Grammatical justice is served
« previous post | next post »
The following is a guest post by Jason Merchant.
Thought the LangLog would like to hear this week's update on the the Supreme Court case involving adverbial modification argued in February: all nine justices agree with the linguists! The decision is posted, but briefly, the money quote is:
"In ordinary English, where a transitive verb has an object, listeners in most contexts assume that an adverb (such as knowingly) that modifies the transitive verb tells the listener how the subject performed the entire action, including the object as set forth in the sentence."
It is so ordered…
J. W. Brewer said,
May 7, 2009 @ 11:41 am
Note also the separate opinion (at the end of the pdf) by Justice Alito, which may display a somewhat higher degree of sophistication from a linguistics standpoint. (The separate Scalia/Thomas opinion seems driven more by distinctive views about the judicial function and the legislative function than by distinctive views about syntax/semantics/pragmatics.)
Breffni said,
May 7, 2009 @ 11:48 am
The right decision in this case, but I think the view expressed in the judgement and in Jason's previous post (and the amicus brief) overgeneralises:
Two examples I gave before:
1. "You knowingly misled my mother" – addressed to an insurance salesman who didn't know me and didn't know that the lady he cheated yesterday even was a mother. In that context it corresponds to "You knowingly misled someone, who, by the way, happens to be my mother".
2. "One of the more serious violations occurred at Tabella's, where security staff at the front door knowingly allowed an undercover officer armed with a firearm to enter their licensed premises” (from here). This doesn't mean the bouncers knew they were admitting an undercover officer.
And a third, with a different adverb: "The driver deliberately ran over a father of three". In a newspaper report, this would not normally be read to mean the culprit sought out a father of three, or necessarily knew the victim was a father, or a father of three, or even male. It can mean just that he deliberately ran over someone.
In these cases, the particular descriptions selected for the objects are there for the hearer / reader's information. They aren't meant to correspond (or not necessarily) to the agent's state of knowledge or intentions.
Adrian said,
May 7, 2009 @ 12:05 pm
re: Breffni
I agree that the decision overgeneralises.
It would be odd to _use_ something without knowing you are using it but
it's perfectly possible to _mislead_ someone without knowing you are doing it.
The example with _allow_ is more complex, and it is not at all clear what the staff knew. The third example is bad journalese.
Breffni said,
May 7, 2009 @ 12:19 pm
Adrian, the second example ("knowingly allowed an undercover officer…") is clear enough in context: the culpable knowledge was the fact that the individual was armed, not that he or she was an undercover officer. In another context (the door staff were guarding an underworld den where guns are de rigueur), the problem might be the fact that they knew it was an undercover officer.
And the third example is certainly journalese (though I don't find it bad – it introduces information relevant to the audience through an NP, same as the other two), but that doesn't disqualify it as a counterexample.
Thomas Westgard said,
May 7, 2009 @ 12:23 pm
This is unquestionably a case of finding a way to justify what they already wanted to do. The US Supreme Court skips gleefully past all manner of other grammatical nuance when it suits their fancy to do so. Mind you, I'm not saying the law would be better if it were run by the grammar police. One hopes that, in this context, language is more a means to an end than an end in itself.
Chris said,
May 7, 2009 @ 1:59 pm
The US Supreme Court skips gleefully past all manner of other grammatical nuance when it suits their fancy to do so.
For example, the Alito concurrence identifies a line of cases holding that to “knowingly transpor[t] an individual who has not attained the age of 18 years in interstate or foreign commerce . . . with intent that the individual engage in prostitution, or in any sexual activity for which any person can be charged with a criminal offense” doesn't require knowing that the person is under 18, which seems pretty indefensible to me.
The use of "knowingly" to describe an action implies that it could have been done unknowingly; while you certainly can transport a person unknowingly (if they hide in your vehicle), you can't transport them *with an intent to involve them in prostitution* unknowingly; your own intentions cannot possibly be unknown to you, and to have intentions that involve the actions of another person implies being aware of that person. (The conjunction of action and intent generally implies purpose – in this case, that you are transporting them *in order to* facilitate their participation in prostitution.) Similarly, you generally can't transport anyone or anything across state lines without being aware you are doing so. Therefore, the only thing that the perpetrator reasonably *might* not know about the offense is the age of the person transported (since it's the only part of the act that is neither open and obvious, nor in the mind of the perpetrator); to read "knowingly" as not applying to that fact is to read it out of the statute altogether, something the courts are not supposed to do.
Unfortunately, given Alito's preconceptions, he uses the cases holding the opposite of the above argument as a reason to argue against a general rule that "knowingly" generally applies to all elements of an offense, rather than to dispute those specific constructions. (I support such a general rule; "knowingly" should be read broadly unless it is specifically limited, not only because that's how it's generally used but also because ambiguity in criminal statutes should be resolved in favor of criminalizing as little conduct as possible, and never in favor of criminalizing conduct whose status as legal or illegal is left in doubt by an ambiguous law. But some of these concerns go a little beyond linguistics.)
In this specific case, there's really no way the court could have ruled otherwise; making up a SSN and having it *purely by chance* duplicate the SSN of an actual person is nothing like "identity theft" and couldn't possibly have been intended to be included in it. Ignoring this would require ignoring not just grammar and usage, but obvious legislative intent as well. Indeed, I'm surprised the case went as far as it did.
"One of the more serious violations occurred at Tabella's, where security staff at the front door knowingly allowed an undercover officer armed with a firearm to enter their licensed premises” (from here). This doesn't mean the bouncers knew they were admitting an undercover officer.
To me, that's exactly what it means on first reading; I have to stop and consciously think about why that interpretation doesn't make sense and only then can I come up with a substitute interpretation. (It probably wouldn't be a "violation" to admit an armed person if the person admitting him/her knew he/she was an undercover officer, and even if it were, you wouldn't mention it in this thread.) If they really meant "security staff knowingly allowed a person armed with a firearm to enter, and that person turned out to be an undercover officer" then what you've quoted was very poorly written.
Collapsing two sentences into one can have significant pitfalls.
