Many textualists see canons of interpretation as a means to deal with statutory ambiguity, while nontextualists are more likely to turn to legislative history. This Essay explores some of the problems with each approach: for canons, determining which context is the best one to analyze, and for legislative history, determining which statements are reliable and which are hot air.

A conference about “best practices” for legal inquiry supposes that there are practices. In the field of legal interpretation, that assumption is doubtful. The Supreme Court is dominantly textualist (perhaps Professors Abbe Gluck and Frederick Schauer would prefer formalist) but has not insisted that all other federal judges use the same approach—and the Justices themselves happily sign pragmatic opinions written by Justice Breyer. They never use legislative history to contradict a statute but sometimes use it to confirm statutory meaning arrived at, they tell us, by other means. No Justice these days is a purposivist after the fashion of the old approach of Professors Henry Hart and Albert Sacks.1 All are delighted to say that the question is not what direction the legislators wanted to go in, but how far, and that you can’t tease that out of a purpose.2 The law is a multidimensional vector, not an arrow in one dimension. At the same time as the Justices tell us to pay heed to the “intent” of Congress, they concede that “intent” is empty and that meaning is objective, if only because we can’t peer into the heads of the median legislators, who usually are silent, even though it is the median voter whose views ought to matter. If we do not have a dominant practice of interpretation, we cannot think about how to test claims.

My goal today is to discuss how textualists approach the subject—or at least how I do. Intents are irrelevant even if discernable (which they aren’t), because our Constitution provides for the enactment and approval of texts, not of intents. The text is not evidence of the law; it is the law. I am leery of pragmatic arguments because they simultaneously depart from the enacted texts and give too much power to judges. That’s fine in a common-law world but not in the domain of statutes and regulations. When texts run out of meaning, we should put them down and go to other sources of law, rather than invent things in their name. I’ve made that argument at length elsewhere and won’t repeat it here.3 This is an argument about jurisprudence rather than linguistics, and I’ve been trying for thirty-one years on the bench to show how it can be done. I leave to others an evaluation of what this experiment shows.

Having worked on textualist interpretation for a long time, I may have seemed to the organizers of this Symposium a good person to describe how the process works—to produce an interpretive algorithm that then may be evaluated free of the problems that arise when evaluating the results of concrete cases. Every case presents a real problem, and people have preferences over the outcomes independent of how the judges produced those outcomes. It would be best to pare away the outcomes and concentrate on the theory of meaning that generated them.

The problem is that even textualism, which is relatively constrained compared to pragmatism or purposivism, does not have an algorithm. A short time ago, Justice Scalia and Bryan Garner produced a book on textualist method that set out fifty-seven approved canons,4 and repeatedly the book says that one or another canon is not definitive but that context matters. No algorithm there—nothing a well-designed expert system could use to replace judges with computer code and a big database. Computers have triumphed at chess, Go, and Jeopardy, but not at the game of legal interpretation.

I want to illustrate my point about the absence of method by looking at two issues. First, can an apparently clear canon, the rule of the immediate antecedent, generate predictable outcomes? Second, can a more sophisticated use of legislative history, stressing how Congress actually behaves, do so? Gluck’s essay for this Symposium champions “The CBO Canon” as such an approach,5 and I have a good deal of sympathy, because meaning depends on how an interpretive community reacts to words and the Congressional Budget Office lies outside the legislature. But instead of analyzing her proposal, I look into the older proposal that judges search for costly promises and ignore the rest.

So I start with a canon. I am skeptical of canons, and not simply because of Professor Karl Llewellyn’s famous list of canons and countercanons.6 One could revise the list, as Justice Scalia and Garner set out to do, to avoid the direct conflict. My concern is different: every canon implicitly begins or ends with the statement “unless the context indicates otherwise,” which potentially leaves so much room for maneuver that the canon isn’t doing much work. Indeed, a reference to “the context” does not even pin down what context. Is it the immediate linguistic context? The context of the whole statute, à la Professor Akhil Amar’s intratextualism?7 Does it matter if the statute was cobbled together by very different legislatures over forty-plus years, like the Clean Air Act? Perhaps it means the economic context. And if it means that, can’t a judge do anything that would be appropriate in a common-law case?

The United States Code has quite a few explicit context clauses, including the one that begins 1 USC § 1, the Dictionary Act. The Justices once split five to four about whether this clause means linguistic or economic context (the five voted for linguistic).8 The definition of “securities” also contains a context clause,9 and in securities cases the Justices have used the economic context,10 without explaining why or trying to justify what amounts to a policy-making role, which definitions, of all things, are supposed to deny to the judiciary.

