The (Not So) Plain Meaning Rule

Introduction

Many tenets of statutory interpretation take a peculiar form. They allow consideration of outside information—legislative history, practical consequences, the statute’s title, etc.—but only if the statute’s text1 is unclear or ambiguous. These tenets are often expressed as a variation of the “plain meaning rule.” If the text’s meaning is “plain,” the other information can’t be considered. If it isn’t plain, the information comes in.

On its surface, the rule has an intuitive appeal. It seems like a safe intermediate position between strict textualism and some form of all-things-considered eclecticism or pragmatism. But if we poke below the surface, we ought to see that the basic structure of the plain meaning rule is quite puzzling. In our normal lives, in most contexts under the rules of evidence, and elsewhere, information is either useful or not. Information that is relevant shouldn’t normally become irrelevant just because the text is clear. And vice versa, irrelevant information shouldn’t become useful just because the text is less than clear.

This puzzling structure—“consider only in case of ambiguity”—deserves investigation. In this Article, we first explain the puzzle more formally, and then begin that investigation. It turns out that we can sketch some conditions under which this puzzling structure could be justified, for certain kinds of evidence. But nobody has shown that the plain meaning rule in fact meets these conditions, and we rather doubt that they could justify the plain meaning rule across the board. More importantly, we suspect that most interpreters have never even asked themselves the question.

Note that we do not take a position on whether one ought to be a textualist or an intentionalist or something else in the first place. That is of course “the big debate”2 in statutory interpretation. Similarly, we take no position here on the correct theory of statutory “meaning.”3 This is not to deny that there are right answers to these questions. But the plain meaning rule attempts to transcend those debates, and our criticisms of it do, too.

Textualists who think they have good reasons to ignore legislative history or the like shouldn’t automatically cave when the statute is ambiguous. Intentionalists who insist that legislative history is relevant shouldn’t automatically discard it when the text by itself seems clear. The plain meaning rule asks both sides to surrender the courage of their convictions. That surrender has not been justified, and perhaps cannot be.

I. The Plain Meaning Rule

The plain meaning rule says that otherwise-relevant information about statutory meaning is forbidden when the statutory text is plain or unambiguous. To see the rule in action, we need not look far. Consider one of the Court’s recent and entertaining statutory interpretation cases, Yates v United States,4 in which the Court split 4–1–4 on interpreting a provision of the Sarbanes-Oxley Act of 20025 that is now codified at 18 USC § 1519.6 Did § 1519’s prohibition on impeding a federal investigation by “knowingly . . . conceal[ing] . . . any record, document, or tangible object”7 apply to a boat captain who threw undersized fish back into the sea?8 In particular, could “tangible object” include things that are quite different from records and documents?9 Five justices said no; four said yes.

There is much to be said about the case,10 but for our purposes the noteworthy exchange was about the relevance of the statute’s title. The plurality, which narrowly construed the statute, started by pointing out that both the provision’s caption11 and the title of its section of the statute12 mentioned only “records” and “documents.”13 While “not commanding,” the plurality said, these headings “supply cues” that “tangible object” should be construed very narrowly.14 Justice Samuel Alito, who concurred in the judgment and provided the fifth vote for the defendant, similarly noted that his view was “influenced by § 1519’s title.”15

Justice Elena Kagan wrote the dissent. She replied that the Court had never before “relied on a title to override the law’s clear terms.”16 Instead, she invoked “the wise rule that the title of a statute and the heading of a section cannot limit the plain meaning of the text.”17 This is an instance of the plain meaning rule, whose key feature is to deny the relevance of other interpretive data if the text’s meaning is “plain” or “clear.”

Invocations of the rule are common. That same term, in King v Burwell,18 the Court dutifully reported that “[i]f the statutory language is plain, we must enforce it according to its terms.”19

Often, the rule is invoked to forbid reliance on a specific kind of source. The statutory titles ignored by Kagan’s opinion in Yates may seem like a minor point, but consider these perhaps more significant examples:

Legislative History. Despite legislative history’s critics, the Supreme Court as a whole has not categorically foresworn the use of legislative history. In some opinions, however, it has said that legislative history can be considered only if the text is ambiguous or unclear. In Tennessee Valley Authority v Hill,20 for instance, the Court said that “[w]hen confronted with a statute which is plain and unambiguous on its face, we ordinarily do not look to legislative history as a guide to its meaning.”21 Similarly, in its more recent decision in United States v Woods,22 the Court dismissed legislative history arguments with a footnote saying: “We do not consider Woods’ arguments based on legislative history. Whether or not legislative history is ever relevant, it need not be consulted when, as here, the statutory text is unambiguous.”23 Similar invocations of the rule are plentiful.24

Policy Considerations. Once again, the Court has certainly deemed the practical consequences or policy implications of interpretation to be relevant in some cases. But it has also said that they need not be considered when the meaning of the text is plain. For example, in Carcieri v Salazar,25 the Court concluded that it “need not consider [ ] competing policy views” in interpreting the statutory language “because Congress’ use of the word ‘now’ . . . speaks for itself.”26 Similarly, in Sebelius v Cloer,27 the Court turned aside the government’s arguments that the Vaccine Act28 “should be construed so as to minimize complex and costly fees litigation,” concluding that such “policy arguments come into play only to the extent that the Vaccine Act is ambiguous,”29 which it was not.30

Practice. The plain meaning rule may also operate to forbid invocations of practice. For instance, in United States v Ron Pair Enterprises, Inc,31 the Court concluded that “pre-Code practice” was relevant to interpreting the Bankruptcy Code only if the text was not clear.32 And in Milner v Department of the Navy,33 the Court rejected the government’s and dissent’s invocations of thirty years of lower court practice “even if true, because we have no warrant to ignore clear statutory language on the ground that other courts have done so.”34

Substantive Canons. Several of the so-called substantive canons of interpretation turn on whether the statute is ambiguous, and so they present instances of the plain meaning rule as well.35 For instance, the Court has rejected an “effort to avoid the plain meaning of the statute” by “invok[ing] the canon of constitutional avoidance,” because “that canon ‘has no application in the absence of statutory ambiguity.’”36 The same may be true of a range of other substantive canons.37

All of them. Other times, the rule is invoked more categorically, as in this oft-quoted statement in Connecticut National Bank v Germain:38

[I]n interpreting a statute a court should always turn first to one, cardinal canon before all others. We have stated time and again that courts must presume that a legislature says in a statute what it means and means in a statute what it says there. When the words of a statute are unambiguous, then, this first canon is also the last: judicial inquiry is complete.39

Again, in each instance the role of the plain meaning rule is not to categorically rule these sources in or out. Rather, it is to make them contingently irrelevant. The category of extraneous information is not considered if the statute is plain, but can be if it is not plain.

Two notes of clarification before proceeding. First, we should note that the word “plain” is (ironically) itself ambiguous. Courts and scholars sometimes use the phrase “plain meaning” to denote something like ordinary meaning—that is, the normal meaning, or the meaning one would normally attribute to those words given little information about their context.40 The ordinary meaning is “plain” in the sense of “plain vanilla.” But the plain meaning rule uses the phrase in a different sense, to denote obvious meaning—that is, the meaning that is clear.41 Here, meaning is “plain” in the sense of “plain to view.” Again, the plain meaning rule uses this latter sense of “plain”—the meaning that is clear or obvious.