"You knowingly misled my mother" *can* have the same ambiguity, but in practice "my mother" is often treated not as a fact about her, but merely as a means of identifying her. "You knowingly misled a police officer" *does* imply that you knew the person being misled was a police officer.
scratchdaddy said,
May 7, 2009 @ 2:04 pm
Wow, that's the first time I've ever tried to read a Supreme Court decision. My mind grew foggy, and I skipped to the end hoping to find just what the decision was. I think they agreed with the defendant and not the government, which in my opinion is the right decision, but man! so much gobbledy-gook and obfuscation in their language. I know lawyers claim they write that way for clarity, but I think it's just to leave themselves wiggle-room for later argument. After all, argument means money to a lawyer. Does anyone else think that laws would be more understandable written in simpler English?
Neal Goldfarb said,
May 7, 2009 @ 6:41 pm
Breffini:
Your examples are not inconsistent with what Breyer's opinion said. In all three examples, the adverb modifies the full verb phrase, by which I mean the verb plus the complete direct object. For example in your first example, even if the sentence is interpreted as You knowingly misled X, where X is my mother, knowingly modifies misled X, not just misled.The same analysis applies to your example 3.
Similarly, the interpretation that you offer for your example 2 is security staff at the front door knowingly allowed a person armed with a firearm to enter their licensed premises, where that person was an undercover officer. Your statement that "the culpable knowledge was the fact that the individual was armed" shows that you interpret knowingly to modify the entire direct object (an undercover officer armed with a firearm) and not just the NP an undercover officer.
The interpretive issues you raise involve de re / de dicto ambiguities that affect, not the sentence's syntactic structure, but rather the interpetation of the direct object. The Supreme Court's decision, and in particular the sentence that's quoted in Jason's quote, deals only with the syntactic issue. The Court didn't discuss the de re / de dicto issue, and I suspect that nobody on the Court thought the statute was amenable to a de re interpretation.
BTW, regarding number 2, you might be interested in this statement from another account of the incident:
Jangari said,
May 7, 2009 @ 6:49 pm
Scratchdaddy, the purpose of deliberately obfuscating legalese is partly to take time; everyone gets paid more that way. But the far more important factor is the legal fraternity's exclusivity.
Law is just the art of interpreting legislation. In fact, 'interpret' is often collocated with 'law'. I still remember an example sentence from my days of learning formal logic: The supreme court cannot misinterpret the law. Imagine if laws were written in simple English. It'd be so unequivocal that no interpretation could happen. The need to a legal fraternity would disappear and no one would need to go to law school; everyone could be their own lawyer.
But laws are convoluted, complex and difficult to parse. Lawyers have to go to law school to learn the language. And who gets to go to law school? Those rich enough to get there; the children of the legal practitioners. Thus, the cycle continues.
In short, if life was a game of monopoly, law is learning the language in which the rules on the back of the box are written.
I know this is particularly broad a generalisation, and has numerous counter-examples, but I believe this is the general paradigm, and has been for hundreds of years.
John Baker said,
May 8, 2009 @ 1:13 am
Although there are exceptions, legal drafting in most cases seeks to achieve precision of meaning, to the exclusion of all other goals. So, yes, laws would be more readily understandable to the average person if they were written in everyday English, but they would also be more ambiguous.
If Jangari, or anyone else, really thinks that less interpretation would be needed if legal language were not used, then let him try rewriting a statute and see how much clarity he can achieve. This very case may be a good example. The statute provides: "Whoever, during and in relation to any felony violation enumerated in subsection c, knowingly transfers, possesses, or uses, without lawful authority, a means of identification of another person shall, in addition to the punishment provided for such felony, be sentenced to a term of imprisonment of 2 years." How would you rewrite it to make interpretation unnecessary?
Aaron Davies said,
May 8, 2009 @ 4:05 am
unfortunately i think the main thing that can be done that statute to fix it is to make it longer. given the apparent abundance of DAs determined to write mens rea out of the law, and the frequent support they receive from judges (e.g. in the "transporting minors" cases cited), "knowingly" is no longer adequate. perhaps "transfers, possesses, or uses, without lawful authority, a means of identification known to him to identify another person" would reduce the chances of ridiculous attempts to pile on bogus charges like this one.
i should note at least one nice thing about this statute: it delegates the definition of "means of identification", presumably to another part of the law. i've seen way too many pieces of legalese (particularly contracts) which would interrupt that sentence with three lines of examples there (e.g., "a social security card, state identity card, driving license, passport, military identity card, …, or any other means of identification").
Breffni said,
May 8, 2009 @ 6:18 am
Chris and Neal: just to be clear, I agree that this judgement was the right one, and I think it was absurd for the government to maintain that “knowingly” didn’t encompass “of another person” in this statute.
I agree that the second example does invite the interpretation that the doormen knew the individual was an undercover cop – and in fact, the article Neal links to leaves open the possibility (I wouldn’t put it any stronger than that) that the writer intended that interpretation. But you can find other examples where I don’t think there’s any such doubt:
Certainly you can argue that these are badly written, in that they leave open or even invite a reading contrary to the author’s intent. But the point is precisely that, as written, they can support both meanings.
Neal: “The Supreme Court's decision, and in particular the sentence that's quoted in Jason's quote, deals only with the syntactic issue.” – yes, and that’s precisely the problem, because in my view nothing follows from the syntactic issue alone. There is no doubt whatsoever that the adverb grammatically modifies the VP as a whole and not just the verb. But to get to the Court’s conclusion from there, you have to further argue that grammatical modification by knowingly (and similar mental-attitude adverbs) necessarily brings with it a de dicto reading of all the constituents within the VP. That is silently assumed in the brief, in Jason’s original post, and in the Supreme Court opinion, but it does not follow from the syntactic relation of modification. To say that a VP is modified by knowingly is not to say that all the constituents of the VP must bear a de dicto reading. The examples I’ve given show that de re readings of NPs within a VP modified by knowingly are entirely possible.
Your brief does address the de re / de dicto issue – separately, not as a development of the adverbial-attachment argument – and it correctly argues that in the case of this statute, a de re reading of the relevant NP isn’t remotely plausible. But it goes on to say “while it is common for knowingly and other mental-attitude adverbs (or negated forms of them) to be used to express a de dicto meaning… we have been unable to find any examples in which they [seem] to be used to express a de re meaning.” What I’m saying is that such examples are actually easy to come by.
