Artis v. District of Columbia, part 2: Units of meaning and dictionary definitions

Sometimes, it’s immediately obvious from the opinions that a case raises questions about interpretation that are interesting, important, or both. Smith v. United States, in which the question was whether trading a handgun for drugs amounts to “using” it, is a classic example. At first glance, the Supreme Court’s decision in Artis v. District of Columbia  doesn’t seem to be in that category. It doesn’t offer interesting linguistic issues that call attention to themselves, except for someone who is familiar with the work of the linguist John Sinclair and the lexicographer Patrick Hanks. But with some digging, Artis yields some issues that I think are  interesting and significant, having to do with new approaches to analyzing questions of word meaning and with how not to use dictionaries.

This is my second post about Artis; the first one discussed Justice Gorsuch’s choice of what dictionaries to cite in his dissenting opinion. It also set out the relevant background information. I won’t repeat everything here, but here are the essentials.

Artis involves the federal supplemental-jurisdiction statute (about which, read my previous post). That statute provides that the limitations period for state-law claims as to which supplemental jurisdiction is invoked “shall be tolled while the claim is pending and for a period of 30 days after it is dismissed unless State law provides for a longer tolling period.” At issue before the Supreme Court was what the statute means in saying that the limitations period is “tolled.” There were two possibilities:

  • The stop-the-clock interpretation. Tolling acts as a “Pause” button that simply stops the running of the limitations clock while the federal case is pending and then for the 30 days after it is dismissed. This is the interpretation advocated by Artis, the plaintiff.
  • The grace-period interpretation. The clock continues to run, but with a 30-day grace period that starts if and when the federal case is dismissed. This was the position of the defendant, the District of Columbia.

The majority adopted the stop-the-clock interpretation, in an opinion written by Justice Ginsburg. Gorsuch (joined by Kennedy, Thomas, and Alito) favored the grace-period reading.

THE MAJORITY OPINION BEGINS by saying that when toll is used “in the context of  a time prescription,” it “ordinarily…means that the limitations period is suspended (stops running) while the claim is sub judice elsewhere, then starts running again when the tolling period ends, picking up where it left off.” The authority cited to support that statement is Black’s Law Dictionary, which the opinion describes as saying that “‘toll,’ when paired with the grammatical object “statute of limitations,” means “to suspend or stop temporarily” This definition, the opinion says, “captures the rule generally applied in federal courts,” The opinion also points to several statutes that use the verb toll in this way, and notes that in contrast, neither the defendant nor the dissent identified any federal statute in which toll was used differently.

The dissent’s textual argument, on the other hand, relies mainly on the dictionary definitions that are the subject of my previous post: “to take away, bar, defeat, or annul” (OED) and “to take away; to vacate; to annul” (Webster’s Second (W2)). In this respect, the dissent tracked the defendant’s argument, which had cited the OED’s definition and similar definitions from other dictionaries, including Webster’s Third (W3).

As an example of toll being used in this sense, the District of Columbia had pointed to an 1822 case that dealt with the tolling of a right of entry upon land: “The remaining consideration under this head is, whether the possession of Joseph Dudley can be considered as an adverse possession so as to toll the right of entry of the heirs, and, consequently, extinguish, by the lapse of time, their right of action for the land[.]” The District argued that toll means the same thing when used with regard to limitations periods as it does when used with regard to rights of entry: “There is nothing special about tolling limitations periods versus tolling any other fact, right, or consequence.”

That view is implicit in the dissent’s reliance on the definitions that I’ve quoted. So the majority’s response to the District of Columbia’s argument also applies to the dissent:

The District offers no reason why, in interpreting “tolled” as used in §1367(d), we should home in only on the word itself, ignoring the information about the verb’s ordinary meaning gained from its grammatical object. Just as when the object of “tolled” is “bell” or “highway traveler,” the object “period of limitations” sheds light on what it means to “be tolled.”

WHEN I READ THAT PASSAGE, it grabbed my attention, because it reflects an approach to word meaning that is similar to what I advocate in my paper from last year’s law-and-corpus-linguistics conference at BYU, A Lawyer’s Introduction to Meaning in the Context of Corpus Linguistics (to be published soon in the BYU Law Review; see also this post for a much shorter summary).