Still, canons are inevitable, because all language depends on them. Canons tell us how to fill in gaps, and all language has gaps. Ludwig Wittgenstein tells us that gaps are inevitable because new cases always arise and speakers can’t give themselves “ought” statements.11 But forget the philosophy of language and ask how we speak. To take an example from a case decided recently,12 if a friend asks you to introduce her to “an actor, director, or producer involved with the new Star Wars movie,” you know immediately that the Star Wars clause modifies all three terms; an actor from an Ingmar Bergman film would not do. This is the principle of parsimony; people use the shortest expression that will convey meaning, because going on at length is hostile.13 What would you think of a person who said “an actor involved with the new Star Wars movie, or a producer involved with the new Star Wars movie, or a director involved with the new Star Wars movie”?

Speakers depend on canons and contexts that listeners will supply, else expression would be interminably long. Imagine a short law: “Anyone who intentionally kills someone else must be put to death.” Every reader knows that this does not mean that a cop on the street can shoot a murderer with impunity. Nor does it mean that the prosecutor must pursue every reported crime. The statute supposes a legal process that will determine who is a murderer. It supposes (or at least does not foreclose) the possibility of defenses, such as self-defense or necessity. Only by making assumptions about how other statutes, and the legal system as a whole, interact with a text is it possible to adopt manageable rules.14 And this means that judges must apply canons and background norms, just as everyone else does. The problem lies in indeterminacy.

The Star Wars hypothetical comes from Justice Kagan’s dissent in Lockhart v United States,15 and I use that case and a predecessor, Paroline v United States,16 to illustrate the difficulty of identifying the appropriate context. Both cases involve the canon usually called the rule of the immediate antecedent, together with related canons that Justice Scalia and Garner call the “Series-Qualifier Canon” and the “Nearest-Reasonable-Referent Canon.”17

These three canons start from the linguistic understanding that, in a list of disparate items, a clause modifies the item nearest to it. To take an example from Barnhart v Thomas,18 if a parent warns a child that a house party or any house-damaging activity will lead to discipline, this means that a party is a cause of discipline whether or not the house is damaged. The parent likely wanted to pick the biggest risk (which might entail drinking) for separate treatment, followed by a general rule. If the damage clause modified both any party and any other activity, then the reference to the party would be pointless, and the parent would just have said, “If you damage the house while we’re out, you’ll be grounded.”

Simple enough in principle. Now try to apply it. Notice the statement “in a list of disparate items.” If you say “statutes, regulations, or Constitution of the United States,” the phrase “of the United States” applies to all three items; that’s how normal people using a principle of parsimony would hear it. They are not disparate. Fine. This sets up Lockhart.

Section 2252 of Title 18 covers some forms of sexual exploitation of minors. Subsection (a)(4) deals with possessing or distributing sexual images of minors on Indian reservations, military bases, and other places quaintly called “the special maritime and territorial jurisdiction of the United States.” Section 2252(b)(2) provides a minimum ten-year sentence for a person convicted under subsection (a)(4)

if such person has a prior conviction under this chapter, chapter 71, chapter 109A, or chapter 117, or under section 920 of title 10 (article 120 of the Uniform Code of Military Justice), or under the laws of any State relating to aggravated sexual abuse, sexual abuse, or abusive sexual conduct involving a minor or ward, or the production, possession, receipt, mailing, sale, distribution, shipment, or transportation of child pornography.

The question in Lockhart was what the phrase “involving a minor” modifies—just the immediately preceding phrase “abusive sexual conduct” or the whole string “aggravated sexual abuse, sexual abuse, or abusive sexual conduct.” Avondale Lockhart had a prior conviction for sexual abuse, but that conviction did not stem from a crime against a minor, and when he was convicted of the child-porn offense under § 2252(a)(4), he opposed the prosecutor’s argument that the minimum sentence was ten years.19

He lost in the Supreme Court—and don’t assume that this was some ideological fight. Justice Sotomayor wrote for the majority and Justice Kagan for the dissent. Both opinions are textualist. No one asked what Congress ought to have done or intended; everyone wanted instead to know what the words meant. But how does one parse those words?

