rewriting” a
statute, which most judges today know they are not supposed to get caught
doing. So, we need an account of severability that makes formal sense.
1
Alaska Airlines, Inc. v. Brock, 480 U.S. 678, 684 (1987) (quoting Buckley v. Valeo, 424
U.S. 1, 108 (1976) (per curiam)).
2
Id. (quoting Buckley, 424 U.S. at 108).
3
Id.
4
Ayotte v. Planned Parenthood of N. New Eng., 546 U.S. 320, 328–29 (2006).
5
See, e.g., Murphy v. NCAA, 138 S. Ct. 1461, 1487 (2018) (Thomas, J., concurring) (“[O]ur
modern severability precedents are in tension with longstanding limits on the judicial
power.”); Kevin C. Walsh, Partial Unconstitutionality, 85 N.Y.U. L. Rev. 738, 742 (2010)
(calling for a “displacement-based approach”); Tom Campbell, Severability of Statutes, 62
Hastings L.J. 1495, 1497 (2011) (proposing a “complete abolition of the severability
doctrine”); Lisa Marshall Manheim, Beyond Severability, 101 Iowa L. Rev. 1833, 1838 (2016)
(advocating for the replacement of severability with a broader inquiry into legislative intent).
This is a natural occasion for a return to first principles, and some have
tried. Several recent articles make promising contributions,6 and recent
opinions by Justices Thomas and Gorsuch have attempted to synthesize
them into a new revisionist account of severability.7 But their work is
incomplete. Justices Thomas and Gorsuch cannot even agree among
themselves in several recent cases, and throughout they may be trying to
squeeze more certainty out of the literature than it can supply. We still
need a clearer account of the first principles that answer the severability
problem and of what those principles do and do not imply.
Returning to first principles also requires us to determine whether
severability analysis comes from the Constitution or instead from
statutory interpretation or other non-constitutional law. In truth, it is both.
Severability principles are a combination of both constitutional and non-
constitutional law. The Constitution tells us that it displaces ordinary law
that is inconsistent with it. It also tells us that judges (among others) are
supposed to apply the law. But these constitutional principles are not all
there is to severability. We also need to know what is the law, when some
part of a statute has been found to be constitutionally repugnant? Ordinary
principles of statutory interpretation fill in this answer. Federal law is
what has been enacted by Congress and not otherwise displaced,
including any fallback law. And, of course, any non-federal legal rules
also continue to apply.
Much of the time, these principles lead to a simple bottom line: judges
should enforce a statute except in the specific cases where its application
is unconstitutional. But this simplicity is deceptive. The bottom line
becomes more difficult to see in the case of unconstitutional
combinations: when two statutory requirements are unconstitutional if
taken together, which one should be disregarded? These difficult cases—
more widespread than many realize—illuminate an aspect of the
Constitution that has been there all along: the Constitution tells us what
the law isn’t, but not always what it is. Solving the severability problem
in these cases—saying what the law is—requires going beyond the text
of the statute, whether formalist judges like it or not.
Other difficulties come up in the context of standing and other
threshold questions. When can a plaintiff establish standing on the basis
6
Especially noteworthy are Walsh, supra note 5; John Harrison, Severability, Remedies,
and Constitutional Adjudication, 83 Geo. Wash. L. Rev. 56 (2014); and Jonathan F. Mitchell,
The Writ-of-Erasure Fallacy, 104 Va. L. Rev. 933 (2018).
7
See infra Section II.B.
of an inseverability argument, and when can a severability argument

defeat standing? These questions have proven difficult for the courts, but
this time it is the difficulty that is deceptive. Once we straighten out our
severability analysis, it drives us to straightforward answers in these
cases.
This Article puts forward the first principles of severability and then
applies them, first to the easy cases and then to the hard ones. Part I argues
that severability is a question of law; that the Constitution displaces
repugnant law; and that all non-repugnant law should be enforced,
including fallback law such as severability and inseverability clauses. Part
II describes how these principles would reframe severability doctrine,
how Justices Gorsuch and Thomas have come close to restating these
principles, and how the principles also clarify facial challenges and
national injunctions. Part III tackles the harder cases, such as
unconstitutional combinations and severability procedure.
I. PRINCIPLES
Fundamentally, severability is a question of law. What is the combined
legal effect of the Constitution and one or more statutory provisions when
there is a conflict between them? It is partly a question of constitutional
law—the Constitution tells us what the law cannot be. And it is partly a
question of statutory or sub-constitutional law—these materials fill out
what the law is.
Throughout, I will elaborate using some broadly formalist premises:
The text is the part of a statute that is law, not its purposes or policies. The
Constitution is also law—indeed, supreme law—and so it controls over a
contradictory statute. And the job of judges is to apply these laws, but not
to change them or to make law of their own. But you don’t have to share
exactly these premises. The same general principles can accommodate
some different approaches to interpretation and adjudication, as I will
discuss on occasion.
A. Severability Is a Question of What the Law Is

Courts and executive officials must enforce the law as a matter of
course. The “severability” question arises when there are conflicting legal
commands. It is common ground that higher law, like the Constitution,
prevails over more ordinary law, like a federal statute. That is, we know
what the law is not—not the unconstitutional thing. The severability
question tries to answer what the law is—what is the law, in light of what
the law is not?8
This means that the question of severability is really a question of legal
interpretation and of the conflict of laws. The judge takes two legal
provisions, a statute and the Constitution, and asks what the combined
legal effect of the two documents is. If a statute is constitutional, the
answer is usually easy. The statute and Constitution are the law. If part of
the statute is unconstitutional, the question is what the law is. The judge’s
answer is supposed to reflect the content of these two legal provisions and
any rules for resolving their conflict.
But many approaches to severability instead start from a different kind
of law—the law of remedies. In our era of judicial supremacy, people
sometimes think of the problem this way: a plaintiff identifies a
constitutional violation, and the court reacts to fix the constitutional
violation by “severing” the bad part of the statute, like slicing a blemish
off a piece of fruit. If this were right, one might incorporate many
principles of the law of remedies into the law of severability.9
But it is not right. First of all, it is inconsistent with basic principles of
legalism and the separation of powers.10 Judges do not actually “strike
down” statutes, and they do not issue “writs of erasure.”11 The invalidity
of an unconstitutional statute is caused by the Constitution; it is not caused
by a judge. Even if the judge’s decision is what causes many people to be
aware of, or to pay heed to, the provision’s invalidity, that invalidity
precedes the decision rather than following it. That must be so, since the
power of judicial review is premised on a judge’s ability to discern and
apply existing legal norms, not on a power to craft them.12
8
I owe this way of thinking about the problem, which recurs throughout, and many other
insights to conversation with John Harrison.
9
See, e.g., Eric S. Fish, Choosing Constitutional Remedies, 63 UCLA L. Rev. 322, 330–33
(2016); Evan H. Caminker, Note, A Norm-Based Remedial Model for Underinclusive
Statutes, 95 Yale L.J. 1185, 1186 n.3 (1986); David H. Gans, Severability as Judicial
Lawmaking, 76 Geo. Wash. L. Rev. 639, 643 (2008); Planned Parenthood of N. New Eng.,
546 U.S. at 328–30; see also Richard H. Fallon, Jr., Facial Challenges, Saving Constructions,
and Statutory Severability, 99 Tex. L. Rev. 215, 257–58 (2020) (“The Supreme Court has
repeatedly used this terminology, as have commentators. But referring to severance as a
remedy invites confusion.” (footnote omitted)); Brian Charles Lea, Situational Severability,
103 Va. L. Rev. 735, 755 n.116 (2017) (“Many scholars, understandably following the Court’s
lead, couch their discussions of severability in terms of remedial discretion . . . .”).
10
See, e.g., Massachusetts v. Mellon, 262 U.S. 447, 488 (1923); Danforth v. Minnesota,
552 U.S. 264, 271 (2008).
11
Mitchell, supra note 6, at 935–36.
12
See Marbury v. Madison, 5 U.S. (1 Cranch) 137, 177–78 (1803).
Second, this orthodox picture is confirmed by the reality of judicial

rulings. Though courts may now label severability as a remedy, even
today, severability doctrine does not actually operate as if judicial
decisions can invalidate and repair unconstitutional statutes. For instance,
when a lower court decision rules that a statute is unconstitutional, that
ruling is not treated as a writ of erasure.13 This is why some lower courts
feel the need to issue nationwide injunctions to expand the effect of their
constitutional decisions, and why people bother to debate the propriety of
such nationwide injunctions14—the injunction, the actual remedy, is
adding something not inherently present in a ruling of unconstitutionality.
The severability question is present in the ruling, not just the remedy.
To be sure, severability questions often arise when a court is deciding
remedial questions. If a court agrees that a plaintiff is entitled to relief on
a constitutional claim, it may need to understand what other legal
provisions are in force before crafting an injunction or the like. Similarly,
to decide standing the court may also need to understand how the
constitutionality of one provision relates to the enforceability of another.
But this does not make severability a question of remedies or a question
of standing. It is a question of legal interpretation, and legal interpretation
is often relevant to standing, remedies, and other parts of a case.15
One final point: the argument that severability is a remedy often
assumes that if severability is a remedy, that will make its application
tailored, discretionary, and generally equitable. But it is not at all clear
that this would follow. Not all remedies are tailored, discretionary, or
equitable.16 There is no writ of erasure, but if there were, why do we
assume it would be an equitable writ instead of a remedy at law? So,
labeling severability as a remedy misunderstands severability, but it
probably also misunderstands remedies.
13
See Harrison, supra note 6, at 97–100 for a great example.
14
See, e.g., Samuel L. Bray, Multiple Chancellors: Reforming the National Injunction, 131
Harv. L. Rev. 417, 418 (2017); Amanda Frost, In Defense of Nationwide Injunctions, 93
N.Y.U. L. Rev. 1065, 1090 (2018); Mila Sohoni, The Lost History of the “Universal”
Injunction, 133 Harv. L. Rev. 920, 921 (2020).
15
See Lea, supra note 9, at 756–57, 756 n.128.
16
See Samuel L. Bray, The System of Equitable Remedies, 63 UCLA L. Rev. 530, 544–50
(2016).
B. Disregarding Repugnant Law

Severability problems begin with validity problems. Ordinarily, we
know what the law is because we take the statute’s word for it. But when
a statute conflicts with the Constitution, the statute cannot be taken at face
value. This is nothing new to anybody who knows about judicial review,
although, as noted above, it is a problem confronted outside the courts,
too.
These basic lessons have implications for severability. The first step in
thinking about severability is understanding this basic rule about what the
law is not: an unconstitutional law simply doesn’t govern in any instance
where it is unconstitutional. It brings with it a related principle about what
the law is: laws generally remain valid and enforceable in the absence of
such a constitutional problem.
In the language of modern severability doctrine, one might say that this
is effectively a very strong rule of severability: every application of a
statute is by default severable from every other application of the statute.
But the point is more fundamental than the doctrine and more
fundamental than the terminology. For instance, Kevin Walsh uses the
terminology of “displacement” and “repugnancy” in his foundational
work on partial unconstitutionality.17 The Constitution displaces any
other legal rules that are repugnant to it—bumping them out of the way
to the extent, but only to the extent, that they are at odds.18
This rule also fits the way that constitutional officers actually apply the
Constitution. Judges were not given the general law-vetoing power of a
council of revision.19 They were given judicial review only as an incident
of their principal power and duty to decide particular cases according to
law. Hence, Chief Justice Marshall explained judicial review by arguing
that those who are sworn to uphold the Constitution must apply it to their
own conduct.20 In any particular instance where judges are called upon to
17
Walsh, supra note 5, at 755. The terms have not always been well-understood. Cf.
Theodore F.T. Plucknett, Bonham’s Case and Judicial Review, 40 Harv. L. Rev. 30, 34 n.17
(1927) (“The meaning of repugnant is not clear; it would almost seem that it meant no more
than distasteful to the court . . . .”) (wrong).
18
For a great formulation, see Walsh, supra note 5, at 765 n.124 (“In the case of . . . a
conflict [with federal law], the Supremacy Clause dictates that ‘to the extent of such collision
and repugnancy, the law of the State must yield, and to that extent and no further, it is rendered
by such repugnancy, inoperative and void.’ ” (quoting Commonwealth v. Kimball, 41 Mass.
(24 Pick.) 359, 361 (1837))).
19
20
Marbury v. Madison, 5 U.S. (1 Cranch) 137, 179–80 (1803).
act, they must examine the Constitution and refuse to do anything

contrary to it.21 These duties are what yield judicial review. So judicial
review, and hence judicial determinations of invalidity, are implemented
in a case-by-case fashion. (And the same goes for executive review under
departmentalist principles, which is similarly incidental to the executive’s
duty to uphold the law in every act it takes.)22
These first principles are also consistent with the practice of
constitutional interpretation in and after Marbury v. Madison. In
Marbury, of course, the Court found that part of Section 13 of the
Judiciary Act provided the Court with original jurisdiction in cases where
the Constitution forbade it. So, the Court refused original jurisdiction in
such cases. It disregarded the Judiciary Act to the extent it was repugnant
to the Constitution. But it continued to apply Section 13 of the Judiciary
Act, and certainly the Act more generally, in cases where the Constitution
did not intercede.23
Cases throughout the first sixty-some years of the Republic reflected
the same understanding.24 One can find occasional explicit statements
about this, for instance from Justice Trimble’s statement in Ogden v.
Saunders:
It is not the terms of the law, but its effect, that is inhibited by the
constitution. A law may be in part constitutional, and in part
unconstitutional. It may, when applied to a given case, produce an effect
which is prohibited by the constitution, but it may not, when applied to
a case differently circumstanced, produce such prohibited effect.25
Let me now provide a more systematic review.26 Most Supreme Court
cases say nothing explicit about severability, but upon closer
investigation, however, they are all consistent with the classic approach
of repugnancy and displacement. There are twelve identified Supreme
21
Id. at 178–79.
22
On executive review and departmentalism, see Frank H. Easterbrook, Presidential
Review, 40 Case W. Rsrv. L. Rev. 905, 905–06 (1990); Michael Stokes Paulsen, The Most
Dangerous Branch: Executive Power to Say What the Law Is, 83 Geo. L.J. 217, 219–20
(1994).
23
24
John Copeland Nagle, Severability, 72 N.C. L. Rev. 203, 212–13 (1993); Walsh, supra
note 5, at 758–61.
25
Ogden v. Saunders, 25 U.S. (12 Wheat.) 213, 316 (1827) (opinion of Trimble, J.).
26
Cf. William Baude, Adam S. Chilton & Anup Malani, Making Doctrinal Work More
Rigorous: Lessons from Systematic Reviews, 84 U. Chi. L. Rev. 37, 51–54 (2017) (describing
systematic reviews of legal doctrine).
Court opinions recognizing constitutional limits on federal statutes before

