Harvard Law Review Harvard Law Review Harvard Law Review

Harvard Law Review Forum

Meta Rules for Ordinary Meaning

Responding to Kevin P. Tobia, Testing Ordinary Meaning

“Ordinary meaning” is a notoriously undefined concept in statutory interpretation theory. Courts and scholars sometimes describe ordinary meaning as the meaning that a “reasonable reader” would ascribe to the statutory language at issue,1×1. See, e.g., WILLIAM N. ESKRIDGE, JR., INTERPRETING LAW: A PRIMER ON HOW TO READ STATUTES AND THE CONSTITUTION 33 (2016); Bostock v. Clayton County, 140 S. Ct. 1731, 1825 (2020) (Kavanaugh, J., dissenting) (quoting ESKRIDGE, supra, at 33); Barker v. Ceridian Corp., 122 F.3d 628, 632 (8th Cir. 1997) (describing ordinary meaning as how “a reasonable person . . . would have understood the words” (quoting Chiles v. Ceridian Corp., 95 F.3d 1505, 1511 (10th Cir. 1996))). but it remains unclear how judges and lawyers should go about identifying such meaning. Over the past few decades, as textualism has come to dominate statutory interpretation, courts increasingly have employed dictionary definitions as (purportedly) neutral, and sometimes dispositive, evidence of ordinary meaning.2×2. See, e.g., James J. Brudney & Lawrence Baum, Oasis or Mirage: The Supreme Court’s Thirst for Dictionaries in the Rehnquist and Roberts Eras, 55 WM. & MARY L. REV. 483, 494–502 (2013) (discussing the recent rise in the Supreme Court’s use of dictionaries and positing that the Justices may have turned to dictionaries as more neutral tools of legal authority in response to criticism of ideological activism on the Court); Jeffrey L. Kirchmeier & Samuel A. Thumma, Scaling the Lexicon Fortress: The United States Supreme Court’s Use of Dictionaries in the Twenty-First Century, 94 MARQ. L. REV. 77, 84–86 (2010) (discussing the high rate of the Supreme Court’s use of dictionaries within decisions since 1999); Samuel A. Thumma & Jeffrey L. Kirchmeier, The Lexicon Has Become a Fortress: The United States Supreme Court’s Use of Dictionaries, 47 BUFF. L. REV. 227, 231 (1999) (discussing the increase in dictionary usage within Supreme Court decisions in the final decades of the twentieth century); Rapanos v. United States, 547 U.S. 715, 732–33 (2006) (plurality opinion) (finding that dictionary definitions precluded agency’s long-standing interpreta- tion of the Clean Water Act); MCI Telecomms. Corp. v. AT&T Co., 512 U.S. 218, 225 (1994) (finding that dictionary definitions showed that the term “modify” does not connote “fundamental change[]”); Taniguchi v. Kan Pac. Saipan, Ltd., 566 U.S. 560, 566–68, 568 n.2 (2012) (citing fourteen different dictionaries to establish the “ordinary meaning,” id. at 566, of the statutory term “interpreter”). And in the past few years especially, some judges and scholars have advocated using corpus linguistics — patterns of usage across various English-language sources — as an objective guide to the ordinary meaning of statutory words and phrases.3×3. See generally Thomas R. Lee & Stephen C. Mouritsen, Judging Ordinary Meaning, 127 Yale L.J. 788 (2018) (arguing that ordinary meaning can be empirically measured and theorized using corpus linguistics principles and methods); Stefan Th. Gries & Brian G. Slocum, Ordinary Meaning and Corpus Linguistics, 2017 BYU L. Rev. 1417 (arguing that corpus analysis and empirical methods can help inform judicial interpretation of ordinary meaning). Professor Kevin Tobia’s illuminating article Testing Ordinary Meaning4×4. 134 Harv. L. Rev. 726 (2020). seeks to test empirically how accurately these two interpretive aids — dictionary definitions and corpus linguistics — reflect ordinary meaning.5×5. See id. at 734.

To do so, Tobia uses a series of experimental studies based on surveys of laypeople gathered through Amazon’s Mechanical Turk,6×6. See id. at 754. as well as surveys of federal and state judges and law students at Harvard, Yale, and Columbia.7×7. See id. at 762. The study uses as a baseline for “ordinary meaning” the unaided collective intuitions of laypeople, federal judges, and law students — and compares those unaided intuitions to the meaning these three groups of interpreters selected when asked to apply dictionary definitions or corpus linguistics to the same set of terms.8×8. See id. at 754–56, 763. Tobia’s experiment constitutes an admirable effort to dissect the concept of ordinary meaning, and one that yields important information and results. In my view, there are three key takeaways from his study: (1) judges and nonexperts assess meaning similarly; (2) ordinary meaning is often unclear; and (3) dictionaries and corpus linguistics provide meanings that diverge from each other and from ordinary meaning, with dictionaries tending to reflect expansive, or “legalist,” word meaning and corpus linguistics tending to reflect “prototypical” meaning.9×9. See id. at 775–76. The distinction between “prototypical” and “legalist” meaning is one that linguist Lawrence Solan has long highlighted in the context of statutory interpretation, although he calls it “definitional” meaning. Lawrence M. Solan, The New Textualists’ New Text, 38 Loy. L.A. L. Rev. 2027, 2039–44 (2005). Solan defines prototypical meaning as the meaning that focuses on the core example that the statute was designed to reach, rather than a meaning that stretches to the conceptual or logical extension of the word at issue. See id. Professor Victoria Nourse has called the latter, logical extension kind of meaning “legalist meaning” — a label that this Response adopts. Victoria F. Nourse, Two Kinds of Plain Meaning, 76 Brook. L. Rev. 997, 1000 (2011) (citing Adrian Vermeule, Law and the Limits of Reason 2–3 (2008)).