Second, while the “rule” is asserted in plenty of cases and has been defended by Justice Antonin Scalia and Professor Bryan Garner as “essentially sound,”42 we do not mean to assert that the plain meaning rule is inviolably observed. For despite the rule that plain text is supposed to be preclusive, the Court has also said that the “meaning—or ambiguity—of certain words or phrases may only become evident when placed in context,”43 and context is sometimes conceived quite expansively. Indeed, Professor William Eskridge Jr has argued that, despite invocations of the plain meaning rule, “[i]n a significant number of cases, the Court has pretty much admitted that it was displacing plain meaning with apparent legislative intent or purpose gleaned from legislative history.”44

But our point stands regardless of whether the plain meaning rule is really a “rule” or merely a common trope. Our inquiry here is fundamentally normative—when does this kind of contingent irrelevance make sense? As we hope to show, the answer to that question is far trickier than most everybody seems to assume.

II. The Puzzle

Upon closer examination, there is something puzzling about the plain meaning rule. There are reasons to consider all pertinent information. There are reasons to categorically discard certain kinds of pertinent information. But why consider it only sometimes?

For examples of considering all pertinent information, think of federal agencies, which “must consider” all “significant comments” received during notice-and-comment rulemaking,45 even if they were very confident in their proposed rule in the first place. The Environmental Protection Agency, likewise, must consider cost when deciding whether to regulate power plants, no matter the degree of noneconomic concern.46 Along the same lines, Jeremy Bentham famously advocated a system of “free proof,” or admission of all logically relevant evidence, on the grounds that it “was simply ordinary epistemology applied to legal matters.”47

Conversely, to Bentham’s likely dismay, the law is also full of cases in which information is excluded, whether because it is intrinsically irrelevant or normatively problematic or too likely to mislead.48 Federal trial courts, for example, exclude most evidence of character or past acts.49 And federal antidiscrimination laws prohibit employers from basing employment decisions on an applicant’s race, religion, or sex.50

These categorical exclusions are easy for most to accept. But why make otherwise-relevant information only conditionally admissible? If legislative history is truly bad evidence of statutory meaning, shouldn’t it be ignored both when the meaning is plain and when it is less than clear? Conversely, if it is good evidence, shouldn’t we always at least look at it, even when the text seems pretty clear on its own? Why should legislative history’s admissibility depend on the evidence we get from another source, like the text?

As we explain below, one might be able to construct a justification for considering pertinent information only sometimes—but such a limit makes sense only if that “sometimes” is connected to some epistemic or other practical end. What makes little sense is a blanket prohibition against considering pertinent nontextual information if statutory language is “clear.” This is especially so if the courts’ main concern is interpretive accuracy—that is, getting it right. Courts justify adherence to the plain meaning rule as a way to avoid interpretive mistakes, but the rule seems ill-suited to the task.

To see the concern more formally, suppose that plain-language clarity is factive: that is, if it is “clear” that some statutory language means that p, then that language means that p in fact. So understood, to say that statutory language’s meaning is “clear” is akin to saying that its meaning is known—if, after all, a court knows that some statutory language means that p, then that language in fact means that p.51 As a linguistic matter, this is a plausible analysis of “clear.” The problem is that if this is what courts mean by “clear,” then the plain meaning rule does no work with respect to accuracy. If a court knows that some statutory language means that p just on the basis of the plain text, then considering, say, legislative history is pointless but also harmless.52 This is because the court knows, in turn, that the corresponding legislative history is misleading to the extent that it indicates the statutory language means something other than that p.

Suppose then that courts in this area are instead speaking loosely, and that when a court says that statutory language is “clear,” what it means is that it has a high degree of confidence in a particular reading just on the basis of the text. On this understanding, there are two possibilities. The first possibility is that the court has such high confidence that nothing in the legislative history could possibly dissuade it. But, if this is the case, then the analysis is the same as above: considering legislative history is pointless but harmless. The second possibility is that, in principle, something in the legislative history could dissuade the court. If this is the case, however, then refusal to consider legislative history amounts to something like willful ignorance. In the case of the second possibility, the prohibitions at issue do have an effect, but the effect is mischievous.53 As Judge Henry Friendly observed, it is “[i]llogical . . . to hold that a ‘plain meaning’ shut[s] off access to the very materials that might show it not to have been plain at all.”54

Implicit in all of this, of course, is the assumption that the nontextual evidence at issue is neither intrinsically irrelevant nor more likely than not to mislead. If courts may consider legislative history if the plain statutory text is not clear, then legislative history sheds light on interpretive questions. But it is easy to see how the same point works for those who categorically reject using legislative history because it is “counterproductive” or error prone.55 If those concerns make legislative history so unreliable that it should not even be considered when the text is clear, the concerns do not go away just because the text is less clear. Better to soldier on with one’s best estimate of the text’s meaning, even if uncertain, than to introduce information that is more misleading than informative.

Our overall point is that the relevance of information is not normally conditional. Either legislative history, statutory titles, or what have you tell us something relevant about meaning or they do not. But whether they do or do not, that does not suddenly change when the text is clear. The puzzle is thus why courts would ignore nontextual evidence only when textual evidence points strongly in one direction.

III. Possible Justifications

Having said all of that, we can imagine some justifications for some applications of the plain meaning rule—that is, some cases in which the conditional relevance of nontextual evidence could make sense. But even so, we stress that these justifications are both conditional and incomplete. They are conditional because they are merely an outline of the circumstances under which a plain meaning threshold might make sense. We do not think adherents to the rule have shown that those circumstances actually obtain, and we are not sure that they do. They are incomplete because, even if those circumstances do obtain in some classes of cases, they are unlikely to result in an across-the-board version of the plain meaning rule.

A. Cost Efficiency

The plain meaning rule might make sense for evidence that is probative but also expensive to collect or consider.

To see this, suppose that considering information A is low-cost and easy whereas considering information B is expensive and cumbersome. Even if A and B are equally reliable, it might make sense, on a cost-efficiency rationale, to start by considering A, considering B only if A leaves you uncertain. After all, if considering A is cheap and good enough for practical purposes, why incur the expense of considering B? The expected marginal benefit is low and the cost high.56

Less abstractly, consider apartment hunting. Seeing an apartment in person is a reliable way of assessing its suitability. It also takes a great deal of time and energy. For that reason, one will often reject an apartment just on the basis of pictures or information contained in a description (for example, square footage, floor, etc.). Because one can consider such information quickly and from the comfort of one’s couch, driving across town is often simply not worth it.57 If an apartment is a clear “no” based just on the ad, why bother? If, by contrast, the apartment is a “maybe,” it is plausibly worth scheduling a visit.

Note that a cost-efficiency story for the plain meaning rule is still a little tricky. In the apartment example above, many people might find it less intuitive to accept an apartment and sign a lease without ever bothering to see it in person, even if the square footage and price are perfect. Yet the plain meaning rule asks interpreters to ignore extraneous information both if the plain meaning of the statute is the equivalent of “yes” and if it is the equivalent of “no.” That may be a tougher sell.

Moreover, the cost-efficiency justification for the plain meaning rule would have to justify the conditional exclusion of evidence. Some scholars have argued, for example, that most nontextual evidence should be categorically excluded in part on cost-efficiency grounds.58 That kind of categorical argument, of course, is too strong to yield the plain meaning rule. Rather, a cost-efficiency justification for the plain meaning rule would require a particular ratio of costs and accuracies such that the extra evidence is too costly when A is clear, but not so costly that it is prohibitive when A is unclear. Again, this is possible, but would require a more precise quantification of the decision costs of considering different kinds of evidence than we have seen.59

A cost-efficiency justification for the plain meaning rule is at least conceivable for some classes of evidence, but for others it is not even plausible. Legislative history, for example, might be time-consuming for courts to consider.60 The relevant documents are often spread out rather than collected in a single place, and even once they are collected it can take some time and mental effort to put them in their proper context—a skill at which many lawyers and law students are not particularly good.61 Thus, even assuming that legislative history is probative with respect to statutory meaning, refusing to consider such history if the text is “clear” might make sense for already overburdened courts. Again, if considering just the text is cheap and good enough for practical purposes, maybe it is sometimes better to move on to the next case rather than to engage in additional, expensive investigation.62 In this respect, legislative history contrasts sharply with, say, titles or section headings, which are easy for courts to consider. It is hard for us to imagine any cost-exclusion justification for excluding those kinds of materials.63

B. Bias

Perhaps the plain meaning rule could make sense for certain kinds of evidence that have both potential value but also a hard-to-assess sort of bias.

