THE CANONS OF LEGAL INTERPRETATION are pithy sayings setting out various ways in which statutes should or should not be interpreted:
Every word in a statute should be given effect.
Expressio unius est exclusio alterius (the expression of one thing suggests the exclusion of others).
Ejusdem generis (interpret a general term to reflect the class of objects reflected in more specific terms accompanying it).
Statutes should be presumed not to apply extraterritorially.
…and so forth.
Karl Llewellyn, a prominent mid-20th century legal scholar, famously said that “there are two opposing canons on almost every point.” On November 3, the Supreme Court will hear argument in Lockhart v. United States, a case that presents about as stark a clash between opposing canons as you could ever hope to find.
Lockhart is of interest to us here at LAWnLinguistics because the canons at issue are among the small group of “linguistic canons”—those that deal with language-related issues. In particular, they deal with resolving ambiguities that arise as a result of the statutory language having more than one plausible syntactic structure. I dealt at some length with the general issue of syntactic ambiguity, and with the specific canons that are now at issue in Lockhart, in my multipart look at Reading Law: The Interpretation of Legal Texts, by Supreme Court Justice Antonin Scalia and legal-writing guru Bryan Garner. (Of my previous posts about the book the ones most relevant here are Syntactic ambiguity, Three syntactic canons, On Garner on Posner on Scalia & Garner, and Last antecedents, series qualifiers, and psycholinguistics.)
In those posts (particularly the last three) I criticized several aspects of how Reading Law treated the canons that are now at issue in Lockhart. Some of those criticisms, as well as others that have occurred to me as I’ve thought about Lockhart, are pertinent to the legal issues in the case. And more broadly, as I’ve continued to think about Reading Law’s handling of the syntactic canons, I’ve come to believe that the framework established by Reading Law is not merely problematic, but deeply flawed. As it turns out, Lockhart provides a good lens through which those flaws can be examined.
LOCKHART INVOLVES A FEDERAL STATUTE (18 U.S.C. § 2252) that prohibits distributing, receiving, or possessing child pornography. It also happens to be the statute that was at issue in United States v. X-Citement Video, Inc., a case that prompted one of the earliest uses of linguistics as a tool in statutory analysis.)
The statute provides (in subsection (b)(2)) that violators are to be sentenced to up to ten years in prison, except that the sentence is increased (to at least ten years and not less than thirty) if the defendant had previously been convicted under various other laws, including—
any state law “relating to aggravated sexual abuse, sexual abuse, or abusive sexual conduct involving a minor or ward[.]”
The defendant in Lockhart was convicted of violating the federal statute, and he had a prior state-law conviction for sexually abusing an adult woman. The question before the Court is whether he is subject to a sentence of up to ten years, or a sentence somewhere in the range of ten to thirty years.
The answer to that question turns on whether the phrase involving a minor or ward modifies only the noun phrase that it immediately follows (abusive sexual conduct) or instead modifies the larger noun phrase in which the smaller noun phrase is embedded (aggravated sexual abuse, sexual abuse, or abusive sexual conduct). The latter interpretation is the one advocated by the defendant. Since his prior conviction was for sexual abuse of an adult, he will be not be subject to the increased sentence if such a sentence is held to be appropriate only if applies only where the prior conviction was for aggravated sexual abuse involving a minor or ward, sexual abuse involving a minor or ward, or abusive sexual conduct involving a minor or ward. The government, on the other hand, naturally argues for the narrowest possible scope for the minor-or-ward provision, so that the increased sentence will be available in all cases where the prior conviction was for aggravated sexual abuse, sexual abuse, or abusive sexual conduct, not just those involving a minor or ward.
IN LINGUISTIC TERMS, THE ISSUE in Lockhart can be rephrased as whether the statutory language has the structure shown below in Figure 1, as the government contends, or the structure in Figure 2, as the defendant contends. (For some background about this kind of ambiguity, see my previous post here.)