I agree with many of the conclusions drawn by Tobia’s thoughtful article. This Response will focus primarily on a few points of disagreement as well as on some methodological lessons that might be gleaned from his findings. First, I discuss an important question that Tobia’s study glosses over — the question of who the appropriate audience (or “ordinary reader”) is for a particular statute10×10. Tobia’s article frames itself as focusing on the interpretation of both contracts and statutes, Tobia, supra note 4, at 728, but the examples and experimental terms it uses — especially “vehicles” from the classic statutory interpretation hypothetical “vehicles in the park” and “carries a firearm” from Muscarello v. United States, 524 U.S. 125 (1998) — tend to derive from the statutory interpretation context, as does the article’s discussion of textualism. See Tobia, supra note 4, at 730–31, 739. Overall, I read the article as more focused on the use of dictionary definitions and corpus linguistics as a guide to ordinary meaning in statutory, rather than contract, interpretation. In any event, because I am significantly more familiar with statutory interpretation law and theory, and with the use of dictionary definitions and corpus linguistics in the statutory rather than the contract law context, my focus in this Response will be on the implications of Tobia’s findings for statutory interpretation. — and I suggest that Tobia’s data do not support the strong version of his claim that different audiences judge statutory meaning similarly. Second, I consider the methodological implications of Tobia’s findings that ordinary meaning often is unclear and that dictionary definitions, corpus linguistics, and collective intuition about ordinary meaning often diverge from each other for a Supreme Court and bench that have moved increasingly toward a textualist approach to statutory interpretation.11×11. See, e.g., Harvard Law School, The 2015 Scalia Lecture | A Dialogue with Justice Elena Kagan on the Reading of Statutes at 8:28, YouTube (Nov. 25, 2015), https://youtu.be/dpEtszFT0Tg [https://perma.cc/KG5P-9PWX] (“I think we’re all textualists now.”); Jonathan R. Siegel, The Legacy of Justice Scalia and His Textualist Ideal, 85 Geo. Wash. L. Rev. 857, 858 (2017) (explaining that, due to Justice Scalia’s “tremendous influence,” “federal courts today generally acknowledge the importance of carefully examining statutory text”); Thomas W. Merrill, Essay, Textualism and the Future of the Chevron Doctrine, 72 Wash. U. L.Q. 351, 355 (1994) (describing textualism’s ascendance on the U.S. Supreme Court). Specifically, I suggest two metarules that courts might adopt to help curb judicial discretion and uncertainty over ordinary meaning: (1) a rule instructing that certain categories of statutes should be construed in light of their prototypical (or, conversely, legalist) meaning; and (2) a rule directing that differences in the ordinary meaning identified by dictionaries, corpus linguistics, different judges, and/or surveys of laypeople should be considered prima facie evidence that a statute is ambiguous and lacks a “plain” meaning.

I. The “Audience” Question: Who Is the Relevant “Ordinary Reader”?

Perhaps the most stunning finding Tobia reports is that judges and nonexperts have similar intuitions about the ordinary meaning of ordinary terms. Tobia bases this claim on survey data indicating that judges, law students, and laypeople were remarkably similar in the rates at which they categorized certain specific items (for example, “car,” “bus,” “airplane,” “canoe,” “roller skates”) as “vehicles” or not “vehicles.”12×12. See Tobia, supra note 4, at 766 & fig.5. But on closer inspection, there are a few problems with basing such a broad claim on this limited data. First, Tobia does not provide data regarding the relative rates at which judges and nonexperts attributed similar (or dissimilar) meanings to any of the other terms used in his study — that is, “carry,” “interpreter,” “labor,” “tangible object,” “weapon,” “animal,” “furniture,” “food,” and “clothing.”13×13. See id. at 764. (To be fair, the study did not gather data from judges and law students for these other terms.14×14. See Email from Kevin P. Tobia, Assistant Professor of L., Georgetown Univ. L. Ctr., to Anita S. Krishnakumar, Mary C. Daly Professor of L., St. John’s Univ. Sch. of L. (July 23, 2020) (on file with author). ) So it is unclear whether the symmetry he observes between judges, law students, and laypeople would be replicated in other contexts, with respect to other words; there could simply be something special about the word “vehicles” that produces greater consistency across interpreters than other terms would.

Second, and perhaps more importantly, Tobia’s data with respect to “vehicles” show noteworthy variation among judges and nonexperts in close or borderline cases, as opposed to easy cases. That is, while there appears to be little disagreement between judges and nonexperts that “cars,” “trucks,” and “buses” are vehicles — or that “drones,” “roller skates,” and “baby carriers” are not vehicles — there is considerable variation between these groups regarding whether borderline items such as an “electric wheelchair,” a “baby stroller,” or a “World War II Truck” that has been decorated as a World War II monument are vehicles.15×15. See Tobia, supra note 4, at 766 & fig.5 (indicating that roughly 66% of ordinary people considered the World War II Truck to be a “vehicle,” whereas only approximately 46%–47% of law students and judges agreed; similarly, roughly 51%–57% of ordinary people and law students considered an electric wheelchair to be a “vehicle,” whereas roughly 71% of judges thought electric wheelchairs were “vehicles”; and approximately 25%–26% of ordinary people and law students considered a baby stroller to be a vehicle, whereas roughly 41% of judges said the same). Not only are these differences between judges and nonexperts sizeable, but they also indicate that there are instances in which a large majority of nonexperts concluded that an entity qualified as a “vehicle,” while judges were unsure or evenly split on the question, and vice versa. Similar variations occur with respect to how judges and nonexperts apply dictionary definitions in borderline cases.16×16. See id. at 768 fig.6 (indicating that approximately 41%–52% of law students and ordinary people concluded that “roller skates” were “vehicles” based on dictionary definitions, while approximately 72%–73% of judges applying dictionary definitions reached the same conclusion; approximately 30%–36% of law students and ordinary people said “pogo sticks” were “vehicles” based on dictionary definitions, while 50% of judges applying dictionary definitions reached the same conclusion; and approximately 44%–51% of law students and ordinary people concluded that a “zip line” is a “vehicle” based on dictionary definitions, while approximately 62% of judges applying dictionary definitions reached the same conclusion). Again, these differences both are sizeable and indicate that there were instances in which a majority (and sometimes a large majority) of judges concluded that dictionary definitions supported considering the entity at issue a “vehicle,” while nonexperts split on the question or even overwhelmingly concluded that the entity was not a vehicle. Why does this matter? Because most legal disputes involve close cases: litigants do not tend to go to court to determine whether a “car” is a vehicle; they tend to go to court to resolve disagreements over whether borderline items like “baby strollers” or “electric wheelchairs” qualify as vehicles.