All eight participating Justices (Justice Scalia had died shortly before the opinion’s release) accepted the immediate­antecedent canon and its relatives. The opinions proceed as if the Scalia-Garner book, Reading Law, had been enacted as an addendum to the Dictionary Act. But the Justices could not agree on whether the list “aggravated sexual abuse, sexual abuse, or abusive sexual conduct” contained like items or unlike items. Justice Kagan thought the items alike, so canonically the qualifier “involving a minor” applied to all three. The majority thought them unlike, and to make that point the majority invoked context.

Fine, but what context? One context was that the series “aggravated sexual abuse, sexual abuse, or abusive sexual conduct” seems to have been lifted from the titles of several statutes, two of which are not limited to minors. The dissent replied that this may show only that the mandatory minimum applies to the subset of the underlying conduct that does involve minors. We have a draw.

The majority chose, as an additional context, the fact that the list of statutes preceding the state-law portion identifies some federal statutes by name, and the “involving a minor” qualifier does not apply to them. What sense would it make, the majority asked, to have a ten-year minimum when a prior federal felony has an adult victim, yet to limit the enhancement to state crimes with minor victims? To this the dissent replied that the majority had slipped from the linguistic to a functional context. The state-law phrase poses a problem that enumeration of federal statutes does not. Congress can identify qualifying federal statutes by citation; that won’t work with state laws. A drafter is charier about a clause with generalities, because who knows what some state might call “sexual abuse”? Far better to apply “involving a minor” across all state laws than to take a chance. I make this another draw.20

I am attracted to the dissent’s approach but get nervous because Justice Kagan wrote (as Justice Scalia often did) about the need to interpret federal statutes as ordinary readers hear the language.21 The problem is that ordinary readers are dealing with newspapers, not statutes. Most norms of ordinary linguistic interpretation assume cooperation between the speaker and the hearer. That’s what gives Justice Kagan’s Star Wars hypothetical its punch. But statutes are written by noncooperative actors; they may reflect compromises between violently opposed blocs in Congress, and this law was designed to control the conduct of people who really want to have child porn. Using the norms of private speech to tell us what legislative texts mean is problematic. Yet we lack any special set of interpretive norms for legislative speech.

Justices sometimes tell Congress how to write, if they want to reduce the vagary of interpretation. Gluck and Professor Lisa Bressman find that legislators don’t pay attention,22 though the sort of problem exemplified by Lockhart is one covered by the legislative drafting manuals in the House and Senate. The problem is solvable by dividing texts into subparts.23

Imagine two possibilities:

the laws of any State relating to (a) aggravated sexual abuse, (b) sexual abuse, or (c) abusive sexual conduct involving a minor or ward.

That form of division makes it clear that “involving a minor or ward” deals only with division (c). Or consider a different option:

the laws of any State relating to

  • aggravated sexual abuse,
  • sexual abuse, or
  • abusive sexual conduct

involving a minor or ward.

By indenting the three subclassifications, and bringing the “minor or ward” language back to the margin, the formatting tells us that the “minor or ward” qualifier applies to all three.

Garner and the drafters of the House and Senate manuals would not be happy with this indent-and-return-to-margin approach, and not only because it makes citation harder.24 And Gluck would not be happy, because drafters ignore the manuals—and, worse, formatting sometimes is applied after the text has been voted on.25 But at least a court could say that if drafters do these things, they get the expected result, and if they don’t, then judges make educated guesses. They just lack a standard method for educating those guesses.

But even this turns out to be too simple. Consider 18 USC § 2259, another statute in the burgeoning list of federal sex crimes. Section 2259 says that a sex offender must make restitution for the full amount of a victim’s losses. Section 2259(b)(3) is a definition clause. It reads:

For purposes of this subsection, the term “full amount of the victim’s losses” includes any costs incurred by the victim for—

(A) medical services relating to physical, psychiatric, or psychological care;

(B) physical and occupational therapy or rehabilitation;

(C) necessary transportation, temporary housing, and child care expenses;

(D) lost income;

(E) attorneys’ fees, as well as other costs incurred; and

(F) any other losses suffered by the victim as a proximate result of the offense.

This time Congress did what I just said would clear everything up. It broke the list into separately numbered sections. So we have, for example, “(D) lost income” with no qualifier and “(F) any other losses suffered by the victim as a proximate result of the offense.” Yet, by a vote of five to four, the Court held that “as a proximate result of the offense” modifies all six things in this list.26 Again, this was not the five you would expect. The majority was Justices Kennedy, Ginsburg, Breyer, Alito, and Kagan; the four dissenters were Chief Justice Roberts and Justices Scalia, Thomas, and Sotomayor. This not only is not the newspapers’ liberal/conservative divide but also does not match the division in Lockhart, in which Justice Sotomayor wrote for the majority and Justice Kagan for the dissent. Justices Alito, Kennedy, and Ginsburg were in both majorities, and no one was in both dissents.27 This is the work of a Court trying hard to interpret enacted texts, not to carry out an agenda—but why didn’t they take the cue from the statutory numbering scheme?