For reasons that I discussed here, research in psycholinguistics suggests that people reading the statute would be more likely to understand involving a minor or ward as modifying only abusive sexual conduct. That research is summed up in an article (p.43) in The Handbook of Psycholinguistics:
One structural factor about which there is striking agreement among researchers is recency: People prefer to attach a new phrase to a more recent than a less recent attachment site.
However, that preference is just one of out of a number of factors that influence readers’ and listeners’ understanding of syntactically ambiguous sentences. And the recency preference’s influence in relation to that of the other factors will vary from one sentence to another. Subject to that caveat, let me explain what the research I’ve referred to means in the context of the statute at issue in Lockhart.
The first thing you need to understand is structure of the phrase aggravated sexual abuse, sexual abuse, or abusive sexual conduct, which I’ve isolated in the tree diagram in Figure 3, below. As that diagram shows, the overall noun phrase (NP1) that is composed of which three smaller noun phrases (NPa, NPb, and NPc):
As you read you read the statute, your mind processes the sentence incrementally, building up a mental representation of what the text you are reading means, without any perceptible delay between perceiving the words and understanding them. The tree diagram in Figure 3 is a visual metaphor for the mental representation of NP1. After you have mentally processed NP1, your brain processes the modifying phrase involving a minor or ward—building a mental representation of that phrase and simultaneously integrating that representation with the existing representation of NP1. That integration process is described with another visual metaphor: the incoming phrase is described as “attaching” to the existing portion of the diagram. In particular, it is described as attaching to one of the places where two of the diagram’s branches meet (called a “node”). These nodes provide what are referred to in the quote above as “attachment sites.” In the case of NP1, there are two nodes to which involving a minor or ward can attach: the node that is labeled NP1 (corresponding to the defendant’s preferred interpretation) or the one labeled NPc (corresponding to the government’s interpretation). Of these two nodes, the “more recent” one is NPc, because it arose later in time than the NP1 node.
So to say that people “prefer to attach a new phrase to a more recent than a less recent attachment site” means that comprehenders—people listening to someone talk or reading a text—tend to understand what they hear or read in a way that can be portrayed on a tree diagram as the incoming item attaching to the most recent available attachment site. As a result, the comprehender will, in the absence of some countervailing factor, be more likely to understand involving a minor or ward as modifying only NPc rather than NP1.
EACH OF THE TWO STRUCTURES I’ve set out above corresponds to the interpretation that would result from applying one or the other of the competing canons.
The interpretation corresponding to the structure in Figure 1, in which involving a minor or ward modifies only abusive sexual conduct, and which is favored by both the recency preference and the government, is the interpretation that would result from applying what is generally known as the Rule of the Last Antecedent. This rule (which is perhaps better thought of as an interpretive strategy) originated in the maxim ad proximum antecedens fiat relatio nisi impediat sententia (“Let reference be to the nearest antecedent, unless the meaning hinders”), which appeared in English case reports at least as long ago as the early 1600s. [Update: I’ve now traced the “proximum antecedens” maxim back to a case report, written in law French, from 1431.)]
Over the years, the rule has been formulated in a variety of different ways; this one (from the first edition of Black’s Law Dictionary edited by Garner) is fairly typical:
An interpretive principle by which a court determines that qualifying words or phrases modify the words or phrases immediately preceding them and not words or phrases more remote, unless the extension is necessary from the context or the spirit of the entire writing.
The fact that this legal principle yields results generally consistent with the findings of modern psycholinguistics is rather mindboggling. But let’s get one thing clear. Although the Rule of the Last Antecedent is sometimes described as a rule of grammar, it is nothing of the sort. Neither of the competing interpretations at issue in Lockhart is ungrammatical. The Rule of the Last Antecedent is nothing more than a legal principle that happens, as a matter of blind luck, to do a decent job of generating interpretations that are linguistically defensible.
