1 Judging Ordinary Meaning 3-23-17
JUDGING ORDINARY MEANING
Thomas R. Lee & Stephen C. Mouritsen*
Abstract
Judges generally begin their interpretive task by looking for the ordinary
meaning of the language of the law. And they often end there—out of respect for
the notice function of the law or deference to the presumed intent of the
lawmaker.
Most everyone agrees on the primacy of the ordinary meaning rule. Yet
scholars roundly bemoan the indeterminacy of the communicative content of the
language of the law. And they pivot quickly to other grounds for interpretation.
We agree with the diagnosis of important scholars in this field—from Richard
Fallon and Cass Sunstein to Will Baude and Steve Sachs—but reject their
proposed cures. Instead of setting aside the threshold question of ordinary
meaning we seek to take it seriously. We do so through theories and methods
developed in the scholarly field designed for the study of language—linguistics.
We identify theoretical and operational deficiencies in our law’s attempts
to credit the ordinary meaning of the law and present linguistic theories and
tools to assess it more reliably. Our framework examines iconic problems of
ordinary meaning—from the famous “no vehicles in the park” hypothetical to
two Supreme Court cases (United States v. Muscarello and Taniguchi v. Kan
Pacific Saipan) and a Seventh Circuit opinion of Judge Richard Posner (in
United States v. Costello). We show that the law’s conception of ordinary
meaning implicates empirical questions about language usage. And we present
linguistic tools from a field known as corpus linguistics that can help to answer
these empirical questions.
When we speak of ordinary meaning we are asking an empirical question—
about the sense of a word of phrase that is most likely implicated in a given
linguistic context. Linguists have developed computer-aided means of answering
such questions. We propose to import those methods into the law of
interpretation. And we consider and respond to criticisms of their use by
lawyers and judges.
*
Thomas R. Lee is Associate Chief Justice of the Utah Supreme Court and a
Distinguished Lecturer of Law at Brigham Young University. Stephen C. Mouritsen is
an Associate at the University of Chicago Law School and Adjunct Professor of Law
and Corpus Linguistics at Brigham Young University.
The authors express thanks to those who commented on earlier drafts or offered
insights in response to presentations in various conferences, symposia, and talks,
including Randy Barnett, Will Baude, Ryan Doerfler, Ed Finegan, Jonah Gelbach, Dan
Klerman, Kurt Lash, John Manning, Noah Messing, Gordon Smith, Larry Solan, Larry
Solum, Eugene Volokh, and John Yoo. Thanks also to Brigham Young University and
to the Georgetown University Law Center for co-sponsoring a conference on law and
corpus linguistics, at which the ideas in this piece were initially vetted, and to the OlinSearle-Smith
Fellows in Law program for making possible Mr. Mouritsen‘s association
with the University of Chicago Law School.
2 Judging Ordinary Meaning 3-23-17
INTRODUCTION
A key component of the meaning we ascribe to law concerns its
―communicative content‖—the ―linguistic meaning‖ of the words of a
statute or regulation,1
encompassing the ―intended‖ meaning of the
lawmaker or the ―contextual meaning‖ understood by the public.2
This is
the threshold question for many problems of legal interpretation. It is the
―standard picture‖ painted by most theories of interpretation, which starts
with a search for the ―ordinary communicative content‖ of the words of
the law.3
―There are excellent reasons for the primacy of the ordinary meaning
rule.‖4
Most of them stem from the purported determinacy of the ordinary
meaning inquiry. We speak of a search for meaning ―not in the
subjective, multiple mind of Congress but in the understanding of the
objectively reasonable person.‖5
And we generally conclude that the
search for such meaning ―matches up well with our understanding of
what the rule of law entails:‖6
It assures notice to the public, protects
reliance interests, assures consistency of application, and respects the will
of the legislative body.7
So although we recognize that ―ordinary
meaning does not always yield predictable answers to statutory issues,‖
we tend to accept that it ―yield[s] greater predictability than any other
single methodology.‖8
1
Lawrence B. Solum, Communicative Content and Legal Content, 89 NOTRE DAME
L. REV. 479, 480 (2013) (distinguishing the ―communicative content‖ of a legal text
from its ―legal content,‖ or in other words ―the legal norms the text produces‖).
2
Richard H. Fallon, Jr., The Meaning of Legal “Meaning” and Its Implications for
Theories of Legal Interpretation, 82 U. CHI. L. REV. 1235, __- __ (2015) (speaking of
these and other conceptions of the communicative or ―conversational‖ content of the
words of the law).
3
See William Baude & Stephen E. Sachs, The Law of Interpretation, 130 HARV. L.
REV. 1079, 1086 (2017) (speaking of the ―standard picture,‖ or ―view that we can
explain our legal norms by pointing to the ordinary communicative content of our legal
texts,‖ or in other words ―an instrument‘s meaning as a matter of language‖); see also
id. at 1082 n.2 (borrowing the ―standard picture‖ terminology from Mark Greenberg,
The Standard Picture and Its Discontents, in 1 OXFORD STUDIES IN PHILOSOPHY OF
LAW 39, 48 (Leslie Green & Brian Leiter eds., 2011)).
4
WILLIAM N. ESKRIDGE, JR., INTERPRETING LAW: A PRIMER ON HOW TO READ
STATUTES AND THE CONSTITUTION 39-40 (2016).
5
Frank H. Easterbrook, The Role of Original Intent in Statutory Construction, 11
HARV. J. L & PUB. POL‘Y 59, 65 (1988).
6
ESKRIDGE, supra note 4, at 35.
7
See id. at (―A polity governed by the rule of law aspires to have legal directives that
are known to the citizenry, that are predictable in their application, and that officials can
neutrally and consistently apply based upon objective criteria.‖).
8
Id.
3 Judging Ordinary Meaning 3-23-17
This premise has taken hold in our courts. ―[W]e‘re all textualists
now.‖9
That holds true at least in the sense that most all judges begin the
interpretive inquiry with the words of a statute—and even end there if
they find the meaning of those words to be ―plain.‖10
Yet the academy has been less sure of the premises of this trend.
Scholars generally endorse the value of determinacy but roundly doubt
the judge‘s ability to find it in mere ―communicative content‖ or
―ordinary meaning‖ of statutory text.11
There are two dimensions to this
skepticism—questions about the meaning of the law‘s search for
―ordinary meaning‖ and concerns about a judge‘s ability to measure or
assess it with any degree of determinacy.12
We share these concerns but offer a different solution. In Part I of this
article we show that the law has done a poor job conceptualizing the
notion of ordinary meaning. And we demonstrate that ―[u]ncertainty and
division‖ in assessing such meaning ―seem inevitable‖ under the methods
resorted to by judges in this field.13
But we do not see these problems as
an invitation to abandon the search for the ordinary communicative
9
Elena Kagan, The Scalia Lecture: A Dialogue with Justice Kagan on the Reading
of Statutes at 8:28 (Nov. 17, 2015), http://today.law.harvard.edu/in-scali-lecture-kagandiscusses-statutory-interpretation
[http://perma.cc/3BCF-FEFR].
10
See JOHN F. MANNING & MATTHEW C. STEPHENSON, LEGISLATION AND
REGULATION 60 (2d ed. 2013) (―Over the last quarter-century, textualism has had an
extraordinary influence on how federal courts approach questions of statutory
interpretation. When the court finds the text to be clear in context, it now routinely
enforces the statute as written.‖); Abbe R. Gluck, The States as Laboratories of
Statutory Interpretation: Methodological Consensus and the New Modified Textualism,
119 YALE L.J. 1750, 1756–58 (2010) (concluding, based on a comprehensive study of
state court approaches to statutory interpretation, that such courts are engaged in an
―effort[] to increase predictability in statutory interpretation,‖ that they give primacy to
text and decline to look to external sources of meaning if they find the text ―plain,‖ and
asserting that ―these state efforts . . . respond directly to the leading academic proposals
advanced to make federal statutory interpretation more determinate‖).
11
See Fallon, supra note __, at 1255–63 (exploring a range of possible meanings of
communicative or ―conversational‖ meaning, including ―semantic‖ or ―literal‖ meaning,
―contextual‖ meaning embraced by ―shared presuppositions‖ of speakers and listeners,
―intended meaning‖ and others, and asserting that there accordingly is ―no single,
linguistic fact of the matter concerning what statutory or constitutional provisions
mean‖); Cass R. Sunstein, There Is Nothing that Interpretation Just Is, 30 CONST.
COMMENT. 193, 194–95 (2015) (identifying possible notions of meaning, including
authorial intention, public meaning, moral reading, and others).
12
See Fallon, supra note ___, at 1272 (noting that ―there can be a multitude of
linguistically pertinent facts, generating different senses of meaning, which in turn
support a variety of claims‖); id. at 1269 (asserting that ―[u]ncertainty and division‖ in
measuring ordinary meaning are ―inevitable,‖ that evidence of ―communicative or
assertive content, understood as a matter of linguistic fact, is often sparse, minimal, or
indeterminate as applied to particular cases,‖ and that ―we cannot proceed by taking or
imagining the outcome of an opinion poll‖ about ordinary meaning).
13
Id. at 1269.
4 Judging Ordinary Meaning 3-23-17
content of the law in favor of ―case-by-case‖ ―interpretive eclecticism‖—
of choosing the ―best interpretive outcome as measured against the
normative desiderata of substantive desirability, consistency with rule of
law principles, and promotion of political democracy, all things
considered.‖14
Nor do we find in the indeterminacy of the search for
ordinary meaning a broad license for ―normative judgments‖ about
whatever ―interpretation‖ ―makes our constitutional system better rather
than worse.‖15
This kind of ―interpretation‖ overrides rather than
protecting the values served by the ordinary meaning rule. It undermines
reliance and fair notice interests and gives voice to the will of judges, not
lawmakers.
Granted, ―we can‘t treat the meaning of [the law‘s] language as the
only source of its legal effect.‖16
Our law of interpretation may have good
reasons to depart from the ―standard picture‖—to substitute ―fake‖
answers to linguistic questions for real ones.17
It is doubtless true,
moreover, that some of our rules of interpretation dictate a ―process‖ that
―often looks nothing like a search for linguistic meaning.‖18
But that is no
reason to abandon the enterprise entirely. The better response—or at least
the first step in the response—is to examine the ―standard picture‖ more
carefully. To do so we should seek a more careful conception of ordinary
meaning and see if we can find better ways to measure it.
We may eventually throw up our hands and conclude that some
questions of ordinary meaning have no good answers. Or we may
conclude that the law has good reason to substitute a non-linguistic
answer that vindicates policies more important than the ones advanced by
the ―standard picture.‖ But it skips the key threshold question to assume
that the ―law of interpretation‖ just is,19
or that its ―fake‖ (non-linguistic)
premises are sufficient to override the search for ordinary meaning.20
The
search for ordinary meaning is hard. But the premises of this inquiry are
too deeply embedded in our law and too clearly rooted in important
policy considerations to give up at the first sight of difficulty or
indeterminacy, or to judge the enterprise on the fuzzy premises or
mistaken methodologies of the past. So we take up the inquiry here.
14
Id. at 1305.
15
Sunstein, supra note __, at 193–94.
16
Baude & Sachs, supra note __, at 1088.
17
Id. at 1096.
18
Id. at 1088.
19
Cf. Sunstein, supra note __, at 193 (asserting that there is nothing about
interpretation that ―just is‖).
20
In other words maybe the ―standard picture‖ doesn‘t ―claim[] to be a picture of
American law.‖ Baude & Sachs, supra note __, at 1089. Maybe American law claims to
be a reflection of the standard picture—of ordinary meaning. We should entertain that
possibility. Or at least we should ask whether there are good reasons for flipping the
picture.
5 Judging Ordinary Meaning 3-23-17
Our thesis is that words have meaning—and that meaning can be
theorized and measured using principles and methods devised in the field
of linguistics. When we speak of ordinary meaning we are asking an
empirical question—about the sense of a word of phrase that is most
likely implicated in a given linguistic context.21
Linguists have developed
computer-aided means of answering such questions. We propose to
import those methods into the law of interpretation. And we identify
problems in the methods the law has been using to address these issues.
We begin by noting the circumstances in which the ―standard picture‖
controls under the law of interpretation, highlighting three exemplary
cases in which the ordinary communicative content of the words of a
statute seem to dictate the court‘s holding. Next we identify shortcomings
in the law‘s attempt to give effect to that communicative content—
shortcomings in both the theory of ordinary meaning and in our attempts
to operationalize (or measure) it. After outlining these two sets of
problems we introduce some theories and empirical methods utilized by
linguists that may help us do a better job of delivering on the promise of
an objective inquiry into ordinary meaning.22
We then apply these
tools—imported from a field called corpus linguistics—to our three
exemplary cases. And we close by responding to actual and anticipated
criticisms of our approach and by highlighting unresolved issues that
must be addressed going forward.
21
Judge Richard Posner framed the ordinary meaning question in this (empirical)
way in his opinion in United States v. Costello, 666 F.3d. 1040, 1042 (7th Cir. 2012).
There he proposed to answer this question using the results of a Google search. We
think Posner‘s instincts were right but his methods fell a bit short, as explained below.
See infra __.
22
Some judges (present company included) are beginning to take note of the
deficiencies we highlight here and to try to address them. In a few recent cases judges
have made a studied effort to define the inquiry into ordinary meaning more precisely.
And, importantly, they have presented empirical analysis in support of their conclusions.
See e.g., State v. Rasabout, 2015 UT 72, ¶¶ __ - __; 356 P.3d. 1258 (Lee, J., concurring
in the judgment and advancing corpus linguistic data in support of his interpretation of
the phrase ―discharge a firearm‖ in a state statute); State v. Canton, 2013 UT 44, ¶ __, n.
__, 308 P.3d. 517 (Lee, J., for the majority; presenting corpus linguistic data in support
of the court‘s construction of the phrase ―out of the state‖ in a tolling provision for
criminal statutes of limitation under Utah law); Baby E.Z. v. T.I.Z., 2011 UT 38, ¶ __ __,
266 P.3d. 702 (Lee, J., concurring in part and concurring in the judgment;
advocating the use of corpus linguistic data in support of his interpretation of ―custody‖
proceeding under the federal Parental Kidnapping Protection Act, 28 U.S.C. § 1738A
(2006)); People v. Harris, 885 N.W.2d 832, 838–39 n.29 (Mich. 2016) (opinion of the
court per Zahra, J.) (citing Utah Supreme Court opinions in support of the methodology
of corpus linguistics; relying on corpus linguistic data in support of the court‘s
interpretation of the term ―information‖ in Michigan statute forbidding use of “
information” provided by law enforcement officer if compelled under threat of
employment sanction); id. at 850 n.14 (Markman, J., dissenting) (also citing Utah
Supreme Court opinions and also relying on corpus linguistic data, but drawing a
different inference from the data).
6 Judging Ordinary Meaning 3-23-17
I. ORDINARY MEANING IN THE LAW OF INTERPRETATION
Everyone agrees that our sense of the ordinary communicative content
of legal language is an important starting point for interpretation. All
agree, moreover, that the law should credit that content at least
sometimes. This holds even for those who doubt our ability to settle on a
single notion of meaning, or to assess it with any degree of consistency.23
Judges generally are even more sanguine about the matter. The
caselaw in this field is marked by numerous references to the ―standard
picture.‖ Judges routinely advert to the idea of crediting the ―ordinary
meaning‖ of statutory text.24
Where such meaning is viewed as ―plain,‖
moreover, the courts consistently declare the interpretive enterprise to be
at an end.25
The general rule is to credit the communicative content of
statutory text where it is ―plain,‖ and in that event to close the door to the
consideration of extra-textual sources of meaning or intent.26
A variation on the theme applies in the realm of substantive canons of
construction or principles of agency deference. The rule of lenity, for
example, says that genuine ambiguities in criminal laws are resolved in
favor of the defendant27
; the converse is the notion that ―the rule of lenity
23
See Fallon, supra note 2, at __; and Baude & Sachs, supra note __.
24
See, e.g., Clark v. Rameker, 134 S. Ct. 2242, 2246 (2014) (―. . . we give the term
its ordinary meaning.‖); Bond v. United States, 134 S. Ct. 2077, 2091 (2014) (―In
settling on a fair reading of a statute, it is not unusual to consider the ordinary meaning
of a defined term . . .‖); Mohamad v. Palestinian Auth., 132 S. Ct. 1702, 1706 (2012)
(―Because the TVPA does not define the term ―individual,‖ we look first to the word's
ordinary meaning.‖); Mac's Shell Serv., Inc. v. Shell Oil Products Co. LLC, 559 U.S.
175, 175–76 (2010) (―Because [the Act] does not further define [―termination‖ and
―cancel‖], they are given their ordinary meanings . . .‖).
25
See William Baude & Ryan Doerfler, The (Not So) Plain Meaning Rule, 84 U.
CHI. L. REV (forthcoming 2017) (characterizing the ―plain meaning rule‖ as a
―compromise‖ in which ―other information can‘t be considered ―[if] the statute‘s
meaning is plain,‖ but in which other information ―comes in‖ ―[i]f it‘s not plain‖).
26
See, e.g., Gluck, supra note 12, at 1758 (stating that the ―modified textualism‖
approach embraced in most state courts ―ranks interpretive tools in a clear order—
textual analysis, then legislative history, then default judicial presumptions—and it
includes legislative history in the analysis‖); KENT GREENAWALT, STATUTORY
INTERPRETATION: 20 QUESTIONS 35 (1999) (―No one seriously doubts that interpretation
of statutes turns largely on textual meaning‖).
27
How much ambiguity, of course, is a difficult question. See Abramski v. United
States, 134 S. Ct. 2259, 2272 (2014) (asserting that the rule applies only if the ―there
remains a grievous ambiguity or uncertainty in the statute‖ that cannot be resolved—if
the court is left to ―simply guess as to what Congress intended‖); id. at 2281 (Scalia, J.,
dissenting) (suggesting that the rule should apply if ―after all legitimate tools of
interpretation‖ have been employed ―a reasonable doubt persists); ANTONIN SCALIA &
BRYAN GARNER, READING LAW: THE INTERPRETATION OF LEGAL TEXTS 299 (2012)
(decrying the ―multiplicity of expressed standards‖ for invoking the rule of lenity,
7 Judging Ordinary Meaning 3-23-17
has no application when the statute is clear.‖28
And Chevron deference is
similar. The courts defer to agencies only where the terms of the statute
are ambiguous.29
So both of these principles likewise implicate the
question of the degree of plainness of statutory language.
This is not an exhaustive list of the courts‘ consideration of the
ordinary communicative content of statutory language. But it is enough to
highlight the point that the law at least sometimes credits that content in
the enterprise of legal interpretation.
That said, commentators are undoubtedly right to question the
determinacy of the inquiry into ordinary meaning. The problem, as noted,
is twofold—going both to the law‘s conception of ordinary meaning and
to our judges‘ attempts to measure it. First is a problem of theory:
Ironically, we have no ordinary meaning of ―ordinary meaning.‖30
The
same goes even for ―plain‖ meaning. ―Courts and commentators
sometimes use the phrase ‗plain meaning‘ to denote something like
ordinary meaning,‖ or in other words ―the meaning one would normally
attribute to [the] words‖ of a statute ―given only very limited information
about the context of the utterance.‖31
―Other times ‗plain meaning‘ is
used to denote obvious meaning—i.e., the meaning that is clear.‖32
(This
is the sense at work in the ―plain meaning rule.‖33
)
Second is a problem of operationalization or measurement. The
concern here is that even if we could settle on a sense of ordinary or plain
meaning we are unsure how to assess it. ―Uncertainty and division seem
inevitable.‖34
That is true because the question of intended or understood
meaning is an empirical one, and judges cannot ―proceed by taking or
imagining the outcome of an opinion poll‖ as to intended or perceived
meaning.35
And the problem is underscored by the tools (mis)used by
―leav[ing] open the crucial question . . . of how much ambiguousness constitutes an
ambiguity‖).
28
See SCALIA & GARNER, supra note __, at 301.
29
See Michigan v. E.P.A., 135 S. Ct. 2699, 2707 (2015).
30
WILLIAM N. ESKRIDGE, JR., PHILIP P. FRICKEY & ELIZABETH GARRETT, CASES
AND MATERIALS ON LEGISLATION: STATUTES AND THE CREATION OF PUBLIC POLICY
792-93 (4th ed. 2007) (noting the irony that ―‗plain meaning‘ is . . . a deeply ambiguous
term‖; highlighting differences in the ways courts use the terms ―plain meaning,‖ often
to refer to a sense that is ―quite clear in a literal sense,‖ and ―ordinary meaning,‖ which
may mean ―the best (most coherent) textual understanding that emerges after close
textual analysis‖); Richard A. Posner, Statutory Interpretation—In the Classroom and
the Courtroom, 50 U. CHI. L. REV. 800, 808 (1983) (observing, as to the ―‗start with the
words‘ canon,‖ that ―[i]t is ironic that a principle designed to clarify should be so
ambiguous‖).
31
Baude & Doerfler, supra note __, at 5.
32
Id.
33
See id.
34
Fallon, supra note __, at 1268.
35
Id.
8 Judging Ordinary Meaning 3-23-17
judges to try to answer this empirical question (resort to a dictionary or a
word‘s etymology, for example, as explored further below).
The theoretical and measurement problems with the ordinary meaning
inquiry are even bigger than most have acknowledged. The depth of the
problem is best illustrated by reference to concrete examples in the case
law. Throughout this article we consider these:
∙ Is a person guilty of carrying a firearm (under a federal
sentencing enhancement provision) in connection with a drug crime if
he merely transports it to a drug deal in a locked glove compartment of
the car he is driving? This was the question presented in United States
v. Muscarello, 524 U.S. 125 (1998). The Muscarello court was sharply
divided. All nine justices agreed that the question came down to the
―ordinary meaning‖ of the notion of carrying a firearm. Yet they
divided 5-4 on whether the ordinary sense of that phrase encompassed
the conveyance of a gun in a glove compartment. And each side
proffered varying senses of the meaning of ―ordinary meaning,‖ and
claimed support for their view in sources ill-suited to providing a
reliable answer to the empirical question presented—looking to
dictionaries, to isolated examples of language from literature, and even
to etymology of the verb carry.
∙ Is a litigation expert who is paid to translate written documents
from one language to another an interpreter under a statute authorizing
an award of costs for prevailing parties who utilize them in litigation?
This question arose in Taniguchi v. Kan Pacific Saipan, Ltd., __ U.S.
__, 132 S. Ct. 1997 (2012). Again the court agreed that the case came
down to a matter of ordinary meaning—of the term interpreter. Yet
again the court was divided, this time 6-3. In Taniguchi the majority
and dissent both agreed that the more common sense of interpreter
referred to a person engaged in simultaneous oral translation. But
again they resorted only to dictionaries and similar sources for their
conclusions. And they also disagreed about what the search for
ordinary meaning ultimately entailed, with the majority insisting that
only the more common sense of the term was covered, and the dissent
asserting that a permissible sense should also count.
∙ Is a woman who allows her ―illegal alien‖ boyfriend to sleep at
her apartment guilty of harboring an alien under a federal statute
criminalizing that act? This question arose in United States v. Costello,
666 F.3d. 1040. 1042 (7th Cir. 2012). Costello is like Muscarello and
Taniguchi: it involves a statutory term broad enough to encompass
both parties‘ positions. Sometimes harbor refers to the mere act of
providing shelter, but it may also indicate the sort of sheltering that is
aimed at concealment. How is the court to decide which sense is the
ordinary one? Judge Richard Posner authored the opinion for the court.
And he recognized the deficiencies of standard methods—principally,
dictionaries—in answering that question. So he proceeded to a search
9 Judging Ordinary Meaning 3-23-17
for data. And he did so using the search engine that is perhaps most
familiar to us today. He performed a Google search.
Below we use these cases to highlight the theoretical and operational
deficiencies in the law‘s search for ordinary meaning.
A. Theoretical Shortcomings
The case law embraces a startlingly broad range of senses of ordinary
meaning. When judges speak of ordinary meaning they often seem to be
speaking to a question of relative frequency—as in a point on the
following spectrum:
POSSIBLE COMMON MOST FREQUENT EXCLUSIVE
At the left end of the spectrum is the idea of a possible or linguistically
permissible meaning—a sense of a word or phrase that is attested in a
known body of written or spoken language. A meaning is a possible one
if we can say that ―you can use that word in that way‖ (as attested by
evidence that other people have used the word in that way in the past).
Yet a possible meaning may be an uncommon or unnatural sense of a
given term; so we might note that a given sense of a term is not common
or usual in a given linguistic setting even if it is possible to speak that
way. And even a common sense of a term might not be the most frequent
use of it in a certain context.
The notion of plain meaning adds one more point to the continuum.
When courts speak of plain meaning (as a construct distinct from
ordinary meaning) they generally mean to ―denote obvious meaning‖ or
―meaning that is clear.‖36
A plain—obvious or clear—meaning would be
more than most frequent. It would be nearly exclusive.
The four points on the continuum can be illustrated by a range of
senses of the term vehicle in the hypothetical ―no vehicles in the park‖
provision.37
One attested sense of vehicle is the notion of a ―carrier,‖
―agent of transmission.‖38
And that sense could sweep broadly. If we are
thinking of the carrier sense of vehicle, the ―no vehicles in the park‖
prohibition could possibly be viewed as covering a pet dog or cat, which
could be referred to as a vehicle (as in a pet as a carrier of infection). Yet
that sense of vehicle would not be viewed as a natural or common one in
this linguistic setting. If the ―no vehicles‖ ordinance is aimed at only
common senses of vehicle, we likely would not deem it to prohibit pets.
