The textualist fallacy: trying to impute meaning to "of"

I've promised an example of the textualist fallacy in legal scholarship. Recall that the SCAT hypothesis holds that the components of meaning of a legal text (and all other texts, really) are structure, context and text, while the textualist fallacy happens when one assumes that meaning comes only from text.

Since today is oral argument day in McDonald v. Chicago, I'll illustrate with an example from commentary on the case. In McDonald, the U.S. Supreme Court will decide whether the Second Amendment's individual right to keep and bear arms is incorporated against the states under the Fourteenth Amendment, or whether it is restricted to the right against the federal government that was found in D.C. v. Heller.

A bit of background explanation may be in order, for readers who know even less constitutional law than I do. Besides the equal protection clause, ection 1 of the Fourteenth Amendment, passed in 1868, contains two clauses that would seem to restrict the ability of states to interfere with individual freedoms of citizens: the privileges or immunities clause (“No state shall make or enforce any law which shall abridge the privileges or immunities of citizens of the United States”), and the due process clause (“nor shall any state deprive any person of life, liberty, or property, without due process of law”). It is generally agreed that the effect of these is to create rights against the states parallel to the rights that exist against the federal government, which is known as incorporating those rights against the states.

Not all rights against the federal government are incorporated against the states. For example, the Seventh Amendment's right to a jury trial has been held not applicable against the states. Other rights have not been adjudicated yet, most saliently the right to keep and bear arms, which was held less than two years ago to include an individual right to possess weapons, in District of Columbia v. Heller. The defendant in that case was the federal government, because the District of Columbia is not a state. McDonald will decide whether one of the Fourteenth Amendment's incorporation clauses protects this right against the states as well.

This is a pretty big question, but McDonald is actually bigger, because the Supreme Court will consider both Fourteenth Amendment clauses as bases for incorporating the right, instead of just the Due Process clause, as it usually does. This is seen by many as an opportunity to correct a historical wrong in the interpretation of the clauses. The privileges or immunities clause may seem, to the ordinary reader, to be more powerful than the due process clause, because it prohibits state interference with substantive rights of citizens, while the due process clause appears to refer only to procedural rights such as the right to a fair trial. In reality, it has almost always been the due process clause that has been used to incorporate rights against the state, including substantive rights. The privileges or immunities clause, on the other hand, has hardly been held to protect anything at all.

Many people are irked by the court's failure to give effect to the privileges or immunities clause, and many of the same and others are irked by the court's creation of substantive rights under the due process clause, where the Constitution seems to only state procedural rights. It is therefore hoped by many that the Supreme Court will “resurrect” the privileges or immunities clause as a source of substantive rights against the states, and perhaps take the offending substance out of the due process clause.

In anticipation of the Court's discussion of the privileges or immunities clause, commentators are discussing how this clause might and should be interpreted. In this connection, Christopher R. Green wrote McDonald v. Chicago, the Meaning-Application Distinction, and “Of” in the Privileges or Immunities Clause, which falls into the textualist trap by claiming that the controversy about the meaning of the privileges or immunities clause is a controversy over the meaning of the word “of.”

What is the controversy over the meaning of the privileges or immunities clause? Green presents the following as some of the possible (and plausible) interpretations of “privileges or immunities of citizens of the United States.”

a. Rights “'which owe their existence to the Federal government, its National character, its Constitution, or its laws.'”

b. Rights “[p]ossessed under the Constitution against the federal government by” citizens of United States.

c. Rights “[g]enerally possessed under state constitutions, statutes and common law by” citizens of the United States.

d. Rights “[g]enerally possessed in 1868 under state constitutions, statutes and common law by” citizens of the United States.

e. Rights “[p]ossessed as a matter of natural right by” citizens of the United States.

f. Rights “[p]ossessed locally by” citizens of the United States.

The controversy is therefore not about whether “privileges or immunities” refers to something other than rights; it is over what “rights” means in this context, and specifically, in Green's words, “the relationship that a right must bear to 'citizens of the United States' to count.” He then continues, “[W]hich is to say, over the meaning of 'of',” which, I argue, is where he errs.

Green's argument can be presented as follows:

1.The phrase “privileges or immunities of citizens of the United States” has several plausible interpretations, and its meaning is controversial.

2.Therefore it is ambiguous.

3.If a sentence is ambiguous, one of its components must be ambiguous.

4.The meaning of “privileges or immunities” is uncontroversial. It means “rights.”

5.The meaning of “citizens of the United States” is also uncontroversial.

6.The only leftover text is “of.” Therefore the ambiguity is in the word “of.”