The paper is based on insights that have come out of using corpus analysis in lexicography. Those insights challenge longstanding assumptions about how word meaning should be analyzed and described, and they are widely regarded as having revolutionized the field. Under the traditional approach to lexicography, which is based on the principle of com­po­sition­ality, the relevant unit of meaning is the word. Each word in a sentence is assumed to have a meaning of its own, and that the meaning of the sentence is seen as being determined by the meanings of the individual words and how they are grammatically combined. And this almost certainly represents the understanding of virtually all lawyers and judges about how language works. But while it is in many respects a useful way of thinking about language, it glosses over a major complication.

Many words, including most of those that are used the most frequently, have multiple possible meanings, and when such a word is used in a sentence, its meaning in that context depends largely on what the rest of the sentence says. So there is a chicken-and-egg problem: how can in­dividual words be regarded as basic units of meaning when the meaning of a word in a particular context is itself affected by the context?

The answer that is suggested by the work in corpus lexicography is that rather than assuming that the relevant unit of meaning is the word, the focus should be on phrases and other multiword expressions. Perhaps the most vivid demonstration of what such an approach would entail is provided by transitive verbs—precisely the category that is relevant in Artis. When we start looking at phrases consisting of a transitive verb and its direct object, we can see that the verb’s meaning in context varies depending on which noun serves as its direct object:

•  throw a football, throw a football game, throw a party, throw a switch
•  drop a dish, drop a course, drop a hint, drop a claim (from a lawsuit)
•  observe a religious holiday, observe somebody’s activities
•  sell a product, sell an idea
•  file a piece of metal, file a lawsuit
•  examine a witness, examine a patient, examine a patent
•  exhibit symptoms, exhibit paintings
•  run a race, run a machine, run a risk, run an ad, run an errand
•  pull a rope, pull rank, pull a prank, pull a gun, pull a switch

This phenomenon was known before corpus linguistics came along, but the introduction of multimillion-word computerized corpora in the 1980s made it clear that this was only one of the ways in which the meaning of a given word in a text was dependent on the words that co-occurred with it and even on the grammatical constructions that it occurred in. (For an example, see my post The semantics of sleeping in railway stations.)

So when the  majority in Artis says that the focus should be on the meaning of toll the period of limitations rather than on the meaning of toll by itself, it acting consistently with advances in lexicography.

THE DISSENT’S ANALYSIS contrasts with the majority’s in several ways.

The most important difference is that whereas the majority’s reasoning was primarily usage-based, meaning that it was driven by how expressions like toll the statute of limitations are typically used, the dissent’s textual analysis was driven almost entirely by dictionary definitions.

Here is the key paragraph from the majority:

Ordinarily “tolled,” in the context of a time prescription like §1367(d), means that the limitations period is suspended (stops running) while the claim is sub judice elsewhere, then starts running again when the tolling period ends, picking up where it left off. See Black’s Law Dictionary 1488 (6th ed. 1990) (“toll,” when paired with the grammatical object “statute of limitations,” means “to suspend or stop temporarily”). This dictionary definition captures the rule generally applied in federal courts. See, e.g., Chardon v. Fumero Soto (Court’s opinion “used the word ‘tolling’ to mean that, during the relevant period, the statute of limitations ceases to run”). Our decisions employ the terms “toll” and “suspend” interchangeably. For example, in American Pipe & Constr. Co. v. Utah, , we characterized as a “tolling” prescription a rule “suspending the applicable statute of limitations”; accordingly, we applied the rule to stop the limitations clock. We have similarly comprehended what tolling means in decisions on equitable tolling. See, e.g., CTS Corp. v. Waldburger, (describing equitable tolling as “a doctrine that pauses the running of, or ‘tolls’ a statute of limitations” (some internal quotation marks omitted)); United States v. Ibarra, (“Principles of equitable tolling usually dictate that when a time bar has been suspended and then begins to run again upon a later event, the time remaining on the clock is calculated by subtracting from the full limitations period whatever time ran before the clock was stopped.”).

The dissent offered nothing comparable—presumably because it was undisputed that in the statute-of-limitations context, toll is never used in the way advocated by the dissent. That point was made explicitly at oral argument by counsel for Artis:

Respondent’s position is that the tolling period consists of the pendency of the petition for rehearing and the entire 90-day period. And the word “tolled” is never used that way. Not in a case, not in a statute, I have not found a single reference to the—to the word being used in that context….So the word “tolling” literally never means what Respondent claims it means. [paragraph breaks omitted]

This assertion can be fact-checked by doing a simple search on Westlaw or Lexis. Since the dissent doesn’t dispute it, it seems fair to assume that the statement is correct.