This variance also matters because if judges and laypeople disagree about the ordinary meaning of statutory terms in borderline cases, then the key question in such cases becomes: Who is the relevant audience or “ordinary reader” of the statute — judges, the average person on the street, or some other group of people? This is a crucial question in statutory interpretation but one that often is ignored.17×17. But see David S. Louk, The Audiences of Statutes, 105 Cornell L. Rev. 137, 159 (2019) (arguing that many disagreements in statutory interpretation cases may be attributed to conflicts in prioritizing competing statutory audiences and that different interpretive methodologies may be appropriate for construing statutes addressed to different audiences). It may be the case, for example, that for criminal statutes or statutes that deal with education, housing, or voting rights, the relevant audience or “ordinary reader” is the average person on the street. Conversely, for statutes that govern cost-shifting among litigants, jurisdiction or other matters of court procedure, or remedies, the relevant audience or “ordinary reader” may instead be judges. There are also potential equality and gender dimensions embedded in this audience question: it may be the case that certain terms have different meanings to men versus women or to people of different racial or ethnic backgrounds.

Tobia’s data, in my view, suggest that it might be worthwhile, in anticipation of the difficult cases that tend to make it to adjudication, for the legislature — or the judiciary — to establish default rules about who the relevant “ordinary reader” is for certain kinds of statutes or even certain kinds of statutory provisions. For example, the audience or relevant “ordinary reader” for those provisions of a statute that govern procedural or legal matters might be judges, whereas the audience for provisions that govern citizen behavior directly may be laypeople. Whether the relevant statutory audience should be determined for entire statutes or for particular kinds of statutory provisions is an important subquestion that courts, legislators, and scholars can and should think deeply about.18×18. One author, David Louk, has suggested that most statutes are directed at multiple audiences and that a “central task” for statutory interpreters therefore “should be to identify the principal audience” to which the statute is directed. Id. at 159; see also id. at 199. My aim in this Response is merely to raise the issue.

Operationally, default rules about the appropriate statutory audience could be established as follows: Congress, ideally, could specify who the relevant audience, or reader, is for individual statutes (or for particular kinds of statutory provisions) when it enacts them. In cases where a particular statutory provision is designed to have multiple audiences,19×19. See, e.g., Meir Dan-Cohen, Decision Rules and Conduct Rules: On Acoustic Separation in Criminal Law, 97 Harv. L. Rev. 625, 631 (1984) (“[A] single statutory provision may simultaneously guide both conduct and decision and may thus function as both a conduct rule and a decision rule.”). Congress could specify a “target audience” for purposes of judicial interpretation or, conversely, indicate clearly that the statute or particular provisions of the statute are directed at multiple audiences — and perhaps list those multiple audiences.20×20. Some may wonder whether it is feasible for Congress to agree on who constitutes a statute’s “target audience.” After all, as textualists have pointed out, the individual legislators who vote to enact a statute often have different intentions about the statute’s substantive meaning. See Daniel A. Farber & Philip P. Frickey, Law and Public Choice: A Critical Introduction 89–90 (1991) (summarizing Justice Scalia’s arguments against conceiving of legislative intent as unitary and coherent); John F. Manning, Competing Presumptions About Statutory Coherence, 74 Fordham L. Rev. 2009, 2029–30 (2006). While it is certainly fair to question whether Congress can muster the consensus necessary to identify a target audience for individual statutes or provisions, the type of consensus needed to establish who constitutes the relevant target audience for a statute (or provision) is different in kind from the type of consensus needed to agree on how a statute should be applied to specific factual circumstances. Indeed, because the interpretive outcomes likely to result from identifying a particular audience are not necessarily predictable ex ante, I suspect it would be easier for legislators to come to a consensus on this subject than on the specific meaning of a statutory provision.

Alternately, courts could establish their own default rules — similar to existing canons of construction that call for liberal or narrow construction of certain categories of statutes or provisions, or that favor certain parties when a statute is ambiguous21×21. Examples include the canon calling for narrow interpretation of exemptions from federal taxation, see United States v. Burke, 504 U.S. 229, 248 (1992) (Souter, J., concurring in the judgment); United States v. Wells Fargo Bank, 485 U.S. 351, 357 (1988), the principle that veterans’ benefits statutes should be liberally construed, see King v. St. Vincent’s Hosp., 502 U.S. 215, 220 n.9 (1991), the rule that ambiguities in criminal statutes should be construed in favor of the accused, see United States v. Davis, 139 S. Ct. 2319, 2333 (2019), and the rule that ambiguities in deportation statutes should be construed in favor of aliens, see INS v. St. Cyr, 533 U.S. 289, 320 (2001). — specifying the audience that courts should bear in mind when seeking to identify a statute’s “ordinary meaning.” In individual cases, a court might choose to ignore the default rule, perhaps because the provision at issue is one that speaks to a different audience than the overall statute, but in so doing the court would at least have to confront the audience question — something that current statutory interpretation canons and rules do not require.22×22. An alternative to default rules, suggested by Louk, is that judges should explicitly state the assumptions about statutory audience that they are making when they construe statutes. See Louk, supra note 17, at 147. In the interest of ensuring greater predictability and minimizing judicial disagreement after the stakes of a particular case are known, I advocate establishing ex ante default rules instead.

To see how these recommendations might work in practice, consider two leading statutory interpretation cases that highlight the importance of the audience question and that demonstrate how default rules could help constrain the judicial search for ordinary meaning.