The answer may be that no one asked them to do so,28 and even at the Supreme Court the advocates’ presentations can knock out valuable options. The second and more important reason is that the context of the legal system has its own consequence. Think back to my example of the murder statute. It does not mention self-defense, but neither does it foreclose that defense, and in this legal system defenses are generally applicable unless knocked out. So, too, causation is the norm unless knocked out. So, when Congress writes “(D) lost income,” it means “income lost as a result of the violation”; anything else would be so unusual that judges would want to see it in writing. One might reply that the express causation requirement in subsection (F) implied its negation elsewhere, but that’s not an application of the immediate-antecedent canon. It would be an inversion of the normal rule about the role of causation in damages. So Paroline isn’t about canons after all; it is about how much the context of the legal system is used to pull meaning from isolated statutes. And this, too, is a subject that can’t be reduced to an algorithm. I don’t think that either the majority or the dissent in either Lockhart or Paroline can be called wrong. We cannot get around the fact that language is incomplete.

I can imagine one approach that would have the mechanical nature required of a rule. The Rule of Lenity as currently understood says, “If ambiguity, then defendant prevails.”29 Yet all interesting language (including most that reaches the Supreme Court) is ambiguous. The Rule of Lenity does not say how serious the ambiguity must be, and the Justices do not agree on what “ambiguity” means for purposes of the rule.30 Suppose we replace “ambiguous” with the rule that, if three Justices think that the defendant prevails on the merits (that is, without regard to the Rule of Lenity), then there is enough uncertainty to require a decision for the defendant. This creates an algorithm. But as far as I know, none of the Justices has ever proposed the adoption of a doctrine that depends on anything other than four votes (to grant certiorari) or a majority (to decide on the merits).

Let me turn from the immediate-antecedent canon to Judge Katzmann’s call for other judges to make more and better use of legislative history.31

Like other textualists, I object to the use of legislative history for two principal reasons.

First, it is unreliable. Statements may be strategic, but I put that possibility aside for today. Suppose all statements are candid disclosures of their authors’ beliefs. Still, they may represent the views of only one person. Statements in committee reports speak for a majority of the committee’s majority, which usually means just a handful of legislators. The committee staff may summarize the views of ten Senators or twenty Representatives, but what judges need to know is the understanding of the median legislator whose vote created the majority on the floor—and the median legislator is silent.

Second, it is illegitimate. Not illegitimate in the sense that the committee staff is off on a lark; Congress thinks that recording this material serves a function, else it would close down the process.32 What I mean by “illegitimate” might be better expressed as “insufficient to constitute legislation under our system of governance.” An opinion poll among legislators does not create a legal obligation, even if we are very confident that the poll is statistically valid. Nor does a majority vote create a law, if before enactment the bill is sabotaged by an unpopular amendment and then killed off by its own sponsors. No judge would dream of declaring the bill enacted, even if the judge were perfectly confident that the unamended text would have passed. Tactics that allow minorities to frustrate majorities are part of the legislative process.

What the Constitution requires for legislation is concurrent, bicameral enactments by the legislature and signature by the President. Legislative history flunks the bicameralism requirement; neither house of Congress actually votes on the stuff.33 And it flunks the requirement that legislation be presented to the President for signature or veto. A President who agrees with the enrolled bill but disagrees with the legislative history has no recourse but to add some “history” of his own—the signing statement.34

Like most other textualists, I am willing to consult legislative history as a cue to linguistic usage, even though not as an authoritative guide to meaning. Wittgenstein showed that words don’t have internal meanings, and that the concept of speakers’ meaning also is empty.35 Meaning depends not on the contents of the speaker’s head, but on the reaction of the contemporaneous interpretive community. This is why when understanding the Constitution it is proper to consider both The Federalist and James Madison’s nemesis “Brutus”; contributions to the public debate show how people at the time used legal words and understood the document’s meaning. For many laws, however, there is no comparable public debate. When Congress in 1922 used the word “agriculture” in the Capper-Volstead Act, exempting agriculture from the antitrust laws, did it mean “agriculture” as understood in 1922, or “agriculture” as it would come to be understood in later decades?36 Legislative history might shed light—not because it would reveal authoritatively one reading of the statute, but because it might illuminate how people in 1922 used the word.