To see this formally, imagine that we can consult information A and/or B on some question. We know that A is 90 percent reliable and that B is only 60 percent reliable. On the other hand, we also know that once we consult B there is a substantial risk that we will be incapable of rationally factoring in A’s response. Somehow source B is so powerful or charismatic that we will start convincing ourselves to prefer it to A, or start using it to reinterpret A’s response—even though A is more reliable than B!

In such a situation, something like the plain meaning rule would be a rational result: consult A first, and consult B only if A is unsure. This is better than always consulting both, because we avoid the biasing effect of B in cases where A is more likely to be correct. And it is probably better than never consulting B because we avoid “throwing away” the relevant information of B in cases where A is unsure.

Now, even in this scenario, it is not certain that consulting B is a good idea, for two reasons. First, it is possible that there could be moral objections to the form of bias at issue in B. (As we discuss, one could imagine that B involves demographic stereotypes or political partisanship, for example.) If so, one might actually prefer never to use B, even at the cost of accuracy. Second, even in cases where A is unsure, A still might yield some information. True equipoise is rare.64 So considering B when A is unsure still risks overpowering the more reliable A with the less reliable B. It is just that this is one of A’s less reliable moments, so the risk is smaller.

To describe this much less formally, think of a job interview. One might well think that a candidate’s suitability for a job is best assessed by reading her resume. One might also think that in-person interviews are a useful, but secondary, source of information about a candidate’s suitability. On the other hand, in-person interviews can introduce subconscious biases, favoring candidates who are more attractive, who have particular demographic characteristics, etc.65 And yet, because the biases are subconscious, one may not be able to fully correct for them either. Indeed, one might worry that the irrelevant factors one learns from the interview will color one’s view of the paper record in a way that one can’t in good faith disentangle.66

In that case, it seems quite sensible for an employer to decide that candidates with very strong or very weak paper records will be in or out on that basis alone. The philosophy department at Princeton, for example, does entry-level hiring without conducting in-person interviews for these and similar reasons.67 (Or for certain jobs, instead of a paper record, one might have a blind audition, as orchestras have discovered.)68 One might still resort to in-person interviews as a tiebreaker, but only in cases where the resumes or other blind qualifications are indeterminate. This obtains useful information while reducing bias.

Is this too far-fetched to be helpful to the plain meaning rule? We are not so sure. It seems at least conceivable to us that something like the practical consequences of a statutory interpretation might fit this model. Judges might well be committed to the view that practical consequences are relevant but of secondary importance to more standard legal materials like text and so on. On the other hand, judges might also worry that once they take into account practical considerations, it is hard to think clearly about anything else, and hard to resist the urge to start reinterpreting the standard materials to match the consequences the judges want to see. Alternatively, even if judges are confident in their own ability to weigh practical consequences appropriately, those same judges might worry about the ability of other judges to do the same.

If this bias is worrisome—and it might be especially worrisome if the practical considerations are ones with deeply contested partisan valences—then we can see why judges might want the plain meaning rule to deal with it: Blind themselves to practical consequences in a range of cases where the text is clear and therefore the likelihood of changing the outcome is small relative to the likelihood of bias. But consider practical consequences when the text is unclear, and the information provided by those consequences is more important.

But again, even so, this bias justification for the plain meaning rule is tricky. The biasing information has to be of a specific type that slightly defies rational thought. And it has to be useful enough to be worth accepting that bias in some cases, and yet not so useful that it is worth accepting that bias in all cases.69 That might happen to be true of something like practical consequences, but we do not really know for sure. (So far as we know, neither the Princeton philosophy department nor orchestras suddenly start resorting to in-person interviews or auditions when it is a close case.)

At the same time, we are fairly confident that most other instances of the plain meaning rule—when dealing with titles, statutory context, legislative history, etc.—could not be justified this way. Judges are likely capable of rationally counting or discounting this material as appropriate. Or, at least, they are as capable of dealing rationally with this material as they are with anything else.

C. Legal Convention

An alternative justification, of sorts, might proceed in a more legalistic way: judges should follow the plain meaning rule because it is a rule, and judges should follow the rules. We recognize that this argument sounds hilariously circular—where did the rule come from?—but we think a version of it can be made to work.

One way is by focusing on the “law of interpretation.” This argument requires us first to accept that rules of statutory interpretation can be set by law, in which case they need not be justified on first-order normative grounds. (Judges should follow the rules of criminal procedure, one might say, not because the rules are necessarily justified on first principles but because judges have assumed an obligation to follow the rules.) The second step is to accept that those rules can also be established by unwritten law, judicial custom that we often call common law.70

Under this argument, maybe the plain meaning rule is simply a common-law rule of statutory interpretation. It might not make perfect logical sense, but judges should apply it just as much as they apply other logically imperfect common-law rules.

There is another variation of this argument: Even if one does not accept that unwritten law can create binding rules of interpretation (though one should!), one could arrive at a version of this argument through expectations. Perhaps Congress knows about the plain meaning rule and intends (or means) that its work should be interpreted through the rule.

These justifications seem logically possible to us, but we still have doubts about them. As to the second, expectations-based version of the argument, recent empirical research by Professors Abbe Gluck and Lisa Bressman has suggested that Congress does not know very much about the Supreme Court’s statutory interpretation rules, suggesting that we should be hesitant to justify interpretive rules purely on the basis of expectations.71 To be sure, the plain meaning rule might turn out to be an exception. The Gluck and Bressman study did report that when asked to name an “interpretive rule[ ] or convention[ ] in particular you think that the U.S. Supreme Court consistently follows,”72 staffers did frequently come up with “the plain meaning rule” by name.73 But because of the strictly consistent, empirical method of the study, we do not know if the respondents specifically had in mind the consider-only-in-case-of-ambiguity version of the plain meaning rule or instead associated “plain” with “ordinary”74 or something else.75

Similarly, we doubt that the plain meaning rule is sufficiently well established to qualify as a binding rule of the unwritten law of interpretation. While such rules can be valid even if Congress is not deeply familiar with them,76 discussion of the plain meaning rule is sufficiently confused that we still doubt that the rule qualifies.77 It is more plausible, however, that some limited instances of the plain meaning rule could be part of the law of interpretation—for instance, one of us has argued that this is the best way to judge the “substantive” canons of interpretation.78

In any event, even if one of these convention-based justifications for the plain meaning rule did hold, it would be a justification of a limited sort. It would justify judges’ invoking and applying the rule now, but it would tell us little about whether we ought to change the convention going forward, or about whether we ought to replicate it when implementing authoritative rules of interpretation in other contexts. Indeed, if one accepts our other normative doubts about the plain meaning rule, then those doubts provide reason to approach these conventions with a wary eye.

D. Public-Facing Explanation

It is also possible that there is a difference between a court’s own reasoning process and the reasoning process it presents to the audience of its opinions. Or, to put a finer point on it, maybe the plain meaning doctrine is a public lie or, more generously, an oversimplification.79 The court does in fact consider all of the evidence, but it does not want the reader to do so because it does not trust the reader to weigh the evidence accurately.

