Adverbial modification at the Supreme Court today

« previous post | next post »

The following is a guest post by Jason Merchant.

The Supreme Court is scheduled today (25 Feb 2009) to hear arguments (Flores-Figueroa v. U.S., No. 08-108) to decide whether Ignacio Flores-Figueroa should have his conviction for aggravated identity theft reversed. The debate centers on the interpretation of a statute, 18 U.S.C. sec. 1028A(a)(1), which states that:

"Whoever … knowingly transfers, possesses, or uses, without lawful authority, a means of identification of another person shall … be sentenced to a term of imprisonment of 2 years."

The facts of the case are undisputed: in 2006, Mr. Flores-Figueroa purchased a forged Social Security card and resident alien card in his own name but with numbers that were not assigned to him by the relevant authorities. His employer, a steel plant in East Moline, Illinois, submitted these numbers to the federal authorities, who determined that the numbers had been issued to other people. Mr. Flores-Figueroa pleaded not guilty to aggravated identity theft, saying that "he didn't know that the ID numbers belonged to anyone" (And Unequal Justice for Some, editorial, New York Times, 25 Feb 2009) but was convicted and sentenced to the mandatory two years.

The question before the Court boils down to the scope of the adverb knowingly: as a law professor friend of mine put it, the question is "whether the term 'knowingly' modifies just 'uses' or also 'of another person'".

As a point of grammar, adverbs modify verb phrases, not just verbs. This can be seen pretty clearly by considering the havoc that would arise if "just verbs" could the target of modification. Consider the sentence in (1):

(1) Oedipus knowingly married his mother.

If we could ignore the object of the verb, and take knowingly to modify just marry, (1) should be ambiguous; on one reading, (1) should be equivalent to (2):

(2) Oedipus knowingly married.

And of course while Oedipus did knowingly marry, he didn't know Jocasta was his mother.

Likewise, we can't just ignore some part of the object either (as is implied by wondering whether knowingly applies also to "of another person"); consider:

(3) Oedipus knowingly married a woman who was his mother.

Again, if we could just ignore the relative clause who was his mother, the sentence in (3) could mean the same thing as (4):

(4) Oedipus knowingly married a woman.

And that would be silly. (3) does not equal (4). The same applies to the verb know of course, to which the adverb knowingly is related:

(5) Oedipus knows he married a woman who is his mother.

(6) Oedipus knows he married a woman.

Again, (5) ≠ (6).

It seems clear that the confusion arose because of a sloppy way of speaking that exists, which is to say things like "an adverb modifies a verb". This is wrong. And because adverbs modify verb phrases, not just verbs, the object of the verb can't be ignored, nor can some arbitrary subpart of the object.

One might also be led into this confusion by an earlier decision: "In a 1985 case, quoting a treatise on criminal law, the Supreme Court summed up the problem this way: 'It is not at all clear how far down the sentence the word knowingly is intended to travel.'" (Justices Take Case on Illegal Workers and Penalties for Identity Theft, Adam Liptak, New York Times, 20 Oct 2008.) Sounds like the Supreme Court is endorsing syntactic movement! And not just movement, but covert movement at that ("Supreme Court Decides in Favor of QR"). But that 1985 decision really was about a structural ambiguity between two adjuncts (one before and one after the VP), and doesn't bear on the identity theft statute here at all. It was parallel to another case whose outcome rode on a structural ambiguity in the language of the statute: "this crime covers anyone who intentionally accesses a federal computer without authorization, and by means of one or more instances of such conduct alters, damages, or destroys information" (18 U.S.C. 1030(a)(5)(A) debated in United States v. Morrison; brought to my attention by a linguistics BA who went on to law school and put his training to good use). In this latter case, the question is whether the adverb intentionally is attached to only the first VP or whether it is attached to the conjunction of the two VPs. The defendant argued in favor of the wide scope reading, while the government argued for the local attachment (the judge agreed with the government).