In Taniguchi v. Kan Pacific Saipan, Ltd.,23×23. 566 U.S. 560 (2012). the Court considered the meaning of the term “interpreters,” which appeared in a statute that listed “compensation of interpreters” as one of several kinds of litigation costs that courts have the power to award prevailing parties.24×24. Id. at 565–66 (quoting 28 U.S.C. § 1920(6)). A majority of the Court held that the term “interpreters” did not encompass written document translation services paid for by the plaintiff, citing fourteen dictionary definitions that referred to “interpreters” as persons who provide oral translation services.25×25. See id. at 566–68, 568 n.2. Justice Ginsburg dissented, citing three legal (and other) dictionaries, as well as several federal court of appeals and district court decisions treating translators of written documents as “interpreters.”26×26. See id. at 576–77 (Ginsburg, J., dissenting). Justice Ginsburg’s dissent did not directly address the audience question, but it did express her view that the “key” context for determining the meaning of the term “interpreters” was “the practice of federal courts both before and after [the statute’s] enactment.”27×27. Id. at 579. In other words, Justice Ginsburg seemed to believe that the relevant audience, or readers, of the cost-shifting statute were federal judges, although she framed her analysis in terms of federal courts’ past “practice” rather than their understanding of the ordinary meaning of the term “interpreters.”28×28. See id. at 576–79. The majority opinion did not grapple with the audience question at all. One wonders whether the case would have come out differently if the Court’s inquiry into the ordinary meaning of “interpreters” had been framed in terms of who — judges or the average person on the street — constituted the relevant audience for the statute.

In another memorable case, McNally v. United States,29×29. 483 U.S. 350 (1987). Justice Stevens argued, in dissent, that there were no due process notice problems with applying a federal mail fraud statute to a Kentucky public official and private individuals engaged in a kickback scheme through which an insurance company hired by the State of Kentucky funneled commissions to the personal accounts of the public official and other politically active party members.30×30. See id. at 375 & n.9 (Stevens, J., dissenting). The federal statute prohibited the use of the mails to execute “any scheme or artifice to defraud, or for obtaining money or property by means of false or fraudulent pretenses, representations, or promises.”31×31. 18 U.S.C. § 1341. The statutory confusion arose because although the Kentucky official and others had been enriched by the kickback scheme, they had not actually defrauded the citizens of Kentucky out of any money or property.32×32. See McNally, 483 U.S. at 360–61. Justice Stevens explained that in determining the meaning and reach of the statutory terms at issue, “it is appropriate to identify the class of litigants” to whom the statute is being applied.33×33. See id. at 375 n.9 (Stevens, J., dissenting). In this case, that class of litigants consisted of “the most sophisticated practitioners of the art of government among us,” and Justice Stevens noted that the “government executives, judges, and legislators who have been accused, and convicted, of mail fraud under the well-settled construction of the statute . . . are people who unquestionably knew that their conduct was unlawful.”34×34. Id. In Justice Stevens’s view, then, the relevant audience, or ordinary reader, of the mail fraud statute was sophisticated government officials — not the average layperson.

As Taniguchi and McNally illustrate, courts have been dancing around questions about the relevant audience or “ordinary reader” of a particular statute for years, although they rarely confront the question squarely. And while Tobia’s data suggest that the audience question may be inconsequential for easy cases, in which different audiences are likely to identify the same ordinary meaning, his data also highlight the necessity and potential benefits of clarifying the relevant audience for those hard cases that inevitably will arise.

Specifying the relevant audience or “ordinary reader” might also help dictate which external sources interpreters should consult to help identify a statute’s ordinary meaning: certain kinds of dictionaries may be appropriate for certain kinds of statutes or terms — for example, legal dictionaries for statutes dealing with court procedure, popular dictionaries or perhaps corpus linguistics for criminal statutes, and possibly even medical or scientific dictionaries for certain statutes. Similarly, certain corpora may be more appropriate for certain kinds of statutes than for others — for example, the TV Corpus or Movie Corpus may be better for criminal statutes (as guides to popular meaning), whereas the Corpus of U.S. Supreme Court Opinions may be more appropriate for procedural statutes.

Finally, Tobia’s study assumes that there is such a thing as a “correct” baseline ordinary meaning for statutory terms — one that can be identified by measuring the collective intuition of laypeople — and against which the accuracy of dictionary definitions and corpus linguistics can be gauged. But this assumption itself elides the audience question — in that it assumes that the collective intuition of ordinary people represents the correct baseline ordinary meaning. (Note, again, that Tobia’s study does not measure the intuitions of judges or law students for terms other than “vehicles” — so his assessments of the accuracy of dictionary definitions and corpus linguistics depend on ordinary people’s intuitions, as surveyed on Amazon’s Mechanical Turk, as the baseline for “ordinary meaning.”) As I have suggested above, that may be appropriate for some statutes; but for others, the relevant audience or reader should be judges and/or lawyers — and the relevant baseline measure for ordinary meaning should be judicial or other legal professionals’ intuitions.

II. Ordinary Meaning Subcanons

Tobia also concludes that the ordinary meaning of specific words often is unclear, based on survey data indicating that laypeople regularly disagreed with each other about the ordinary meaning of several different statutory terms35×35. Tobia’s study also showed, as discussed above, that judges and law students disagreed among themselves about the ordinary meaning of the term “vehicle.” See Tobia, supra note 4, at 766 fig.5. and that dictionary definitions, corpus linguistics, and collective intuition about ordinary meaning often diverged from each other.36×36. See id. at 762–64. He further concludes that dictionary definitions tend to encompass a broad, or expansive, reading of the term at issue, whereas corpus linguistics analysis tends to reflect a word’s prototypical meaning; this was evidenced by the fact that study subjects who applied dictionary definitions concluded that nearly every entity tested (from cars to zip lines) qualified as a “vehicle,” whereas subjects asked to apply corpus linguistics took a narrower view of the meaning of “vehicle” and were less likely to conclude that any of the tested entities qualified.37×37. See id. at 756–58, 757 fig.1, 770–72, 770 fig.8, 771 fig.9.