For instance, when interpreting a statute, maybe a judge really does consider the title of the statute, or the legislative history, even when the text seems plain to that judge. (Indeed, one would often have to go out of one’s way not to consider it, and we doubt that Justice Kagan gets mad at her clerks if they tell her the title of the statute or present her with other information outside of a clear text.)80 Perhaps judges do not think those things are really irrelevant, the same way they might think that a litigant’s race or criminal history is irrelevant.81 At the same time, they want to encourage the reader not to worry herself about them.

Under this justification, then, it is not actually true that outside information is ignored when the meaning is plain. Rather, judges think that the outside information will change the purely textual result only in an unusual case, and when the information does not change the result, it is better to pretend that it could not have changed the result. In other words, the court considers both text and other materials, but in cases in which the text wins, the court pretends it did not look at the other materials in the first place.

Why might a court do this? Perhaps it does not fully trust its audience. When presenting its textual argument to nonjudges and even lay people, who are not as steeped in the court’s conventions of statutory interpretation, it makes sense to speak in accessible shorthand. Courts really mean something like, “When the text is 80 percent clear, it is almost impossible for even 100 percent clear legislative history to outweigh it.” But it is easier for courts to say, as they do, that we should not even consider the legislative history at all in cases of textual clarity.

Or, to put it slightly less nobly, maybe judges worry that acknowledging the countervailing factor will make their interpretation seem much weaker. Better to invoke a legalistic-sounding reason that the title and legislative history do not matter rather than to candidly say: “True enough, but still . . . the text wins.”

As a descriptive matter, these accounts seem plausible to us, at least some of the time. But as a normative matter, they raise the usual questions about a duty of judicial sincerity.82 And, in any event, they “justify” invocations of the plain meaning rule only in the pyrrhic sense of saying that none of them should be taken seriously by legally sophisticated readers.

E. Predictability and Consistency

Additionally, the plain meaning rule might make sense—under certain extremely specific assumptions—if one were willing to trade accuracy for predictability. Suppose, for example, that a regulated private party cares not very much about whether she has the meaning of the statute “right” in the abstract, but cares a great deal about whether she correctly guesses how a judge will interpret the statute. That party might prefer that the range of considerations for a judge be limited in cases in which one consideration—the text—points clearly in one direction. “If I look up the statute and see that it says clearly what I can do,” she might reason, “I want to be able to take that to the bank.”

A judge might reason similarly, regarding consistency in decision-making across courts as worth promoting so long as the accuracy trade-off is minimal. Thus, a judge might think it best to stop if the text is “clear” because she is confident that her colleagues would read the statute in the same way. More still, because plain meaning is reasonably probative of statutory meaning, the resulting gains in consistency would be accompanied by only minimal losses in accuracy.

As before, note that this justification requires some tricky assumptions. It is not enough to argue—as many have83 —that text is a useful coordinating point. That argument would be more likely to point toward textualism across the board. Rather, it requires an argument that text is only sometimes useful as a coordinating point. The underlying intuition seems to be that when the text is plain, the coordinating function is strong and the loss in accuracy is weak, but when the text is less plain, we should flip to emphasizing accuracy over coordination.84

Maybe that argument could work, but it rests on several empirical assumptions. And these assumptions seem even more questionable than in the examples above.

The first required assumption is that the plain meaning threshold is itself reasonably plain—in other words, that most interpreters can agree on which textual meanings are plain. Consider the regulated private party who wants to “take it to the bank.” For her to do so sensibly, her perception of the plainness of the statute’s meaning would itself need to be widely shared. If, by contrast, interpreters frequently disagree over whether a statute’s meaning is plain, then the private party cannot be sure that what is plain to her will be plain to others.

Worse, if courts do not agree on the plainness thresholds in particular cases, the plain meaning rule can actually exacerbate unpredictability. Courts that are 80 percent sure from the text that the statute means X will sit resolute in their convictions, because the plain meaning rule tells them to consider no other evidence. Courts that are merely 54 percent sure from the text that the statute means X, however, will open the door to other evidence, which in turn increases the risk that they will move from X to something else. Because the plain meaning rule creates an interpretive cliff between “plain” and “nonplain” meaning, the predictability of that threshold becomes important to predicting what courts will do.

The current evidence suggests that this assumption is false—that is, the plain meaning threshold is highly vulnerable to dispute (good faith and otherwise85 ). The leading empirical study showed that different interpreters attribute ambiguity to the same text at quite different rates,86 and that those “simple judgments about ambiguity are entwined with policy preferences, and . . . there may well be a causal relationship between them.”87 It also showed that even when asked a less policy-laden question, to predict whether others will find a text ambiguous, interpreters remained divided.88 Another study, this one of contract interpretation, found that both judges and laypeople overestimate the extent to which their interpretation is widely shared.89 “Thus,” the study concluded, “a judge may consider language to be plain when in fact different people do not understand it the same way, and this may happen even when the judge’s understanding is shared only by a minority of people in general.”90

Even worse, Judge Brett Kavanaugh, a sitting judge on the DC Circuit, has asserted that his colleagues cannot even agree on what the plain meaning threshold is. He reported:

In practice, I probably apply something approaching a 65-35 rule. In other words, if the interpretation is at least 65-35 clear, then I will call it clear and reject reliance on ambiguity-dependent canons. I think a few of my colleagues apply more of a 90-10 rule, at least in certain cases. Only if the proffered interpretation is at least 90-10 clear will they call it clear. By contrast, I have other colleagues who appear to apply a 55-45 rule. If the statute is at least 55-45 clear, that’s good enough to call it clear.

Who is right in that debate? Who knows?91

Kavanaugh also went on to agree that “even if my colleagues and I could agree on 65-35” as the threshold for clarity, it would be “difficult” for them to apply it “neutrally, impartially, and predictably.”92 Rather, “the magic wand of ipse dixit is the standard tool for deciding such matters.”93 For these reasons, Kavanaugh advocated “eliminating or reducing threshold determinations of clarity versus ambiguity.”94

A second assumption required for the consistency-and-accuracy argument is that nontextual evidence is substantially less probative of statutory meaning than is text viewed in isolation. To see why, consider a case in which text in isolation points plainly in one direction but nontextual evidence points plainly in another. In such a case, a court that considers just the text will have reasonably high confidence as to statutory meaning.95 If nontextual evidence has only limited weight, going on to consider that evidence will predictably leave the court only less confident. The reason is that considering this less weighty evidence will not alter the court’s confidence very much, certainly not enough to make it as or more confident in the alternate reading that the nontextual evidence supports.96 If, by contrast, nontextual evidence has significant weight, going on to consider it may leave the court as or more confident in the alternate reading—particularly so if the nontextual evidence has an undercutting effect, undermining the evidential connection between the text in isolation and the initial reading (for example, by making apparent a previously unrecognized ambiguity).97 The predictability of the plain meaning rule thus depends on the nontextual evidence that is being excluded having limited probative value.98

This empirical assumption is also open to question. As both of us have argued elsewhere, a court’s perception of what Congress is trying to say depends in large part on that court’s understanding of what Congress is trying to do.99 By unsettling a court’s priors about what Congress is plausibly trying to do, nontextual evidence can thus alter significantly a court’s assessment of what Congress is attempting to say. For example, taken in isolation, a rule that reads, “No police officers are permitted,” would seem to tell law enforcement officials to stay away. Add as additional context, however, that the text appears beneath the heading, “Costume restrictions for Halloween party,” and that interpretation becomes much less obvious. Whether nontextual evidence can unsettle priors in this way—and, hence, exert significant evidentiary weight—is difficult to assess on a categorical basis. Sometimes bad practical consequences will, for instance, reveal a particular interpretation as implausible.100 Other times, though, such consequences will show only that the most plausible interpretation is also bad policy.101

It may well be that some instances (or even all instances) of the plain meaning rule could be shown to satisfy these assumptions. But the current evidence makes that unlikely, and in any event we are pretty sure that those who invoke the plain meaning rule have rarely satisfied themselves of it.