A nice argument preview exists, and briefs and documents are available at Flores-Figueroa v. US, including the excellent amici curiae brief by Tom Ernst, Georgia Green, Jeffrey Kaplan, and Sally McConnell-Ginet.

(Thanks to Prof. Todd Henderson of the University of Chicago Law School who originally brought the case to my attention and asked me about this point of grammar.)

[The decision? Here.]


  1. jfruh said,

    February 25, 2009 @ 11:52 pm

    Er, the problem with your examples is that in the Oedipus myth, Oedipus doesn't know that Jocasta is his mother (he was abandonned as an infant and raised elsewhere). It's fairly clear from context what you mean but the fact is that Oedipus did not knowingly marry his mother.

    [(myl) This is exactly the point of Jason's example: apparently you have accepted his conclusion in just the same sense that Oedipus married his mother. ]

  2. Russell said,

    February 25, 2009 @ 11:58 pm

    Best sentence:

    The problem is that in these structures, of another person is forced against its will to function adverbially, since it is combined with a verb phrase. That is not a function it can perform, even under duress.

    Also: One unspoken part of the argumentation (especially but not only when we get to the trees) is that semantic composition and scope may be read off of syntactic structure (see above: since it is combined with a verb phrase). This is often the case (for reasonable meanings of "often"), but not always. What immediately comes to mind are focus-sensitive adverbs like only; quantifier raising (mentioned by Jason); and various sorts of "misplaced" modification (smoke a languid cigarette, and a few other choice examples in one of the final chapters of James McCawley's "Syntactic Phenomena" text).

    What's interesting is not the fact that these went unmentioned (that would be supremely out of place), but rather that it is so intuitive and unproblematic that there should be such a degree of isomorphism, that it need not be mentioned.

  3. dgraham said,

    February 26, 2009 @ 12:40 am

    jfruh: I believe that that was the point. If the narrow-scope reading were an option then that sentence would (on one reading) make sense within the context of the myth. BUT since the narrow-scope reading is not an option, the sentence can never make sense in the context of the myth.

  4. Graham said,

    February 26, 2009 @ 12:51 am

    Also at the Supremes this week: CARCIERI v. SALAZAR, in which the term "now" in section 479 of the Indian Reorganization Act of 1934 (which defines "Indian" to "include all persons of Indian descent who are members of any recognized tribe now under Federal jurisdiction") is ruled to unambiguously denote the time of speech (e.g. 1934) and not some other time (say, 1998), thus depriving the federal government of the authority to take 31 acres in trust for Rhode Island's Narragansett Tribe (who – as it happens – weren't under Federal jurisdiction until 1983). The court notes that this accords both with case law and "with the word’s natural reading in the context of the IRA."
    (Also interesting: In his dissenting opinion, Justice Stevens questions the compositionality of the phrase "Indian tribe", suggesting that the "Indian" here may not be that defined in section 479 of the IRA.)

  5. Neal Goldfarb said,

    February 26, 2009 @ 1:26 am

    As the author of the amicus brief in Flores-Figueroa, I'd like to thank Jason and Russell for their comments.
    For those who are interested in reading the transcript of the argument, it's available here.

    I also filed an amicus brief in United States v. Hayes, a case that the Supreme Court decided on Tuesday. The amici in that case were Ted Gibson, Georgia Green, Ray Jackendoff, Jeff Kaplan, and Language Log's own Roger Shuy. That brief is available here and the court's opinion is here.

    Hayes involved the interpretation of a statute that prohibited ownership of firearms by anyone who had been convicted of a "misdemeanor crime of domestic violence"; it's gotten a fair amount of media attention because it's the first gun case the court has decided since Heller.

    From the point of view of trying to promote greater linguistic sophistication on the part of the courts, the decision in Hayes is disappointing. (Translation: The brief didn't make much headway with the majority.) From the point of view of the people who don't much care about how courts deal with interpretive issues, the decision is either good (for those who want to keep guns away from people who have been convicted for committing acts of domestic violence) or bad (for those who have been or may be convicted of such acts, and for those who are proponents of Second Amendment rights). But this being Language Log, that's not the viewpoint that matters here.

