The Incompatibility of Substantive Canons and Textualism

Introduction

When judges interpret statutes, they often invoke general rules or presumptions known as “canons of construction.” Commentators typically distinguish two main families of canons.¹ First, so-called “semantic” canons (also known as “linguistic” or “descriptive” canons) are generalizations about how particular linguistic constructions are used and understood by competent speakers of English.² According to the “last-antecedent canon,” for example, when some limiting phrase follows an item in a list, it usually modifies only that immediately preceding item, not others further upstream.³ In contrast, “substantive” canons (also termed “normative” or “policy-based” canons) are nonlinguistic considerations that weigh in favor of particular legal results. Examples include the rule that ambiguities in criminal statutes are resolved in favor of the defendant (the “rule of lenity”),⁴ the principle that courts should avoid interpreting statutes in ways that raise serious doubts about their constitutionality,⁵ and the presumption against interpreting federal statutes to intrude on the traditional prerogatives of state governments.⁶

Judges who embrace a “textualist” approach to statutory interpretation have little difficulty reconciling their interpretive philosophy with semantic canons, at least in principle. Roughly speaking, these “canons” are just shorthand labels for ordinary inferences drawn from linguistic common sense.⁷ But what about substantive canons? Whenever one of these canons does any work, it must be leading a court to a result different from the one that the same court would have reached based only on the apparent linguistic import of the text.⁸ But if, as textualists hold, a court’s job in a statutory case is “to apply, not amend, the work of the People’s representatives,”⁹ how can that ever be appropriate?

In a series of major cases over the past two years — starting with West Virginia v. EPA¹⁰ — Justice Kagan has repeatedly posed precisely this challenge to her textualist colleagues.¹¹ West Virginia concerned a provision of the Clean Air Act requiring the EPA to set emission limits that are “achievable through the application of the best system of emission reduction” that satisfies certain criteria.¹² In the Clean Power Plan, the EPA had determined that the “best system of emission reduction” involved not only efficiency improvements at individual coal plants, but also substitution of cleaner energy sources for coal.¹³ The EPA therefore calculated emissions limits that were premised on such “generation shifting” taking place.¹⁴ Justice Kagan (joined by Justices Breyer and Sotomayor) thought that the statutory text clearly allowed this approach.¹⁵ After all, she argued, “generation shifting fits comfortably within the conventional meaning of a ‘system of emission reduction,’” in light of the dictionary definition of “system,” the use of that term in other parts of the statute, and the Clean Air Act’s overall structure and design.¹⁶

But while the majority conceded that “[a]s a matter of ‘definitional possibilities,’ generation shifting can be described as a ‘system,’” it rejected that interpretation.¹⁷ The EPA was asserting an unprecedented authority to transform the U.S. electricity sector, Chief Justice Roberts explained, and its action thus triggered the “major questions doctrine” — a canon presuming that federal statutes do not delegate extraordinary regulatory authority to agencies except through a specific statement to that effect.¹⁸

To Justice Kagan, this amounted to little more than an opportunistic deviation from the textualist approach that the Justices in the majority usually preach. “The current Court,” she wrote, “is textualist only when being so suits it.”¹⁹ “When that method would frustrate [the Court’s] broader goals, special canons like the ‘major questions doctrine’ magically appear as get-out-of-text-free cards.”²⁰ And in the year since West Virginia, this indictment has become a refrain. When the Court interpreted the Clean Water Act narrowly in Sackett v. EPA²¹ — based, in part, on a substantive canon disfavoring interpretations that would “alter the balance between federal and state power and the power of the Government over private property”²² — Justice Kagan again charged the majority with “non-textualism.”²³ So, too, when the Court rejected the Biden Administration’s student loan forgiveness program in Biden v. Nebraska,²⁴ with the major questions doctrine again playing a role. Once more, the dissenters (led, again, by Justice Kagan) faulted the majority not just for getting the statutory interpretation question wrong, but for relying on a “made-up”²⁵ canon that “works not to better understand — but instead to trump — the scope of a legislative delegation.”²⁶

Justice Kagan is not the first to observe a tension between modern textualists’ foundational commitments and the use of substantive canons. Over the years, several leading textualist theorists have recognized and wrestled with this problem. Justice Scalia once observed that substantive canons amount to “dice-loading rules” that pose “a lot of trouble” for “the honest textualist.”²⁷ In an influential 2010 article, then-Professor Amy Coney Barrett similarly recognized that substantive canons are “at apparent odds with the central premise from which textualism proceeds.”²⁸ Yet, as the Court’s recent practice highlights, textualist judges have frequently deployed substantive canons anyway.²⁹ If anything, in fact, this tendency seems to have mounted with textualism’s rise in recent years. And while the major questions doctrine looms especially large at the moment,³⁰ the embrace of substantive canons by avowed textualists is not limited to that context (as the invocations of federalism and property-rights canons in cases such as Sackett confirm).³¹ For that matter, Justice Gorsuch has even urged his colleagues to breathe new life into the rule of lenity.³²

So what accounts for this disconnect? Justice Kagan’s dissenting opinions offer one answer: a too-fervent commitment to an ideological agenda (more politely, to “broader goals”³³ or “policy preferences”³⁴) that trumps the textualist Justices’ ordinary conception of their proper role, leading them to “shelve[] the usual rules of interpretation.”³⁵ Maybe so. But another important part of the explanation, we think, is textualist theorists’ longstanding insistence that while the tension here may be real, even awkward, it is ultimately manageable within their theory’s own terms. Indeed, after acknowledging the difficulty that we have described, Justice Scalia and then-Professor Barrett each went on to explain why, at the end of the day, it is probably not as acute as it appears.³⁶ Their arguments did not purport to save all substantive canons — and, until very recently, they did not make the leap from extrajudicial writings to judicial opinions. Still, the sheer existence of respectable and avowedly textualist defenses of substantive canons, together with a convenient vagueness about the scope of those defenses, has allowed textualist judges to deploy these canons with a relatively clean conscience. And now that the challenge to the integrity of this practice has been sharply posed, a new round of such defenses is emerging — including separate opinions by Justice Gorsuch³⁷ and Justice Barrett³⁸ that seem poised to play a similar legitimating role within and beyond the Court.

With the debate thus gaining steam, this Article systematically and closely assesses each of the leading efforts to square modern textualist theory with substantive canons. We take these proposed reconciliations to be of three essential types. The first line of argument — most closely associated with Justice Scalia, but now prominently deployed by Justice Barrett as well — recharacterizes at least some “substantive” canons as background conventions that a reasonable reader would consider in discerning what a lawmaker actually meant by the enacted text.³⁹ A second argument — developed most fully in then-Professor Barrett’s 2010 article and now embraced by Justice Gorsuch in his West Virginia concurrence — purports to reconcile substantive canons with textualism by grounding them in judges’ higher duty to “act as faithful agents of the Constitution.”⁴⁰ And a third approach — often paired with one of the other two, and endorsed by several of the Justices — suggests that textualist judges can legitimately rely on substantive canons if, but only if, the text itself provides only an ambiguous or uncertain answer.⁴¹

After considering each of these arguments, we conclude that substantive canons are generally just as incompatible with textualists’ jurisprudential commitments as they first appear. If we are right about that, then principled textualists face a choice. They could abandon or drastically curtail the use of substantive canons. Or they could concede that important claims about those jurisprudential commitments are inaccurate and oversimplified, in ways that undermine textualists’ traditional critiques of nontextualist interpretive methods.

The Article is organized as follows: Part I provides more rigorous characterizations of both “textualism” and “substantive canons,” and in so doing makes the prima facie case for their fundamental incompatibility. Parts II through IV, the heart of the Article, evaluate each of the three attempts at reconciliation and conclude that none is satisfactory. Part V offers some concluding thoughts on the different paths forward for textualism that we have just described.

I. The Problem

The basic tension between textualism and substantive canons is intuitive — sufficiently so that first-semester law students can grasp it and ask perceptive questions about how textualist judges can justify their reliance on these canons. But because both “textualism” and the idea of a “substantive” canon can prove quite slippery — and because our aim here is to make the nature and extent of the conflict inescapably clear, even to those who might be inclined to resist or dismiss it — we will start by nailing down each of these notions somewhat more precisely. As we will see, laying out the essential features of “textualism” and of “substantive canons” also lays bare the conceptual roots of the contradiction between them.⁴²

A. Textualism

1.  The Central Idea. — Although “modern textualism” resists easy definition, the beating heart of this interpretive philosophy has always been a normative thesis about the proper exercise of government power, at least under the U.S. Constitution.⁴³ That claim has both a legislature-facing aspect and a court-facing one. First, textualists believe that members of Congress (and the President) have the power to make binding law only through a constitutionally prescribed process that involves saying publicly what that law shall be. As Justice Scalia put it, “it is simply incompatible with democratic government, or indeed, even with fair government, to have the meaning of a law determined by what the lawgiver meant, rather than by what the lawgiver promulgated.”⁴⁴ Call this the promulgation principle.⁴⁵ Second, textualists believe that courts are obligated to take valid statutes as they find them, rather than seeking to improve upon them in the course of giving them effect. “Congress can enact foolish statutes as well as wise ones,” Justice Scalia continued, “and it is not for the courts to decide which is which and rewrite the former.”⁴⁶ For, again, “[i]t is simply not compatible with democratic theory that laws mean whatever they ought to mean, and that unelected judges decide what that is.”⁴⁷ Call this the fidelity principle.⁴⁸ Taken together, the two principles hold that Congress makes law only by formally enacting texts (promulgation) and that judges ought to faithfully apply the law thereby created (fidelity).

Over the past few decades, these basic theses have been elaborated and defended in many ways. For example, an important strand of textualist thought stresses that legislation often embodies a bargained-for compromise among legislators with competing aims, and that for this reason the statutes that emerge may not entirely square with the purposes of any of the individual legislators. Requiring courts to adhere to the promulgated text, even when it seems obvious that some deviation would better serve the aims of some or all lawmakers, preserves for members of Congress the opportunity to specify a compromise that they can trust will be honored rather than revised.⁴⁹ Relatedly, this kind of fidelity leaves to Congress the higher-order choice of how to allocate institutional authority for solving a given problem: Congress can leave discretion to courts by using conspicuously open-ended, purposive language, but it can also choose to resolve a matter itself by using more determinate wording.⁵⁰ Many textualists also believe that fidelity to the promulgated text can reduce the actual or perceived role of judges’ policy preferences in their decisionmaking⁵¹ and will promote predictability and clear notice for regulated parties.⁵²

But if all of this suggests that courts should faithfully interpret and apply the promulgated law — that they should show “fidelity to the text as it is written”⁵³ — it remains to ask what any given text, as it is written, means. Sophisticated textualists appreciate that a text is just an assemblage of signs, and that talk of fidelity to “the text itself” can thus only be a figure of speech; the real object of fidelity is some content that a text is used or understood to convey.⁵⁴ What content ought to count from a textualist point of view?⁵⁵

Most modern textualists answer that the relevant content is what a certain kind of reasonable reader would take a lawmaker who promulgated the statute to have said (or, perhaps, to have intended to say).⁵⁶ We will call this third commitment the reasonable reader principle.⁵⁷ Rendering this principle more determinate raises a number of complexities, but for now, the key point is a simple and familiar one: the communicative content that a reasonable reader would impute to the statute may differ from the legal rule that legislators actually intended to impose when they voted for the bill. Indeed, that gap is what gives textualism its bite. The classic example is Church of the Holy Trinity v. United States.⁵⁸ When Congress prohibited any U.S. person from bringing a foreigner into the United States “to perform labor or service of any kind,”⁵⁹ some or all members of Congress might have intended to ban only the hiring of manual laborers from abroad. But because the statute said what it said, textualists explain, any legislators with that limited aim evidently “overlegislated.”⁶⁰ The possibility that Congress employed “labor or service of any kind” in an idiosyncratic sense is ruled out by the premise that Congress was trying to communicate with a reasonable reader (or by the roughly equivalent premise that the statute’s communicative content should be discerned from such a reader’s point of view).⁶¹ And given that understanding of the prohibition that Congress actually established, the fidelity principle bars a judge from rewriting the statute so as to cover only the evils that the judge thinks Congress really intended, or ought to have intended, to address.

2.  Hard Cases and “Ambiguity.” — This repudiation of Holy Trinity–style reasoning illustrates the essence of what John Manning terms “second-generation” textualism: “[J]udges in our system of government have a duty to enforce clearly worded statutes as written, even if there is reason to believe that the text may not perfectly capture the background aims or purposes that inspired their enactment.”⁶² To this extent, moreover, “textualism” is now largely uncontroversial.⁶³ But what about cases where the communicative content of the statute is not so straightforward? Read one way, formulations such as Manning’s might suggest that textualism has little to say about statutes that are not “clearly worded” in the first place.⁶⁴ Indeed, Manning and others have observed that “when modern textualists find a statutory text to be ambiguous, they believe that statutory purpose — if derived from sources other than the legislative history — is itself a relevant ingredient of statutory context.”⁶⁵ Significantly for our purposes, such a restricted textualism might open the door for substantive canons as well, at least when the text is less than pellucid.⁶⁶

But we do not understand the jurists and theorists most prominently associated with textualism to take the modest view suggested by the most concessive reading of these provisos. To clarify textualists’ commitments, it will help to consider three different circumstances in which a statute might be deemed “unclear” or “ambiguous” and to observe how textualists characteristically approach each.