1850.27 All of them are consistent with the repugnancy framework. We
have already discussed Marbury. Beyond that:
Mossman v. Higginson narrowed the jurisdictional grant in the
Judiciary Act of 1789, which gave jurisdiction “where an alien is a
party.”28 The Court concluded this must be confined to the limits of the
Constitution, which extended federal jurisdiction only between one alien
and one citizen, not two aliens.29 It did not, of course, eliminate the
jurisdictional grant entirely as inseverable. Hodgson v. Bowerbank30 and
Jackson v. Twentyman31 did the same.
United States v. Cantril rejected as “repugnant” a federal bank fraud
indictment emanating from a badly worded federal statute.32 It is
ambiguous whether this was really a case of unconstitutionality, though
Whittington classifies it as one, and either way, the reporter noted that the
law had already been amended by Congress.33 There was no occasion for
the Court to use inseverability for the same task, and indeed, the point of
error described the statute as being unconstitutional only “so much thereof
as relates to the charge set forth in the indictment.”34
Reynolds v. M’Arthur refused to give retroactive effect to a law dealing
with a military land grant in the new state of Ohio.35 But that refusal did
not stop the law from having prospective effect, assuming Congress had
the power to pass the law in the first place (a question the court
27
The cases are identified in the future-canonical Keith E. Whittington, Repugnant Laws:
Judicial Review of Acts of Congress from the Founding to the Present (2019) [hereinafter
Whittington, Repugnant Laws], and the database available at Keith E. Whittington, Judicial
Review of Congress Database, Princeton (Aug. 14, 2022), https://scholar.princeton.
edu/kewhitt/publications/judical-review-congress-database [https://perma.cc/2J6Y-8WKK]
(filter spreadsheet). I pick 1850 because it roughly marks the general origins of inseverability
doctrine, see Eason v. State, 11 Ark. 481, 501–03 (1851); Washington v. State, 13 Ark. 752,
763–64 (1853); Warren v. Mayor of Charlestown, 68 Mass. (2 Gray) 84, 99–100 (1854)
(Shaw, C.J.); Walsh supra note 5, at 769 n.145, though the Supreme Court’s first inseverability
decision was not until United States v. Reese, 92 U.S. 214 (1875).
28
4 U.S. (4 Dall.) 12, 14 (1800) (quoting Judiciary Act of 1789, ch. 20, § 11, 1 Stat. 73, 78).
29
Id.
30
9 U.S. (5 Cranch) 303, 304 (1809).
31
27 U.S. (2 Pet.) 136, 136 (1829).
32
8 U.S. (4 Cranch) 167, 168 (1807). The two points of error were that the indictment was
inconsistent with the statute and that the statute was inconsistent with the Constitution. Id. at
167–68.
33
Id. at 167, 168 n.*.
34
Id. at 168.
35
27 U.S. (2 Pet.) 417, 434–35 (1829).
reserved).36 Similarly, United States v. Percheman refused to read federal

laws as abrogating previously vested property rights in the acquired
territory of Florida, with no hint that the laws were otherwise invalid.37
Parson v. Bedford concluded that a law about federal procedure in the
civil law state of Louisiana had to be limited by the principles of the
Seventh Amendment, which insulated jury findings from appellate
review.38 The law otherwise remained operative.39
United States v. Phelps (Ex Parte United States)40 concluded that a law
requiring a government collection suit to be tried promptly41 could not
stop the courts from granting additional continuances to obtain evidence
from overseas—otherwise, according to Justice McLean, “It would be
depriving the party of his right to a trial by jury.”42
New Orleans v. United States concluded that the federal government
could not retain jurisdiction over a quay in New Orleans after Louisiana’s
statehood.43 This decision was “echoed”44 by a more notable decision a
decade later in Pollard’s Lessee v. Hagan, which recognized limits on
Congress’s power to grant land post-statehood.45 In both cases, the rest of
the cessions and reservations took effect as proscribed.
With Marbury, that makes eleven. None of these eleven cases even
paused over the question of the validity of other parts of the laws—the
grant and cession laws, the procedural and jurisdictional provisions, or
the other statutes at issue.
That leaves the earliest example, the 1794 unreported case of United
States v. Yale Todd.46 Because the decision was only rediscovered
36
Id. at 435.
37
32 U.S. (7 Pet.) 51, 85–89 (1833).
38
Parsons v. Bedford, 28 U.S. (3 Pet.) 433, 447–48 (1830).
39
Id. at 449.
40
33 U.S. (8 Pet.) 700, 703 (1834).
41
Act of March 2, 1799, ch. 22, § 65, 1 Stat. 627, 677 (“[A] continuance may be granted
until next succeeding term and no longer.”).
42
Phelps, 33 U.S. (8 Pet.) at 702 (statement of McLean, J.).
43
Mayor of New Orleans v. United States, 35 U.S. (10 Pet.) 662, 736–37 (1836).
44
Whittington, Repugnant Laws, supra note 27, at 99.
45
44 U.S. (3 How.) 212, 223–24, 230 (1845). For discussion, see William Baude,
Rethinking the Federal Eminent Domain Power, 122 Yale L.J. 1738, 1771–74 (2013).
46
The case documents are now available in 6 Documentary History of the Supreme Court
of the United States, 1789–1800, at 370–86 (Maeva Marcus et al. eds., 1998) [hereinafter
Documentary History]. The record was reprinted and the case was discussed earlier in Wilfred
J. Ritz, United States v. Yale Todd (U.S. 1794), 15 Wash. & Lee L. Rev. 220 (1958).
decades later47 and the Court wrote no opinion, we do not know exactly
what happened. But its resolution gives us some further clues about the
application of partly unconstitutional statutes.
As for the background, in an earlier set of correspondence and opinions
reported as part of Hayburn’s Case, the Supreme Court Justices generally
concluded that the Pension Act was partly unconstitutional.48 The
Supreme Court’s resolution of Hayburn’s Case itself confronted some
ancillary procedural and jurisdictional issues. But three groups of judges
sitting on circuit—including among them Justices Jay, Cushing, Wilson,
Blair, and Iredell—had all concluded that they couldn’t be required to
perform duties under the Pension Act because it assigned non-judicial
duties to federal judges. One of the groups, however, sitting as the circuit
court for the district of New York, expressed willingness to volunteer to
“execute this act in the capacity of commissioners.”49
Yale Todd brought the merits issue back to the Supreme Court by
considering the consequences of this choice. The Connecticut Circuit
Court did indeed act as volunteer commissioners under the Act and
awarded a pension to a wounded veteran named Yale Todd.50 The
Attorney General then sued Todd in the Supreme Court to recover the
money on the theory that it was invalidly paid. On February 17, 1794, the
Supreme Court agreed with the Attorney General.51
The best guess from the historical record is that the Court had
concluded that the statute was partly unconstitutional and that, as a
47
Secretary of War Henry Knox reported it to Congress at the time, but it seems to have
then been forgotten. 6 Documentary History, supra note 46, at 381–82. For instance, Chief
Justice Taney only belatedly learned about the case and had a discussion of it inserted into
United States v. Ferreira, 54 U.S. (13 How.) 40, 52–53 (1851). The editors of the
Documentary History of the Supreme Court noted that they “have had to use the copy of the
case papers in United States v. Yale Todd that is filed with the records of United States v.
Ferreira because the original papers no longer exist.” 6 Documentary History, supra note 46,
at 380.
48
Hayburn’s Case, 2 U.S. (2 Dall.) 409, 410 (1792). For discussion, see William Baude,
The Judgment Power, 96 Geo. L.J. 1807, 1818–20 (2008).
49
That was Chief Justice Jay, Justice Cushing, and District Judge Duane. 2 U.S. (2 Dall.) at
410 n.†; see also id. (“As, therefore, the business assigned to this court, by the act, is not
judicial, nor directed to be performed judicially, the act can only be considered as appointing
commissioners for the purposes mentioned in it, by official instead of personal descriptions.
That the Judges of this court regard themselves as being the commissioners designated by the
act, and therefore as being at liberty to accept or decline that office.”).
50
6 Documentary History, supra note 46, at 378–79. This time Jay and Cushing were joined
by a district judge named Richard Law. Id. at 379.
51
Id. at 381.
consequence, the United States was entitled to recover.52 The statute

could not constitutionally assign non-judicial duties to judicial officers,
as all of the Justices had already said, and for whatever reason,53 the
Supreme Court refused to accept the circuit court’s workaround.
And it has lessons for severability: As with the other cases, there was
no hint that the constitutional problem in Yale Todd invalidated anything
else in the pension act.54 But the remedy that Yale Todd did recognize is
also noteworthy. In modern cases, there is some dispute about whether
the action of a government official should be treated as void if there are
constitutional difficulties surrounding his statutory authority.55 In the
context of Yale Todd, the Justices apparently thought the answer was
“yes.”
To be sure, we should not squeeze too much precedent out of the Yale
Todd case. We do not know the Court’s precise reasons. But for whatever
weight it is worth, the Supreme Court seemed to think that an adjudication
made pursuant to an invalid grant of power should be set aside, though
the rest of the statute need not be.
Finally, stepping back, it is worth noting that the Court’s general
framework for judicial review during this time fit seamlessly with the
repugnancy framework. The Court was frequently ambiguous or
equivocal about whether it was holding a statute unconstitutional and then
refusing to apply it or instead holding that the statute should not be
interpreted to do something because that thing would be
unconstitutional.56 Under modern doctrine, these two things—
constitutional avoidance and severability analyses—may be quite
52
Ritz, supra note 46, at 227; Whittington, Repugnant Laws, supra note 27, at 68–69.
53
This reason might have been another constitutional objection, that the judges hadn’t
received separate appointments and commissions that complied with the constitutional
requirements, or a statutory objection, that the statute simply didn’t permit this kind of
workaround. See William Michael Treanor, Judicial Review Before Marbury, 58 Stan. L. Rev.
455, 537 n.423 (2005); James E. Pfander, Judicial Compensation and the Definition of Judicial
Power in the Early Republic, 107 Mich. L. Rev. 1, 36 & n.185 (2008).
54
See Robert L. Nightingale, How to Trim a Christmas Tree: Beyond Severability and
Inseverability for Omnibus Statutes, 125 Yale L.J. 1672, 1705 (2016) (“The history of
Hayburn’s Case and Todd demonstrates that the judges of the new federal courts understood
limits to exist on their power of judicial review: severability was the default. The Justices did
not put into question the validity of the rest of the 1792 pension scheme; they only nullified
the unconstitutional eligibility determinations.”).
55
See infra Section III.A.
56
Whittington, Repugnant Laws, supra note 27, at 23–24; see also Stephanie H. Barclay,
The Historical Origins of Judicial Religious Exemptions, 96 Notre Dame L. Rev. 55, 90–103
(2020) (giving examples of this practice in early nineteenth-century cases).
different.57 But at the time, under the classic doctrine of avoiding

unconstitutionality and under repugnancy analysis instead of modern
severability doctrine, these two holdings were functionally equivalent.58
This further evidences the deep roots of the repugnancy principle.
C. Applying Fallback Law

If the repugnancy of a statutory provision is an obstacle to Congress’s
goals, it can always attempt a new legislative solution. But it also does
not need to wait. If Congress anticipates constitutional challenges, it can
preemptively pass what Michael Dorf calls “fallback law,” a statutory rule
that kicks in on some contingency, such as the unconstitutionality of
another provision of law.59 Just as judges should enforce valid law that is
not repugnant to the Constitution, they should enforce valid fallback law
as well.
The simplest cases of fallback law are severability and inseverability
clauses. Severability clauses are the most common and the simplest.60 In
providing that some provisions or applications of the statute should be
severed from some others, they generally restate part of the first principle
of severability. Were the first principle universally followed, they might
be entirely unnecessary. And even under modern severability doctrine
they are often unnecessary since modern doctrine also employs a
presumption of severability.
Inseverability clauses are less common but more interesting. These
deviate from the classical rules of severability by yoking two provisions
together. In essence, an inseverability clause makes legal provisions
contingent, providing that a given rule will cease to have legal force if the
unconstitutionality of another is discovered. The trigger can be
formulated in different ways. It could be triggered by one part of a law
being unconstitutional. Or more commonly the trigger is having one part
of a law found unconstitutional by a court.61 These differences have
57
Adrian Vermeule, Saving Constructions, 85 Geo. L.J. 1945, 1946 (1997) (“Although
older forms of the two doctrines were indeed compatible, as currently articulated avoidance
and severability stand in severe reciprocal tension.”).
58
Id.
59
Michael C. Dorf, Fallback Law, 107 Colum. L. Rev. 303 (2007).
60
Caleb Nelson, Statutory Interpretation 144 (2011).
61
See, e.g., Michael D. Shumsky, Severability, Inseverability, and the Rule of Law, 41 Harv.
J. on Legis. 227, 243 n.76 (2004) (offering two examples that take this form); Abbe R. Gluck,
Reading the ACA’s Findings: Textualism, Severability and the ACA’s Return to the Court,
130 Yale L.J.F. 132, 159 & n.97 (2020) (offering a few more).
important consequences for retroactivity, for executive adjudication, and

other things. But either way, they reflect Congress’s choice about what
kind of contingencies to legislate.62
Finally, and rarer still, is the possibility of what Dorf calls
“‘substitutive’ fallback law.”63 This is fallback law that goes beyond
tying and untying particular statutory provisions from one another and
instead provides a new rule that takes effect only after a contingency
occurs. The most famous example (whose details are too complicated and
irrelevant to recount here) was enacted as part of the Balanced Budget and
Emergency Deficit Control Act and given effect by the Supreme Court’s
opinion in Bowsher v. Synar.64
All of these forms of fallback law should be enforced to the extent that
they are constitutionally permissible. And they are generally
constitutionally permissible.
To be sure, there may be some specific constitutional constraints on
Congress’s ability to enact fallback law, but any such constraints are
likely quite broad. One possible constraint is the non-delegation doctrine.
Contingent fallback law is unlikely to violate today’s non-delegation
doctrine (because almost nothing does).65 And even the most plausible
revisionist theories of the non-delegation doctrine would also uphold a
law whose effect is simply contingent on a future fact, such as a
determination of unconstitutionality.66
62
Tobias Dorsey argues that there are important differences between “what [he] would call
a sunset clause” and a true “nonseverability clause,” focusing on retroactivity and
prospectivity as a possible example. Tobias A. Dorsey, Sense and Severability, 46 U. Rich. L.
Rev. 877, 892 n.64 (2012). He also argues that true nonseverability clauses violate the
separation of powers, id., but it seems more accurate to see them simply as another form of
contingent legislation with retroactivity issues.
63
Dorf, supra note 59, at 305.
64
478 U.S. 714, 735 (1986) (falling back to Balanced Budget and Emergency Deficit
Control Act of 1985, Pub. L. No. 99–177, § 274(f), 99 Stat. 1038, 1100).
65
See Gundy v. United States, 139 S. Ct. 2116, 2123–24 (2019); see also Dorf, supra note
59, at 326 (“Accordingly, there is no plausible argument that Congress violates federal
nondelegation principles whenever it enacts a substitutive fallback provision.”). Dorf also
notes that courts’ occasional skepticism of severability and severability clauses may reflect
submerged non-delegation concerns, id. at 326–27, but I think those concerns are in fact
misplaced, see infra Subsection II.A.2.
66
Gundy, 139 S. Ct. at 2136–37 (Gorsuch, J., dissenting); Ilan Wurman, Nondelegation at
the Founding, 130 Yale L.J. 1490, 1516–17 (2021); Michael B. Rappaport, A Two-Tiered and
Categorical Approach to the Nondelegation Doctrine, in The Administrative State Before the
Supreme Court 195, 203–04 (Peter J. Wallison & John Yoo eds., 2022); Gary Lawson,
Delegation and Original Meaning, 88 Va. L. Rev. 327, 363–64 (2002); Michael W.
McConnell, The President Who Would Not Be King 107–08, 326–35 (2020).
The other major category of possible constraints are seemingly

coercive fallback rules. For instance, Dorf argues that using fallback law
to retaliate for an undesired judicial ruling might be unconstitutional.67 In
Heckler v. Mathews, the Court reserved the question of whether an anti-
funding fallback provision would be unconstitutional if it destroyed the
plaintiff’s standing to sue.68 And in United States v. Klein, the Court held
unconstitutional a jurisdictional provision that can be seen this way: as
stripping the Court’s jurisdiction if the Court made particular substantive
decisions that Congress did not want.69
On a formalist account of judicial duty, however, some of this coercion
can be brushed aside. True, there may be something unseemly about a
legislature saying: uphold our dubious statute, or we will take money from
needy children.70 But if judges are obligated to decide the
constitutionality of the statute without fear or favor, then the threat can
have no lawful effect. If the legislature is allowed to take money from
needy children, then the judges must not worry about that when deciding
the constitutionality of a separate law. And it may be true that the
legislature will dissemble, claiming that the funding cuts are really the
courts’ fault, but that is not a justification for a constitutional constraint.71
Even on this account, there would be some constraints. If the legislature
is not allowed to take money from needy children because that violates
some independent constitutional requirement, then it is unconstitutional
regardless of why it is threatened. This is why the legislature couldn’t
67
Dorf, supra note 59, at 328–42; see also Fred Kameny, Are Inseverability Clauses
Constitutional?, 68 Alb. L. Rev. 997, 997–99 (2005) (criticizing a state law that made
inseverable judicial salary increases and a controversial legislative expense increase).
68
Heckler v. Mathews, 465 U.S. 728, 739 n.5 (1984). The lower court judge had found the
clause “an unconstitutional usurpation of judicial power by the legislative branch of the
government.” Mathews v. Schweiker, No. 79-G-5251-NE, 1982 U.S. Dist. LEXIS 18124, at
*11 (N.D. Ala. Aug. 24, 1982). (Hat tip to Bruce K. Miller, Constitutional Remedies for
Underinclusive Statutes: A Critical Appraisal of Heckler v. Mathews, 20 Harv. C.R.-C.L. L.
Rev. 79 (1985).)
69
See United States v. Klein, 80 U.S. (13 Wall.) 128, 145–46 (1871). For more on this
reading of Klein, see Brian Kulp, Counteracting Marbury: Using the Exceptions Clause to
Overrule Supreme Court Precedent, 43 Harv. J.L. & Pub. Pol’y 279, 289–90 (2020).
70
Dorf, supra note 59, at 332–33 (describing a hypothetical fallback that cuts school lunch
funding).
71
Dorf, by contrast, argues that these accountability concerns, plus concerns about judicial
independence, justify a rule against coercive fallback law. Id. at 335–36. I think judicial
independence simply requires judges to ignore lawful but undesirable consequences, as noted
above, and I do not think accountability concerns justify an expansion of judicial power.
threaten summary executions of the deciding judges or anybody else.72