Alternatively, we could say that the meaning of vehicle in this setting is
36
Baude & Doerfler, supra note __, at 5.
37
H.L.A. Hart, Positivism and the Separation of Law and Morals, 71 HARV. L. REV.
593, 606–15 (1958).
38
WEBSTER‘S THIRD NEW INTERNATIONAL DICTIONARY 2538 (1961).
10 Judging Ordinary Meaning 3-23-17
plain or clear, or in other words that the exclusive sense of vehicle is one
that eliminates the possibility of its extension to pets.
Is a bicycle a vehicle covered by the ordinance? Perhaps so—as
encompassed by the sense of vehicle as ―a means of carrying or
transporting something; conveyance.‖39
This sense of vehicle could easily
be viewed as a common sense of vehicle—certainly more common than
the infection carrier sense noted above. But even this sense may not be
the most common—the statistically most frequent sense of vehicle in this
linguistic setting (an outdoor public park). If we are looking for the most
frequent sense of vehicle in this context,40
we might understand the term
to encompass only motor vehicles, and thus not to cover the bicycle.
(Here we are making some suppositions on the points on the
continuum—on which senses of vehicle are possible, common, and most
frequent. We do so to illustrate the range of senses of ordinary meaning.
We will move from supposition to empirical analysis of these questions
later.41
)
The four points on the frequency continuum do not completely capture
the range of senses of ordinary meaning embraced by our courts.
Sometimes our judges seem to have reference to a fifth notion of
ordinary—a notion of linguistic prototype.42
A prototype is a sense, or
example of a sense, that is viewed as most strongly associated with a
given term in a given context. And that may jibe with the way we
separate senses or definitions in our minds. A difference in word meaning
may ―be represented in cognition not as a set of criterial features with
clear-cut boundaries‖ the way a dictionary would represent things, but
instead ―in terms of prototype (the clearest cases, best examples) of the
category.‖43
Thus, prototype analysis has shown that people consider
chair to be a more prototypical example of furniture than stool,44
39
Id.
40
This notion of ―ordinariness‖ is attested in the Oxford English Dictionary. See 10
OXFORD ENGLISH DICTIONARY 912 (2d ed. 1989) (defining ordinary as ―Of language,
usage, discourse, etc.: that most commonly found or attested . . . .‖); see also Lawrence
M. Solan & Tammy A. Gales, Finding Ordinary Meaning in Law: The Judge, the
Dictionary or the Corpus? 1 INT‘L J. OF L. DISCOURSE 253, 263 (2016) (―‗Ordinary
meaning,‘ especially as applied to particular words and phrases, is a distributional fact.
A usage is ‗ordinary‘ when it predominates.‖).
41
See infra Part III.C.
42
See McBoyle v. United States, 238 U.S. 25, 27 (1931) (per Holmes, J.)
(determining whether an ―airplane‖ was a ―vehicle‖ for the purposes National Motor
Vehicle Theft Act of 1919, and stating: ―When a rule of conduct is laid down in words
that evoke in the common mind only the picture of vehicles moving on land, the statute
should not be extended to aircraft . . . .‖ (emphasis added)). This notion of a ―picture‖
―evoke[d] in the common mind‖ maps very well onto the concept of prototype.
43
Eleanor Rosch, Cognitive Representation of Semantic Categories, 104 J.
EXPERIMENTAL PSYCHOL. 192, 193 (1975).
44
Id. at 229.
11 Judging Ordinary Meaning 3-23-17
automobile to be a more prototypical vehicle than yacht,45
and robin to be
a more prototypical bird than ostrich.46
And that is another way to conceive of the notion of ordinary meaning
in the law.47
A judge who approaches the question of ordinary meaning
by attempting to determine the most prototypical example of a given
sense of a term is searching for a linguistic prototype.48
Under this
approach, the ordinary (prototype) sense of vehicle would be the one that
is most ―vehicle-like,‖ perhaps encompassing a passenger vehicle with
four wheels and an engine. If that is our sense of the ordinary meaning of
vehicle, we might conclude that the hypothetical ordinance prohibits cars
and trucks but not motorized scooters.49
This range of meaning of ordinary meaning can also be illustrated
through our three feature cases. We turn to them here.
1. Muscarello v. United States
In Muscarello the court was asked to interpret a statute calling for a
five-year mandatory prison term for a person who ―uses or carries a
firearm‖ ―during and in relation to‖ a ―drug trafficking crime.‖50
Frank
Muscarello was convicted on drug charges in a case in which he was
shown to have transported a gun to the drug deal in the locked glove
compartment of his car. The question presented was whether that counted
as carrying. Both the majority (Breyer) and dissenting (Ginsburg)
opinions agreed that the proper interpretation of ―carries a firearm‖ came
45
Id. at 230.
46
Id. at 232.
47
See Lawrence M. Solan, Why Laws Work Pretty Well, but Not Great: Words and
Rules in Legal Interpretation, 26 L. & SOC. INQUIRY 243, 258 (2001) (citing Smith v.
United States, 508 U.S. 223 (1993) (―In the realm of statutory interpretation, judges
often evoke the canon that they are to give words in a statute their ‗ordinary‘ meaning.
Prototype analysis tells us that the notion of ordinary meaning has a cognitive basis.‖);
see also ESKRIDGE ET AL., supra note 27, at 850 (discussing prototypical meaning in the
context of statutory interpretation); Lawrence Solan, Terri Rosenblatt & Daniel
Osherson, False Consensus Bias in Contract Interpretation, 108 COLUM. L. REV. 1268,
1276–80 (2008); Lawrence M. Solan, The New Textualists‟ New Text, 38 LOY. L.A. L.
REV. 2027, 2042–46 (2005); Note, The Supreme Court 1997 Term, Leading Cases, 112
HARV. L. REV. 355, 361–62 (1998) (―[W]hen a legislature uses non-technical terms . . .
it is likely that both the legislature and the general public interpret the term in
accordance with its prototypical meaning.‖).
48
Lawrence M. Solan, Law, Language, and Lenity, 40 WM. & MARY L. REV. 57,
67–68 (1998).
49
This conclusion, however, cannot be derived with mere intuition. The discovery of
a prototype for a given word in a given context requires the application of empirical
methods, as we will discuss below.
50
18 U.S.C. § 924(c)(1).
12 Judging Ordinary Meaning 3-23-17
down to the ―ordinary English meaning‖ of that ―phrase.‖51
Yet neither
opinion settled on a single sense of ―ordinary.‖ Both opinions slide back
and forth along the continuum, without acknowledging that they are
doing so.
At one point Breyer seems to employ a merely ―common‖ sense of
ordinary; he does so in asserting that the transport in a vehicle sense of
carry is ordinary given that ―many‖—―perhaps more than one-third‖—of
the instances of carrying a firearm in a New York Times database reflect
that sense,52
and in concluding that ―the word ‗carry‘ in its ordinary sense
includes carrying in a car.‖53
Yet Elsewhere Breyer seems to speak of the
car-carrying sense as most frequent. He does so in asserting (a) that the
―ordinary English‖ sense of carry is to transport it in a vehicle; (b) the
bear personally sense is ―special‖; and (c) ―we believe Congress intended
to use the word in its primary sense and not in this latter, special way.‖54
Justice Ginsgurg‘s dissent is also inconsistent. In asserting that the
personally bearing sense is ordinary Justice Ginsburg asserts that it is
―hardly implausible, nor at odds with an accepted meaning‖ of the
statutory terms.55
That is the language of possibility or commonality.56
Elsewhere, however, Justice Ginsburg seems to speak in terms of
personally bearing as the most frequent sense of the term—in noting, in
response to Justice Breyer‘s statistics, ―what meaning showed up some
two-thirds of the time.‖57
2. Taniguchi v. Kan Pacific Saipan
Taniguchi was a personal injury case.58
The plaintiff was a Japanese
baseball player suing for medical expenses and lost income from
contracts he was unable to honor as a result of injuries at the defendant‘s
resort.59
The defense ―paid to have various documents translated from
Japanese to English.‖60
And when the district court dismissed
Taniguchi‘s case on summary judgment, the defense submitted a request
for compensation for the amounts it paid for document translation. As in
Muscarello the Taniguchi case came down to ordinary meaning. Here the
51
Muscarello v. United States, 524 U.S. 125 at 127–28 (1998).
52
Id. at 129.
53
Id. at 131 (emphasis added).
54
Muscarello, 524 U.S. at 128 (emphasis added).
55
Id. at 148 (Ginsburg, J., dissenting) (emphasis added).
56
See id. at 143 – 44 (asserting ―that ‗carry‘ is a word commonly used to convey
various messages,‖ and that it ―could mean‖ either personally bear or transport in a
vehicle).
57
Id. at 143.
58
Taniguchi v. Kan Pac. Saipan, 132 S.Ct. 1997 at 2000 (2012).
59
Id.
60
Id.
13 Judging Ordinary Meaning 3-23-17
operative language was from a statute allowing the prevailing party in
federal litigation to recover certain costs, including those incurred by an
―interpreter.‖61
And here the case seemed to turn on the operative notion of ordinary
meaning. Justice Alito wrote for the majority and concluded that the
ordinary sense of interpreter is oral translator. Alito finds written
translator possible but ―hardly a common or ordinary meaning.‖62
He
says that ―an interpreter is normally understood as one who translates
orally from one language to another.‖63
And he concludes that the written
translator sense is not ―truly common,‖ but is rather ―obsolete,‖ citing
dictionaries to support that conclusion.64
Justice Ginsburg‘s dissent acknowledges that interpreter ―commonly
refers to translators of oral speech‖ but concludes that the term ―more
than occasionally‖ is ―used to encompass those who translate written
speech as well.‖65
This is a core basis of the view of the Taniguchi
dissenters. They do not expressly disagree with Alito‘s assertion that the
oral translator notion is most common; they are simply saying that either
of two common senses of a term should count as ordinary.66
3. United States v. Costello
The defendant in Costello was charged with knowingly ―conceal[ing],
harbor[ing], or shield[ing] from detection‖ an ―alien in any place,
including any building or any means of transportation.‖67
Her alleged
crime was essentially in ―having permitted‖ her ―illegal alien‖ ―boyfriend
to live with her.‖68
And the principal question presented was whether the
ordinary meaning of the verb harbor required proof of concealment.
As in Taniguchi, the difference between the majority and dissent in
Costello seems to come down largely to the conception of the meaning of
ordinary meaning. Judge Posner, for the majority, warns of the perils of
overreliance on the dictionary to resolve questions of ordinary meaning.
(More on that below.) And he directs the ordinary meaning analysis to an
empirical inquiry—which he proposes to resolve by means of a Google
search.
61
Id. at 2003.
62
Id.
63
Id. at 2004.
64
Id.
65
Taniguchi, 132 at 2007 (Ginsburg, J., dissenting).
66
See id. at 2008 (asserting that the written translator sense is an ―acceptable usage‖
even if it is ―not ‗the most common usage‘‖).
67
8 U.S.C. § 1324(a)(1)(B)(ii).
68
United States v. Costello, 666 F.3d 1040, 1043 (2012).
14 Judging Ordinary Meaning 3-23-17
Posner‘s reliance on his Google results places his sense of ordinary
meaning on the frequency continuum. He uses Google to look for relative
numbers of ―hits‖ for phrases like ―harboring fugitives‖ and ―harboring
guests.‖ And because Posner found more hits for the former than for the
latter, he concludes that ―‗harboring,‘ as the word is actually used, has a
connotation . . . of deliberately safeguarding members of a specified
group from the authorities.‖69
This is a ―most frequent‖ sense of
ordinariness—and a blatantly empirical sense of that inquiry.
The Costello dissent takes a different tack. In concluding that the
providing shelter notion of harbor falls within the statute, Judge Manion
asserts that ―the ordinary meaning of ‗harboring‘ certainly includes
‗providing shelter to.‘‖70
In support of this point Manion cites definitions
from dictionaries in print at the time of the statute‘s enactment. He says
that these dictionaries show that ―[t]his was a common understanding of
the term when the term ‗harbor‘ was first added to the statute in 1917,
and when the statute was amended and the term retained in 1952.‖71
* * *
Our judges purport to be speaking of a consistent, common sense of
ordinary meaning. But we switch back and forth between different senses
of ordinary meaning, usually without acknowledging the inconsistency.
Sometimes (as in Muscarello) our judges embrace varying senses of
ordinary meaning within a single opinion. Elsewhere (as in Taniguchi
and Costello) the seemingly nuanced distinction between different senses
of ordinary meaning becomes outcome-determinative—with the majority
selecting one point on the frequency continuum that leads to one result
and the dissent opting for a different point that yields the opposite
outcome. This is problematic—not just for statutory interpretation, but
for the rule of law.72
B. Operational Shortcomings
The theoretical deficiencies identified above are one element of the
problem. Another is operational—in the way we seek to identify or
69
Id.
70
Id. at 1052 (Manion, J., dissenting) (emphasis added).
71
Id. (citing WEBSTER‘S NEW INTERNATIONAL DICTIONARY OF THE ENGLISH
LANGUAGE 981 (1917) (with a definition of harbor of ―[t]o afford lodging to; to
entertain as a guest; to shelter; to receive; to give refuge to‖)); WEBSTER‘S NEW
COLLEGIATE DICTIONARY 376 (John P. Bethel et al. eds., 1953) (including a definition
of harbor as ―to entertain as a guest; to shelter; to give a refuge to‖).
72
See William Baude & Stephen E. Sachs, The Law of Interpretation, 130 HARV. L.
REV. 1079, 1089 (2017) (―[W]e have to decide which meaning, produced by which
theory of meaning, we ought to pick.‖).
15 Judging Ordinary Meaning 3-23-17
measure the ordinary meaning of statutory terms. Typically this
assessment is made at a gut level, on the basis of a judge‘s linguistic
intuition, without recognition of the empirical nature of the question.
A judge considering the prohibition on vehicles in the park, for
example, would reject out of hand the notion that the ordinance extends
to pets, insisting (without further analysis or support) that the infection
carrier sense of vehicle is an outlier—an extraordinary meaning. A
parallel conclusion would be likely in response to an attempt to extend
the no vehicles ordinance to bicycles. We understand vehicle to
encompass a conveyance on wheels, but again a court seems likely to
jump to the conclusion that the ordinary sense of vehicle is motor vehicle,
and that a bicycle doesn‘t count.
These conclusions seem uncontroversial. But the judge who makes
them is making an empirical assessment. Gut-level empirics probably
won‘t bother us if they go only to a holding that a pet or a bicycle is not a
vehicle prohibited in the park. But what about a motorized scooter or a
golf cart? Are they covered? These are harder questions. And here we
may have more cause for concern about the lack of transparency and
determinacy.
With this in mind, judges sometimes turn to other grounds for their
assessment of ordinary meaning. Two primary grounds are dictionaries
and etymology. A common use of a dictionary involves simple cherrypicking.
―Instead of acknowledging and rejecting contrary senses of a
statutory term, judges tend to ignore them—identifying only the sense of
a word they deem ordinary without acknowledging any others.‖73
As to
vehicle, for example, a judge might simply cite a definition referring to an
automobile and assert, without more, that the term‘s ordinary meaning
does not encompass a motor scooter, or maybe even a golf cart. That is of
course troubling; a judge who cherry-picks a preferred dictionary
definition while ignoring an alternative is clearly misusing the dictionary.
Some judges, to their credit, are more transparent. Instead of ignoring
a contrary definition—the conveyance on wheels notion of vehicle, for
example—a judge may acknowledge both the automobile and
conveyance notions but find a basis for embracing one as ordinary. We
see two principal means of doing so in the cases: (a) preferring the
definition that appears first in a dictionary‘s list of senses, and (b) citing
73
State v. Rasabout, 2015 UT 72, ¶ 53, 356 P.3d 1258 (Lee, J., concurring in part
and concurring in the judgment). See also e.g., Kovach v. Zurich Am. Ins. Co., 587 F.3d
323, 346 (6th Cir. 2009) (McKeague, J., dissenting) (criticizing the majority for ignoring
other definitions in basing its presentation of ―ordinary meaning‖ of ―accidental‖ on one
definition without regard to others); Konop v. Hawaiian Airlines, Inc., 302 F.3d 868,
878 (9th Cir. 2002) (ignoring broader definitions in favor of a narrow definition as
―ordinary meaning‖ of ―intercept‖); United States v. Warner Bros. Well Drilling, 899
F.2d 15, 15 (6th Cir. 1990) (citing only one definition of ―operator‖ in determining the
ordinary meaning even though opposing definitions existed).
16 Judging Ordinary Meaning 3-23-17
the etymology of the statutory term. Neither of these approaches is
defensible, however, for reasons explained below in our critique of the
court‘s ordinary meaning analysis in Muscarello and Taniguchi. Costello,
on the other hand, acknowledges some of the problems we identify here
and turns to Google, albeit in a manner raising a new set of problems.
1. Muscarello v. United States
The Muscarello majority invokes both sense-ranking and etymology
in support of its holding. Justice Breyer acknowledges that carry can be
understood to mean either transport in a vehicle or bear on your person.74
But he embraces the former sense as the ―primary‖ one and dismissed the
latter as ―special.‖75
And his first argument in support of that conclusion
is that ―[t]he Oxford English Dictionary gives as its first definition
‗convey, originally by car or wagon, hence in any vehicle, by ship, on
horseback, etc.‘‖76
The italicized emphasis on ―first‖ is Breyer‘s.
Breyer‘s opinion takes a similar tack in citing the ―first definition‖ in
Webster‘s Third (―‗move while supporting (as in a vehicle or in one‘s
hands or arms)‖77
) and the ―first definition‖ in the Random House
Dictionary (―‗to take or support from one place to another; convey;
transport.‘‖78
Justice Breyer reinforces his reliance on sense-ranking in his reference
to the personally bear sense of carry in the Oxford English Dictionary—
noting that this is ―twenty-sixth definition‖ in the OED.79
This seems
clearly to be the threshold basis of Breyer‘s conclusion that ―[t]he
relevant linguistic fact[]‖ is that ―the word ‗carry‘ in its ordinary sense
includes carrying in a car.‖80
Breyer also turns to etymology, in asserting that ―[t]he origin of the
word ‗carries‘ explained why the first, or basic, meaning of the word
‗carry‘ includes conveyance in a vehicle.‖81
Breyer states that carry
traces from ―Latin ‗carum,‘ which means ‗car‘ or ‗cart,‘‖ and from ―Old
French ‗carier‘ and late Latin ‗carricare,‘ which meant to ‗convey in a
car.‖82
The precise premises of Breyer‘s analysis are left implicit. But the
implicit point seems clear—the etymology of the verb carry confirms that
74
524 U.S. at 128.
75
Id.
76
Id. (citing 2 OXFORD ENGLISH DICTIONARY 919 (2d ed. 1989)).
77
Id. (citing WEBSTER‘S THIRD NEW INTERNATIONAL DICTIONARY 343 (1986)).
78
Id. (citing RANDOM HOUSE DICTIONARY OF THE ENGLISH LANGUAGE
UNABRIDGED 319 (2d ed. 1987)).
79
524 U.S. at 130 (quoting 2 OXFORD ENGLISH DICTIONARY at 921).
80
Id.
81
Id. at 128 (citing BARNHART DICTIONARY OF ETYMOLOGY 146 (1988)).
82
Id.
17 Judging Ordinary Meaning 3-23-17
the transport sense of the term is ordinary and the personally bear sense
is unusual.
This is problematic. If the ordinary meaning question in Muscarello is
an empirical question going to frequency or prototype analysis, neither
the dictionary nor etymology is useful. The dictionaries typically cited by
our courts (including those cited by Justice Breyer) make no claims about
relative frequency of listed senses of a given word.83
Many ―dictionaries
simply rank their definitions according to evidence of historical usage.‖84
Others openly ―disavow[] any attempt to establish a hierarchy of
ordinariness in the ranking of . . . senses, admitting that sometimes an
‗arbitrary‘ listing of senses is used.‖85
Thus, historical ordering is
unhelpful in the assessment of ordinariness. If anything, an older sense
may be less likely to be the more common one today.
That is also the problem with etymology. If our usage and
understanding of a word have evolved over time, as they often will, the
historical pedigree of a word may direct us to an outmoded or even
obsolete definition—to the notion that December is the tenth month, or
that an anthology is a bouquet of flowers.86
That is entirely possible as to
83
See Stephen C. Mouritsen, The Dictionary Is Not a Fortress: Definitional
Fallacies and the Corpus-Based Approach to Plain Meaning, 2010 BYU L. REV. 1915,
1924-25 (discussing problems with dictionary usage by courts; identifying the ―senseranking
fallacy‖). The Random House Dictionary of the English Language, Unabridged
appears to be an exception. Its front matter states that ―the most frequently encountered
meaning generally comes before less common ones. Specialized senses follow those in
the common vocabulary, and rare, archaic, and obsolete senses are listed last.‖ RANDOM
HOUSE DICTIONARY OF THE ENGLISH LANGUAGE UNABRIDGED xxii (2d ed. 1987). But
we see grounds for skepticism of these sorts of claims. See infra ___. And Random
House acknowledges that its sense-ranking based on frequency holds only ―generally.‖
RANDOM HOUSE DICTIONARY OF THE ENGLISH LANGUAGE UNABRIDGED, supra, at xxii.
Without more (and there is no more in this dictionary), the reader is left to guess about
which senses are ordered according to frequency and which ones follow some other
organizing principle.
84
Id.; see also 1 OXFORD ENGLISH DICTIONARY xxix (2d ed. 1989) (―That sense is
placed first which was actually the earliest in the language: the others follow in order in
which they have arisen.‖); WEBSTER‘S THIRD NEW INTERNATIONAL DICTIONARY 17a
(1971) (indicating that the order of senses is ―historical,‖ in that ―the one known to have
been first used in English is entered first‖; also stating that its ―system of separating
senses‖ is ―only a lexical convenience,‖ and not an ―enduring hierarchy‖).
85
Id.
86
December, THE BARNHART CONCISE DICTIONARY OF ETYMOLOGY 188 (Robert K.
Barnhart ed,, 1995) (―1122, borrowed from Old French decembre, from Latin
December, from decem TEN, this being originally the tenth month of the early Roman
calendar (which began with March).‖); Anthology, THE BARNHART CONCISE
DICTIONARY OF ETYMOLOGY 29 (Robert K. Barnhart ed., 1995) (―1640, collection of the
‗flowers‘ of verse (i.e. small, choice poems) by various authors; borrowed, perhaps by
influence of French anthologie, from Greek anthologíā flower-gathering (ánthos flower
+ légein gather).‖).
18 Judging Ordinary Meaning 3-23-17
Breyer‘s analysis of carry; the etymology of the word tells us nothing
about its ordinary meaning today.
2. Taniguchi v. Kan Pacific Saipan
The Taniguchi opinion appears, at first glance, to employ dictionaries
in a less arbitrary way. Justice Alito does not turn to sense ranking or
etymology. He presents an informal ―survey‖ of dictionary definitions,
asserting that ―only a handful‖ of dictionaries include the written
translator sense of interpreter, but ―all‖ of them speak of oral
translator.87
And he says that the ―sense dividers‖ in the cited
dictionaries confirm the court‘s holding—in designating the oral
translator notion one that is ―especially‖ indicated and flagging the
written translator sense as ―obsolete.‖88
Yet Alito‘s approach is still problematic. The ―survey‖ of dictionaries
is far from systematic. Alito presents his own set of preferred
dictionaries. And within the cited dictionaries the court sometimes cites a
definition of the noun interpreter and sometimes cites a definition of the
verb interpret. We cannot tell from the opinion whether the written
translator sense of interpreter is less often listed in a real ―survey‖ of
dictionaries because we are not presented with an actual survey of
dictionaries. We have only the definitions that Justice Alito presented for
our review.89
Alito‘s sense dividers are also insufficient. First, not all dictionaries
designate written translator as obsolete or oral translator as special. At
least one definition mentioned in the majority opinion explicitly
encompasses the written sense of the term, without any indication of
obsolescence.90
Second, sense dividers are not reliable measures. Dictionaries tell us
very little about the basis for the ―obsolete‖ sense designation. And
ultimately, such a designation must be made on the basis of some
underlying data that is unavailable to the reader of the dictionary. So the
―obsolete‖ designation tells us only that the lexicographers who compiled
the dictionary in question deemed a particular sense to be a matter no
87
132 S.Ct. at 2003.
88
Id. at 2002–03; id. at 2003 (noting that the Oxford English Dictionary ―designated
[the written translator] meaning as obsolete‖); id. at 2003 & n.2 (noting that the Concise
Oxford Dictionary of Current English, Webster‟s Third New International Dictionary,
the World Book Dictionary, and Cassell‟s English Dictionary designate the oral
translator meaning as ―especially‖ indicated).
89
Even a documented survey of every known dictionary might not be sufficient,
moreover, for reasons explained below. See infra ___.
90
Id. at 567 (citing BALLENTINE‘S LAW DICTIONARY 655 (3d ed. 1969) (defining
―interpreter‖ as ―one who interprets, particularly one who interprets words written or
spoken in a foreign language‖)).
