Fun with commas

« previous post | next post »

For your reading pleasure this morning: Kenneth Adams, "Bamboozled by a Comma: The Second Circuit’s Misdiagnosis of Ambiguity in American International Group, Inc. v. Bank of America Corp.", 16 Scribes J. Legal Writing 45 (2014–15):

In its opinion in American International Group, Inc. v. Bank of America Corp., the United States Court of Appeals for the Second Circuit invoked the principle of construction that this article refers to as “the comma test under the rule of the last antecedent”: if in a sentence a series of nouns, noun phrases, or clauses is followed by a modifier and the modifier is preceded by a comma, the modifier applies to the entire series, not just the final element in the series.

But as the opinion inadvertently demonstrates, that principle of construction is inconsistent with English usage and should be rejected. The opinion also serves as a reminder that judges cannot always be counted on to understand how ambiguity operates; courts should permit expert-witness testimony on ambiguity.

We can also add this to the list of reasons that legal education should include a course or two in practical linguistic analysis.

You can find some background on Ken Adams' blog here: "My Forthcoming Article, 'Bamboozled by a Comma: The Second Circuit’s Misdiagnosis of Ambiguity in American International Group, Inc. v. Bank of America Corp.'", 10/7/2013.

See also Ken's newpaper piece about the million-dollar comma: "Behind the scenes of the comma dispute", Globe and Mail 8/28/2007.

And for another — alas more complex — comma dispute, see a couple of LLOG posts, "The indubitable equivalent of such claims", 12/16/2009, and "Civilization, Congress, and punctuation", 12/16/2009.


  1. John Baker said,

    August 17, 2015 @ 9:14 am

    The court's decision,, actually appears to come to the correct outcome, even as it burdens its opinion with a page and a half of nonsense based on the court's misapprehension of grammar. It's another example showing that easy cases make bad law: the judgment may have been correct, but the bad grammatical precedent lingers and will influence other courts, particularly those district courts from which appeal lies to the Second Circuit (those in Connecticut, New York, and Vermont).

  2. Pete said,

    August 17, 2015 @ 9:28 am

    Fantastic article. It shows how simple it is to test the validity of any linguistic claim with a few examples.

    However, most laypeople still labour under the delusion that English grammar exists not in the brains of English speakers but in a canon of usage guides and dictionaries.

  3. Stephen Goranson said,

    August 17, 2015 @ 9:30 am

    "A Comma in Catullus," an Aug. 15 post by Patrick Finglass, relates an emendation by A. E. Housman that has been called (by C. J. Fordyce) "the most spectacular contribution of modern scholarship to the understanding of Catullus."

  4. Guy said,

    August 17, 2015 @ 10:38 am

    It's not clear to me that statutory interpretation interpretation must always match the way that people express themselves in ordinary writing. Laws are a specialized form of writing, and it could be argued that there is a use for court-invented interpretive rules that remove ambiguity in a category of cases that would ordinarily be ambiguous. Of course, it's a bit unfair for a court to spring such a rule on someone while purporting to only be interpreting according to ordinary principles of usage.

    The particular rule in question looks to me like an example of a common fallacy made by people when discussing language that could be called "symmetry imposition". If there are two forms, A and B, and A is ambiguous between meanings X and Y, but B either unambiguously or typically means Y, then it's tempting to conclude that A always means X (and that B always means Y, to the extent it doesn't already). The which/that rule is a good example of this kind of reasoning. One reason why an author might insert a comma before the adjunct is to show that it is outside the scope of coordination, but there are many other reasons to do so that are consistent with it being inside that scope.

  5. J. W. Brewer said,

    August 17, 2015 @ 11:14 am

    Something else funny about the "list" at issue here: the Edge Act applies to lawsuits (I've added line breaks):

    "arising out of transactions involving international or foreign

    or banking in a dependency or insular possession of the
    United States,

    or out of other international or foreign financial

    either directly or through the agency, ownership,
    or control of branches or local institutions in dependencies or
    insular possessions of the United States or in foreign countries . . . ."

    Note that "out of" appears neither once nor thrice but twice, which is not what you would expect in a "A, B, or C" list drafted with strict syntactic parallelism. The question is whether to conclude that this lack of parallelism is supposed to connote a not-strictly-parallel conceptual structure — perhaps "either (A or B) or C," which might cut against the court's ruling by making it more plausible that the "either directly" qualification following item C applies only to C, or whether instead to conclude that the lack of parallelism just suggests sloppy drafting, thereby making it even less plausible that the presence vel non of a comma connotes anything.

    I'm too lazy to look them up myself, but given that Ken Adams was getting a professional publication out his critique, it might have added value for him to pull up the parties' appellate briefs and see what they did or didn't say about the comma issue. It would be helpful to know if the excursion into comma-theorizing was an instance of sua sponte judicial free-lancing (one possible cause of subpar analysis in published judical opinions), or the result of one side making a dubious argument that the other side failed to adequately debunk (another possible cause), or some third thing.

  6. Guy said,

    August 17, 2015 @ 12:34 pm

    @J.W. Brewer

    I don't know enough about the subject matter to judge pragmatically, but syntactically it doesn't look to me like slopping drafting: we have multiple levels of coordination. I would read the coordination of the two "banking" elements as the complement of "involving" – they're both specifying types of "transactions". Then "out of other … operations" is coordinated to expand the scope beyond just "transactions". Or at least that's what it seems to me.

  7. Guy said,

    August 17, 2015 @ 12:43 pm

    My "slopping" instead of sloppy, however, was (autocorrect induced) sloppy drafting.