First, statutory language is sometimes “unclear” in the sense that it expresses a conspicuously vague or open-textured concept. Consider, for instance, the provision of the Patent Act that imposes liability on those who supply “a substantial portion” of the components of a patented invention for assembly abroad.⁶⁷ No reasonable reader would think that a lawmaker who gave that directive intended thereby to provide a determinate instruction as to whether liability should follow from each possible fraction. (Because any sane lawmaker would recognize that their chosen language could not do that, they could not have used that language with the intention that it would do so.) So, insofar as textualist provisos about “clarity” or “ambiguity” are meant to acknowledge the role for textually undetermined judgment in these sorts of cases, those provisos are not really qualifying textualism’s core theses at all. To the contrary, textualists who draw on extratextual resources to resolve these cases of blatant underdeterminacy — a task now often denominated “construction,” as distinct from “interpretation”⁶⁸ — are just being faithful textualists. They are taking the lawmaker’s use of blatantly indeterminate language as its own “crucial signal about the choice of means” that the lawmaker made.⁶⁹

Second, some cases present a court with language that is apparently intended as precise, but that in fact might bear either of two (or more) meanings. Limtiaco v. Camacho⁷⁰ offers a good illustration.⁷¹ That case concerned the Guam Organic Act, which caps Guam’s indebtedness at ten percent of “the aggregate tax valuation of the property in Guam.”⁷² Did “tax valuation” here refer to the appraised value of property (that is, its market value) or to its assessed value (as determined by local tax law)?⁷³ In contrast to the Patent Act provision discussed above, here it seems clear that a speaker would have meant something specific by “tax valuation”; they would not have meant the words to express some general concept that might later be concretized in either of the two ways. And, tellingly, the Court’s avowed textualists opted for the “assessed value” reading on the ground that it was the “most natural[]” linguistic fit — even though they agreed that the statutory term “has no established definition,”⁷⁴ and even though the dissenters (who found the case a “coin toss”⁷⁵ on the text) advanced powerful arguments that using assessed value ill served the statute’s purpose.⁷⁶ In other words, the Court’s textualists resolved Limtiaco in much the same way that they would have resolved Holy Trinity, even though they recognized that the linguistic import of the text in Limtiaco was much less clear cut.

We think this approach is characteristic of modern textualist jurists and fits naturally with their foundational commitments. By the same token, we think that Limtiaco and similar cases caution against overreading textualists’ provisos confining the duty to follow the text’s linguistic import to “clearly worded” statutes.⁷⁷ Leading textualists do not generally treat those provisos as relevant when faced with statutory language that, in their view, Congress intended to be determinate, even if Congress failed to express its instructions clearly.⁷⁸ Most notably, textualist interpreters do not look to Congress’s substantive policy objectives for guidance in such cases — perhaps because Congress’s apparent intention to speak clearly implies that it cannot be taken to have relied on any mutual understanding of those objectives to convey whatever content it intended.⁷⁹ In this sort of case, in other words, the lawmaker’s worldly purposes (or other elements of “policy context”⁸⁰) might speak to what they should have said, but, in light of the style of communication in which the lawmaker clearly signaled an intent to engage, those purposes cannot displace an otherwise-better candidate for what the lawmaker actually did say.⁸¹

Third (and finally), an intermediate kind of case presents a court with more modestly indefinite or polysemous language — language more like “vehicle” or “attorney’s fee” than either “substantial” (on the one hand) or “tax valuation” (on the other). Justice Scalia’s discussion of the classic “no vehicles in the park” hypothetical helpfully illustrates the prevailing textualist thinking here.⁸² The judge’s task in this case, he said, is to settle on “a selection from among the permissible meanings of vehicle.”⁸³ Significantly, Justice Scalia thought that doing so required the judge to consider “[t]he context of the [ordinance] here at issue, which includes its purpose of excluding certain things from the park — presumably things that would otherwise commonly be introduced.”⁸⁴ According to Justice Scalia, that purposive inference explains why “vehicles” in the sense of “substances used as mixing media” should not be deemed covered.⁸⁵ Looking to dictionaries and their own sense of colloquial usage, Scalia and his coauthor ultimately concluded that the best interpretation of “vehicle” in this context is “sizable wheeled conveyance.”⁸⁶ So “airplanes, bicycles, roller skates, and toy automobiles” are allowed into the park; “ambulances, golf carts, mopeds, [and] motorcycles” are not; and Segways pose the rare close case.⁸⁷

We take two related lessons from this discussion. First, faced with what is usually thought to be a paradigm case of uncertainty or indeterminacy, Justice Scalia maintained that textualism yields a bevy of determinate results.⁸⁸ Interpreting the park ordinance may not be “easy,” he said, but “the relevant line of inquiry is pretty straightforward” and “judges who use [his] method will arrive at fairly consistent answers.”⁸⁹ Second, Justice Scalia did not think that the uncertainty about the ordinance’s contours invited or authorized a judge to consider which interpretation would better serve what he called the lawmaker’s “more general” purposes, such as quiet or safety.⁹⁰ To be sure, Scalia did consider what he called the “textually apparent purpose” of the ordinance.⁹¹ But he defined that purpose as minimally as possible, and he used it only as a guide to settling on an apt, across-the-board understanding of “vehicle.”⁹² He did not look to “why things are excluded”⁹³ in order to decide, on a case-by-case basis, which particular objects should be deemed “vehicles” in the relevant sense.⁹⁴

Although we recognize the diversity within textualist thought, we think that Justice Scalia’s approach here is representative of mainstream modern textualism and fits well with most textualists’ avowed commitments.⁹⁵ Taking account of the “textually apparent purpose,” but not more, makes sense if one presumes that Congress intends even readers who might not know of or agree about its larger goals to converge on a common understanding of what it has said.⁹⁶ This approach also honors the promulgation principle by refusing to collapse the objective meaning of the text with what any given interpreter thinks the legislators were really trying to do. Moreover, considering a statute’s “textually apparent purpose” in discerning how a word or phrase was used does not seem to threaten legislators’ ability to bargain confidently and see their decisions enforced. If anything, this sort of attention to context seems essential to reconstructing the bargain that they would have understood themselves to have struck.⁹⁷

We do not want to overstate the point. “Textualism” can mean different things in the hands of different theorists and jurists, and some avowed textualists might reject the particular approach advocated by Justice Scalia in the “vehicle” case (or, for that matter, the approach taken in Justice Thomas’s opinion for the Court in Limtiaco).⁹⁸ For example, Tara Leigh Grove has recently described a “flexible” brand of textualism that “authorizes interpreters to make sense of the statutory language by looking at social and policy context, normative values, and the practical consequences of a decision.”⁹⁹ If that approach is accepted as a genuine form of textualism, then the view that we have in our sights might be singled out as “hard textualism” (or perhaps “formalistic textualism”).¹⁰⁰ But because we take the approach that we have described to represent the orthodox conception of textualism among most self-described textualists, particularly those on the bench, we will often refer to it simply as “textualism” as well.

Labels aside, the upshot of our discussion in this section is that, both in theory and in practice, textualists take their interpretive philosophy regularly to yield determinate answers even in hard cases (even if not in all such cases).¹⁰¹ Textualists recognize that Congress sometimes legislates standards as well as rules, and that reading a statute with a view to its manifest purpose may be necessary to identify the content that a lawmaker intends a reasonable reader to take the text as conveying. Neither of these acknowledgments imposes a serious limitation on either the promulgation principle or the fidelity principle, however. They simply flesh out how, at least according to leading textualists, a “reasonable reader” approaches a statutory text. Even when a judge is faced with a “prima facie ambiguity,”¹⁰² therefore, the judge’s job remains to try to determine, as best as they can, how such a reasonable reader would understand the statute — a task that requires respecting the distinction between what Congress actually said and what Congress ought to have said in order best to advance any given substantive aim.¹⁰³

B. Substantive Canons

Although the idea of a “substantive” canon seems straightforward enough, common usage of the term has tended to run together two different ideas. On the one hand, substantive canons are often characterized as “policy-based presumptions.”¹⁰⁴ But, on the other, it is also common to understand such canons as promoting “values external to a statute”¹⁰⁵ (in contrast to interpretive norms aimed at “decipher[ing] the legislature’s intent”¹⁰⁶). The problem is that these definitions do not necessarily come to the same thing. In fact, as we have seen already, a reader might make suppositions about the kinds of policies that a lawmaker is likely to pursue precisely in order to discern what the lawmaker intended the statutory text at hand to convey.¹⁰⁷

Our purposes here require a definition that is somewhat more precise — and our account of textualism naturally suggests one.¹⁰⁸ For textualists, the obvious distinction to draw among putative canons is between those that are justified by their probative value with respect to a statute’s communicative content — that is, by the light that they cast on what a reasonable reader would understand a lawmaker to have said — and those that are not.¹⁰⁹ We thus propose that a canon is “substantive” in the sense that is of interest here when it purports to speak to a statute’s proper legal effect in a way that is not mediated by its evidentiary bearing on what a reasonable reader would take a lawmaker to have said in enacting the statute.¹¹⁰

To make the meaning and implications of this definition clear, however, we will first need to distinguish not two categories, but four. Each of these captures one salient kind of reason why a judge might see fit to read a statute as a given canon prescribes (apart, that is, from any precedential weight held by the canon itself). Put another way, the four lines of thought that we will now distinguish amount to four different ways of justifying a canon in the first place. Because the first two of these justifications ground the canon’s force in what it reveals about a statute’s communicative content, they are not “substantive” in the sense that concerns us here (and so are not within the ambit of our critique). The final two types, however, are substantive — and, as we will see, each breaks with one of the central principles of textualism as we have described it.

1.  Inferences from Manner of Expression to Communicative Content. — Consider, first, the assorted canons that are traditionally classified as “semantic” or “linguistic.” What these canons most clearly have in common is that they speak to a statute’s communicative content — to what a reasonable reader would take a statute to convey — without presuming anything about what the lawmaker was more or less likely to have intended the statute, once interpreted, to accomplish. That explains the widely shared sense that, as Justice Kagan once put it, these canons “formaliz[e] . . . intuitions” about “how language works and how the people who write things think that language works.”¹¹¹ The presumption that a word or phrase bears the same meaning throughout a statute is a good example.¹¹² Regardless of the content that a lawmaker intends to convey, the lawmaker is (we might suppose) unlikely to express that content in a way that requires readers to assign different meanings to different instances of the same word. All else equal, a reasonable reader seeking to reconstruct the statute’s communicative content should therefore favor interpretations that would show the statute to be consistent with this pattern. This “content-neutral” focus on how Congress is apt to express itself makes these canons safe for textualists and makes the “semantic” or “linguistic” moniker understandable (even if technically imprecise).¹¹³

2.  Inferences from Purpose to Communicative Content. — Some canons trade on a richer body of contextual information than we have just described, but nonetheless remain aimed at deciphering the communicative content that a reasonable reader would take a statute to bear (rather than, say, at discerning the disposition of a case that the lawmaker would have favored). To get a fix on this intermediate category, consider again Justice Scalia’s supposition that the park ordinance was supposed to exclude “vehicles” in the sense of conveyances rather than of mixing media.¹¹⁴ He explained that this follows from the ordinance’s apparent “purpose of excluding certain things from the park — presumably things that would otherwise commonly be introduced.”¹¹⁵ Indeed, he observed, “[t]here is no more reason to address intrusion into the park of mixing media than to address intrusion of elephants.”¹¹⁶ Now, we are taking as a fixed point that this reasoning is permissible from a textualist point of view. But it is plainly not true that the purpose Scalia imputed here is entirely “derived from the text, not from extrinsic sources such as . . . an assumption about the legal drafter’s desires.”¹¹⁷ After all, a lawmaker certainly could intend to guard against a rare occurrence rather than a “common[]” one; nothing in the text favors one assumption over the other. And even granting Justice Scalia’s assumption on that point, there is no basis in the text for supposing that mixing media are not commonly introduced into parks. So the lesson of Justice Scalia’s discussion is that the textualist’s reasonable reader must bring some general background knowledge — not just about language use, but also about lawmakers and even about parks — to the interpretive inquiry.

When a canon trades on that sort of information — information about what lawmakers are likely to be trying to accomplish in the world, rather than about how they are likely to express themselves in pursuing whatever worldly aims they may have — the canon is not aptly labeled “semantic” or “linguistic.” But neither is such a canon inherently “substantive” in the sense that concerns us here, since it may aim only to decipher what the lawmaker is best understood to have said, rather than what the lawmaker ought to have said to best further its purposes. Such canons are compatible with textualism so long as the background information on which they trade is of the sort that the textualist’s reasonable reader, rightly understood, possesses (and would take the lawmaker to have known they would possess as well).¹¹⁸ The presumption against extraterritoriality is one plausible example of a canon of this kind.¹¹⁹ We will return to this type of canon (and say more about that possible example) shortly, because one natural way for textualists to accommodate some canons that are traditionally thought substantive is to recast them as falling in this nonsubstantive, nonsemantic category instead.