Green thus falls into the textualist trap by using a sensible-seeming principle: when you have eliminated the uncontentious, whatever remains, however semantically empty, must be the source of contention. The problem is that eliminating all the text but one word doesn't just leave the one word. Under the SCAT hypothesis, It leaves one word plus all of the invisible, non-text components of meaning: the structure and context.

I would locate the ambiguity not in the word “of” but in the context. “Right” is a modal word, and modals are notoriously dependent on context for their meaning. “Right” is modal because whether something is a right or not must be resolved by reference to possibility, rather than the actual world. You can't tell if something is a right by determining whether someone does it, since rights don't have to be exercised. Philosophers and semanticists model modals using possible worlds, so that something is possible if it happens in some possible world with the relevant characteristics, and something is necessary if it happens in every possible world with the relevant characteristics. The indeterminacy of the privileges and immunities clause arises from what the relevant characteristics are in the set of possible worlds that we consider when evaluating the word “right.”

That may sound circular, but it is actually not. The semantic meaning of the word “right,” if I'm theorizing it correctly, is a function which (in essence) takes as an input an individual X and an action E, and returns the value “TRUE” if and only if there is a world W in which X does E, and W is in a certain set of possible worlds. The particular meaning of “right” will emerge from the choice of sets of worlds. Green's meaning 1, above, might pick out the set {w1, w2, w3}, where the w's represent possible worlds, while his meaning 2 might pick out {w1, w2, w3, w4}. The word “rights” doesn't show up at this level - it is inherent in the set; and the different possible meanings for “right” are the criteria by which the membership in the sets is picked out.

The alternative, that “of” is many ways ambiguous, just doesn't work. We all have the intuition that “of” has little, if any, semantic content. I know of no evidence that “of” is modal, so it is not possible to blame the indeterminacy of the privileges or immunities clause on a function that selects possible worlds for placement in a set, as we can do with “right.” If “of” means anything, it might mean something like “a function from a whole to the set of its parts,” which at least accounts for one type of phrase containing “of” - the type seen in “three of the boys.”

More likely, “of” is just like the pleonastics “it” and “there,” which are believed to carry no meaning in sentences like “it is better to have loved and lost than never to have loved at all,” and “there lived a country boy named Johnny B. Goode.” “Of” is probably just a placeholder like the other pleonastics, which must be vocalized for syntactic or morphological or phonological reasons, but which contribute no meaning of their own.

None of this is to pick on Green, who appears, unlike some jurists who write about language, to have some sophistication in the subject. Nor is it to disagree with his conclusions about what the plausible interpretations of the privileges or immunities clause are, or about his conclusion of which interpretations are the best options – topics I am not competent to broach. I wrote this post simply to illustrate the textualist fallacy, and I used Green's paper as illustration because I happened to have recently read it. And the fact that someone like Green can fall for the textualist fallacy suggests that everyone should be on the alert for it.

The linguistics of God

Is God a linguistic object? The results of two very brief explorations suggest different conclusions.

In one of his pornolinguistic papers, James D. McCawley, writing as Quang Phuc Dong of the South Hanoi Institute of Technology (beating Accepted to the joke by about 40 years) explores subjectless sentences like "fuck you" and "damn you." He argues convincingly that they are not imperatives, but is unable to advance a meaningful alternative analysis.

One of the alternatives that he considers and discards posits that such sentences have God as the subject in their underlying representations, followed by a deletion transformation creating the appearance of subjectlessness in their surface representations. Like so:

UR: God fuck you.
DT: God --> Ø /subject position
SR: Fuck you.

Besides accounting for the absence of a subject, this analysis has the advantage of being acceptable to both atheists and theists. Atheists can call it "God-deletion," while theists can invoke the hidden hand of God.

Ultimately, this analysis doesn't work. Besides not being very explanatory, there is the fact that you cannot say *"Fuck himself" to mean "fuck God.," but you can say "Fuck God" to mean "Fuck God." This should not be possible if God is the hidden subject.

So subjectless sentences are not good evidence for God. But there is better evidence from Hebrew negative imperatives.

As I think I pointed out in my masters thesis (I haven't organized my books since I moved last month, and finding the thesis would be a hassle), the normal way to express negation in Hebrew suppletive imperatives is with the use of the negative al:

(1) al tircax oto!
neg murder.2.masc.sg.fut him.DO
"Don't murder him!"

However, if God is issuing a commandment, the proper form of the negative is lo:

(2) a. lo tircax!
b. ?al tirtzach!
"Thou shalt not kill!"