As noted above, I am unsurprised by these findings — in part because my own empirical research is consistent with them. In a recent study on the extent of judicial dueling over interpretive resources in the U.S. Supreme Court, for example, I measured how often majority and dissenting opinions in the same case used the same interpretive tool to reach opposing outcomes.38×38. See Anita S. Krishnakumar, Dueling Canons, 65 Duke L.J. 909, 912 (2016). For plain or ordinary meaning analysis, I found a 42.7% rate of judicial dueling — meaning that in over 40% of the Court’s divided vote cases in which at least one opinion argued that the statute had an ordinary or plain meaning, an opposing opinion countered that the statute had a different ordinary meaning.39×39. See id. at 929–30 & tbl.1. Perhaps even more interestingly, I found that in 41.2% of the cases in which majority and dissenting opinions disagreed about a statute’s plain meaning, one opinion advocated adopting the “core” or “prototypical” meaning of the word at issue while the other focused on the broad or legalist meaning of the word.40×40. See id. at 962–63, 962 nn.146–47.

One solution, or response, to Tobia’s experimental findings (or mine, for that matter) is to acknowledge that dictionaries and corpus linguistics cannot serve as dispositive determinants of ordinary meaning and to urge that interpreters use both of these sources in tandem or that they consult other contextual clues, such as statutory purpose, alongside such sources to determine whether a statutory term should be given its prototypical or legalist meaning (dictionary or corpus meaning). These are the solutions Tobia suggests toward the end of his article.41×41. See Tobia, supra note 4, at 803–04. But there are other possible responses to these experimental findings about the indeterminacy of ordinary meaning. One is to seek to reduce, ex ante, the universe of possible ordinary meanings among which judges can choose — on the theory, subscribed to by many textualists in the context of interpretive tools such as legislative history and statutory purpose, that if given multiple options, judges will find it far too easy to choose an ordinary meaning that fits their ideological policy preferences.42×42. See, e.g., Antonin Scalia & Bryan A. Garner, Reading Law: The Interpretation of Legal Texts 18–19 (2012) (arguing that each statute has multiple purposes and that purposivism empowers judges to frame the statute’s purpose at the level of generality that suits their desired outcome); Antonin Scalia, Common-Law Courts in a Civil-Law System: The Role of United States Federal Courts in Interpreting the Constitution and Laws, in A Matter of Interpretation 3, 36 (Amy Gutmann ed., 1997) (“In any major piece of legislation, the legislative history is extensive, and there is something for everybody. . . . The variety and specificity of result that legislative history can achieve is unparalleled.”). Another possible response is to treat disagreement about ordinary meaning in dictionary definitions versus corpora, or in laypersons’ survey responses, or in judicial decisions, as establishing a prima facie case of ambiguity that would in turn trigger a move to second-order interpretive canons or tools — and would prompt courts to abandon the search for ordinary meaning altogether. This section explores these latter two possibilities.

A. Prototypical vs. Expansive Meaning

One possible lesson from Tobia’s study is that it might be helpful for interpreters to know in advance — ex ante, before a particular dispute is before them — whether a given statute, or particular kinds of statutory provisions, should be interpreted in light of their prototypical or their expansive meaning. That is, it might be helpful if the ordinary meaning canon or rule contained metarules, or subcanons, dictating that certain statutes, or certain categories of statutes or provisions, should be given their prototypical or expansive meaning.43×43. For example, ex ante rules might dictate that antitrust statutes should be given an expansive meaning, that statutory provisions that create exceptions should be given their prototypical meaning, or that provisions that contain lists should be given their prototypical meaning. Such metarules might be established in a number of ways:

1. Congress, when enacting individual statutes, could specify that a statute should be interpreted in light of its prototypical meaning or, conversely, in an expansive or legalist manner.44×44. Cf. Abbe R. Gluck, The States as Laboratories of Statutory Interpretation: Methodological Consensus and the New Modified Textualism, 119 Yale L.J. 1750, 1824–27 (2010) (discussing general interpretive rules mandated by state legislatures and the possibility that Congress could do something similar). Alternately, Congress could enact a more general statute, along the lines of the Dictionary Act,45×45. 1 U.S.C. §§ 1–8. dictating that certain categories of statutes should be given either their prototypical or expansive meaning — for example, terms in criminal statutes should be interpreted in light of their prototypical meaning, while terms in antitrust statutes should be interpreted in light of their expansive, or legalist meaning.

2. Another option would be for courts, and the Supreme Court in particular, to establish ex ante rules dictating that certain kinds of statutes, or provisions within statutes, should be interpreted based on their prototypical or legalist meaning. If the Court does not wish to make such broad pronouncements, it could make a determination about whether a statute should be interpreted in light of its prototypical or legalist meaning on a case-by-case basis, for each individual statute that comes before it — and that determination could then bind lower courts in the future. Lower courts also could analogize from those statutes with respect to which the Court has articulated such a metarule to other similar statutes the Court has not yet evaluated.

3. A third, path-of-least-resistance alternative is that courts could turn existing liberal or narrow construction rules into metarules about expansive versus prototypical meaning.

Courts already recognize a number of interpretive rules, often called “canons of construction,” that essentially tell interpreters to read statutory terms broadly or narrowly. Examples include the canon calling for narrow interpretation of exemptions from federal taxation,46×46. See United States v. Burke, 504 U.S. 229, 244 (1992) (Scalia, J., concurring in the judgment); id. at 248 (Souter, J., concurring in the judgment); United States v. Wells Fargo Bank, 485 U.S. 351, 357 (1988). the canon directing that veterans’ benefits statutes should be liberally construed,47×47. See King v. St. Vincent’s Hosp., 502 U.S. 215, 220 n.9 (1991). and the whole act rule directive that provisos (statutory provisions that create exceptions) should be narrowly construed to cover only those items that clearly fall within the exception.48×48. See Maracich v. Spears, 570 U.S. 48, 60 (2013); Comm’r v. Clark, 489 U.S. 726, 739 (1989). Such canons rather easily could be transformed into metarules, or ordinary meaning subcanons, dictating that tax exemptions should be given their prototypical rather than expansive meaning, that veterans’ benefits statutes should be construed in terms of their expansive meaning, and that provisos should be given their prototypical meaning. A canonical direction to construe a statute narrowly is, in essence, a command to give the statute’s terms only those meanings that clearly fall within its core (or prototypical) coverage; likewise, a directive to construe a statute liberally is, in essence, a command to construe the statute’s terms expansively to encompass all that they reasonably can cover.49×49. Cf. William N. Eskridge Jr. et al., Cases and Materials on Legislation and Regulation: Statutes and the Creation of Public Policy 690 (5th ed. 2014) (equating liberal construction canons with a directive to apply the statute “expansively to new situations”). Accordingly, courts should be able to convert many existing canons into metarules about prototypical versus expansive meaning in a relatively straightforward manner.