F. Contract Analogies

Finally, we think it instructive to contrast the plain meaning rule with seemingly analogous arguments in interpretation of private law. Consider first Professor Eric Posner’s argument in favor of the parol evidence rule in contract law,102 which has some analogies to the predictability justification canvassed above. The parol evidence rule forbids courts from considering extrinsic evidence of a contract’s meaning unless that contract is incomplete or ambiguous on its face.103 Posner defended the rule on the ground that “parties derive advantage from being able, in their contract, to limit the evidence a court can use to decide a dispute should one arise,” because, among other things, limiting admissible evidence reduces variance in judicial outcomes.104

For our purposes, what is particularly instructive about Posner’s argument is that it highlights several important differences between contracts and legislation, and therefore between the parol evidence rule and the plain meaning rule. First, Posner’s claim concerning derived advantage rests in large part on the empirical observation that contracting parties make frequent use of so-called “merger” clauses, which are clauses directing courts not to consider extrinsic evidence, whereas “anti-merger” clauses, which are clauses directing courts to consider such evidence, are more or less unknown.105 In legislation, by contrast, analogues of merger clauses are rare,106 whereas analogues of antimerger clauses are, if anything, slightly less so.107 Neither is common.

Second, Posner’s prediction of reduced variance under the parol evidence rule assumes the relative unpredictability of judicial responsiveness to extrinsic evidence. Because the extrinsic evidence parties might introduce in a given case is so varied—anything from “excerpts from the general chit-chat” to “pages of scrawled notes”—Posner inferred that how a given court will respond to the evidence introduced is much less predictable than how that court will respond to contractual language in isolation.108 By contrast, the major categories of nontextual evidence in statutory interpretation are more systematic, and so judicial responsiveness to such evidence is more predictable. At a practical level, for example, we think many lawyers have a good guess as to how different justices on the Supreme Court would respond to the invocation of a committee report.

Finally, Posner’s argument assumes that contracting parties are responsive to judicial interpretive rules. Plausible as that assumption might be for contracts, there are two reasons to doubt it in the case of legislation. One reason is recent empirical work that suggests that legislative drafters do not know much about the judicial interpretive rules.109 The other is that the transaction costs for negotiating legislation are much higher than they are for contracts because of both the complexity and the conventions of legislation.110 (Related elements of the legislative process may give legislators more reason than contracting parties to be strategically vague in drafting.)111

As the example of the parol evidence rule suggests, practical differences between contracting and legislating will often differentiate the plain meaning rule from superficially similar rules of contract law. Such practical differences also explain away another superficially similar rule, the so-called best evidence rule, which conditions the admissibility of secondary evidence (for example, facsimile or oral description) of the contents of a document on the unavailability of the original copy.112 As Professor Frederick Schauer has observed, strict application of the best evidence rule “imposes cumbersome requirements on the introduction of reliable secondary evidence.”113 Nonetheless, Schauer reasoned, the best evidence rule is plausibly justifiable as an evidence-generating rule, because, by making it difficult for parties to introduce (presumably reliable) secondary evidence, the rule incentivizes parties to preserve and produce (presumably more reliable) primary evidence.114

As noted above, that legislators respond to incentives set by interpretive rules is, at best, questionable. More to the point here, though, is that the practical problem the best evidence rule is designed to solve—failure to preserve and produce primary evidence—does not exist with respect to legislation, at least not today.115 In the modern era, primary evidence of the contents of legislation (for example, the Statutes at Large and the United States Code) is available at the click of a mouse. There is thus no need to cajole legislators or litigants to further preserve and produce that evidence.

Conclusion

There is much to be said about the comparative superiority of text, statutory context, legislative history, consequences, and so on in statutory interpretation. In this Article, we’ve tried to make a related but different intervention, about the relationship between those things.116 Whatever one thinks of the probative value of text and other evidence, it’s not at all obvious why one source’s probative value should depend on the other.

The plain meaning rule reflects that kind of puzzling interdependence. There are indeed conditions under which such a rule would make sense, but they are more complicated and less universal than most uses of the plain meaning rule seem to assume.117 It may well be that most interpreters should simply have the courage of their convictions—either to consider nontextual evidence in all cases or to ignore it across the board.

Ultimately, though, we come neither to praise the plain meaning rule nor to bury it. Our main aim is to challenge those who use the rule to consider and explain why they think nontextual evidence is relevant at some times but not at others—and to show all readers that the challenge is harder to answer than they might have first thought.