3.  Intended Effects. — A third type of canon — which is substantive — favors statutory interpretations thought to accord with Congress’s usual preferences or intentions regarding the effects of the statutes it enacts. Consider, for example, the Charming Betsy canon, which favors interpretations of statutes that would put them on the right side of international law.¹²⁰ Under one influential conception of this canon, it “does not claim that Congress considered international law in enacting the statute, just that Congress would not have wanted to violate international law if it had considered it.”¹²¹ That is a claim about the effects that Congress would have wanted a statute to have, not a claim about what a reasonable reader would think Congress actually said (or even intended to say) in enacting the text at issue. Even if the claim undergirding the canon were true, therefore, the promulgation and reasonable reader principles would rule it out of bounds as a consideration relevant to a textualist’s statutory interpretation. As Justice Scalia would put it, “we are governed by what the legislators enacted, not by the purposes they had in mind”; “[w]hen what they enacted diverges from what they intended, it is the former that controls.”¹²²

4.  Superimposed Values. — Finally, a second and distinct type of substantive canon simply favors interpretations that the court deems more consistent with some important value, regardless of whether that fact might support a further inference that Congress would therefore have favored those interpretations. For instance, some interpreters might endorse the Charming Betsy canon on the straightforward ground that compliance with international law is, in their view, a good thing. In fact, many canons appear to rest on this sort of justification, from the rule of lenity (which is often justified simply as promoting fairness)¹²³ to the major questions doctrine (which has been justified partly on the basis of the “separation of powers principles” that it allegedly serves).¹²⁴ When canons are justified in this way, their force is independent of whether or how much Congress actually values the principles at issue. Therefore, deploying a canon justified in this way would seem a blatant breach of the fidelity principle. If textualists have been clear about any­thing, it is that a judge’s job is to give effect to the law that Congress actually made, not to improve upon it.¹²⁵

C. The Problem Restated

The upshot of our discussion thus far is straightforward: When canons are justified in the ways that warrant dubbing them “substantive,” they appear flatly at odds with textualism as commonly practiced and traditionally justified. Although we think this conflict has received too little attention, we readily acknowledge that several textualists have grappled with just this concern.¹²⁶ As then-Professor Barrett summarized the problem in her 2010 article:

Substantive canons are in significant tension with textualism . . . . [When a] judge applying a substantive canon . . . exchanges the best interpretation of a statutory provision for a merely bearable one . . . [,] she abandons not only the usual textualist practice of interpreting a statute as it is most likely to be understood by a skilled user of the language, but also the more fundamental textualist insistence that a faithful agent must adhere to the product of the legislative process, not strain its language to account for abstract intention or commonly held social values.¹²⁷

In short, we agree.

The question then becomes whether there is any way out of this dilemma. Before turning to that question in earnest, however, we should make explicit what is probably already obvious: our definitions of textualism and of substantive canons entail that deploying a genuinely substantive canon cannot comport with unadulterated textualism. To be “substantive,” we have suggested, just is to bear on something other than the meaning that a reasonable reader would impute to the text — and “textualists,” we have also said, are committed precisely to the notion that judges must faithfully seek to reconstruct that meaning in statutory cases. Because we do not want to define the debate away, we have to be alert to the possibility that what is in substance a reconciliation of textualism and substantive canons will look less like a means of dissolving this conflict and more like a way of circumventing it.

And, indeed, the proposed reconciliations that we will entertain in the balance of this Article all respond to the dilemma in much that way. The first possibility, stated broadly, is that some or all of the canons that are commonly denominated “substantive” actually do speak to how a reasonable reader would understand what Congress said; they are thus not “substantive” in our sense at all, and textualists are within their rights to employ them. A second argument concedes the substantive quality of many substantive canons, and perhaps even their inconsistency with textualism as such, but defends many of them as legitimate exceptions to textualism — exceptions that are justified on constitutional grounds. And a third possible reconciliation likewise suggests that at least some substantive canons — now the “ambiguity-dependent” ones — operate outside textualism’s domain. Over the next three Parts, we assess each of these possibilities in turn.

II. “Substantive Canons” as Nonsubstantive Canons

The idea that substantive canons could be recast in nonsubstantive terms comes in two basic forms. Both versions of the argument claim that these canons (or at least some of them) actually capture how a reasonable reader would understand the content of a lawmaker’s communication. But they differ in why they take this to be the case — specifically, in the direction of the causal arrow that they would draw between the canons and the meaning of a lawmaker’s speech. According to one story, the canons simply distill general background knowledge about how lawmakers would naturally tend to express themselves, at least when they intend to convey certain content; the canons would thus speak to a statute’s communicative content even if that statute had been enacted in a canon-free world. According to the other account, the canons enjoy a kind of bootstrapped validity: they are probative of a statute’s communicative content precisely because, and to the extent that, their very existence can be presumed to have shaped that communicative content.

Of course, these ideas are not mutually exclusive, and in practice they often come bundled together. For purposes of analysis, though, it proves important to tease them apart. We will thus take each in turn.

A. Canons as Guides to the “Natural” Meaning of Legal Texts

“[T]he good textualist is not a literalist”; they understand that the meaning of a statutory text depends on what, in context, a reasonable person would think a lawmaker had said.¹²⁸ As we have already seen, moreover, the textualist’s reasonable reader makes that judgment armed with the sort of background knowledge that might loosely be described as common sense. This opens the door to recasting some seemingly substantive canons as simply default inferences that a reasonable reader would draw — not about what Congress intended a statute to do (either to the world or to the law), but about what, given a commonsense understanding of Congress’s aims, Congress should be understood to have said.¹²⁹

The presumption against extraterritoriality is a possible example.¹³⁰ Statutes often say that it shall be unlawful for “any person” to perform a particular act, or that “whoever” performs the act shall be punished. Read literally, these sentences pertain equally to all people who perform the act in question, everywhere in the world. Yet it is common to interpret this language not to reach conduct outside the United States. Imposing this limitation might seem at odds with textualism — an instance of narrowing a broadly worded statute to capture only the results that, one imagines, Congress would really have favored.¹³¹ But the application of the canon here need not be understood in that way. In ordinary speech, the practical context in which an assertion is made often tacitly restricts its domain. For example, if a store manager posts a sign that says “Everything Is On Sale,” they are asserting that all of the goods that are displayed in the store are on sale — not that the cash register or the shelves are, and not that goods in other stores are either.¹³² Perhaps the communicative content of a lawmaker’s stipulations is similarly restricted by default to the lawmaker’s territorial jurisdiction. In other words, perhaps the reasonable reader simply understands that, given what are mutually understood to be the ordinary concerns or objectives of lawmakers, this restricted context is the one “under discussion” when Congress speaks about what is allowed or forbidden.¹³³ Although this requires a supposition about lawmakers’ typical purposes, that supposition does not seem different in kind from the assumption, noted above, that lawmakers intend to exclude from parks items that would otherwise be commonly introduced into them.¹³⁴

Justice Barrett recently undertook to defend the major questions doctrine in parallel terms.¹³⁵ According to Justice Barrett, that doctrine is not actually a substantive canon at all, but instead simply affords due interpretive weight to a putative shared understanding (embodied in “common sense”)¹³⁶ that major delegations to agencies are exceptional or anomalous. A reasonable grocery clerk, she pointed out, would not take an instruction to “buy apples” as a grant of authority to buy an unlimited number of them; the clerk would appreciate that “the grocer would have spoken more directly if she meant to authorize such an out-of-the-ordinary purchase.”¹³⁷ Likewise, a reasonable reader would not take Congress as making an extravagant delegation through language that it would have known could also be taken as expressing something more routine.¹³⁸ So if a statute can be read in either of two ways, only one of which says that a major question should be resolved by an agency, then Congress’s sheer failure to clarify — together with the premise that major delegations are mutually understood as improbable — favors the reading whereby Congress is not delegating a major issue. Schematically, this argument resembles the defense of the extraterritoriality canon that we sketched above. And like that argument, this one does not seem to differ in kind from Justice Scalia’s more ad hoc resort to the premise that Congress usually tackles recurring rather than niche problems.¹³⁹

But while there is nothing inherently improper about this way of squaring some (seemingly) substantive canons with textualism, Justice Barrett’s argument illustrates how such attempts are apt to come up short. First of all, the story that one has to tell in order to vindicate a canon in nonsubstantive terms often just is not persuasive (and contrary claims then appear to mask, deliberately or not, the interpreter’s appeal to their own ideas about how Congress should be in the habit of legislating).¹⁴⁰ Notwithstanding Justice Barrett’s assertions, for example, we see little reason to think that “major” delegations are anomalous, especially in statutes specifying the authorities of a regulatory agency charged with addressing some complex and evolving problem.¹⁴¹ Although Justice Barrett insists that her “baseline assumptions”¹⁴² about how Congress empowers federal agencies are not “normative,” they do appear mostly to reflect her sense of what “the Constitution’s structure” contemplates — namely, a regime in which Congress “make[s] the big-time policy calls itself” — rather than even her own sense of what Congress actually does.¹⁴³ And unless a lawmaker thought that a reasonable reader would start from the same delegation-light picture that Justice Barrett does, that lawmaker would not naturally feel under pressure to specially flag any major delegations that would be effected if the statute’s terms were given their otherwise-most-natural meanings, for fear of being taken as implicitly clawing those delegations back.¹⁴⁴ At most, then, any interpretive inference that could be drawn on that basis would be a very weak one.¹⁴⁵

The depth of the challenge here bears emphasis. In order for a textualist interpreter to be justified in drawing the sort of inference about statutory meaning that we have described, it is not enough for them to be correct in endorsing the premises of a given canon — for example, correct in the belief that Congress rarely makes major delegations — or even to be correct in thinking that the lawmaker shares their perspective on that point. What matters, really, is whether the interpreter thinks that the lawmaker knows that this same perspective is widely shared among the audience by whom they intend to be understood. Only then would a reasonable reader expect the lawmaker to take their (presumed) knowledge of some proposition into account in shaping a message to them, as the argument above imagines.¹⁴⁶ But, especially in the legislative context, this sort of common knowledge is hard to come by. After all, there are many propositions that many generally informed people might know (say, about the pro-business tendencies of the U.S. Congress) but that are nonetheless irrelevant to deciphering the communicative content of a statute — precisely because a reasonable reader would not think that Congress, faced with the task of communicating clearly with a diverse audience, could have counted on the reader’s knowing those propositions. This is what textualists are recognizing, we take it, when they speak of a presumption “that Congress communicates with the regulated according to the conventions that the two share as skilled users of English.”¹⁴⁷ If that restriction is taken to suggest that the context must be (or even could be) strictly linguistic, it is overstated. But it accurately reflects that the relevant context does not extend to anybody’s sectarian ideas; “common sense” is fair game only insofar as it is genuinely common.

Returning to the major-questions example, then, even if the (putative) fact that Congress rarely makes major delegations were accepted by many, it is an exceedingly unlikely candidate for an element of the shared context on which Congress could rely (or on which reasonable readers would therefore take Congress to rely) in legislative communication.¹⁴⁸ When you interpret a parent’s instruction to a babysitter, it might well make sense to draw on your “intuition” about the “balance of power” between the two.¹⁴⁹ But that is only because you can trust that the parent had — and would have taken the babysitter to have had — the same intuition about the terms of the relationship that you do. For better or worse, intuitions about the terms of Congress’s relationship to federal agencies, including the typical frequency and scale of delegations, vary widely — even among competent and reasonable people.¹⁵⁰ And if the putative fact about Congress’s “habits”¹⁵¹ on which Justice Barrett’s argument depends is thus not a matter of common ground, then even a reasonable reader who did accept it would have to take it only as a reason for suspecting that — insofar as Congress used language that most naturally suggests a major delegation — Congress may well have failed to say what it really should have said to best further its own purposes.¹⁵² For a judge bound by the fidelity principle, that may be cause for regret, but it is not a license for “rewrit[ing] the statute so that it covers only what [the court] think[s] . . . Congress really intended.”¹⁵³

Finally, we cannot help but note an astute law professor’s observations about the predictable “temptation to rationalize ostensibly substantive canons” in nonsubstantive terms.¹⁵⁴ Even though “argument[s] that substantive canons capture what ordinary language means or what Congress would want”¹⁵⁵ have “been largely rejected,”¹⁵⁶ then-Professor Barrett explained, they “surface[] repeatedly in the cases and literature” — a tendency that “almost surely reflects discomfort with the application of substantive canons in a legal climate where a strong vision of legislative supremacy is the dominant view.”¹⁵⁷ We agree that the “motivation for making this move” is “easy to see.”¹⁵⁸ And so too, we imagine, Justice Barrett might agree that we should greet with skepticism arguments that purport to salvage controversial canons by imputing to Congress (and to all “reasonable interpreter[s]”) a judge’s own views about what “makes eminent sense in light of our constitutional structure.”¹⁵⁹

B. Canons as Stipulated Linguistic Conventions

All of this suggests that seemingly substantive canons rarely start out as reasonable generalizations about the communicative content of statutes. But, as we have noted, one could accept that much and still insist that, once the error of “canonizing” these canons has been made, they come to enjoy a kind of bootstrapped validity. In fact, this is probably the most developed and time-honored defense of textualists’ reliance on substantive — or, from this point of view, formerly substantive — canons. As Justice Scalia put the point, when norms like the rule of lenity “have been long indulged, they acquire a sort of prescriptive validity, since the legislature presumably has them in mind when it chooses its language.”¹⁶⁰ He later elaborated the same reasoning (writing together with Bryan Garner) as follows:

It might be said that rules like these, so deeply ingrained, must be known to both drafter and reader alike so that they can be considered inseparable from the meaning of the text. A traditional and hence anticipated rule of interpretation, no less than a traditional and hence anticipated meaning of a word, imparts meaning.¹⁶¹

Scalia illustrated the idea by positing a hypothetical scenario in which “the Supreme Court . . . announce[d] and regularly act[ed] upon the proposition that ‘is’ shall be interpreted to mean ‘is not.’”¹⁶² In that world, he suggested, it would be appropriate for the Court to interpret legislation as it had said it would. The reason is that if, in this hypothetical context, Congress enacts a statute containing the word “is,” the content it intended to convey is more likely the one traditionally expressed (in ordinary speech) by “is not.” So too, the thought goes, with substantive canons that are sufficiently entrenched to shape Congress’s expressive choices.¹⁶³

John Manning has integrated much the same thought into his larger explication and defense of “textualists’ practice of reading statutes in light of established background conventions.”¹⁶⁴ As Manning observes (and as we noted above), “[e]ven the strictest modern textualists properly emphasize that language is a social construct,” and they are therefore interested not in literal meaning, but in “how a reasonable person, conversant with the relevant social and linguistic conventions, would read the text in context.”¹⁶⁵ Once one makes that move, the thinking goes, at least some substantive canons can be thought of as settled conventions that dictate how legal language is interpreted — such that a reasonable interpreter familiar with that context would understand statutory terms in the ways indicated by these canons, even if that is not how an ordinary reader would understand that same language in some other context.¹⁶⁶

While Scalia and Manning have done the most to elaborate this idea in theoretical terms, the Court itself has sometimes gestured in this direction as well. Before bringing a substantive canon to bear, for example, the Justices will often recite that they “assume that Congress legislates against the backdrop” of the canon at issue,¹⁶⁷ or more broadly, that the Court “presume[s] congressional understanding of . . . [the Court’s] interpretive principles.”¹⁶⁸ In context, these pronouncements seem to be stating a rule about how the Court interprets statutes more than expressing a belief about how Congress legislates.¹⁶⁹ But, one way or another, they imply that the canon at issue is relevant to statutory interpretation not (or not only) as a kind of superimposed value judgment, but because the canon was part and parcel of the context that Congress itself would — or at least could — have considered in deciding how best to get across whatever it was that it intended to communicate.