Similarly, the fallback rules in Klein and Mathews have the effect of
denying litigants access to the very court attempting to adjudicate their
claim—the constitutionality of such rules depends on the power to
regulate jurisdiction and access to courts.73 But the bottom line is that the
threat of merely undesirable policies would be permissible while the
threat of illegal ones would not.
Regardless of where precisely one draws these lines, they leave plenty
of space for the enactment of valid fallback law. This returns us to the
more fundamental point: if fallback law is not repugnant to some
provision of the Constitution, a court must enforce it, just as it enforces
other law.
II. IMPLICATIONS
A. Severability Doctrine
These basic principles provide a better way to approach the legal effect
of partly unconstitutional statutes. They have not always been followed
by our courts. At times the distance has been great, but at other times our
courts, using the modern language of “severability,” have come to
approximate some of them. If we tried to restore severability to first
principles, it would be much closer to a conclusive rule of severability
rebutted only by fallback law.
1. The Presumption of Severability

The current rule is that a statute is presumed to be severable, even if
Congress did not enact a severability clause.74 The fact that this is a
presumption rather than a rule reflects a turn-of-the-century meander.
After the Court first followed the fundamental principles of displacement
and repugnancy, it veered into inseverability doctrines in the late 1800s
and early 1900s, at one point even implying that statutes were presumed
72
But see Kameny, supra note 67, at 1003 (assuming, quite contrary to the above, that a
judicial salary inseverability clause would be “impermissible even if there had been no
constitutional provision specifically forbidding reductions in judicial salaries”).
73
Lea, supra note 9, at 760 n.142 (addressing Matthews). For my views on those topics, see
William Baude, Reflections of a Supreme Court Commissioner, 106 Minn. L. Rev. 2631,
2643–47 (2022).
74
See Alaska Airlines, Inc. v. Brock, 480 U.S. 678, 684 (1987); Nagle, supra note 24, at
220.
inseverable. It then rightly rejected that rule and emerged with the more
modern presumption of severability.75
This modern presumption of severability resembles the fundamental
principle that courts should simply refuse to enforce a law to the extent it
is repugnant to higher law. Both of these formulations permit judicial
review of unconstitutionality while leaving most of a law enforceable.
But the fundamental principles of repugnancy and displacement are
both clearer and more consistent with the separation of powers. For one
thing, they are clearer about scope. Under severability doctrine, people
wrestle with what exactly they are supposed to sever from what—
individual sections of the statute, words of the statute, applications of the
statute?76 Under repugnancy and displacement, it is clearer that courts
should focus on applications: apply law whenever it is valid, and do not
when it is not.
Indeed, under these fundamental principles, it is also clearer that the
judge does not actually “sever” anything. There is no surgery, no
amending, and no making law. The judge is still in the traditional posture
of applying law—higher law, lower law, and the rule for conflicts
between them.77
The fundamental principles are thus also more clearly a rule about
judicial authority and not a fiction about legislative intent. Judges do not
“presume” a statute to be severable because they have any particular
reason to know what the legislature would have wanted them to do with
the statute. They treat the statute this way because that is all the legal
authority they have in the ordinary case.
Finally, these principles make it clearer what it would take to rebut any
presumption of severability or displacement: law.
2. Severability Clauses
Current doctrine gives great weight to severability and inseverability
clauses. But it does not treat them as conclusively binding. That is, while
courts will usually follow severability and inseverability clauses,
sometimes they don’t.78 The first principles of severability would go
75
For an account of this history, see Nagle, supra note 24, at 213–19.
76
See, e.g., Mark L. Movsesian, Severability in Statutes and Contracts, 30 Ga. L. Rev. 41,
41 & n.1 (1995); Nagle, supra note 24, at 208 & n.24.
77
See supra Section I.B.
78
Shumsky, supra note 61, at 234–45. For a subsequent example, see Whole Woman’s
Health v. Hellerstedt, 136 S. Ct. 2292, 2319 (2016).
further. Because these clauses are a kind of fallback law, they should be
enforced absent a specific constitutional infirmity, just like other law.
Courts apply law—they don’t just give great weight to it.
The plurality opinion in Barr v. American Ass’n of Political
Consultants (“AAPC”) is especially auspicious on this point.79 There the
Court dealt with a severability clause contained in the Federal
Communications Act, invoking it to justify enforcing one part of the Act
(a ban on robocalls) after finding another part unconstitutional (an
exception to the ban for government-backed debt).80 In using the
severability clause, the Court specifically noted and disavowed the
possibility of “overrid[ing] the text of a severability or nonseverability
clause on the ground that the text does not reflect Congress’s ‘actual
intent’ as to severability.”81 We will no longer do that kind of thing, said
Justice Kavanaugh:
That kind of argument may have carried some force back when courts
paid less attention to statutory text as the definitive expression of
Congress’s will. But courts today zero in on the precise statutory text
and, as a result, courts hew closely to the text of severability or
nonseverability clauses.82
This reasoning treats severability and inseverability clauses as law, as it
should.
Justice Kavanaugh’s plurality opinion treated severability clauses as
law in another more subtle and more technical way. The severability
clause in the case had been enacted as part of the Federal Communications
Act of 1934.83 The two substantive provisions in the case were not added
until 1991 and 2015.84 The new provisions did not mention the old
severability clause, but the old severability clause applied to those new
79
140 S. Ct. 2335 (2020).
80
Id. at 2352–56. This is not the only way to frame the analysis, which gets into a more
complicated question of “unconstitutional combinations,” discussed infra at Section III.B.
81
AAPC, 140 S. Ct. at 2349.
82
Id. The plurality did still say that it should adhere to severability clauses “absent
extraordinary circumstances.” Id. It’s not clear what that exception was about. Maybe the
doctrine of absurdity? Cf. Brett M. Kavanaugh, Fixing Statutory Interpretation, 129 Harv. L.
Rev. 2118, 2156–57 (2016) (reviewing Robert A. Katzmann, Judging Statutes (2014))
(endorsing a narrow absurdity doctrine). Or the possible instances of unconstitutional fallback
law? See supra Section I.C.
83
Communications Act of 1934, ch. 652, § 608, 48 Stat. 1064, 1105 (codified at 47 U.S.C.
§ 608). (It was called a “separability” clause, as was then sometimes common.)
84
AAPC, 140 S. Ct. at 2344–45.
provisions precisely because it was binding law: it governed “this Act,”

and therefore continued to govern any new amendments or additions to
the Act over time.85 As the plurality put it, “a severability clause must be
interpreted according to its terms, regardless of when Congress enacted
it.”86
Similarly, an earlier majority opinion in Seila Law v. CFPB had
invoked a severability clause despite the objection that it was “non-
probative ‘boilerplate’” that “‘appears almost 600 pages before the
removal provision at issue.’” 87 These objections could matter if
severability clauses are just pieces of evidence about some underlying
congressional policy. But they have much less force if the clause is law,
which generally applies over long distances.
Seila Law was not as clear as the AAPC plurality about the legal status
of severability clauses. It defended a “boilerplate” clause as “tried-and-
true language to ensure a precise and predictable result,” defended the
“logical and prominent” placement of the severability clause, and
speculated about the disruption that would occur under alternative
approaches to severability.88 But it is a healthy step toward treating
clauses about severability as fallback law, which should be applied
whenever it is constitutionally valid.
Doing so should also mean putting to rest some of the spurious
separation-of-powers challenges raised against such clauses and against
severability more generally. For instance, the Supreme Court has
sometimes complained about applying severability analysis to laws that
are written in an overbroad fashion. Consider this passage from United
85
See William Baude & Stephen E. Sachs, The Law of Interpretation, 130 Harv. L. Rev.
1079, 1102–04 (2017) (explaining how previously enacted interpretive rules apply to future
statutes). There is one additional wrinkle: the plurality quoted the version of the clause
contained in the U.S. Code, which said “this chapter” instead of “this Act.” AAPC, 140 S. Ct.
at 2352 & n.10 (emphasis added) (quoting 47 U.S.C. § 608, 48 Stat. at 1105). Perhaps that
made the analysis seem slightly easier, but as a technical matter, it is the public law, not the
paraphrasing of the U.S. Code, that is the law. See Jesse M. Cross & Abbe R. Gluck, The
Congressional Bureaucracy, 168 U. Pa. L. Rev. 1541, 1656–74 (2020).
86
AAPC, 140 S. Ct. at 2352.
87
140 S. Ct. 2183, 2209 (2020) (quoting Brief for the Petitioner at 45, Seila Law, 140 S. Ct.
2183 (No. 19-7)). They also argued more generally that “severability ‘rarely turn[s] on the
presence or absence of such a clause,’ ” and that the clause only applied to severing one title
of the Act from another. Brief for the Petitioner, supra, at 45 (quoting United States v. Jackson,
390 U.S. 570, 585 n.27 (1968)).
88
Seila Law, 140 S. Ct. at 2209–10.
States v. Reese, an aggressively anti-severability decision that rejected a

civil rights prosecution during Reconstruction:
It would certainly be dangerous if the legislature could set a net large
enough to catch all possible offenders, and leave it to the courts to step
inside and say who could be rightfully detained, and who should be set
at large. This would, to some extent, substitute the judicial for the
legislative department of the government.89
The same passage has been quoted as authority in modern cases involving
free speech and abortion as a reason to disregard a statute’s severability
clause.90
But the separation of powers objection to this scenario is confused. The
legislature has decided to prohibit as much conduct as it can, and the
judiciary has decided that the Constitution imposes limits on that.
Respecting the judiciary’s prerogative, the legislature has acknowledged
these limits and made clear that the judiciary can enforce them in every
case where they are relevant, but only in such cases. It doesn’t seem too
much to ask that the judiciary decide exactly what those limits are and
apply them to the law it believes to be so limited. The Court’s real
complaint in these cases is not that the judiciary is doing the legislature’s
job, but that the legislature is refusing to do the judiciary’s.91
To be sure, there is a potentially more valid complaint about some
severability clauses in these scenarios, which is that they could be
unconstitutionally vague. If the Constitution imposes some restriction on
the vagueness of a law,92 and if the judiciary’s own doctrinal tests flunk
that vagueness test, then maybe there is a vagueness problem with using
89
92 U.S. 214, 215, 221 (1875).
90
See Whole Woman’s Health v. Hellerstedt, 136 S. Ct. 2292, 2319 (2016) (quoting Reno
v. ACLU, 521 U.S. 844, 884 n.49 (1997) (in turn quoting Reese, 92 U.S. at 221)); Ayotte v.
Planned Parenthood of N. New Eng., 546 U.S. 320, 330 (2006) (quoting Reese, 92 U.S. at
221).
91
Peter Salib tries to rehabilitate the Reese principle by arguing that in some of these cases,
such as Reese, the statute has “no separate, constitutional commands” and so “severing the
unconstitutional portions of a law means severing the entire law.” Peter N. Salib, Ban Them
All; Let the Courts Sort Them Out.: Savings Clauses, the Texas Abortion Ban, and the
Structure of Constitutional Rights, 100 Tex. L. Rev. Online 13, 26 (2021). If so, he concludes,
“the only way for the law to continue requiring anything is for the Court to make up a new
rule from scratch,” which it shouldn’t do. Id. I am not sure I agree with the premise, but I agree
that if it is true, the conclusion follows.
92
For instance, another part of Reno, 521 U.S. at 870–74, dealt with First Amendment
vagueness.
judicial doctrine to mark the bounds of what is prohibited.93 But that

would be a somewhat ironically pessimistic assessment of judicial
doctrine, and presumably the Court would still need to enforce whatever
parts of the law were not repugnant to the vagueness doctrine.94
3. Fully Operative Law and Counterfactual Intent

Severability doctrine also sometimes allows the presumption of
severability to be rebutted, even in the absence of a severability clause.
According to the cases, the presumption can be rebutted if “it is evident
that the Legislature would not have enacted those provisions which are
within its power, independently of that which is not.”95 The part that
remains must also be “fully operative as a law.”96 The first of these
questions is a counterfactual about how Congress would have reacted to
the constitutional problem; the second is an objective question about the
statute.97
The counterfactual test made the difference in Murphy v. NCAA. In that
case, the Supreme Court concluded that two provisions of a federal anti-
gambling statute violated the anti-commandeering doctrine because they
purported to stop a state government from authorizing gambling.98 It then
turned to the other provisions of the statute, which fell like a string of
dominos. The statute’s direct prohibitions on sports gambling schemes
would result in “a scheme sharply different from what Congress
contemplated” if left to stand on their own.99 These prohibitions on
private actors “were obviously meant to work together” with the
commandeering provisions.100 And that left only a stand-alone ban on
advertising, which Congress rarely enacts, and might be
93
Larry Alexander, There Is No First Amendment Overbreadth (But There Are Vague First
Amendment Doctrines); Prior Restraints Aren’t “Prior”; and “As Applied” Challenges Seek
Judicial Statutory Amendments, 27 Const. Comment. 439, 440–41 (2011).
94
This is contrary to Johnson v. United States, 576 U.S. 591, 603–04 (2015), and perhaps
United States v. L. Cohen Grocery Co., 255 U.S. 81, 89–91 (1921), on which Johnson relied.
95
Alaska Airlines, Inc. v. Brock, 480 U.S. 678, 684 (1987) (quoting Buckley v. Valeo, 424
U.S. 1, 108 (1976) (per curiam)).
96
Id.
97
Lea argues that these “are not independent, standalone tests. Rather, they are both aspects
of the search for legislative intent.” Lea, supra note 9, at 745 n.38. Regardless, one aspect is
more objective than the other. Accord id. at 745.
98
Murphy v. NCAA, 138 S. Ct. 1461, 1478–82 (2018).
99
Id. at 1482.
100
Id. at 1483.
unconstitutional.101 The result was that an anti-commandeering problem

with part of the law resulted in a completely unenforceable statute.
The more objective part of the test rarely seems to be dispositive. But
perhaps something like it made the difference in the 1922 case of Hill v.
Wallace, where the Court held unconstitutional a tax on futures contracts
that taxed contracts unless they were “made by or through a member of
the Board of Trade designated by the Secretary of Agriculture” and
consistent with various regulations.102 The Court concluded that this tax
was impermissibly regulatory (like the child labor tax it invalidated the
same day).103 And it then concluded that if there were no tax on contracts
that did not comply with the regulations, there would be no point to the
regulations themselves: “Section 4 with its penalty to secure compliance
with the regulations of Boards of Trade is so interwoven with those
regulations that they cannot be separated. None of them can stand.”104
Both of these aspects of severability doctrine should be viewed with
suspicion. They are hard to square with first principles, because they seem
to call on courts to do something other than disregarding repugnant law
and enforcing valid law. Instead, they ask courts to set aside valid law
absent specific legal instructions to do so.
To be sure, it is possible that these tests could be seen by anti-textualists
as a sort of fallback law. If one believes that valid law can be found in
unwritten congressional intent or inferred congressional policy, one might
ask oneself “what Congress would have enacted” as a way to find legal
instructions in the absence of any sort of fallback law. But most modern
judges are not willing to say that unwritten counterfactual intent is law,105
yet the test hangs around, seemingly legitimated by repetition in the case
reports. Those who do not think that counterfactual intent is law should
probably stop treating it as if it were.
The second, more objective part of the test—asking oneself if the
provision remains “operative as law”—may be less problematic. Indeed,
if applied narrowly, the test is almost tautologically unobjectionable. If a
provision is no longer “operative as law,” how could a judge continue
101
Id. at 1484.
102
259 U.S. 44, 63–68 (1922); see Nightingale, supra note 54, at 1710–11 (discussing this
example and drawing it from Alaska Airlines).
103
Hill, 259 U.S. at 67 (citing Bailey v. Drexel Furniture Co. (Child Labor Tax Case), 259
U.S. 20, 37–38 (1922)).
104
Id. at 70.
105
Though some judges today may still be willing to treat it this way in a pinch. See infra
Section III.B.
applying it anyway? If one provision is logically or linguistically