    It's worth noting that there is further coordination inside each of these levels. And the "either" adjunct is complicated enough to require its own internal commas. I think adjuncts are more likely to be set off with commas when they have their own internal commas, so it isn't necessary to posit that the comma before "either" was placed to indicate that it was outside the top-level coordination to explain its presence.

    The whole section is complicated enough that I think the legislature would have been well-advised to make use of sub-headings (e.g. i) ii) or A) B) or whatever would be appropriate for the level of subsection they were at) to clarify how the various levels of coordination were interacting with each other.

  8. Sergey said,

    August 17, 2015 @ 2:30 pm

    It sounds like there should be a sort of an algebraic notation for the legal documents that would allow the unambiguous formulation. And it would be much easier to read as well, just as the mathematical problems are much easier to read in the algebraic notations than in words.

  9. Eric P Smith said,

    August 17, 2015 @ 5:13 pm

    The general "rule of the last antecedent" is well illustrated by published today, where we read:

    And the Scotsman, who moves to second in the ATP rankings on Monday, revealed the news from his French mentor, the only woman to coach a major men's player who gave birth to a baby boy on Sunday morning.

    The missing comma gives a ridiculous result.

  10. Guy said,

    August 17, 2015 @ 7:19 pm

    @Eric P Smith

    The problem there is the unintegrated appositive needs a comma at the end as well as the beginning, and the "who" clause is obviously meant to be an unintegrated relative so it isn't part of the appositive, not that the relative needs a comma to enable it to attach to the higher head noun in general. For example in "the only animal without legs that I like" the relative "that I like" is most naturally read to be modifying "animal", not "legs", and indeed a comma is not just not required but prohibited after "legs". But in "John, my brother went to the store", this would be read as a vocative to John with "my brother" referring to someone other than John. If "my brother" is intended as an appositive, a comma must be placed after it for reasons entirely unrelated to pronoun reference or later modifier attachment.

  11. George Williams said,

    August 18, 2015 @ 9:28 am

    Some decades ago I began studying how legal reasoning might provide useful insight into semantic theory in general. This led to law school and the practice of law. From time to time, others have done the same.

    The issues touched on here are omnipresent in the law, despite the widespread sensitivity to language use on the part of lawyers. How much it would help to have law students take linguistic courses is unclear, although I have taught some and a number of law schools periodically or regularly teach courses that might count. Some philosophers of law also deal routinely with language-related issues, as do some sociologists and anthropologists.

    Even the issues of comma usage can be complex. The fact that there is a rule of statutory interpretation dealing with this issue suggests judicial desire for closure when more than one answer is possible. Current drafters of statutes and documents, including different regulatory agencies, have learned or imbibed different comma styles. Documents drafted in different places or at different times may vary in their placement of commas without any difference in meaning being intended, and sometimes drafters are just sleepy or careless. It may also be difficult in a particular case to bring in an expert witness who will be able to demonstrate to everyone's satisfaction that the current syntax of English must mean that only one reading is intended.

    The harder cases are even more interesting, and often important issues ride on their resolution.

  12. Eric P Smith said,

    August 18, 2015 @ 5:25 pm

    @Guy: you're absolutely right, thankyou.

  13. Jeff W said,

    August 18, 2015 @ 5:33 pm


    …in "the only animal without legs that I like" the relative "that I like" is most naturally read to be modifying "animal", not "legs"…

    Ha, that reminded me of the following exchange with Gracie Allen, George Burns's wife:

    Man #1: I can't wait to get home and sit by the fireplace, my dog at my feet, smoking a pipe, reading the paper.
    Man #2: [After a pause] Mrs Burns…aren't you going to ask him why he lets his dog smoke a pipe?
    Gracie Allen: Heavens, no! [The men nod and go back to what they were doing.] After all, if his dog is smart enough to read the paper, why shouldn't he be allowed to smoke a pipe?

    Yeah, it’s obviously different comma-wise but I still get a kick out of it.

  14. Neal Goldfarb said,

    August 20, 2015 @ 1:49 pm

    Although I don't have a very high regard for the ability of judges and lawyers to analyze the kind of grammatical issues that Ken's article deals with, I don't think think the case he discusses is a very good illustration of that problem.

    The court's decision is based on the "comma exception" to the rule of the last antecedent:

    Referential and qualifying phrases, where no contrary intention appears, refer solely to the last antecedent. . . . Evidence that a qualifying phrase is supposed to apply to all antecedents instead of only to the immediately preceding one may be found in the fact that it is separated from the antecedents by a comma.

    As legal rules go, the rule of the last antecedent doesn't do that bad a job of lining up with what is known about sentence processing. It's been noted by various people (going back to Larry Solan in The Language of Judges) that the rule results in the interpretation that would be favored by Lyn Frazier's "late closure" parsing preference. And even though her view of the processing architecture seems to have been overtaken by different approaches, I think those other approaches recognize a tendency similar to late closure (e.g., what Ted Gibson refers to as "locality").

    The comma exception to the rule is also linguistically defensible. If you start from the assumption that there's a preference for attaching the modifier low—

    [[item][item] [[item][modifier]]]
    rather than

    [[item][item][item]] [modifier]]

    —the effect of the comma separating the modifier from the last item is to insert an implicit prosodic break before the modifier, which makes it more likely that the modifier isn't parsed as a continuation of the last item, and therefore diminishes the tendency toward low attachment.

    Or at least, that's what I'd argue.

RSS feed for comments on this post