Gluck says that the CBO Canon would not run afoul of these concerns, and I have some sympathy with that view, though I need to give it more thought. Here I want to discuss an older argument for identifying reliable legislative history—an argument advanced by McNollgast, a portmanteau of Professors Mathew McCubbins, Roger Noll, and Barry Weingast.

McNollgast have proposed that judges could check the reliability of legislative history not by trying to deduce whether it spoke for a majority, or for swing voters, but by asking whether it represented a costly commitment.37 The proposal rests on two ideas. First, some legislators occupy positions that, as a result of agenda control, exercise inordinate influence; their statements may be informative, while what everyone else says is just wind. Second, statements by the influential legislators affect their reputations as dealmakers. If their statements are accurate, then deals they propose will be believable, and they will keep their influence (and perhaps acquire more); if they make false promises, their reputations will decline, and so will their influence. McNollgast suggested searching for the people with agenda influence or other holdup powers and then asking whether their statements were reliable. Influential legislators might make statements that are known by their colleagues to be hot air—stuff to fool interest groups, perhaps, or to take in credulous judges, but not capable of fooling other members of Congress. But when key legislators make concrete deals with other members, their words must be reliable because their reputations are at stake.

The proposal sounds nice in the abstract, but the original articles were short on particulars. Weingast, together with Daniel Rodriguez (now dean of Northwestern Pritzker School of Law), set out to rectify that by conducting a case study.38 They chose United Steelworkers v Weber,39 in which the Supreme Court held that Title VII of the Civil Rights Act of 1964, which prohibits racial discrimination in employment, permits an employer to give a preference to black applicants for jobs.40 The statute provides that it is unlawful “to fail or refuse to hire . . . any individual . . . because of such individual’s race.”41 The court of appeals had held that, when applicants of one race receive credit simply because of their race, applicants of another race necessarily will not be hired “because” they have the wrong race. The Court concluded in Weber that the statute does not enact a system of formal equality; it was designed to prevent the majority from discriminating against a minority, not to prevent the majority from discriminating against itself. The Court used legislative history to support this atextual reading.42

One might respond that the Court missed the lessons of public-choice theory. There is no “majority”; there is only a coalition of minorities, and an intense minority may be able to obtain preferential treatment by vote trading on other issues.43 Or one might respond, in a traditional legal fashion, that the Court looked over the heads of the crowd and picked out its friends among the speakers in Congress,44 so legislative history was used only to mislead the public about a decision reached on other grounds.45 But Rodriguez and Weingast proceeded differently. They asked whether the Court had identified the legislators with deal-making, or deal-breaking, power. Then they asked whether the Court distinguished between statements offered as fodder for interest groups and statements made as commitments to other members of Congress. Rodriguez and Weingast found abject judicial failure on both issues.

Justice Brennan’s opinion for the Court in Weber quoted freely from legislators who favored a pro-preference approach, or at least were neutral, whether or not that person had any influence. Many quotations were from the law’s most ardent supporters and opponents—as if the Justices did not know, or did not care, that to turn aspirations into law these boosters had to mollify the moderates. Rodriguez and Weingast, by contrast, concluded that enactment of Title VII depended on a deal between the northern Democrats (for whom Senators Case and Humphrey were the main spokesmen) and the northern Republicans (of whom Senator Dirksen and Representative Halleck were the leaders). There were no southern Republicans in Congress at the time. The Southern Democrats, and some moderates in both parties, were the opposition. Looking at the statements by people who could make deals paints a different picture from the statements of legislators who don’t play a role.

But, as Rodriguez and Weingast recognize, Senator Humphrey and the others who had deal-killing power were playing to three audiences: voters, the press, and their colleagues. Only the last set of statements represents commitments, breach of which would be costly. Rodriguez and Weingast tracked down the actual commitments, recorded in formal representations by the floor managers. According to Rodriguez and Weingast, these commitments show the adoption of a rule of formal equality: employers are forbidden to give race any weight, except via a seniority system that could lock in yesteryear’s use of race—and the permissible role of seniority was the subject of extensive negotiations and a complex compromise. In other words, Rodriguez and Weingast concluded that the Justices in the majority were wrong for the wrong reasons and the Justices in dissent were right for the wrong reasons.