Because this idea — which we will term the “bootstrapping argument” — has gained more traction than the last one, we will consider it in somewhat greater depth. In what follows, we first highlight several limitations on the results that the argument can deliver, and then argue that it is difficult to square with textualist premises even within its limited domain.

1.  Inherent Limitations. — Even if we were to stipulate to the bootstrapping argument’s coherence and compatibility with textualist principles, that justification is subject to a couple of important limitations. First, the bootstrapping argument would do nothing to justify the Court’s positing any new substantive canons, or indeed the use of any canon that was not well established at the time the relevant statutory text was enacted.¹⁷⁰ That is no minor caveat. Justice Scalia may have been able to assert that a venerable canon like the rule of lenity is validated by its “sheer antiquity,”¹⁷¹ but many of the canons that courts apply, and that have proved most consequential in recent years, are of recent vintage.¹⁷² Often the Court applies substantive canons to statutes enacted before the Court had clearly articulated those canons, rendering the claim that Congress drafted the statute with a given canon in mind absurd on its face.¹⁷³ And even in the case of statutes enacted after the Court announced the relevant canon, it is not at all clear that actual legislators are sufficiently aware of or focused on the Court’s interpretive practices to craft statutes in light of them.¹⁷⁴ If you sent your friend a special code book to use when sending you messages, and then you learned that they didn’t read it, it would make no sense to go on deciphering their meaning with reference to that lexicon rather than the standard one. So too here.¹⁷⁵

Furthermore, the bootstrapping argument could justify at most retrospective fidelity to heretofore settled canons; it does not explain why the Court should not renounce even those canons going forward.¹⁷⁶ Absent such an explanation, the substantial costs of specialized communicative conventions provide a strong prima facie case for repudiating them. As leading textualists have argued, transparency and democratic accountability strongly favor giving words and phrases their ordinary meanings, not special stipulative ones.¹⁷⁷ (For this reason alone, if the Justices found themselves in the bizarre world where their predecessors had made “is” into legalese for “is not,” they would presumably repudiate that confusing anomaly for purposes of interpreting statutes enacted in the future.) Additionally, even well-settled interpretive conventions will often make legislating in particular ways more difficult; indeed, that effect on the legislative process is part of the (nontextualist) case for imposing substantive canons in the first place.¹⁷⁸ From a textualist point of view, therefore, the Court presumably ought to clear out any residue left behind by judge-made, formerly substantive canons and leave it to Congress alone to specify (as through the Dictionary Act) any special interpretive conventions that it intends to employ.¹⁷⁹

2.  Can Substantive Canons Be Bootstrapped? — The domain limitations described above may be sufficient to dispose of the bootstrapping argument entirely: The set of canons for which that argument is plausible may be empty, or so limited that the argument could not sustain more than a tiny number of unusually entrenched interpretive conventions. But let us now set that point aside and ask whether, within the domain where the argument might have purchase, it is compatible with textualist principles. If a substantive canon really was well established at the time of a given statute’s enactment, doesn’t that context bear on the content that any given text should be understood to convey?

We see the allure of this thought, but we do not think a textualist can embrace it. To explain why, it will help to draw a familiar distinction between two kinds of canons: (a) “ambiguity-dependent” canons, such as the rule of lenity, that tell interpreters how to construe an otherwise-unclear text; and (b) “clear statement” or “implied limitation” rules, such as the presumption against abrogation of state sovereign immunity, that narrow the apparent sweep of statutory language even though that language is not “unclear” in the usual sense.¹⁸⁰ The bootstrapping argument fails in each of these settings for different, but related, reasons.

(a)  Ambiguity-Dependent Canons. — Consider, first, canons that purport to resolve textual ambiguities, such as the rule of lenity or constitutional avoidance. For a concrete example, we will use Commonwealth v. Davis,¹⁸¹ a Kentucky case that Justice Scalia and Bryan Garner offer as a paradigm of where lenity ought to apply.¹⁸² In Davis, two teenagers pooled their money to buy whiskey, and one was then charged with the crime of “sell[ing], loan[ing], or giv[ing]” alcohol to a minor (the other).¹⁸³ Scalia and Garner suggest that, in light of the rule of lenity, “give” should be construed narrowly in this case — to mean “bestow a gift” (which the defendant did not do), rather than “furnish, provide, or supply” (which he did).¹⁸⁴

But could a court that refused to give genuinely substantive force to the rule of lenity follow this advice? We think not. Suppose the court believes that the rule of lenity is “so deeply ingrained” that it “must [have] be[en] known to” the Kentucky legislature.¹⁸⁵ And presume, for the moment at least, that the relevant “reasonable reader” (whose perspective the court must adopt) also knows of both the rule of lenity and the legislature’s acquaintance with it.¹⁸⁶ Even so, that reasonable reader could hardly think that the legislature intended to convey the narrower meaning of “give” and opted to rely on the combination of ambiguity and lenity to get that content across. After all, if a lawmaker both intended to convey “bestow a gift” and realized at the time that “give” was ambiguous between that meaning and another one, why would the lawmaker not have simply opted for “bestow a gift” over “give” in the first place? It seems far more likely that the lawmaker either overlooked the ambiguity or spotted it but chose not to resolve it. Either way, surely the least likely inference is that the lawmaker gambled on a reader later deeming the statute gravely ambiguous, and thus turning to the rule of lenity, in order to arrive at the content that the lawmaker actually did intend all along.¹⁸⁷

At least with respect to ambiguity-dependent canons, then, the bootstrapping argument is better construed differently. True, a lawmaker could not plausibly aim to communicate some specific content (such as “bestow a gift”) indirectly via a mutual awareness of an ambiguity-dependent canon. But still, the lawmaker might reasonably be assumed to have relied upon — or, at least, acquiesced to — the future use of any then-established canons that the lawmaker did not disturb or override. So, for instance, the Kentucky legislature might be taken as intending to convey both some particular content (whatever it actually sought to prohibit) and also that, as per usual, the prohibition should be deemed unenforceable to the extent of any grave ambiguities in the language used to express it. In that event, a court that applied lenity would be doing as it was (implicitly) told; it would not need to credit any of the substantive justifications for lenity.¹⁸⁸

The problem with this alternative argument is that nothing actually supports imputing to a legislature a tacit endorsement of whatever interpretive rules it leaves undisturbed. Indeed, once we separate out the legislature’s implicit endorsement of the relevant canons from the meanings of the words it used — so that we take the legislature to have intended to convey both some content that is expressed by those words and a kind of rider authorizing the employment of lenity — it becomes mysterious why we would think the statute expressed the latter content at all.¹⁸⁹ Rather than positing statute-specific lenity clauses (all written in “invisible ink”¹⁹⁰), we should presumably say that the rule of lenity exists apart from any particular statute, and that a lawmaker who says nothing about that canon simply leaves it as it stood.¹⁹¹ But then, if the rule of lenity starts out as a substantive canon at odds with textualism, it does not become part of “what the lawgiver promulgated”¹⁹² just because the lawgiver knew about it in advance. To the contrary: What the lawmaker knew in advance — and left unchanged — is that the standing judicial practice is not fully textualist.

Given all of this, we suspect that the intuition driving the bootstrapping argument is not really that standing interpretive practices shape what a lawmaker is reasonably taken to have said — or, really, anything to do with linguistic meaning at all. Rather, the intuition is that courts should try to vindicate the expectations that lawmakers or others would justifiably have had about the effects that a statute’s enactment would have on the law.¹⁹³ Here again, one natural response is that, even if that prescription were sound, it would not bring seemingly nontextualist norms within the fold of textualism; it would just identify a reason, akin to stare decisis, for deviating from textualism.¹⁹⁴ Indeed, the same sort of thinking would have pressed powerfully against the rise of modern textualism in the first place.¹⁹⁵ And while one could certainly redefine “textualism” as this line of argument requires, that change would come at steep rhetorical and argumentative costs.¹⁹⁶ It would mean dropping the refrain that textualism inherently privileges the communicative contents of promulgated texts (i.e., “what the text means”), and it would sap the intuitive motivation for elevating so-called “semantic” context over other context in determining a statute’s legal effect — since, from this point of view, there is nothing so special about linguistic meaning, at least as far as textualism goes, to begin with.

Finally, even putting aside whether an indiscriminate rule of vindicating justified expectations could fairly be cast as “textualism,” it is not clear that such a rule actually does cut in favor of sticking with lenity or similar norms.¹⁹⁷ Consider the Kentucky legislature again. If we are to impute to the legislature a grasp of the existing law of interpretation, including the rule of lenity, we should presumably also impute to it an awareness that lenity is a substantive canon — in particular, a canon of the “superimposed value” variety that we distinguished above — rather than a conjecture on anyone’s part about what a lawmaker intended or expected.¹⁹⁸ Given that full context, however, the only expectation that a lawmaker could be justified in maintaining is conditional: that the prohibitions they enact will be tempered by lenity if, and to the extent that, the courts maintain their own policy of “tenderness for the accused”¹⁹⁹ — about which, remember, the lawmaker has chosen to say nothing either way. So if the promulgation and fidelity principles, rightly understood, mean that courts really have no business showing tenderness toward the accused, then we do not see how a textualist could conclude that anybody’s legitimate expectations would be thwarted if courts simply modified their approach, even with respect to old statutes, accordingly.²⁰⁰

An analogy may help to underscore this point. Suppose that, at Time 1, Congress enacts some statutory cause of action that allows “any person” to bring suit in federal court. At that time, the Supreme Court’s Article III case law nonetheless denies standing to most would-be plaintiffs. At Time 2, the Court liberalizes about standing to some extent. Would it make sense for a court then to hold, as a matter of statutory interpretation, that the cause of action conferred in this particular statute is restricted to those who would also have had constitutional standing at Time 1? We doubt it — and we think textualists would be especially unlikely to think as much. Yes, they would say, the ultimate legal effect of the statute now departs from what the smart money would have predicted at the time of its enactment. But — they would explain — that just reflects Congress’s choice to leave any narrowing of the universe of potential plaintiffs to be done by the Court, pursuant to its Article III case law, rather than to prescribe any desired limits itself. In other words, the smart money is usually right about a statute’s legal effect — but it is not always right, no matter what, just by dint of having been smart.

(b)  Clear Statement Rules. — Turn, then, to clear statement or implied limitation rules. When it comes to these canons, the first construal of the bootstrapping argument that we considered (and dismissed) above — the notion that the canon actually shapes what the lawmaker means in speaking as they do — is significantly more plausible. Suppose, for example, that Congress imposes liability on “any recipient of Federal assistance” who violates certain requirements.²⁰¹ If Congress chose that language knowing that courts would deem the provision not to apply to states that receive federal funds, then it does not seem fanciful to infer that Congress used the phrase with that same restricted meaning. In effect, we might think, the term “recipient of Federal assistance” has simply become a “term of art” with a specialized legal meaning.²⁰² And just as with Justice Scalia’s hypothetical where “is” has come to mean “is not”²⁰³ — the argument would go — the courts ought to gauge the statute’s communicative content accordingly.

Even here, however, the same basic rejoinder applies. Assume that Congress (and others) did foresee that courts would respond to the statute at issue by carving out states. Just as with the ambiguity-dependent canons, it does not follow that Congress should be understood as telling courts to do this — as opposed to acquiescing to (or, for that matter, capitalizing on) the prospect that, for their own reasons, courts probably would do it. For all the courts can tell, in fact, the bargain struck and recorded in the text might be precisely an agreement to say that any recipient of federal assistance is liable while eschewing any mention of states as such. Opponents of state liability might have been willing to make that agreement not because they justifiably believed the courts would take Congress as actually saying something narrower than it did, but rather because they were optimistic that courts’ nontextualist solicitude for federalism would lead them to deviate from the apparent content of Congress’s instruction.²⁰⁴

We can illustrate the point here with an example set in a more familiar context. Imagine that two parents — you and your partner — have different hopes for when your family’s dinner will start. You would like it to start at 6:30, but your partner would like it to start at 6:15. Both of you also know from experience that your child rarely cleans up until he thinks he is already supposed to be at dinner, so he is usually 15 minutes late. After some negotiation, you and your partner decide to tell your child, “dinner at 6:15.” Under this plan, you are probably going to get your way (i.e., dinner at 6:30), but your partner can hold out hope that the child will actually do as he is told for once (i.e., be ready for dinner at 6:15). Now suppose that the child does decide that, given his role in the family, he really should try to follow his parents’ instructions. That would mean trying to be ready at 6:15, not 6:30 — even if this comes as an unwelcome surprise to you (and a pleasant surprise to your partner). The same goes for a court that, having employed substantive canons (and thus deviated from the apparent communicative content of statutes) in the past, belatedly embraces the conception of its role prescribed by textualism.