Although the quote above from Black’s Law Dictionary is fairly typical of how the Rule of the Last Antecedent has been stated in the U.S. for more than a century, in Reading Law the authors decided to split the rule into two separate canons. The first one, which they call “the Last-Antecedent Canon,” applies only to pronouns, relative pronouns, or demonstrative adjectives, and it provides that such words “generally refer to the nearest reasonable antecedent.” That canon is irrelevant to Lockhart, in which the language at issue doesn’t consist of those kinds of words. (From here on, I will refer to this canon by the name Scalia and Garner use, and will use the term “Rule of the Last Antecedent” to refer to the principle that is defined in the quote from Black’s.)
The other canon that Scalia and Garner extracted from the Rule of the Last Antecedent is one that they call the “Nearest-Reasonable-Referent Canon.” As they frame it, this canon applies to “prepositive or postpositive modifiers” (i.e., modifiers that come either before or after item(s) that they modify—meaning all modifiers), but only when “the syntax involves something other than a parallel series of nouns or verbs[.]”
This formulation presents several problems.
First, the exclusion of cases in which “the syntax involves a parallel series of nouns or verbs” is ambiguous. Is the phrase a parallel series of nouns or verbs limited to only nouns or verbs, without any subsequent modifiers or complements, as in example (1), or does it also include noun and verb phrases, as in (2)?
Nouns: cities, counties, and states
Verbs (in passive voice): used, possessed, or transferred
Noun phrases: aggravated sexual abuse, sexual abuse, or abusive sexual conduct [Note that a noun phrase can also consist of just a noun: cities, the county]
Verb phrases: assault another person or threaten to do so
As you might have suspected when you read the noun-phrase examples in (2), this ambiguity could play a role in Lockhart if the Court decides to apply the Reading Law framework. In that event, the outcome of the case might turn on how the ambiguity is resolved. Under one interpretation of the Nearest-Reasonable-Referent Canon (the one in which the phrase a parallel series of nouns or verbs also covers noun phrases and verb phrases), the canon would not cover the statute at issue in Lockhart. The statute would instead be covered by the Series-Qualifier Canon, which is discussed below. That would be what the defendant advocates. But if the phrase a parallel series of nouns or verbs is interpreted NOT to include noun phrases or verb phrases, then the statute WOULD fall be covered by the Nearest-Reasonable-Referent Canon. That would yield the same outcome as would result from the Rule of the Last Antecedent, which is what the government advocates.
If you’re starting to get confused, I don’t blame you. We are at the point in the discussion where, in order to use Reading Law‘s canons to resolve a statutory ambiguity, we need to first resolve an ambiguity in the canons. Are there such things as Canons of Canonical Interpretation?
Probably not, I’m afraid. So if the Supreme Court decides to follow the Reading Law framework, and therefore has to resolve the ambiguity on its own, what should it decide? The answer seems to me to be a no-brainer. To treat a series of nouns or verbs differently than a series of noun phrases or verb phrases would be utterly arbitrary. It would draw a distinction that wouldn’t be supported by any conceivable rationale, either legal or linguistic. So the reference in the Nearest-Reasonable-Referent Canon to a parallel series of nouns or verbs should be interpreted as a parallel series of noun phrases or verb phrases.
Hell, the Court shouldn’t just interpret it that way—it should rewrite the damn thing. Reading Law is just a book, not a statute. So even a hard-core textualist should have no compunction about doing that.
Of course, this assumes that that the Court is going to follow the Reading Law framework. Before it decides whether to do that, it might want to consider the other issues that the framework presents. So let’s turn to…
THE SECOND PROBLEM WITH the Nearest-Reasonable-Referent Canon. Although Bryan Garner has described the canons in Reading Law as all being “well established and…frequently applied,” this one actually reflects two major changes from how the Rule of the Last Antecedent has been formulated and applied.
The first of these changes is the ambiguous limitation in the canon’s scope, which we have just looked at. That limitation (whatever its scope might be) is at odds with the pre-existing law.