19 Judging Ordinary Meaning 3-23-17
longer in use; but without more such designation gives us only the
opinion of those lexicographers, and not a hard basis for an empirical
conclusion.91
An ―especially‖ designation may be even more unreliable. Such a
designation suffers from all of the problems inherent in the ―obsolete‖
designation. And it also masks another deficiency, going to the
arbitrariness of the distinction between two senses listed in a dictionary
(described further below). The fact that a given sense, or subsense, of a
term is a special application of another highlights the interrelationship
between the two senses.92
It suggests the notion that the two senses are
not highly distinct from each other, but instead are more exemplars or
prototypes of a broader category. That is what the Webster‘s definition
cited in Taniguchi seems to convey. The cited Webster‘s Third definition
of interpreter is ―‗one that translates; esp.: a person who translates orally
for parties conversing in different tongues.‘‖93
This is an indication that
the lexicographers who formulated this definition for Webster‘s viewed
the especially designated notion not as a separate sense but as an
exemplar of it—perhaps a common, prototypical example.
So for these reasons the Taniguchi opinion also employs inadequate
tools of measurement. Justice Alito‘s ―survey‖ and sense designations
seem more sophisticated but they are not ultimately good tools of
assessing empirical questions of ordinary meaning.
3. United States v. Costello
Judge Posner rejects a dictionary-based approach to ordinary meaning
in Costello. He rightly notes that ―[d]ictionary definitions are acontextual,
whereas the meaning of sentences depends critically on context,
including all sorts of background understandings.‖94
And for that reason
Posner turns to Google to get a ―rough index of the frequency of
[harbor‘s] use.‖95
This approach is innovative. But it is far from perfect.
Google might seem to be a good source for data-driven analysis of
language usage. ―The web is enormous, free, immediately available, and
largely linguistic.96
And ―it is appealing to use the web as a data source‖
91
See also DOUGLAS BIBER, CORPUS LINGUISTICS: INVESTIGATING LANGUAGE
STRUCTURE AND USE 39 (1998) (observing that ―citation slips‖ used by lexicographers
―represent only those contexts that a human reader happens to notice‖).
92
See infra Part III.
93
132 S.Ct. at 2003 (citing WEBSTER'S THIRD NEW INTERNATIONAL DICTIONARY
1182 (1976)).
94
Id.
95
Costello, 666 F. 3d at 1042 (2012).
96
Adam Kilgarriff, Googleology Is Bad Science, 33 COMPUTATIONAL LINGUISTICS
147 (2007) (discussing the limitations of Google as a corpus).
20 Judging Ordinary Meaning 3-23-17
because ―language analysis and generation benefit from big data.97
Google has low entry costs, moreover. Even the most Luddite lawyer or
judge is likely to be able to perform a basic Google search. Yet we still
see a range of problems in Posner‘s approach.
First is the black box of the Google algorithm. Google searches ―are
sorted according to a complex and unknown algorithm (with full listings
of all results usually not permitted) so we do not know what biases are
being introduced. If we wish to investigate the biases, the area we
become expert in is googleology not linguistics.‖98
Google returns can
vary by geography, by time of day, and from day to day.99
So Google
search results are rather unscientific.
Second are problems with the Google search engine: the fact that it
does not allow us to search only for verb forms of harbor and that it will
not allow us to look at a particular speech community or period of time
(but only contemporary web pages). If we are interested in knowing the
ordinary use of harbor as a verb among ordinary English speakers at the
time of the enactment of the statute at issue (1917), Google cannot give
us that kind of parsed data.
In light of these search engine problems Posner formulated his own set
of search terms—comparing hit counts for phrases like ―harboring
fugitives‖ and ―harboring guests,‖ and comparing relative hit counts for
his two sets of search terms. But that introduces another set of problems.
Posner gives no basis for his chosen set of search terms, and the terms he
chose seem likely to affect the outcome.
Finally, even assuming away the above problems, the hit counts that
Posner relies on may not be indicative of ordinariness in the sense of
frequency of usage. Posner implies that relative hit counts are an
indication of frequency of usage in our ordinary language. But that may
not hold. Google hit counts are based on the total number of web pages,
not the total number of occurrences of a given phrase.100
A single web
page may have 10s, 100s or 1,000s of uses of an individual word or
phrase that would only register in a Google search as a single hit. So hit
counts may not be a reliable indication of ordinariness even if we could
overcome the other problems identified here.
For all these reasons we think Judge Posner was onto something in
seeking an empirical method of measurement. But we think his Google
search was inadequate.
II. THEORIZING ORDINARY MEANING
97
Id.
98
Id.
99
Id.
100
Id.
21 Judging Ordinary Meaning 3-23-17
Before we turn to the measurement problem we must start by refining
our theory of ordinary meaning. Legal scholarship posits a range of
conceptions of ordinary meaning. Professor Richard Fallon‘s catalog is
perhaps the most extensive. He speaks of the ―semantic‖ or ―literal‖
meaning of the words of the law; the ―contextual‖ meaning informed by
―shared presuppositions‖ of speakers and listeners (which we take to
align with Sunstein‘s notion of ―public meaning‖101
and the Baude-Sachs
idea of the ―reader‘s understanding‖102
); the ―intended meaning‖ of the
lawmaker; the ―reasonable‖ or ―imputed‖ meaning attributed to
―hypothetical, reasonable legislators,‖ and the ―interpreted meaning‖ of
laws in judicial precedent.103
Not all of these conceptions fit within the ordinary meaning construct
that is our subject here. Certainly there is a case for respecting statutory
meaning embedded in judicial precedent. If judges have deemed a statute
to have a certain meaning in the past, the law of interpretation—informed
by principles of stare decisis—can (and should) yield due deference to
the ―interpreted meaning‖ established by precedent. But our reasons for
respecting such meaning have nothing to do with the rule of law premises
behind the law‘s search for the ordinary communicative content of the
words of the law. They stem from principles of stare decisis.104
Fallon‘s notion of ―reasonable‖ or ―imputed‖ meaning is also, but less
obviously, a conception of extra-ordinary meaning. This construct is
related to the ―fair reading‖ method advanced by Scalia and Garner in
Reading Law. And the inquiry is framed in objective-sounding terms—in
a search for ―objectified intent.‖ But on closer review this notion of
meaning has nothing to do with actual communicative content of the
words of the law (or, accordingly, of intentions to attribute to lawmakers
or reliance interests developed by the public). It is a highly constructive
inquiry aimed at an impossibly well-informed ideal legislator—one with
―aptitude in language, sound judgment, the suppression of personal
preferences regarding the outcome, and, with older texts, historical
linguistic research‖; ―an ability to comprehend the purpose of the text,
which is a vital part of its context,‖ and to glean it ―only from the text
itself‖; and even an understanding of ―a word‘s historical associations
acquired from recurrent patterns of past usage‖ and ―a word‘s immediate
syntactic setting—that is, the words that surround it in a specific
utterance.‖105
101
Sunstein, supra note __, at 198.
102
Baude & Sachs, supra note __, at 1090 (distinguishing the ―author‘s intent‖ and
the ―reader‘s understanding‖).
103
See Fallon, supra note __, at 1255–63.
104
See id. at 1251 (articulating stare decisis arguments in support of the law‘s
acceptance of ―interpreted meaning‖).
105
Id.
22 Judging Ordinary Meaning 3-23-17
This notion of meaning has little or nothing to do with the actual
meaning intended by a legislator or understood by the public. We may
well have reasons to credit this sort of idealized, constructive meaning.
But if we do it will not be because we think that any actual legislator is
likely to have read the words of a law and understood it in this
―reasonable‖ way, much less that an ordinary member of the public
gleaned that understanding. It will be because we deemed other
policies—policies having nothing to do with vindicating linguistic
meaning—to be of greater significance.106
That leaves (a) semantic meaning, (b) contextual meaning (public
meaning or the reader‘s understanding); and (c) intended meaning.
Semantic meaning, in Fallon‘s taxonomy, is meaning that the language of
the law would have ―for someone operating solely with dictionary
definitions, rules of grammar, and other general propositions bearing on
how the meaning of a sentence emerges from the combination of its
elements.‖107
And Fallon attributes this sort of meaning to a sort of
―literalist‖ textualism, asserting that ―[p]articipants in legal discourse
frequently assume or argue that a legal provision‘s semantic or literal
meaning determines its legal meaning.‖108
The point here is the notion that ―literalist‖ textualism excludes nonsemantic
context. In contrasting his notion of ―contextual‖ meaning, for
example, Fallon distinguishes ―semantics, which is concerned with the
context-independent meaning of words, phrases, and sentences, and
pragmatics, which involves the meaning of utterances in particular
contexts.‖109
Here he cites an example from Reading Law: ―‗Nail in a
regulation governing a beauty salon has a different meaning from nail in
a municipal building code.‘‖110
Literalist semantic meaning alone is not an indication of ordinary
communicative content. Real human beings do not derive meaning from
dictionary definitions and rules of grammar alone. Everyone takes nonsemantic
context—pragmatics—into account in deriving meaning from
language.111
And for that reason we see no basis to credit semantic
106
See Gary Lawson & Guy Siedman, Originalism as a Legal Enterprise, 23 CONST.
COMMENT. 47, 48 (2006) (stating that the ―touchstone‖ of this approach to interpretation
―is not the specific thoughts in the heads of any particular historical people, but rather
the hypothetical understandings of a reasonable person who is artificially constructed by
lawyers‖).
107
Fallon, supra note __, at 1245.
108
Id.
109
Id. at 1246.
110
Id. (quoting SCALIA & GARNER, supra note __, at 20).
111
In the law, pragmatism can refer to ―basing judgments (legal or otherwise) on
consequences, rather than on dedication from premises in the manner of syllogism.‖
Richard A. Posner, How Judges Think 40 (2010). In linguistics, pragmatics can refer to
the study of specific linguistic phenomena like conversational implicature or deixis.
23 Judging Ordinary Meaning 3-23-17
meaning without consideration of pragmatic context. If no lawmaker
would read the text that is voted into law purely semantically—devoid of
pragmatic context—then there is no reason to credit that kind of meaning
as a means of vindicating the intent of a lawmaker. The same would go
for the public governed by the law. If no one reads laws literally by pure
semantics, we should have no reason to protect reliance interests or notice
concerns rooted in that kind of understanding.
That does not make the elements of semantic meaning irrelevant. We
humans do take verbal, semantic context into account in interpreting
language. It‘s just that we also take nonverbal, pragmatic context into
account. And we think it important to highlight each of these elements of
context that might affect our understanding—further to clarify the
determinants of ordinary meaning that our law might seek to measure.
Before we turn to that endeavor, however, we first finish our treatment
of Fallon‘s taxonomy. Is the law‘s search for meaning aimed at finding
the ―public‖ meaning inferred by a ―reader‖ of the law or a more private
―intended‖ sense of a lawmaker? On this we agree with Baude and Sachs.
―There may be good reasons for a legal system to prefer‖ either public
meaning or intended meaning.112
And ―neither has to win every time,‖
because the ―right‖ answer ―depends on our reasons‖ for the resort to
ordinary meaning in the first place.113
Intended meaning is an appropriate construct to the extent we are
aiming to vindicate the preferences of lawmakers. This is a viable,
distinct basis for crediting ordinary meaning. And a decision to credit
such meaning is appropriate to the extent we find the policy of protecting
the intentions of lawmakers to predominate. We may say, as does
Professor Larry Alexander, ―that the reason we should seek the actual
authors‘ intended meaning is that the actual authors possessed the legal
authority to promulgate norms, and their texts just are their
communications of the norms they intended to promulgate.‖114
And if
that is our premise for looking to the ordinary communicative content of
the law then we will certainly look to the intended meaning of lawmakers
(informed by relevant elements of context, as presented below). Even the
reader, at least arguably, would seek for this meaning.115
Alan Cruse, Meaning In Language: An Introduction to Semantics and Pragmatics 317-
97 (2004). Here we use the term generally with reference to non-verbal context that may
affect meaning.
112
Baude & Sachs, supra note __, at 1091.
113
Id. at 1090.
114
Larry Alexander, Telepathic Law, 27 CONST. COMMENT. 139, 140 (2010).
115
See Larry Alexander, Originalism, the Why and the What, 82 FORDHAM L. REV.
539, 540 (2013) (asserting that ―our job is to determine the uptake the legislator(s)
intended us to have). We also agree with Professor Ryan Doerfler, however, that
legislative intent is ultimately a fiction—not only because ―Congress is a ‗they,‘ not an
‗it,‘‖ or because ―Members of Congress share no . . . intention to treat as authoritative
24 Judging Ordinary Meaning 3-23-17
But there is also a case for the public or ―reader‘s‖ understanding. This
sort of meaning makes sense to the extent we are seeking to vindicate the
notice rationale for the ―standard picture‖—the protection of reliance
interests and the avoidance of unfair surprise.116
Enforcing ―hard-to-find
intentions‖ of lawmakers ―would make the law unpredictable and
arbitrary.‖117
So to the extent our search for ordinary meaning is aimed at
protecting these interests, we should seek to assess the public‘s
understanding of the law (again, as properly informed by relevant
elements of semantic and pragmatic context—as described below).
So again this is a viable construct for our inquiry into ordinary
meaning. Before framing the theory of meaning in manner that may allow
us to measure it, we must first delineate the components of such meaning.
At a broad level those components encompass semantic meaning and
pragmatic meaning. To assess meaning, linguists would also tell us that
we must take into account the relevant speech community (whose
meaning?) and the relevant timeframe (meaning as of when?). We
explore each of these components below.
A. Semantic Meaning
Semantic meaning encompasses a range of components: lexicography,
syntax, and semantics.
1. Lexicography
the views of a statute‘s ‗principal sponsors‘ or ‗others who worked to secure
enactment,‘‖ but also because language must be understood in light of context
consisting of ―information salient to both author and audience.‖ Ryan Doerfler, Who
Cares How Congress Really Works?, 66 DUKE L. J. 979, 982–83 (2017). This suggests
that the line between intended meaning and public meaning is thin or perhaps
nonexistent, a point we return to below. See supra __.
116
See Antonin Scalia, Common-Law Courts in a Civil-Law System: The Role of
United States Federal Courts in Interpreting the Constitution and Laws, Address at the
Tanner Lectures on Human Values at Princeton University (March 8–9, 1995) at 92,
available at http://tannerlectures.utah.edu/_documents/a-to-z/s/scalia97.pdf. (asserting
that it is ―incompatible with democratic government—or indeed, even with fair
government—to have the meaning of a law determined by what the lawgiver meant,
rather than by what the lawgiver promulgated‖); Note, Textualism as Fair Notice, 123
HARV. L. REV. 542, 542 (2009) (―Perhaps the most intuitive and straightforward
argument for textualism is that it promotes fair notice of the law.‖); Michael Herz,
Purposivism and Institutional Competence in Statutory Interpretation, 2009 MICH. ST.
L. REV. 89, 102 (2009) (arguing that ―the case for textualism‖ is in part ―[t]he claim . . .
that if legal rules are embedded in publically available texts, affected person will be able
to know, understand, and comply with those rules. . . . [T]he fair notice argument
for textualism in statutory interpretation presupposes, and seeks to ensure the full benefit
of, a shift from the common law to statutes‖).
117
Baude & Sachs, supra note __, at 1091.
25 Judging Ordinary Meaning 3-23-17
The search for ―semantic‖ meaning often distills to a question of word
sense. In Muscarello, Taniguchi, and Costello, for example, the courts
were considering a problem of competing word senses—senses numbered
separately from each other in the macrostructure of the cited dictionaries.
Judges tend to assume that a dictionary‘s division of senses (by
numbers and letters) represents an immutable linguistic fact about the
universe. We tend to ―ignore the fact,‖ as Professor Larry Solan put it,
―that someone sat there and wrote the dictionary, and we speak as though
there were only one dictionary, whose lexicographer got all the
definitions ‗right‘ in some sense that defies analysis.‖118
But that is not
linguistic reality. Dictionaries may differ sharply in the number of senses
they assign to a given term or in the divider they use to distinguish
senses. Dictionaries do not ―emerge from some lexicographical Sinai;
they are the products of human beings. And human beings, try as they
may, bring their prejudices and biases into the dictionaries they make.‖119
The question of ―what is a word sense‖ turns out to be a very
challenging one in lexical semantics. Linguists and lexicographers lack
―decisive criteria for defining word senses and clearly discriminating
between them.‖120
This ―has always been a burning issue of lexical
semantics.‖121
And linguists also acknowledge that the sense distinctions
reflected in dictionaries are ―more of a descriptive device rather than a
claim about psycholinguistic reality.‖122
In traditional lexicography, words are defined first by determining the
class of things to which they belong (their genus) and second by
distinguishing them from all other things in their class (their species).123
Words are then divided into senses based on a variety of factors,
including their part of speech, pronunciation, inflection, etymology, and
shades of meaning.124
This approach to defining words and dividing them
118
Lawrence Solan, When Judges Use Dictionaries, 68 AM. SPEECH 50, 50 (1993)
(―[O]ur society‘s reverence for dictionaries is not driven by the latest discoveries in
psycholinguistic research. Rather, it is deeply embedded in our culture.‖).
119
JONATHAN GREEN, CHASING THE SUN: DICTIONARY MAKERS AND THE
DICTIONARIES THEY MADE xiv (1997).
120
Nikola Dobric, The Predictive Power of the (Micro)Context Revisited—
Behavioral Profiling and Word Sense Disambiguation, UDC 77 (2014),
https://www.researchgate.net/publication/270285812_The_Predictive_Power_of_the_M
icroContext_Revisited_-_Behavioral_Profiling_and_Word_Sense_Disambiguation.
121
Id.
122
Dylan Glynn, Polysemy and Synonymy, in CORPUS METHODS FOR SEMANTICS:
QUANTITATIVE STUDIES IN POLYSEMY AND SYNONYMY 10 (Dylan Glynn & Justyna A.
Robinson eds., 2014).
123
SIDNEY I. LANDAU, DICTIONARIES: THE ART & CRAFT OF LEXICOGRAPHY 153 (2d
ed. 2004).
124
BO SVENSÉN, PRACTICAL LEXICOGRAPHY: PRINCIPLES AND METHODS OF
DICTIONARY-MAKING 204–05 (John Sykes & Kerstin Schofield trans. 1993).
26 Judging Ordinary Meaning 3-23-17
into senses can be highly impressionistic and has a number of limitations.
There is no agreed-upon formula for sense division—some
lexicographers make very fine-grained distinctions between senses (they
are sometimes called splitters), while others tend to make broader, more
coarse-grained distinctions (they are sometimes called lumpers.)125
Moreover, the citation or quotation files from which many contemporary
dictionaries are derived were collected without the benefit of modern
sampling methods. Accordingly, as noted in Part I above, they cannot not
be relied upon for information about the frequency of a given word or
word sense.126
Contemporary lexicographers have moved past reliance on citation
files and have begun to rely on electronic ―corpora‖—large bodies or
databases of naturally occurring language—to gather linguistic data.
Corpus analysis has allowed lexicographers to address the problem of
sense division with greater granularity. Lexicographers can now view a
more complete range of potential uses of a given word and collect
statistical information about the likelihood of a given word appearing in a
particular semantic environment.127
We should be sensitive to the challenges presented by sense division
in our assessment of ordinary meaning. We should not jump immediately
to the conclusion that two senses listed separately in a dictionary reflect
―psycholinguistic reality‖—that a distinct listing in a dictionary tells us
that human beings will understand distinctly listed senses as mutually
exclusive constructs. Nor should we assume the converse—that that the
lack of distinct listing tells us that human beings will not perceive a
distinction.
Instead we should look to what is less arbitrary and more readily
measurable—to patterns that emerge from corpus linguistic data, or to
what we can glean about human perception from psycholinguistic
surveys and experiments. The point can be made by reference to
competing senses of carry in Muscarello or the alternative notions of
interpreter in Taniguchi. Lexicographers may disagree about where to
draw the lines between senses of these terms, or whether the two
alternatives are distinctly separate from each other.128
But if the question
of the dividing line is in some sense arbitrary and not reflective of
125
ANN O‘KEEFFE & MICHAEL MCCARTHY, THE ROUTLEDGE HANDBOOK OF
CORPUS LINGUISTICS 433 (2010).
126
LANDAU, supra note __, at 153.
127
JOHN LYONS, LINGUISTIC SEMANTICS: AN INTRODUCTION 80 (1995) (noting that
another way to think about word senses is as ―the set, or network, of sense-relations that
hold between [a word] and other expressions of the same language.‖).
128
See DOUGLAS BIBER, SUSAN CONRAD, RANDI REPPEN, CORPUS LINGUISTICS:
INVESTIGATING LANGUAGE STRUCTURE AND USE 40 (1998) (documenting the
differences in the definition and sense distribution of the noun ―deal‖ as recorded in five
general-use dictionaries).
27 Judging Ordinary Meaning 3-23-17
psycholinguistic reality, then we ought to seek to measure
psycholinguistic reality rather than relying on the sense-divisions in the
dictionaries before us.
This problem is most acute as to two senses that are viewed as closely
related to each other. The two notions of interpreter in Taniguchi are
illustrative. We can find distinct definitions encompassing oral translator
and written translator. But that may not tell us that these two senses
represent psycholinguistic reality—that one sense would be viewed as
excluding the other. The notion of oral translator could simply be
perceived as a more common ―prototype‖ of the more general notion of
―one who translates.‖ And the written translator idea could certainly be
viewed as an atypical example. That may be all that dictionaries are
telling us by the indication that written translator is ―obsolete.‖ And if so,
that sort of obsolescence would not tell us that an ordinary person would
not understand text providing for compensation for an interpreter to
cover a written translator. (More on solving this puzzle in Part III.C.)
A dodo, after all, is an obsolete bird. But it is still a bird. And a person
who happened to discover a remaining dodo on a remote island would
certainly be understood to be in possession of a bird. Such a person
would be covered, for example, by the terms of a rental agreement
prohibiting tenants to keep ―dogs, cats, birds, or other pets‖ in their
apartments. If you are found in possession of a caged dodo, you are not
likely to escape the wrath of the landlord by insisting that a dodo is an
―obsolete‖ sort of a bird.
2. Syntactic and Semantic Context
The need to consider context is a staple element of the judicial inquiry
into ordinary meaning. Courts often reference the notion of ―context‖
when they invoke the ordinary meaning canon.129
Yet they rarely say
what they mean by context. Linguistic theory can help identify which
elements of context may matter.
Context can be viewed as encompassing both verbal and non-verbal
components.130
The verbal context of a word or phrase in a statute
129
See Johnson v. United States, 559 U.S. 133, 139 (2010) (observing that in the
ordinary meaning inquiry, ―[u]ltimately, context determines meaning‖); Braunstein v.
C.I.R., 374 U.S. 65, 70 (1963) (defining the ordinary meaning of ―gain‖ in a particular
context); Chisom v. Roemer, 501 U.S. 380, 404 (1991) (Scalia, J., dissenting) (arguing
that the ―regular method for interpreting the meaning of language in a statute‖ was to
―first, find the ordinary meaning of the language in its textual context‖). We are
referring to linguistic context, which is a somewhat different concept than seeing if the
statutory context precludes turning to ordinary meaning altogether. See, e.g., Gonzales
v. Carhart, 550 U.S. 124, 152 (2007) (―In interpreting statutory texts courts use the
ordinary meaning of terms unless context requires a different result.‖).
130
Charles Goodwin & Alessandro Duranti, Rethinking Context: An Introduction, in
RETHINKING CONTEXT: LANGUAGE AS AN INTERACTIVE PHENOMENON 6–9 (1991).
28 Judging Ordinary Meaning 3-23-17
includes its syntactic and semantic environments. Syntax is a set of rules
and principles that governs sentence formation and determinates which
sentences will convey meaning to members of the same speech
community.131
One aspect of syntax is argument structure, a linguistic
term of art that has reference to the participants in the action of a verb.132
A transitive verb, like carry has two arguments—the subject and the
object. If we are interested in examining the meaning of phrases like
carries a firearm, we would look for phrases that have a similar argument
structure. And those arguments may affect our understanding of the
meaning of carry.
Semantic context may also affect our perception of meaning.
Semantics is the study of meaning at the word or phrase level.133
Embedded within the words and phrases we use are a number of concepts
that are sometimes referred to as the semantic features or semantic
components of a word.134
These features include concepts like number,
animacy, gender, humanness, and concreteness (i.e., tangibleness).135
In
semantic theory, words can also be understood with reference to their
functional role in an utterance. A word has an agentive function if it is
instigator of the action of a verb, or an objective function if it is the entity
that is affected by the action of the verb. A word may also serve an
instrumental function if it is a force or object involved in, but not
instigating, the action. 136
An illustration of these linguistic concepts may be made by reference
to the Muscarello statute‘s requirement of a mandatory minimum
sentence for ―anyone who . . . carries a firearm.‖ Our understanding of
anyone who carries a firearm is informed by the syntactic arguments—
131
17 OXFORD ENGLISH DICTIONARY 487 (2d ed. 1989). Syntax is also the study of
these rules and principles. Noam Chomsky, Syntactic Structures 11 (1957) (―Syntax is
the study of the principles and processes by which sentences are constructed in
particular languages.‖).