The sentences with al is shown with a question mark to indicate that the utterance, while not unacceptable, is ungodly. That is, if God uttered (2b), it would be interpreted as carrying less than the full authority and timelessness of one of God's commandments.

The pattern in (1)-(2) is decent prima facie evidence of the existence of God, at least as a morphological phenomenon. I propose to formalize this by subscripting either [+God] or [-God] to morphemes, depending on whether they are or are not godly. Thus al would be represented

Neg_{[+imp, -God]}

while lo would be represented

Neg_{[+imp, +God]}.

Can a non-God speaker use lo? My intuition as a semi-native speaker is that this is possible, but it would be interpreted as a "godly" statement. So for example an emperor might use lo to issue an edict, but it would be seen as extremely arrogant language, such as that of a megalomaniac who thinks he is as great as God. It would not be used by someone like a judge or a democratic political figure, even if it the statement was completely authoritative.

Can God use al? Yes. If God was speaking to a person in a more private capacity, rather than issuing universal commandments, Gods would use al. In fact, he did so all the time in the Bible. So for example, when God played a practical joke on Abraham, and had him bind Isaac to an altar and get ready to sacrifice him, he used al to tell Abraham not to harm Isaac after all. The key to godly language is that it expresses a universal, principled prohibition, not just a particular "don't do this."

I know of no evidence of the existence of God in nonimperative morphology. There are many mysteries in the morphology of imperatives, both in Hebrew and more universally. Whether the presence of God in imperative morphology advances linguistic research remains to be seen. But it sure is fun to think about.

Global Language Map

Here’s a nice-looking, if very coarse-grained, map of the languages of the world. It shows the most widely spoken languages, but continent. I’m posting it mostly because I had not been aware, until I saw the data on the map, just how much less linguistically diverse North America is than any other continent shown.

But I guess that result is only due to the fact that Australia is not counted as a separate continent, but rather as part of “Asia and Pacific,” so that its overwhelmingly English-speaking population is drowned out by the languages of China, Indonesia, Japan and Indonesia.

Shooting into a structure from within it

The Statutory Construction Blog points out this Pennsylvania opinion which asks whether one can shoot into a building that one already occupies. The court holds that one can't, since "the plain meaning of the term “into” requires that the original location is outside of the destination."

The statute in question declares the following:

A person commits an offense if he knowingly, intentionally, or recklessly discharges a firearm from any location into an occupied structure.

Unfortunately, the court resorts to looking up "into" in the dictionary, but in this case the practice did not lead the court to the wrong result.

The government's main argument was that "from any location" is so broad that it must include any place inside the occupied structure.

I think the court got it right. If "from any location" is taken literally, we have a contradiction between this phrase and "into," which either makes the law self-contradictory, or requires that the law be constructed or interpreted in a non-contradictory way. In this case, there is no need for the court to construct a rule, because pragmatic principles of interpretation plainly allow us to limit "from any location" to mean "from any location outside the structure." The domain of "any" is more flexible than the requirement that the path described by "into" start outside of the locational reference point.

Transcript from Flores-Figueroa

Here. Argument recap here.

What is Language and Law?

Peter Tiersma asks: What is Language and Law? And Does Anyone Care?

I care! The paper is from the volume LAW AND LANGUAGE: THEORY AND SOCIETY, which I don’t know much about, because the product descriptions online are in a foreign language.

Solum on Possibility and Necessity in Law

Prof. Solum suggests that arguments about legal possibility and necessity can be sharpened by using the possible worlds approach to modality. Quite possibly so. The possible worlds approach is fascinating (to me, at least) and it's useful philosophical knowledge in any case. In my note, I argued that it was Justice Scalia's failure to attend to the difference between universal and existential modality (the feature that distinguishes possibility from necessity) that led him to reason wrongly about the meaning of the Second Amendment in D.C. v. Heller.

Another Linguists' Brief

The Supreme Court has agreed to hear Flores-Figueroa v. United States. Oral argument is set for next Wednesday, Feb. 25th. As in Hayes, there is a linguists' amicus brief on behalf of neither party (authored by the same attorney). Amici this time are Tom Ernst, Georgia Green, Jeffrey Kaplan and Sally McConnell-Ginet, and the subject matter is adverbs. In particular, the interpretation of the phrase knowingly transfers, possesses, or uses, without lawful authority, a means of identification of another person, and whether a means of identification of another person is in the scope of knowing. The conclusion: unambiguously yes, unless the reader adopts the not very natural de re reading.

By the way, oral argument in Hayes was full of linguistic discussion. Also, Chief Justice Roberts learned a new word, "romanette."