• 1To be sure, considering “just” the text requires attention to minimal information about the text (for example, that it is a legislative text). See generally John R. Searle, Literal Meaning, 13 Erkenntnis 207 (1978).
• 2William N. Eskridge Jr, Book Review, The New Textualism and Normative Canons, 113 Colum L Rev 531, 532 (2013). See also Trevor W. Morrison, Constitutional Avoidance in the Executive Branch, 106 Colum L Rev 1189, 1241 (2006) (observing the “lively and ongoing academic debate over whether it is legitimate for courts to rely on extratextual sources when construing statutes”) (emphasis omitted).
• 3See Richard H. Fallon Jr, The Meaning of Legal “Meaning” and Its Implications for Theories of Legal Interpretation, 82 U Chi L Rev 1235, 1243–52 (2015). See also Frederick Schauer, An Essay on Constitutional Language, 29 UCLA L Rev 797, 798–99 (1982).
• 4135 S Ct 1074 (2015).
• 5Pub L No 107-204, 116 Stat 745.
• 6Yates, 135 S Ct at 1079 (Ginsburg) (plurality).
• 718 USC § 1519. We might be accused of stacking the deck in the government’s favor by omitting the other verbs from our quotation. See Yates, 135 S Ct at 1086 (Ginsburg) (plurality) (relying on the verbs); id at 1089–90 (Alito concurring in the judgment) (same). But we’re not actually concerned here with the many other interpretive moves in Yates, so we assure you they are omitted without prejudice.
• 8See Yates, 135 S Ct at 1078–79 (Ginsburg) (plurality).
• 9See id at 1079 (Ginsburg) (plurality).
• 10See generally, for example, Tobias A. Dorsey, Some Reflections on Yates and the Statutes We Threw Away, 18 Green Bag 2d 377 (2015); Richard M. Re, The New Holy Trinity, 18 Green Bag 2d 407 (2015).
• 11See Yates, 135 S Ct at 1083 (Ginsburg) (plurality). See also 18 USC § 1519 (captioning the section “Destruction, alteration, or falsification of records in Federal investigations and bankruptcy”).
• 12See Yates, 135 S Ct at 1083 (Ginsburg) (plurality). See also Sarbanes-Oxley Act § 802, 116 Stat at 800 (entitling the section “Criminal penalties for altering documents”).
• 13See Yates, 135 S Ct at 1083 (Ginsburg) (plurality).
• 14Id (Ginsburg) (plurality).
• 15Id at 1090 (Alito concurring in the judgment).
• 16Id at 1094 (Kagan dissenting).
• 17Yates, 135 S Ct at 1094 (Kagan dissenting) (emphasis added), quoting Brotherhood of Railroad Trainmen v Baltimore & Ohio Railroad Co, 331 US 519, 528–29 (1947). See also Trainmen, 331 US at 528 (“Where the text is complicated and prolific, headings and titles can do no more than indicate the provisions in a most general manner; to attempt to refer to each specific provision would often be ungainly as well as useless.”).
• 18135 S Ct 2480 (2015).
• 19Id at 2489. As many probably know, the Court found that the language was not plain. Id at 2490.
• 20437 US 153 (1978).
• 21Id at 184 n 29.
• 22134 S Ct 557 (2013).
• 23Id at 567 n 5.
• 24See, for example, Milner v Department of the Navy, 562 US 562, 572 (2011) (refusing to “allow[ ] ambiguous legislative history to muddy clear statutory language”); Lamie v United States Trustee, 540 US 526, 533–34 (2004); Barnhart v Sigmon Coal Co, 534 US 438, 457 (2002).
• 25555 US 379 (2009).
• 26Id at 392.
• 27133 S Ct 1886 (2013).
• 28National Childhood Vaccine Injury Act of 1986, Pub L No 99-660, 100 Stat 3755, codified at 42 USC § 300aa-1 et seq.
• 29Cloer, 133 S Ct at 1895 (brackets and quotation marks omitted).
• 30Id at 1896.
• 31489 US 235 (1989).
• 32Id at 245–46.
• 33562 US 562 (2011).
• 34Id at 575–76.
• 35These are an apt example only to the extent that substantive considerations are used as evidence of statutory meaning, and not as policy tools used to fill statutory gaps. See Lawrence B. Solum, The Interpretation-Construction Distinction, 27 Const Commen 95, 105–07 (2010).
• 36Department of Housing and Urban Development v Rucker, 535 US 125, 134 (2002), quoting United States v Oakland Cannabis Buyers’ Cooperative, 532 US 483, 494 (2001).
• 37See Brett M. Kavanaugh, Book Review, Fixing Statutory Interpretation, 129 Harv L Rev 2118, 2146–56 (2016).
• 38503 US 249 (1992).
• 39Id at 253–54 (quotation marks and citations omitted). See also John F. Manning, The New Purposivism, 2011 S Ct Rev 113, 126–27 (suggesting that “the Court’s new approach” to statutory interpretation “is perhaps best captured by the Court’s oft-cited opinion in Connecticut National Bank v Germain”).
• 40See, for example, Frederick Schauer, Statutory Construction and the Coordinating Function of Plain Meaning, 1990 S Ct Rev 231, 251. See also David A. Strauss, Why Plain Meaning?, 72 Notre Dame L Rev 1565, 1565 (1997) (treating “‘ordinary’ or ‘plain’ meaning” as “interchangeable”). As Professor Frederick Schauer has observed, the “plainness” of statutory meaning in this sense is complicated by the question whether statutory language is ordinary or technical. See generally Frederick Schauer, Is Law a Technical Language?, 52 San Diego L Rev 501 (2015). To the extent that statutory language consists of terms of art familiar to lawyers, such language might have a “plain” technical meaning—the meaning a lawyer would normally attribute to the words upon knowing they were uttered by another lawyer.
• 41See, for example, Ron Pair Enterprises, 489 US at 242 (“The language and punctuation Congress used cannot be read in any other way. By the plain language of the statute, the two types of recovery are distinct.”) (citation omitted and emphases added).
• 42Antonin Scalia and Bryan A. Garner, Reading Law: The Interpretation of Legal Texts 436 (Thomson/West 2012). Scalia and Garner acknowledge the difficulty of determining what is unambiguous. See note 94.
• 43Food and Drug Administration v Brown & Williamson Tobacco Corp, 529 US 120, 132 (2000).
• 44William N. Eskridge Jr, The New Textualism, 37 UCLA L Rev 621, 628 (1990).
• 45Perez v Mortgage Bankers Association, 135 S Ct 1199, 1203 (2015), citing Citizens to Preserve Overton Park, Inc v Volpe, 401 US 402, 416 (1971).
• 46See Michigan v Environmental Protection Agency, 135 S Ct 2699, 2707–08 (2015).
• 47Frederick Schauer, On the Supposed Jury-Dependence of Evidence Law, 155 U Pa L Rev 165, 169 (2006), discussing generally Jeremy Bentham, Rationale of Judicial Evidence, Specially Applied to English Practice (Hunt & Clark 1827).
• 48See Schauer, 155 U Pa L Rev at 194 (cited in note 47) (“[T]he idea of Free Proof may have more cognitive and epistemic disadvantages than Bentham thought almost two centuries ago.”).
• 49See FRE 404.
• 5042 USC § 2000e-2(a)(1).
• 51See Timothy Williamson, Knowledge and Its Limits 34 (Oxford 2000).
• 52This is putting aside for a moment the cost of gathering and considering the evidence, which we discuss in Part III.A.
• 53This is what differentiates the “plain meaning rule” invoked in the cases we cite from a more sensible “presumption” in favor of the plain meaning that “is subject to rebuttal,” as described by John F. Manning, The Absurdity Doctrine, 116 Harv L Rev 2387, 2399 (2003). This also differentiates it from rules of cumulativeness, or “marginal probative value.” Old Chief v United States, 519 US 172, 185 (1997).
• 54Henry J. Friendly, Mr. Justice Frankfurter and the Reading of Statutes, in Henry J. Friendly, Benchmarks 196, 206 (Chicago 1967). See also Pacific Gas and Electric Co v G.W. Thomas Drayage & Rigging Co, 442 P2d 641, 645 (Cal 1968) (in bank) (Traynor) (“The exclusion of parol evidence . . . merely because the words do not appear ambiguous to the reader can easily lead to the attribution to a written instrument of a meaning that was never intended.”) (emphasis added).
• 55See, for example, Adrian Vermeule, Legislative History and the Limits of Judicial Competence: The Untold Story of Holy Trinity Church, 50 Stan L Rev 1833, 1838 (1998) (arguing that “problems of judicial competence create grave risks that judicial resort to legislative history to gauge legislative intent will prove counterproductive”); Antonin Scalia, A Matter of Interpretation: Federal Courts and the Law 31–32 (Princeton 1997) (Amy Gutmann, ed) (“[T]he use of legislative history . . . is much more likely to produce a false or contrived legislative intent than a genuine one.”).
• 56See generally, for example, Remco Heesen, How Much Evidence Should One Collect?, 172 Philosophical Stud 2299 (2015) (discussing the trade-off between accuracy and cost in the context of scientific research). See also Joseph Raz, Practical Reason and Norms 60 (Princeton 1990) (observing that “[f]act-finding and evaluating the different reasons for action consume time and effort,” and that these costs “will often outweigh the marginal benefits” that “ensue from engaging in a complete assessment of the situation on its merits”).
• 57See Tim Logan, Apps, Sites Aim to Transform Apartment Rental Listings (LA Times, Nov 30, 2014), archived at http://perma.cc/NNW3-ZX6L; Jonah Bromwich, Apartment Hunting with a Mobile App (NY Times, Mar 14, 2014), online at http://www.nytimes
  .com/2014/03/16/nyregion/apartment-hunting-with-a-mobile-app.html (visited Mar 24, 2017) (Perma archive unavailable).
• 58See, for example, Adrian Vermeule, Judging under Uncertainty: An Institutional Theory of Legal Interpretation 189–205 (Harvard 2006).
• 59See id at 159 (“[S]o little work has been done to assess the empirical consequences of interpretive choice.”).
• 60See Scalia, A Matter of Interpretation at 36 (cited in note 55) (“The most immediate and tangible change the abandonment of legislative history would effect is this: Judges, lawyers, and clients will be saved an enormous amount of time and expense.”).
• 61See Vermeule, 50 Stan L Rev at 1863–77 (cited in note 55). See also generally Frederick Schauer, Our Informationally Disabled Courts, 143 Dædalus 105 (Summer 2014). But see Victoria F. Nourse, A Decision Theory of Statutory Interpretation: Legislative History by the Rules, 122 Yale L J 70, 91, 135 (2012) (arguing that “no one should try to understand legislative history without understanding Congress’s own rules,” but then asking: “Is it really ‘too complex’ or difficult for judges and academics to learn a dozen congressional rules?”).
• 62Professor Adrian Vermeule has noted this possibility. See Vermeule, Judging under Uncertainty at 195 (cited in note 58) (“Intermediate solutions include a rule . . . that consults legislative history only if the statute lacks a plain meaning.”). But he is skeptical. See id (“[I]n practice such intermediate solutions prove highly unstable over any extended period and inevitably dissolve back into plenary consideration of legislative history.”).
• 63We also note that costs might vary over time. The cost of considering other statutes, for example, has decreased significantly with the development of electronic search tools. See Ellie Margolis and Kristen E. Murray, Say Goodbye to the Books: Information Literacy as the New Legal Research Paradigm, 38 U Dayton L Rev 117, 121–26 (2012).