conditional on another provision, and the latter cannot be given effect,
then the former cannot either. But precisely because the test is so narrow,
it is largely irrelevant and adds little to the more basic principles.
One can see how the doctrine got to its current form, and how it may
have been asking the right questions for the dominant judicial fashions at
the time it took hold. But those ways of thinking of law have largely fallen
out of favor, and for good reason. To figure out the law of severability
today, we should not just recite the too-familiar phrases from severability
doctrine but rather ask the more fundamental question: is there any
binding law that is triggered by the statute’s partial unconstitutionality? If
so, that instruction should be followed. But if not, courts should enforce
all law that is not repugnant to the Constitution.
4. State Law
Most of the recent Supreme Court cases about severability have
centered on federal statutes, which this piece has focused on so far. The
same principles apply—with some important modification—to
interpreting state law.
The issue of state law is more complicated because federal judicial
power is a question of federal law while the meaning and interpretation
of state law is a question of state law. Because severability blends issues
of judicial power and issues of interpretation, it blends state and federal
law in a way that can be confusing. But the severability first principles let
us disentangle the relevant roles of federal and state law.
The principle that federal courts should disregard repugnant law is no
different for repugnant state law than for repugnant federal law. That
principle is simply an application of the basic Marbury rule, which is a
rule about the scope of federal judicial power.106 If anything the point may
be even easier for state laws, because the Supremacy Clause explicitly
106
States might conceivably choose to understand judicial power differently or to vest their
courts with a kind of power that federal law would not call “judicial.” But it is not clear that
states have in fact done so, see Caleb Nelson, The Legitimacy of (Some) Federal Common
Law, 101 Va. L. Rev. 1, 26–28 (2015) (questioning the assumption that state courts have
different powers from federal courts), nor whether the Federal Constitution would permit those
choices, see Amnon Lehavi, Judicial Review of Judicial Lawmaking, 96 Minn. L. Rev. 520,
546–52 (2011).
emphasizes that state laws that are contrary to federal law must be
disregarded.107
However, the principles that federal courts should apply all non-
repugnant law, and apply all fallback law, are slightly more delicate. As
to applying all non-repugnant state law, the Constitution is less explicit
about this obligation than it is about applying federal law. Federal law is
the “supreme Law of the Land,” according to the Supremacy Clause,
while there is nothing explicit about an obligation to apply state law.108
Some have inferred from this that “the Constitution permits the justices
to subordinate state law to judge-created doctrines.”109
But we probably should not read too much into these gaps in the
Supremacy Clause. The idea that courts should apply law where it was
not displaced was a widely shared understanding of judicial power and
duty.110 It is most likely that the Supremacy Clause addresses particular
permutations of this duty (like the duty of state judges to apply federal
law) because they were seen to pose a particular question at the time.111
So, the general principle remains that judges should apply whatever law
has not been displaced by higher law.112
As to applying state fallback law, this can become more complicated
in practice because there is much more potential diversity in the content
of a state’s fallback law. For instance, state legislation or state common
law might provide for rules of inseverability much more commonly than
federal law does.113 The state might make much greater use of legislative
intent or even counterfactual intent in finding implicit fallback law. So
federal courts should not be so quick to assume that there is no fallback
107
U.S. Const. art. VI, § 2.
108
Id.
109
Jonathan F. Mitchell, Stare Decisis and Constitutional Text, 110 Mich. L. Rev. 1, 45
(2011) (emphasis omitted).
110
See Walsh, supra note 5, at 755–57.
111
Caleb Nelson, Preemption, 86 Va. L. Rev. 225, 249, 251, 256 (2000).
112
There is also the obligation of 28 U.S.C. § 1652, originally enacted as § 34 of the
Judiciary Act of 1789, which now reads: “The laws of the several states, except where the
Constitution or treaties of the United States or Acts of Congress otherwise require or provide,
shall be regarded as rules of decision in civil actions in the courts of the United States, in cases
where they apply.” For discussion, see Mitchell, supra note 109, at 51–55 (offering different
constructions of the Act); see also Stephen E. Sachs, The Unlimited Jurisdiction of the Federal
Courts, 106 Va. L. Rev. 1703, 1721 (2020) (noting limits to the force of the Act).
113
See Michael C. Dorf, Facial Challenges to State and Federal Statutes, 46 Stan. L. Rev.
235, 295–304 (1994) (cataloguing state severability doctrines as of 1994, which were
“remarkably uniform,” none of which included the categorical severability rule and many of
which referenced legislative intent).
law if they cannot find it on the text of the page. And while the content of
state fallback law may be more complicated, the federal courts also have
various tools—abstention, certification, remand, deference doctrines—to
defer those questions to state courts that may be better equipped to answer
them.
So, while the Supreme Court has sometimes said that “[s]everability is
of course a matter of state law,”114 and other times ignored this statement
to apply federal severability principles,115 there are some elements of both
that should be carefully separated. Federal courts must disregard a legal
provision in cases where it is repugnant, and only in those cases. This is
a principle of federal law regarding the power of federal courts. They
must then apply any relevant fallback law, whatever form it may take.
The content of this law is a question of state law, and it may well deviate
from federal norms of textualism or whatever else. Severability is thus a
question of both federal and state law, at different steps.
B. Justices Gorsuch and Thomas

Some of the Justices are getting these basics right. In Murphy, Justice
Thomas wrote separately, correctly recognizing that something had gone
wrong with the Court’s severability doctrine. Along the very lines now
sketched here, Justice Thomas wrote that the Court’s severability doctrine
is “in tension with traditional limits on judicial authority.”116 “[C]ourts do
not have the power to ‘excise’ or ‘strike down’ statutes,” he wrote, but
rather have “the negative power to disregard an unconstitutional
enactment” in a particular case.117 Beyond this, any further decisions
about severability ought to be an exercise in statutory interpretation, but
Justice Thomas feared that current severability doctrine focuses
incorrectly on things such as hypothetical congressional intent.118 In sum,
Justice Thomas endorsed the traditional approach of repugnancy and
displacement, plus fallback law.119
114
Leavitt v. Jane L., 518 U.S. 137, 139 (1996).
115
Ayotte v. Planned Parenthood of N. New Eng., 546 U.S. 320, 328–30 (2006).
116
Murphy v. NCAA, 138 S. Ct. 1461, 1485 (2018) (Thomas, J., concurring).
117
Id. at 1486 (quoting Massachusetts v. Mellon, 262 U.S. 447, 488 (1923)).
118
Id. at 1486–87. Justice Thomas also argued that severability doctrine is in tension with
principles of standing, id. at 1487, an issue that comes up again in Seila Law, California v.
Texas, and other cases. See infra Section III.C.
119
Slade Mendenhall & Brian Underwood, To Sever or Not to Sever: Mixed Guidance from
the Roberts Court, 69 Drake L. Rev. 273, 278, 287–89 (2021), seem to read Justice Thomas
differently, believing him to want the entire statute to fall as an inseverable whole. I do not
Justice Thomas has repeated these views, often joined by Justice

Gorsuch, in subsequent cases. He reiterated them in the separation of
powers case of Seila Law.120 And Justice Gorsuch discussed them, joined
by Justice Thomas, in AAPC.121 This is cause for celebration: Their
general approach is quite right, it is high time that somebody on the Court
put it forward, and in some ways the Court itself seems to be gravitating
closer to some of their observations and terminology.122
But canonization would be premature: First, both Justices have
exaggerated how simple their view is, or how clearly it can resolve some
of the cases the Court faces. This became especially apparent in United
States v. Arthrex and Collins v. Yellen, where Justice Gorsuch and Justice
Thomas took their theory in very different directions. Further
investigation also calls into question their separate opinions in Seila Law
and AAPC. These cases will be discussed in Part III, which discusses the
complications of ultra vires acts and unconstitutional combinations.
Second, it is too soon to tell if the Justices can stick to their own principles
even in simpler severability cases, such as the inseverability of the
Affordable Care Act (“ACA”).
Let us briefly consider that ACA severability question. In NFIB v.
Sebelius, Justice Thomas joined a jointly bylined dissent that found the
individual mandate to buy health insurance unconstitutional, as was the
expansion of Medicaid.123 That opinion also concluded that the entire
ACA was inseverable from these two provisions. The Justices rejected
“uncritical severance” as “assum[ing] the legislative function,” applied a
version of the Alaska Airlines test,124 and declared unenforceable even
provisions “such as requiring chain restaurants to display nutritional
content” that they conceded “appear likely to operate as Congress
think that is the right reading. See Murphy, 138 S. Ct. at 1486 (Thomas, J., concurring)
(maintaining that courts’ rulings are limited to “the case before them”); Collins v. Yellen, 141
S. Ct. 1761, 1795 (2021) (Thomas, J., concurring).
120
140 S. Ct. 2183, 2219–24 (2020) (Thomas, J., concurring in part and dissenting in part).
121
140 S. Ct. 2335, 2365–67 (2020) (Gorsuch, J., concurring in the judgment in part and
dissenting in part). Interestingly, Justice Thomas joined only Justice Gorsuch’s severability
section, not the merits; he joined the plurality’s merits holding and did not write separately to
explain how he stitched the two together. Id. at 2343 (plurality opinion).
122
See, e.g., supra notes 79–87; see also AAPC, 140 S. Ct. at 2352 n.8 (comparing plurality
approach to Justice Thomas’s and concluding that “in many cases, the different paths lead to
the same place”).
123
567 U.S. 519, 646–61, 671–91 (2012) (joint dissent).
124
Id. at 691–92.
intended” because the Justices found “no reason to believe that Congress
would have enacted them independently.”125
This analysis cannot be squared with Justice Thomas’s later writings in
Murphy and the subsequent cases. The NFIB joint dissent found two
provisions of the ACA unconstitutional. So, as Justice Thomas later
recognized, the ordinary course would be to disregard those provisions,
and then turn to statutory interpretation as needed.126 The ACA also
contained no explicit fallback law or inseverability provision making the
entire statute conditional on the individual mandate. So, the rest of the
statute would remain presumptively enforceable.
There would be two further questions. One question is whether the
ACA’s express textual conclusion about the importance of the individual
mandate could be read as a kind of fallback law—an implicit
inseverability clause. The answer is no, as I will discuss shortly.127 The
other question would be whether any of the other provisions of the ACA
had separate constitutional infirmities, which would of course render
them unenforceable for separate reasons. The joint dissent did not have to
confront this question, but by Justice Thomas’s lights, it is plausible that
he would find other parts of the Act to exceed Congress’s powers to tax,
spend, and regulate commerce.128 But, again, this would be a case-by-case
inquiry that would not dispose of the whole Act.
None of this is to criticize Justice Thomas. The joint dissent was no
doubt written quickly and had to cover a great deal of shifting ground.
There was likely some incentive for the four to hang together. At the time,
Justice Thomas had not developed his more recent wisdom about
severability. He was no doubt making a good faith attempt to apply the
Court’s confused doctrine.129 It is just to note that after Murphy, Justice
125
Id. at 705. Indeed, in passages like this it is possible that the dissenters were applying
something like Alaska Airlines but without the presumption of severability.
126
As Justice Thomas later recognized in Murphy v. NCAA, many of the inseverability
arguments should not have been addressed at all in that case, because they were not part of the
controversy before the Court. See 138 S. Ct. 1461, 1487 (2018) (Thomas, J., concurring); Lea,
supra note 9, at 788–803 (discussing the standing issues these types of “gratuitous severability
rulings” present); see also infra Section III.C (addressing severability and procedure).
127
See infra notes 139–43 and accompanying text.
128
Cf. NFIB, 567 U.S. at 707–08 (Thomas, J., dissenting) (reminding readers that he
continued to oppose modern Commerce Clause doctrine).
129
Cf. Center for the Study of Constitutional Originalism, First Paper Pt. 3: 13th Annual
Hugh and Hazel Darling Works-in-Progress Conference, YouTube, at 39:32 (Mar. 29, 2022),
https://www.youtube.com/watch?v=41l5JI5T4z4 [https://perma.cc/SFY9-RUKM]
Thomas’s analysis of the ACA would surely have looked different.

Indeed, in Murphy, Justice Thomas seemed to distance himself from the
NFIB dissent.130
Justice Gorsuch is more of a puzzle. He was not on the Court when
NFIB v. Sebelius was decided, so he did not have occasion to voice his
views on severability until the most recent case of California v. Texas.131
In that case, Justice Thomas denied standing and did not take a stand on
the severability of the statute.132 But Justice Gorsuch joined Justice
Alito’s dissent concluding that the states had standing to make
inseverability arguments, which required addressing inseverability.133
For these two Justices to join was somewhat tricky, because Justice
Alito—the author of the dissent—had already reached inseverability in
the joint dissent in NFIB v. Sebelius, and he had not joined any of the
subsequent Gorsuch/Thomas revisionist opinions. So, Justice Alito
inserted a paragraph to explain why Justice Gorsuch could reach the same
result as Justice Alito had:
The same result follows under the new approach to questions of partial
unconstitutionality that some Members of the Court have adopted in the
years since NFIB. They have suggested the severability analysis should
track ordinary rules of statutory interpretation. Seila Law, 591 U.S., at
––––, –––– – ––––, 140 S.Ct., at 2199–2200 (THOMAS, J., concurring
in part and dissenting in part). In their view, Congress decides whether
the provisions it enacts are linked to one another or not, and the answer
lies in the ordinary tools of statutory construction. And everything the
NFIB dissenters said points to the same conclusion as a matter of the
ACA’s text, history, and structure. The relevant provisions were passed
as a comprehensive exercise of Congress’s Commerce Clause and
(arguably) Taxing Clause powers. Those powers cannot justify the
individual mandate. The statutory text says the individual mandate is
“essential” to the overall scheme, 42 U.S.C. § 18091(2)(I), and it
repeatedly states that the various provisions work “together,” NFIB,
567 U.S., at 694–696, 132 S. Ct. 2566 (joint dissent). It does not matter
that this language appears in a section entitled “findings” as opposed to
(comments of Randy Barnett) (describing how the severability issue was litigated and offering
subsequent reflections).
130
Murphy, 138 S. Ct. at 1487 (Thomas, J., concurring).
131
141 S. Ct. 2104 (2021).
132
Id. at 2120–23 (Thomas, J., concurring).
133
Id. at 2123–24 (Alito, J., dissenting).
a section entitled “severability.” Congress can link distinct provisions

in any number of ways, on this view, so long as it does so in the text.
The broader statutory history and structure, moreover, reinforce that
conclusion. The NFIB dissent explained how the ACA’s provisions
work in tandem to alter the insurance market. 567 U.S., at 691–706, 132
S. Ct. 2566. Here, the individual mandate requires individuals to obtain
“minimum essential coverage.” 26 U.S.C. § 5000A(f). The reporting
requirements, in turn, implement the mandate—indeed, they explicitly
cross-reference § 5000A—by requiring employers to provide
information about such coverage. §§ 6055(e), 6056(b)(2)(B). And the
adult-children coverage requirement works as part of a cohesive set of
insurance reforms central to the ACA’s overall structure, which turns
on healthy persons’ entry into the market via the individual mandate.
See 42 U.S.C. § 300gg–14(a). The individual mandate is thus
inseverable from the provisions burdening the States under either
approach to severability.134
This passage almost135 asks the right question—does the Affordable
Care Act provide that if the individual mandate is unconstitutional, the
reporting requirements and adult-children coverage requirement should
not be enforced? But its answer is wrong.
Focusing on the specific statutory requirements first: The individual
mandate requires people to buy a particular kind of insurance, or pay a
penalty, and defines what kind.136 The reporting requirements require
employers to say whether they have provided that kind of insurance.137 If
the mandate is now unconstitutional, nobody has to buy it, and nobody
has to pay. But that does not mean nobody has to report it. The connection
between the reporting requirement and the mandate was that they used the
same criteria for what made an insurance plan covered. But the
unconstitutionality of the mandate did not make the criteria
unconstitutional or forbid all cross-references to those criteria.
The adult-children coverage requirement has even less explicit
connection to the individual mandate. It simply says:
134
Id. at 2139–40.
135
To ask whether the provisions are linked “in any number of ways” by “text,” id., is
imprecise. The question is whether their legal force is linked by law.
136
26 U.S.C. § 5000A(a)–(b), (f).
137
26 U.S.C. §§ 6055(e), 6056(b)(2)(B).
A group health plan and a health insurance issuer offering group or