I admire this work. One might even call it a compelling demonstration that legislative history can be used reliably. But I don’t think that the McNollgast program in general, or the Rodriguez and Weingast exemplar in particular, should lead to an increase in the use of legislative history. I can express my reason in six words: the article is 125 pages long. Perhaps I may be permitted a few more words: the article was published twenty-four years after Weber.

One of the academy’s most persistent illusions is a belief that judges have essentially unlimited time to decide cases and bring to bear universal expertise. Nothing else could support a system that allows judges to design the institutions of government, and to resolve society’s most complex moral problems, at the same time as they prescribe the details of automobile door handles. Some political scientists treat judges as automata who vote the platforms of the political parties that appointed them; for these judges, time is irrelevant, and so is McNollgast. But for those scholars who think that briefs, argument, and reasoning matter to courts, if only at the margin, the tradition is to assume that judges can devote as much time as is necessary to reach correct decisions and can apply the most sophisticated models drawn from economics, political science, engineering, and medicine.

What a bunch of baloney! Judges are lawyers, not scholars; they are trained in making arguments, not in forming and testing hypotheses. Yes, some judges are scholars, but with the exception of Judge Richard Posner none has universal expertise. I claim none outside of economics and physics. Anyway, my point is that most judges lack any “law and . . . ” training. They can’t replicate the sophisticated work of Rodriguez and Weingast—and you can be sure that if one side hires an expert to do so, the other side will hire an expert to refute it. Professor George Stigler was fond of saying that there’s no proposition in economics so absurd that you can’t locate one reputable economist to vouch for it. He was right. And the proposition holds for political science, too. Lawyers can’t adjudicate such a contest.

It took Rodriguez and Weingast twenty-four years after Weber to do their work. Actually, the time is two years from when they began, but even that is too long for litigation, which on issues of law can reach the Supreme Court less than two years after the suit is filed. And Rodriguez and Weingast were not preparing six hundred other articles at the same time.

Why six hundred? Because a federal appellate judge hears that many cases, or more, during a two-year span. Judges of the Seventh Circuit, where I sit, hear oral argument in about 150 cases a year, and handle another 150 on briefs. Judges have lots of other duties, including attending conferences, resolving contested motions, and handling the administrative side of the courts, which are large institutions. Suppose we ignore all of that time. A judge who spends two thousand hours a year to resolve three hundred appeals has less than seven hours per case to read the district court’s opinion and essential portions of the record, read the briefs (which can exceed one hundred pages per appeal), read any relevant precedents, hear argument, discuss the appeal with colleagues, write a draft (or, if not writing, read a colleague’s draft), and participate in the give-and-take among judges that turns drafts into published opinions. The Supreme Court hears fewer cases on the merits and spreads the opinion writing more widely, but it also has more discussion per case because it is harder to forge agreement among nine judges than among three. Even on the most optimistic view, a Justice can devote less than twenty hours to each argued case.

Yet Rodriguez and Weingast needed two years and 125 pages to analyze the legislative history of one clause of Title VII. That took more than twenty hours. Judges, who lack training in social science, would need more time than specialists; but judges actually have substantially less.

Rules of interpretation must reflect the resources available to the task. I argued over twenty years ago that this implies a relatively simple and mechanical approach to interpretation,46 and nothing I have seen since has changed my mind. Schauer elaborated on that view by contrasting the way in which experts approach the task of interpreting legal texts with the way in which amateurs should proceed.47 Experts try for the best and most nuanced understanding; amateurs should use the approach that causes the least damage when it goes wrong. And judges are amateurs. I may be a specialist in judging, but I’m an amateur in tax law, labor law, immigration law, and the other fields that federal courts handle. Judges are experts only in criminal law, which occupies perhaps a third of our docket. I know something about corporate and securities law, and Judge Posner knows about antitrust and torts; we’re not amateurs across the board. But when judges are amateurs, which is most of the time, we should use an appropriately modest interpretive strategy.48

The modest treatment of legislative history is the one that ignores it, in favor of the actual enacted text. Relying on text does the least harm, for the text is visible to everyone, while legislative history can take people by surprise (especially given judicial discretion about which history to emphasize, a choice that judges lack when dealing with enacted texts). So, much as I admire McNollgast, I don’t think that political science has shown how legislative history can be used beneficially.