  6. Jon said,

    February 26, 2009 @ 5:16 am

    As a criminal lawyer myself, I'd be interested in hearing opinions about whether the substitution of a defined term for the phrase "means of identification of another person" would alter the mental element of the offence. Say the statute read:

    (1) "Whoever … knowingly transfers, possesses, or uses, without lawful authority, a false ID shall" etc
    (2) A "false ID" means a means of identification of another person

    Would that change the scope of "knowingly"?

  7. linda seebach said,

    February 26, 2009 @ 8:33 am

    This issue was central to a debate in the Colorado legislature several years ago, about when developers were liable for treble damages in cases of construction defects because the defects were "willful and knowing." The plaintiffs' bar argued that no one ever put a roof on a house other than "knowingly" and therefore if it leaked the treble damages provision applied (and moreover, applied to every house in a development even if only one roof leaked). Because the financial risk was intolerable, insurers settled claims of this kind and then refused to write any more policies. The law was modified, although not fully in the direction of common sense.

  8. jfruh said,

    February 26, 2009 @ 8:49 am

    Argh, everybody, of course you're right; I missed the point. That's why you shouldn't read (or comment on) blogs late at night?

  9. Nathan said,

    February 26, 2009 @ 9:32 am

    @Jon: IANAL (lawyer nor linguist), but it seems to me your definition of "false ID" is a problem. The defendant knowingly used an ID not legitimately his. His defense is that he didn't know he was using someone else's numbers, but thought they were made up numbers not assigned to anyone–along with the claim that the law as written treats this differently, which is the matter of interpretation under discussion here.

  10. Chris said,

    February 26, 2009 @ 9:32 am

    ISTM that the prosecution's interpretation here is absurd. You can't unknowingly use an identification ("use" implies purposeful behavior); you can only use an identification without knowing whether or not it belongs to another person, which is exactly what the defendant in this case claims to have done. Therefore, the prosecution's interpretation would effectively read the word "knowingly" out of the statute altogether, which the court is not permitted to do. (Well, it's the SC, so the permission is sort of self-imposed. But they theoretically have an ethical principle against deleting words from statutes.)

    Isn't it customary in criminal statutes for "knowingly" to cover all the remaining elements of the offense? The use of "knowingly" indicates a legislative intent to exclude people who accidentally commit acts that would be criminal if committed intentionally – which is exactly the present case.

    P.S. In retrospect, it seems absurd not to include an intent element – something like "with the intent to induce in another person, organization, government or legal entity the belief that the offender is the person whose identification is used" – which, if it had been included, would clearly have excluded this defendant. But maybe the legislature believed that the scienter element would adequately cover this, as long as its scope wasn't unreasonably reduced by strained interpretations.

  11. Jason said,

    February 26, 2009 @ 11:07 am

    @Neal: Thanks for the comment and the links (and sorry for not identifying you in my post)! The transcript of oral argument makes for fascinating reading. I was especially struck by the govt's argument mentioning Chin (p.34 l.23ff.):

    Or take the Chin statute. The Chin statute says "knowingly and intentionally uses, hires or employs a person under the age of 18 to avoid detection of drug trafficking crime."
    In Chin the D.C. Circuit said in every other court of appeals who have considered the question has said the defendant does not need to be specifically aware that the individual in question is less than 18 years old.