One more (invented) example: you say to me, “You deliberately poisoned my pet carp.” For this to be true, minimally it has to be the case that I knew I was poisoning something. World knowledge tells us that I probably knew it was a fish. It may or may not be the case that I knew it was anyone’s pet, or that it was your pet (or even that you existed), or that it was a carp, any combination of these facts. Clearly deliberately grammatically modifies the whole VP, but that does not settle the question of which elements need to be read de dicto; at a minimum, the verb does, but only shared knowledge and context can tell us which other elements should be read de dicto
So my point is, interpreting knowingly+VP is not a question of syntax, it’s a question of pragmatics. Even within a VP modified by a mental-attitude adverb, particular descriptions may be selected for the information they provide to the hearer / reader, and not necessarily because they correspond to the mental state of the VP’s agent.
Mark Liberman said,
May 8, 2009 @ 7:41 am
Breffni: There is no doubt whatsoever that the adverb grammatically modifies the VP as a whole and not just the verb. But to get to the Court’s conclusion from there, you have to further argue that grammatical modification by knowingly (and similar mental-attitude adverbs) necessarily brings with it a de dicto reading of all the constituents within the VP. That is silently assumed in the brief, in Jason’s original post, and in the Supreme Court opinion, but it does not follow from the syntactic relation of modification.
This is both true and crucial to the issue under discussion. For readers who are not already familiar with the de dicto/de re distinction, I'll link to some older LL posts — "Pernicious ambiguity at Davos", 2/8/2005; "Rarely better than de re", 10/23/2005; "Lips that touch Lampchops, shall never touch mine", 7/31/2007 — and a couple of relevant entries in the Stanford Encyclopedia of Philosophy, notably "Propositional Attitude Reports and The De Re/De Dicto Distinction.
Joshua said,
May 8, 2009 @ 9:28 am
While I agree with the judgment here, I think the concurrences both make important linguistic arguments, most of which have already been discussed here. But Scalia makes an interesting point about previous cases in which the Court improperly (in his view, with which I concur) applied "knowingly" well beyond the phrase it modifies in normal grammar to all subsections, where the subsections were plainly a kind of legal modifier of "commits a felony" not for "knowingly transmits an image."
Another interesting case from a linguistic standpoint is "Watson v. United States", in which Watson was charged with use of a firearm in the commission of a felony when he used a gun in lieu of cash to purchase drugs from an undercover officer. Justice Souter wrote the unanimous opinion, which included this very fine defense of both language and the law:
Joshua said,
May 8, 2009 @ 9:30 am
Errr, I misstated the facts of the case. Watson was selling drugs to an undercover agent and accepted the gun in lieu of more orthodox payment, something the Souter quote probably made plain.
[(myl) See Smith v. United States for an earlier case that conforms more closely to your original description, and where the ordinary-language meaning of "use" was championed by Associate Justice Scalia, among others, as discussed in "A result that no sensible person could have intended", 12/8/2005.]
Neal Goldfarb said,
May 8, 2009 @ 10:10 am
Breffini:
I don't disagree with what you say about your examples, but I think they differ in important ways from the statute involved in Flores-Figueroa.
In all of your examples, the the verb's direct object is a description of a particular individual (or a group of particular individuals), and the possibility of alternative readings derives from the fact that there are a variety of different ways by which those individuals can be identified or described. (In the case of undercover agent this potential for supporting alternative descriptions is built into the meaning of the phrase.) And it occurs to me that a de re interpretation is by definition possible only where there is some particular individual or group of particular individuals who the expression can be interpreted as picking out. De re means "of the thing," which presupposes the existence of a thing.
In the statute, OTOH, the phrase of another person doesn't describe or identify any particular individual. Rather, it describes a property that is predicated the type means of identification. So let me suggest the hypothesis that a de re reading isn't possible in that ontological environment: an expression describing a type doesn't pick out any specific tokens of that type as its referent. A de re reading becomes possible only in the context of an individual case, when one is dealing with a particular entity in the world and trying to decide whether that entity is in fact a token of the type.
More broadly, in your examples (as well as all the examples of de re / de dicto ambiguity that I can think of at the moment), the sentence makes an assertion of fact of some sort. Again, that's not true of the statute here. The statute says, "Whoever knowingly transfers, possesses, or uses…a means of identification of another person [shall be punished]." The act of knowingly transferring, etc. is being predicated of whoever, a relative pronoun that can't be interpreted as picking out any particular individual. So the modality* of the phrase in the statute is different from the modality of your examples. (*I think that's the term I'm looking for, but if it's not, hopefully someone will correct me.)
Mark F. said,
May 8, 2009 @ 11:01 am
Jangari said that laws are deliberately obfuscated largely to preserve "the legal fraternity's exclusivity". This ignores the value of copying turns of phrase that you know have already been interpreted in court in the way that you mean.
John Baker, in response, said that "legal drafting in most cases seeks to achieve precision of meaning, to the exclusion of all other goals." I imagine they do seek that, but, considering how often they continue to use "knowingly" in the face of its amply demonstrated scope problems, they don't seem to be trying as hard as they should.
J. W. Brewer said,
May 8, 2009 @ 11:34 am
Two points:
1. There is a feedback thing going on because statutory drafting occurs with knowledge of how courts tend interpret statutory language. Since American courts adopt multiple approaches, which may in some instances lead to different results, often without predictable meta-rules as to which approach will end up being adopted in a particular case, difficulties arise, especially given the sausage-making & horse-trading dynamics of the legislative process. In some instances, statutes may be deliberately ambiguous because no clear consensus on a particular issue could be obtained in Congress so the issue is being punted to the courts (perhaps with Congressional partisans for rival views each trying to get some evidence that can later be argued to support their interpretation). My own guess is that, perhaps ironically, the only way to fix this problem (if it's considered a problem, and one worth fixing) and create effective incentives for Congressional clarity would be for courts to turn away from what sensible linguistics scholars would say about how to interpret English as it occurs in ordinary contexts and instead rigidly follow various arbitrary "prescriptivist" type rules of the form "we will never interpret the statute to mean X unless you use specific lexical item or syntactic construction Y, regardless of the fact that there are lots of other ways to communicate X in ordinary English usage. Oh, and don't tell us that sometimes in ordinary English lexical item or syntactic construction Y means Z instead of X, because we won't care about that either."