To begin with, it is inconsistent with cases that describe the Rule of the Last Antecedent as applying to precisely the kinds of constructions that the Nearest-Reasonable-Referent Canon excludes. For example, the Utah Supreme Court has said, “The [last antecedent] rule provides in general terms that when there is a modifier following a series of nouns, the modifier will apply only to the immediately prior antecedent[.]” Similarly, there are cases in which (although there is no similarly explicit statement) such constructions have been analyzed under the Rule of the Last Antecedent—something that would have been unnecessary if they were not within the rule’s scope. (For example, this case, this one, and this one.)
Moreover, I have run a number of Westlaw searches looking for evidence suggesting that the exclusion has ever been part of the Rule of the Last Antecedent, and I have come up dry. It’s of course possible that I’ve missed something; the Rule of the Last Antecedent has apparently been cited in more than 1,000 U.S. cases. But consider this. Before Reading Law was published, none of the three editions of Black’s Law Dictionary that were edited by Garner (the 7th, 8th, and 9th) said anything to suggest that the Rule of the Last Antecedent has the limited scope that Reading Law attributes to the Nearest-Reasonable Referent Canon. And nowhere in Reading Law is any authority cited to explain or justify the limitation in scope.
Taking all of this together, it is reasonable to conclude that the restriction on the scope of the Nearest-Reasonable-Referent Canon has no basis in the prior caselaw.
And that’s not all. The Nearest-Reasonable-Referent Canon also reflects yet another change from the prior law. The Rule of the Last Antecedent concerned modifiers that appear after the words or phrases that they modify. But the Nearest-Reasonable-Referent Canon applies both to modifiers that follow the items they modify and those that precede them. In other words, it applies to all modifiers—subject, of course, to the carve-out of parallel series of nouns and verbs.
As is the case with respect to that carve-out, Reading Law makes no effort to explain or justify this change from the existing caselaw; it doesn’t even acknowledge that there was a change in the first place. This won’t be at issue in Lockhart, since the modifier there comes after the item it modifies. But this change, along with the others, nevertheless cries out for an explanation.
SO MUCH FOR LAST ANTECEDENTS and Reasonable Referents. What about the Series-Qualifier Canon, on which the defendant relies?
Before Reading Law came out, there was no such thing as a Series Qualifier Canon. There was, instead, a 1920 Supreme Court case named Porto Rico Railway, Light & Power Co. v. Mor, in which the Court said this:
When several words are followed by a clause which is applicable as much to the first and other words as to the last, the natural construction of the language demands that the clause be read as applicable to all.
In his book The Language of Judges, Larry Solan called this “the across-the-board-rule,” and that’s as good a name as any, so I’ll use it, too. Neither this rule nor the case where it originated is cited in Reading Law. However, I speculated in one of my earlier posts about Reading Law that this rule was the source of what became the Series-Qualifier Canon. And since that time, pretty much everyone who has discussed the canon has assumed the same thing. So what I’m calling the across-the-board rule is often referred to now as the Series-Qualifier Canon. But I’m sticking with calling it “the across-the-board rule,” because what the Court said in Porto Rico Railway, Light & Power Co. differs substantially from what Reading Law calls the Series-Qualifier Canon.
I’ll get to the Series-Qualifier Canon in a moment, but first I want to say this about the across-the-board rule: It is completely circular. It applies “[w]hen several words are followed by a clause which is applicable as much to the first and other words as to the last.” But whether the clause is equally applicable to all the words is precisely the interpretive question that needs to be answered. The rule itself provides no guidance, which suggests that its only function is to rationalize a decision that was made for some other reason. [On edit: See update here.]
With that in mind, it’s easy to see why, if the Series-Qualifier Canon is in fact derived from the across-the-board-rule, the former looks nothing like the latter. This is the Series-Qualifier Canon as set out in Reading Law:
When there is a straightforward, parallel construction that involves all nouns or verbs in a series, a prepositive or postpositive modifier normally applies to the entire series.
This formulation does an admirable job of avoiding the across-the-board rule’s empty circularity. It provides a clear and easy-to-follow algorithm for deciding on the provision’s presumptive meaning. But it’s not enough just to have a rule that is easy apply. There has to be—well, there ought to be—some justification for the particular rule that you decide on. And that’s where the Series-Qualifier Canon falls short.