    Here the courts have sided with a de re reading of the indefinite "a person under the age of 18", an issue addressed in your amicus brief but not in my post. For the present case, the question then becomes the trickier one of whether the descriptive content of "a means of identification of another person" could be read de re (or 'anchored' to the knowledge-world of someone other than the subject)…

  12. Richard Humphrey said,

    February 26, 2009 @ 3:46 pm

    I am a drafting lawyer who has drafted legislation in Australia, the UK and Ireland. The approach to drafting offences in the Australian (Commonwealth) Criminal Code avoids this problem–it requires the drafter to identify precisely both the facts or circumstances that constitute the physical elements ("actus reus") of the offence, and the requisite state of mind of the accused in relation to each–there can be different states of mind for different acts or circumstances. In this case, the drafter would probably not have used "knowingly" at all–he or she would have said something like "uses a social security number that was not issued to him or her and knows that the number was not issued to him or her", depending on what the physical elements of the offence were. The drafter might well need to use a lot more words, but drafting time is cheaper than appellate court time.

  13. Noetica said,

    February 26, 2009 @ 5:09 pm

    Here the courts have sided with a de re reading…

    Surely the distinction between de re and de dicto readings is at the core. I was surprised not to find even an oblique reference to these terms before this point in the discussion, though I note that the matter is covered admirably in the splendid amici curiae brief.

    And indeed the term acteus reus (not occurring in the amici brief) is worthy of the airing Richard has given it. I believe it is connected etymologically with the re of de re; in Roman jurisprudence the case tried was the res. For completeness we should also mention the converse mens rea (occurring in the amici brief). Surely it all goes like this, in the case of Oedipus:

    Oedipus knew de re that he was marrying his mother; that act of marrying was an actus reus.

    Oedipus did not know de dicto that he was marrying his mother; Oedipus lacked mens rea.

    A pity for Oedipus that strict liability applied, which effectively removes the requirement for mens rea.

    Have I got it right?

  14. dr pepper said,

    February 26, 2009 @ 7:16 pm

    I would imagine that the Erinyes have little patience for legal niceties.

  15. Dan T. said,

    February 26, 2009 @ 10:18 pm

    A "criminal lawyer" can be parsed as a lawyer who's a criminal, can't it? Though, in both this and the more standard interpretation of the phrase, the adjective "criminal" modifies the noun "lawyer", so the difference is not a matter of which word or phrase to which a modifier is attached.

  16. Kenny Easwaran said,

    February 26, 2009 @ 11:24 pm

    Am I the only one with the intuition that the reading on which the sentence comes out true is an actual possible reading of the sentence, even if not the preferred one? But maybe my intuitions are off because of philosophical training.

  17. Noetica said,

    February 27, 2009 @ 1:10 am

    Kenny Easwaran, am I the only one struggling to make sense of your post?

    You report having this intuition:

    …that the reading on which the sentence comes out true is an actual possible reading of the sentence, even if not the preferred one.

    Questions for you:

    1. Which sentence do you mean? There have been quite a few! I have scanned back and forth in vain to see if one stands out as an obvious candidate. Is it the sentence cited from the statute? Is it one of the initial poster's six numbered sentences?

    2. What do you mean by "an actual possible reading"? For philosophers, all that is actual is also possible, though not all that is possible is actual, so I would have thought that "an actual possible reading" is simply a way of saying "an actual reading". But then, what is "an actual reading"? One that has in fact been canvassed? Where? By whom?

    3. Assuming (charitably) that you intend "a genuinely possible reading", why would you assert this:

    The reading on which the sentence comes out true is a genuinely possible reading of the sentence.

    4. What is "the preferred" reading of the sentence (whichever that sentence turns out to be)? Why is it the preferred reading?

    5. Generally, how might your post be more lucidly paraphrased so that it genuinely communicates what you intend?

  18. Chris said,

    February 27, 2009 @ 9:47 am

    The Chin interpretation is even worse than the government's interpretation in Flores-Figueroa – there's no other word that "knowingly" could meaningfully modify in the statute at issue in Chin. "Avoid" is a verb of purposeful behavior and therefore can't be done other than knowingly and intentionally; if "knowingly" doesn't apply to the fact of age then it can't reasonably apply to anything, which amounts to removing it from the statute.