Summary of my second amendment paper

I've made some promises to some people about trying to summarize my paper about the linguistics of the second amendment. Here is my attempt. The full paper is available here (the website makes you wait 90 seconds before downloading, since I signed up for a free membership; the delay is their way of motivating people to get paid memberships).

The issue is as follows: In the Supreme Court case of D.C. v. Heller (discussed previously on this blog here, here and here), Justice Scalia, author of the majority opinion, and Justice Stevens, author of one of the two dissents, disagree over whether the absence of the word to, bracketed in the following, is semantically significant.

The right of the people to keep and [to] bear arms shall not be infringed.

Justice Stevens reports the intuition that the absence of to supports the conception that the amendment does not protect two separate rights, but rather a single right,

a right to have arms available and ready for military service, and to use them for military purposes when necessary.

Justice Scalia disagrees:

JUSTICE STEVENS resorts to the bizarre argument that because the word "to" is not included before "bear" (whereas it is included before "petition" in the First Amendment), the unitary meaning of "to keep and bear" is established. Post, at 16, n. 13. We have never heard of the proposition that omitting repetition of the "to" causes two verbs with different meanings to become one. A promise "to support and to defend the Constitution of the United States" is not a whit different from a promise "to support and defend the Constitution of the United States."

Both Stevens and Scalia appear to be correct in the judgments that they report, which means that as far as interpreting the Constitition, Stevens is correct and Scalia is incorrect. Scalia makes the error of dismissing an English speaker's intuition about the language of the second amendment by theorizing about linguistics. The purpose of my paper is to show why Scalia's reasoning about linguistics is wrong.

The task, then, is to identify the property of the sentences that makes the presence of to significant in the Constitution, but not significant in Scalia's example sentences. I argue that of the various differences between the sentence pairs (such as the definiteness or indefiniteness of the phrase, and the definiteness and plurality of the object [arms vs. the Constitution of the United States]), the key difference is between right and promise, and the important difference between these two modal nouns is the quantificational force - the fact that right denotes an existential modal noun while promise denotes a universal modal noun.

That the difference is due to right and promise is evidenced by the following sentence pairs, where the sentences are to the extent possible held constant except for right vs. promise, and the presence vs. absence of to.

Right:

a. Workers have the right to unionize and strike

b. Workers have the right to unionize and to strike

Promise:

a. The workers made a promise to unionize and strike

b. The workers made a promise to unionize and to strike

In the case of right, there is a clear truth-conditional difference between the sentences. The (a) sentence does not seem to be asserting an absolute right to strike; rather, the right to do so is contingent on their first having unionized. In the (b) sentence, the right to strike is unconditional.

In the case of promise, there is no such difference. In both (a) and (b) there are two independent promises. If the workers made the promise in (a) and then broke their promise to unionize, their commitment to strike still stands.

So the question is why right is different from promise. As I've suggested, my approach is that it has to do with the quantificational difference between the terms. Right is an existential modal: X has a right to do R if in some possible world consistent with X's rights, X does R. Promise is a universal modal: if X has made a promise to do P, then X does P in every world consistent with X's commitments.

What difference does the quantificational force make? Well, consider the following truth conditions.

a. "A has the right to keep and bear arms" is true iff in some world w in the set W, where W is the set of worlds consistent with A's rights, A keeps arms in w and bears arms in w.

b. "A has the right to keep and to bear arms" is true iff in some world w in the set W, where W is the set of worlds consistent with A's rights, A keeps arms in w, and in some world w' in the set W, where W is the set of worlds consistent with A's rights, A bears arms in w'.

These are different truth conditions. (a) unifies the keeping of arms and the bearing of arms by requiring them to take place in the same possible world, while in (b) they can occur in separate worlds, and are therefore independent rights. But if the quantificational force is existential, the truth conditions come out logically equivalent:

a. "A has made a promise to support and defend the Constitution" is true iff in every world w in the set W, where W is the set of worlds consistent with A's promises, A supports the Constitution in w and A defends the Constitution in w.

b. "A has made a promise to support and to defend the Constitution" is true iff in every world w in the set W, where W is the set of worlds consistent with A's promises, A supports the Constitution in w, and in every world w' in the set W, where W is the set of worlds consistent with A's promises, A defends the Constitution in w'.

This gets us the semantic difference between right and promise, and therefore accounts for our intuitions about the second amendment and Scalia's attempted counterexample. The rest of the paper aims to justify the truth conditions just shown, by arguing for syntactic structures and semantic analysis of these sentences.