individual health insurance coverage that provides dependent coverage
of children shall continue to make such coverage available for an adult
child until the child turns 26 years of age. Nothing in this section shall
require a health plan or a health insurance issuer described in the
preceding sentence to make coverage available for a child of a child
receiving dependent coverage.138
No word about the mandate, nothing saying that judges should stop
enforcing the provision if other economic premises of the law are false.
That leaves only the argument that the Affordable Care Act contained
what is effectively an inseverability clause because it repeatedly finds that
the individual mandate is “essential,” to the larger regulatory scheme and
to creating effective health insurance markets.139 One response is that this
finding applied only to the 2010 mandate but not to the 2017 amended
mandate. This response, however, would be unavailing against a true
inseverability clause. If Congress enacts a severability or inseverability
clause into law, it can amend the subjects of that law just as any other.140
The more fundamental problem is that the “essential” finding is not an
inseverability instruction. Every time Congress makes such findings to
invoke its necessary-and-proper powers, it is stating its view about the
importance and relevance of what it is doing.141 It is not thereby making
fallback law. Indeed, nobody seems to have taken the
essential = inseverable argument truly seriously: The ACA findings
declare the individual mandate as “essential” not only to other provisions
of the ACA, but to all of the Employee Retirement Income Security Act
138
42 U.S.C. § 300gg-14(a).
139
42 U.S.C. § 18091(2)(H), (I), (J); see also Josh Blackman, Unreviewable: The Final
Installment of the “Epic” Obamacare Trilogy, 2020–2021 Cato Sup. Ct. Rev. 109, 129–30
(agreeing with this argument).
140
See Barr v. AAPC, 140 S. Ct. 2335, 2349 n.6 (2020) (plurality opinion) (“When Congress
enacts a law with a severability clause and later adds new provisions to that statute, the
severability clause applies to those new provisions to the extent dictated by the text of the
severability clause.”); see also id. at 2352 (“To get around the text of the severability clause,
plaintiffs point out that the Communications Act’s severability clause was enacted in 1934,
long before the TCPA’s 1991 robocall restriction and the 2015 government-debt exception.
But a severability clause must be interpreted according to its terms, regardless of when
Congress enacted it.”); Baude & Sachs, supra note 85, at 1102–04 (explaining how previously
enacted interpretive rules apply to future statutes).
141
See Gluck, supra note 61, at 155–58 (giving examples of similar findings).
and the Public Health Service Act.142 If the “essential” finding were an
inseverability clause, it would condemn these laws as well, which nobody
was willing to argue or accept.143
At bottom, many aspects of Justice Alito’s dissent could hold up if the
Affordable Care Act contained an inseverability clause.144 So, too, Justice
Gorsuch might well be able to join such a dissent if there were an
inseverability clause. But there was not, and so this was a mistake.
C. Other Doctrines
1. Facial Challenges
In modern litigation, there is a frequently drawn distinction between
challenging a statute “on its face” or challenging it “as applied” to the
litigant in question.145 This is the difference between a challenge that
attacks the entire statute, and would imply the invalidity of the entire
statute, and a challenge that pertains only to the challenger’s case.
The Court has said that “facial challenges are disfavored” under current
doctrine,146 and the Court has sometimes explicitly connected this
principle to severability.147 There is something to that. The first principles
of severability indicate that as-applied challenges are a presumptive and
normal way of thinking about constitutional litigation. But it is also more
complicated. As Richard Fallon has shown, in practice, the Supreme
Court itself does not simply disfavor facial challenges,148 nor does a
presumption of severability do all of the work here.149
Nothing stops a litigant from making an argument that logically implies
that the statute would be unconstitutional in a range of cases or in all
142
42 U.S.C. § 18091(2)(H) (“Under the Employee Retirement Income Security Act of 1974
(29 U.S.C. 1001 et seq.), the Public Health Service Act (42 U.S.C. 201 et seq.), and this Act,
the Federal Government has a significant role in regulating health insurance. The requirement
is an essential part of this larger regulation of economic activity, and the absence of the
requirement would undercut Federal regulation of the health insurance market.”).
143
Florida ex rel. Att’y Gen. v. U.S. Dep’t of Health & Hum. Servs., 648 F.3d 1235, 1326–
27 (11th Cir. 2011) (making this point), aff’d in part, rev’d in part sub nom. NFIB v. Sebelius,
567 U.S. 519 (2012).
144
For an endorsement of his theory of standing, see Section III.C.
145
See, e.g., Wash. State Grange v. Wash. State Republican Party, 552 U.S. 442, 449 (2008).
146
Id. at 450–51 (2008).
147
See, e.g., Ayotte v. Planned Parenthood of N. New Eng., 546 U.S. 320, 328–29 (2006).
148
Richard H. Fallon, Jr., Fact and Fiction About Facial Challenges, 99 Calif. L. Rev. 915,
935–48 (2011).
149
Id. at 955–59.
cases. For instance, if Congress passes a law regulating spending on

political campaigns, and A wishes to spend money and believes that all
such legislation is unconstitutional, one might shorthand A’s argument as
a facial challenge. So too for a law that facially discriminates on the basis
of race, for instance.150
It is just that the consequences of A’s argument will not result in the
statute being erased, invalidated, or judicially repealed. The consequences
of the argument will be to ignore the statute in A’s case and then issue
whatever judicial relief is appropriate in light of the non-existence of the
statute. (For instance, to dismiss the criminal charges against A, or to
enjoin a government official who was threatening to punish A for her
spending, or to issue damages against an official who did so punish A in
the past.)
Similarly, nothing in the severability framework says that all
constitutional rules must operate at the level of enforcement. It is possible
for the Constitution to grant “Rights against Rules,”151 to say that no rule
about speech can be enforced unless it has a certain property or that no
law can be enforced if it was enacted with a certain kind of intent. This
does not change the severability framework, but it can effectively create
a rule of fallback law in particular situations. For instance, if one believes
that the validity of a rule about free speech turns on whether it is
unconstitutional as applied to a substantial number of others,152 then that
substantive free speech doctrine effectively creates a fallback principle of
inseverability for certain free speech claims. One could make similar
arguments about other constitutional provisions, such as the enumerated
powers, though this has gotten less attention.153
150
Id. at 921. Caleb Nelson notes that, in recent decades, the Supreme Court has become
more willing to engage in judicial review of constitutionally forbidden intent, possibly
explaining the rise in such facial challenges. Caleb Nelson, Judicial Review of Legislative
Purpose, 83 N.Y.U. L. Rev. 1784, 1876–79 (2008).
151
See Matthew D. Adler, Rights Against Rules: The Moral Structure of American
Constitutional Law, 97 Mich. L. Rev. 1 (1998). Adler’s theoretical analysis notwithstanding,
I agree with Michael Dorf that “[e]ven if some rights are rights against rules, others are rights,
simpliciter.” Michael C. Dorf, The Heterogeneity of Rights, 6 Legal Theory 269, 270 (2000).
152
This is the overbreadth doctrine. But see United States v. Sineneng-Smith, 140 S. Ct.
1575, 1585 (2020) (Thomas, J., concurring) (criticizing basis for this doctrine); Ams. for
Prosperity Found. v. Bonta, 141 S. Ct. 2373, 2390–91 (2021) (Thomas, J., concurring in part
and concurring in the judgment) (same). Larry Alexander argues that First Amendment
overbreadth is instead another instance of constitutional doctrine being too vague to be used
as a rule for primary conduct. Alexander, supra note 93, at 441.
153
This is one reading of United States v. Reese, discussed supra in the text accompanying
notes 27, 89–90. The Court said that because the law was overbroad, it was not “appropriate
These situations aside, the first principles of severability likely would

lead to fewer facial challenges. Fallon observes that some of the Court’s
tolerance for facial challenges reflects the fact that “the Court’s practices
in treating severability as a bar to declarations of facial invalidity fail to
conform to consistent rules.”154 By calling for more consistent
severability, except where provided by law, a return to first principles
would probably lead to a more consistent focus on as-applied challenges
as well.
2. National Injunctions
The first principles of severability also cast some light on the extensive
debates about national injunctions.155 These are cases where a court,
especially a lower court, forbids the federal government from enforcing a
statute or rule against anybody anywhere in the country. (The same kind
of remedy against state governments is conceptually similar,156 although
arguably distinguishable.157) Until 1939, such injunctions were unheard
of, and until recently, they were rare and unnatural.158
The principles of severability explain why nationwide injunctions were
so rarely thought necessary or pertinent. The consequence of a rule’s
unconstitutionality as to a particular plaintiff—even on broad
constitutional arguments—do not require the statute to be erased or
forcibly taken off the ledger. The normal consequence is simply that the
statute is not applied to those to whom it cannot be applied. And this
normal consequence does not call for anything but the normal remedy, of
enjoining whatever it is that threatens to unlawfully harm the plaintiff.
legislation” under the Fifteenth Amendment. 92 U.S. 214, 218, 221 (1875). Thanks again to
John Harrison for this point. Fallon argues that other enumerated powers cases should be seen
as overbreadth cases too. Fallon, supra note 148, at 945. For another version of this argument,
see Nicholas Quinn Rosenkranz, The Subjects of the Constitution, 62 Stan. L. Rev. 1209,
1273–88 (2010).
154
Fallon, supra note 148, at 921–22. Fallon does offer some insightful generalizations
about the Court’s behavior in such cases. Id. at 955–59.
155
See supra note 14.
156
Sohoni, supra note 14, at 975–77.
157
See Samuel Bray, A Response to The Lost History of the “Universal” Injunction, Yale
J. on Regul.: Notice & Comment (Oct. 6, 2019), https://www.yalejreg.com/nc/a-response-to-
the-lost-history-of-the-universal-injunction-by-samuel-bray/ [https://perma.cc/M43G-UBK
Y].
158
Bray and Sohoni tangle over whether the date is 1913, see Sohoni, supra note 14, at 943,
or 1939, see Bray, supra note 157.
The principles of severability also shed light on the question of whether

and how Congress can authorize nationwide injunctions. Congress can
enact fallback law, including rules that say “if this rule is found to be
unenforceable in any particular case, it should be found unenforceable in
all cases.” In other words, Congress can give federal judges the de facto
power to completely invalidate a rule nationwide, as a consequence of a
particular plaintiff’s claim. Congress can legislate something other than
the normal rules.
Indeed, many maintain that the Administrative Procedure Act (“APA”)
has done exactly that. The Act says “[t]he reviewing court shall . . . hold
unlawful and set aside agency action,”159 and some have relied on that
language to say that nationwide injunctions are permissible under the
APA, whatever may be true of them more generally.160 A few revisionists
dispute this interpretation of the APA.161 But my point is that it is possible.
Even if national injunctions are contrary to general principles, that does
not preclude Congress from creating fallback law that creates a similar
effect.
3. Contract Law
These first principles may also have parallels in contract law. Under
current doctrine, the severability of contract terms tracks the severability
of statutes. If a term of a contract is illegal, it will generally be severed
unless it is “an essential part of the agreed exchange.”162 Courts use the
parties’ intent to determine if the part was essential.163 This parallels the
rebuttable presumption of statutory severability.
If the same first principles of severability applied to contract law, they
might refine the doctrine in a similar way. And indeed, Omri Ben-Shahar
159
5 U.S.C. § 706.
160
Nat’l Mining Ass’n v. U.S. Army Corps of Eng’rs, 145 F.3d 1399, 1409–10 (D.C. Cir.
1998).
161
See, e.g., John Harrison, Section 706 of the Administrative Procedure Act Does Not Call
for Universal Injunctions or Other Universal Remedies, Yale J. on Regul.: Bull. (Apr. 12,
2020), https://www.yalejreg.com/bulletin/section-706-of-the-administrative-procedure-act-
does-not-call-for-universal-injunctions-or-other-universal-remedies/ [https://perma.cc/RC8W
-PF3K]; see also Bray, supra note 14, at 438 n.121, 454 n.220 (arguing that the APA did not
call for national injunctions). But see Mila Sohoni, The Power to Vacate a Rule, 88 Geo. Wash.
L. Rev. 1121, 1126 (2020) (arguing that the APA does allow universal injunctions).
162
Movsesian, supra note 76, at 48; see also 8 Samuel Williston & Richard A. Lord, A
Treatise on the Law of Contracts § 19:70 (4th ed. 1998) (describing severance).
163
Movsesian, supra note 76, at 48.
has proposed this, arguing that courts should respond to unconscionable

contract terms by enforcing the “minimally tolerable term.”164 As he
writes:
[I]f a term is considered unfair, it can be broken down to two distinct
components: the allowable portion, and everything beyond it. Once the
second component—the excessive increment—is eliminated, the
remainder is no longer unfair or unconscionable (even if still relatively
one-sided), and does not necessitate further intervention. This
remainder—the minimally tolerable term—would be enforceable.165
This happens to parallel the stronger rule of statutory severability
produced by first principles.
That said, there are arguments for different first principles for contracts
than statutes. First, contract law focuses more centrally on the parties’
intent, for reasons that will likely be familiar. Statutes bind third parties
who are not party to the bargain and are made by collective entities whose
intent is hard to find.166 Second, different kinds of law govern contracts.
In a public law case, the effect of a statute is governed by a combination
of constitutional law and the general law of statutory interpretation.167 In
a contract law case, external statutory provisions and common law rules
govern its validity and interpretation.168
These interpretative doctrines, and severability doctrines, have a
similar formal structure at an abstract level: higher law displaces certain
parts of a legal instrument (i.e., says what the law is not), and other law
determines what takes effect in light of that displacement (i.e., says what
the law is). But the substance of the first principles of contract severability
is ultimately beyond the scope of this paper.
164
Omri Ben-Shahar, Fixing Unfair Contracts, 63 Stan. L. Rev. 869, 876–78 (2011).
165
Id. at 877–78.
166
Movsesian, supra note 76, at 69–71.
167
Baude & Sachs, supra note 85, at 1099. This means that, in principle, a jurisdiction could
develop some general law principles of severability that would supplement or replace some of
those described here, see generally Ryan Scoville, The New General Common Law of
Severability, 91 Tex. L. Rev. 543 (2013) (arguing that the Supreme Court has created such a
general common law), but I am skeptical that our federal system has done so, apart perhaps
from the special-purpose canons described infra at Subsection III.B.1.
168
Baude & Sachs, supra note 85, at 1083, 1094–95 (noting parallels to contract law).
III. DIFFICULTIES
To be sure, not all cases are as easy as California v. Texas.169
A. Ultra Vires
The revisionist view of severability has been most prominent in recent
separation of powers cases. When faced with an executive officer whose
tenure or independence is inconsistent with the Court’s interpretations of
Article II, what is to be done? Justices Gorsuch and Thomas have rejected
conventional severability analysis in ways that seem both intuitive and
arresting. But upon further investigation, some of these intuitions do not
pan out.
In Seila Law v. CFPB,170 the head of the Consumer Financial
Protection Bureau served a five-year term during which he could be
removed—said the statute—only “for inefficiency, neglect of duty, or
malfeasance in office.”171 The Court held this restriction unconstitutional
and also severable.172 Justice Thomas, joined by Justice Gorsuch, rejected
severability and wrote that they “would simply deny the Consumer
Financial Protection Bureau (CFPB) petition to enforce the civil
investigative demand.”173 They elaborated:
As the Court recognizes, the enforcement of a civil investigative
demand by an official with unconstitutional removal protection injures
Seila. Presented with an enforcement request from an
unconstitutionally insulated Director, I would simply deny the CFPB’s
petition for an order of enforcement. This approach would resolve the
dispute before us without addressing the issue of severability.174
This may seem intuitive enough, but the next year the two revisionists
suddenly divided among themselves over a seemingly identical question.
In Collins v. Yellen,175 the Court dealt with the head of the Federal
Housing Finance Agency (“FHFA”), whose governing statute claimed to
impose a similarly impermissible tenure in office.176 Though the majority
169
Supra notes 123–43 and accompanying text.
170
140 S. Ct. 2183 (2020).
171
12 U.S.C. § 5491(c)(1), (3).
172
Seila Law, 140 S. Ct. at 2192.
173
Id. at 2219 (Thomas, J., concurring in part and dissenting in part).
174
Id. at 2220 (citation omitted).
175
141 S. Ct. 1761 (2021).
176
12 U.S.C. § 4512(b)(2).
opinion did not talk explicitly of severability, it seemed to presume the

provisions severable once again, noting the lack of a connection between
the removal restriction and the challenged actions of the agency.177
But this time, Justice Thomas abandoned Justice Gorsuch. Justice
Gorsuch reiterated their shared position from Seila Law, that “a court
would normally set aside the Director’s ultra vires actions.”178 But Justice
Thomas took a step back. He agreed that the removal provision was
unconstitutional, but worried “that the Court and the parties have glossed
over a fundamental problem with removal-restriction cases such as these:
The Government does not necessarily act unlawfully even if a removal
restriction is unlawful in the abstract.”179
Indeed, in a fundamental sense the constitutional problem in Collins
was self-correcting: the statute purporting to insulate the director was
unconstitutional; it could not be enforced, and it never had been enforced.
So, “the President has always had the power to fire the Director for any
reason.”180 Hence, the unlawfulness of the statute did not imply the
unlawfulness of the director’s actions. When Justice Gorsuch complained
that Justice Thomas had thought otherwise in Seila Law,181 Justice
Thomas responded that the point had been conceded there182 so he hadn’t
focused on it.
This split between Justices Thomas and Gorsuch mirrored their split
two days earlier in United States v. Arthrex.183 Again, the Court had found
a separation of powers problem in agency structure—this time, the
excessive authority assigned to administrative patent judges, who had
been appointed as “inferior officers” under Article II.
And again, the two revisionists divided. Justice Gorsuch would have
taken the “traditional path” of “‘setting aside’ the PTAB decision in this
case.”184 Justice Thomas dissented on the merits, so he did not explicitly
address severability. But one passage in his dissent resembled his and the
majority’s analysis in Collins:
177
Collins, 141 S. Ct. at 1788.
178
Id. at 1795 (Gorsuch, J., concurring in part).
179
Id. at 1789 (Thomas, J., concurring).
180
Id. at 1793.
181
Id. at 1798 n.2 (Gorsuch, J., concurring in part).
182
Id. at 1793 n.5 (Thomas, J., concurring).
183
141 S. Ct. 1970 (2021).
184
Id. at 1990–91 (Gorsuch, J., concurring in part and dissenting in part).
If we accept as true the Court’s position that the Appointments Clause