Once such metarules are established — whether by Congress or the courts — those rules in turn could be used to guide interpreters’ reliance on dictionaries or corpus linguistics as external aids to ordinary meaning. That is, where metarules dictate that a statute should be interpreted in light of its prototypical meaning, interpreters might use corpus linguistics as an interpretive aid; whereas where metarules indicate that a statute should be interpreted in light of its expansive meaning, interpreters might instead consult dictionary definitions as an external aid.

Ordinary meaning metarules, or subcanons, of the kind recommended above would have a number of interpretive advantages. First, they would cabin the universe of potential ordinary meanings among which judges can choose — and thus reduce judicial discretion and opportunities for decisionmaking based on ideology or personal policy preferences. Second, and relatedly, metarules would help ensure consistency in the interpretation of the same statute. Under our current system of statutory interpretation, there is no methodological stare decisis (at least in federal courts) dictating that once one provision of Statute A has been construed using X, Y, and Z canons or tools, other provisions of that same statute must also, in the future, be construed using the same canons or interpretive tools.50×50. See, e.g., Gluck, supra note 44, at 1822–23; Sydney Foster, Should Courts Give Stare Decisis Effect to Statutory Interpretation Methodology?, 96 Geo. L.J. 1863, 1878 (2008) (describing the Court’s shift, in a series of cases involving the implication of private rights of action, from ignoring legislative intent as an interpretive resource, to considering intent as one among several factors, and finally to treating intent as the single determinant factor). Thus, a court may interpret one provision of Title VII using legislative history and statutory purpose but later interpret another provision (or even another term in the same provision!) using the whole act rule and dictionary definitions. And there is no guidance dictating that if a court gives a term in a particular statutory provision its prototypical meaning in Case 1, it must give another provision in that same statute or even another term in that same statutory provision its prototypical meaning in Case 2.51×51. Cf. Tobia, supra note 4, at 789 (noting that Republican appointees often interpret definitions of words broadly in the Second Amendment context and narrowly in the Eighth Amendment context, while Democratic appointees, when they use dictionaries, tend to do the reverse). Thus, courts — including the U.S. Supreme Court — are currently free to construe the term “national origin” in Title VII’s list of prohibited bases for employment discrimination in light of its prototypical meaning but to construe the term “religion” in that same list of prohibited bases for discrimination expansively, in light of its legalist meaning.52×52. See 42 U.S.C. § 2000e-2(a)(1). Title VII makes it unlawful for employers to engage in employment practices that discriminate against any individual “because of such individual’s race, color, religion, sex, or national origin.” Id. That is, courts may limit the term “national origin” to its core applications, leaving out borderline cases, while interpreting the term “religion” — which appears just three words away from “national origin” in the same statutory sentence — broadly to encompass borderline applications.53×53. See id.

If, however, Congress or the Court were to establish a metarule dictating that Title VII (or civil rights or antidiscrimination statutes generally) should be interpreted in light of its legalist (or prototypical) meaning, then all of the terms appearing in Title VII’s employment discrimination provision would have to be construed using the same kind of ordinary meaning — that is, prototypical or legalist. We do not have to look far to realize the practical implications of this kind of metarule: one way of understanding Justices Gorsuch’s and Kavanaugh’s opposing opinions in the recently decided Bostock v. Clayton County54×54. 140 S. Ct. 1731 (2020). case is as a clash over precisely this kind of legalist versus prototypical meaning analysis of the statutory phrase “because of sex.” In that case, Justice Gorsuch (and a majority of the Court) concluded that the phrase should be given an expansive, or legalist meaning, to include discrimination based on sexual orientation or gender identity; whereas Justice Kavanaugh (and two other dissenters) concluded that the phrase should be limited to its prototypical meaning, which would not include such discrimination.55×55. Compare id. at 1739, 1741–42, with id. at 1755, 1767, 1769 (Alito, J., dissenting), and id. at 1824–28 (Kavanaugh, J., dissenting). Justice Alito’s dissent was joined by Justice Thomas. Id. at 1754 (Alito, J., dissenting). Notably, Justice Gorsuch’s opinion framed itself as interpreting “sex” in light of dictionary definitions that equate “sex” with biological distinctions — arguing that one cannot make distinctions on the basis of sexual orientation or gender identity without considering, at some level, biological differences. Id. at 1739, 1741–42. Justice Kavanaugh’s dissent countered that this is true only in a highly literal (or legalist) sense, pointing out that no one in common conversation would say that a gay person was fired because of his or her “sex.” See id. at 1828 (Kavanaugh, J., dissenting). Justice Gorsuch’s majority opinion is in many ways a classic legalist opinion, with its insistence that “conversational conventions do not control Title VII’s legal analysis” and that such analysis depends on whether “sex” was a but-for cause of the employer’s discrimination. Id. at 1745 (majority opinion). Conversely, Justice Kavanaugh’s dissenting opinion is, at bottom, an argument that the prototypical meaning of discrimination “because of sex” does not include discrimination based on sexual orientation or gender identity. Indeed, the “literal” meaning criticized by Justice Kavanaugh is, essentially, the expansive, or legalist meaning of a term; while the “common parlance,” id. at 1828 (Kavanaugh, J., dissenting), meaning is another way of getting at the core, or prototypical (most common), meaning of the term.