Now, there is a natural response for proponents of the bootstrapping argument to offer: If the child had come to understand that when a parent says “dinner at 6:15,” the parent actually intends to convey that dinner should start at 6:30, then dutiful compliance would require being ready at 6:30. But notice that (a) intending dinner to start at 6:30, and (b) meaning (or intending to communicate) that dinner should start at 6:30, are fundamentally different things.²⁰⁵ When you say “dinner at 6:15,” you may intend that dinner start at 6:30 (counting on the child’s lateness), but you do not intend to communicate that start time.²⁰⁶ In fact, if the child took that as your meaning, he probably would not be ready until 6:45. And while it is certainly possible for an inside joke to develop within the family such that “dinner at 6:15” does refer to 6:30, the key point for our purposes is that conventions of that sort require mutuality.²⁰⁷ In our example, the child would not have reason to think that you meant anything nonstandard by “6:15” unless he thought that you thought that he would understand you in some nonstandard way. But as long as his lateness appeared to be due to his own disregard for his parents’ instructions — as it would if, for instance, he explained it that way — you would have no reason to think that he actually would interpret you in a nonstandard way, and hence no reason to try to communicate with him in that way, leaving him no reason to take you as doing so.

The very same lack of mutuality explains why it seems doubtful that many, if any, implied limitation rules actually shape the content that Congress could be understood to intend to communicate (even to courts, let alone — as we will soon discuss — to the “reasonable reader”). Again, suppose that Congress intends to impose liability on recipients of federal funds other than states (but not on states). If Congress believed the courts actually think that when Congress refers to “any recipient of Federal assistance” in the context of imposing liability, Congress intends to be taken as referring only to recipients other than states, then Congress might intend to accomplish its ultimate aim simply by means of being understood in that way. But since Congress (like the Kentucky legislature) would know that implied limitation rules are justified mainly or entirely on substantive grounds, we do not see why Congress would think that courts are behaving as they do only because of what the courts think Congress more likely means.

What Congress would see as part of the backdrop, in other words, is not a linguistic convention of which it could avail itself, but a “requirement,” rooted in the Court’s concern to protect the “usual constitutional balance between the States and the Federal Government,” that “[w]hen Congress chooses to subject the States to federal jurisdiction, it must do so specifically.”²⁰⁸ Given that understanding on Congress’s part of what the courts are doing, there would be little reason (and perhaps no room) for Congress to try to get courts to exempt states from liability by an alternative, roundabout means instead — that is, by courts’ taking Congress’s use of “any recipient of Federal assistance” as an effort on Congress’s part to communicate that narrower meaning without saying it outright.²⁰⁹ At a minimum, the description of Congress as intending to use the words in their ordinary senses, while simply expressing no view about the courts’ possible departure based on a substantive canon, seems equally plausible.

Finally, if all of this seems somewhat technical, there is also a more basic and more straightforwardly normative dimension to the problem. We have been entertaining a tacit premise that Congress is engaged in a kind of dialogue with the courts, akin to the imagined dialogue between a parent and child. But textualists are largely and under­stand­ably opposed to that whole picture of the relevant communication. As we noted earlier, textualists take statutes to be addressed to the “reasonable reader” — “a skilled, objectively reasonable user of words,”²¹⁰ not a legislator or a court.²¹¹ Indeed, Justice Scalia insisted that nearly all “legislation is an order not to the courts but to the executive or the citizenry.”²¹² And, as then-Professor Barrett argued, “[i]f . . . a legislative command is directed to the citizenry, it is both sensible and fair for the courts to interpret that command as its recipients would.”²¹³ Ryan Doerfler thus observes that “because context consists of mutually salient information,” the background that informs the communicative content of a statute will often be “limited to information of which citizens should be aware.”²¹⁴ In all likelihood, that does not include the various nonnatural canons ordained by courts.²¹⁵ For courts and Congress to settle into a practice of using these norms to fix and discern legal content would thus be akin to “the trick the emperor Nero was said to engage in: posting edicts high up on the pillars, so that they could not easily be read.”²¹⁶

Even when the parties directly regulated by a statute are sophisticated, moreover, the bootstrapping argument’s departure from ordinary meaning ought to make textualists uneasy — especially when it comes to implied limitation rules. Because these rules allow Congress to convey exemptions on which its members need not directly vote (and, indeed, the very existence of which need not even be acknowledged), they appear at odds with the same accountability-related concerns that leading textualists have invoked in condemning judicial reliance on legislative history.²¹⁷ Congress can give every appearance of imposing some liability on all recipients of federal funds, for example, while actually exempting state governments.²¹⁸ If reliance on legislative history is unacceptable because it allows legislators to effectively make law while ducking tough votes, recognizing a set of counter-textual defaults that Congress can select without openly addressing the matter at all does not seem better.²¹⁹

For all of these reasons, “substantive” canons cannot be squared with textualism by pointing to their alleged nonsubstantive, court-made offshoots. Not only would the bootstrapping argument, on its own premises, cover only a narrow domain, but even within that domain the argument does not hold up to scrutiny. To be sure, we acknowledged in the prior section that some canons traditionally deemed “substantive” might be misclassified: they might capture an inference about communicative content that is drawn from a (genuinely) commonsense understanding of the practical context in which Congress speaks.²²⁰ (Recall our example of the presumption against extraterritoriality.) But when a canon is indeed substantive, it cannot be converted into a determinant of communicative content — a rule of legalese, as it were — simply by being “long indulged”²²¹ by courts. Rather, genuinely substantive canons would need to be justified, and squared with textualism, on their own, substantive terms.

III. Substantive Canons as Constitutionally Inspired Exceptions to Textualism

If substantive canons cannot be recast in nonsubstantive terms, perhaps they can still be reconciled with textualism because — and to the extent that — they derive their authority from the Constitution itself. After all, textualism is an account of how courts should discern and give effect to Congress’s exercise of its lawmaking powers,²²² but Congress’s lawmaking authority is limited by the Constitution — and a court’s duty to faithfully enforce the texts that Congress promulgates is thus limited in the same way.²²³ Substantive canons may not be fatally at odds with textualism, then, as long as they are themselves grounded in what Barrett called (in her influential 2010 article) the courts’ higher obligation to “act as faithful agents of the Constitution.”²²⁴

But grounded in that obligation how? The answer to that question is remarkably elusive, as Justice Gorsuch’s rendition of this argument in West Virginia nicely illustrates. “Like many parallel clear-statement rules in our law,” his concurring opinion declared, the major questions doctrine “operates to protect foundational constitutional guarantees.”²²⁵ He then explained:

One of the Judiciary’s most solemn duties is to ensure that acts of Congress are applied in accordance with the Constitution in the cases that come before us. To help fulfill that duty, courts have developed certain “clear-statement” rules. These rules assume that, absent a clear statement otherwise, Congress means for its laws to operate in congruence with the Constitution rather than test its bounds. In this way, these clear-statement rules help courts “act as faithful agents of the Constitution.”²²⁶

Notice how the key claims here are expressed in terms of “accordance” and “congruence” with the Constitution — language that straddles the line between a looser notion of harmony and a more rigid notion of compatibility. That leaves the argument open to two quite different interpretations. By parsing this passage closely, we can both disentangle these two possibilities and highlight the problematic tendency to run them together.

The first interpretation emphasizes the first sentence in this passage — the one invoking the courts’ “solemn dut[y] . . . to ensure that acts of Congress are applied in accordance with the Constitution.” As Justice Gorsuch surely anticipated, that reads as little more than a bromide about the judicial duty to “disregard[]” a statute when a “conflicting rule[]” imposed by the Constitution so requires.²²⁷ If that is what Justice Gorsuch intended, then applying laws “in accordance with the Constitution” must mean denying them effect if (or to the extent that) they violate the Constitution. And insofar as Justice Gorsuch was saying that clear statement rules “help fulfill that duty,”²²⁸ he would have to have meant that they help to ensure that courts deny effect to unconstitutional laws. Substantive canons would thus “protect foundational constitutional guarantees”²²⁹ in the straightforward sense of reducing the actual incidence of genuine violations of those guarantees.

But the passage is also susceptible to another interpretation. The specific “way” that clear statement rules help a court discharge its duty, Justice Gorsuch said, is by presuming that Congress intends to enact laws that “operate in congruence with the Constitution rather than test its bounds.”²³⁰ That contrast implies that any law that “test[s] [the Constitution’s] bounds” must not be “operat[ing] in congruence with” the Constitution — even though that law might ultimately be within Congress’s constitutional power to enact. Justice Gorsuch can thus be understood as touting substantive canons as tools for promoting this demanding sort of “congruence” — in other words, as backstops or buttresses that “protect” the values served by “foundational constitutional guarantees” against even technically lawful encroachments. This interpretation implies that Justice Gorsuch’s opening allusion to the “solemn duty” of invalidating unconstitutional laws is something of a red herring.²³¹ But Justice Gorsuch’s choice of examples does lend support to this reading, since some of the examples disfavor legal results that, while perhaps not “congruent” with the Constitution in the sense of sharing its scheme of values, Congress has an undoubted power to impose.²³² And his reliance on Barrett’s article cuts in this direction as well: her rendition of the case for (some) substantive canons casts them as measures to “promote[] constitutional values” and ensure that Congress does not “inadvertently exercise extraordinary constitutional powers” that it does possess, rather than merely as means of giving effect to the Constitution’s actual requirements.²³³

For our purposes, figuring out which of these two arguments Justice Gorsuch (or Justice Barrett, or anyone else) actually intended is less important than identifying and distinguishing the arguments themselves. In fact, we suspect that Justice Gorsuch was really giving voice to both of these ideas at once, whether or not he appreciated their distinctness. More broadly, these ideas both seem to fuel the “intuition,” apparently shared by many textualists, that “constitutionally inspired canons” are “more consistent with the principle of faithful agency.”²³⁴ Yet we also suspect that — as often happens in such cases — unappealing features of each idea are being obscured by the convenient presence of the other. We will thus treat the two lines of argument as distinct and take each in turn, even though this has the consequence that our presentation will diverge somewhat from the way that our interlocutors frame their own position.

A. Canons as Enforcing Constitutional Requirements

As we have just explained, one strand in the defense of “constitutionally inspired” substantive canons casts them as means of averting actual constitutional violations. At first blush, that might seem a surprising claim. Suppose, for instance, that a federal antidiscrimination statute would most naturally be read as applying to state governments. If that application of the statute is unconstitutional, a court can simply hold as much pursuant to its power of Marbury-style judicial review. And if extending the statute to state governments is not unconstitutional, then it is difficult to see how nullifying that application — based on constitutional avoidance, a federalism canon, or the like — could possibly be justified as a means of averting a constitutional violation that does not exist.²³⁵ The basic challenge for proponents of this first line of argument is to explain why this apparent dilemma is false.

We can think of three such explanations. First, perhaps the Constitution itself prescribes that only clear or explicit statutes can effect certain changes in the law. Second, constitutional doctrine might properly “underenforce” certain constitutional requirements, leaving a gap for substantive canons to fill. Third, the Court might believe that some mistaken constitutional doctrine should be preserved as a matter of stare decisis; then, too, a substantive canon could serve to conform a statute’s legal effect to what the Court thinks the Constitution really requires. We will consider each of these possible justifications in turn.

1.  Constitutional Clarity Requirements. — Perhaps some norms that are described as substantive canons of statutory interpretation are in fact constitutional rules that tie the extent of Congress’s power to the clarity with which that power is exercised.²³⁶ Consider, as a possible example, the rule that when a federal statute conditions funding to states on compliance with certain requirements, any ambiguities in those conditions are resolved in favor of the states.²³⁷ One explanation for this rule is that, because the Constitution prohibits Congress from imposing the obligations in question on nonconsenting states, the Constitution itself renders all federal conditions ineffective except insofar as they would have been clear to — and thus validly consented to by — the states.²³⁸ The rule of lenity could be justified in a similar fashion, if the Due Process Clause bars the government from punishing conduct whose criminality was insufficiently clear.²³⁹ The major questions doctrine could be susceptible of this kind of defense as well: Just as it is unconstitutional for Congress to delegate a decision to an agency without providing an “intelligible principle,” perhaps it is also unconstitutional for Congress to delegate a major question without making clear enough that it is doing so.²⁴⁰

Our response to this argument is twofold. First, even if there are constitutional requirements of this kind, they are not “canons” in a meaningful sense.²⁴¹ What a court is doing in enforcing such a rule is simply invalidating the statute at issue (or a particular application of that statute) on the grounds that the statute is not clear enough to satisfy a constitutional clarity requirement.²⁴² So the argument does not establish that substantive canons are compatible with textualism so much as it asserts that some rules of constitutional law have been misclassified as interpretive canons.

Second, and more importantly, those who would recharacterize a given substantive canon in this way must justify the supposed constitutional clarity requirement by the lights of their favored theory of constitutional interpretation. They cannot shirk that obligation by framing the requirement as a mere statutory canon: the premise of this defense is that the so-called “statutory canon” is in fact a bona fide rule of constitutional law.²⁴³ Notwithstanding the illustrative examples we noted above, however, we doubt that almost any existing substantive canons could actually be justified on originalist grounds (as most textualist interpreters would require²⁴⁴). For example, while some have sought to defend a robust nondelegation doctrine in originalist terms, we know of no originalist argument that Article I’s Vesting Clause permits major delegations if, but only if, the statutes that make these delegations also make especially clear that they are doing so. In fact, the text of the Constitution imposes hardly any requirements with respect to the form that legislation must take.²⁴⁵ So while the argument from constitutional clarity requirements is both logically coherent and theoretically interesting, it cannot do much real work. Put another way, insofar as some claim that substantive canons serve to avert constitutional violations, they presumably mean that these canons forestall other constitutional violations — not that the absence of a clearer statement itself would render some exercise of Congress’s powers unconstitutional.