• 64See, for example, O’Neal v McAninch, 513 US 432, 435 (1995) (noting that in “unusual” cases the judge might “feel[ ] himself in virtual equipose”).
• 65See generally, for example, Regina Pingitore, et al, Bias against Overweight Job Applicants in a Simulated Employment Interview, 79 J Applied Psychology 909 (1994); David C. Gilmore, Terry A. Beehr, and Kevin G. Love, Effects of Applicant Sex, Applicant Physical Attractiveness, Type of Rater and Type of Job on Interview Decisions, 59 J Occupational Psychology 103 (1986); Comila Shahani, Robert L. Dipboye, and Thomas M. Gehrlein, Attractiveness Bias in the Interview: Exploring the Boundaries of an Effect, 14 Basic & Applied Soc Psychology 317 (1993).
• 66See David Hausman, Note, How Congress Could Reduce Job Discrimination by Promoting Anonymous Hiring, 64 Stan L Rev 1343, 1349–55 (2012).
• 67See Farewell to the Eastern APA, Redux? (Leiter Reports: A Philosophy Blog, Sept 13, 2011), archived at http://perma.cc/JJR4-LVBQ?type=image (discussing Princeton’s interview policy in the comments).
• 68See generally Claudia Goldin and Cecilia Rouse, Orchestrating Impartiality: The Impact of “Blind” Auditions on Female Musicians, 90 Am Econ Rev 715 (2000).
• 69We would add that this justification is even harder to sustain if the plain meaning threshold is itself subject to bias, as some have argued. See notes 85–94 and accompanying text.
• 70For an extended argument in defense of these claims, see William Baude and Stephen E. Sachs, The Law of Interpretation, 130 Harv L Rev 1079, 1104–18 (2017).
• 71See 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); 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 (2014).
• 72Abbe R. Gluck and Lisa Schultz Bressman, Statutory Interpretation from the Inside: Methods Appendix *40 (Stanford Law Review, May 2013), archived at http://perma.cc/SR6K-6UPA.
• 73Gluck and Bressman, 65 Stan L Rev at 995 (cited in note 71).
• 74See note 40 and accompanying text.
• 75See Abbe Gluck, Professor of Law, Yale Law School, E-mail to William Baude (Jan 21, 2016) (on file with authors).
• 76See Baude and Sachs, 130 Harv L Rev at 1108–09 (cited in note 70).
• 77See Part I.
• 78Baude and Sachs, 130 Harv L Rev at 1121–22 (cited in note 70).
• 79See, for example, Return of the Jedi (Lucasfilm 1983) (“So, what I told you was true—from a certain point of view.”).
• 80Indeed, in several cases in which the Court has invoked the plain meaning rule, it has gone on to discuss the evidence it just declared irrelevant. See, for example, Milner, 562 US at 572, 578–80 (Kagan).
• 81But see Justin Driver, Recognizing Race, 112 Colum L Rev 404, 432–38 (2012) (noting “gratuitous” judicial references to the race of litigants and others).
• 82See generally Micah Schwartzman, Judicial Sincerity, 94 Va L Rev 987 (2008); David L. Shapiro, In Defense of Judicial Candor, 100 Harv L Rev 731 (1987). But see generally Scott Altman, Beyond Candor, 89 Mich L Rev 296 (1990).
• 83Professor Frederick Schauer, for example, has argued that a general presumption in favor of ordinary meaning has a coordinating function, because it is likely to serve as “common ground” among members of a linguistic community. Schauer, 1990 S Ct Rev at 250–56 (cited in note 40). For that reason, Schauer has argued, a “group of [otherwise-] diverse decisionmakers might suppress some of that diversity and achieve agreement” by substantially restricting the basis for decision-making to that which is common to the group, namely, ordinary meaning. Frederick Schauer, The Practice and Problems of Plain Meaning: A Response to Aleinikoff and Shaw, 45 Vand L Rev 715, 724–25 (1992). This argument does not seem to point toward the conditional evidence rules of the plain meaning rule.
• 84An intriguing suggestion that might work along these lines comes from Professor Adam M. Samaha, who has argued that the need for a random tiebreaker can be reduced by “moving a relevant variable into a lexically inferior position. The reduction in ties from lexical ordering is usually greater than the reduction from adding the same variable to the mix of other relevant considerations. The cost, however, is a higher error rate.” Adam M. Samaha, On Law’s Tiebreakers, 77 U Chi L Rev 1661, 1664 (2010). He went on to discuss the plain meaning rule as an example of lexical priority, though he acknowledged that other tiebreakers are available in some cases (such as lenity in criminal cases). See id at 1708–10.
• 85See Ryan D. Doerfler, The Scrivener’s Error, 110 Nw U L Rev 811, 840–41 (2016) (discussing willful or motivated mischaracterization of the clarity of legislative texts).
• 86See generally Ward Farnsworth, Dustin F. Guzior, and Anup Malani, Ambiguity about Ambiguity: An Empirical Inquiry into Legal Interpretation, 2 J Legal Analysis 257 (2010).
• 87Id at 271.
• 88See, for example, id at 272 (“All respondents considering that case are 55 percent likely to say the statute is ambiguous when asked for an external judgment.”).
• 89See Lawrence Solan, Terri Rosenblatt, and Daniel Osherson, False Consensus Bias in Contract Interpretation, 108 Colum L Rev 1268, 1285–94 (2008).
• 90Id at 1294.
• 91Kavanaugh, Book Review, 129 Harv L Rev at 2137–38 (cited in note 37) (citation omitted).
• 92Id at 2138.
• 93Id at 2140 (quotation marks and brackets omitted), quoting Farnsworth, Guzior, and Malani, 2 J Legal Analysis at 276 (cited in note 86).
• 94Kavanaugh, Book Review, 129 Harv L Rev at 2134 (cited in note 37) (emphasis omitted). Even Justice Scalia and Professor Garner, after describing the plain meaning rule as “essentially sound,” conceded that it is “largely unhelpful, since determining what is unambiguous is eminently debatable.” Scalia and Garner, Reading Law at 436 (cited in note 42).
• 95This should be the case regardless of whether a court regards the text as “clear.” See text accompanying note 91. Even if a court takes the text to be less than clear, it should still be reasonably confident in the interpretation supported by the text in isolation (for example, a confidence level of 0.7 as opposed to 0.9), at least absent additional, nontextual evidence.
• 96Again, this should be the case regardless of perceived clarity. As a practical matter, this is crucial to the attractiveness of the plain meaning rule because otherwise the rule would allow for wildly divergent outcomes depending on whether a text is regarded as clear or slightly less than clear.
• 97See John L. Pollock, Contemporary Theories of Knowledge 37–39 (Rowman & Littlefield 1986) (distinguishing between “rebutting” defeaters, which prevent evidence E from supporting belief in proposition P by supporting not-P more strongly, and “undercutting” defeaters, which prevent E from supporting belief in P by undermining the apparent rational connection between E and P) (emphasis omitted).
• 98Philosopher Lara Buchak has argued that, as a matter of both instrumental and epistemic rationality, a “risk-avoidant agent” sometimes does best not to consider all available evidence before making a decision, even if considering additional evidence is cost free. See Lara Buchak, Instrumental Rationality, Epistemic Rationality, and Evidence-Gathering, 24 Philosophical Persp 85, 96–101 (2010). On the standard picture of instrumental rationality, an instrumentally rational agent maximizes expected utility, which is to say that, of two acts, an instrumentally rational agent prefers the one with the higher expected-utility value. In a classic paper, Professor I.J. Good showed that, assuming away behavioral irrationalities of the sort discussed in Part III.B, so long as considering additional evidence is cost free, it always maximizes expected utility to do so before making a decision. See generally I.J. Good, On the Principle of Total Evidence, 17 Brit J Phil Sci 319 (1967). From this, Good inferred that it is always instrumentally rational to consider such evidence and, from this, that to do so is always epistemically rational as well—that is, rational in one’s capacity as an agent concerned with truth or knowledge. See generally id. In her work, Buchak has offered an alternative to the standard picture of instrumental, and in turn epistemic, rationality, arguing that actual agents are—according to Buchak, reasonably—risk avoidant in the sense that such agents are unwilling to accept the possibility of a loss in exchange for an equivalently sized possibility of a gain. See generally, for example, Lara Buchak, Risk and Rationality (Oxford 2013). Needless to say, assessing the merits of Buchak’s alternative, risk-avoidant picture of rationality goes well beyond the scope of this Article. See Buchak, 24 Philosophical Persp at 96 (cited in note 98) (conceding that there are “those who are inclined to think that theories like [hers] are theories of predictable irrationality”) (emphasis added). Of special interest here, though, is Buchak’s observation that, to the extent that risk avoidance is rational, a risk-avoidant agent ought not to consider additional cost-free evidence under certain conditions. Specifically, Buchak showed that a risk-avoidant agent should refuse to consider such evidence if she is “antecedently fairly confident that X” and if that evidence could “tell somewhat in favor of ~X but not strongly in favor of ~X,” that is, if the evidence at issue has only limited weight. Id at 100 (emphasis omitted). Thus, to the extent that risk avoidance recommends ignoring additional evidence, it does so under similar conditions as does the predictability-and-consistency rationale articulated here.
• 99See Baude and Sachs, 130 Harv L Rev at 1144–45 (cited in note 70); Ryan D. Doerfler, Who Cares How Congress Really Works?, 66 Duke L J 979, 994–98 (2017).
• 100See, for example, United States v X-Citement Video, Inc, 513 US 64, 69–70 (1994).
• 101See, for example, Pavelic & LeFlore v Marvel Entertainment Group, 493 US 120, 126 (1989).
• 102See generally Eric A. Posner, The Parol Evidence Rule, the Plain Meaning Rule, and the Principles of Contractual Interpretation, 146 U Pa L Rev 533 (1998).
• 103See id at 534. Posner referred to this as “hard-PER,” distinguished from softer versions of the rule. Id at 534–35.
• 104Id at 570–71. See also id at 543. Posner also argued that contracting parties share an intention about the conventions of contract interpretation, see id at 570, somewhat analogous to the argument we discuss in Part III.C.
• 105Id at 570–71.
• 106For the rare example, see Civil Rights Act of 1991 § 105(b), Pub L No 102-166, 105 Stat 1071, 1075, codified at 42 USC § 1981 note:

  No statements other than the interpretive memorandum appearing at Vol. 137 Congressional Record S 15276 (daily ed. Oct. 25, 1991) shall be considered legislative history of, or relied upon in any way as legislative history in construing or applying, any provision of this Act that relates to Wards Cove—Business necessity/cumulation/alternative business practice.

  See also Lawrence M. Solan, The Language of Statutes: Laws and Their Interpretation 187 (Chicago 2010) (observing that “[e]ven such small limits” on what evidence of statutory meaning courts can consider “are not easy to find”).

• 107See, for example, 16 USC § 831dd (“This chapter shall be liberally construed to . . . provide for the national defense, improve navigation, control destructive floods, and promote interstate commerce and the general welfare.”); 18 USC § 3731 (“The provisions of this section shall be liberally construed to effectuate its purposes.”).
• 108Posner, 146 U Pa L Rev at 572 (cited in note 102).
• 109See generally Gluck and Bressman, 65 Stan L Rev 901 (cited in note 71). See also generally Victoria F. Nourse and Jane S. Schacter, The Politics of Legislative Drafting: A Congressional Case Study, 77 NYU L Rev 575 (2002).
• 110See Posner, 146 U Pa L Rev at 553–55 (cited in note 102) (observing that the parol evidence rule is least likely to be useful when transaction costs are high (as when the text is complex) and when the form of the contract at issue is conventional (as is the case with, for example, ordinary consumer contracts)). See also Jarrod Shobe, Intertemporal Statutory Interpretation and the Evolution of Legislative Drafting, 114 Colum L Rev 807, 816 (2014) (describing the professionalization of the legislative drafting process, noting the emphasis on “consistency of legislative drafting”); Josh Chafetz, The Phenomenology of Gridlock, 88 Notre Dame L Rev 2065, 2075 (2013) (“The United States federal government has a relatively [ ] cumbersome process for enacting laws.”).
• 111See Christine Kexel Chabot, Selling Chevron, 67 Admin L Rev 481, 522–25 (2015).
• 112See, for example, Sirico v Cotto, 324 NYS2d 483, 485–86 (NY City Civ 1971).
• 113Schauer, 155 U Pa L Rev at 198 (cited in note 47).
• 114Id.
• 115By contrast, in the decades after the Founding, statutes “were not regularly published,” and “[e]ven when copies of records could be found, the copies themselves were highly unreliable.” Stephen E. Sachs, Full Faith and Credit in the Early Congress, 95 Va L Rev 1201, 1209–10 (2009). As Professor Stephen E. Sachs showed, this made the best evidence rule highly important to statutes at the Founding (and central to the Constitution’s Full Faith and Credit Clause). See id at 1209–12.
• 116Ironically, the plain meaning rule may in fact cause courts to devalue the statutory text. The plain meaning rule requires text to be considered first, to decide what other sources can be considered. But as Professor Samaha recently observed, empirical evidence suggests that, “[o]ften enough, last matters more than first” in that, as a psychological matter, decision-makers often attribute greater significance to evidence considered at the end of a sequence than at the beginning. Adam M. Samaha, Starting with the Text—on Sequencing Effects in Statutory Interpretation and Beyond, 8 J Legal Analysis 439, 456 (2016). Thus, textualists who do decide to retain the plain meaning rule might do well to counsel “circling back” to the unclear text after other sources have been let in. Id at 481.
• 117See Kavanaugh, Book Review, 129 Harv L Rev at 2135 n 87 (cited in note 37) (“[E]ach ambiguity-dependent canon should be independently evaluated. I am not proposing a one-size-fits-all solution.”).