132
ALAN CRUSE, MEANING IN LANGUAGE: AN INTRODUCTION TO SEMANTICS AND
PRAGMATICS §§ 14.1-5 (2011). An intransitive verb (fall, die, yawn) has one
argument—the subject. A di-transitive verb (throw, send) has three arguments—the
subject, the direct object, and the indirect object.
133
MICHAEL MORRIS, AN INTRODUCTION TO THE PHILOSOPHY OF LANGUAGE 152
(2007) (―Semantics is the attempt to give a systematic explanation of how the meaning
of sentences depends upon the meaning of their parts.‖).
134
JOHN I. SAEED, SEMANTICS 260–62 (20XX).
135
Id.
136
Following Charles Fillmore, Alan Cruse lists a number of functional roles for
words, including agentive—the instigator of the action; instrumental—the force or
object involve in the action of the verb; dative/experiencer—the animate being affected
by the action of the verb; factitive—the result of the action of the verb; locative—
location or spatial orientation of the verb; and objective—the inanimate entity affected
by the action of the verb. CRUSE, supra note ___, at § 14.5.
29 Judging Ordinary Meaning 3-23-17
with who as the subject and firearm as the object of the verb. With
respect to semantic features, we can characterize the relevant subject of
the statute in Muscarello as animate, human, and concrete. Similarly, we
can characterize the relevant object as inanimate, concrete, non-human
and even weapon. With respect to functional roles, who performs the
agentive function, instigating the action of carry and firearm serves an
objective function. In the phrase anyone who carries a firearm in a car,
car serves an instrumental function—it is involved in the action of the
verb, but doesn‘t instigate it and doesn‘t receive it.
By looking to the argument structure and semantic features of the
relevant statute, we are able to perform a more targeted search for
language data to inform our inquiry into ordinary meaning. When we
seek to measure language usage, we may wish to limit our search to uses
of the verb carry that share the above-noted features—the syntax of a
transitive verb, with the semantic features of a human subject and a
weapon object.137
B. Pragmatic Meaning
The meaning of an utterance will not always be expressly
communicated in its semantic content. Non-verbal (pragmatic) contextual
considerations will also be taken into account.138
Such considerations
may encompass the physical or social setting of an utterance, and even an
inference about the intent of the speaker.139
Judge Richard Posner‘s ―Keep off the grass‖ problem is a good
illustration. As Posner notes, a sign in a park that says ―‗Keep off the
grass‘ is not properly interpreted to forbid the grounds crew to cut the
137
See Muscarello, 524 U.S. at 143 (Ginsburg, J., dissenting) (noting that the issue
presented ―is not ‗carries‘ at large but ‗carries a firearm‘‖).
138
See Goodwin & Duranti, supra note __, at 6-9; CRUSE, supra note __. Pragmatics
includes concepts like conversational implicature, where the meaning of an utterance is
strongly implied but not expressly stated—as where a spouse who says ―there sure are a
lot of dishes in the sink‖ is not just making an observation about the state of the universe
but is reminding somebody about whose turn it is to do the dishes).
139
Careful scholars have recognized this point; and they have identified it as a basis
for concluding that the space between textualism and intentionalism is small. See Larry
Alexander & Saikrishna Prakash, “Is That English You're Speaking?”: Why Intention
Free Interpretation is an Impossibility, 41 SAN DIEGO L. REV. 967, 979 (2004) (―[T]he
commonplace truth that all understandings of texts are contextual just demonstrates that
all texts qua texts acquire their meaning from the presumed intentions of their
authors.‖); Scalia, Response, in A MATTER OF INTERPRETATION: FEDERAL COURTS AND
THE LAW 144 (1998) (conceding that ―what the text would reasonably be understood to
mean‖ and ―what it was intended to mean‖ are concepts that ―chase one another back
and forth to some extent, since the import of language depends upon its context, which
includes the occasion for, and hence the evident purpose of, its utterance‖).
30 Judging Ordinary Meaning 3-23-17
grass.‖140
Our understanding of the meaning of this sign is informed by
more than just its semantic and syntactic content. We understand it in
light of its pragmatic context, which includes inferences about the place
and manner of the utterance and presumed intentions of the speaker.
Pragmatic considerations are of relevance to any attempt to assess the
ordinary meaning of a statutory phrase. An utterance that merely
describes a person carrying a firearm might be understood to convey one
ordinary meaning. But a criminal prohibition—more precisely, a
requirement of a mandatory minimum criminal sentence—may be
understood differently. At least that‘s possible. And we may need to take
such context into account in assessing ordinary meaning.
C. Meaning as of When?
Human language is in a constant state of change.141
But it does not
change at a predictable rate.142
Nor do different linguistic features change
at the same time.143
Our theory of ordinary meaning must take account of
this variation and allow us to examine the linguistic norms prevailing at
different historical periods.
We are used to thinking about timeframe in constitutional
interpretation. There we often acknowledge that original meaning may
differ from modern meaning. But we often ignore the problem in
statutory interpretation. Or sometimes we just assume it away. In
Costello, for example, Judge Posner noted that the harboring an alien
statute was enacted in 1917, but looked for modern data as to the ordinary
sense of the verb harbor.144
Perhaps he did so out of convenience or
necessity, given that his Google search framework would not have
allowed historical analysis. But his stated reason was reflective of the
reality of much litigation over statutory interpretation: The parties simply
didn‘t bother to consider the possibility that the term harbor may have
evolved over time—both sides presented dictionary definitions from
140
See RICHARD A. POSNER, LAW, PRAGMATISM, AND DEMOCRACY (2005) (making
the case for a pragmatic theory of government); RICHARD A. POSNER, HOW JUDGES
THINK 40 (2010) (―[Pragmatism] refers to basing judgments (legal or otherwise) on
consequences, rather than on dedication from premises in the manner of syllogism.‖).
141
JOHN LYONS, INTRODUCTION TO THEORETICAL LINGUISTICS 43 (1968) (―All
language are subject to constant change. This is an empirical fact . . . . All living
languages . . . are of their nature efficient and viable systems of communication serving
the different and multifarious social needs of the communities that use them. As these
needs change, languages will tend to change to meet the new conditions.‖).
142
TERRY CROWLEY & CLAIRE BOWERN, AN INTRODUCTION TO HISTORICAL
LINGUISTICS 149-51 (2011) (discussing criticisms of attempts to quantify the rate of
language change).
143
Id.
144
Costello, 666 F.3d at 1043–44.
31 Judging Ordinary Meaning 3-23-17
modern times—so Judge Posner appears to have concluded that that gave
him license to do the same thing.
That phenomenon is at least sometimes reflected in our theory of
statutory interpretation. At least a few courts have looked to the ordinary
meaning of a statute as of the time it was enacted.145
That seems
appropriate to the extent we are seeking ―intended‖ meaning. This is the
point of the originalists who argue for the vindication of intended original
meaning—that the ―ratifiers of the Constitution . . . are the persons with
authority to make and change constitutional norms,‖ and thus that we are
bound by their views.146
If we ―‗interpret the Constitution as if it had
been authored by someone other than its ratifiers,‖ these originalists
argue, we are ―mak[ing] constitutional ‗law‘ without authority to do
so.‖147
And the same point can be made as to statutes. If intended
meaning is the relevant construct, we must be bound by meaning as of the
time of the statute‘s initial enactment. Otherwise we are vindicating
intentions at other times and by other people.
The ―public‖ meaning construct at least arguably leads in a different
direction. If we are seeking to protect reliance interests and fair notice we
are at least arguably opting for contemporary (not historical) evidence of
ordinary meaning. ―Normal‖ English speakers are guided by their
contemporary understanding. And they lack a sophisticated
understanding of historical usage. So if we are trying to assess public
meaning we may be assuming away the temporal premises of
originalism, and searching only for contemporary meaning.
If we seek to measure historical meaning how can we do so? One
common means of assessing historical ordinary meaning is to consult an
old dictionary. That is an approach that courts often take in seeking the
original meaning of the constitution. But that practice is fraught with all
of the difficulties highlighted above as to contemporary dictionaries:
Historical dictionaries, no more than contemporary ones, cannot yield
reliable information about which of various senses is more ordinary. And
the problems are compounded as to historical dictionaries.148
145
See, e.g., Carcieri v. Salazar, 555 U.S. 379, 388 (2009); Norfolk S. Ry. Co. v.
Perez, 778 F.3d 507, 512 (6th Cir. 2015); State v. Ziska, 334 P.3d 964, 967 (Or. 2014);
State Bd. of Nursing v. Ruebke, 913 P.2d 142, 157 (Kan. 1996).
146
Alexander, supra note __, at 141.
147
Id.
148
See Rickie Sonpal, Old Dictionaries and New Textualists, 71 FORDHAM L. REV.
2177, 2209 (2003) (―Supreme Court Justices are sometimes very scrupulous about
choosing the dictionary and edition with a publication date close to the date the statute
was enacted; yet, this practice is often of deceptively limited value. This practice is of
even less value when old dictionaries are used because some popular older dictionaries
were not only reprinted but even appeared in new editions without any substantive
change to the body of the dictionary. . . . Accordingly, judges who carefully choose the
printing or edition of an old dictionary that is most closely contemporary with the statute
risk relying on a dictionary the substance of which far antecedes the statute.‖).
32 Judging Ordinary Meaning 3-23-17
This is not to say that historical dictionaries do not have value.
Historical dictionaries can be useful for defining unknown terms and
attesting contested uses. But we ought to regard them with the skepticism
when they are offered as evidence of ―ordinary‖ or ―original‖ meaning.
D. Whose Meaning?
Our understanding of meaning is also shaped by our speech
community—the group of people with whom we share a set of linguistic
norms, conventions, and expectations about linguistic behavior.149
Meaning may also vary across different linguistic registers—varieties of
texts, ranging from spoken communications, to newspapers, academic
prose, or even congressional committee reports that tend to share
linguistic features. Our theory of ordinary meaning must be able to
account for the speech community we are evaluating and must be able to
address the differences in various linguistic registers.
The choice between ―public‖ meaning and ―intended‖ meaning may
have implications for our identification of the relevant speech
community. The public meaning construct seems to dictate a speech
community consisting of a broad cross-section of the public. The
intended meaning inquiry, on the other hand, could at least arguably point
to a more limited community. Members of Congress are generally not
common, ordinary people. And their usage may not be colloquial. So if
our search for ordinary meaning is aimed at deriving intended meaning,
we may wish to assess the usage or understanding of a more sophisticated
group of English speakers. And we may also wish to take into account the
pragmatic consideration that the more formal nature of legal language can
affect human understanding of meaning.150
149
MARCYLIENA H. MORGAN, SPEECH COMMUNITIES: KEY TOPICS IN LINGUISTIC
ANTHROPOLOGY 1 (2014) (―Speech communities are groups that share values and
attitudes about language use, varieties and practices. These communities develop
through prolonged interaction among those who operate within these shared and
recognized beliefs and value systems regarding forms and styles of communication.‖);
SANDRA LEE MCKAY, SOCIOLINGUISTICS AND LANGUAGE TEACHING 49 (1996) (―A
conglomeration of individuals who share the[] same norms about communication is
referred to as a speech community. A speech community is defined as a community
sharing a knowledge of the rules of the conduct and interpretation of speech.‖); John
Sinclair, Meaning in the Framework of Corpus Linguistics, 20 LEXICOGRAPHICA 1, 22
(2004) (―The differences in interpretation between members of a speech community are
small and they do not interfere much with normal communication.‖); Reed Dickerson,
Statutory Interpretation: Dipping Into Legislative History, 11 HOFSTRA L. REV. 1125,
1154 (1982) (defining speech community as the ―group of people who share a common
language (or sublanguage) and thus a common culture (or subculture), which in turn
defines the context that conditions the utterances that occur within it‖).
150
But see Doerfler, supra note __, at 983–84 (articulating a ―conversation‖ model
of ―fictionalized‖ legislative intent, in which ―[a]n interpreter occupies the position of
33 Judging Ordinary Meaning 3-23-17
III. OPERATIONALIZING ORDINARY MEANING
The above sets the stage for a more careful formulation of the law‘s
assessment of the ordinary communicative content of the language of the
law. We may choose to measure either public meaning or intended
meaning. And however we choose to frame the inquiry we should
account for all of the semantic, pragmatic, temporal, and speech
community considerations that may be relevant.
That leaves the question of measurement or operationalization. Again
we can find help in linguistics. We propose the use of tools employed in
corpus linguistics. Corpus linguistics is an empirical approach151
to the
study of language that involves large, electronic databases of text known
as corpora (the plural of corpus).152
A corpus is a body or database of
naturally occurring language.153
Corpus linguists draw inferences about
language from data gleaned from real-world language in its natural
habitat—in books, magazines, newspapers, and even transcripts of
spoken language.154
The defining characteristic of corpus linguistics is
―the claim that it is possible to actually ‗represent‘ a domain of language
conversational participant, hearing statements directed at her and other participants‖ and
credits ―information salient to both members of Congress and to citizens‖).
151
TONY MCENERY & ANDREW HARDIE, CORPUS LINGUISTICS: METHOD, THEORY
AND PRACTICE 49 (2011) (―Empiricism lies at the core of corpus linguistics‖); Paul
Baker, Glossary of Corpus Linguistics 65 (2006) (―In linguistics, empiricism is the idea
that the best way to find out about how language works is by analyzing real examples of
language as it is actually used. Corpus linguistics is therefore a strongly empirical
methodology.‖).
152
MCENERY & HARDIE, supra note ___, at i.
153
Douglas Biber, Corpus-Based and Corpus-driven Analyses of Language
Variation and Use, in THE OXFORD HANDBOOK OF LINGUISTIC ANALYSIS 159 (2010)
(―Corpus linguistics is a research approach that has developed over the past several
decades to support empirical investigations of language variation and use, resulting in
research findings that have much greater generalizability and validity than would
otherwise be feasible . . . . It utilizes a large and principled collection of natural texts,
known as a ―corpus‖, as the basis for analysis.‖).
154
Id. at 160–61 (―[Corpus linguistics] depends on both quantitative and qualitative
analytical techniques. . . . [T]he major contribution of corpus linguistics is to document
the existence of linguistic constructs that are not recognized by current linguistic
theories. Research of this type—referred to as a ―corpus-driven‖ approach—identifies
strong tendencies for words and grammatical constructions to pattern together in
particular ways, while other theoretically possible combinations rarely occur. . . .
[C]orpus-based research investigates the patterns of variation among the full set of
spoken and written registers in a language. In speech, these include casual face-to-face
conversation, service encounters, lectures, sermons, political debates, etc.; and, in
writing, these include email messages, text-messaging, newspaper editorials, academic
research articles, etc.‖).
34 Judging Ordinary Meaning 3-23-17
use with a corpus of texts, and possible to empirically describe linguistic
patterns of use through analysis of that corpus.‖155
Through corpus analysis we can test our hypotheses about language
through rigorous experimentation with observable and quantifiable data.
And the results of a corpus-based conclusion will be replicable and
falsifiable.156
Corpus data can tell us the relative frequency of different senses of
vehicle (or of carrying a firearm, of interpreter, or of harboring an alien)
in naturally occurring language.157
And if the search for ordinary
meaning entails an analysis of the relative frequency of competing senses
of a given term, then corpus linguistics seems the most promising tool.
Such data can also inform our assessment of linguistic prototype.158
If
the corpus data reveal that most vehicles that we speak of are
automobiles, or that most instances of carrying a firearm involve bearing
it on your person, we may infer that those senses are more likely to be
prototypical senses of the operative terms.
Below we drill down further on the proposed means of measurement.
First we present linguistic tools and means of measurement of the
components of ordinary meaning identified above. Then we illustrate the
utility of those tools by applying them to the cases and examples
discussed throughout the article. And we conclude this part of the article
with some observations about inferences that can be drawn from the data
about the ordinary meaning of vehicle, carry a firearm, interpreter, and
harbor.
155
BIBER & REPPEN, supra note ___, at [***].
156
McEnery & Hardie, supra note __, at 66 (―As a key goal of corpus linguistics is
to aim for replicability of results, data creators have an important duty to discharge in
ensuring the data they produce is made available to analysts in the future.‖).
157
Assuming, of course, the corpora used are properly constructed such that they
enable us to make generalizations about a larger population. See Douglas Biber,
Representativeness in Corpus Design, 8 LITERARY & LINGUISTIC COMPUTING 243
(1993); JESSE EGBERT, B. GRAY, & DOUGLAS BIBER, DESIGNING AND EVALUATING
LANGUAGE CORPORA (forthcoming).
158
Intuitively, we might assume that frequency and prototype would map onto one
another with some precision, but this is not always the case. See John R. Taylor,
Prototype Theory, in 1 SEMANTICS: AN INTERNATIONAL HANDBOOK OF LANGUAGE
MEANING 649–50 (Claudia Maienborn, Klaus von Heusinger & Paul Portner eds., 2011)
(―In response to the question ‗where does prototypicality come from?‘ (Geeraerts 1988),
many people are inclined to say that prototypes (or prototypical instances) are
encountered more frequently than more marginal examples and that that is what makes
the prototypical. Although frequency of occurrence certainly may be a factor (our
prototypical vehicles are now somewhat different from those of 100 years ago, in
consequence of changing methods of transportation) it cannot be the whole story.‖).
35 Judging Ordinary Meaning 3-23-17
A. Tools
Corpus linguistic tools can be employed to measure ordinary meaning
as conceptualized in this article. Here we explore the range of available
corpora and the functionalities they encompass.
1. Varieties of Linguistic Corpora
Linguistic corpora come in a number of varieties, each tailored to suit
the needs of a particular set of empirical questions about language use.
Corpora may be general or special. A general corpus endeavors to
represent the language used by a broad (often national) speech
community. Special corpora are limited to a particular genre, register, or
dialect.159
There are monitor corpora that are continuously updated with
new texts in order to track contemporary language use, and there are
historical or sample corpora that reflect the language use of a particular
period. We will rely on both monitor and historical corpora in the
analysis below.
Corpora may also be raw, tagged, or parsed. A raw corpus contains
almost no linguistic metadata (e.g., a .txt file containing the complete
works of Shakespeare would be a raw corpus). Tagged corpora typically
contain metadata from a grammatical ―tagging‖ program that
automatically marks each word that with a part of speech. A tagged
corpus can dramatically improve corpus analysis by allowing a researcher
to look for all different forms of a single word in a single search (e.g., a
search for the verb carry would automatically include every verb
inflection carries, carrying, and carried) and to search for results related
to a particular part of speech (e.g., the verb harbor not the noun harbor).
This type of search is called a lemmatized search—a search that searches
for the base form of a word and reveals its permutations. Parsed corpora
contain phrase-, clause-, or sentence-level annotation, revealing the
syntactic relationships among the words in the corpus. While automated
tagging is highly accurate, automated parsing is not. Thus, parsed corpora
tend to require a significant amount of human editing and annotation,
which increases the costs of their production dramatically. For that reason
parsed corpora tend to be smaller than tagged corpora. The corpora we
rely on in the analysis below are tagged, but not parsed. 160
159
For example, the Linguistic Data Consortium at the University of Pennsylvania
produced a corpus of recorded Egyptian Arabic telephone calls. See Alexandra Canavan,
George Zipperlen & David Graff, CALLHOME Egyptian Arabic Speech, LINGUISTIC
DATA CONSORTIUM (1997),https://catalog.ldc.upenn.edu/LDC97S45.
160
The corpora relied on in this paper were tagged by the Constituent Likelihood
Automatic Word-tagging System (CLAWS-7) program. Mark Davies, The 385+ Million
Word Corpus of Contemporary American English (1990-2008+): Design, Architecture,
and Linguistic Insights, 14 INT‘L J. OF CORPUS LINGUISTICS 159, 164 (2009).
36 Judging Ordinary Meaning 3-23-17
2. Corpus Tools—Frequency, Collocation, and KWIC
Linguistic corpora can perform a variety of tasks that cannot be
performed by human linguistic intuition alone. For example, as noted
above, corpora can be used to measure the statistical frequency of words
and word senses in a given speech community and over a given time
period.161
Whether we regard the ordinary meaning of a given word to be
the possible, common, or the most common sense of that word in a given
context, linguistic corpora can allow us to determine empirically where a
contested sense of a term falls on that continuum.
Corpora can also show collocation, ―which is the tendency of words to
be biased in the way they co-occur.‖162
As we have seen, words are often
interpreted according to the semantic environment in which they are
found. And a collocation program show the possible range of linguistic
contexts in which a word typically appears and can provide useful
information about the range of possible meanings and sense divisions.163
Corpora also have a concordance or key word in context (―KWIC‖)
function, which allows their users to review a particular word or phrase in
hundreds of contexts, all on the same page of running text. This allows a
corpus user to evaluate words in context systematically. This sort of
systematic observation of a word in context can allow us to gain
meaningful and quantifiable insight about the range of possible uses of a
word and the frequency of its different senses.
3. Representing Speech Community and Register in a Corpus
Linguistic corpora can be built from the ground up using text or
speech from any given speech community or register. This is a defining
characteristic of corpus linguistics—―the claim that it is possible to
actually ‗represent‘ a domain of language use with a corpus of texts, and
possible to empirically describe linguistic patterns of use through analysis
of that corpus.‖164
As professor Larry Solan has noted, ―[w]hen the legal
system decides to rely on the ordinary meaning of a word, it must also
determine which interpretive community‘s understanding it wishes to
adopt. This choice is made tacitly in legal analysis, but becomes overt
when the analysis involves linguistic corpora because the software
displays the issue on a screen in front of the researcher.‖165
In this article
161
TONY MCENERY & ANDREW WILSON, CORPUS LINGUISTICS: AN INTRODUCTION
82 (2d ed. 2001).
162
See SUSAN HUNSTON, CORPORA IN APPLIED LINGUISTICS 78 (2002).
163
Id. at 69.
164
BIBER & REPPEN, supra note __..
165
Solan, The New Textualists‟ New Text, supra note ___, at 2059.
37 Judging Ordinary Meaning 3-23-17
we rely on a pair of broad-based corpora of standard written American
English (one contemporary and one historical). But a corpus can be
constructed to represent the language use of any speech community or
register.166
4. Representing Historical Language Use
Finally, a linguistic corpus can be built from texts representing the
language use from any period in history. To the extent our understanding
of ordinary meaning should be informed by the linguistic norms and
conventions prevailing at the time that a given legal text was drafted,
corpus linguistics can provide powerful evidence of historic language
use.
5. The BYU Corpora
Below we will tackle the interpretive problems posed by the
Muscarello, Taniguchi and Costello cases using data from two linguistic
corpora: the News on the Web (―NOW‖) Corpus and the Corpus of
Historical American English (the ―COHA‖), both developed at Brigham
Young University and referred to here as the BYU Corpora. Here we
outline the parameters of each corpus and highlight their differences.
a. NOW Corpus
The NOW Corpus is a database of ―3.7 billion words of data from
web-based newspapers and magazines from 2010 to the present time.‖167
It is a monitor corpus that ―grows by about 4-5 million words of data
each day (from about 10,000 new articles), or about 130 million words
each month.‖168
The NOW Corpus downloads content every night from
dozens of websites listed on Google News, using an automated software
program.169
These texts are then automatically tagged and lemmatized
(adding part-of-speech metadata to each word) and integrated into the
existing corpus.170
Because of this extraordinary rate of growth, the NOW
Corpus is currently the largest tagged corpus of English in the world and
seems likely to keep that title for the foreseeable future.
166
While corpora vary in size and sophistication, anyone can build a corpus using
freely available software like AntConc. See AntConc Homepage, LAURENCE ANTHONY,
http://www.laurenceanthony.net/software/antconc/ (last visited Mar. 8, 2017).
167
See NOW Corpus, BYU, http://corpus.byu.edu/now/ (last visited Mar. 8, 2017).
168
Id.
169
Id.
170
Id.
38 Judging Ordinary Meaning 3-23-17
―There is no data like more data,‖171
and the chief virtues of the NOW
Corpus are its size and immediacy. With the NOW Corpus, the user is
able to examine what is happening in the language at the moment. And
because of the size and scope of the corpus, lower frequency linguistic
phenomena (words, word senses, and syntactic structures, etc.) are more
likely to be attested, while the distribution of higher frequency
phenomena will be better and more completely represented.
The NOW Corpus has a few limitations. First, even with searches
limited to U.S. sources, the NOW Corpus records the language use of a
single, large speech community (the United States) in a single linguistic
register (newsprint). But if the interpretation of a federal statute requires
us to consider the linguistic norms and conventions of the citizens subject
to that statute, then U.S. newsprint may be the appropriate speech
community and register. Spoken dialects of American English show
sharp (and increasing) differences in vocabulary, grammar, and
phonology,172
but the norms and conventions of the written variety of
American English (sometimes called standard written American English)
tend to be more uniform. Since we are interpreting a written text,
evaluating that text through the lens of a corpus standard written
American English from newsprint may be the right approach.
The NOW Corpus is also limited with respect to timeframe. NOW
tracks the linguistic norms and conventions of now (and over the past
decade). So if we want to evaluate interpretive problems against the
backdrop of linguistic norms prevailing at the time of enactment, we will
need to turn elsewhere.
b. Corpus of Historical American English (―COHA‖)
The COHA is ―the largest structured corpus of historical English.‖173
It contains ―more than 400 million words of text from the 1810s-2000s
(which makes it 50-100 times as large as other comparable historical
corpora of English) and the corpus is balanced by genre decade by
decade.‖174
Using data from the COHA, we can gather linguistic
information from the decade that a statute was enacted, going back
approximately 200 years.