2. This is a funny sort of statute, in that it only applies to defendants who are already committing another "predicate" crime which presumably has its own mens rea requirement. A number of people, including some judges and legislators, turn out to have different intuitions (as a matter of moral philosophy or public policy or whatever) about how state-of-mind requirements should work in that context. I.e. they may think the line between non-criminal and criminal activity should be a lot harder to blunder over by accident than the line between more minor crime A and more major crime B, with those in the latter category perhaps proceeding at their own risk ("In for a penny, in for pound.") For example, drug laws that impose punitive mandatory minimum sentences when more than a specified quantity of the drug is involved often turn not to be construed to require actual knowledge of the weight, because we turn out as a society not to be particularly sympathetic to the plight of the fellow who conscientiously tried to keep the amount of cocaine in his possession at 499 grams or less at all times but "unknowingly" ended up in possession of 501 grams thereby exposing himself to a much longer sentence. This doesn't mean the Court wasn't totally right in how it construed the language of the statute before it, only that the argument that the contrary reading would be absurd because it would lead to a comparatively trivial mens rea requirement isn't necessarily as compelling as some seem to think.
John Baker said,
May 8, 2009 @ 12:12 pm
Mark F. – Achieving precision is really hard. Using legal language makes it somewhat easier, but only somewhat. And, yes, it's often the case that we can look back and see that the draftsman should have done a better job.
We do sometimes see cases where new and simpler language turns out to be more precise than the old legalese. That is not language that is just pulled out of the air, though; it's the result of craft and effort.
Breffni said,
May 9, 2009 @ 8:09 am
Neal:
I don't think this analysis works. Statutes by their nature tend not to pick out particular individuals, only types. But such generic provisions then have to be applied to individuals, either hypothetical or real-world ones. When you have knowingly etc., the question then resolves to one of whether a specific individual, the accused, knew that a specific thing was a token of the type named in the statute. Imagine a statute that says
Anyone who knowingly causes damage to a national monument shall be guilty of…(etc.)
It's possible to cause damage unwittingly, so knowingly minimally applies to cause damage – i.e., the verb has to be read de dicto (and as far as I can see this is always the case). But does "national monument" have to be read de dicto? It certainly picks out a type rather than a specific individual, so your argument would suggest that it must be interpreted de dicto. But it can be read de re: it would be possible to argue that if you vandalise a particular building that, to your dismay, turns out to be a national monument, then you have infringed the statute. (Is this called "absolute liability"?) It would also be possible to argue for a de dicto reading, of course, so the statute would indeed have the de re / de dicto ambiguity, even though, as in most legislation, it's framed in generic rather than specific terms.
Neal Goldfarb said,
May 9, 2009 @ 12:29 pm
While I’m not certain that this does display the ambiguity that Breffini sees, let’s assume that it’s at least potentially ambiguous. I’d suggest that the source of the ambiguity isn’t necessarily the distinction between de re and de dicto. Rather, I think there may be a syntactic explanation. Here are two possible syntactic structures for the phrase knowingly causes damage to a national monument
(To convert these examples into tree diagrams, copy them (all three at once), follow this link, and paste them into the white box in place of what’s already there. Then click on Draw and you'll see the three trees.)
The structure in (1) would correspond to what Breffini describes as the de re reading, while the structures in (2) would correspond to a de dicto reading. Of course, this assumes that the structures (1) and (2a)—in which to a national monument is treated as an adjunct rather than a complement of damage—is valid, because otherwise there’s no structural ambiguity. I realize that such a structure might be problematic, since national monument is a semantic argument of damage and therefore should arguably be shown as part of the NP. But the latter structure, too, is problematic: it doesn’t fully capture the semantic relation between causes and damage.
Whereas the type of event described by a verb phrase is ordinarily encoded in the verb itself (e.g., throw a baseball describes an event of throwing), in the example here the description of the specific event type is spread across the verb and its immediate direct object. The expression cause damage describes an event of damaging. This is shown by the fact that to cause damage to a national monument means essentially the same thing as to damage a national monument So as used in Breffini’s example, damage simultaneously has the properties of a verb (in its semantic relations) and of a noun (in its syntactic relations). And that, I think, explains why the structures in (1) and (2a) are simultaneously defensible and problematic.
If I’m right about this, the (potential) ambiguity in Breffini’s example can be explained in terms that don’t rely on the concept of de dicto versus de re. And if that’s the case, the example isn’t necessarily inconsistent with my hypothesis.
Breffni said,
May 9, 2009 @ 5:00 pm
Neal, I don't think there's any structural ambiguity in that example. I think the correct analysis is
(3) [VP [Adv knowingly] [VP [V causes] [NP damage] [PP to a national monument]]]
…on the grounds that the to-PP is licensed by cause. But whatever the facts about that, it seems to me that – as you said yourself – it's semantically equivalent to
Anyone who knowingly damages a national monument shall be guilty…
…which would definitely be
[VP [Adv knowingly] [VP [V damages] [NP a national monument]]
And this – I believe – has the same de re / de dicto ambiguity as "You knowingly misled my mother".
To come at it from another direction, if your hypothesis (regarding generic versus specific reference) were correct, then you should feel that (4) below is ambiguous while (5) is not:
(4) You knowingly damaged a national monument.
(5) It is an offence to knowingly damage a national monument.
I just can't see any difference. They both seem ambiguous, and for the same reason: de re versus de dicto readings of the direct object NP.
To be clear, I do think the de dicto reading is more salient – but not because of syntax.
Neal Goldfarb said,
May 9, 2009 @ 6:51 pm
I don't think that generic-versus-specific is really the distinction I was getting at when I talked about types and tokens (though that may not have been clear from what I said). I think the more important question is whether the NP that acts as the direct object has a referent. (Or maybe it's a question of modality. I don't know; I'm in way over my head here.)
In any case, I do see a difference between (4) and (5), and my best explanation for that is that it's because in (4), a national monument refers to a specific thing, while in (5) it does not.