The dissent therefore had little choice but to rely on dictionary definitions. Here is the heart (and practically the entirety) of the dissent’s textual analysis:

Start with the statute’s key term. Where, as here, a law instructs us to “toll” a limitations period, we know it may be telling us to do one of (at least) two different things. The dictionary informs that to “toll” means “[t]o take away, bar, defeat, [or] annul.” See 18 Oxford English Dictionary 204 (2d ed. 1989); Webster’s New International Dictionary 2662 (2d ed. 1957) (“[t]o take away; to vacate; annul”); Oxford Latin Dictionary 1947 (1982) (“tollere,” the Latin origin, means to “remove” or “lift”). So when a statute speaks of tolling a limitations period it can, naturally enough, mean either that the running of the limitations period is suspended or that the effect of the limitations period is defeated.

What I just quoted was essential to the dissent’s analysis. The remainder of that analysis, arguing in favor of the second of the two interpretations, is premised on the conclusion that toll is ambiguous, and that conclusion, in turn, is based entirely on the two definitions.

So in evaluating the dissent, the question is whether the definitions it relied on can reasonably be regarded as statements of what toll means in the context of statutes of limitations. The answer, I think, is clearly no.

TO UNDERSTAND WHY I SAY THAT, it’s necessary to understand a few things about lexicography. Let’s start with this statement from the Oxford Guide to Practical Lexicography (which is also the source of the quotes in the next paragraph): “A reliable dictionary is one whose generalizations about word behavior approximate closely to the ways in which people normally use (and understand) language when engaging in real com­muni­ca­tive acts (such as writing novels or business reports, reading newspapers, or having conversations).”

Attaining that kind of reliability requires that lexicographers rely on objective evidence: data from “observing language in use.” The data comes from real-world texts, originally in the form of excerpts collected by hand (known as “citations”), and beginning in the mid-1980s, in the form of corpus data. And then, generalizing from that data lexicographers attempt “to make explicit the mean­ing dis­tinctions which—in normal communication—humans deal with uncon­sciously and effort­lessly.”

Thus, dictionary definitions are generalizations based on actual instances of use.

On the reasonable assumption that toll has never been used in the way that was argued for by the District of Columbia and the dissent, the definitions cited by the dissent couldn’t possibly have been written with the intent of covering such uses. And beyond that, it’s doubtful that the definitions were written with statutes of limitations in mind at all. In the OED’s definition, none of the example sentences involved the tolling of a statute of limitations. And there’s a good chance that as of the time the OED’s definition was written, toll had never been used in England with respect to statutes of limitations. In an effort to find out whether it had been used that way, I did some quick historical research in Google Books and English case law. I found nothing to suggest that toll was used in that way in the period predating the OED’s definition. I’m not sure how much weight to put on that, since I have no feel for whether there  might be something else out there that I didn’t find. But I did find something that gives reason to think that, at least as of the early-19th century, toll wasn’t used with respect to statutes of limitations. What I found was an English law dictionary published in 1820, which contained a 10,500-word entry on limitation of actions. That entry included only one use of toll, and it had to do with the tolling of rights of entry. While that’s not conclusive, on balance, the sources I looked at gave more reason to think that toll was not used in connection with statutes of limitations than that it was. And if those sources are representative of the state of English usage, the OED’s definition was not intended to apply to the kind of usage that was at issue in Artis.

So much for the OED’s definition. What about the one from Webster’s Second? Here, too, there’s no clear answer. But if I were a betting man, I’d put my money on the theory that W2’s definition was based on the OED’s and therefore contributes nothing new. In the frontmatter to W2 is a statement saying that in compiling the dictionary, the editors relied not only on the 1,655,000 citations in the Merriam-Webster files, but also on “about 2,000,000 citations in other dictionaries.” The vast majority of those additional citations almost certainly came from the OED. Indeed, while W2 didn’t explicitly acknowledge having used the OED for that purpose, such an acknowledgement can be found in the frontmatter of  W2’s predecessor (Webster’s New International Dictionary) and of its successor, W3.

To sum all of this up, there is no reason to think that either of the definitions cited by the dissent was intended to cover expressions such as the one in dispute in Artis, and there are pretty good reasons to think the opposite.