Like the NOW Corpus, the COHA is limited in terms of speech
community and register. Though it has texts from a wider variety of
registers than the NOW Corpus (including fiction, magazines, non-
171
Kenneth Church, Has Computational Linguistics Become More Applied? in
Computational Linguistics and Intelligent Text Processing: 10th INT‘L CONF., CICLING
1, 3 (2009) (attributing the statement to Robert Mercer).
172
WILLIAM LABOV, DIALECT DIVERSITY IN AMERICA: THE POLITICS OF LANGUAGE
CHANGE 1–2 (2012).
173
Id.
174
Id.
39 Judging Ordinary Meaning 3-23-17
fiction), these tend to fall within the ambit of standard written American
English. In addition, the 400 million words of the COHA are spread out
over 200 years. Consequently, the COHA is essentially a collection of
twenty separate corpora (one for each decade from 1810 to 2010)
averaging just over twenty million words.
There is a lot of linguistic information to be gleaned from a twentymillion
word corpus. But as we have seen, in the specialized setting of
statutory interpretation it is important to evaluate words in context. These
contexts may be poorly represented (or not represented at all) in the
corpus with limited data for a given period. And the earliest texts in the
COHA date from the period of 1810 to 1820. These texts come in twenty
to thirty years shy of the Founding Era, leaving us without a data source
for the prevailing linguistic norms during the drafting and ratification of
the Constitution.175
With all of that said, the COHA remains the largest corpus of
historical American English. And it contains significant linguistic
information relevant to the statutes at issue in Muscarello, Taniguchi, and
Costello.
B. Applications
1. Vehicles in the Park
The ―no vehicles‖ problem seems a mandatory subject for any serious
treatment of statutory interpretation. It was introduced initially by
Professor HLA Hart176
in his famous debate with Professor Lon Fuller.177
But seemingly everyone has treated the problem since then.178
And there
is no shortage of extensions of the hypothetical. Hart says that ―[p]lainly‖
the rule ―forbids an automobile,‖ but asks ―about bicycles, roller skates,
[and] toy automobiles.‖179
(The airplane example invokes an actual
175
There are good reasons for this omission. Prior to the 1806 publication of Noah
Webster‘s influential text, A Compendious Dictionary of the English Language,
American spelling was very much in disarray, with many common words having as
many as a half dozen potential spellings. This makes corpus the construction of a corpus
interface and the automated tagging of corpus data very difficult (and expensive).
Moreover, because of widely varied orthographic practices, many historical texts are
difficult, if not impossible, to subject to optimal character recognition (―OCR‖). The
BYU Law School, moreover, is seeking to fill this gap. It is currently working on a
Corpus of the Founding Era American English (―COFEA‖).
176
Hart, supra note ___, at 608.
177
Lon L. Fuller, Positivism and Fidelity to Law—A Reply to Professor Hart, 71
HARV. L. REV. 630, 663 (1958).
178
See, e.g., SCALIA & GARNER, supra note __, at 37–38; Fallon, supra note __, at
1260–61; ESKRIDGE, supra note __, at 45–46.
179
Hart, supra note __, at 607.
40 Judging Ordinary Meaning 3-23-17
case—McBoyle v. United States,180
in which the United States Supreme
Court held that an airplane was not a vehicle under the National Motor
Vehicle Theft Act, which prohibited transporting stolen ―vehicles‖ across
state or national borders.181
)
The scholars cited throughout this article have offered their own views
on the scope of ―vehicle.‖ Scalia and Garner‘s READING LAW says that
the Hart prohibition should extend to any ―sizable wheeled conveyance,‖
and thus to automobiles—including ―ambulances, golf carts, mopeds,
motorcycles, and (perhaps) Segways‖—but not ―remote-controlled cars,
baby carriages, tricycles, or perhaps even bicycles.‖182
Professor Richard
Fallon objects to the extension to ambulances. He says the ―reasonable
meaning‖ of vehicle should not be understood to extend to ambulances—
at least those responding to emergencies.183
And Professor William
Eskridge disagrees with Scalia‘s assertion as to bicycles; he says that
―bicycles are commonly considered vehicles,‖ a conclusion he claims to
confirm using corpus data (more on this below).184
Yet no one has attempted to measure our assessment of vehicle with
any data. We present some relevant data below—data of relevance to the
frequency or prototypicality of various senses of this term.
a. Lexical Collocation of Vehicle through Time
One way to examine the most common context in which a word
appears is collocation. The collocation function of the corpus can show us
the words that are statistically most likely to appear in the same context
as vehicle for a given period. We can use collocation to get a snapshot of
the semantic environment in which vehicles appears and the kinds of
vehicles that tend to appear in that environment.
We can view the most common contemporary collocates185
of vehicle
in the NOW Corpus.186
In NOW the fifty most common collocates of
vehicle are as follows:
180
283 U.S. 25, 27 (1931).
181
18 U.S.C. § 408 (1919).
182
SCALIA & GARNER, supra note __, at 37–38.
183
See Fallon, supra note __, at 1260 – 61.
184
ESKRIDGE, supra note 4, at 45–46.
185
The NOW corpus and other BYU corpora are available without a subscription. To
access NOW you go to http://corpus.byu.edu/now/. To generate a list of collocates in
NOW you take the following steps: (1) Select ―Collocates‖ on the NOW Corpus
homepage; (2) Enter ―VEHICLE_n‖ in the ―Word/phrase‖ field (capitalization makes
the search lemmatized—assuring that we find all inflections of the word; the ―_n‖ is to
limit the search noun forms); (3) Enter an asterisk ―*‖ (a wildcard) in the ―Collocates‖
field; (4) Select ―Sections‖ and select ―United States‖; (5) Select ―Sort/Limit‖ and set
the ―Minimum‖ to ―Mutual Info‖; and (6) Click ―Find collocates.‖
186
The following link will reproduce the search above, except that the user would
need to repeated step four (4), ―Select ‗Sections‘ and select ‗United States,‘‖ which
41 Judging Ordinary Meaning 3-23-17
electric, motor, plug-in, unmanned, armored, connected, cars,
aerial, charging, pure, launch, owners, hybrid, traffic, fuel,
driving, gas, autonomous, struck, operating, road, safety,
accidents, battery, ownership, emergency, batteries, emissions,
seat, advanced, driver, primary, demand, gmv, commandeered,
fuel-efficient, uavs, automakers, demonstrators, excluding,
lunar, passenger, fleet, gasoline, luxury, drove, parking,
retirement, vehicles, infrastructure187
Many of the collocates of vehicle in the NOW Corpus strongly
indicate automobile as a likely candidate for vehicle‘s most common use
of vehicle. The NOW Corpus lists a number of automotive collocates like
motor, car, traffic, fuel, driving, gas, battery, batteries, emission, driver,
fuel-efficient, automakers, gasoline, drove, and parking. It also includes
more recent automotive collocates of vehicle like electric, plug-in,
connected, charging, and hybrid. Some of the collocates by themselves
have a range of possible uses (owners, operating, safety, accidents,
ownership, emergency, seat, primary, infrastructure), but when examined
in the context almost always indicate an automotive meaning.188
Airplane
doesn‘t appear, though two particular types of aircraft are attested in the
collocates—unmanned aerial vehicles (drones) and spacecraft.189
Similarly, bicycle doesn‘t appear among the collocates of vehicle in
contemporary usage.
We can also examine the collocates of vehicle during the 1950s, the
decade of the Hart / Fuller debate, in the COHA. These collocates are
listed below and can be viewed at the link below.190
motor, space, trucks, moving, wheeled, tax, self-propelled,
passenger, unit, tracked, orbit, test, b.g., launching, highways,
tanks, license, robot, emergency, units, taxes, streets,
equipment, manned, armored, vehicles, fees, vehicle, traveling,
operate, loaded, fuel, commercial, driver, ride, traffic,
designed, weight, speed, cars, carrying, operation, unsafe,
doesn‘t repopulate automatically. See NOW Corpus, BYU,
http://corpus.byu.edu/now/?c=now&q=54596680 (last visited Mar. 8, 2017).
187
The search results are saved at the following link, see NOW CORPUS, BYU,
http://corpus.byu.edu/now/?c=now&q=52902048 (last visited Dec. 21, 2016).
188
To the extent that there is any doubt that any of these collocates suggest the
automotive meaning of vehicle, clicking on any of the listed collocates in the NOW
Corpus interface will display the context in which it appears and confirm the automotive
meaning is intended. For example, it is possible to speak of bicycle traffic or airplane
emissions, but in the context of the word vehicle, the words traffic and emissions are
used in the automotive sense.
189
2 OXFORD ENGLISH DICTIONARY at 920.
190
See COHA, BYU, http://corpus.byu.edu/coha/?c=coha&q=52600298 (last visited
Dec. 21, 2016).
42 Judging Ordinary Meaning 3-23-17
horse-drawn, high-powered, amphibious, administrators,
tactical, registration, delivery
We can see from this data that the meaning of vehicle evolved
significantly from this period, though the automotive use of vehicle still
predominated. The decade of the 1950s is remarkable as the first decade
in which the spacecraft sense of vehicle appears, but also the last decade
in which the horse-drawn collocate of vehicles appears.191
Unmanned no
longer appears, but manned vehicles appear (spacecraft in this case).
Still, the overwhelmingly most common use of vehicle is the automotive
sense, while a number of context specific possible senses are attested.
And, again, none of the top fifty collocates of vehicle include the notion
of airplane and bicycles.
We can also use the COHA to examine the collocates of vehicle from
the period relevant to the McBoyle case. Because the statute at issue in
McBoyle was enacted in 1919,192
and because the COHA only allows us
to search in 10-year increments, it may make sense to include data from
the decades of 1910 through 1930.
Whether or not the use of the word vehicle ―evoke[s] in the common
mind only the picture of vehicles moving on land,‖ as Justice Holmes
suggests, may not be a question that can be addressed with a corpus. But
the collocate data from this period (consistent with the collocate data
above) allow us to draw a similar inference that the automotive use is the
most common use of vehicle, and that the airplane sense remains
unattested.
motor, horse-drawn, wheeled, horses, pedestrians, kinds, expression,
sqdriver, passing, moving, various, horse, automobiles, tax, heavy,
drawn, carry, roadless, rickety, trucks, communication, approaching,
traffic, electric, mental, physical, 3500000, astral, belonging, steam,
transportation, commissioner, rear, total, carrying, propulsion,
propelled, oncoming, carriages, registration, ego, conceivable, tires,
drivers, vehicle, carriers, 45, loaded, halted, manufacturers
The collocates from this period add a few interesting vehicles to our
growing list, including astral vehicle (a reference to the theosophical
notion of a ―supersensible substance‖ that ―accompanies the individual
through life . . . and survives the individual after death‖).193
It should also
191
An additional vehicle is added to our collection with amphibious vehicle, and
tanks makes an appearance again. And two collocates from (the abbreviation b.g. for
background; and robot) both are the result of the inclusion of science fiction screenplays
in the corpus. In both cases the vehicles in question are spacecraft.
192
See McBoyle v. United States, 283 U.S. 25, 25–26 (1931) (quoting National
Motor Vehicle Theft Act , 18 U.S.C. § 408 (1919)).
193
This notion that also explains the presence of ego and mental in the collocates of
vehicle.
43 Judging Ordinary Meaning 3-23-17
be noted that only a few of the collocates in their period occur more than
once, and only four—motor, horse-drawn, wheeled, and horses—occur
ten times or more, with motor occurring twice the number of times as the
other three combined.194
From the collocates of vehicle displayed by the NOW Corpus and the
COHA, we can make the following preliminary observations
(observations that we can later confirm by reviewing Key Word In
Context data.) First, the collocates of vehicle strongly suggest that the
most common use of vehicle is with reference to automobiles. Second,
the absence of airplane and bicycle in the top fifty collocates of vehicle
raises an important question for our frequency continuum. If we accept
that the necessary and sufficient conditions of vehicle are ―[a]ny means of
carriage, conveyance, or transport‖195
or ―a means of carrying or
transporting something,‖196
then there seems little question that both an
airplane and a bicycle are possible readings of vehicle. But if vehicle is
never used to refer to bicycle or airplane in the corpus data, then we may
end up with an even further extension of our frequency continuum from
possible but rare to possible but unattested. Before jumping to the
conclusion that the airplane and bicycle uses of vehicle are entirely
unattested in the corpora or the language at large, however, we should
evaluate the use of vehicle in the concordance data.
b. Vehicle as a Key Word In Context
We can extract concordance data from the NOW Corpus.197
A NOW
search for concordance lines of vehicle will yield an output along these
lines:
194
See COHA, BYU, http://corpus.byu.edu/coha/?c=coha&q=53847214 (last visited
Feb. 8, 2017).
195
19 OXFORD ENGLISH DICTIONARY 480 (2d ed. 1989).
196
WEBSTER‘S THIRD NEW INTERNATIONAL DICTIONARY (1961).
197
The concordance line search in NOW is executed as follows: (1) Select ―KWIC‖
on the NOW Corpus homepage; (2) Enter ―VEHICLE_n‖ in the ―Word/phrase‖ field;
(3) Click on ―Sections‖ and select ―United States‖; (4) Click ―Keyword in Context
(KWIC).‖ See NOW Corpus, BYU, http://corpus.byu.edu/now/?c=now&q=54499369
(last visited Mar. 5, 2017). By selecting ―Options‖ and ―# KWIC,‖ the corpus user can
select the number of randomized concordance lines to be reviewed.
While the search parameters can be saved in a link, the corpus randomizes the
results, and, in the case of the NOW Corpus, the corpus updates with millions of new
words on a nightly basis. So until the BYU corpora develop the ability to save the exact
content of a particular randomized search, it is useful to copy the results of the search
into a spreadsheet.
44 Judging Ordinary Meaning 3-23-17
Table No. 1—KWIC of Vehicle
the driver, Bhaskar Jha, apparently lost control of the vehicle because he was traveling too fast for the wet road conditions .
of the troopers. Parrott says the suspects in the vehicle began showing aggression and shots rang out. Corporal Shane
injury and leaving a child under 12 unsupervised in a motor vehicle but released on a written promise to appear.) Risk
Hybrid electric vehicles use regenerative braking (when the vehicle captures energy that would be otherwise lost from braking ) and
pushed onto the property because of the speed of which these vehicles collide,‖ said Dr. Tom Lawrence , of Clinical Nutrition
,2009. That day the two officers saw a vehicle connected to a domestic violence case in which shots had been
say automakers would be better. Wakefield says autonomous vehicles could erode the image of certain brands more than others . Brands
biogas, and Daimler, which supplies a number of experimental vehicles designed to run on natural gas. The German Federal Ministry of
is that they aren‘t kept on file with the Motor Vehicle Division or any other entity. By contrast, beneficiary
The KWIC output in the NOW Corpus will allow us to select
anywhere from 100 to 1,000 randomized sample uses of vehicle(s) and
display them in their semantic environment. To the extent that the snippet
view above fails to provide sufficient evidence of usage, the corpus
interface allows us to click through to an expanded context in the article
referenced in a given concordance line.
In order to examine the sense distribution of vehicle, we reviewed one
hundred (100) randomized concordance lines of vehicle in the NOW
Corpus. Of those, ninety one (91) were automobiles. There was a single
reference to a bus, and one reference to an ambulance, but in every other
instance, a passenger car was referenced. Of the remaining vehicles, there
was one cargo ship, one jet ski, and an ambiguous reference to a military
ground vehicle of an unknown type. There were three metaphorical uses
of vehicle (e.g., the role of the city as a vehicle for
development, celebration, connectivity and more), and there was a
reference to the military‘s efforts to create a flying Humvee / helicopter
hybrid.198
There were no airplanes, bicycles, tricycles, skateboards, rollerskates,
toy cars or any of what Hart and others have characterized as
penumbral, disputed cases in the NOW Corpus data. To the extent our
notion of ordinary meaning has a frequency component, this data
suggests that automobile is the overwhelmingly most common use of the
word vehicle in the contemporary, modern written American English
represented in the NOW Corpus. The corpus data also suggest that there
are numerous possible (if much less common) uses of vehicle, and that
some seemingly possible meanings are unattested and may not be current.
A similar review of data from the COHA for the period of the 1950s
showed a wider range of vehicles. Still approximately sixty-five percent
198
We are not making this up. See Aerial Reconfigurable Embedded System
WIKIPEDIA,https://en.wikipedia.org/wiki/Aerial_Reconfigurable_Embedded_System
(last visited Mar. 5, 2017).
45 Judging Ordinary Meaning 3-23-17
of the instances of vehicles during this timeframe had reference to
automobiles. Another thirty percent had reference to the space program
or missile defense, while the remaining five percent had reference to
metaphorical uses of vehicle (e.g., a film as a starring vehicle for an
actor). For the period of the 1910s and 1920s, automobiles made up
approximately sixty percent of the instances of vehicle. References to
carriages or horse drawn vehicles were more common, and there were a
significant number of cases where the choice between automobile and
horse-drawn vehicle was not clear. (When a text from 1915 says that 5th
Avenue was crowded with vehicles, it is not clear from context whether
automobiles or carriages or both were intended.) Finally, there were a
number of references to theosophy and the notion of an astral vehicle.
Still, in the COHA data, there were no references to airplanes, bicycles,
tricycles, skateboards, roller-skates, or toy cars in the data for either
period.
c. Searching for Vehicles in the Context of a Park
Hart‘s interpretive puzzle is not simply about vehicles at large, but
vehicles in the park. As we have discussed, with the corpus we can
examine the question of ordinary meaning in the relevant semantic and
syntactic context. We can search for vehicles that collocate with the term
park.199
A review of the concordance data from this search reveals at least one
limitation of the corpus. We can search for specific parts of speech
(nouns, verbs), but not specific senses. That means that our search for
vehicles in the park must begin by eliminating the approximately forty
(40) percent of the concordance lines that have reference to vehicles that
are in park as opposed to in reverse, neutral, or drive. There are
approximately five percent of the concordance lines that refer to
recreational vehicles in recreational vehicle parks. Of the remaining
instances of vehicle, more than fifty (50) percent have specific reference
to automobiles. Bicycles are not attested in this context, nor are airplanes,
skateboards, or roller skates.
Our understanding of a prohibition on vehicles in the park may depend
largely on the physical and spatial characteristics of the park itself. If a
municipal park has no means of ingress or egress for automobiles, then
we might assume that cutting across the grass in a car would be
prohibited. It is not surprising then that where municipal parks are
concerned, the vehicle most likely to show up in the context of park in
199
(1) Select ―Collocates‖ on the NOW Corpus homepage; (2) Enter ―VEHICLE_n‖
in the ―Word/phrase‖ field; (3) Enter an asterisk ―PARK_n‖ in the ―Collocates‖ field;
(4) Select ―Sections‖ and select ―United States‖; (5) Select ―Sort/Limit‖ and set the
―Minimum‖ to ―Mutual Info‖ and ―3‖; (6) Click ―Find collocates‖; and (7) Click
―PARK‖ or ―PARKS.‖
46 Judging Ordinary Meaning 3-23-17
the corpus data (i.e., automobiles) is often not in the park, as in (1) and
(2) below:
(1) juvenile[s] were taken into custody Wednesday, accused of discharging
a BB gun at passing vehicles near Sunset Park.
(2) two males in another vehicle near a park on Toledo‟s west side when
one of those males opened fire
In the very rare circumstance in which there is any actual debate about
vehicles in municipal parks, such debates tend to center around closing
off existing roads through the park, as in (3) below:
(3) A revived plan to remove vehicle traffic from the center of San Diego‟s
Balboa Park was moved forward Monday by the City Council, which
agreed to spend $1 million to complete planning and documentation.
Still, even in the specific park context, where the physical and spatial
features of a park might seem to preclude the entrance of an automobile,
it is the automobile usage of vehicle that predominates.
d. Is Bicycle a Vehicle? Is Airplane a Vehicle?
We can use the KWIC function of the corpus to perform targeted
searches for concordance lines featuring two key terms raised in the Hart
/ Fuller debate—bicycle and airplane.200
Professor William Eskridge has asserted that ―[a] corpus search
reveals that bicycles are commonly considered vehicles—a quantitative
result in striking contrast to the understanding advanced by linguist Bryan
Garner, who joined Justice Scalia in opining that the ordinary meaning of
‗vehicles‘ excludes bicycles.‖201
Professor Eskridge is certainly correct
that there are numerous instances of the co-occurrence of bicycle with
vehicle. Some of these instances establish that the bicycle sense of vehicle
is, at the very least, attested, as in (1) and (2) below:
(1) There are a lot of potholes. It is hard to ride bicycles and other
vehicles.
(2) In New Jersey, bicycles are considered vehicles and must follow the
same laws as motorists.
200
(1) Select ―Collocates‖ on the NOW Corpus homepage; (2) Enter ―VEHICLE_n‖
in the ―Word/phrase‖ field; (3) Enter an asterisk ―BICYCLE_n‖ in the ―Collocates‖
field; (4) Select ―Sections‖ and select ―United States‖; (5) Select ―Sort/Limit‖ and set
the ―Minimum‖ to ―Mutual Info‖ and ―3‖; (6) Click ―Find collocates‖; and (7) Click
―BICYCLES.‖ See NOW Corpus, BYU,
http://corpus.byu.edu/now/?c=now&q=54497865 (last visited Mar. 5, 2017).
201
Eskridge, supra note 4, at 45–46.
47 Judging Ordinary Meaning 3-23-17
Yet other instances show that bicycle is often used in contrast to the
word vehicle, as in (3) and (4) below:
(3) there were 68 collisions between bicycles, pedestrians and vehicles
(4) side mirrors to detect hazards (bicycles, humans, vehicles, pets, etc.)
The conclusion that bicycle is included in the ordinary meaning of
vehicle depends on our view of ordinary meaning. Based on the corpus
data reviewed above, bicycle is certainly a possible sense of vehicle, but
from the stand point of statistical frequency, it is not a common meaning
and certainly not the most common.
With respect to airplane, the answer is simpler. In both the
contemporary NOW Corpus and the COHA (for the relevant periods of
the 1910s, 1920s, and 1950s), we were unable to find a single collocation
or concordance line that reflected the use of vehicle to mean airplane.
Airplane is neither the most common or even a common meaning of
vehicle, and based on its absence from any of our corpus data we might
well ask if it is even a possible sense of vehicle. To the extent that
airplane fits what some lexicographers have regarded as the necessary
and sufficient conditions for inclusion in the class of vehicles (i.e.,
anything that is a ―means of carriage, conveyance, or transport‖) all that
can be said of airplane is that it may be a possible meaning of vehicle,
but it is unattested in the corpus data.
2. Muscarello and Carries a Firearm
The Muscarello question—of the meaning of carry—is likewise
susceptible to measurement. We can assess the relative frequency and of
the personally bear sense and the transport sense using corpus analysis.
a. The Collocates of Carry
We can view collocation data for carry in the NOW corpus.202
The
fifty most common collocates of carry in the NOW Corpus are listed as
follows:
out, yards, concealed, weight, gun, attacks, weapons, guns, sentence,
weapon, exchange, maximum, margin, passengers, heavy, penalty, bag,
signs, opinions, firearm,203
express, burden, permit, thoughtful, load,
bags, plane, firearms, virus, tradition, flag, capable, torch, handgun,
202
Follow the same steps set forth in note __ above, substituting ―CARRY_v‖ for
―VEHICLE_n.‖ See NOW Corpus, BYU,
http://corpus.byu.edu/now/?c=now&q=54015027 (last visited Feb. 15, 2017).
203
Collocates are not lemmatized in the BYU corpora, so the singular and plural
form of a given collocate care counted separately. In this case, that means that firearm
and firearms both make the list.
48 Judging Ordinary Meaning 3-23-17
cargo, openly, permits, duties, pipeline, mosquitoes, touchdowns, ships,
executions, loads, trucks, felony, tasks, handguns, experiments, knife
These collocates suggests that a number of uses of carry don‘t fit
neatly into the syntactic structure and the semantic relationships we have
previously identified. There are instances in which an inanimate object
serves as carry‘s subject (planes carrying passengers, trucks carrying
loads, ships carrying cargo). And there are a number metaphorical uses
of carry (felonies carrying certain penalties, people carrying opinions).
There‘s carrying out of attacks and executions, and there is a sporting
references to carry as well (carrying so many yards for so many
touchdowns).204
Yet ―[a]t issue here is not ‗carries‘ at large, but ‗carries a firearm.‘‖205
And a list of collocates simply tending to show that there are a variety of
small, inanimate, concrete objects (including weapons) that can be
carried on your person or in your car doesn‘t get us much closer to
determining which of these senses of carry is the most frequent.206
But as
we will see, this search reveals common collocates of carry that have the
similar semantic features to firearm (i.e., pistol, handgun, rifle, gun) that
will help us better evaluate the contexts in which carry a firearm occurs.
b. Carry as a KWIC
The KWIC data will give us a clearer picture of the use of carry.207
The NOW corpus will give us a randomized sample of concordance lines
featuring carry.208
And we can review these concordance lines to
determine both the range of possible meanings of carry and the
204
We can see similar results in the COHA, using the same instructions in note __
above, except that when we click on ―Sections‖ we select ―1960.‖ The results of this
search in the COHA can be viewed at the link below. See COHA, BYU,
http://corpus.byu.edu/coha/?c=coha&q=54015512 (last visited Feb. 15, 2017). As we
have noted, the COHA, when divided by decade results in a functionally smaller corpus
for that decade. As a consequence, it is more susceptible to being offset by unusual
collocations.