Another point, and this gets to the issue of pragmatics, is that (4) can (and typically would) be uttered in a particular discourse and real-world context, and with respect to a particular situation. What the speaker means in uttering (4), and what the hearer understands him to mean will be sensitive to those details, which can vary each time the sentence is uttered. That makes it possible to conceive of a context in which a de re interpretation might be intended.
Statutes, OTOH, are very different. They are uttered only once, when they are enacted. The occasion for a statute to be interpreted comes later, when it has to be applied to a specific case, and at that point it is as if a recording of the original utterance is played back. But this playback doesn't bear the same kind of relationship to the situation at issue as an utterance of a sentence such as (4). The sentences in the statute aren't being uttered about the situation. Therefore, the sort of pragmatic processes that can go on with respect to interpreting an utterance of (4) can't take place with respect to interpreting a statute.
Does any of this make sense?
David said,
May 10, 2009 @ 1:56 am
"The statute provides: "Whoever, during and in relation to any felony violation enumerated in subsection c, knowingly transfers, possesses, or uses, without lawful authority, a means of identification of another person shall, in addition to the punishment provided for such felony, be sentenced to a term of imprisonment of 2 years." How would you rewrite it to make interpretation unnecessary?"
I'll give Baker's challenge a try. While I think my attempt is poor, I do think it would be easier to understand laws if they wrote them in multiple parts instead of squeezing everything into one sentence. So I would rewrite as (in this instance meaning the accused does not have to know that the identity belonged to another person):
Required Cicrumstances: Punishable action occurs while committing a felony as found in subsection C, and punishable action must happen in relation to said felony.
Required Object: Another person's means of identification.
Punishable Action: Knowingly transfers, possesses, or uses Required Object without lawful authority.
Punishment: 2 years imprisonment
Aaron Davies said,
May 10, 2009 @ 9:17 am
perhaps we should just give up and write all our laws in lojban…
Bob Moore said,
May 11, 2009 @ 3:52 pm
Wow, this takes me back to many a discussion I had in the 1970s and 1980s, and also a good bit of my dissertation "Reasoning about Knowledge and Action". The Supreme Court reached the right decision, I think, but gave an incorrect linguistic analysis. This is a simple case of scope ambiguity of the noun phrase "a means of identification of another person." De dicto / de re is an old fashioned, and also misleading description of this ambiguity. The misleading bit is that it implies there is a binary distinction here, whereas in principle you can have an n-way ambiguity by stacking up n-1 propositional operators. In a binary case like this, "wide scope" and "narrow scope" are a more perspicuous way of describing it, since if there were another propositional operator, we could add "middle scope", etc. Both indefinite and definite determiners have scope in this sense, so the issue of how an indefinite could be "de re" need not arise. See Bertrand Russell, "On Denoting" (1905) for the origins of this way of looking at things.
The narrow-scope reading of the noun phrase is the one that puts it within the scope of "knowingly". This is the one that makes the most sense here. As someone pointed out in reply to the earlier post, the wide scope reading doesn't really make sense, because under that interpretation it is hard to see what "knowingly" contributes, since one can hardly use a piece of identification without knowing that you are doing it. The only element that there could reasonably be a question about is whether you knew it belonged to another person.
Breffni said,
May 12, 2009 @ 6:49 am
Neal [sticking with the terms of the discussion so far; I’ll get back to Bob Moore’s point below] – I think I may have some insight into your intuition about a difference between (4) and (5) with regard to ambiguity. Tell me if I have this right.
What lies behind most of my argument has been the idea that NPs can be used by a speaker (i) to identify a referent using knowledge shared between the speaker and the hearer – I mention “my pet carp” in a situation where you’re already aware of my having a pet carp – or (ii) to identify a referent while also introducing information new to the hearer. In all of the real-life examples I’ve used so far, the de re reading is underwritten by this second use. The journalist refers to “undercover agents” to tell the reader something about how the offence was uncovered, the angry son uses “my mother” to assert the relevance to him of the salesman’s cheating. And it’s true that it doesn’t seem possible for an NP to function in such a way in the context of a statute, where there is no shared local context between the framer and the interpreter (lawyer, judge…) and no particular referent is yet in question.
But actually, I don’t think an NP has to be used in this information-introducing fashion for it to be readable de re. It can be used simply for its denotation, i.e., to delimit the class of things to which it may refer. This is what I called function (i). But function (i) is just as audience-oriented as function (ii); the NP “a national monument” is selected for the benefit of readers, lawyers, judges and so on, and therefore need not be taken as a characterisation of the content of the (hypothetical) offender’s state of knowledge. It can be read as doing no more than identifying for the audience the class of objects to which the knowingly-damage provision applies. And if you accept that, then, since it’s possible to unwittingly damage [something], “It is an offence to knowingly damage a national monument” can – it still seems to me – mean “knowingly (de-dicto)-damage (de-re)-a national monument”.
To look at it from another angle, if you did want to frame a single-sentence provision that made it an offence to knowingly damage a national monument regardless of whether the accused knew the structure was a national monument – how would you do it? Wouldn't it be natural to frame it along the lines of "A person knowingly damaging a national monument shall be guilty of an offence, regardless of whether that person was aware, at the time of the offence, of the structure's status as a national monument"?
I think a lot of the difficulty is the specialness of mental-attitude adverbs (and verbs), which is easy to lose sight of. Usually an NP is just a device for connecting the knowledge of the speaker and that of the hearer, taking into account the state of shared knowledge between them. But when an NP occurs within the scope of knowingly (etc.), you’ve now got the additional question of the match (de dicto), or lack of it (de re), between the precise linguistic expression (the description) and the state of knowledge of the semantic agent of the clause, who need not be a party to the discourse at all.
And – Bob – that’s why I don’t understand how the de re / de dicto distinction can be ignored. If intensional statements necessarily bring in questions of the match between a particular linguistic expression and the semantic agent’s state of knowledge, then how can that be reduced to NP scope? I’m prepared to be persuaded, because I'm not sure what you mean by the wide- vs narrow-scope readings of the NP “a means of identification of another person”. Does this hinge on the presence of the PP? Because the ambiguity that I perceive is there even in simpler noun phrases, like in the “knowingly damage a national monument” examples. It's good to have a semanticist's view, incidentally.