2.  Offsetting Underenforcement. — Return, then, to the question of how substantive canons could help to avert actual constitutional violations when a court already has the traditional power of judicial review. At a high level, the answer would have to be that the Constitution imposes more stringent limits on Congress than the courts actually would enforce in the context of Marbury-style review. If that were the case, then even though “constitutionally inspired” canons might be superfluous when it comes to enforcing the constitutional doctrine that the Court applies in the Marbury context, the same canons might nonetheless vindicate the Constitution itself in at least some circumstances where that body of doctrine would not.

Building on earlier scholarship, Barrett sketched an argument of just this sort in her 2010 article.²⁴⁶ As she noted, courts sometimes purposely underenforce a constitutional requirement in the context of traditional judicial review.²⁴⁷ The standard justification for such underenforcement is a felt lack of what the Court generally calls “judicially manageable standards.”²⁴⁸ According to this familiar line of thought, a court should not deny effect to a law — even one the court thinks is unconstitutional — unless the court can also explain that conclusion with reference to a standard that satisfies certain thresholds of intelligibility, predictability, public acceptability, and the like. If courts do restrain themselves in that way, they will leave a gap between what they think the Constitution truly requires and the more modest set of requirements that they actually insist upon.²⁴⁹ “Constitutionally inspired” substantive canons could then avert genuine constitutional violations by, in effect, filling in that gap — not with ironclad constitutional barriers, but with defeasible statutory interpretations.²⁵⁰ This judicial practice would make it more difficult for Congress to enact certain rules and policies that, if push came to shove, the Court would permit. But, the thought goes, Congress can hardly complain about having to jump through extra hoops in order to exercise authority that Congress does not really have at all.²⁵¹

Before we turn to the merits of this argument, note that it (like others we have considered) could only justify a fraction of the Court’s existing practice. As Barrett rightly acknowledged, the Court’s adoption of a clear statement rule often “has more to do with [the Court’s] determination that some constitutional principle merits heightened protection than with a decision to underenforce a norm through judicial review.”²⁵² Congress’s power to preempt state law, for example, is not generally thought to result from any judicial underenforcement — yet the Court still presumes that Congress does not intend to exercise that power (and this presumption is often said to be “constitutionally inspired”).²⁵³ The same goes for Congress’s power to abrogate a state’s sovereign immu­nity when it enforces the Reconstruction Amendments, and so on.²⁵⁴ To be sure, some substantive canons might operate in areas of purposeful underenforcement: the major questions doctrine, cast as a complement to assertedly underenforced limits on legislative delegations, is again a salient possibility.²⁵⁵ But many plainly do not.

Even when a substantive canon might be cast as a complement to underenforcement, moreover, there is another problem that further narrows the argument’s scope — and that might render it untenable altogether, at least for a textualist. The root of the problem is that substantive canons do not (and generally cannot) operate only within the gap left by underenforcement. After all, the very reason for underenforcement is usually that a would-be constitutional limit cannot be specified precisely in the first place.²⁵⁶ As Barrett recognized, therefore, substantive canons actually operate in much the same way as the “overenforcing” or “prophylactic” rules that courts sometimes employ to “implement” the Constitution.²⁵⁷ Much as Miranda v. Arizona²⁵⁸ tells courts to answer a more manageable question about warnings rather than the ultimate constitutional question about compulsion, substantive canons offset underenforcement only by telling courts to answer a different question from the question that the Constitution itself asks.²⁵⁹ And just as Miranda therefore requires the exclusion of some fully voluntary confessions, these substantive canons necessarily instruct courts to deny effect to the law that Congress made — or, at least, the law that a textualist would take Congress to have made — in some cases in which even the true, undiluted meaning of the Constitution permits Congress’s choice. For example, the Court has adopted a strong presumption against reading a statute to regulate in a domain traditionally left to the states, and this rule is sometimes defended as a complement to the Court’s purported forbearance from enforcing the true limits on the commerce power (or the mirror-image protection for states’ rights in the Tenth Amendment).²⁶⁰ But even if that is the canon’s purpose, nobody contends that “the commerce power is wholly devoid of congressional authority to speak on any subject of traditional state concern,”²⁶¹ so the effect of the presumption is to nullify some laws that all would agree that the Constitution did authorize Congress to make.

That means that any substantive canon that is meant to offset underenforcement would need to be justified based on an assessment, presumably undertaken at the level of the particular canon, of the relative costs and benefits of under- and overenforcing the relevant constitutional requirements. Among other things, such an assessment would need to take into account the costs of requiring Congress to “override” a misinterpretation in order to exercise constitutional authority that it legitimately possessed all along — and the substantial likelihood that, for any number of reasons, Congress would fail to do so.²⁶² So, at a minimum, the legitimate sweep of the argument from offsetting underenforcement is further hemmed in, and justifying a “constitutionally inspired” canon in these terms would require the courts to show their work in a manner that they never have.

More fundamentally, we think the very need to resort to such a calculus reveals how awkward — really, untenable — this entire line of argument is likely to be for most textualists. One way to resist the objec­tion that substantive canons necessarily go beyond the Constitution’s requirements — and that they therefore nullify valid statutes — is to resist the baseline of a “Constitution-free” or “otherwise preferred” statutory interpretation in the first place.²⁶³ For textualists, however, that route is unavailable: Insofar as the statute’s legal effect is fixed by its communicative content, the statute’s effect can only be overridden, not shaped, by a “constitutionally inspired” canon. Yet the very idea that a court may knowingly deny effect to a constitutional statute — as long as the judge-made rule pursuant to which the court acts maximizes some conception of net benefits — would strike most textualists as bizarre.²⁶⁴ It may be true, as Barrett says, that this sort of nullification “clips congressional prerogatives much less than [does] overenforcement in the context of judicial review,” because Congress remains free to enact the same rule again more clearly.²⁶⁵ But the real challenge is to explain why any judicial “clipping” of Congress’s constitutional authority is permissible. We do not see how such an intervention could be squared with the “classical conception of the judiciary, . . . especially revered by formalists, in which the court’s job is to simply apply the law in cases that come before it.”²⁶⁶ And, at the end of the day, Justice Barrett may not either: Despite her earlier argument that “the judicial power to safeguard the Constitution can be understood to qualify the duty that otherwise flows from the principle of legislative supremacy,”²⁶⁷ she declined to join Justice Gorsuch’s West Virginia concurrence — and then flatly stated that, “if the major questions doctrine were a newly minted strong-form canon, [she] would not embrace it.”²⁶⁸

3.  Concessions to Precedent. — Still, there is a final variation on the same argumentative theme that merits separate consideration — one that might parry some of the difficulties we have just noted. Suppose that a majority of Justices believes, not that some restraint on Congress should go underenforced for lack of manageable standards, but rather that the Court’s precedents have simply erred in recognizing some power on the part of Congress that it does not really possess. Yet suppose that the same majority believes that stare decisis (or some other felt imperative) makes overruling those precedents inappropriate. In that circumstance, applying a “constitutionally inspired” substantive canon might provide the Court with a middle option: Congress may still exercise the power that the Court’s precedents mistakenly gave it, but Congress must at least do so clearly or explicitly. This argument has much in common with the argument from offsetting underenforcement; in both cases the idea is that, while the resulting canons distort the substance of Congress’s lawmaking, they distort it in the direction of what the Constitution actually requires. But there is an important difference: the “underenforcement” that is being offset here is due not to a lack of judicially manageable standards, but to the Court’s unwillingness to overrule its erroneous precedent. That means that a substantive canon that aims to bridge the resulting gap could, at least in principle, be confined to cases in which the putatively correct constitutional rule would invalidate the statute outright.²⁶⁹ If that constraint were observed, canons justified in this way would not commit the Court to nullifying any laws that it thinks constitutional.²⁷⁰

This precedent-centric argument for “constitutionally inspired” canons is rarely stated plainly — a fact that, as we will see, points toward some of its difficulties.²⁷¹ Nonetheless, we suspect that something like this idea does figure importantly in the real-world motivations of the textualist jurists who embrace such canons. Very often, after all, they are finding “inspiration” for the sorts of restraints on Congress that, to their chagrin, existing doctrine neglects. And while these Justices have rarely spelled out their use of substantive canons to blunt the impact of precedents that they are reluctant to overrule, some scholars have mounted or suggested more explicit defenses of specific canons in similar terms.²⁷²

Although we think this is probably the strongest defense of “constitutionally inspired” canons that textualists could mount, it comes with a heavy burden of justification — one that the Court has rarely, if ever, acknowledged or carried. For one thing, a satisfactory explanation for invoking this theory would have to include an explanation of why, in a given case, the balance of stare decisis considerations tips far enough in favor of an erroneous precedent that the precedent should be preserved for purposes of Marbury-style review, but not so far that the Court should simply take Congress’s apparent invocations of that power at face value. And more fundamentally, a substantive canon could only be justified in this way if the existing constitutional doctrine is wrong and the “shadow” doctrine that one is applying instead (via the statutory canon) is right. After all, a textualist judge is justified in departing from the statute’s ordinary meaning, under this theory, only because the statute, interpreted in that ordinary way, is actually unconstitutional. When it is justified in these terms, then, coining or invoking a substantive canon is not really a way of postponing a full reckoning with a hard constitutional question. It is just a concession that one might make to stare decisis in operationalizing one’s answer to that constitutional question.²⁷³

To be sure, the Court’s textualists could acknowledge all of this and undertake to justify their embrace of one or another substantive canon accordingly. But that would entail forging a majority for a full-blown holding that, under the soundest interpretation of the Constitution, Congress lacks authority to enact some legal rule. Such a determination would presumably come only after adversarial briefing addressed to that merits question. And presumably the majority would need to articulate a doctrinal structure to implement its newly announced vision of the Constitution’s actual requirements as well. Only then, at the last analytic moment, would the majority demote its bottom-line conclusion to a defeasible statutory interpretation in avowed deference to stare decisis.

In practice, we are not aware of any majority opinion that has ever reasoned in this fashion. And, while any explanation of that fact will inevitably be speculative, we doubt that it is an accident. First and foremost, the very same concerns that tell against overruling a precedent will usually also tell against declaring it erroneous — in a majority opinion, at least — while affording it a stare decisis lifeline. A precedent that received that treatment would surely seem marked for a future demise, perhaps after some implicit notice period had elapsed,²⁷⁴ and much of the upheaval (and institutional blowback) that might be triggered by a simple overruling could thus be triggered by the majority’s open proclamation of error, too. From the point of view of the Court’s felt legitimacy and integrity, moreover, it is one thing to decline to revisit an issue due to the weight of stare decisis, but quite another to rule that the Constitution requires X and then go on enforcing not-X indefinitely anyway.²⁷⁵ That is one reason why, as Barrett once observed, the “technique of assuming, and therefore not investigating, a precedent’s validity . . . is a critical means of keeping law stable.”²⁷⁶

Conceivably, the Justices could avoid these problems by reasoning in the fashion that we sketched above, but keeping that logic to themselves. That is, they could: (1) determine that a statute actually would be invalid under (what they take to be) the correct understanding of constitutional law; but then (2) forbear from announcing as much; and (3) cite a hazier, “constitutionally inspired” canon as justification for reaching the same result. In effect, they would then be using ill-defined substantive canons as covers for a kind of “off-the-books originalism.”²⁷⁷ The Justices’ conspicuous obscurity about how “constitutionally inspired” canons are supposed to relate to the actual exercise of judicial review makes plausible that something like this may be afoot (even if they are not fully self-conscious about what they are doing).²⁷⁸ But, in any case, there are (or would be) glaring problems with such a practice. After all, this logic requires the Justices not to articulate the reasons that they actually endorse as legally sufficient to warrant their decisions. That means both that those actual reasons are not exposed to public view and that the Justices are meanwhile spinning out rationales (which then come to bear legal weight) that they do not actually accept as legally sufficient.²⁷⁹ Perhaps most troubling is that, by pushing underground the determination on which the Justices would really be relying — that is, the determination of a statute’s actual constitutionality — this practice leaves the Justices to make that determination in a highly undisciplined way. They are often denied the benefit of on-point briefing, and they are freed of the obligation to answer countervailing arguments, or even to reduce their intuitions to a determinate analysis that will “write.”²⁸⁰

The bottom line is that, if “constitutionally inspired” canons do help the Court to avert actual constitutional violations, they do so mainly by enabling the Court to “make constitutional law on the cheap,”²⁸¹ a pernicious form of “stealth constitutionalism.”²⁸² If the Court were instead required to make its tacit determinations of unconstitutionality explicit, we doubt that it would ever opt for the awkward middle ground that the theory we have described represents. Yet nobody has defended a regime in which the Court routinely makes important constitutional decisions “off the books,” and we doubt that anyone would. Pending such a defense, then, we take the theoretical possibility of substantive canons that are justified as concessions to stare decisis (and hence are compatible with textualism) to be a theoretical possibility and nothing more.