205
Mucarello, 524 U.S. at 143 (Ginsburg, J., dissenting).
206
As we will see, in the case of carry, the collocates do help us in identifying words
with similar semantic features as firearm—gun(s), weapon(s), handgun(s), rifle(s),
pistol(s)—and which would serve similar functional roles in a sentence. This will help
us locate relevant concordancing data, but does not answer the question of which sense
of carry is the most common.
207
One way to examine carry in context is simply to enter a search similar to that in
note __ above, but substituting ―CARRY_v‖ for ―VEHICLE_n.‖
208
As we have already seen, carry has a transitive argument structure and in the
relevant context of Section 924(c)(1) carry has a human subject and a non-human,
inanimate, weapon object. If we are going to take context into consideration, we should
be looking for uses of carry that reflect the same or similar syntactic structure and
semantic relationship.
49 Judging Ordinary Meaning 3-23-17
comparative frequency of those meanings. We can also locate (and
determine the comparative frequency) of instances of carry with the same
syntactic and semantic features as Section 924(c)(1).
Yet we might be able to eliminate a lot of irrelevant uses of carry by
searching instead for carry within a few words of firearm.209
A search for
concordance lines containing these terms will require coding. Because
every interpretative question is different, the process of coding
concordance lines will vary with each task. After examining only a few
concordance lines, a problem emerges: A significant majority of the
instances of carry in the context of firearm in the NOW Corpus refer
back to the statutory prohibition in Section 924(c)(i) or similar statutes. In
order to ensure that we have sufficient data from outside of a legal
context, we have also examined instances of carry in the context of a
number of common synonyms of firearm that were listed among the most
common collocates of carry—gun(s), pistol(s), handgun(s), and
rifles(s).These synonyms share the same semantic features with firearm,
but less commonly appear in statutory prohibitions against carrying a
firearm.
Our search parameters will eliminate a number of irrelevant uses of
carry.210
All that is left is to review the concordance lines and determine
in how many instances carry a firearm has reference to carry on one‟s
person or carrying in a car. Here, the physical and spatial context can be
helpful, as when the physical location in (1), (2), and (3) below:
(1) Dressed in body armor and carrying two handguns, [the suspect] tried
to flee out a back door . . . .
(2) adults with the proper permits no longer need to hide the handguns
they carry in their shoulder or belt holsters . . . .
(3) the crowd was sedate and well behaved with those carrying guns
checking their ammunition at the door.
A number of concordance lines were unclear and a number of them, as
noted, had reference to statutory provisions similar to Section 924(c)(1).
All told, we reviewed two hundred seventy-one (271) concordance
lines from the NOW Corpus in which carry co-occurred with firearm(s),
gun(s), pistol(s), handgun(s), and rifles(s). Of these instances of carry we
found that one hundred and four (104) instances of that indicated a sense
209
Such a search can be executed as follows: (1) Select ―Collocates‖ on the NOW
Corpus homepage; (2) Enter ―CARRY_n‖ in the ―Word/phrase‖ field; (3) Enter
―FIREARM_n‖ in the ―Collocates‖ field; (4) Click on ―Sections‖ and select ―United
States‖; (5) Select ―Sort/Limit‖ and set the ―Minimum‖ to ―Mutual Info‖ and ―3‖; and
(6) Click ―Find collocates.‖ See NOW Corpus, BYU,
http://corpus.byu.edu/now/?c=now&q=54477447 (last visited Feb. 15, 2017).
210
The metaphorical sense (carry a tune) or senses where the subject / agent is an
inanimate objection (the ship carries cargo) or the sporting sense (carried the football
nine yards).
50 Judging Ordinary Meaning 3-23-17
of carry a firearm on one‟s person, while only five (5) instances
suggested a carry a firearm in a car sense. The remaining senses were
either unclear or senses of carry or unrelated to the question at hand. As
would be expected, much less data was available for carry in the COHA.
We found twenty-eight (28) concordance lines from the COHA, in which
carry co-occurred with firearm(s), gun(s), pistol(s), handgun(s), and
rifles(s). Of these instances of carry we found that eighteen (18) instances
of carry on one‟s person, two (2) instances of carry in a car. The
remaining instances were either unclear or reflected a different sense of
carry.
To the extent that we view the question of ordinary meaning has
involving statistical frequency, the analysis above tells us that carry on
one‟s person is overwhelmingly the most common use while the carry in
a car is a possible, but far less common use.
3. Taniguchi and the Meaning of Interpreter
We can also measure the relative frequency of the written translator
and oral translator senses of interpreter. And we can do so using
collocation and concordance analysis.
a. The Collocates of Interpreter
The fifty most common collocates of interpreter in the NOW Corpus
are as follows:211
an, through, language, sign, spanish, via, speaking, afghan,
translators, iraqi, certified, served, english, qualified, translator, army,
basic, deaf, spoke, moderator, sign-language, asl, costumed,
interpreter, translate, full-time, dream, trained, soldiers, yun,
interpreters, arabic, translated, translation, freelance, certification,
courts, maladies, requests, spanish-language, communicate, cespedes,
languages, troops, carlotto, simultaneous, somali, listened,
proceedings, employed
A number of the collocates tend to support the Taniguichi majority‘s
position that interpreter most commonly has reference to an interpreter
of spoken language. These include speaking, spoke, listen. A number of
the collocates have reference to battlefield interpreters (Afghan, Iraqi),
where context would suggest their role is primarily as spoken
interpreters. The collocates an and through both come from the very
common phrase that a public figure is speaking through an interpreter.
These collocates stand in contrast to the collocates of translator in the
NOW Corpus, which make a number of references to a writing and
211
See NOW Corpus, BYU, http://corpus.byu.edu/now/?c=now&q=54018483 (last
visited Feb. 15, 2017).
51 Judging Ordinary Meaning 3-23-17
publishing context, including bible, writer, poet, editor, literary,
publisher, journalist, Borders, and even Wycliffe.212
In addition, the
collocates from the decade of the 1970s, when the Court Interpreters Act
was passed, suggest a similar conclusion.213
b. Interpreter as a KWIC
With respect to Key Words in Context, we reviewed 188 concordance
lines from the NOW Corpus, in which interpreter occurred.214
A number
of instances, interpreter had reference to artist expression or the
interpretation of works of art (a noted interpreter of modern music).
Another common sense has reference to the interpretation of documents
the primary language (interpreters of the Constitution). There were
numerous instances in both corpora of cases of the spoken language
conversion from a primary language to a second language notion of
interpreter, as in (1) and (2) below:
(1) civil rights violations for not providing professional interpreters
for patients who do not feel comfortable speaking English
(2) Motto was speaking in French, through a volunteer interpreter
In addition, there were numerous transcripts of spoken interviews
from news sites with the annotation ―through interpreter,‖ referencing a
spoken interview facilitated by an interpreter. And there was one instance
of an interpreter translating a foreign language document into spoken
English, in (3) below:
(3) In 1992, during a top-level meeting in Moscow, Russia finally released
the cockpit voice recorder transcript. It was 10 p.m. in a dimly lit
meeting room of the Presidential Hotel when an interpreter for the U.S.
ambassador translated the Russian transcript into English for
Ephraimson-Abt and other delegates.
212
See NOW Corpus, BYU, http://corpus.byu.edu/now/?c=now&q=54609539 (last
visited Mar. 9, 2017).
213
Interpreter has very few frequent collocates from during the decade the 1970.
The two most common collocates of interpreter from this period are an and through,
function words that mutual information scoring typically eliminates if other options are
available. A review of the concordance lines associated with these collocates reveal their
origin in the extremely common phrase speaking through an interpreter, or related
phrases. See COHA, BYU, http://corpus.byu.edu/coha/?c=coha&q=54495283 (last
visited Mar. 5, 2017).
214
The statute at issue in Taniguchi states: ―A judge or clerk of any court of the
United States may tax as costs the following: (6) . . . compensation of interpreters . . . .‖
28 U.S.C. § 1920. The noun phrase compensation of interpreters is part of a standalone
enumeration that has an attenuated relationship to the argument structure of the verb to
tax. What we can say about the relevant context for interpreter is that we are looking for
individuals who are capable of decoding a foreign language into a native one. The
operative variable being whether the language at issue is spoken or written.
52 Judging Ordinary Meaning 3-23-17
Absent from all of these concordance lines was a single instance of
anyone referred as an interpreter performing a text-to-text translation
from a foreign language into a primary language like English. To the
extent that our notion of ordinary meaning has a frequency component,
we can say from this data that the text-to-text translation sense of
interpreter is neither the most common nor even a common use of
interpreter. We might question whether it is even a possible sense of
interpreter as the text-to-text translator sense of interpreter is entirely
unattested in our data.
4. Costello and Harboring an Alien
The interpretive issue in Costello bears some similarity to the question
at issue in Muscarello. In both cases the question turns on the meaning of
a transitive verb and its relation to its object, though in the case of
harbor, our object has the semantic features of human, animate, etc. We
would therefore look to the corpus data to tell us which senses of harbor
are the most frequent, common or possible senses of harbor and we
would look to the data to help us make informed decisions about sense
division. We will look at the use of harbor in contemporary English,
using the NOW Corpus and in the decade of 1910 through 1919, the
period during which the relevant statute was enacted.
a. Collocation of Harbor
With respect to the collocation data, it is immediately apparent from a
review of the collocates of harbor that the overwhelmingly most common
use of the term harbor has reference to harboring feelings.
bacteria, feelings, resentment, doubts, terrorists, species, secret,
mariners, views, ambitions, immigrants, fugitive, planets, illusions,
hatred, dreams, cells, mutations, ocean, hopes, animosity, virus,
secrets, anger, grudge, suspicions, fantasies, planet, fears, sentiments,
desire, pathogens, galaxy, viruses, suspicion, persons, thoughts,
fugitives, germs, mutation, tumors, aliens, moon, bias, genes, gene,
hole, diversity, grudges, resentments215
Such uses of harbor do not match the semantic features in the relevant
statute. We are looking for objects of harbor that are human, animate,
215
See NOW Corpus, BYU, http://corpus.byu.edu/now/?c=now&q=54496834 (last
visited Mar. 5, 2017). This search examines only the nominal (noun) collocates of
harbor. Harbor is a low frequency verb and as such instances of harbor are rare in the
COHA for the period of 1910 through 1919. Even expanding the search through the
1920s reveals only a sparse number of collocates. While some of these are relevant to
our present inquiry (alien, refugee), no other relevant collocate appears more than once
in the COHA. See COHA, BYU, http://corpus.byu.edu/coha/?c=coha&q=54496926 (last
visited Mar. 5, 2017).
53 Judging Ordinary Meaning 3-23-17
concrete, etc. With that in mind, we tailored our searches to those
nominal objects of harbor reflected in the collocates listed above that had
these same semantic features—fugitives, terrorists, criminals, aliens,
refugees.
b. Harbor as a KWIC
In the NOW Corpus, we examined one hundred and forty concordance
lines in which harbor occurred in the same environment as fugitives,
terrorists, criminals, aliens, refugees. Of these, twenty three (23)
instances of harbor had explicit reference to concealment while thirty
two (32) had reference to shelter. In additional eighty three (83)
instances, the distinction could not be determined by context. There were
also three (3) instances of unrelated senses of harbor. In the COHA, there
were only three (3) clear cut cases of the shelter sense. The remaining
five (5) instances of harbor could not be determined by context.
This data raises more questions than answers. With respect to
frequency, we would be hard pressed to say that either the shelter
meaning or the conceal meaning of harbor are the most common. We
might say that both are common meanings, and they are both certainly
possible and attested meanings. But where more than half of the instances
of harbor are unclear as to whether they include shelter or concealment
or both, it is hard to state from the standpoint of frequency what the
ordinary meaning actually is.
C. Conclusions
Above are the data. But what to make of them? Do corpus data yield
means of measuring ordinary meaning? We think the answer is a
resounding yes—with a few caveats. Certainly the answer is yes by
comparison to existing means of measurement. If ordinary meaning is an
empirical construct—and we think it is—then corpus analysis is superior
to an intuitive guess (or, worse, crediting a dictionary or a word‘s
etymology).
And we also think that corpus data are well-suited to give reliable
answers to the question of ordinary meaning. To support this conclusion
(as applied to the problems analyzed throughout the article) we begin
with a more careful synthesis of the theory of ordinary meaning
introduced above. And we then offer some conclusions about what the
corpus data tell us about the ordinary meaning of vehicle, carry a firearm,
interpreter, and harbor.
1. Corpus Data and Meaning
54 Judging Ordinary Meaning 3-23-17
Corpus analysis may be addressed to the range of senses of ordinary
meaning identified above—to semantic meaning, to intended meaning,
and to public meaning. Through data from the COHA, the NOW Corpus,
or other corpora we can assess the relative frequency of competing senses
of a statutory term or phrase. And from frequency and collocation data
we can draw inferences about the semantic meaning of the language of
the law and even about intended or public meaning.
a. Semantic meaning
Frequency data from a corpus may give us useful evidence of the
ordinariness or plainness of a given semantic sense of a particular term.
In constructing a corpus search we can control for relevant components
of semantic meaning—of elements of syntax like argument structure and
components of semantics like number, animacy, humanness, and
concreteness. We can do that, as above, by constructing a controlled
search for uses of the verb carry (a) as a transitive verb, (b) by a human
subject, (c) acting on an inanimate, concrete object, and (d) that is a
weapon or form of firearm. And the results of our search will show us
that carry in this context almost always (90% of the time in the above
data set) has reference to the bear on your person sense of the verb.
That provides an empirical ground for the conclusion that the ordinary
semantic meaning of carry in this setting aligns with the view espoused
in Justice Ginsburg‘s dissent. If a database of naturally occurring
language tells us that our references to carrying a firearm in this linguistic
setting are most always references to bearing on your person and very
rarely to transporting in a vehicle, then that is an indication that the
ordinary semantic sense of carry in this setting is the former sense.
Certainly this is a better way to solve the Muscarello puzzle than by
resort to sense ranking or etymology. And if what we are looking for is
semantic meaning—as the law sometimes says we are—then this may be
an acceptable answer to the question.
That said, we see some possible limitations on the strength of the
inference to be drawn from this sort of data.
Sense division. One possible limitation stems from the vagaries of
word sense division. Sense division is subjective.216
Linguists, as noted
above, have no agreed-upon formula for distinguishing senses of a
word.217
They concede that distinctions among senses may be ―more of a
216
Nikola Dobric, Word Sense Disambiguation Using ID Tags—Identifying Meaning
in Polysemous Words in English, in PROCEEDINGS OF THE 29TH INTERNATIONAL
CONFERENCE ON LEXIS AND GRAMMAR/LGC 97, 97 (Dusko Vitas & Cvetana Krstev
eds., 2010) (explaining that polysemy—multiple word meaning—is ―[o]ne of the
persisting issues in modern lexicography‖).
217
No one is quite sure where to draw the line—research ―show[s] that different
polysemy criteria (i.e., criteria that may be invoked to establish that a particular
55 Judging Ordinary Meaning 3-23-17
descriptive device rather than a claim about psycholinguistic reality.‖218
This seems particularly true as regards closely related or ―finegrained‖
sense distinctions. The space between some senses will be
sufficient to justify a strong inference from clear corpus data. Consider
the above-cited example of the use of the term nail in Reading Law:
―‗Nail in a regulation governing a beauty salon has a different meaning
from nail in a municipal building code.‘‖219
Surely we could confirm that
using corpus data. We could show that the term nail as used in the
context of a beauty salon is almost always by reference to a fingernail or
toenail. And we would likely feel confident concluding that such data
supports the conclusion that the ordinary understanding of nail in this
semantic setting is not a piece of metal used to attach pieces of wood.
But what about more closely related senses? The two competing
notions of carry in Muscarello are related. Both get at the idea of
transport; the difference concerns the mechanism—on one‘s person or in
a vehicle. And it accordingly seems hard to know whether this difference
reflects ―psycholinguistic reality.‖ The bear personally sense seems to be
the notion of carry that we speak of most always. And for that reason it
may also be the sense we think of most often. But if pressed we might
well concede that the transport by vehicle sense is also encompassed
within the notion of carrying a firearm. Psycholinguistic reality could be
such that most ordinary people first think of the bear personally sense but
on reflection agree that the transport sense is included.
A variation on this problem has been highlighted by Professors Larry
Solan and Tammy Gales. They have observed that corpus data along
these lines may reflect only the fact that a given sense of a certain term is
a more factually common iteration of that term in the real world.220
And
interpretation of a lexical item constitutes a separate sense rather than just being a case
of vagueness or generality) may be mutually contradictory, or may each yield different
results in different contexts.‖ DIRK GEERAERTS, THEORIES OF LEXICAL SEMANTICS 196
(2009). And there is no agreed-upon taxonomy of polysemy; some linguists speak of
senses and sub-senses, Glynn, supra note __, at 17, others of more or less prototypical
exemplars of senses, see, e.g., Dagmar Divjak & Antti Arppe, Extracting Prototypes
from Exemplars: What Can Corpus Data Tell Us about Concept Representation?, 24
COGNITIVE LINGUISTICS 221 (2013), and others of hyponomy and hypernomy in
polysemy, see Glynn, supra note __, at 10.
218
Stefan Th. Gries, Polysemy, in HANDBOOK OF COGNITIVE LINGUISTICS 482–83
(Ewa Dabrowska & Dagmar S. Divjak eds., 2015). The ―problem of an apparent lack of
decisive criteria for defining word senses and clearly discriminating between them has
always been a burning issue of lexical semantics to the point that it fundamentally
questions the possibility to provide a clear account of polysemy.‖ Dobric, supra note __,
at 77.
219
SCALIA & GARNER, supra note __, at 20.
220
Lawrence M. Solan & Tammy Gales, Corpus Linguistics as a Tool in Legal
Interpretation, Paper presented at Brigham Young University Law School‘s Law and
Corpus Linguistic Conference (Feb. 3, 2017).
56 Judging Ordinary Meaning 3-23-17
if that is true, there may be reason to doubt the probity of the data in
establishing the semantic meaning221
perceived by lawmakers or the
public.
These are important concerns. And anyone turning to corpus analysis
would do well to consider these limitations before jumping too quickly to
an inference about ordinary meaning. But we do not view the sensedivision
problems noted here to be fatal to the probity of corpus linguist
analysis (even for related senses of a statutory term). Instead we propose
a range of responses to this concern.
First, the Solan-Gales point seems overstated. Let‘s apply it to the
carry data. It may be likely, as Solan and Gales might suggest, that the
corpus data we found is indicative of the fact that most iterations of
carrying a firearm in the real world involve personally bearing it. Yet we
don‘t see that as depriving the data of probative value. If most iterations
of firearm carrying involve personally bearing then that sense of carrying
seems likely to be the one that first comes to mind when we think of this
term. That top-of-mind sense, as noted, may not exhaust the breadth of
human perception of this term. If pressed some people might concede that
the term encompasses the transport sense too. (We could conceivably test
that through a psycholinguistic experiment; more on that later.)
That raises the question of whether to credit only the top-of-mind
sense or a possibly broader, ―reflective‖ sense as ordinary. But this is not
a deficiency in corpus data—or even in linguistic theory. It is a question
for law—―we have to decide which meaning, produced by which theory
of meaning, we ought to pick.‖222
And we think the answers to these
questions are dictated in part by the rationale that drives us to consider
ordinary meaning. A concern for fair notice and protection of reliance
interests may well direct us to stop at the top-of-mind sense of a statutory
term. If the personally bear sense of carry is the first one that comes to
mind, then that may be the sense that the public will have in mind upon
reading the terms of a statute. And if we are interested in protecting
reliance interests and avoiding unfair surprise, we may want to stop short
of including the broader transport sense that the public might concede to
be covered upon reflection. Meaning arrived at upon reflection is
meaning that could catch the public unawares. So there may be reason to
stick to top-of-mind meaning and not encompass reflective meaning if we
are concerned about fair notice.
The legal authority rationale could at least arguably lead us in a
different direction. An ordinary legislator could expect to have his
221
The point is not to suggest that mere semantic meaning is the right framing.
Above we conceded that the pragmatic context of relevance to so-called intended or
public meaning is the correct focus. But for now we are speaking only of semantic
meaning. We add the wrinkle of pragmatic context below.
222
Baude & Sachs, supra note__, at 1089–90.
57 Judging Ordinary Meaning 3-23-17
legislative enactments sweep more broadly than the first sense to come to
mind. (To answer that question we would need to proceed to a sense of
intended meaning; more on that below.) And where that holds, it may
make sense to treat the broader sense of carry that occurs to us on
reflection to count as ordinary semantic meaning.
Inconclusive data. Above we were considering data at the right end of
the frequency continuum—an indication that one of two senses is clearly
the most frequent, or even almost exclusive. But what if the data are less
clear? What if the data suggest that each of two senses is about equally
possible? Or that one is a bit more frequent but not clearly so?
Sometimes an indication that both senses of a term are relatively
frequent will be telling. If two senses are closely related and both appear
relatively equally in the data, that may tell us that both are about equally
likely to be called to mind. And in that event it may be difficult to
exclude either as extraordinary.
The salience of inconclusive data may also depend on the nature of the
question presented. We have been speaking here of a raw question of
ordinariness. But sometimes the question is one of plainness—of whether
the semantic meaning of a word or phrase is sufficiently clear to close the
door on legislative history (or on a substantive canon or agency
deference).223
And where that is the question inconclusive data may be
quite conclusive—it may tell us that there is ambiguity sufficient to
proceed past the threshold ―standard picture.‖ So corpus data can bring
rigor to this range of questions too; instead of guessing about plainness
we can summon some data.
In some cases the data may just be too mixed to yield any helpful
answers. Even then that does not require us to abandon the standard
picture. We could look to another empirical tool for assessing semantic
meaning (see the discussion below on psycholinguistic experiments).
Barring that kind of help, we can fall back on a principle of interpretation
framed by something other than a view of the standard picture—as in a
rule of interpretation that has to do with ―legal content‖ of the law, like
the rule of lenity (more on that below too). But we see no reason to fall
back too quickly. Again, the law‘s commitment to the standard picture is
for good reason. We think the courts should try their best to find real
answers to linguistic questions before falling back on fake ones.
b. Intended Meaning
Semantic meaning is a good starting point for getting at intended
meaning. But linguistic theory tells us that a lawmaker‘s intent can
properly be inferred only by also considering pragmatic context. So if we
are interested in vindicating not just word meaning but perceived
223
See supra at __.
58 Judging Ordinary Meaning 3-23-17
intended meaning of the lawmaker we must proceed beyond semantics.
We will have to consider pragmatics.
Classic illustrations are Posner‘s ―Keep off the grass‖ sign at a park
and Fallon‘s extension of the ―no vehicles‖ rule. Posner rightly says that
the park sign would ―not properly be interpreted to forbid the grounds
crew to cut the grass.‖224
And Fallon understandably asserts that the
lawmaker adopting the ―no vehicles in the park‖ rule would ―reasonably‖
be understood to intend to allow the ―gatekeeper‖ at the park to allow in
an ambulance in the event of an emergency.225
We can think of these
conclusions as departures from the standard picture—as the vindication
of consequentialist concerns overriding the communicative content of the
law.226
But we think they can easily fit into the standard picture. The pure
semantic meaning of ―Keep off the grass‖ would apply to the
groundskeeper, just as the ―no vehicles‖ prohibition would encompass an
ambulance (which on any semantic understanding would count as a
vehicle). But the pragmatic context of these rules implies clear limitations
on this semantic meaning that leaves room for the groundskeeper and the
ambulance.
The author of the ―Keep off the grass‖ rule in Posner‘s example is the
property owner. And that owner undoubtedly also hired the
groundskeeper. That pragmatic context gives us a clear indication of
intended meaning: If the property owner hired the groundskeeper to cut
the grass he obviously intended for the groundskeeper to cut the grass
(and not keep off it).
The ambulance hypothetical is a little harder. But here again we don‘t
think we need pure consequentialism to justify the exception to the
semantic rule. Again the rulemaker is the property owner. And any
property owner can be expected quite clearly to intend an exception for
ambulances dealing with emergencies. Whether out of altruism or
concern for premises liability, the park owner in Fallon‘s hypothetical
can be presumed to intend to allow emergency vehicles in to rescue fallen
park-goers.
Fallon says that the rulemaker likely was ―thinking about automobiles
driven for recreational purposes, not ambulances.‖227
And he accordingly
says that ambulances responding to emergencies are not a ―specifically
expected application.‖228
Fair enough. But that doesn‘t mean that ―moral
224
RICHARD A. POSNER, REFLECTIONS ON JUDGING 180 (2013)..
225
Fallon, supra note __, at 1260–61.
226
See POSNER, HOW JUDGES THINK, supra note __, at 40 (describing ―pragmatism‖
as a process of ―basing judgments (legal or otherwise) on consequences, rather than on
dedication from premises in the manner of syllogism‖).
227
Fallon, supra note __, at 1261.
228
Id. at 1262.