  • 1. See generally Henry M. Hart Jr and Albert M. Sacks, The Legal Process: Basic Problems in the Making and Application of Law (Cambridge tentative ed 1958).
  • 2. See Rodriguez v United States, 480 US 522, 526 (1987) (per curiam) (“Deciding what competing values will or will not be sacrificed to the achievement of a particular objective is the very essence of legislative choice—and it frustrates rather than effectuates legislative intent simplistically to assume that whatever furthers the statute’s primary objective must be the law.”). This has become one of the Court’s more frequently cited and quoted propositions about statutory interpretation.
  • 3. See generally, for example, Frank H. Easterbrook, Statutes’ Domains, 50 U Chi L Rev 533 (1983); Frank H. Easterbrook, Textualism and the Dead Hand, 66 Geo Wash L Rev 1119 (1998); Frank H. Easterbrook, Judicial Discretion in Statutory Interpretation, 57 Okla L Rev 1 (2004).
  • 4. See generally Antonin Scalia and Bryan A. Garner, Reading Law: The Interpretation of Legal Texts (Thomson/West 2012).
  • 5. See generally Abbe R. Gluck, Congress, Statutory Interpretation, and the Failure of Formalism: The CBO Canon and Other Ways That Courts Can Improve on What They Are Already Trying to Do, 84 U Chi L Rev 177 (2017).
  • 6. Karl N. Llewellyn, Remarks on the Theory of Appellate Decision and the Rules or Canons about How Statutes Are to Be Construed, 3 Vand L Rev 395, 401–06 (1950).
  • 7. See generally Akhil Reed Amar, Intratextualism, 117 Harv L Rev 747 (1999).
  • 8. See Rowland v California Men’s Colony, Unit II Men’s Advisory Council, 506 US 194, 199–201 (1993).
  • 9. 15 USC § 77b(a).
  • 10. See generally, for example, United Housing Foundation, Inc v Forman, 421 US 837 (1975) (concluding that shares of stock representing rights to apartments are not securities); Marine Bank v Weaver, 455 US 551 (1982) (concluding that notes for indebtedness to a neighbor are not securities); Reves v Ernst & Young, 494 US 56 (1990) (concluding that corporate demand notes are securities, though personal checks would not be, notwithstanding the statutory language).
  • 11. See Ludwig Wittgenstein, Philosophical Investigations §§ 185–90 at 80–83e (Wiley-Blackwell 4th ed 2009) (P.M.S. Hacker and Joachim Schulte, trans).
  • 12. Lockhart v United States, 136 S Ct 958, 969 (2016) (Kagan dissenting).
  • 13. See generally Geoffrey P. Miller, Pragmatics and the Maxims of Interpretation, 1990 Wis L Rev 1179.
  • 14. For additional discussion of this point, see my “opinion” in The Case of the Speluncean Explorers: Revisited, 112 Harv L Rev 1876, 1913–17 (1999).
  • 15. 136 S Ct 958, 969 (2016) (Kagan dissenting).
  • 16. 134 S Ct 1710 (2014).
  • 17. Scalia and Garner, Reading Law at 144–53 (cited in note 4).
  • 18. 540 US 20, 27–28 (2003).
  • 19. Lockhart, 136 S Ct at 961–62.
  • 20. See id at 968; id at 975, 978 (Kagan dissenting).
  • 21. See id at 969, 978 (Kagan dissenting).
  • 22. See generally Abbe R. Gluck and Lisa Schultz Bressman, Statutory Interpretation from the Inside—an Empirical Study of Congressional Drafting, Delegation, and the Canons: Part I, 65 Stan L Rev 901 (2013).
  • 23. See Legislative Drafting Manual § 112 at *9–12 (Office of the Legislative Counsel, US Senate, Feb 1997), archived at; House Legislative Counsel’s Manual on Drafting Style §§ 311–12 at *23–25 (Office of the Legislative Counsel, US House of Representatives, Nov 1995), archived at
  • 24. See Bryan A. Garner, Guidelines for Drafting and Editing Legislation § 3.4(E) at 96 (RosePen 2016) (“Avoid unnumbered ‘dangling’ sections.”); Legislative Drafting Manual § 112(d) at *10 (cited in note 23).
  • 25. See Lisa Schultz Bressman and Abbe R. Gluck, Statutory Interpretation from the Inside—an Empirical Study of Congressional Drafting, Delegation, and the Canons: Part II, 66 Stan L Rev 725, 750–52 (2014).
  • 26. See generally Paroline, 134 S Ct 1710.
  • 27. Compare Lockhart, 136 S Ct at 961, with Paroline, 134 S Ct at 1716.
  • 28. At least none of the parties’ briefs did. I have not read all of the amicus briefs.