B. Prima Facie Ambiguity

Tobia’s experimental findings also suggest two other possible metarules that could help clarify, or standardize, the role played by ordinary meaning analysis in statutory interpretation. In particular, his findings that laypeople (and judges) often disagree about the ordinary meaning of specific words and that dictionaries and corpora often produce conflicting ordinary meanings suggest that courts perhaps should be more willing, more often, to conclude that a given statute is ambiguous. There are at least two ways in which such a suggestion could be operationalized. First, interpreters could check dictionary definitions and corpus linguistics research results against each other: if and when they find conflicting meanings, that divergence itself could be taken as prima facie evidence that the statute has no plain or ordinary meaning — and that interpreters should move on to other interpretive tools to determine the meaning of the term at issue. This approach would have the effect of shifting the court’s focus, in cases where dictionary definitions and corpus linguistics produce different meanings, from identifying a statute’s ordinary meaning to identifying the meaning that makes the most sense in light of the statute’s other provisions or structure (the whole act rule), logical deductions encompassed in language canons, its purpose or legislative history, policy norms embodied in substantive canons of construction, and so on.

A second, alternate approach would be for courts to use the lack of a clear or internally consistent collective intuition among laypeople or judges as prima facie evidence that a statute has no clear meaning. That is, if laypeople or judges demonstrate substantial disagreement about a term’s ordinary meaning, that fact itself could be taken as prima facie evidence that the statutory term at issue is ambiguous — and that courts must determine statutory meaning based on other canons or interpretive tools. Judicial disagreement is relatively easy to measure; divergence among the judges on an appeals court or lower court splits over a statute’s ordinary meaning could be used as gauges. Indeed, a few state court judges have suggested, in opinions issued in specific cases, that this kind of judicial disagreement should be taken as conclusive evidence of ambiguity.56×56. See, e.g., Peace v. Nw. Nat’l Ins. Co., 596 N.W.2d 429, 449 (Wis. 1999) (Abrahamson, C.J., dissenting) (“When numerous courts disagree about the meaning of language, the language cannot be characterized as having a plain meaning. Rather, the language is ambiguous; it is capable of being understood in two or more different senses by reasonably well-informed persons even though one interpretation might on careful analysis seem more suitable to this court.”); Nedlloyd Lines B.V. v. Superior Ct., 834 P.2d 1148, 1170 (Cal. 1992) (Kennard, J., concurring and dissenting) (“[T]he very fact that [two other justices] disagree with the majority regarding the meaning of the clause, and that both the majority and these two justices find the clause clear, but conclude it has opposite meanings, ironically and convincingly demonstrates that the clause is ambiguous.”); see also E. Sanderson Hoe & Mary E. Buxton, The Plain Meaning Rule in the Federal Circuit: An Update of Case Law After Coast Federal Bank, FSB v. United States, 42 Pub. Cont. L.J. 151, 153 (2012) (“[T]he fact that judges reach conflicting conclusions on the meaning of contract language should be the best evidence that an ambiguity exists.”). Conversely, lower court consensus could be taken as evidence that there is substantial agreement among judges about the statute’s ordinary meaning.57×57. Again, there is some precedent for this. See Aaron-Andrew P. Bruhl, Following Lower-Court Precedent, 81 U. Chi. L. Rev. 851, 859 (2014).

With respect to laypeople, Tobia’s experimental study and others like it suggest the possibility that courts might take robust survey data into account as evidence of the lack of a clear ordinary meaning.58×58. See, e.g., James A. Macleod, Ordinary Causation: A Study in Experimental Statutory Interpretation, 94 Ind. L.J. 957, 962–64 (2019); Omri Ben-Shahar & Lior Jacob Strahilevitz, Interpreting Contracts via Surveys and Experiments, 92 N.Y.U. L. Rev. 1753, 1758 (2017) (arguing that courts should rely on survey evidence in contract interpretation); Shlomo Klapper, Soren Schmidt & Tor Tarantola, Ordinary Meaning from Ordinary People 42–43 (July 17, 2020) (unpublished manuscript), https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3593917 [https://perma.cc/3L7D-6VNZ] (suggesting that courts and legislators could use survey data to identify the ordinary meaning of disputed statutory terms). That is, if a survey of laypeople indicates that there is substantial disagreement over the ordinary meaning of a statutory term — perhaps defined by data indicating that less than sixty-five percent of those surveyed agreed on the same meaning — courts should consider that prima facie evidence that the statute lacks a clear meaning. There is an emerging literature suggesting that survey data should be used to determine the actual ordinary meaning of terms in contract and statutory interpretation;59×59. See sources cited supra note 58. my suggestion is to instead use such data to determine whether a statutory term or phrase has a readily identifiable ordinary meaning at all.

There is, of course, a risk of opportunistic survey design or data presentation once the stakes of a particular survey question are known in an actual, live case. This suggests that litigants and their attorneys should not be the source of such survey data. But there are other, more reliable ways to obtain such data. For example, academics could conduct surveys similar to Tobia’s and submit them to the court in amicus briefs. Academics could even seek to head off potential allegations of bias by testing several terms in well-known statutes ex ante, ahead of litigation — and then publishing or making those survey results available independent of pending lawsuits. Alternately, law clerks or long-term court employees or special masters could be trained to perform such surveys and could conduct them when cases arise.