59 Judging Ordinary Meaning 3-23-17
reasonableness‖ is the only way to make room for the ambulance
exception, as Fallon suggests.229
In ordinary communication we often
make inferences about a speaker‘s intent. And where (as here) such intent
is clear, we naturally credit it as a matter of normal conversational
meaning.
So in our view this is still ordinary meaning. (It is just inferred
intended meaning that takes pragmatic context into account.) We are still
on the standard picture. The question for present purposes is whether we
can measure this kind of meaning.
In our view the answer is that it may be difficult but not impossible. If
we had an unlimited corpus with a sufficiently targeted register (genre),
we could find ways to glean data on the sort of pragmatic, intended
meaning described above. If we had a database of park owners talking
about groundskeepers we would undoubtedly find examples of statements
indicating their intent to allow them to tread on the grass. Presumably we
could find something similar for ambulances. If park owners talk
routinely about ambulances responding to emergencies in a manner
indicating approval (and never disapproval), we could infer park owner
intent to allow such access. Perhaps the relevant corpus would have to be
even more targeted to be probative. To find an intended exception to the
―Keep of the grass‖ or the ―no vehicles‖ rule we would need to be
looking at park owners who have these rules in place. But the point is that
corpus analysis is at least theoretically capable of getting at pragmatic,
intended meaning.
In many cases such meaning may be beyond the reach of most corpora
that are currently available. As to Muscarello, for example, it might be
impossible to find a corpus sufficient to get at the pragmatic components
of the intended meaning of a sentencing enhancement for carrying a
firearm in connection with a drug crime. Theoretically, the right corpus
would reflect dialogue among the 535 members of the Congress that
voted on the sentencing enhancement in 18 U.S.C. § 924(c)(1). If we had
such a corpus, and if it recorded extensive discussion among the members
of Congress about the kind of gun carrying they were talking about when
they enacted this statute, we might be able to get data of relevance to the
intended meaning of this provision. Perhaps it would reveal only
examples of personal bearing of firearms and never of transporting in a
vehicle. If so that might tell us that the intended meaning is limited to the
former.
Even then, the might qualifier is necessary. The limitation here is the
same one presented above as to semantic meaning—as to whether a
predominance of examples of uses of one sense of carry may indicate
only that this is the first sense to come to mind, and whether a broader
229
Id.
60 Judging Ordinary Meaning 3-23-17
sense that might occur to a lawmaker on reflection should count as
ordinary.230
How might a judge answer this question? For some such questions the
question may be framed within the standard picture. On questions of
likely intended meaning of rules like the ―Keep of the grass‖ sign or the
prohibition on ―vehicles‖ we think judges are in a good position to assess
likely intended meaning (even absent hard data about actual usage). We
say that because we think the relevant pragmatic context of these rules is
likely to be apparent in the cited circumstances. It seems difficult to think
of a legislative ―compromise‖ that would call into question the inference
of uniform legislative intent to allow groundskeepers on the grass or
ambulances in the park.231
And if so, it seems safe to conclude that the
intended communicative content of these rules would sustain the noted
exceptions (for groundskeepers and ambulances).
That will not always be so, however. Muscarello may be a good
example. If we lack confidence in the corpus data on carry we may be
left to make an inference about likely legislative intent. And here that
seems hard. As the majority and dissenting opinions in that case
demonstrate, it is easy to contemplate legislative intent running in either
of two directions—to call for a sentencing enhancement (a) whenever a
gun is available to the defendant in a drug deal, since an available gun
may always be used in a harmful way if it is available,232
or (b) only if
the gun is being carried on the defendant‘s person, since that kind of
availability is even riskier.233
230
See supra Part IV.B.1.a.
231
See Frank H. Easterbrook, Text, History, and Structure in Statutory
Interpretation, 17 HARV. J.L. & PUB. POL‘Y 61, 68 (1994) (noting that if a particular
outcome under a statute seems ―unprincipled,‖ it may be the ―way of compromise‖ in
the legislative process; noting that ―[l]aw is a vector rather than an arrow,‖ ―[e]specially
when you see the hand of interest groups‖); MANNING & STEPHENSON, supra note __, at
54 (stating that ―laws will be messy, uneven, and ill-fitting with their apparent purposes
not because Congress is short-sighted or imprecise, but rather because legislation entails
compromise, and compromise is untidy by nature‖); SCALIA & GARNER, supra note __,
at 39 (―Not only is legal drafting sometimes imperfect, but often the imperfection is the
consequence of a compromise that is not the function of the courts to upset—or to make
impossible for the future by disregarding the words adopted.‖).
232
Muscarello v. United States, 524 U.S. 125 at 133 (1998) (―How persuasive is a
punishment that is without effect until a drug dealer who has brought his gun to a sale
(indeed has it available for use) actually takes it from the trunk (or unlocks the glove
compartment) of his car? It is difficult to say that, considered as a class, those who
prepare, say, to sell drugs by placing guns in their cars are less dangerous, or less
deserving of punishment, than those who carry handguns on their person.‖).
233
Id. at 145 (Ginsburg, J., dissenting) (―It is reasonable to comprehend Congress as
having provided mandatory minimums for the most life-jeopardizing gun-connection
cases (guns in or at the defendant's hand when committing an offense), leaving other,
less imminently threatening, situations for the more flexible guidelines regime.‖).
61 Judging Ordinary Meaning 3-23-17
So Muscarello is an example of a case in which pragmatic judgments
about likely legislative intention are likely to be difficult. If we lack a
reliable basis to decide the case on either semantic or likely intended
meaning, we may be left to resolve it on other grounds. Here we could
simply turn to the law of interpretation—giving the law legal content that
doesn‘t pretend to be based on communicative content (because we
haven‘t been able to find it). We could do so, for example, on the basis of
a substantive canon like the rule of lenity. Reliance on that canon may
make sense doctrinally, as lenity kicks in in the face of genuine
ambiguity about statutory meaning. Such a move, moreover, would be
more open and transparent than a false assertion about communicative
content. And for that reason we would favor it even though it might not
obviously vindicate the principles motivating the law‘s threshold
devotion to ordinary meaning.
c. Public meaning
Much of what we said above about intended meaning applies also to
public meaning. Ideally the two senses of meaning overlap completely.
They overlap, for example, if the groundskeeper or ambulance driver
accurately anticipates the intent of the park owner in the above
hypotheticals, or if would-be drug dealers perfectly assess the sort of
firearm carrying that Congress has in mind in enacting 18 U.S.C. §
924(c)(1). If that happens then ―what the speaker means and what the
hearers take the speakers to mean‖ are the same thing.234
And in that
event the pragmatic meaning question implicates many of the same issues
highlighted above.
Yet the overlap will not always be complete. Sometimes the public
will interpret statutory language in pragmatic context differently than a
legislative body would interpret it. And that implicates the ―whose
meaning‖ and ―speech community‖ questions.
The speech community question, as we have noted, has implications
for the selection of a relevant corpus. If we are trying to measure intended
meaning we might want to gather data from a corpus of a community of
speakers who look demographically like Congress. Yet if we are
interested in public meaning we would want to turn to a broader corpus.
Data from a general, balanced corpus could tell us something about
the way the way the human mind conceptualizes the notion of carrying a
firearm. But that might not be the right question to ask. We might be
missing an important element of pragmatic context if we ask only about
234
Soames, supra note __, at 598; see also Doerfler, supra note __, at 983–84
(getting at a similar conclusion through his premise of a ―conversation‖ model of
meaning and ―fictionalized‖ legislative intent).
62 Judging Ordinary Meaning 3-23-17
―carrying a firearm‖ in the abstract. Another relevant element of such
context may be the legal nature of the language of this law.
The human mind may react differently to a criminal prohibition—a
law imposing harsh consequences like a sentencing enhancement—than it
would to a mere statement of description. Thus, we may form one
understanding when listening to a descriptive narrative of a person
carrying a firearm in connection with a drug crime and another when
warned that the punishment for a drug crime could be significantly
enhanced if we carry a firearm in that circumstance.
That sort of context seems impossible to suss out with corpus analysis.
To get at this kind of nuance we may only be left with psycholinguistic
study. (More on that below.)
What if our sense of public meaning differs from our sense of intended
meaning? If that happens we would need to decide which data set to rely
on. Again that is a problem for legal theory—and essentially for a choice
of which of two sets of justifications for the ―standard picture‖ we seek to
vindicate. In the Muscarello setting the answer may well be the fair
notice rationale. The law of interpretation may already have given that
answer—in the rule of lenity. In criminal cases the rule of lenity suggests
that the notice rationale predominates. It indicates that a criminal
defendant is entitled to the benefit of the doubt in cases of ambiguity as to
the law‘s communicative content.
The question may be harder to answer in civil cases. But again that is a
problem for legal theory. As above we will simply say for now that
transparent answers are better than opaque ones. Further thinking on this
problem is needed. Yet surely we will be better off with an open,
transparent discussion about whether (and when) to given primacy to
intended meaning and when to credit public meaning. Once we speak
more carefully about the meaning we are looking for and proceed more
reliably in trying to measure it we can have a better dialogue about these
difficult questions of legal theory.
2. Corpus Data and Our Test Cases
Now we can assess the above data in light of the principles of
semantic, intended, and public meaning developed above. Here we offer
some data-backed conclusions about the ordinary sense of vehicles in the
park, carrying a firearm, interpreter, and harboring an alien. In so doing
we highlight strengths of the corpus analysis while also acknowledging
some drawbacks and unresolved questions.
On each of the test cases we start with a premise of ordinary meaning
that is susceptible of both definition and measurement. The premise is
that the ordinary sense of a term is that which occurs most frequently in a
properly controlled linguistic context—a context that controls for relevant
syntactic and semantic considerations, that is aimed at the relevant speech
63 Judging Ordinary Meaning 3-23-17
community, and that is limited to the appropriate time frame. Where the
data identify a sense that is most frequent in this setting we conclude that
there is at least a prima facie basis for concluding that that sense is the
ordinary one that would protect against unfair notice and vindicate the
intent of the lawmaker.
Yet we also recognize some possible grounds for questioning the
prima facie showing. One possible response would be to question the
viability of the relevant sense division—to suggest that the less frequent
sense is just the dodo bird (an unusual example of a bird, but no less a
bird). Another would be to identify pragmatic considerations that are not
adequately assessed through a corpus search. In circumstances in which
either of these concerns is established we think the conclusion that the
most frequent sense of a term is the ordinary one may be in doubt. And in
that event we identify other considerations that may be brought to bear,
such as alternative means of empirical analysis of the standard picture
(through a psycholinguistic experiment) or, ultimately, considerations
that go to legal content rather than communicative content.
a. Vehicles
Our data yield helpful grounds for assessing the meaning of vehicle.
We can use corpus analysis to make conclusions about whether the
ordinary sense of this term extends to the various examples discussed in
the literature—automobiles, bicycles, toy cars, airplanes, ambulances,
and golf carts.
Based on the common collocates of vehicle and our analysis of its use
in concordance lines, we can conclude that the most common sense of
this term is in reference to automobiles. And we can make a strong case
for crediting that sense as the ordinary one, in that it will best avoid
unfair surprise and vindicate the presumed intent of the lawmaker.
Our review of the corpus data also lead us to conclude that a number
of other potential candidates for the ordinary meaning of vehicle are out
of the running. Airplane (and its synonym plane) are unattested in the
collocation data, and a review of the concordance data didn‘t turn up any
instances in which the word vehicle was used to describe what we might
think of as a conventional airplane. The corpus data did show a few wellattested
uses of vehicle in reference to unmanned aerial drones. This was
not the most common or even a comparatively common use of vehicle, but
the data demonstrated that it was an attested and possible use of vehicle.
We can accordingly conclude that the ordinary meaning of vehicle is
automobile—at least to the extent that we accept the most common use of
a word and exclude other possible uses.
A similar story holds for bicycle. Bicycle (and its synonym bike) are
unattested in the collocation data and the bicycle use of vehicle didn‘t
show up in our randomized collection of concordance data. Yet a specific
64 Judging Ordinary Meaning 3-23-17
search for bicycle (and bike) turned up a number of instances in which
vehicle appears to be used to describe a bicycle. Here again, we can
conclude that the ordinary meaning of vehicle is automobile if we accept
the most common use of the word as the ordinary meaning, and exclude
the rare, but possible use of bicycle.
There may be very good reasons to exclude rare, but possible uses of a
word from coverage under the ordinary meaning rule. A decision to
extend the law to these terms could upset reliance interests of those
who—according to the data—are likely to think of automobile when they
read the law prohibiting vehicles. And the data give us no reason to think
that those who enacted this prohibition were thinking of airplanes,
bicycles, or toy cars.
Airplanes and bicycles appear on our frequency continuum: They are
attested in the data as possible examples of vehicle. But they are
unusual—not the most frequent and not even common. In our view this
weighs against treating these examples as falling under the ordinary sense
of vehicle. But again that is a question for legal theory.
A similar question for legal theory concerns the ambulance question.
Again ambulance is attested as a vehicle in the corpus data. And
ambulance also easily fits within the ordinary (automobile) sense of
vehicle. (It shares a variety of features with other automobiles and
appears in many similar contexts. It is simply a less common example.)
So the question here is one of intended meaning or pragmatic public
meaning—another question for legal theory.
What about golf carts? We found no examples of golf carts as vehicles
in the corpus. But does that mean they do not qualify under the ordinary
meaning of vehicle? Like the ambulance, a golf cart shares a number of
features with the most common vehicles—automobiles. It has four
wheels, a motor, carries passengers and baggage, and may even be
subject to licensure by a state department of motor vehicles.235
On the
other hand, licensed or not, we would not expect to see a lot of golf carts
on the Autobahn. The question whether a golf cart fits into the ordinary
meaning of vehicle (an ordinary meaning that the corpus data tells us in
the automotive use of vehicle) is accordingly a difficult one. It turns on
the viability of the sense divisions at work—on whether the golf cart is an
unusual example or a distinct psycholinguistic construct. That is not an
easy question to answer. It depends, as noted in Part II above, on the
sufficient conditions for the automobile sense of vehicle.
We see two ways of answering that question. One would be through
further corpus analysis. With sufficient corpus data we could assemble a
235
See Nevada Department of Motor Vehicles, Low Speed Vehicles & Golf Cars,
http://www.dmvnv.com/lowspeed.htm (last visited March 23, 2017).
65 Judging Ordinary Meaning 3-23-17
list of criteria for things we speak of as an automobile. And then we
could ask whether a golf cart has those criteria.236
An alternative would be to construct a psycholinguistic study.
Psycholinguists study the human acquisition, use, and comprehension of
language.237
They may do so by means of a human subject survey or
experiment. One possible experiment would be aimed at identifying the
sufficient conditions for the automobile sense of vehicle—to test whether
the golf cart is an unusual example of the automobile sense or a distinct
psycholinguistic construct.238
We may ultimately find no determinate answer in the above kind of
analysis for the golf cart question—or for related questions about go-carts
or four-wheelers. But psycholinguistics could yet yield empirical data of
some relevance. An alternative approach would be to construct a survey
aimed at assessing not just the first sort of vehicle that comes to mind but
also the range of meanings encompassed within a prohibition on vehicles
in the park. Survey data could give us quantitative information about
these notions of ordinary meaning—and whether, for example, a golf cart
would be understood to come within the ordinary sense of vehicle.
Yet there are barriers to psycholinguistic survey data, and reasons to
doubt the viability of this sort of analysis. One problem is timing-based—
to the ―meaning as of when‖ question. If we want to find meaning as of
the date of a statute‘s enactment we will never be able to measure it
psycholinguistically. And the older the statute the more concerning that
problem might seem.
Another problem concerns the artificiality of the psycholinguistic
study and its susceptibility to context effects and response bias.239
236
Possible criteria, for example, would likely include a steering wheel, motor,
wheels for passage on land, and seats for passengers. If those are the criteria then a golf
cart might count. But we can also imagine other criteria—like usual usage on paved
roads or highways, or licensure by the state motor vehicle division. And if those are the
criteria then a golf cart might not count.
237
JOSEPH F. KESS, CURRENT ISSUES IN LINGUISTIC THEORY: PSYCHOLINGUISTICS 6
(1992) (describing the ―domain of psycholinguistics‖ as ―defined by the activities of
comprehension, production, and acquisition‖).
238
See Eleanor Rosch, Cognitive Representation of Semantic Categories, 104 J.
EXPERIMENTAL PSYCHOL. 192, 193 (1975) (explaining the basis for this kind of study—
that word meaning may ―be represented in cognition not as a set of criterial features
with clear-cut boundaries‖ the way a dictionary would represent things, but instead ―in
terms of prototype (the clearest cases, best examples) of the category‖).
239
There are very few ways to (ethically) conduct a linguistic experiment without
letting the test subjects know about it. And once test subjects know they are being
evaluated there is no way to prevent them from self-correcting and self-evaluating while
they are participating in the experiment. People naturally, and unconsciously, attempt to
give the answer they think the experiment administrator wants or the answer that they
perceive to be the most ―socially desirable.‖ They try to figure out the ―right‖ answer.
This presents a problem when using psycholinguistic experiments in evaluating ordinary
meaning.
66 Judging Ordinary Meaning 3-23-17
Psycholinguists do their best to test real human perception that is
unbiased by flaws in the study design. But there will always be flaws in
study design. As to vehicle, for example, we would need to figure out
what kind of question to ask study recipients. Do we ask an open-ended
question or provide a list of possibilities to choose from? If we ask an
open-ended question we may end up with complicated coding problems
that could make the analysis less determinate. But if we provide a list the
study design may bias the answers.240
These limitations may be prohibitive. They may lead us to conclude
that we cannot give a conclusive answer to the question of whether the
ordinary understanding of vehicle extends to the golf cart. At that point it
may be time to abandon the standard picture—to fall back on ―fake‖
answers giving legal content to the law that is not necessarily in line with
its communicative content. That seems fine as a fallback. As our sense of
the law‘s communicative content becomes less clear the reasons for
crediting it are much weaker. Our point is just that this should not be the
law‘s first instinct.
b. Carrying a Firearm
The corpus data tend to support the dissenting position in Muscarello.
In the NOW Corpus we found 109 concordance lines involving carry and
firearm (or a synonym) in which we could discern one of the two senses
at issue. Most all of those (104) involved the bearing on your person
sense of carry. The COHA concordance lines were more limited but
supported the same conclusion: in 18 out of 20 lines in which we could
discern the relevant sense of carry the verb was used in the packing on
your person sense of the verb.
That gives us some meaningful empirical data about language usage. It
tells us that when people speak of carrying a firearm they are almost
always talking about carrying it on their person. And that provides a
prima facie basis for concluding that the ordinary communicative content
of the mandatory minimum sentencing provision in 18 U.S.C. §924(c)(1)
is limited to the personally bearing notion of carry.
Solan and Gales might observe that the data may merely be an artifact
of the greater commonality of personally bearing carrying in the real
world. That is probably correct but not necessarily a reason to distrust the
data. If most every time we speak of carrying a firearm we are talking
240
This is hard. Perhaps that‘s why Larry Alexander spoke (hypothetically) of a
telepathic basis for finding meaning. See Alexander, supra note __, at 142. That would
be easier. And presumably this is why Fallon and Posner, and to a lesser extent Baude
and Sachs, choose to fall back on a non-standard picture—on answers to questions other
than the ordinary communicative content of the law. We concede the difficulty. But
again suggest that we should try harder to conduct this threshold inquiry better before
we fall back to other premises.
67 Judging Ordinary Meaning 3-23-17
about personally bearing it, then the first sense of carrying to come to
mind is likely to be that sense. And extending the statute to the
transporting in a car sense may jeopardize significant reliance interests.
That leaves, as above, the question whether bearing and transporting
are two distinct psycholinguistic constructs or just alternative examples
within the same construct. Again we could test that by further empirical
analysis—by finding (through corpus or psycholinguistic study) the
sufficient conditions of carrying, and asking whether bearing and
transporting both qualify.
Perhaps we won‘t ultimately find a satisfactory answer to this question
in any empirical data. But even then the data will have been helpful. They
will allow us to avoid the smokescreen grounds for assessments of
ordinariness articulated by the competing opinions in Muscarello, and a
sufficient basis for turning to other means of assessment.
One such basis could be to make an attempt to assess intended
meaning. This inquiry may be a difficult one, as noted above. But again
at least a decision on this basis will be a transparent one—rooted in a
disagreement about whether Congress was likely concerned only about
firearms on a drug dealer‘s person, or might also have been concerned
about guns within relative reach in the dealer‘s vehicle. That sort of
debate may seem an empty one to a judge seeking determinacy in the
ordinary meaning of the text; but where such meaning is indeterminate
this debate seems preferable to a completely fabricated answer (as in one
rooted in a dictionary or etymology).
c. Interpreter
The data seem to provide support for Justice Alito‘s majority view in
Taniguchi. Of the 188 concordance lines we reviewed from the NOW
corpus, we found not a single instance of interpreter in the context of
text-to-text written translation. That seems strongly to indicate, at least,
that this is the kind of interpreter that first comes to mind when we use
this term.
That leaves the same question highlighted above in the other
examples—whether the written translator sense is psycholinguistically
separate from the oral translator notion.241
Here we see reason to suspect
that these are just alternative examples of a single psycholinguistic
construct. There is at least some indication of that in the fact that some
lexicographers treat these as just alternative examples of a single sense.
And again that question could be tested empirically through a
psycholinguistic study.
241
See Taniguchi, 132 S.Ct. at 2010 (Ginsburg, J., dissenting) (asserting that
―[d]istinguishing written from oral translation‖ is a ―dubious‖ endeavor, noting that
―some translation tasks do not fall neatly into one category or another,‖ and asserting
that an oral interpreter ―may be called upon to ‗sight translate‘ a written document‖).
68 Judging Ordinary Meaning 3-23-17
We have not sought to study intended meaning in our corpus analysis.
But as noted above we think such a study is possible. One approach
would be to think of ―interpreter‖ as a term used by lawmakers, and to
look for evidence of usage in this speech community. If we assembled
such evidence then we could have the debate flagged above—as to
whether intended meaning should win out over public meaning, or
whether they ought to collapse together as a matter of theory.
d. Harbor
Our ―harbor‖ data seems inconclusive.242
For a number of
concordance lines, it simply was not possible to determine from context
whether a conceal sense or only the shelter sense was applicable. Yet we
found a number of instances of both the conceal sense (23) and the
shelter sense (32) of this term.
That suggests that both senses are common and attested. To the extent
we regard the ordinary meaning as the most common sense of a word,
however, the data indicates that neither sense is ―ordinary.‖
It is hard to know what conclusion to draw from these inferences
(even accepting that we have a statistical basis for doing so). One
possibility is to say that both senses are ordinary in the sense that they are
both commonly attested. This is presumably the dissenting view in
Costello, and in line with the approach (at least sometimes) taken on the
―carry‖ question in Muscarello (that both personal carrying and car
carrying count as ordinary).
Another alternative is simply to abandon our search for the ―standard
picture.‖ If we lack probative data on the most frequent sense of a given
term we may simply conclude that we cannot determine the ordinary
communicative content of the law—and thus that we need a ―fake‖
answer, or an answer rooted in legal content. One such answer may be
found in the rule of lenity. Yet even here the data will have proven
useful. We have not fallen back on the law of interpretation (legal
content) just because the ordinary meaning question seemed difficult, or
that our intuition told us there was an ambiguity. We have identified data
to support our determination of ambiguity. The standard picture here
yields to the law of interpretation, but only after the necessary work has
been done.
IV. OBJECTIONS AND RESPONSES
We have little doubt of the need and basis for corpus linguistic
analysis of ordinary meaning. But we anticipate—and already have
seen—significant objections to the use of these new tools of
242
See infra CONCLUSION.
69 Judging Ordinary Meaning 3-23-17
interpretation. In a few recent cases judges have proffered corpus
linguistic analysis in support of their assessment of the ordinary meaning
of statutory terms.243
Some of these attempts have prompted doubt and
criticism from fellow judges.244
And even the judges who have advocated
for this approach (present company included) have acknowledged cause
for concern and care in this endeavor.245
The criticisms that we have heard or considered fall into three
categories: proficiency, propriety, and practicality. Each concern has a
kernel of viability but crumbles under careful scrutiny.
A. Proficiency: Judges (and Lawyers) Can‟t Do Corpus Linguistics
Judges and lawyers are not linguists. Most all of us, at least, are not
professionally trained ones. From that premise it is easy to jump to the
conclusion that judges and lawyers should leave the linguistic analysis to
professional linguists—to expert witness reports or testimony. A majority
of the Utah Supreme Court has so concluded in a couple of recent
cases.246
The ―proficiency‖ critique has some bite to it. For reasons noted above
we must concede that corpus linguistics is not ―plug and play‖ analysis.
Corpus data can be gathered and analyzed properly only with care and a
little background and training in the underlying methodology. And a
judge who proceeds willy-nilly may, either consciously or unwittingly, be
proffering data that has only the appearance of careful empiricism.247
For
243
See, e.g., State v. Rasabout, 2015 UT 72, 356 P.3d 1258 (Lee, J., concurring in
part and concurring in the judgment); Baby E.Z. v. T.I.Z., 2011 UT 38, 266 P.3d. 702
(Lee, J., concurring in part and concurring in the judgment); People v. Harris, 885
N.W.2d 832 (Mich. 2016).