B. Canons as Promoting Constitutional Values

What, then, of the alternative formulation of the “constitutional inspiration” argument that we distinguished at the outset of this Part?²⁸³ Recall that this second variant of the argument casts the same canons as efforts to advance certain constitutional values by disfavoring interpretations that would encroach on them, even though Congress would in no sense violate the Constitution by legislating in those disfavored ways. Despite Justice Gorsuch’s blurring of this line in West Virginia, the two arguments are quite distinct: they envision substantive canons “operat[ing] to protect foundational constitutional guarantees”²⁸⁴ in fundamentally different senses of those words.²⁸⁵ And when the opaque rhetoric is stripped away, both then-Professor Barrett’s article and Justice Gorsuch’s concurrence do seem to rely on the promotion of constitutionally favored values as a distinct justification for substantive canons. For her part, Barrett casts substantive canons partly as “‘stop and think’ measures” that ensure that Congress does not “inadvertently cross constitutional lines or inadvertently exercise extraordinary constitutional powers,” and she deems this judicial check on Congress’s use of its conceded authority potentially justifiable in part because the courts are “act[ing] only in service of values enshrined in the Constitution.”²⁸⁶ And while Justice Gorsuch’s language is more ambiguous, his account of how “clear-statement rules” serve as “corollar[ies]” of “constitutional rules” ultimately trades on the same idea, whether wittingly or not.²⁸⁷

If we put all methodological commitments to the side for a moment, it is easy to see the appeal of accounting for constitutional values in statutory interpretation in this way. What Ernest Young has termed the “binary” model of judicial review — where a court enforces some set of “lines” that the government may not cross — is inevitably somewhat artificial.²⁸⁸ The substantive concerns that justify various constitutional limits are usually not dichotomous; they often phase in over some range, gaining strength in proportion with whatever circumstances implicate them in the first place. At some point in that range — when the facts fall within the “penumbra”²⁸⁹ of a constitutional guarantee, one might say — it could make sense to give effect to Congress’s apparent decision if, but only if, a court is especially certain that Congress intended that result, or has not evaded political accountability for choosing it, or values it highly enough to pay an extra price for it.²⁹⁰ “Constitutionally inspired” substantive canons can thus be understood as “resistance norms” that, as Young explains, “mak[e] it harder — but not impossible — to achieve certain legislative goals that are in tension with the canon’s underlying value,” and these norms, in turn, can be defended as valuable features of a system of judicial review.²⁹¹

But, of course, there is a reason that Justice Gorsuch did not make this argument forthrightly in West Virginia: as a textualist and originalist, he expressly rejects the premises from which it proceeds. As Justice Gorsuch stressed in a recent Fourth Amendment dissent (joined by Justices Thomas and Alito), he believes that the Court is “tasked only with applying the Constitution’s terms”; the Justices “have no authority to posit penumbras of ‘privacy’ and ‘personal security’ and devise whatever rules [they] think might best serve the Amendment’s ‘essence.’”²⁹² “The Fourth Amendment allows this Court to protect against specific governmental actions,” he explained, “and that is the limit of our license.”²⁹³ Indeed, Justice Gorsuch clearly relishes deriding opposing positions as resting on “some penumbra emanating somewhere,” in contrast to a rule stated in some authoritative text.²⁹⁴ So it would be quite something for him to openly assert the authority to deny effect to the law that (under textualist premises) Congress actually made, in the name of furthering some value that “underl[ies]” or is “embodied in”²⁹⁵ a constitutional provision that is, strictly speaking, inapplicable.

Lest we seem to be picking on Justice Gorsuch, it bears emphasis that the basic incompatibility here is closely tied to the core commitments of textualism itself. As Justice Scalia, Justice Barrett, and John Manning have all argued, “the Constitution is, at its base, democratically enacted written law,” and textualists thus ought to “approach[] the Constitution like any other legal text.”²⁹⁶ Just as “[a textualist] hews closely to the rules embedded in the enacted text,” Barrett has explained, “an originalist submits to the precise compromise reflected in the text of the Constitution.”²⁹⁷ From this point of view, the Constitution does not grant any legal status to “federalism” or the “separation of powers” as such; the Constitution simply imposes a suite of legal rules that allocate authority between different governmental entities in particular ways (and that might have been intended by their proponents to advance those abstract values to some degree).²⁹⁸ Whether or not this view is sound as a general matter, we agree with these leading textualists that their textualism appears to commit them to it. But then how can Barrett claim that, when substantive canons “reflect constitutionally derived values,” they “promot[e] . . . norms that have been sanctioned by a super-majority as higher law”?²⁹⁹ To the contrary: what the canons promote has not been sanctioned as higher law, and, as we demonstrated above, what has been sanctioned as higher law — the Constitution’s actual requirements — the canons do not promote. Thus, as Caleb Nelson asks: “Why should the policy behind the Constitution’s specific protections of federalism spill over to justify special rules of statutory construction that are nowhere intimated in the Constitution itself?”³⁰⁰

In fact, this quintessentially textualist line of thought might make even nontextualists uneasy about efforts to justify substantive canons by tying them to “constitutional values.” As Manning has argued especially forcefully, the relevant values in this context are almost always in tension with one another, and efforts to promote one or another of them thus inevitably favor one side of a compromise or trade-off at the expense of the other.³⁰¹ In West Virginia, for example, Justice Gorsuch emphasized that “the framers deliberately sought to make lawmaking difficult” in order to “ensure that any new laws would enjoy wide social acceptance [and] profit from input by an array of different perspectives.”³⁰² The major questions doctrine, he reasoned, protects those objectives.³⁰³ But, as Chief Justice Marshall famously observed, the Constitution also reflects a determination to subject Congress’s exercise of its powers to “no other limits than are prescribed in the constitution,” and thus to “rely solely” on the political checks inherent in “representative government[]” to protect against most possible abuses.³⁰⁴ If one takes the Constitution to “value” the broad discretion it affords to the political branches, handicapping Congress is never an unqualified good from the perspective of “constitutional values.”³⁰⁵ At a minimum, therefore, justifying a substantive canon in these terms requires determining how different values ought to be prioritized relative to one another in the context at hand. And that, in turn, calls for a weighing of either our own values or those reflected in our constitutional practice — an exercise that is probably only obscured and rendered less accountable when it is portrayed as an effort to enforce normative judgments already somehow embedded in the Constitution itself.³⁰⁶

Whether we are right about that or not, much the same point explains why Barrett’s own prescription for identifying the canons with a valid claim to “constitutional inspiration,” advanced in her 2010 article, cannot succeed within textualist parameters.³⁰⁷ In order for a canon to be legitimate, she suggested, the Constitution must be “best understood” as “biased” against the same exercises of power that the canon resists.³⁰⁸ So, for example, the federalism canons could not qualify a textualist’s ordinary obligation of faithful agency “[i]f the Constitution is best understood as neutral or favorable to federal preemption of state law,” rather than as inclined against it.³⁰⁹ But perhaps textualists’ most forceful insight — and, in any event, one of their basic commitments — is that when lawmakers negotiate a compromise among conflicting values, the question whether the resulting text is “biased” in a given direction lacks a meaningful answer. Indeed, Barrett’s claim that the validity of constitutionally inspired canons “depends upon one’s view of the substantive meaning of the [constitutional] provision involved” cannot be squared with textualists’ understanding of what legal language means.³¹⁰ The “substantive meaning” of the Supremacy Clause, a committed textualist would say, is simply that federal law preempts inconsistent state law; its meaning does not incorporate anyone’s real or imagined enthusiasm or reluctance about the power that the clause confers. And that leaves interpreters with no basis for concluding that, in resisting some exercise of federal power, they would be going with the grain of the Constitution rather than against it. From a textualist’s point of view, the Constitution has no grain at all.

C. A Note on Historical Conceptions of the Judicial Power

Textualists have one final card to play. In a striking passage in the same article, Barrett forthrightly acknowledged that because “[c]onstitutionally inspired canons . . . constrain Congress more than the Constitution does,” they are “in tension with the structural limitations upon statutory interpretation,” and that her defense of them thus could not be “rock solid.”³¹¹ In fact, she said, “[i]t would be reasonable, and as a formal matter, perhaps more satisfying, to take the position that overenforcing canons of construction exceed the limits of judicial power.”³¹² So far, so good. Yet she then pointed out that “[t]he federal courts have long asserted the power to employ language-pushing substantive canons,” and she concluded that this venerable “pedigree” favored construing the Article III “judicial Power” as including the power to create and apply such canons.³¹³ “If the Marshall Court thought itself empowered to adopt the Charming Betsy and avoidance canons,” she reasoned, “that is evidence that it believed that ‘the judicial Power’ encompassed the authority to do so.”³¹⁴ Along the same lines, Justice Gorsuch’s West Virginia concurrence asserted that the Court had acted permissibly because “our law is full of clear-statement rules and has been since the founding.”³¹⁵

There are two problems with this appeal to historical practice. The first follows directly from our earlier point about the sheer impossibility, from a textualist’s point of view, of interpreting the Constitution as expressing a residual favoritism toward some of its provisions (or the values they serve) over others.³¹⁶ Even if the Constitution authorizes the use of substantive canons to further penumbral rights or limitations, a textualist judge will simply have no way to exercise that authority. It is telling in this regard that Barrett failed to give a single example of a canon that, in her considered judgment, actually is rendered legitimate by its alignment with the Constitution’s normative thrust; she said only that “such an inquiry must be undertaken.”³¹⁷ If that inquiry is undertaken within textualist parameters, we do not see how it could ever bear fruit.³¹⁸

Second, the appeal to historical practice cannot shore up Barrett’s position if the reason that courts have long employed substantive canons is that the courts have long operated from premises that textualism repudiates. In that event, past practice would lend support to substantive canons only because, and to the extent that, it undercuts textualism itself. And, indeed, Barrett’s historical study of canons of construction in the early Republic shows that these interpretive norms were justified largely on grounds that are flatly at odds with modern textualism.³¹⁹ We fail to see why one would take from this history the lesson that “formulating and applying substantive canons is a permissible exercise of the judicial power,” as opposed to, say, “favoring statutory interpretations that further widely shared societal values is a permissible exercise of the judicial power.”

In the end, even Barrett seemed to place relatively little stock in the historical practice on which she relied. The canon-making power that she found in early cases, after all, was not conditioned on identifying “constitutional inspiration” for a given canon.³²⁰ Yet she still would throw overboard all canons that place “extraconstitutional values” above “the best interpretation of a statute”; despite their historical pedigree, she reasoned, these canons are not compatible with the textualist conception of faithful agency.³²¹ As we have explained, the so-called “constitutionally inspired” canons are also squarely at odds with that conception. There is no reason why history should trump the merits in one context but not the other.

IV. Substantive Canons as Last Resorts

We now turn to the third and final family of explanations as to why some substantive canons might be consistent with textualism’s core commitments. Broadly speaking, the idea here is that substantive canons are permissible only when a judge has first tried and failed to resolve a question based on the statute’s communicative content alone. This claim, too, is susceptible of multiple interpretations. In one form, it addresses cases in which the statute’s communicative content, although expressed clearly, simply does not prescribe a determinate resolution of the case at hand. In another, more consequential form, the argument qualifies the fidelity principle by authorizing judges to make interpretive close calls on substantive grounds.³²²

We will unpack and assess these possibilities in turn, but we first note an important limitation that pertains equally to both. Substantive canons are often applied in cases where, but for the canon, the proper textualist resolution of the case would be clear. This is especially common in cases involving “clear statement” or “implied limitation” rules, at least where they function as such. To return to one of our earlier examples, we very much doubt that it is indeterminate or uncertain whether, as a matter of ordinary meaning, states that receive federal funds would be counted as “recipient[s] of Federal assistance.”³²³ To take another example, we think the Court’s stated rationale for invalidating the Biden Administration’s vaccine-or-test requirement would collapse under this approach as well, though the explanation (like the statutory text) is admittedly more complex.³²⁴ In short, even if one of the arguments we will now consider were valid, this would not vindicate a sizable swath of existing practice.

A. Construction

As we noted in our opening discussion of textualism and “ambiguity,” judges often face cases for which lawmakers could not have thought they had prescribed any determinate resolution at all.³²⁵ This sort of legal incompleteness arises largely because, even when the statutory communication is perfectly successful, the concepts successfully expressed are vague, in the sense that they admit of borderline cases,³²⁶ or “open-textured,” in the sense that they comprise multiple dimensions that lack common units.³²⁷ A statute that assigns consequences to “neglect,” “violence,” or “safety,” for example, cannot possibly convey a complete mapping from all possibly relevant states of the world to the consequences that they are supposed to have.³²⁸ As a result, a judge faced with a statutory case involving one of these concepts will often have more work to do, even once all of the “interpretation” — in the sense of recovering the content encoded by language — is done. This work, as earlier noted, is often denominated as the distinct task of “construction.”³²⁹

Because textualism can neither guide nor constrain that inherently nontextual work, it follows straightforwardly that a judge’s commitment to textualism would not prevent them from turning to substantive canons to carry it out. In fact, any norm or consideration that a judge might take into account at the construction stage would qualify as “substantive” in our sense, since it could not be casting further light on the statute’s already-known communicative content.³³⁰ Still, we think the necessity of looking beyond the text when making judgments within such a “construction zone”³³¹ does not actually create much of an opening for textualists to employ substantive canons as conventionally understood.

For one thing, we have struggled to identify any real-world cases employing a substantive canon to guide the application (or necessary elaboration) of the content that, on nonsubstantive grounds, Congress is taken to have expressed. That makes sense: canons generally take the form of priority rules (roughly, “prefer interpretations (or norms) that have property X over those that have property Y”), and the drawing of a line in a continuous spectrum — as one must in settling what is “reasonable,” “substantial,” or the like — does not frame that sort of choice in the first place. To be sure, courts sometimes invoke implied limitation rules to lop off some applications of textually expressed concepts. A court might hold, for instance, that an agency’s general authority to impose “reasonable” requirements is not sufficient to authorize rules that intrude on state authority in particular ways. But the Court also imposes such limitations when the concepts at issue are not vague or open-textured in any relevant respect.³³² By all appearances, then, what the Court is doing when it deploys substantive canons of this sort is not answering the question that it takes to be framed by the textually expressed concept (such as, “Is imposing this particular requirement reasonable?”), but rather imposing an extrinsic restriction on the statute’s invocation of that concept, one at odds with the unrestricted content expressed by the text. The necessity of construction does not license that sort of subtraction any more than it would license the addition of applications wholly outside the statute’s communicative content.