A second, less radical metarule suggested by Tobia’s data is that judicial, lay, or dictionary-versus-corpora disagreement over a statute’s ordinary meaning could be used to make threshold determinations about statutory clarity with respect to canons that are triggered only when a statute is deemed to be ambiguous. Sometimes referred to as “clarity doctrines,” such canons include interpretive rules like the rule of lenity, the canon of constitutional avoidance, and the first step of the Chevron60×60. Chevron U.S.A. Inc. v. Nat. Res. Def. Council, Inc., 467 U.S. 837 (1984). deference test.61×61. The rule of lenity directs that ambiguities in criminal statutes be resolved in favor of the defendant. See 3 Norman J. Singer & J.D. Shambie Singer, Statutes and Statutory Construction § 59:3 (7th ed. 2008); United States v. Wiltberger, 18 U.S. (5 Wheat.) 76, 95 (1820). The canon of constitutional avoidance instructs that if there are two or more plausible readings of a statute (that is, if the statute is ambiguous), and one of these raises serious constitutional concerns, the Court should adopt the reading that avoids the constitutional problem. See, e.g., Rapanos v. United States, 547 U.S. 715, 737–38, 738 n.9 (2006) (plurality opinion); Edward J. DeBartolo Corp. v. Fla. Gulf Coast Bldg. & Constr. Trades Council, 485 U.S. 568, 575 (1988); United States v. Del. & Hudson Co., 213 U.S. 366, 407–08 (1909). The first step of the Chevron test, often referred to as Chevron Step One, directs courts to determine “whether Congress has directly spoken to the precise question at issue” and whether Congress’s intent “is clear.” Chevron, 467 U.S. at 842. Clarity doctrines have received a fair amount of academic attention recently. Justice Kavanaugh and Professor Lawrence Solan, for example, have argued forcefully for eliminating (or at least curbing) reliance on clarity doctrines, on the grounds that the question of whether a statute is clear or ambiguous is often itself ambiguous.62×62. See Brett M. Kavanaugh, Fixing Statutory Interpretation, 129 Harv. L. Rev. 2118, 2144 (2016) (book review); Lawrence M. Solan, Pernicious Ambiguity in Contracts and Statutes, 79 Chi.-Kent L. Rev. 859, 859 (2004) (“The problem, perhaps ironically, is that the concept of ambiguity is itself perniciously ambiguous.”). Others have suggested that courts establish clarity thresholds akin to confidence levels — for example, ninety percent confidence that they have identified the correct reading of the statute — before a statute may be declared “clear.”63×63. See, e.g., Richard M. Re, Clarity Doctrines, 86 U. Chi. L. Rev. 1497, 1502–03, 1502 n.19 (2019).

Tobia’s experimental study, by contrast, suggests a possible measure of clarity that is far more concrete than any previously recommended measure and one that could mitigate concerns about “the ambiguity of ambiguity”64×64. See, e.g., Solan, supra note 62, at 883 (emphasis omitted). that have motivated some to advocate abandoning clarity doctrines altogether.65×65. See Re, supra note 63, at 1504. That is, courts could use one or more of the indicators of disagreement over ordinary meaning discussed above as prima facie evidence of ambiguity sufficient to trigger the relevant clarity doctrine. In other words, where judges, laypeople, or dictionaries-versus-corpora are split over a statutory term’s ordinary meaning, that disagreement itself could be taken as presumptive evidence that the statute is ambiguous — and that the rule of lenity, avoidance canon, Chevron deference, or other relevant clarity doctrine should be applied to decide the statutory question at issue. Of course, courts would have to establish ex ante what level of disagreement among judges, laypersons, or dictionaries-versus-corpora would suffice to trigger the prima facie presumption. But once that determination is made, Tobia’s experimental study suggests several ways to measure the level of agreement. As a starting point, I would propose something like a two-thirds majority rule — requiring that at least sixty-five percent of judges or laypersons who have considered the statutory question agree on its ordinary meaning in order for the statute to be considered “clear.” This would mean that where judges or laypersons split anywhere in the range from 50%–50% to 64%–36% about the statute’s meaning, interpreters should turn to default rules designed to apply in the absence of textual clarity.

For dictionaries-versus-corpora, the measure could be simpler — for example, determining (1) whether definitions from different dictionaries produce consistent outcomes; (2) whether the meanings identified across different relevant corpora are consistent; and (3) if different dictionaries and corpora both produce internally consistent meanings, comparing the meaning produced by dictionaries to the meaning produced by corpora. In cases where different dictionaries or different corpora suggest inconsistent meanings, that internal inconsistency itself could be taken as evidence of ambiguity. Where different dictionaries and different corpora overwhelmingly (again, sixty-five percent could be used as a measure) produce a consistent meaning, the meaning generated by each source could then be compared for consistency; where consistency is found, the statute would be declared “clear,” and where inconsistency is found, it would be declared ambiguous.


The experimental study described in Tobia’s Testing Ordinary Meaning is rich and full of valuable data about how laypersons judge ordinary meaning — and about the specific kinds of ordinary meaning that dictionaries and corpus linguistics tend to measure. In my view, it oversells a bit regarding the breadth of the information it provides about how judges judge ordinary meaning, as judges were asked to evaluate the meaning of only one of ten terms about which laypeople were surveyed. As a result, it is difficult to tell whether judges and nonexperts really do judge ordinary meaning similarly, or if the way they evaluated the one statutory term studied across all three groups (“vehicle”) was anomalous. This observation is important because it implicates significant questions about who the relevant audience, or ordinary reader, is for specific statutes: for some statutes, the answer may be ordinary citizens, or laypeople, but for others it may be judges. The lack of deeper data about judges should not, however, detract from what is otherwise a splendid article. Tobia’s data regarding the intersection between how laypersons, dictionaries, and corpus linguistics define ordinary meaning is incredibly useful and important. Indeed, both the data and methodological approach of his study suggest numerous possible ways that courts or legislatures might narrow the universe of potential ordinary meanings, ex ante, for specific statutes — in order to cabin judicial discretion and promote greater predictability in statutory interpretation. I have suggested a few such possibilities in this Response; it is my hope that Tobia’s article will prompt further discussion about how best to use dictionaries and corpus linguistics — and how best to define what is meant by “ordinary meaning” — in the future.


* Mary C. Daly Professor of Law, St. John’s University School of Law, Visiting Professor of Law, Georgetown University Law Center. I owe deep thanks to Deborah A. Widiss and Tara Leigh Grove for valuable comments on earlier versions of this Response. Special thanks also to Dean Michael A. Simons and St. John’s University School of Law for generous research support. All errors are my own.