244
See, e.g., Rasabout, 2015 UT 72 (majority opinion); Baby E.Z., 2011 UT 38
(majority opinion).
245
See, e.g., Rasabout, 2015 UT 72 at ¶¶ 94–101 (Lee, J., concurring in part and
concurring in the judgment)
246
Rasabout, 2015 UT 72, at ¶ 18 (majority opinion) (―Linguistics is a scientific
field of study that uses empirical research to draw findings. And just as with other fields
of scientific study, simply trying harder will not lead us to a better answer. The
knowledge and expertise required to conduct scientific research are ‗usually not within
the common knowledge‘ of judges, so ‗testimony from relevant experts is generally
required in order to ensure that [judges] have adequate knowledge upon which to base
their decisions.‘‖) (quoting Bowman v. Kalm, 179 P3d. 754, 755–56 (Utah 2008)); Baby
E.Z., 2011 UT 38, at ¶ 19 n.2 (―Unless this linguistic ‗context‘ is placed in its proper
context, it is of little analytical or persuasive value.‖).
247
See Ben Zimmer, The Corpus in the Court: „Like Lexis on Steroids‟, ATLANTIC
(Mar. 4, 2011), http://www.theatlantic.com/national/archive/2011/03/the-corpus-in-thecourt-like-lexis-on-steroids/72054/
(―While the corpus revolution promises to put
judicial inquiries into language patterns on a firmer, more systematic footing, the results
are still prey to all manner of human interpretation.‖); Michael Stubbs, Corpus
Semantics 107, in ROUTLEDGE HANDBOOK OF SEMANTICS (Nick Riemer ed., 2015)
70 Judging Ordinary Meaning 3-23-17
these and other reasons we whole-heartedly agree that the judicial
analysis of ordinary meaning will be improved in cases in which the
parties or their experts proffer corpus analysis that can be tested by the
adversary system.248
So we take the ―proficiency‖ critique as an appropriate word of
warning. Judges should acknowledge the pitfalls and limitations of the
tool of corpus linguistics. They should not overstate its utility, ignore the
care required to use it properly, or overlook the potential for subjectivity
or even strategic manipulation.
But that is as far as this critique can take us. The inevitable fact of the
matter is that judges and lawyers are linguists. We may not be trained in
linguistic methodology. But the enterprise of our work puts us
consistently and inevitably in the business of resolving ambiguities in
legal language.
Judges are experts, in other words, in interpreting the law.249
So the
question, ultimately, is not whether we trust judges to engage in linguistic
analysis. It is whether we want them to ―do so with the aid of—instead of
in open ignorance of our rebellion to—modern tools developed to
facilitate that analysis.‖250
Judges are likewise not historians. And it may rightly be said that
many lawyers and judges are even ―bad historians‖ with a tendency to
―make up an imaginary history and use curiously unhistorical
methods.‖251
―Yet judges of all stripes engage in historical analysis,
particularly in their interpretation of the constitution.‖252
―So the response
to our lack of historical training is not to back away from the enterprise; it
is to arm ourselves with the tools necessary to do the best history we
can.‖253
(noting that ―a constant background question is whether a corpus can ever, strictly
speaking, provide semantic data, since intuition is always required to interpret data,‖ but
concluding that while ―[s]emantic analysis can never be entirely objective . . . corpora
allow us to study language ‗with a degree of objectivity [. . .] where before we could
only speculate‘‖); id. at 110 (asserting that while ―software can automatically extract
data from large corpora, . . . intuition is necessary to interpret these data‖); Rasabout,
2015 UT 72, ¶ 21 (noting that a potentially significant portion of corpus data ―require[s]
an interpretive assumption‖ or retains some level of ambiguity).
248
Rasabout, 2015 UT 72 at ¶ 97 (Lee, J., concurring in part and concurring in the
judgment) (agreeing that judicial analysis of any kind is ―better when adversary briefing
is complete and in-depth‖).
249
Id. at ¶¶ 107–08 (explaining that although they do not have ―the kind of training
possessed by ‗linguistics experts‘ . . . judges are experts on one thing—interpreting the
law‖).
250
Id. at ¶ 111.
251
MAX RADIN, LAW AS LOGIC AND EXPERIENCE 138 (1940).
252
Rasabout, 2015 UT 72 at ¶ 109 (Lee, J., concurring in part and concurring in the
judgment).
253
Id.
71 Judging Ordinary Meaning 3-23-17
The same goes for linguistic analysis. ―We could continue to judge the
ordinary meaning of words based on intuition, aided by the
dictionary.‖254
―But those tools are problematic, for reasons noted
above.‖ ―So it is our current methodology and tools that involve bad
linguistics produced by unscientific methods.‖ For that reason, ―[i]f the
concern is reliability, the proper response is to embrace—and not
abandon—corpus-based analysis.‖255
The potential for subjectivity and arbitrariness is not heighted but
reduced by the use of corpus linguistics.256
Without this tool, judges will
tap into their linguistic memory to make assessments about the frequency
or prototypicality of a given sense of a statutory term. Such recourse to
memory and judicial intuition, moreover, is neither transparent, nor
replicable. Nothing is statistically worse than one data point—especially
a biased one. The potential for motivated reasoning is evident. A
common critique of the use of legislative history comes to mind. The
judge who considers only the linguistic evidence that she can summon to
mind may be looking over a crowd of language to pick out her friends.257
Corpus linguistics, by contrast, facilitates transparency and scrutiny.258
It is ―an empirical check on our (imperfect) linguistic intuition.‖259
And it
is not, ultimately, a terribly complex or difficult endeavor. ―Corpus
analysis is like math‖—everyone can do it as some basic level; at more
advanced levels it becomes too complicated for all but the experts.260
We‘re advocating rudimentary linguistic analysis that most anyone can
254
Id. at ¶ 112.
255
Id.
256
As one of us has noted, while a judge may ―go looking for supporting evidence in
a corpus‖ it is possible that ―after reviewing hundreds of concordance lines, a salient
meaning contrary to the judge‘s initial conclusion becomes harder to ignore.‖ Stephen
C. Mouritsen, Hard Cases and Hard Data: Assessing Corpus Linguistics as an
Empirical Path to Plain Meaning, 13 COLUM SCI. & TECH. L. REV. 156, 202 (2012).
257
See Patricia M. Wald, Some Observations on the Use of Legislative History in the
1981 Supreme Court Term, 68 IOWA L. REV. 195, 214 (1983) (―It sometimes seems that
citing legislative history is still, as [Judge] Harold Leventhal once observed, akin to
‗looking over a crowd and picking out your friends.‘‖). A parallel problem appears in
cases in which judges summon examples of word usage in literary works. Whitfield v.
United States, 135 S. Ct. 785 (2015) (interpreting the ordinary meaning of ―to
accompany‖ using a host of sources, including quotes from a Jane Austen as well as a
Charles Dickens novel). That kind of data cherry-picking if fraught with risk of
hindsight bias or motivated reasoning.
258
Zimmer, supra note 73 (―at least these ideological arguments can proceed on a
basis of concrete facts about how we use language, rather than on a welter of
idiosyncratic assumptions, as has too often been the case‖); Mouristen, supra note __, at
203 (―corpus analysis brings these subconscious assumptions about language and
meaning out in the open‖).
259
Rasabout, 2015 UT 72 at ¶ 21 (Lee, J., concurring in part and concurring in the
judgment).
260
Id. at ¶ 115.
72 Judging Ordinary Meaning 3-23-17
do.261
We ―just think we should be using a calculator instead of doing it
in our heads.‖262
The path forward is for judges and lawyers to identify the sort of
corpus analysis that we can perform sufficiently and reliably to
supplement the tools we are now using (and the sort of analysis we must
leave to linguists). In time the law and corpus linguistics movement263
will develop standards and best practices for this field. But until then we
should proceed cautiously and carefully in a direction that will allow us
to be the best linguists we can. Paraphrasing an observation made by
Justice Scalia and his co-author Bryan Garner regarding judges
performing historical analysis, we may or may not be able to do corpus
linguistics with the precision of an expert, but ―[o]ur charge is to try.‖264
B. Propriety: Judges Shouldn‟t Do Corpus Linguistics
The law puts limits on judicial analysis of matters that exceed the
bounds of the briefing and record in a particular case. Our rules of
judicial ethics say that ―[a] judge shall not investigate facts in a matter
independently,‖ but shall ―consider only the evidence presented and any
facts that may properly be judicially noticed.‖265
With this in mind, a
majority of the Utah Supreme Court has challenged the ―sua sponte‖ use
of corpus linguistics as falling beyond the proper domain of the judge.266
The analogy here may arguably be to cases in which judges perform
their own experiments to assess the factual assertions of the parties in a
particular case. A prominent example is in Judge Posner‘s opinion in
Mitchell v. JCG Industries.267
A question in that case was how long it
took poultry processing workers to change in and out of the safety
clothing they used to do their jobs. And Judge Posner‘s opinion included
261
While the COCA and similar linguist-designed corpora are more foreign than
Google or Westlaw, they are being employed in the same way. In short, we are
advocating using a corpus as a search tool or database to find uses of language that are
as easy to read as a Google search result. The results are just more sound.
262
Rasabout, 2015 UT 72at ¶ 115.
263
Perhaps it is a bit ambitious—or self-aggrandizing—to call this a ―movement.‖
But we are convinced that the future of the judicial analysis of ordinary meaning must
include a consideration of this linguistic tool. It is a bit embarrassing, in fact, that the
law has so much to do with the construction of language but we do not have a field
called law and linguistics. That will change in time. And we like to think of it as a
movement.
264
SCALIA & GARNER, supra note ___, at 400 (quoting KENT GREENAWALT, LEGAL
INTERPRETATION: PERSPECTIVES FROM OTHER DISCIPLINES AND PRIVATE TEXTS 168
(2010)).
265
MODEL CODE OF JUDICIAL CONDUCT r. 2.9(C) (AM. BAR ASS‘N 2011).
266
Rasabout, 2015 UT 72 at ¶ 17 (majority opinion).
267
754 F.3d 837 (7th Cir. 2014).
73 Judging Ordinary Meaning 3-23-17
a reference to an experiment he conducted on that question in
chambers—in which he ordered the clothing in question and asked ―three
members of the court‘s staff‖ to change in and out of it ―as they would do
if they were workers at the plant.‖ ―Their endeavors were videotaped.‖
And ―[t]he videotape automatically recorded the time consumed in
donning and doffing and also enabled verification that the ‗workers‘ were
neither rushing nor dawdling.‖ Posner referred to the results of this
experiment in support of ―the commonsense intuition that donning and
doffing a few simple pieces of clothing and equipment do not eat up half
the lunch break.‖268
Judge Diane Wood, in dissent, asserted that the Posner majority went
―beyond the proper appellate role‖ in conducting an experiment of
relevance to a factual question in the case. Wood complained that the
results of Posner‘s experiment ―cannot be considered as evidence in the
case,‖ which is limited to matters placed in the record pursuant to
applicable rules of civil procedure.269
This may be the paradigm that critics of corpus linguistics have in
mind when they question the viability of ―sua sponte‖ use of this tool.
But the analogy is inapt. A judge who considers corpus data in assessing
the ordinary meaning of a statute is not investigating the adjudicative
facts of a case; he is considering facts of relevance to the proper
interpretation of the law. These are known as legislative facts. And the
investigation of those facts is the inevitable—and quite proper—domain
of the judge‘s ―sua sponte‖ analysis. The governing rules of judicial
ethics expressly carve this out; they allow the courts to consider ―facts
that may properly be judicially noticed.‖270
The point is supported by the law of evidence. Governing rules of
evidence typically state that limitations on the judge‘s judicial notice
power is addressed to ―an adjudicative fact only, not a legislative fact.‖271
The distinction is this: ―[L]egislative facts are matters that go to the
policy of a rule of law as distinct from the true facts that are used in the
adjudication of a controversy.‖272
Such facts ―are not appropriate for a
268
Id. at 842.
269
Id. at 849 (Wood, C.J., dissenting).
270
MODEL CODE OF JUDICIAL CONDUCT r. 2.9(C) (AM. BAR ASS‘N 2011); UTAH R.
J. ADMIN. CODE r. 2.9(C) (2016).
271
FED R. EVID. 201(a) (―This rule governs judicial notice of an adjudicative fact
only, not a legislative fact.‖); Utah R. Evid. 201(a) (same).
272
UTAH R. EVID. 201(a) advisory committee‘s note. See also Kenneth Culp Davis,
An Approach to Problems of Evidence in the Administrative Process, 55 HARV. L. REV.
364, 402–03 (1942) (―When an agency wrestles with a question of law or policy, it is
acting legislatively, just as judges have created the common law through judicial
legislation, and the facts which inform its legislative judgment may conveniently be
denominated legislative facts. The distinction is important; the traditional rules of
evidence are designed for adjudicative facts, and unnecessary confusion results from
attempting to apply the traditional rules to legislative facts.‖).
74 Judging Ordinary Meaning 3-23-17
rule of evidence.‖ They are ―best left to the law-making considerations by
appellate and trial courts.‖273
And that is precisely what is involved in the
corpus linguistic analysis of the meaning of statutory text. Corpus
analysis has nothing to do with adjudicative facts—with the who, what,
when, or where of an underlying controversy. It has only to do with the
proper construction of the applicable law. And for that reason there is no
ethical or evidentiary prohibition on ―sua sponte‖ corpus analysis by a
judge.274
―A contrary conclusion would call into question a wide range of‖
inquiries routinely conducted by our courts, including the use of
dictionaries.275
―If we were foreclosed from considering outside material
that informs our resolution of open questions of law, we would be barred
from engaging in historical analysis relevant to a question of original
meaning of a provision of the constitution, or from considering social
science literature in resolving a difficult question under the common
law.‖276
And ―[l]inguistic analysis is no different.‖ ―[T]o the extent we
charge our judges with resolving ambiguities in language, we cannot (and
do not) reasonably restrict their ability to do so on a well-informed
basis—even on grounds not presented by the parties, and not within the
domain of judges‘ professional training.‖277
For better or worse, judges do that all the time. State court judges
decide questions of common law that require us to consider and weigh
questions implicating literature in fields of social science about which we
are hardly experts.278
No one bats an eye when judges do their own
273
UTAH R. EVID. 201(a) advisory committee‘s note.
274
Another way to frame this is the distinction between adjudicative and legislative
facts. Corpus-based analysis to determine the meaning of legal language falls under the
latter, and is appropriate and expected for judicial inquiry. See, e.g., Bulova Watch Co.
v. K. Hattori & Co., 508 F. Supp. 1322, 1328 (E.D. N.Y. 1981) (concluding that
―whether we explore the economic, political, or social settings to which the law must be
applied explicitly, or suppress our assumptions by failing to take note of them, we
cannot apply the law in a way that has any hope of making sense unless we attempt to
visualize the actual world with which it interacts,‖ and that the ―court‘s power to resort
to less well known and accepted sources of data to fill in the gaps of its knowledge for
legislative and general evidential hypothesis purposes must be accepted because it is
essential to the judicial process‖); see also ROBERT E. KEETON, JUDGING 38–39 (1990)
(discussing the distinction between legislative and adjudicative facts).
275
Rasabout, 2015 UT 72 at ¶ 106 (Lee, J., concurring in part and concurring in the
judgment).
276
Id.
277
Id.
278
Recent examples in the Utah Supreme Court include the question whether a
medical practitioner owes a duty to third parties who are foreseeably injured by the
negligent prescription of pharmaceuticals, B.R. ex rel. Jeffs v. West, 2012 UT 11, 275
P.3d 228; and the question of the appropriate age at which a child may appropriately be
required to defend against a claim for negligence, Nielsen ex rel. C.N. v. Bell ex rel.
B.B., 2016 UT 14, 370 P.3d 925. No one bats an eye when judges consult additional
75 Judging Ordinary Meaning 3-23-17
research and thinking on a broad range of ―legislative facts.‖ The matter
should be no different for linguistic analysis of ordinary meaning.
C. Practicality: Corpus Linguistics Will Impose an Unbearable Burden
The widespread use of corpus linguistics could put a strain on parties
and the courts. This is another criticism that has appeared in majority
opinions in the Utah Supreme Court. The argument is that turning the
analysis of ordinary meaning into an empirical, data-driven enterprise
will introduce the ―dueling expert‖ problem—and will make statutory
cases more costly and time-consuming.279
The effects of the proliferation of expert testimony are a matter
meriting careful consideration. For every question on which we require
expert analysis, we compound the expense and time it takes for a case to
be resolved. And we should not do that without a good reason.
Yet we find this objection to corpus analysis a hollow one for several
reasons. First is the fact that not all problems of statutory interpretation
lend themselves to corpus linguistic analysis. The utility of this tool, as
currently conceived, is limited to problems of lexical ambiguity—of a
contest between two meanings of the terms of the statutory text. That
excludes a category of problems of semantic or structural ambiguity—a
problem, for example, as to whether a modifier is understood to apply to
all items in a statutory list or only the ―last antecedent.‖280
Even as to the
class of cases of lexical ambiguity, moreover, not all cases will call for
corpus analysis. In our view ―[c]orpus analysis is something of a last
resort.‖281
―It comes into play only if we find that the legislature is not
materials—such as social science studies or materials—of relevance to the policy
determination implicated in these cases. That is because the inquiry involved in these
cases concerns ―legislative facts.‖
279
Rasabout, 2015 UT 72 at ¶ 19 (majority opinion) (declaring that the use of corpus
linguistics ―would require ad hoc linguistics research that could only be reliably
conducted by dueling linguistics experts,‖ and that ―[i]mposing such a significant
financial burden on so many of the litigants coming through the doors of our courts
would be tantamount to locking those doors for all but the most affluent‖).
280
See Lockhart v. United States, 136 S. Ct. 958 (2016) (describing the ―rule of the
last antecedent‖ and applying it to interpret a statute imposing a mandatory minimum
sentence on defendants who violate the federal child pornography statute and have
previously been convicted of certain crimes); CRUSE, supra note __, at107–08
(―Ambiguity has been presented here as a lexical phenomenon; it is important to
emphasize, however, that there are other sources of ambiguity. One of these, of course,
is syntax, as in Mary saw the man with the telescope. Many syntactic ambiguities arise
from the possibility of alternative constituent structures, as here: with the telescope is
either a manner adverbial modifying saw, or a prepositional phrase modifying the
man.‖).
281
Rasabout, 2015 UT 72 at ¶ 118 (Lee, J., concurring in part and concurring in the
judgment).
76 Judging Ordinary Meaning 3-23-17
using words in some specialized sense, and only if we cannot reject one
of the parties‘ definitions based on the structure or context of the
statute.‖282
This yields a limited but important domain for corpus
linguistics. We turn to an empirical analysis of frequency only in cases in
which we have ―no better way‖ of resolving a contest between
probabilities of meaning.283
That is a relatively rare case.284
Second, corpus-based analysis won‘t always require an expert. This
―isn‘t rocket science.‖285
Lawyers are crafty, ingenious creatures with the
capacity to learn and even master new tools, technologies, and
methodologies. Witness the way attorneys have learned to parse historical
materials and present them when litigating the original meaning of the
constitution. In a way, lawyers have been doing corpus analysis for a long
time; they scour Westlaw or Lexis to determine how courts have
interpreted a phrase or concept. So it‘s undoubtedly true that lawyers will
have to bone up on some basic linguistic methodology. But continuing
education is an ongoing element of the legal profession. And a familiarity
with and capacity for corpus analysis can take root just like Westlaw and
Lexis searches did.286
The rising generation of millennials is particularly
suited to the task. They have never known life without a computer, and
are constantly embracing new applications and tools for computer
analysis. In time we will see competing corpus presentations as a matter
282
Id.
283
Id.
284
See id. at ¶ 118 (asserting that in ―five years‖ on the Utah Supreme Court, he had
―employed such analysis only a very few times,‖ and that ―[i]n the many other statutory
cases‖ that have arisen, he ―disposed of the matter using more traditional tools of
interpretation‖).
285
Id. at ¶ 117.
286
The advent of computer-aided legal research is now an accepted staple. But it
wasn‘t always thought to be so. Early reactions paralleled some of the responses to
corpus linguistics. See, e.g., Barbara Bintliff, From Creativity to Computerese: Thinking
Like a Laywer in the Computer Age, 88 LAW. LIBR. J. 338, 339 (1996) (warning that
computer-aided legal research will undermine the ability to think like a lawyer); Molly
Warner Lien, Technocratism and the Soul of the Common Law Lawyer, 48 AM. U. L.
REV. 85, 85–86 (1998) (arguing that computer-aided legal research ―may be harmful to
the process of legal reasoning‖ and that lawyers should be aware of the ―negative
impacts‖ of using technology in this way); Robert C. Berring, Legal Research in the
World of Thinkable Thoughts, 2 J. APP. PRAC. & PROCESS 305, 316 (2000) (declaring
that it ―scares‖ the author ―[i]f search engines like Google move into legal
information‖); Scott P. Stolley, Shortcomings of Technology: The Corruption of Legal
Research, ABA‘S APP. ADVOC. COMMITTEE NEWSL. 39 (Apr. 2004) (viewing the likes
of Lexis-Nexis and Westlaw as leading to a generation of lawyers who can‘t find cases
on point). Most of us view this criticism as downright silly today. We realize that
computer research tools can be misused, and may be improved if supplemented by more
traditional methods. But they cannot properly be rejected on the basis of their
unfamiliarity.
77 Judging Ordinary Meaning 3-23-17
of course in adversary briefing.287
There will often be no need for dueling
experts, just as there is often no need for dueling historical experts in
constitutional litigation, or dueling dictionary experts on a statutory
question.
Finally, if in the rare case there is a need for the parties to retain
corpus linguistic experts, that is hardly cause for alarm. Where the issue
is complex enough and the stakes are high enough, expert analysis could
be helpful—and certainly preferable to deciding a matter as significant,
say, as the applicability of a federal sentencing enhancement on the basis
of an unreliable source like a dictionary or an opaque one like a judge‘s
intuition. Some problems are important enough to merit expert analysis.
We should leave that matter to the marketplace—to the clients and
lawyers who decide how best to formulate and present a legal position.
CONCLUSION
Some points of analysis outlined herein are necessarily tentative. That
seems inevitable in the course of breaking new ground on refined theories
and new tools for assessing the ordinary communicative content of the
language of the law. We trust that some of the value in our contribution
will be to spark further analysis and scholarship on the questions we have
raised.
Moving forward, judges, lawyers, and linguists will need to
collaborate to settle on some best practices in this emerging field. Some
important questions to answer are methods for selecting the best corpus
for a given type of ambiguity, standards for the appropriate sample size
for a given search, standards for determining appropriate search terms
and search methods for various types of inquiries, and identification of
suitable coding methods. Scholars have begun to explore these and other
related questions.288
Further work is in order. But we are confident that
lawyers and linguists can work together to develop an orthodox set of
methods that will refine an approach that now stands at its infancy.
287
This will hold if but only if our courts continue to embrace this methodology—as
has happened recently in Michigan. If we (judges) build it, they (attorneys) will surely
come. See In re Estate of Cliffman, 2016 WL 4480882 (Mich. 2016) (Appellate Brief)
(submitting supplemental authority with exhibits ―show[ing] the relative frequency with
which words or word combinations appear in the COCA database,‖ and noting that ―[i]n
Harris, this Court approved the use of the corpus linguistics in determining the common
usage and meaning of statutorily undefined words‖).
288
See James C. Phillips & Jesse A. Egbert, Improving Corpus Design and CorpusBased
Analysis for Linguists and Lawyers: Principles and Practices from Survey and
Content-Analysis Methodologies, Paper presented at Brigham Young University Law
School‘s Law and Corpus Linguistic Conference, (Feb. 3, 2017) (arguing that principles
and methodologies from survey and content-analysis methodologies need to be applied
to corpus design, selection, and coding).
78 Judging Ordinary Meaning 3-23-17
Linguists have observed that corpus linguistics generally ―has not yet
reached the stage where it can present a stable set of methodological
procedures coupled to specific descriptive questions.‖289
That
undoubtedly holds true even more for an application of this tool to a
brand new field. The law, after all, asks questions that linguists
historically have not deemed important—concerning the average or
―ordinary‖ understanding of a given term in a given linguistic setting. So
the methodology of corpus linguistics will undoubtedly experience
growing pains as it is employed for new purposes. Yet linguists have
noted elsewhere (more generally) that ―[t]he observation that
distributional corpus analysis has not reached‖ the stage at which we
have embraced a set of widely accepted norms ―is certainly not a reason
to abandon the approach; rather, it defines a promising and exciting
research [program].‖290
And that is certainly true as to the application of
corpus linguistics to the enterprise of judging ordinary meaning.
Whatever its current limitations, ―semantic analysis can, and indeed,
should, turn to corpus methods.‖291
The need is acute when the interpretive task involves questions of law.
Too much rides on the resolution of legal ambiguity to resolve the matter
by means ―fraught with a potential for bias and error.‖292
If and when the
law turns on an assessment of ordinary communicative content we must
at least try to define and operationalize the inquiry with greater care. We
see the approach outlined here as a step in that direction.
289
GEERAERTS, supra note __, at178.
290
Id.
291
Glynn, supra note __, at 7.
292
See State v. Rasabout, 2015 UT 72, ¶ 134 (Lee, J., concurring in part and
concurring in the judgment).