    That's the problem with de re readings in this case and in Chin – they conflict with the legislature's use of the specific word "knowingly", which indicates legislative intent to include mens rea as an element of the offense.

    Overall, I think this proves Richard Humphrey's point – more explicitness would have been preferable – but allowing courts to strip elements out of a criminal statute strikes me as very dangerous.

  19. mollymooly said,

    February 27, 2009 @ 10:40 am

    Laws are still written in human language. A formal language like XML would remove a lot of these ambiguities. Have there been any moves in this direction?
    I guess porting existing statutes to the new format would be much less difficult than porting precedents and judgments. So maybe such a change would be more feasible in civil law than common law jurisdictions.

  20. David Schwartz said,

    March 2, 2009 @ 7:03 am

    If knowingly covers "without lawful authority", as it pretty much has to, then it must also cover "of another person". I don't think there's a linguistic issue here though and I think the original post doesn't make any sense.

  21. Noetica said,

    March 2, 2009 @ 8:48 pm

    David Schwartz, you write:

    If knowingly covers "without lawful authority", as it pretty much has to, then it must also cover "of another person".

    In what sense of cover does knowingly "cover" without lawful authority?

    I'm sure others would like to see an answer that responds to the nuanced analysis knowingly receives in the amici brief, in the original post, and in subsequent posts here. You write:

    I think the original post doesn't make any sense.

    That's a pretty weighty assertion. Please give your evidence and your arguments.

  22. Dan_ad_nauseam said,

    March 3, 2009 @ 1:11 am

    In modern statutory drafting, use of the adverbs "intentionally," "knowingly," "recklessly," and "negligently" in the criminal context derives from the influence of the Model Penal Code. MPC @ 2.02(4) provides that if one of these adverbs appears in a section clearly applying to one or more elements of an offense, it also applies to all of the unmodified elements.

    Unfortunately, although Congress was influenced by the MPC in drafting many of the current federal criminal statutes, it did not adopt Section 2.02(4).

  23. Breffni said,

    March 3, 2009 @ 8:40 am

    I think Noetica is right in saying the de re / de dicto distinction is the key. It’s true that there’s no structural ambiguity, but once that’s established, the question remains: is it really the case that “knowingly” semantically modifies the whole VP in the sense that the subject's knowledge state must correspond to every aspect of the state of affairs in the VP as described (de dicto)? And, contra Jason and the amici, I think the answer to that is no. The same applies to the argument about the indivisibility of object NPs: the subject does not necessarily have to know that the referent was as described in the NP.

    Jason’s Oedipus examples, and the relevant ones from the brief, are unhelpful, because they use actions that can't be undertaken unknowingly: marry, hire, submit, use, etc. In other words, the reason "Oedipus knowingly married his mother" doesn't seem ambiguous is because the nature of marrying rules out one of the interpretations. In a world where you could unwittingly marry, it would indeed be ambiguous. If we substitute a verb that can bear an "unknowing" interpretation, the ambiguity is obvious: "You knowingly misled my mother." It would be perfectly natural to say this to an insurance salesman who didn't know me and didn't know that the lady he cheated even was a mother. In that context it corresponds to "You knowingly misled someone, who, by the way, happens to be my mother".
    So "my mother" has a de re reading in that case. A similar case, though with a different mental-attitude adverb, would be the journalistic "He deliberately ran over the father of three" – this doesn't have to mean that he sought out a father of three, or ran over this one because he was a father of three: the expression chosen for the NP slot is just providing the reader with additional information. So in cases like these, the verb is read de dicto but the object needn't be.

    Then there's this case: "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.” This doesn't mean the bouncers knew they were admitting an undercover officer; presumably that's not the case. So "an undercover officer" has to be read referentially, de re. The point here is that they knew the individual was "armed with a firearm" – the postmodifier is the element that corresponds to the culpable knowledge, i.e., has to be read de dicto. So even at the level of the NP, it isn't a case of all or nothing, as implied by Jason and the brief.