At this point, the question arises whether the dissent’s interpretation can be justified as being based on a generalization extending the definitions’ scope so that it covers the limitations context. The answer again is no.

For starters, the dissent’s interpretation cannot be derived by simply generalizing from the definitions it relied on. Taking the two definitions together, they define toll as ‘take away, bar, defeat, annul, or vacate.’ Generalizing from that definition (or rather, from that set of supposed synonyms, since that what the “definitions” really are), we get the following:

toll X = take away X, bar X, defeat X, annul X, or vacate X,


toll the period of limitations = take away the period of limitations, bar the period of limitations, defeat the period of limitations, annul the period of limitations, or vacate the period of limitations,


toll the statute of limitations = take away the statute of limitations, bar the statute of limitations, defeat the statute of limitations, annul the statute of limitations, or vacate statute period of limitations.

Some of these readings are awkward at best—I don’t know what it means to “vacate” a period of limitations or statute of limitations—and that is itself a reason to be dubious about this exercise in generalization. But the meaning of “taking away” or “defeating” a period of limitations or statute of limitations is fairly straightforward. The problem for the dissent is that the most obvious interpretation of the phrase doesn’t support the dissent’s overall interpretation of the statute. To take away or defeat a limitations period or a statute of limitations, it seems to me, would be to relieve plaintiff of the need to file suit within the time period that would apply if the statute hadn’t been taken away or defeated.

The dissent concedes that its interpretation requires that the rather thin semantics that I’ve described be inferentially fleshed out: “The grace period approach construes the term ‘period of limitations’ as directing us to the ‘effect of the period of limitations’ that is tolled or taken away.” While there’s nothing unusual about a sentence communicating a meaning that is richer or more complex than what it expressly says, I don’t see any principled reason to think that the dissent’s construal is appropriate. There is certainly no basis for it in the definitions the dissent relies on, which are the foundation on which its analysis is built. Nor, as far as I know, is there any type of pragmatic enrichment that could generate the dissent’s reading. And the dissent all but admits that it is reasoning backward from a desired result: it describes that construal as being implicit in the grace-period interpretation, rather than as arising naturally from how toll is used in the context of the statute.

The dissent contends that the majority’s interpretation similarly relies on inferential enrichment: “For its part…, the stop clock approach construes ‘period of limitations’ to refer to the ‘running of the period of limitations’ that is tolled or taken away.” I disagree. To begin with, the dissent’s reference to the period of limitations being “tolled or taken away” takes it for granted that the majority’s interpretation is compatible with understanding toll to mean ‘take away.’ But the majority doesn’t accept that reading; it interprets toll to mean ‘suspend.’ And while even that interpretation arguably requires some enrichment, the necessary enrichment is of a kind that is very familiar in English. Think of phrases such as start a book or finish lunch, which are understood to refer to specific types of action that are not spelled out explicitly. Depending on the context, start a book  would probably be understood to mean ‘start reading a book’ or ‘start writing a book,’ and finish lunch to mean ‘finish eating lunch’ or ‘finish making lunch.’ Other meanings are also possible, given an appropriate context. The inference that the majority’s interpretation requires is analogous to the inferences in these examples: it provides detail about the semantic representations that contribute to the way in which the expression is understood.

(I should note that although I’ve been talking in terms of inferences of pragmatic enrichments, the kinds of phenomena I’m discussing can also be analyzed under theoretical approaches that treat them differently. However, for purposes of this discussion, these differences in linguistic theory aren’t important.)

SO FAR I HAVE BEEN FOCUSING on issues specifically relating to the definitions at issue; there are also two issues that have wider relevance and are more important for legal interpretation generally

First, to the extent that the dissent’s approach appears at first to be plausible, that is a byproduct of the style in which the definitions are written. Each definition consists of a series of verbs that are (implicitly) presented as being synonymous with toll. Except for the examples in the OED, which show toll being used mainly in sentences in which the thing being tolled is a right of entry, the definitions give no information about the kinds of contexts in which the word typically appears. That may lead foster the belief that toll will mean the same when used in expressions such the period of limitations is tolled. Certainly, the definitions do little (OED) or nothing (W2) to discourage such a belief. But that’s a problem, because as I’ve noted, the meaning of a verb in context will often be affected by which noun combines with it.