John Baker said,
May 12, 2009 @ 10:49 am
David – While your rewriting does not seem to me to be any more ambiguous than the original statutory language, I don't think it's any less ambiguous either. In that sense, it's actually a pretty good job of rewriting, from a lawyer's perspective, in that it preserves the original meaning – ambiguities and all. But, more to the point, it doesn't address Jangari's concern, since it is not a rewriting in everyday language. In fact, I would say your wording is even further removed from everyday language than the original is.
Bob Moore said,
May 12, 2009 @ 12:28 pm
In response to Breffni:
In logic, statements involving determiners are normally analysed in terms of quantifiers. The indefinite determiners, "a", "an", and "some" normally correspond to existential quantifiers. For example,
John kissed a girl
might be analyzed as
Exists(X, (girl(X) and kissed(John,X)))
There are cases where this kind of analyis is problematic, but the statue in question is not one of them.
If we add an intensional adverb like "knowingly", there is a semantic ambiguity in how far out to pull the quantified expression representing the noun phrase. Hence,
John knowingly kissed a girl
has two readings that might be represented as
Exists(X, (girl(X) and knowingly(John,kissed(John,X))))
and
knowingly(John,Exists(X, (girl(X) and kissed(John,X)))
[I have included "John" in these formulas twice, to make it clear that John is doing both the knowing and the kissing.] The first of these formulas would be the wide-scope reading of "a girl", and the second would be the narrow-scope reading of "a girl".
Semantically, an ambiguity like this exists whenever a noun phrase with a determiner is in a position to interact with an intensional adverb or verb. No syntactic ambiguity needs to exist for the semantic ambiguity to exist, so it is not a question of what modifies what syntactically. There is interaction with syntax, however, in that in different syntactic contexts the preference to give a wide-scope or narrow-scope reading can vary, but pragmatic considerations can almost always over-ride the syntactic preferences.
I consider the de dicto / de re distinction old-fashioned and somewhat confusing for a couple of reasons. First, as I hinted at in my previous post, it suggests a fundamentally binary distinction, while in fact we can produce more than two possibilities by embedding the noun phrase in question in additional intensional operators. For example,
Bill said that John knowingly kissed a girl.
It could be that Bill said "John knowingly kissed a girl," with either interpretation of the quoted material that we discussed above, or it could be that Bill said something like "John knowingly kissed Mary," and the speaker is using "a girl" instead of identifying Mary by name. This reading may be a bit hard to see in this example, but it is easy to make it more plausible by further elaborating the ambiguous noun phrase:
Bill said that John knowingly kissed a girl whose name I won't disclose.
The second reason that I do not like to talk in terms of de dicto / de re is that it is sometimes conflated with the pragmatic "attributive / referential" distinction, which I believe is what Breffni is aluding to in his latest post. The attributive / referential distinction arises only for noun phrases that semantically are interpreted as having the widest possible scope with respect to any intensional operators the sentence may contain. It is a purely pragmatic distinction rather than a semantic one, and it can arise whether or not there are any intensional operators to create a semantic scope ambiguity. Saul Kripke's paper "Speakers Reference and Semantic Reference" (anthologized in a number of places) clearly explains why this is a pragmatic and not semantic distinction.
If I read "de re" in Breffni's latest post as "having widest possible scope" then I think he and I may fundamentally be in complete agreement. In the case of the statute under discussion, I would assert that placing the adverb "knowingly" between the subject "whoever" and the verbs "transfers, possess, or uses" creates a context where either scope for the noun phrase "a means of identification of another person" is equally plausible syntactically and semantically; so only pragmatic considerations can resolve the intended interpretation, and pragmatically, the wide-scope interpretation is not plausible for reasons already discussed.
Neal Goldfarb said,
May 12, 2009 @ 1:53 pm
Let me respond to Bob Moore’s comment; I’ll get to Breffni’s latest in another post.
First of all, am I correct inunderstanding that when you talk about NPs having wide or narrow scope, you’re talking about scope in logical form rather than in the morphosyntactic form? And if so, am I correct in understanding that a wide-scope NP is (in this case) an NP that has scope over the functions corresponding to transfers, possesses, or uses, but has narrow scope if it is within the scope of those functions? For example, for the sentence Sam wants to marry a lawyer, the two possible scopes for lawyer are shown here (please forgive the bastardized and inconsistent notation):
Assuming I’m right about what you mean, the question that comes up is, how does the logical form you are positing adverb knowingly? More specifically, are you following a (neo-)Davidsonian approach to event semantics? Because if you are, why wouldn’t the appropriate logical form follow this template
where p stands for the logical form of the sentence without the adverb? There’s support for this approach in Davidson’s The logical form of action sentences (pp. 121-22), Parson’s Events in the Semantics of English (ch. 4, p. 64), and Ernst’s The Syntax of Adjuncts (p. 55).
In the case of the statute we’re discussing, this would translate into something like this (again, excuse the lousy notation):
In this structure, the sort of scope ambiguity you’re referring to can’t arise, because knows that (corresponding to knowingly in the actual sentence) has scope over the logical form of the entire underlying event.
Neal Goldfarb said,
May 12, 2009 @ 1:59 pm
I see that while I was composing my last comment, Bob Moore posted a comment that answered most of my questions. Based on what he says, it looks like (1) I did understand him correctly and (2) the logical form he is positing does not take explicit account of events.
Bob Moore said,
May 12, 2009 @ 3:37 pm
In response to Neal Goldfarb:
As a matter of fact, I do like the neo-Davidsonian approach to certain types of adverbials, including manner adverbials. I didn't include the event analysis here, because I didn't want to make things more complicated for readers who might not have encountered the notion of quantifier scope before.
In any case, it doesn't matter, because the ambiguity of the scope of noun phrase quantifiers arises whether or not a Davidsonian account of events is adopted. You seem to be assuming that a noun phrase quantifier cannot out-scope the event quantifier of the main verb of the clause, but why not? If that were the case, there would be no accounting for the ambiguity of "Ralph believes that someone is spying," to paraphrase Quine's famous example, using an action verb to make sure there is a Davidsonian event.
On a related matter, for a discussion of when adverbs should be treated as predicates of events, and when they should be treated as propositional operators, please take a look at "Events, Situations, and Adverbs," Chapter 9 of my book "Logic and Representation," which is now online at
http://sul-derivatives.stanford.edu/derivative?CSNID=00003780&mediaType=application/pdf
and don't believe the line break after the ? in the middle of that URL!