    So I think the point is that some element of the VP – V, N, a modifier – has to be read de dicto for "knowingly" to make sense, but it's open for some or all other elements to be read de re, with the precise form determined by considerations of audience design. Which are de re and which de dicto is a problem of inference: if that last example came from an internal report on security at an underworld den, the problem might be knowingly admitting undercover cops rather than knowingly admitting people with guns. What's going on here, I think, is pragmatics, not just syntax.

    I agree, incidentally, that the most natural reading of the statute in question is the one advocated by the petitioner.

  24. Joe Shuren said,

    April 14, 2009 @ 8:59 pm

    I take it your brief interprets the "Whoever … knowingly transfers, possesses, or uses, without lawful authority, a means of identification of another person" as

    A. (The prosecution must prove wilful intent)(to commit serious fraud)(unless there is circumstantial evidence)(that the accused knows that he does all of:)(1. transfer, possess, or use)(2. without lawful authority)(3. a means of identification)(4. of another person).

    Here is an alternative interpretation:

    B. (Given the facts proven of)(1. transfer, possession, or use)(2. of a means of identification)(3. of another person)(4. without lawful authority)(unless there is circumstantial evidence)(then the jury is warranted to conclude beyond a reasonable doubt there is wilful intent to commit serious fraud).

    Here are two problems. 5. It is unfair to convict two people of the same offense and punish one more than another if the SSN in one case is random and the other is that of another person. 6. The Congressional committee drafted the act so that the penalty would be more severe for those who used the ID of a real person than a random number. A favors 5, B 6.

    I wonder if your brief was understood by the Court and if in the end the problems can be resolved, no matter the decision.

    FF first used a SS card with a fake name and SS number, apparently not used by anyone else. Six years later he went to his employer to use his real name and a SSN that belonged to a different name. The system works this way: SSS enters the name and SSN in their computer, if they do not match what is in the computer, the employer is notified. Since it was his real name and this time the SSN was attached to a different name, it fell out, and he was arrested and was guilty of immigration fraud. The question then was whether it was aggravated identity fraud, and whether to apply A or B.

    I think the key first is that circumstantial evidence shows prima facie he could not have intended identity fraud. He used his real name, not the name of the person attached to that SSN. He could still be guilty of immigration fraud of course.

    The second key is that it is really not relevant whether the SSN belonged to another person or not, or whether that person was dead or alive, or whether FF intended to commit fraud. The difference is that a law intended to protect against ID fraud is being used to elevate the penalty for ordinary immigration fraud. This is exactly what was done in Postville. The workers there could also have shown circumstantial evidence they could not have intended ID fraud (many of them, Mayans, could hardly speak Spanish and had no idea what a SS card was) but to do that they would have to stay in jail much longer and risk the same deportation after much expense. With the Postville success, ICE uses this in other jurisdictions.

    The third key is in the EPIC brief. "Means of identification" are not IDENTITY, they are "identity attributes." The Social Security Card used to say on the back it must not be used for identification. If someone shows you a SS card with the name "Jon Burrows" and number "409-52-2002" it might even pass the SS computer but there are several names attached to it including "Elvis Presley." Many people have many identity attributes and even identities such as the pseudonym I use here. That does not prove fraud, or aggravated fraud, on either A or B. It does prove that B makes grammatical sense just as much as A, but is just as wrong.

    It would be best if the Court could throw out the law here, but for reasons other than what each side gave in arguments. Congress needs to make sure the Social Security number does not pretend to be a secure identification and that there is a better way to determine immigration status. The best way to prevent identify fraud is security, and that means multiple identity attributes, flexibility in a distributed system, not fragility in a centralized system.

    I hope linguists keep contributing to briefs to inform the legal system. But language is a form of communication, speech acts involve not only statements of mental states of speakers, but also predictions of responses of the audience in the light of what Justice Roberts called the "context" of the real world underneath our grammar.

RSS feed for comments on this post