inherently grants the Director power to reverse Board decisions, then
another problem arises: No constitutional violation has occurred in this
suit. The Board had the power to decide and lawfully did decide the
dispute before it. The Board did not misinterpret its statutory authority
or try to prevent direct review by the Director. Nor did the Director
wrongfully decline to rehear the Board’s decision. Moreover, Arthrex
has not argued that it sought review by the Director. So to the extent
“the source of the constitutional violation is the restraint on the review
authority of the Director,” his review was not constrained. Without any
constitutional violation in this suit to correct, one wonders how the
Court has the power to issue a remedy.185
In these cases, it became clear that the revisionist account of
severability is incomplete. Without more, it does not tell us what to do in
cases like these: nor does it even tell us what exactly is the right category
to focus on.
Justice Gorsuch’s argument is that courts would traditionally disregard
ultra vires actions, and that the executive actions in Seila, Collins, and
Arthrex, are ultra vires in light of the constitutional issues. It sounds
formalist, not only because it is Latin. And let us start by conceding that
there is a core insight to this view. Though he did not cite it, Justice
Gorsuch might have derived force from the precedent of United States v.
Yale Todd.186 That is the case, recall, where the Supreme Court
disregarded the adjudications made by judge/commissioners under a
constitutionally flawed veterans’ relief statute. The best explanation
seems to be the ultra vires principle in action. The statute could not confer
non-judicial power on judicial officers, so their adjudications were
beyond their power. This is the traditional approach that Justice Gorsuch
would apply in other separation of powers cases.
But the more recent cases are different in a crucial way. Yale Todd
featured a lack of constitutional authority. The same logic would carry
over to challenges under the Appointments Clause.187 But Seila Law and
185
Id. at 2006 (Thomas, J., dissenting) (citation omitted).
186
See supra notes 46–54 and accompanying text.
187
Indeed, Yale Todd probably should have been mentioned to the parties who were arguing
about the “de facto officer doctrine” in the recent appointments case of Financial Oversight
and Management Board for Puerto Rico v. Aurelius Investment, 140 S. Ct. 1649, 1656 (2020).
Collins, at least,188 did not present the same kind of problem. The directors
of the CFPB and the FHFA were validly appointed under Article II, and
validly vested with executive power, which was the kind of power they
were exercising. The problem was that the statute purported to limit the
President’s power to remove them.189 We must stretch the ultra vires
doctrine some length to get it to cover removals.
Now, Justice Gorsuch did try to stretch it. Justice Gorsuch argued that
there was no historical precedent for a distinction between appointment
and removal.190 But as John Manning has observed, until the 1980s, “The
Court's leading removal cases merely involved claims for back pay.”191
And Justice Gorsuch argued that “[i]f anything, removal restrictions may
be a greater constitutional evil than appointment defects.”192 But the
question is not whether a constitutional defect is “great” but rather
whether it goes to the authority of the officer. So, while Justice Gorsuch’s
approach has a superficial formalist appeal, it glossed over the “vires” in
ultra vires.
Justice Gorsuch also may have committed the first half of the writ-of-
erasure fallacy. That fallacy is to think that judges have the power to strike
down or erase unconstitutional statutes.193 It is a fallacy because in truth
it is the Constitution that makes unconstitutional statutes irrelevant, a fact
judges simply recognize. That means that an unconstitutional removal
restriction is irrelevant and should be treated as such by all in the
executive and judicial branch. Yet Justice Gorsuch would have treated the
unconstitutional statutory provision as quite relevant, indeed he would
188
Arthrex is a more ambiguous case. It is not clear whether the problem was one of
appointment, and hence of authority, or one of supervision, which is more analogous to
removal. The majority refused to say. United States v. Arthrex, Inc., 141 S. Ct. 1970, 1985
(2021) (“The principal dissent repeatedly charges that we never say whether APJs are principal
officers who were not appointed in the manner required by the Appointments Clause, or
instead inferior officers exceeding the permissible scope of their duties under that Clause. But
both formulations describe the same constitutional violation: Only an officer properly
appointed to a principal office may issue a final decision binding the Executive Branch in the
proceeding before us.” (citation omitted)). Presumably the Court did not think it mattered, but
this line of thinking shows why it did.
189
Collins, 141 S. Ct. at 1787 (stressing this point).
190
Id. at 1795–96 (Gorsuch, J., concurring in part).
191
John F. Manning, The Independent Counsel Statute: Reading “Good Cause” in Light of
Article II, 83 Minn. L. Rev. 1285, 1293 n.35 (1999) (first citing Wiener v. United States, 357
U.S. 349, 349–50 (1958); then citing Humphrey’s Ex’r v. United States, 195 U.S. 602, 612
(1935); and then citing Myers v. United States, 272 U.S. 52, 106–07 (1926)).
192
Collins, 141 S. Ct. at 1796 (Gorsuch, J., concurring in part).
193
Mitchell, supra note 6, at 937.
have effectively given it great force by holding that its presence—even if

ignored in practice—precluded the enforcement of other provisions of
law.
Another way to see this mistake: Suppose Congress were to repeal the
statutory section that deals with tenure protection. Then surely the director
would be allowed to carry on with no problem. Nobody would call his
actions ultra vires, because the unconstitutional statute would be “gone.”
Now suppose we were to instead repeal the tenure protection by
constitutional amendment, thanks to some modern groundswell of
support for executive supervision. Surely the director would also be
allowed to carry on with no problem. The fact that the repeal was by
amendment rather than statute would not weaken it. Now suppose that we
already have adopted a constitutional provision that repeals the
unconstitutional tenure protection—which we have, Article II. That
should produce the same outcome, because the Constitution displaces
inconsistent statutes regardless of their relative dates.
An unconstitutional statute is void. An unconstitutional removal
restriction is therefore void. Nobody should apply it, nobody should
enforce it, and if nobody does,194 all is right with the legal world. That is
really what the Court said in Collins, and that is all fine and good.
But there is another way to look at the problem that has been splitting
Justice Gorsuch and Justice Thomas. Unfortunately, that problem is
ubiquitous, and it is not amenable to an easy solution. But until we
recognize it, we will be doomed to worse confusion. It is the problem of
unconstitutional combinations.
B. Combinations
It is easy enough to say the Constitution displaces unconstitutional laws
and requires the others to be enforced. But sometimes a law is
194
To be sure, the analysis is more complicated if an executive officer does give effect to
an unconstitutional statute. But this calls for a nuanced remedial approach. In Collins v. Yellen,
the Court remanded for examination of whether the unconstitutional statutory provision had
somehow been given legal effect that injured the plaintiffs. 141 S. Ct. at 1788–89. And John
Harrison had argued that the validity and consequences of this kind of executive action should
be addressed through the Administrative Procedure Act. See Brief for Professor John Harrison
as Amicus Curiae in Support of Respondents, at 4, 30, Collins v. Yellen, 141 S. Ct. 1761
(2021) (No. 19422); see also Collins, 141 S. Ct. at 1794 n.7 (Thomas, J., concurring)
(contemplating this possibility as “colorable,” but also pointing out that “we would need to
consider the interaction between this statutory claim and the [FHFA’s] anti-injunction
provision”).
unconstitutional only because it is combined in a particular situation with

another law. In these cases, it is obvious that the easy saying is
incomplete. Which law is to be displaced, and why?
Consider, for instance, Free Enterprise Fund v. Public Company
Accounting Oversight Board (“PCAOB”),195 where the Supreme Court
held that it was unconstitutional for an agency to be insulated from
presidential control by two layers of protection. The PCAOB could be
removed for cause by the SEC, whose members could be removed for
cause by the President. One such layer, between the President and the
SEC, was thought to be fine.196 One layer between the SEC and the
PCAOB would also have been fine.197 But two layers—from the President
to the SEC, then from the SEC to the PCAOB—“contravene[d] the
President’s ‘constitutional obligation to ensure the faithful execution of
the laws.’” 198 In such a case, the Court must say more to explain which
layer will be disregarded as repugnant.
These combinations problems generally take something like this form.
There are Statutory Requirement A, Statutory Requirement B, and a
Constitutional Requirement. Any two of these can be enforced.
Requirements A and B would work fine together were it not for the
Constitution. The Constitution and A can be enforced, but not B. Or vice
versa. Moreover, we know from basic principles of constitutional
supremacy that the Constitution must be enforced. So, the question
remains what to make of A and B.
The problem may seem quirky, but it is a recurring one. In Arthrex, the
separation of powers problem was a combination of the way that the
administrative patent judges were appointed, the significance of the
power they were given, and the lack of control of that power by superior
officers.199 Seila Law and Collins, discussed earlier, are combinations
problems too. The statutes there did two things—vest executive power in
an appointed official, and tell the President there were limits on his ability
to remove that official. Either of these things alone is permissible. Vesting
executive power in removable officials is okay. Limiting the power to
195
561 U.S. 477 (2010).
196
Id. at 483 (citing Humphrey’s Ex’r v. United States, 295 U.S. 602 (1934)).
197
Id. (citing United States v. Perkins, 116 U.S. 483 (1886); and then citing Morrison v.
Olson, 487 U.S. 654 (1988)).
198
Id. at 484 (quoting Morrison, 487 U.S. at 693). It was also “contrary to Article II’s vesting
of the executive power in the President.” Id. at 484.
199
United States v. Arthrex, Inc., 141 S. Ct. 1970, 1985–86 (2021).
remove non-executive officials is okay. But not both together. The

disagreement over severability in those cases can be seen as simply
another application of the combinations problem. Some Justices thought
that it was the removal restriction that must be ignored, allowing the
official to exercise enforcement authority if it was. Other Justices thought
that it was the enforcement authority which must be nullified in light of
the removal restriction.
A different setting for the combinations problem came in in Barr v.
American Ass’n of Political Consultants, another recent severability case.
There the Court concluded that the First Amendment forbade a
combination of two rules: a general ban on robocalls, and a permission
for robocalls for government-backed debt.200 Indeed, as AAPC reminded
many lawyers, there is a whole class of cases dealing with the question of
whether to “level up” or “level down” when there is a constitutionally
impermissible discrimination between two classes of persons or
activity.201 All of these level up/down cases are unconstitutional
combinations: the higher level rule for one class, and the lower level rule
for the other.202
The examples do not stop there. Shelby County v. Holder dealt with the
preclearance requirements of the Voting Rights Act, whose coverage
formula was not adequately justified.203 This might seem like a standalone
constitutional problem. But the problem also partly came from adjacent
provisions, such as the limited ability to add and remove jurisdictions
from coverage based on new developments. The Solicitor General argued
that these adjacent provisions were enough to save the statute,204 and even
if they weren’t, that suggests that if the Court had instead said that the
Constitution required more vigorous “bail in” and “bail out,” the
preclearance formula could have been saved. Indeed, we have evidence
200
140 S. Ct. 2335, 2343–44 (2020).
201
Id. at 2354–55 (plurality).
202
Lea, supra note 9, at 776–77, has a very good discussion of these problems but with
different terminology: he uses the term “statutory convergences” to capture this set of cases,
using “combinations” only for the subset that excludes some of the antidiscrimination cases.
Calling them all “combinations” is more straightforward.
203
570 U.S. 529, 550–51 (2013). For discussion of the merits, see William Baude, The Real
Enemies of Democracy, 109 Calif. L. Rev. 2407, 2414–15 (2021).
204
Transcript of Oral Argument at 34–35, 52, Shelby County, 570 U.S. 529 (No. 12-96); see
also id. at 52 (“General Verrilli: . . . if the tailoring mechanism doesn’t work, then jurisdictions
that could make such a claim may well have an as-applied challenge.”).
of this: in a previous case, the Court had broadened the “bail out”
provision to help avoid holding it unconstitutional.205
Recognizing the ubiquity of the unconstitutional combinations
problem is clarifying. But it is also daunting. For at this point, the simple
model of repugnancy and enforcement seizes up: Courts should disregard
the two statutory provisions because they are unconstitutional. But once
both provisions are disregarded there is actually no need to disregard one
of them, because it is permissible on its own. Or alternatively there is no
need to disregard the other, because it is permissible on its own. So, on
what warrant can courts disregard the first rather than the second, or the
second rather than the first?
Thus, the problem of combinations drives home the way in which
simple formalist accounts of severability are incomplete.206
Unconstitutional combinations, which have always been possible,
highlight that when the Constitution tells us what the law isn’t, it does not
always tell us enough about what it is. This is most obvious in cases like
Arthrex or Free Enterprise Fund, but it is an instance of the general point
that what the sub-constitutional law is depends at least in part on the sub-
constitutional law. That’s always true, and easier cases involving partial
unconstitutionality just obscure the point because it’s so clear what the
sub-constitutional law is in those cases. Even seemingly easy cases of
severability actually rest on the conclusion (usually implicit) that no
separately constitutional rule or application is dependent on a separately
unconstitutional rule or application.207
With all of the difficulties in mind, let us focus on possible solutions.
1. Fallback Law Solutions

If the legal system is complete and well-functioning, some principle of
law will tell us what the law is in combinations cases. Congress could
enact a statute simply telling us whether it would prefer Requirement A
205
Nw. Austin Mun. Util. Dist. No. 1 v. Holder, 557 U.S. 193 (2009); see also Justin Levitt,
Section 5 As Simulacrum, 123 Yale L.J. Online 151, 156 n.16 (2013) (“It is theoretically
possible that the bailout criteria are unconstitutionally stringent . . . . But . . . if the bailout
provision is not working as intended—as a symmetric counterpart to coverage—the
congressional purpose can be better effectuated by construing the bailout provision than by
discarding the entire statutory scheme.”).
206
Fish goes so far as to use the combinations problem to argue that formalist accounts of
severability, and the view that severability is a matter of interpretation rather than remedy, are
“untenable.” Fish, supra note 9, at 330. As this section shows, that goes too far.
207
Thanks again to John Harrison for all of this.
to prevail over Requirement B, if only one can prevail. But despite the
ubiquity of severability clauses, they do not answer this question. Still,
there are at least three other potential sources that could provide a legal
answer to combinations problems—congressional intent, special-purpose
canons, and general-purpose canons. None of these are simple to use and
to justify. But they are the best bets we have before we turn to more
incomplete solutions.
a. Hypothetical congressional intent