Neal Goldfarb said,
May 12, 2009 @ 10:28 pm
@Bob Moore:
I think I'd need to see the sort of logical-form structure you've got in mind (specifically, a (neo-)Davidsonian structure for knowingly transfer, possess, or use…a means of identification of another person in which knowingly (as represented by knows that) is within the scope of of another person or a means of identification of another person.
The Quine example doesn't strike me as particularly relevant, because it seems to me that its ambiguity can be explained by the polysemy (or underspecification?) of someone. Also, the Quine example involves a significantly different sentence structure than is involved in the statute. In the former the relevant NP (someone) is the subject of the proposition that Ralph believes, while in a sentence such as Ralph knows [or believes] that Ralph used a means of identification of another person (corresponding to the second major part of the logical form I proposed for the statute), the relevant NP is embedded in the predicate. Isn't that going to result in there being a big difference between the logical forms?
Bob Moore said,
May 13, 2009 @ 2:25 am
In reply to Neal:
To simplify some of the irrelevant complexity, let's consider "whoever knowingly uses a means of identification of another person shall be sentenced." I hope you agree this is enough to get to the heart of the matter. The logical analysis for the wide-scope reading, which you doubt the possibility of, would be something like the following (in notation modeled after your last):
For all X, if (there exists Y, such that (Y is a means of identification of another person) and (there is an event Z such that (Z is a using event) and (X is the agent of Z) and (Y is the instrument of Z)) and (X knows that (there is an event Z such that (Z is a using event) and (X is the agent of Z) and (Y is the instrument of Z)))), then (X shall be sentenced).
So, I hope you can see that introducing a Davidsonian treatment of "uses" in no way blocks the quantifier scope ambiguity of "a means of identification of another person". Leaving aside the Davisonian treatment of action verbs, this has been the standard treatment of this sort of ambiguity since Quine's "Quantifiers and Propositional Attitudes," Journal of Philosophy 53. Reprinted in his 1976 Ways of Paradox. Harvard Univ. Press: 185-96. You are bucking 50 years of tradition here, so I think the burden is on you to show that there is something wrong with this analysis, if you still don't like it.
Breffni said,
May 13, 2009 @ 3:31 am
Bob, thanks for the explanation. As you say, we are in agreement: the argument from syntactic scope (alone) doesn't work, there is a potential de re / de dicto (wide / narrow scope) ambiguity, it has to be resolved by pragmatic inference, and in the case of this statute that's easily done.
The brief does in fact make the argument from de re / de dicto, but it overgeneralises by implying that examples of de re readings of ordinary sentences modified by knowingly are vanishingly rare. That, and the focus on syntactic scope, are my main criticisms of it.
I'd like to defend de re / de dicto, though. Bob has two reservations. One is that it can be mistakenly conflated with attributive / referential. (I was going to stick my neck out and use those terms for what I called functions (i) and (ii), but I wasn't sure – I'm still not – that they're exactly the same thing.) Fair enough, but that's an argument for greater conceptual clarity and care in using terminology. The second reservation was that it gives you a binary distinction where in fact relative scope ambiguity can be n-way. But the claim that de re/ de dicto is binary rests on the assumption that it's all or nothing – that a given intensional sentence is entirely de re or entirely de dicto. I've been assuming all along that the terms can be applied to individual constituents, so that you can say this NP must be read de dicto, this PP (e.g., "of another person") de re, and so on. And if that is the case, then your n-way ambiguity is captured in different constellations of de re and de dicto readings (like in my carp examples above). (The brief correctly dismisses what it calls a "half-and-half" reading of the statute, but that doesn't generalise.) So if that's the case, what's wrong with using de re / de dicto as a label for certain scope ambiguities in intensional contexts?
Bob Moore said,
May 13, 2009 @ 6:22 am
If you want to use de re systematically to mean "widest possible scope with respect to intensional operators" and de dicto to mean "any scope with respect to intensional operators besides the widest possible" then I have no objection. My objection is based only on my perception that many people use the term under the misimpression that it gets at some fundamentally binary distinction that is not fully captured by scope ambiguities.
Neal Goldfarb said,
May 14, 2009 @ 10:11 am
Bob: OK, I see your point, but let me ask a different but related question.
You've given a logical form for a wide-scope reading of the entire NP a means of identification of another person, under which the statute in Flores-Figueroa could apply even if the defendant didn't know that what he was transferring, possessing, or using was a means of identification. E.g., if he gave someone a sealed envelope containing a social security card, but he didn't know what was in it, the wide-scope reading would allow you to say that he knowingly transferred a social security card. I don't think there was ever any chance that the Supreme Court would adopt that interpretation.
But what about an interpretation that would require the government to prove that the defendant knew that what he transferred, possessed, or used was a means of identification, but not that he knew that the means of ID belonged to someone else? This would correspond to giving wide scope to of another person but not to the matrix NP a means of identification. The government argued for that interpretation as a fallback position, and while we argued in the brief that such an interpretation would be unnatural, we didn't try to give a theoretical basis for that intuition.
After having thought about what you've said here, it occurs to me that your analysis would suggest this interpretation isn't possible. First, if the de re / de dicto ambiguity is described as an ambiguity relating to quantifier scope, the ambiguity is possible only with respect to quantified phrases. But unless I'm mistaken, of another person, like any other PP, is not quantified. It therefore can't have scope over knowingly.
Does this sound right?
(BTW, Bob, what's your email address? If you don't want to post it here, you can email it to me. You can find my address by Googling my name.)
Bob Moore said,
May 14, 2009 @ 3:44 pm
Neal,
According to my theory, the interpretation that he knew it was a means of identification, but didn't know that it belonged to another person, is not linguistically available. On my theory, "belonging to another person" cannot float outside the scope of "knowingly" and leave "a means of identification" inside the scope of "knowingly". Interestingly, "another person" is also a quantified noun phrase meaning "a person, not the same as (some contextually specified person)", so it could take wide scope, leaving "a means of identification belonging to _" with narrow scope. But this would have made the government's case even more difficult, because the standard interpretation of *that* would be that he would not just have to know that it belonged to someone else, he would have to know who it belonged to.