In the syntax section of my paper, I argue that the presence or absence of to is probative of a substantial structural difference. So the reason that the right to keep and bear arms means something different from the right to keep and to bear arms is not the lexical semantics of the word to, but the fact that in the former sentence it is just the verbs keep and bear that are conjoined, while in the latter, the full infinitival clause is conjoined. The conjunction is indicated in the following:

a. The right of the people PRO to [keep and bear] arms
b. The right of the people [PRO to keep arms and PRO to bear arms]

[Explanations: PRO is the invisible subject of the infinitival clause; the first instance of "arms" in (b) is phonetically deleted by a process that I can only guess at - it's not a fantastic analysis, but it beats the alternatives.]

Given this syntactic analysis, the logical forms are determined by the semantics as follows. Keep and bear each denote functions from entities (their direct object) to sets of events. So in the sentence John kept his house, the function denoted by keep takes as an argument the entity John's house, and returns the set of events of keeping John's house. I think the semanticists are calling this the neo-Davidsonian or the semi-Davidsonian approach to verb meanings - I've always had trouble keeping these terms straight.

In (a), keep and bear are conjoined. The semantic result is that the truth conditions of each are required. So the meaning of the conjunction is a function from an entity (the direct object of the conjoined verb) to a set of events - the set of events in which the object is both kept and borne.

To takes a set of events as input, and returns a function from entities to sets of worlds in which the entity is the agent of the appropriate type of event in the world. For example, in PRO to keep and bear arms, to takes as its first input the set of events of keepoing and bearing arms, then takes PRO as its entity argument, and returns as the meaning of the infinitival clause the set of worlds in which there is an event of keeping and bearing arms of which PRO is the agent. Ultimately, the truth condition of (a) makes the people the referent of PRO, and compares this set of worlds with the set of worlds consistent with the rights of the people.

The semantic analysis of (b) is more complicated, and requires some creativity of analysis. The meanings of PRO to keep arms and PRO to bear arms are determined straightforwardly - they denote the sets of worlds in which there are events of keeping or bearing arms, respectively, with PRO as the agent of the events.

The challenge is dealing with the conjunction. The ordinary approach to conjoining two sets of worlds would be to create a new set of worlds such that the conditions of both of the sets are met. In this case, the result would be the set of worlds that contain an event of PRO keeping arms and an event of PRO bearing arms. The rest of the semantic analysis would be the same as for (a): PRO would be assigned the meaning "the people," and the set would then be compared with the set of worlds consistent with the rights of the people.

The problem is that this does not match our intuition about the meaning of the sentence. We tend to interpret the right to keep and to bear arms as equivalent to the right to keep arms and the right to bear arms, yet this is not the result of the most straightforward semantic analysis.

The problem seems to be this: in conjunctions such as this one, the conjunction appears to take very wide scope, in effect to be conjoining entire propositions, even if the syntax makes it look like the conjunction is more local. In this case, it is only the infinitival clause that is conjoined syntactically, though semantically it is the propositions that are conjoined. My suggestion is that this syntax/semantics disparity takes place where the conjoined object is an argument rather than a function; that is, when the combinatory operation that the conjoined object participates in is one in which it is "fed into" the meaning of the adjacent object, as in (b), where the set of worlds denoted by the infinitive is one of the arguments of the (denotation of the) word right.
Not, as in (a), one in which the conjoined object takes the adjacent element as an argument, as in (a) where the conjoined verb takes the direct object arms as an argument.

What I propose, then, is that we need a different rule for interpreting conjunctions in situations in which the conjunction is an argument. What I suggest, without presenting a real formal analysis, is that whenever the semantics encounters a conjunction, it assigns it the meaning of the set of conjuncts. So in the case of keep and bear, the meaning is the set consisting of two functions from entities to sets of events, while in the case of PRO to keep arms and PRO to bear arms, it is the set consisting of two sets of worlds.

This set then has the option of either undergoing the traditional conjunction rule and combining with an argument, or of undergoing a new operation and feeding into a function. The new operation consists of creating two separate semantic derivations that are conjoined at the proposition level. So the two sets of worlds in (b) will each, independently, combine with right, yielding a proposition, and those two propositions will be conjoined.

I haven't seen such a procedure suggested before, but it looks like it can handle the syntax/semantics mismatch in the scope of coordination, and it does it without any "looking ahead" by the semantic derivation to see the future interactions of the coordinated object. The derivational mechanism always treats a conjunction the same way - it creates a set of the conjuncts; and what eventually happens to this set is determined by what possibilities exist at the time the conjunction goes to combine with an adjacent object.

Whew. Any questions?