Defining words by using synonyms, with little or no contextual information, is a common defining strategy in the kinds of dictionaries that are cited by judges and lawyers, which is to say, monolingual dictionaries intended for native speakers of English. But a different defining strategy can be found in a category of dictionaries that judges and lawyers don’t generally consult. I’m talking about learners’ dictionaries, which are intended for people who are not native English-speakers, and in particular I’m talking about dictionaries for advanced learners.

You might be surprised to learn that I think that such dictionaries are useful not only for people learning English as a second language but also for people litigating and deciding lawsuits. Unlike familiar standbys such as Merriam-Webster and American Heritage, learners’ dictionaries reflect insights that have been arisen out of the analysis of corpus data. In fact, the analysis I’m referring to was conducted in compiling the first dictionary based entirely on corpus data, which was a learner’s dictionary: the Collins COBUILD Dictionary of the English Language. The insights that came out of that work, which I talked about earlier in this post, have to do with the ways in which the meaning of a word in context can be influenced by that context.

The Cobuild dictionary pioneered a defining style that aimed to incorporate contextual information into the definition itself (although Cobuild called the entries “explanations” rather than “definitions”). An example of this is the following excerpt from the entry for carry in Cobuild’s first edition. As you will see, the various possible meanings are linked, based on what the corpus data showed, to the lexical and grammatical patterns with which each meaning is typically associated. Specifically, the entry shows that the meaning of carry in any given use will depend on who or what is portrayed as doing the carrying and/or as being carried:

1. If you carry something, you hold it or support it so that it does not touch the ground, and take it with you as you go somewhere.

2. If something such as a river or a wind carries something somewhere, it moves it there by its force.

3. When a vehicle carries people, they travel inside it from one place to another.

5. If a person, animal, or thing carries a germ or a disease, they are infected with it and are capable of giving it to someone and making them ill.

6. If someone carries something such as a new idea or a message from one place or another, they spread it or take it there.

(If the Supreme Court had been made aware of this definition when it decided Muscarello v. United States (discussed here), the outcome might well have been different.)

Learners’ dictionaries don’t include entries covering the legal uses of toll. But such a definition, written in the Cobuild style, would look something like this:

1. If a statute of limitations is tolled, the running of the limitations period is suspended or starts over.

2. If a right of entry is tolled, it may no longer be exercised.

A definition such as this one would not have afforded any traction for the dissent’s approach of generalizing from dictionary definitions. And it would at the same time have offered support for the majority’s conclusion that the meaning of toll can vary depending on what noun it is paired with.

THIS BRINGS ME TO what is the most abstract of my criticisms of the dissent in Artis, but also possibly the most important. It has to do with the dissent’s overall conception of the role of dictionaries in legal interpretation. In the dissent itself, that conception is only implicit. However, it is visible, although still not fully explicit, in a question that Justice Alito asked the plaintiff’s counsel at oral argument:

Do you admit that there are definitions of the term “toll” that are consistent with Respondent’s argument? If we look in dictionaries, are there not definitions that are consistent with their argument? [Transcript at 15-16.]

Underlying this question, and the dissenting opinion, is an assumption that even if toll  has never been used in a way consistent with the dissent’s interpretation, that interpretation is permissible due its being consistent with something written in two books that have lots of pages and weigh a lot. The dictionary definitions are in a sense treated as if they were statutory definitions but without binding effect. If a statute says, “In this chapter, ‘red’ means ‘blue,’ and ‘blue’ means ‘red,’” courts have to interpret the statute that way, linguistic reality be damned. Under the dissent’s analysis, the dictionary definitions don’t require courts to interpret toll consistently with the definition, but they nevertheless authorize the courts to do so, despite linguistic reality to the contrary.

Statutory definitions, which specify how certain words in the statute are to be understood, are what is known as stipulative definitions. Such definitions “impart a meaning to the defined term, and involve no commitment that the assigned meaning agrees with prior uses (if any) of the term.” They could also be called Humpty-Dumpty definitions, because Humpty is famous for saying, “When I use a word, it means just what I choose it to mean—neither more nor less.” When Alice raised the question “whether you can make words mean so many different things,” Humpty Dumpty responded that the real question is “which is to be master—that’s all.” With a statutory definition, the legislature is master. In the dissent in Artis, the dictionaries are treated as the masters.

That approach to interpretation has nothing to do with determining ordinary meaning—whether with regard to the general usage of nontechnical words or, as in Artis, regarding the usage within the legal community of legal terms of art.


Some of the quotations in this post have been cleaned up.

Leave a Reply