I found your contact info on the web, and I will send you my email address privately.
Breffni said,
May 16, 2009 @ 4:31 am
Bob:
As I understand it, the same argument applies to other modifiers in the NP – they'll all fall within the scope of the determiner, so that the entire NP must be read either de re (widest possible scope) or de dicto (narrower scope). But what about examples like this? –
"You deliberately ruined my new hearthrug."
In the context of the play, this is addressed to someone unknown to the speaker, and it seems clear that she means "knowingly" to apply to "ruin my hearthrug", but not to "new". Similarly:
"He was ordered to pay £7,000 in fines and costs for knowingly breaching a prohibition notice that had been issued by the Fire Authority under the new Regulatory Reform (Fire Safety) Order which was introduced in October last year." (From here)
I don't imagine "knowingly" was meant to apply to the relative clause, certainly not in every detail. It wouldn't be a plausible defence to argue that you didn't know the prohibition notice was issued under that particular Order, or that it was introduced in October.
Breffni said,
May 16, 2009 @ 4:33 am
Correction – in my comment on the first example, for "knowingly" read "deliberately".
Bob Moore said,
May 16, 2009 @ 9:10 am
In those examples, I would say that, in theory, the ambiguity is between a semantic narrow scope reading, knowing everthing within the syntactic scope of "deliberately" and "knowingly", and a semantic wide scope reading
"deliberately ruined X"
"knowingly breaching Y"
where X and Y are described by the speaker (not the knower) as "my new hearthrug" and "a prohibition notice … last year."
In the wide scope reading, it's completely undetermined how much of the content of the descriptions the speaker gave of X and Y might be known to the knower.
My guess about the second example is that the journalist has made a complete hash of the court's decision by inserting "under … last year" where it doesn't belong. I would bet that the person in question was charged with "knowingly breaching a prohibition notice issued by the Fire Authority," where "a prohibition notice issued by the Fire Authority" is intended (in the statute) to be read with narrow scope. The journalist then comes along, trying to add more information without adding an extra sentence, and inserts "under … last year" without even considering how that interacts with "knowlingly".
Breffni said,
May 17, 2009 @ 4:16 am
Bob,
This means that under a wide-scope analysis, ambiguity remains. Either none of the NP's description is known to the agent ("pure" de re), or some of it (some elements de re, some de dicto), or all of it – but this last possibility is indistinguishable from the narrow-scope reading. This means that there are two different analyses corresponding a "pure" de dicto reading: either the NP has narrow(er) scope, or the NP has widest possible scope but on pragmatic grounds all elements are understood to be "known".
So wouldn't it be more parsimonious to construe the de dicto / de re phenomenon as pragmatic from start to finish? For any given case, simply presuppose the wide-scope analysis and leave everything to contextual inference (which may or may not lead to a reading which happens to correspond to the narrow-scope analysis). That way you don't have to assume two layers of ambiguity, and two successive resolution processes, and you have a unified account of all possible combinations of known and unknown elements in the VP.
On a related point, if wide-scope readings are radically underdetermined, is there really any non-pragmatic way of excluding a reading of the statute in which the PP "of another person" sits outside the scope of "knowingly"?
Bob Moore said,
May 17, 2009 @ 12:51 pm
Breffni,
I wouldn't say that ambiguity remains under the wide-scope reading. Rather, the wide-scope reading just doesn't have anything to say about whether the knower knows anything of the descriptive content of the NP, just as it doesn't have anything to say about the proverbial price of tea in China. I made a comment about that because of your wanting "hearthrug" to be within the scope of knowingly and "new" to be outside the scope. I would say instead that if "new" is outside the scope, "hearthrug" also has to be outside the scope. because scope ambiguities generally apply to whole NPs, not pieces of NPs, unless they also happen to be whole NPs. I think your intuition about "new" and "hearthrug" is driven by imagining a scenario in which it would be obvious that the thing ruined is a hearthrug, but not obvious that it is new. Make sense, but that is what the sentence actually entails on either reading.
On the (standard) theory I am espousing, on the narrow-scope reading the knower would have to know that he ruined some new hearthrug, but might have no idea what particular individual object he ruined; on the wide-scope reading, he would have to know what particular object he ruined, but not that it was a new hearthrug. To see that particular reading, imagine substituting some esoteric object for the commonplace hearthrug, e.g.
You deliberately ruined my new 1-gigabit 8-port 802.11N wireless router.
That reading would be satisfied if the person saw the object and deliberately smashed it. In that case, he would know what particular object he had ruined, but whether he knew any part of the fact that it was a new 1-gigabit 8-port 802.11N wireless router is not merely underdetermined, it is irrelevant.
So, the wide-scope reading is not just an underdetermined version of the narrow-scope reading. It substitutes a different positive requirement, that one know something about a particular object. Of course, what it actually means to know something about a particular object is itself not always clear, but that is fundamentally a question for epistemology rather than linguistic semantics.
The broader point here is that this is an ambiguity in which the conditions that are required to be satisfied by each of two readings are not incompatible with each other. Hence, on either reading, it is possible that some or all of the conditions of the other reading also are satisfied, but that is not required by either reading.
Getting back to the statute that started all this discussion. I would say that there is no available reading that, in order for the statute to be violated, requires the violator to know that what he is using is a means of identification, but does not require him to know that it belongs to another person. Instead, the wide-scope reading would require him to know that he was using some particular object, which (possibly unknown to the violator) happens to be a means of identification belonging to another person. As we have both observed, however, it is pragmatically implausible that this requirement would be met without him also realizing that it as a means of identification, but that is not what the statue would actually require on that reading.
Bob Moore said,
May 17, 2009 @ 10:52 pm
The last sentence of the first paragraph of my last comment should read "Makes sense, but that is *not* what the sentence actually entails on either reading."
Breffni said,
May 18, 2009 @ 4:46 am
Bob, thanks for this very helpful elaboration. I think what has clicked with me (with regard to the wide-scope reading) is the difference between silence and ambiguity. That, and the point that "on either reading, it is possible that some or all of the conditions of the other reading also are satisfied, but that is not required by either reading". It's a fascinating topic; I'll have to chase up some of the references you gave.