As noted above, hypothetical congressional intent is part of black-letter
severability doctrine today. And in some ways the combinations cases
prove its usefulness. Despite the many valid complaints about the use of
hypothetical congressional intent, it is no worse at solving the problem of
unconstitutional combinations than at solving any other severability
problem. We simply try to ask which statutory provision Congress would
have preferred to keep if it knew it could only have one.
Of course, as noted, more formalist approaches to interpretation tend
to reject hypothetical congressional intent in interpretation. The
combinations problem puts pressure on that tendency. Is legislative intent
a categorically forbidden source in statutory interpretation, or is it simply
a disfavored source compared to enacted text? If it is categorically
forbidden, formalists will have to turn elsewhere. But if it is simply
disfavored, then perhaps formalists can turn to it in a pinch.
Some formalists are more categorical than others. There are some
accounts of formalism that seem to categorically reject legislative intent.
For instance, if one takes the view that it is theoretically or practically
impossible for judges to determine collective intent, then it can’t be used
to solve the combinations problem, no matter how useful it would be.208
In Frank Easterbrook’s memorable turn of phrase, if a judge picks up a
statute in search of fallback law and cannot find it, maybe he can do
nothing more than to “put it down.”209
208
Frank H. Easterbrook, Statutes’ Domains, 50 U. Chi. L. Rev. 533, 547–48 (1983)
[hereinafter Easterbrook, Statutes’ Domains]; Frank H. Easterbrook, The Absence of Method
in Statutory Interpretation, 84 U. Chi. L. Rev. 81, 82 (2017); John F. Manning, Textualism
and Legislative Intent, 91 Va. L. Rev. 419, 430 (2005) (“[T]extualists deny that a legislature
has any shared intention that lies behind but differs from the reasonable import of the words
adopted; that is, they think it impossible to tell how the body as a whole actually intended (or,
more accurately, would have intended) to resolve a policy question not clearly or satisfactorily
settled by the text.”).
209
Easterbrook, Statutes’ Domains, supra note 208, at 535.
But many other formalists allow legislative intent to play some role in
interpretation. Some of them go so far as to permit the consideration of
legislative history—that is, evidence of subjective legislative intent—so
long as we understand that it is strongly outranked by the text.210 Other
formalists go less far, but still farther than Judge Easterbrook might seem
to. They restrict themselves to objective legislative intent. For instance,
formalists have defended the use of fictionalized legislative intent,211
elements of context such as the mischief rule,212 or other objective
versions of collective intent.213
Formalists of any of these stripes could use subjective or objective
evidence of legislative intent to decide which provision in an
unconstitutional combination should be enforced. At the same time, in
doing so, formalists must be mindful of their own arguments about the
error costs of judicial inquiries into legislative intent. Judges do not
become any better at non-textual analysis just because they have no other
choice.
b. Special-purpose canons
Another possibility is the use of substantive canons of construction. In
normal statutory interpretation, canons are a frequent tactic to deal with
textual ambiguities instead of falling back to legislative history or other
kinds of intent-based analysis. And while the legitimacy of this tactic is
well debated,214 one can see it on display in some of the combinations
cases the Court has confronted.
The best example is AAPC, where the Court confronted the
combination of the 1996 ban on robocalls and a 2015 permission for
robocalls for government-backed debt.215 A majority of the Justices
210
Caleb Nelson, What is Textualism?, 91 Va. L. Rev. 347, 360–62 (2005); see also id. at
405 (observing, though in 2005, that “courts conducting severability analysis routinely have
to speculate about how the enacting Congress would have answered a question that it did not
actually face. Textualist judges regularly join opinions taking this approach, and they have
voiced no fundamental objection to it.”).
211
Ryan D. Doerfler, Who Cares How Congress Really Works?, 66 Duke L.J. 979, 1020–
31 (2017).
212
Samuel L. Bray, The Mischief Rule, 109 Geo. L.J. 967 (2021).
213
Richard Ekins, The Nature of Legislative Intent 9–10 (2012). Though Manning, supra
note 208, at 425, does not deny the construct of objective legislative intent it is not clear to me
whether he would countenance these kinds of inquiries.
214
See, e.g., Baude & Sachs, supra note 85, at 1121–28; Ryan D. Doerfler, Late-Stage
Textualism, 2021 Sup. Ct. Rev. 267, 269 (2022).
215
Barr v. AAPC, 140 S. Ct. 2335, 2352–56 (2020).
concluded that the 1996 ban should be enforced while the 2015
permission disregarded, and a plurality invoked two related principles
from the Court’s severability cases. One was a principle that where there
is a constitutionally permissible rule that is rendered unconstitutional
because of an amendment, the amendment should be junked.216 The other
was a principle that where there is a rule and an exception that are
unconstitutional together, the rule should be retained rather than the
exception broadened.217
The plurality called these “general severability principles.”218 They are
effectively canons. They are recurring statements about how to deal with
severability in a range of contexts that could be found in previous cases.
They do not particularly follow from more abstract principles of
severability, and the plurality did not really derive them from any source
of law other than the practice of the cases. Indeed, in response to the
revisionist dissent by Justice Gorsuch, the plurality agreed that “there is
no magic solution to severability that solves every conundrum” but called
its “current approach as reflected in recent cases . . . constitutional, stable,
predictable, and commonsensical.”219
Of course, perhaps that statement doth protest too much. For instance,
there is an alternative framing available, which would have called for an
opposite canon. In other equality cases, one often frames the problem as
a choice between leveling up or leveling down, where the general rule
seems to be level up rather than to level down.220 The plurality’s use of
the rule/exception and amendment canons made the level up rule more
problematic.221 So here as elsewhere, the use of canons gives rise to a
debate about which canons are the most fitting. And it does not free us
from the more general debate about where these canons come from and
what warrant the Court has for treating them as law.222
216
Id. at 2353–54; see also James Durling & E. Garrett West, Severing Unconstitutional
Amendments, 86 U. Chi. L. Rev. Online 1, 3 (2018) (defending this principle).
217
AAPC, 140 S. Ct. at 2354–55.
218
Id. at 2349.
219
Id. at 2356.
220
See Caminker, supra note 9, at 1186–90; Fish, supra note 9, at 349.
221
Louis Michael Seidman, The Ratchet Wreck: Equality’s Leveling Down Problem, 110
Ky. L.J. 59, 77–79 (2021) (describing AAPC’s impact on the “level up” presumption).
222
See Amy Coney Barrett, Substantive Canons and Faithful Agency, 90 B.U. L. Rev. 109
(2010).
c. General-purpose canons
AAPC demonstrated the ability of special-purpose, severability-
specific canons to provide at least a semblance of fallback law. But it
would be equally possible, and perhaps more legitimate, to make use of
the most relevant general-purpose canons of construction. This would
jive more with the treatment of severability questions as ordinary
questions of law. But it would also lead to some disorienting outcomes.
For instance, couldn’t AAPC have been solved through application of
the last-in-time rule? The Court was faced with a statute enacted in 1996
and an amendment adopted in 2015. The ordinary last-in-time rule says
that if one can’t have both, one is supposed to prefer the one adopted
later.223 That would mean that the permission for some robocalls
implicitly repealed the ban.
Note, though, that this is the opposite of the special-purpose canon the
Court adopted. And it would thus have led to the opposite result in the
Court’s mid-century cases that generated the canon as well.224 In Eberle
v. Michigan, the introduction of an exemption for local wine and cider
would have implicitly repealed Michigan’s dry-county law.225 In Truax v.
Corrigan, the state ban on labor injunctions would have implicitly banned
all injunctions.226 In Frost v. Corporation Commission, a special
regulatory exemption for cooperative corporations would have implicitly
repealed a broader permitting scheme.227
Now, one reason these results seem so implausible is that our law has
a very strong presumption against implied repeal. So, these implied
repeals are instinctively hard to swallow. Yet we may have to bite back
that instinct. The reason we have such a strong presumption against
implied repeal is that the two enactments should be harmonized if they
can be.228 And the premise of an unconstitutional combinations problem
is that the two enactments cannot be harmonized. It may well be that
Congress wanted them to be and thought they could be (in whatever sense
Congress wants and thinks), but we know that Congress was wrong about
that. So, our instincts about implied repeals may have to be set aside in
favor of the more fundamental principle—to which the presumption
223
See The Federalist No. 78 (Alexander Hamilton).
224
See AAPC, 140 S. Ct. at 2353 (noting and relying on these cases).
225
232 U.S. 700 (1914).
226
257 U.S. 312 (1921).
227
278 U.S. 515 (1929).
228
See Nelson, supra note 111, at 241.
against implied repeals is an exception—that more recent legislation

trumps earlier inconsistent legislation.
Another general-purpose canon we might turn to in other cases is the
canon of avoiding unconstitutionality.229 Perhaps one of the two
provisions that seems to produce the unconstitutional combination can be
“interpreted” into constitutional compliance. If one of the two provisions
is more ambiguous, hence more vulnerable to such interpretation, that
could provide a resolution.
Here, too, the canon might counsel the opposite of what the Court does
in practice. Consider Free Enterprise Fund again.230 The two layers of
removal protection that produced the constitutional problem were the rule
that members of the SEC could be removed only for cause, and that they
could remove members of the PCAOB only for cause. It turns out that
using the canon of avoidance could have neatly solved the combinations
problem. The second layer of protection, the PCAOB’s, appeared
explicitly in the statute.231 But the first layer of protection, the SEC’s, was
nowhere to be found!232
When this embarrassing fact emerged during argument of the case,
there was a fair response—that under interpretive presumptions in place
at the time, an agency like the SEC was simply presumed to have
protection from presidential removal.233 That is debatable on its own.234
But in light of the unconstitutionality of the two layers of removal
together, it would have been simple to say that any such presumption was
229
See generally Caleb Nelson, Avoiding Constitutional Questions Versus Avoiding
Unconstitutionality, 128 Harv. L. Rev. F. 331, 334 (2015) (describing this canon and
distinguishing it from the more novel canon “about avoiding constitutional questions”
(emphasis added)).
230
Free Enter. Fund v. PCAOB, 561 U.S. 477 (2010).
231
15 U.S.C. §§ 7211(e)(6), 7217(d)(3).
232
15 U.S.C. § 78d.
233
Wiener v. United States, 357 U.S. 349, 356 (1958) (inferring protection from removal
when statute was silent); see also Transcript of Oral Argument at 18–19, Free Enter. Fund,
561 U.S. 477 (No. 08-861) (petitioner’s counsel invoking Wiener).
234
See Free Enter. Fund, 561 U.S. at 546–47 (Breyer, J., dissenting) (arguing that because
the SEC was created during the nine-year period between the decisions of Myers and
Humphrey’s Executor, Congress would not have intended the SEC to be independent); Note,
The SEC Is Not an Independent Agency, 126 Harv. L. Rev. 781, 782 (2013) (same); Adrian
Vermeule, Conventions of Agency Independence, 113 Colum. L. Rev. 1163, 1219 (2013)
(“All this is defensible, if at all, only on the ground that the Court was implicitly recognizing
and incorporating by reference an extrajudicial convention about the independence of the
SEC.”).
rebutted by the need to construe the statute to avoid constitutional

problems.235
Perhaps the counterintuitive consequences of these canons explain why
nobody has been inclined to use them. But for those skeptical of special-
purpose severability canons and legislative intent, they provide another
possible source of established fallback law.
d. Final thought
The first principles of severability on their own simply do not commit
a judge to any one of the above solutions. All of them are consistent with
the premises that Justice Thomas, Justice Gorsuch, and others claim to
share. What will determine which solution is most appropriate is a judge’s
more specific interpretive commitments. Judges that find some form of
congressional intent analysis permissible may be able to use hypothetical
congressional intent. Judges that do not will likely use one of the canons-
based approaches, just as those judges use canons to avoid other
ambiguities and difficulties in text-based analysis.236 And the kind of
canons those judges will be willing to use will depend on their approach
to precedent and pedigree in using canons of interpretation. So, while the
first principles of severability on their own do not resolve these cases,
they do indicate the question of interpretive theory that does resolve them.
2. Non-Fallback Approaches
Still, what if a judge wishes to avoid committing to any of these forms
of fallback law? Perhaps a judge is simply unwilling to make use of
congressional intent or canons in saying what the law is. Or perhaps the
judge is simply not sure of the best interpretive approach and wishes to
find a way to avoid such seemingly abstruse questions. Is it possible to
handle combinations cases without a theory of fallback law?
Sometimes yes, sometimes no.
235
Free Enter. Fund, 561 U.S. at 548 (Breyer, J., dissenting) (“The Court then, by
assumption, reads into the statute books a ‘for cause removal’ phrase that does not appear in
the relevant statute and which Congress probably did not intend to write. And it does so in
order to strike down, not to uphold, another statute. This is not a statutory construction that
seeks to avoid a constitutional question, but its opposite.”).
236
See Doerfler, supra note 214, at 268.
a. . . . that do not work

Some judges’ have attempted to do so in ways that do not hold up to
scrutiny. These approaches, while well-intentioned, should be rejected.
i. Standing bootstrapping
One approach, championed by some judges, is to focus on the provision
that is necessary to redress plaintiff’s injury. This approach traces back to
a lower court opinion by then-Judge Scalia and was championed more
recently by two esteemed Fifth Circuit judges in the litigation that reached
the Supreme Court as Collins v. Yellen.237 As those judges put it: “Which
statute should the court refuse to apply when either one would be
constitutional in isolation? . . . [T]he statute that allegedly authorizes the
injury-in-fact that confers standing upon the plaintiff.”238 In Collins, those
judges concluded, this principle called for the invalidation of the
government’s enforcement action—thus reaching the same result that
Justice Gorsuch later reached on other grounds.
This approach is attractive on the surface. It focuses correctly on the
federal courts’ duty to decide the controversies before them and issue
judgments that resolve those controversies. But this tactic relies on an
unintentional sleight of hand. The posture is something like this: first,
conclude that plaintiff has standing; second, observe that if Provision A is
the unconstitutional one, plaintiff’s injury will be redressed, but if
Provision B is the unconstitutional one, plaintiff’s injury will not be
redressed; third, conclude that because plaintiff has standing, Provision A
must be the unconstitutional one.
But the steps here are in the wrong order. If the choice between
Provision A and Provision B will determine whether plaintiff has
standing, then one cannot conclude that the plaintiff has standing before
deciding which provision to choose. If courts proceed in this order, they
have at least two logical approaches: make the standing inquiry looser, in
which case it no longer provides a reason to pick Provision A over
Provision B; or conduct the strict standing inquiry more accurately, in
237
See Synar v. United States, 626 F. Supp. 1374, 1393 (D.D.C. 1986); Collins v. Mnuchin,
938 F.3d 553, 609–10 (5th Cir. 2019) (Oldham & Ho, JJ., concurring). It was also espoused
by Justice Thomas in Seila Law v. CFPB, 140 S. Ct. 2183, 2221 (2020) (Thomas, J.,
concurring in part and dissenting in part), before his different opinion in Collins v. Yellen, 141
S. Ct. 1761, 1789–99 (2021) (Thomas, J., concurring), the next year.
238
Collins, 938 F.3d at 609 (Oldham & Ho, JJ., concurring) (internal quotation marks
omitted).
which case a court must decide at the standing stage which provision is at
stake.
But what a court cannot logically do is breeze through the standing
problem without noticing the unconstitutional combinations problem,
then attempt to backfill the combinations problem by relying on an
implicit but arbitrary assumption made when breezing through the
standing problem. That is how Wile E. Coyote ended up stuck in mid-air
and then falling off a cliff.239
ii. Avoiding “editorial freedom”

On other occasions, courts have tried to avoid the question by simply
picking the solution that seemed simplest. To return to Free Enterprise
Fund, consider what the Court said in dealing with another related
combinations problem. Having concluded that the PCAOB was
unconstitutionally insulated from presidential control, the Court noted
that it had a lot of options for how to respond:
It is true that the language providing for good-cause removal is only
one of a number of statutory provisions that, working together, produce
a constitutional violation. In theory, perhaps, the Court might blue-
pencil a sufficient number of the Board’s responsibilities so that its
members would no longer be “Officers of the United States.” Or we
could restrict the Board’s enforcement powers, so that it would be a
purely recommendatory panel. Or the Board members could in future
be made removable by the President, for good cause or at will. But such
editorial freedom—far more extensive than our holding today—
belongs to the Legislature, not the Judiciary. Congress of course
remains free to pursue any of these options going forward.240
The problem the Court describes here prefigures the kind of problem the
Court later confronted in Arthrex, Seila, and Collins. But the opinion,
written by Chief Justice Roberts, tries to avoid that problem through a
generally minimalist approach. It rejects the idea that it might exercise
“editorial freedom” in deciding which provision to treat as invalid, also
invoked by refusing to “blue-pencil” the statute.241
239
The Canyon Fall Gag, Looney Tunes Wiki, https://looneytunes.fandom.com/wiki/The
_Canyon_Fall_Gag [https://perma.cc/8TVM-XV4Z] (last visited Sept. 19, 2022).
240
Free Enter. Fund, 561 U.S. at 509–10 (Roberts, C.J.).
241
Id.
But this minimalist attitude does not actually avoid the problem,
because it does not tell the Court which provision should be treated as
invalid. The Court is of course right to say it should act judicially, not
legislatively. But saying what the law is is judicial business. When two
statutory provisions cannot both be the law, saying which one is the law
is thus judicial business. Neither selection involves more “editorial
freedom” than the other.242 This is not to say that the majority’s selection
is wrong, but it would need to be justified on some actual basis.
By contrast, the Court’s approach in Arthrex—by the same author as
Free Enterprise Fund—is a marked improvement.243 There the Court
again resolved an unconstitutional combinations problem through a form
of minimalism—it held that one statutory section “cannot constitutionally
be enforced to the extent that its requirements prevent the Director from
reviewing final decisions rendered by APJs” rather than eliminating the
tenure protection of the APJs or eliminating entirely their ability exercise
executive power.244 But this time it justified this resolution of the
combinations problem by reference to things like the structure and
historical development of the statutory scheme.245 This is effectively
using the hypothetical congressional intent fallback law solution
described above. It is perfectly justifiable for any judge who is not a
hardcore textualist to do so, and the author was not one.246
b. . . . that work better

That said, in some combinations cases a judge might be able to avoid
thinking about fallback law.
i. Stipulation
There is widespread confusion about whether parties can “stipulate the
law,”—i.e., whether judges should assume that the law is X if both parties
agree that it is.247 But there is some support for the view that they can, and
242
See also Lea, supra note 9, at 781 n.231, and Mannheim, supra note 5, at 1858, agreeing
that this approach is mistaken. But see Ryan M. Folio, Constitutional Avoidance, Severability,
and a New Erie Moment, 42 Harv. J.L. & Pub. Pol’y 649, 679 (2019) (seemingly endorsing
it).
243
United States v. Arthrex, 141 S. Ct. 1970 (2021) (Roberts, C.J.).
244
Id. at 1987.
245
Id. at 1986–87.
246
See, e.g., King v. Burwell, 576 U.S. 473, 497–98 (2015).
247
See Gary Lawson, Stipulating the Law, 109 Mich. L. Rev. 1191, 1191 (2011).
to the extent that they can, a judge might use that stipulation to solve a
combinations problem.
In the end, that was how the Court tried to deal with the choice between
the two layers of removal in Free Enterprise Fund. It observed that
everybody agreed that the SEC was protected from removal, and implied
that everybody agreed that this limitation was constitutional.248 Similarly,
in many other cases like Shelby County v. Holder, the Court might have
justified its choice to focus on Provision A rather than Provision B simply
on the grounds that nobody was telling it to focus on Provision B.249
Of course, this approach will not always work. It depends on what the
parties argue and whether those arguments implicate both parts of the
combination. This approach also relies on the ability to stipulate the law,
which is itself contested. But when this approach does work, it may be
good enough to at least resolve the case at hand.
ii. Factual particularity