Indeed, in virtually all cases, when the Court employs substantive canons, it treats them as bearing on what Congress should be deemed to have said — which is precisely the territory also claimed by genuine textual interpretation. Often, some word in the statute is polysemous, and the question is in which sense (or with what subset of its semantic range) Congress used that word.³³³ West Virginia v. EPA is again a salient example.³³⁴ When Congress referred to the “best system of emission reduction,” did it mean a “system” in the more concrete sense of a technology or apparatus, or did it mean any “system” in the broad sense of “an organized and coordinated method; a procedure”?³³⁵ On its face, that is a question about the statute’s communicative content (hence the dueling definitions); it is not a question about what some particular content that the statute clearly conveyed should be understood to entail with respect to generation-shifting.³³⁶ So the Justices who invoked the major questions doctrine here (and understood it as a substantive canon in our sense) were not using it to flesh out the implications of the statute’s known communicative content, but to supplant that content (whatever it might have been) with a legally imputed one. In this case as in nearly all others, then, the canon was not being deployed within a “construction zone” (at least of the sort that we described above) at all.³³⁷

Of course, the fact that substantive canons are not used as aids to translating communicative content into legal effect does not mean that they could not be used in that way.³³⁸ And as we have said, something substantive is bound to happen in these cases regardless. For present purposes, therefore, our main point is simply that the inevitability of some textually ungoverned judgments in these cases should not be mistaken as supporting textualists’ current practice of employing substantive canons. Still, it bears noting that the case for using anything like the existing crop of substantive canons as construction tools would not be clear-cut. We doubt that this hodgepodge of judge-made, off-the-rack normative principles would fare better in a textualist’s evaluative ranking of such tools than, say, trying to resolve a matter in a way that furthers the objectives that a provision was apparently designed to serve,³³⁹ or, when possible, simply deferring to the judgments of politically accountable decisionmakers.³⁴⁰ In any case, the problem of giving legal effect to clearly expressed but underdeterminate content is not really where the action is when it comes to substantive canons today.

B. Managing Uncertainty

The more important question, then, is whether resorting to substantive canons can be justified in cases where the very communicative content of a statute is uncertain. We noted early on a difficulty for this kind of argument: for understandable reasons, textualists do not abandon their textualism merely because a case is “hard” or presents a “prima facie ambiguity.”³⁴¹ Still, perhaps there comes a point where the textual evidence is so nearly balanced that even a committed textualist could reasonably look elsewhere. In this spirit, Caleb Nelson suggests that “[i]n cases of genuine ambiguity (that is, cases in which our primary interpretive tools have simply identified a range of possible meanings, none of which is significantly more likely than the others to reflect the enacting legislature’s intent), the textualists’ general view of legislative supremacy does not rule out reliance upon normative canons.”³⁴²

As Nelson recognizes, this formulation “conceals some difficult questions.”³⁴³ Most importantly, “[h]ow big a gap must exist between the leading interpretation and the next most likely alternative for the Court to say that the statute permits only one [interpretation]?”³⁴⁴ On the one hand, if substantive canons were confined to cases that textualist judges experience as genuine “ties,” there would arguably be nothing nontextualist about resorting to them — but such cases of perfectly matched arguments are so unusual as to be practically irrelevant.³⁴⁵ On the other hand, if the text yields to substantive canons whenever multiple interpretations are plausible, textualism will turn out to mean relatively little in hard cases after all. Indeed, as then-Professor Barrett argued, any departure from what the judge takes to be the likelier candidate for the statute’s communicative content would seem to “conflict with the obligation of faithful agency” (at least unless the canon at issue serves a “constitutional value”).³⁴⁶

Formally, we think that Barrett is probably right about this (apart from the “constitutional values” proviso, that is), but the question remains whether textualists could reasonably moderate their approach, consistent with the values animating their textualism, to rely on substantive canons at least in very close cases. The thought here would be that, at least when two alternatives entail nearly equal risks of error, one ought to minimize expected error costs instead. For example, mistakenly deciding that an agency does have the power to take some major action might seem much worse than mistakenly deciding that it does not have that power. In that event, a court might “play it safe” by holding the action unauthorized, even if this means modestly increasing the chance of an incorrect decision.³⁴⁷

This sort of thinking is familiar and appealing in other domains of life. Suppose the commander of an artillery unit stationed at a tense international border receives a garbled radio communication from headquarters, which either said to fire on the enemy positions across the border or said to hold her fire, despite recent provocations.³⁴⁸ After listening to the distorted message carefully, the commander concludes there is a 51% chance that headquarters ordered her to open fire, and a 49% chance that headquarters ordered her to hold her fire. But the commander knows that mistakenly opening fire will start an avoidable war, while mistakenly holding fire will have only a minor adverse effect on her side’s strategic position. Even if the commander cares exclusively about serving as the faithful agent of her generals, she might properly decide not to open fire in this scenario. Though a dutiful soldier, she also applies a kind of ambiguity-dependent substantive canon: Do not take action that will start a war unless the orders to do so are clear and unmistakable.

But any textualist who sought to justify their resort to substantive canons on these grounds would face daunting obstacles — and we doubt that their textualism would remain recognizable at the end of the journey. The first problem is who gets to decide what constitutes “playing it safe.” In general, an agent might treat errors in one direction as much worse either because of a conjecture about what the principal values or because of her own values. (In our example, the artillery commander might reasonably presume that her superiors view the costs of erroneously starting a war as greater than the costs of a delay in opening fire, or she might hold her fire until she gets clearer orders because she holds that view.) Under textualists’ conception of the judicial role, it is hard to see how the judge’s (or even the judiciary’s) own normative judgments about error costs — reflected in substantive canons of the “superimposed value” variety — could properly form the basis of their decisions. Who empowered Article III courts to decide, say, that mistakenly upholding the Clean Power Plan (thereby aggrandizing the federal bureaucracy) would be worse than mistakenly invalidating it (thereby interfering with efforts to thwart a global catastrophe)? Moreover, the question is particularly awkward for courts because, in either scenario, the main determinant of the “error costs” will be the costs (or benefits) of the result itself, not any costs specific to having brought it about mistakenly. Indeed, the balance of expected costs could favor an interpretation that almost certainly does not capture the statute’s communicative content, for the simple reason that this content was a very bad idea.

In light of all this, the relevant benchmark would probably have to be the enacting Congress’s (imputed) valuations of different mistakes, rather than the court’s own. In principle at least, substantive canons could perhaps be reglossed as entrenched generalizations about that issue. For example, the Charming Betsy canon would be justified (from this perspective) if Congress generally prefers that, when courts lack confidence in any proposed interpretation of a statute, they prioritize not misinterpreting it in a way that violates international law. We are skeptical that this way of thinking can be squared with widespread textualist premises about the nature of multi-member legislatures, however.³⁴⁹ (What does it mean to say, apart from any text enacted by Congress, that Congress “generally prefers” one thing or another? How could textualists square a practice of allowing these unenacted preferences of legislators to shape a statute’s legal effect with the promulgation principle?) Because existing substantive canons were not developed or justified with this function in mind, moreover, there is no particular reason to see them as well-suited to that task.

Even apart from all of that, we expect that textualists would find this line of thought hard to accept because of the standardless discretion it affords to courts and the unpredictability it would inject into the lawmaking process as a result. The judgment of whether a statute is “ambiguous” is notoriously difficult to regiment or discipline — yet, with these error-cost canons in effect, it would become a ticket to flipping the apparent meaning of a statutory text.³⁵⁰ Meanwhile, legislators seeking to bargain over matters of policy would face a complex and imbalanced set of incentives: nobody could rely on the expectation that courts would adhere to what even the courts themselves thought the words more likely expressed; moreover, savvy proponents of some outcomes would know that even language that tips modestly in the other direction would (probably) actually be construed in their favor, yet ordinary citizens presumably would not.³⁵¹

None of this amounts to a deductive argument that nobody claiming the mantle of textualism could possibly afford substantive canons any role as tools for managing uncertainty about a statute’s communicative content.³⁵² But we think it is fair to say that the thrust of such a move would be contrary to the vision of judging and of statutory interpretation that has animated modern textualism thus far, and we do not see how this line of argument could vindicate either the present use of substantive canons or anything much resembling it.

V. Implications and Conclusion

We began our analysis by highlighting what appears to be a fundamental incompatibility between the tenets of textualism — or at least the “hard” textualism to which a majority of the Court officially subscribes — and the use of substantive canons to resolve statutory cases. That tension, though sometimes ignored or obscured, is straightforward: If substantive canons are justified, it is for reasons that textualists otherwise rule out of bounds.³⁵³ These canons might illuminate what Congress would or should have said if only it had foreseen certain circumstances — for example, a conflict with international law, or the pros­pect that the Clean Air Act could be used to transform the electricity sector — that Congress actually failed to anticipate.³⁵⁴ Or the canons might push the law in the direction of values and policies — from lenity to federalism — that the courts deem worth promoting even if Congress cares little for them.³⁵⁵ Either way, these substantive canons are flatly at odds with textualists’ insistence that “[o]nly the written word is the law”³⁵⁶ and that judges ought to enforce the statutes that Congress actually enacted.³⁵⁷

Although we have tried to put the conflict especially sharply and precisely, this is not really news. Yet the contradiction is usually framed as a “tension” or cause for “discomfort” — a kind of awkward fit that might yet be managed through one or more theoretical maneuvers.³⁵⁸ The main contribution of this Article has been to provide a thorough and — we hope — fair evaluation of each of those efforts to date. What that analysis shows, we submit, is that the jury is not really out on this question: None of the proposed reconciliations can justify anything like the current practice among the Court’s textualist Justices. In a few instances, to be sure, we have identified conceivable justifications that could square at least some kind of substantive canon with textualism. But in each of these cases, the potential justification identifies a mere possibility space, with little reason to think that it is occupied (let alone as crowded as vindicating current practice would require). In fact, the justifications that appear most coherent — such as the argument for using canons to moderate the effect of stare decisis, or to guide “construction” in the absence of textual constraint — are ones that have received little defense, perhaps because these arguments would justify few existing canons and would entail other commitments that many textualists would find unattractive.

How might a self-identified textualist jurist (or scholar) reply to our critique? One possibility, of course, is that there is a flaw in one or more of our arguments, or that we have overlooked a potential reconciliation. We would be happy if textualist theorists took this Article as a friendly but pointed challenge to offer a rebuttal. Another possibility is that the handful of possible arguments we acknowledge as coherent might be explicitly embraced and expanded to cover a broader domain. As we have just noted, however, we are dubious that most textualists would go this route. The only remaining possibility, and the one we view as most likely, is that a principled textualist would need to concede the contradiction. Having made that concession, textualists would then be faced with two options.

First, they might abandon substantive canons, or at least restrict their use to the very narrow set of circumstances in which one of the justifications we have considered can actually be sustained. In other words, a principled textualist might respond to the challenge that Justice Kagan posed in her recent dissents — that her colleagues have taken to using substantive canons as “get-out-of-text-free cards” — with contrition. They might commit to relegating substantive canons to the dustbin of history, just as they largely succeeded in doing with legislative history. This approach not only would preserve textualism’s philosophical purity by purging it of an embarrassing inconsistency, but might meaningfully advance the values that make textualism attractive to many in the first place.

For many textualists, however, preserving textualism’s integrity in that way would come at too steep a cost. Realistically, textualist judges would lose a valuable tool for furthering a favored vision of the law. As we have noted, substantive canons appear particularly attractive as a way of simulating the constitutional limits — say, in the name of federalism and nondelegation — that many jurists and scholars think are given short shrift by current constitutional doctrine.³⁵⁹ Furthermore, while existing statutory precedents decided partly on the basis of substantive canons would presumably survive on stare decisis grounds, suddenly ditching the whole collection of substantive canons would be a substantial disruption. For both principled and other reasons, therefore, even if interring substantive canons would be a salutary change, it seems unlikely.

Barring such a change, the remaining course would be to acknowledge that textualism as practiced is a good deal less pure and more complicated than textualism as advertised. If avowed textualists give weight to substantive canons, they are already pluralists about statutory interpretation; they do not actually think that “[t]he text of the law is the law”³⁶⁰ or that “[f]idelity to the law means fidelity to the text as it is written.”³⁶¹ And once that is brought to the surface and conceded, the question of what ought to figure in the pluralist mix, and why, is squarely posed.³⁶² Perhaps courts ought to apply substantive canons that reflect the normative aspirations of the majority of the political community as it exists today, whatever those might be, rather than those of generations (or judges) past. Or perhaps courts should give more weight to what, with the benefit of “policy context” and all, it seems likely that a given statute’s proponents intended it to do.³⁶³ Our point here is not to advocate any particular reforms to statutory interpretation. But we think that if textualists maintain their attachment tosubstantive canons — notwithstanding those canons’ evident incompatibility with textualism — they have opened the door for a wide-ranging discussion of what other nontextualist considerations courts ought to entertain as well.

---

* Professor of Law, Harvard Law School.

∗∗ Henry L. Shattuck Professor of Law, Harvard Law School. For valuable comments anddiscussion, we thank Mitchell Berman, Ryan Doerfler, Richard Fallon, Sherif Girgis, Abbe Gluck, John Goldberg, Jonathan Gould, Mark Greenberg, Tara Grove, Emma Kaufman, Anita
Krishnakumar, Daryl Levinson, Jamie Macleod, Neysun Mahboubi, John Manning, Andrei Marmor, Larry Solum, Cass Sunstein, Chris Walker, and Bill Watson, as well as participants in the Harvard Law School Faculty Workshop and the NYU Colloquium on Constitutional Theory. We also thank Thomas Nielsen and Thomas Koenig for excellent research assistance.