Reading Law doesn’t try to justify the Series-Qualifier Canon as a restatement of existing caselaw. And I doubt that the canon could be justified on that basis. As far as I have been able to determine, no principle similar to the Series-Qualifier Canon had ever been mentioned until Reading Law was published. [On edit: This statement is too broad. As I explain here, there is caselaw support for the principle expressed in the canon, with respect to modifiers that precede the items they modify. However the statement remains valid for modifiers that follow those items.]
Nor does Reading Law offer any linguistic justification for the generalization stated in the Series-Qualifier Canon. Contrary to what the defendant argues in Lockhart (p. 11 of his brief), the canon does not express “a familiar principle of syntax.” It doesn’t even express an unfamiliar principle of syntax. And from the standpoint of psycholinguistics, the Series-Qualifier Canon is a mixed bag.
In one of my previous posts, I considered whether the interpretations generated by the canon would generally be consistent with the interpretations that one would expect based on the results of psycholinguistic research. What I concluded was that the answer would be “yes” in some cases and “no” in others. With regard to modifiers that precede the item(s) that they modify, applying the canon would yield results consistent with the recency preference that I discussed above. But in cases involving modifiers that follow the item(s) they modify, the results would be inconsistent with that preference.
This raises the question of what would be the net effect of applying the canon to the overall universe of relevant cases. Would results consistent with the recency preference outnumber those inconsistent with it? Or vice versa? Or would they balance each other out? The answer would presumably depend on whether the instances of premodification outnumber the instances of postmodification—something about which I don’t have the slightest idea.
THE SERIES-QUALIFIER CANON ALSO presents two other problems. One is that it contains the same ambiguity as the Nearest-Reasonable Referent Canon: the type of construction to which it it applies is described as “a parallel construction that involves all nouns or verbs in a series[.]” So once again, we are faced with the question whether nouns or verbs should be understood to mean noun phrases or verb phrases.”
The second problem, which is probably more serious, is that while the Series-Qualifier Canon generates different results than the Last-Antecedent Canon, the coverage of the two canons overlaps. The Series-Qualifier Canon covers parallel constructions consisting of nouns or verbs (or, I assume, noun phrases or verb phrases), and that category includes parallel constructions that are modified by relative clauses. And relative clauses, in turn, generally begin with relative pronouns and therefore come within the Last-Antecedent Canon. In fact, two such constructions are provided as examples illustrating the Series-Qualifier Canon (Reading Law pg. 148):
Institutions or societies that are charitable in nature
A wall or fence that is solid
[Contra The Cambridge Grammar of the English Language, Scalia and Garner treat that as a relative pronoun; see Reading Law, pg. 142. And in any event, the nouns are appropriately regarded as the relative clauses’ antecedents (although linguists would probably use different terminology ).]
In these examples, the Series-Qualifier Canon would presumptively require the italicized modifiers to apply to both of the nouns in the sentence, while the Last-Antecedent Canon would presumptively require them to refer only to the second noun.
Reading Law says nothing about this conflict, and presumably the authors didn’t realize the conflict existed. The book therefore provides no help in deciding which of the two inconsistent canons to apply.
Here’s what I think should be the solution: when both canons apply, follow the Last-Antecedent Canon. And this is why. In the cases falling under both canons, applying the Last-Antecedent Canon is more defensible linguistically. Moreover, the caselaw support for Last-Antecedent Canon is much stronger than that for the across-the-board rule or the Series-Qualifier Canon. My Westlaw research indicates that in the time since the across-the-board rule was first stated in Porto Rico Railway, it has been cited in U.S. cases less than 10% as often as the Rule of the Last Antecedent was cited during that same period (97 times as opposed to more than 1,100). As for the Series-Qualifier Canon: it was first stated in 2012, in Reading Law, and before that it didn’t exist.
However, the more fundamental question is whether the courts should follow the Reading Law canons at all. Given the morass of complexities that following them would entail, the smartest course might be to Just Say No.