Another option is to focus on what has actually happened. As discussed
above, an unconstitutional statute should be disregarded to the extent of
its unconstitutionality. If it is disregarded, then it is no different than a
statute that has been subsequently repealed. But sometimes it has not been
ignored. Indeed, that is often why a plaintiff is in court complaining about
the statute rather than simply ignoring it. These basic facts provide a way
to resolve some combinations problems. If Provision A and Provision B
cannot both be enforced, and one of them has been or is, then that
provides good reason for a court to ignore the other one.
To see how this would work, consider the separation of powers cases
where an official cannot have both executive enforcement power and
insulation from removal, like Collins and Seila. By the time the Court
confronted those cases, one of those things had plainly happened—the
248
Free Enter. Fund, 561 U.S. at 487 (“The parties agree that the Commissioners cannot
themselves be removed by the President except under the Humphrey’s Executor standard of
‘inefficiency, neglect of duty, or malfeasance in office,’ and we decide the case with that
understanding.” (citations omitted)). It is not clear how firm these stipulations were. See
Transcript of Oral Argument at 21–22, Free Enter. Fund, 561 U.S. 477 (No. 08-861) (“Mr.
Carvin: . . . If this Court wants to say that—that those people are subject to the President’s
plenary—
Justice Scalia: I’d love to say that. That would be wonderful.
Mr. Carvin: I’m not going to stand in your way . . . .
Justice Scalia: This is not an argument you have made anyway.”).
249
But see the oral argument statements cited supra note 204.
officials had been vested with and exercised executive power. And the
other had arguably not—the officials had perhaps not been insulated from
removal, because the President had not tried to remove them, nor was it
clear whether they had done anything because of any (mistakenly)
anticipated insulation. That supported the Court’s choice to disregard the
unconstitutional tenure protection, not the exercise of executive power.250
Similarly, in a claim involving a challenge to past discrimination in the
awarding of some benefit, a Court might avoid the more general level-
up/level-down question. If the benefit has already been awarded to some,
and if the state is in no position to claw that benefit back from them
retroactively, then it has no choice but to “level up” at least with respect
to past benefits.251
This approach has the appeal of avoiding resort to controversial sources
of fallback law. But it does have some counterintuitive consequences. For
one thing, it may effectively empower the executive branch to elect which
of two statutes to enforce. For instance, imagine the executive branch had
tried to get out in front of the constitutional challenges to the Voting
Rights Act by arranging for bailout for many jurisdictions even if they did
not strictly comply with the statutory criteria. Doing so early enough and
often enough might have created facts on the ground that diffused the
challenge to the preclearance formula in Shelby County.252
This approach will not work in every case, and it is probably better for
judges to just try to figure out how to say what the law is in
unconstitutional combinations cases. But it has the great virtue of
allowing courts to focus on their judicial task, which is to decide cases by
applying all and only valid law and leaving matters outside the case for
others to worry about. And again, in many cases, it may be good enough
for government work.
250
See supra notes 170–82 and accompanying text.
251
See Lea, supra note 9, at 785–86 n.240; Ruth Bader Ginsburg, Address, Some Thoughts
on Judicial Authority to Repair Unconstitutional Legislation, 28 Clev. St. L. Rev. 301, 307
(1979). Of course, relief going forward would still pose the general question.
252
For instance, while Shelby County was pending, the Attorney General consented to the
bailout of the State of New Hampshire, despite imperfect compliance and over the objection
of an attempted intervenor. New Hampshire v. Holder, 293 F.R.D. 1, 3, 8 (D.D.C. 2013);
Consent Judgement and Decree, New Hampshire v. Holder, No. 1:12-CV-01854 (D.D.C. filed
Mar. 1, 2013). One amicus alleged that this was a nefarious attempt to “save[]” the Voting
Rights Act by “attempting to make bailout more widely available by bending the rules.” Brief
of Amicus Curiae the State of Alaska in Support of Petitioner Shelby County, Alabama, at
18–20, Shelby County v. Holder, 570 U.S. 529 (2013) (No. 12-96). If so, it was too little too
late.
C. Severability Procedures
The confusing nature of severability has also left courts confused about
how severability impacts procedure. When should courts address
severability? When do plaintiffs have standing based on inseverability?
When do they lack standing based on severability?
But once the proper view of severability is before us, the procedural
inquiries become fairly simple. Severability arguments are simply
arguments about what the law is. So, courts should refrain from
gratuitously opining about severability just as they do from gratuitously
opining about other law. Plaintiffs can premise their standing on legal
arguments that derive from fallback law, just as from other kinds of law.
And, if their legal arguments about severability are wrong, their claims
can fail for that reason.
In terms of gratuitous severability rulings, we should not simply expect
that Part I of a judicial opinion about a law’s constitutionality should be
followed by a Part II that addresses severability. That would be the natural
order of things if judges had a general power of constitutional-law-
opining, backed up with a remedy of severability. But they do not. Judges
opine about the constitutionality of various rules—“say what the law
is”—as needed to “apply the rule to particular cases.”253 Severability is
just another part of that saying what the law is, and therefore also
something courts should do only as needed to apply a statute to particular
cases.
Thus, Brian Lea and Justice Thomas are right to question the
occasional practice of “gratuitous severability rulings,” where the Court
assumed that after holding that a plaintiff was correct about a
constitutional claim that it should also go on to talk about the severability
of provisions that didn’t affect the plaintiff.254 This is just gratuitous
dictum and is no more justifiable than other forms of gratuitous dictum.255
At the same time, in other cases courts have been too reluctant to
confront severability questions integral to a plaintiff’s standing. In
California v. Texas, the plaintiffs tried to get the Supreme Court to say
that most of the Affordable Care Act was invalid because it was
inseverable from the unconstitutional individual mandate.256 The case
253
Marbury v. Madison, 5 U.S. (1 Cranch) 137, 177 (1803).
254
Lea, supra note 9, at 789–803; Murphy v. NCAA, 138 S. Ct. 1461, 1485 (2018) (Thomas,
J., concurring).
255
I take no position here on whether such dictum is unconstitutional, or simply bad practice.
256
See supra notes 131–39.
thus presented controversial questions of constitutionality (was the $0

mandate unconstitutional?), of severability (was the mandate indeed
inseverable from the rest of the Act?), and of standing (could the plaintiffs
raise this argument?). The Court resolved the case on standing grounds,
but standing might in fact have been the least vulnerable part of the
plaintiffs’ case.
The best argument for Texas’s standing was “bank-shot” standing—
that Texas was entitled to have an injunction against the plausible
enforcement of Provision A, if Provision B is invalid and inseverable.257
This may seem like a strange form of third-party standing, but if
inseverability is limited to fallback law it is actually unremarkable first-
party standing. The plaintiff is effectively saying that Congress has
instructed for Provision A not to be enforced if Condition X obtains, and
that Condition X obtains.258 Such a plaintiff has an orthodox legal injury,
an orthodox claim for why that injury is illegal, and an orthodox claim for
redress. A Court might be annoyed if the determination of Condition X
involves an important or awkward question and the current case feels too
unimportant to justifying answering it. But a judge’s duty is to answer the
questions necessary to apply the law to decide the cases before him, not
the questions he would like to answer.
This is not to say that the bank-shot argument should have succeeded.
The bank-shot theory of standing rested on the premise of inseverability,
and that premise was false.259 And because inseverability is a pure
question of law, it can be resolved at a very early stage of the litigation—
it would even be permissible to resolve it before considering the
constitutional merits argument. So even if an inseverability argument can
be used to produce standing, using a bad inseverability argument to
produce standing has little consequence. A plaintiff who uses a bad
inseverability claim to get into court and then lose has gained nothing
more than a plaintiff who invents a fictitious cause of action to enforce a
fictitious right. Perhaps the plaintiff has standing,260 but it is simply
standing to lose on the merits a few minutes later.
257
See Stephen Sachs, ACA: The Lay of the Land, Reason: Volokh Conspiracy (Nov. 13,
2020, 7:32 AM), https://reason.com/volokh/2020/11/13/aca-the-lay-of-the-land/ [https://per
ma.cc/JA2M-ZJL9].
258
Accord Lea, supra note 9, at 765–66.
259
See supra notes 136–39.
260
See Steel Co. v. Citizens for a Better Env’t, 523 U.S. 83, 89 (1998).
In the course of denying Texas’s claim to standing, the Court did not
fully address “bank-shot” standing. It treated the argument partly as
waived, and partly as a different kind of causation argument.261 Perhaps
that was for the best. It could well be that the majority was fractured, both
on whether to recognize bank-shot standing (which it should) and whether
the inseverability arguments were correct (which they were not). But if
the Court confronts the question again in cooler air, it should accept this
kind of argument if there is inseverability. Because real inseverability
clauses are so rare,262 however, perhaps it never will.
Thus, because inseverability is simply a claim about fallback law, it
can be an ingredient in a plaintiff’s theory of why a statutory provision
against him is unenforceable. For the same reason, a severability
argument can be an ingredient in defeating a plaintiff’s legal theory. If a
plaintiff’s only constitutional complaint is about statutory provisions that
did not injure him, and if those provisions are severable from the
provisions that did, his claim fails.263 That is why federal courts classes
properly recognize the conventional wisdom that a plaintiff may not
challenge the constitutionality of a statute that has not been applied to
him, absent some unusual circumstance like an inseverability
argument.264
Perhaps this principle needs to be dusted off to deal with some of the
recent separation of powers cases. As noted above, since the 1980s, the
Supreme Court has been willing to assume that a regulated party is injured
by the presence of an unconstitutional removal restriction. When the
restriction is actually relevant to a plaintiff’s case—if the president tried
and failed to protect the plaintiff from administrative overreach, or if the
261
California v. Texas, 141 S. Ct. 2104, 2116, 2119–20 (2021); see also id. at 2122 (Thomas,
J., concurring) (stating that the argument was not raised early or often enough and also
required a theory of statutory interpretation that plaintiffs did not propose). The bank-shot
theory was more clearly articulated in the Federal Government’s brief, see Brief for the
Federal Respondents at 16–22, California, 141 S. Ct. 2104 (Nos. 19-840 & 19-1019); see also
Blackman, supra note 139, at 133–35, but elements of the theory could also be found in
Texas’s, see Brief for Respondent/Cross-Petitioner States at 26–27, California, 141 S. Ct.
2104 (Nos. 19-840 & 19-1019).
262
See Gluck, supra note 61, at 159 & n.97 (canvasing the U.S. Code and Public Laws and
finding only nine inseverability clauses).
263
It is tempting to say that such a plaintiff has no standing, but it is more accurate to say
that the claim fails on the merits, see Lea, supra note 9, at 760.
264
See Richard H. Fallon, Jr., John F. Manning, Daniel J. Meltzer & David L. Shapiro, Hart
and Wechsler’s The Federal Courts and the Federal System 168–74 (7th ed. 2015)
(reproducing and discussing Yazoo & Miss. Valley R.R. v. Jackson Vinegar Co., 226 U.S.
217 (1912)).
restriction changed the decisions of the official265—this is quite plausible.

But in cases where the removal restriction has never come into play, this
is more akin to Yazoo & Mississippi Valley Railroad v. Jackson Vinegar
Co.266
The recent separation of powers severability decisions drive this home.
If the end results of Free Enterprise Fund, Seila Law, and Collins were
simply to expand the rights of the President, and to deny plaintiffs the
actual relief against enforcement that they sought, perhaps then each of
those suits should have been dismissed at an early stage.267
As a final point of emphasis and clarification, it is important to
remember that, ultimately, severability should drive standing rather than
the other way around. Severability is a question of the meaning of law. It
preexists the courts, courts do not have the power to change it, they are
just supposed to discern it and recognize it. Also, the executive branch
has to resolve questions of severability even outside the courts, as it too
must grapple with the consequences of statutes that have unconstitutional
applications.
This means that, in general, the answers to severability questions do
not depend on the standing of particular litigants, and its contours should
not be reshaped in order to produce a judge’s desired resolution of a
standing question. That is why standing is addressed here at the end of the
article. The fact that judges have to address it earlier in their cases should
not confuse them into letting it drive their views of the law.
CONCLUSION
The Constitution does not tell us everything we need to know about
severability. But it tells us two important things: that the task of a judge
is to apply law, and that the Constitution displaces ordinary law that is
repugnant to it. Ordinary law tells us the rest: when the Constitution does
not stand in the way, it tells us what the rest of the law is. And it tells us
265
See supra note 194 and accompanying text.
266
See Fallon et al., supra note 264, at 169–74 (discussing Jackson Vinegar Co.).
267
See Tyler B. Lindley, The Writ-of-Erasure Fallacy and the Balance of Powers, 17–19,
28–37 (Aug. 31, 2022) (unpublished manuscript) (on file with author); Tyler B. Lindley,
Justiciability and Remedies in Administrative Law Challenges, U. Chi. L. Rev. Online (Apr.
1, 2021), https://lawreviewblog.uchicago.edu/2021/04/01/lindley-justiciability/ [https://perm
a.cc/5NKL-JC6N]; cf. Aziz Z. Huq, Standing for the Structural Constitution, 99 Va. L. Rev.
1435, 1477–78 (2013) (questioning standing in Free Enterprise Fund v. PCAOB, 561 U.S.
477 (2010), and Morrison v. Olson, 487 U.S. 654 (1988)).
whether to apply any fallback rules in cases of unconstitutionality. Taken

together, those are the principles of severability.

Severability First Principles

Uploaded by

Copyright:

Available Formats

Severability First Principles

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Severability First Principles

Uploaded by

Copyright:

Available Formats

University of Chicago Law School

Severability First Principles

Follow this and additional works at: https://chicagounbound.uchicago.edu/journal_articles

Part of the Law Commons

VIRGINIA LAW REVIEW

SEVERABILITY FIRST PRINCIPLES

* Professor of Law, Faculty Director of the Constitutional Law Institute, University of

2 Virginia Law Review [Vol. 109:1

2023] Severability First Principles 3

b. . . . that work better ................................................ 53

4 Virginia Law Review [Vol. 109:1

2023] Severability First Principles 5

of an inseverability argument, and when can a severability argument

A. Severability Is a Question of What the Law Is

6 Virginia Law Review [Vol. 109:1

2023] Severability First Principles 7

Second, this orthodox picture is confirmed by the reality of judicial

8 Virginia Law Review [Vol. 109:1

B. Disregarding Repugnant Law

2023] Severability First Principles 9

act, they must examine the Constitution and refuse to do anything

10 Virginia Law Review [Vol. 109:1

Court opinions recognizing constitutional limits on federal statutes before

2023] Severability First Principles 11

reserved).36 Similarly, United States v. Percheman refused to read federal

12 Virginia Law Review [Vol. 109:1

2023] Severability First Principles 13

consequence, the United States was entitled to recover.52 The statute

14 Virginia Law Review [Vol. 109:1

different.57 But at the time, under the classic doctrine of avoiding

C. Applying Fallback Law

2023] Severability First Principles 15

important consequences for retroactivity, for executive adjudication, and

16 Virginia Law Review [Vol. 109:1

The other major category of possible constraints are seemingly

2023] Severability First Principles 17

threaten summary executions of the deciding judges or anybody else.72

1. The Presumption of Severability

18 Virginia Law Review [Vol. 109:1

2023] Severability First Principles 19

20 Virginia Law Review [Vol. 109:1

provisions precisely because it was binding law: it governed “this Act,”

2023] Severability First Principles 21

States v. Reese, an aggressively anti-severability decision that rejected a

22 Virginia Law Review [Vol. 109:1

judicial doctrine to mark the bounds of what is prohibited.93 But that

3. Fully Operative Law and Counterfactual Intent

2023] Severability First Principles 23

unconstitutional.101 The result was that an anti-commandeering problem

24 Virginia Law Review [Vol. 109:1

applying it anyway? If one provision is logically or linguistically

2023] Severability First Principles 25

26 Virginia Law Review [Vol. 109:1

B. Justices Gorsuch and Thomas

2023] Severability First Principles 27

Justice Thomas has repeated these views, often joined by Justice

28 Virginia Law Review [Vol. 109:1

2023] Severability First Principles 29

Thomas’s analysis of the ACA would surely have looked different.

30 Virginia Law Review [Vol. 109:1