  • 29. Scalia and Garner, Reading Law at 296–302 (cited in note 4).
  • 30. Judge Brett M. Kavanaugh thoughtfully develops this point in Fixing Statutory Interpretation. See Brett M. Kavanaugh, Book Review, Fixing Statutory Interpretation, 129 Harv L Rev 2118, 2134–44 (2016).
  • 31. See Robert A. Katzmann, Judging Statutes 39 (Oxford 2014).
  • 32. This implies, as Professor Nicholas Quinn Rosenkranz concludes in Federal Rules of Statutory Interpretation, that Congress could lift any bar on the use of legislative history, other than the constitutional-legitimacy one discussed below, by specifying what force any particular history has—and could lift the Constitution’s bar by enacting the history, as it did in one section of the Civil Rights Act of 1991. See Nicholas Quinn Rosenkranz, Federal Rules of Statutory Interpretation, 115 Harv L Rev 2085, 2109 (2002).
  • 33. Often for good reasons. See Kavanaugh, Book Review, 129 Harv L Rev at 2122–28 (cited in note 30). See also generally John F. Manning, Book Review, Why Does Congress Vote on Some Texts but Not Others?, 51 Tulsa L Rev 559 (2016).
  • 34. That many members of Congress who contribute barrels of ink to the Congressional Record deem presidential signing statements illegitimate is one of government’s ironies.
  • 35. See Wittgenstein, Philosophical Investigations § 693 at 181e (cited in note 11) (“[N]othing is more wrong-headed than to call meaning something a mental activity!”).
  • 36. For a decision addressing this question, see generally National Broiler Marketing Association v United States, 436 US 816 (1978). This case arose from the provision of the Capper-Volstead Act that authorized those “engaged in the production of agricultural products” to “act together in associations . . . in interstate and foreign commerce.” Capper-Volstead Act, 42 Stat 388, 388 (1922), codified at 7 USC § 291.
  • 37. See generally McNollgast, Legislative Intent: The Use of Positive Political Theory in Statutory Interpretation, 57 L & Contemp Probs 3 (Winter 1994); McNollgast, Positive Canons: The Role of Legislative Bargains in Statutory Interpretation, 80 Georgetown L J 705 (1992).
  • 38. See generally Daniel B. Rodriguez and Barry R. Weingast, The Positive Political Theory of Legislative History: New Perspectives on the 1964 Civil Rights Act and Its Interpretation, 151 U Pa L Rev 1417 (2003).
  • 39. United Steelworkers of America, AFL–CIO–CLC v Weber, 443 US 193 (1979).
  • 40. Id at 209.
  • 41. 42 USC § 2000e–2(a)(1).
  • 42. Weber, 443 US at 200, 203–07.
  • 43. See Daniel A. Farber and Philip P. Frickey, Law and Public Choice: A Critical Introduction 40–62 (Chicago 1991); Dennis C. Mueller, Public Choice III 104–08
    (Cambridge 2003).
  • 44. Judge Patricia M. Wald memorialized the phrase “looking over a crowd and picking out your friends,” but its origin is a quip by Judge Harold Leventhal. See Patricia M. Wald, Some Observations on the Use of Legislative History in the 1981 Supreme Court Term, 68 Iowa L Rev 195, 214 (1983). For a more empirical approach, see generally David S. Law and David Zaring, Law versus Ideology: The Supreme Court and the Use of Legislative History, 51 Wm & Mary L Rev 1653 (2010).
  • 45. This is the theme of Justice Rehnquist’s dissent, amplified in a careful study by my late colleague Professor Bernard D. Meltzer—who was no conservative. See Bernard D. Meltzer, The Weber Case: The Judicial Abrogation of the Antidiscrimination Standard in Employment, 47 U Chi L Rev 423, 444–47 (1980).
  • 46. See generally Frank H. Easterbrook, What’s So Special about Judges?, 61 U Colo L Rev 773 (1990); Easterbrook, 50 U Chi L Rev 533 (cited in note 3).
  • 47. See Frederick Schauer, Statutory Construction and the Coordinating Function of Plain Meaning, 1990 S Ct Rev 231, 231–32; Frederick Schauer, The Practice and Problems of Plain Meaning: A Response to Aleinikoff and Shaw, 45 Vand L Rev 715, 733 (1992); Frederick Schauer, The Dilemma of Ignorance: PGA Tour, Inc. v Casey Martin, 2001 S Ct Rev 267, 268–69.
  • 48. See Adrian Vermeule, Judging under Uncertainty: An Institutional Theory of Legal Interpretation 289–90 (Harvard 2006).