Robocalls, legal interpretation, and Bryan Garner (the first in a series)
« previous post | next post »
A few weeks ago, Mark’s post “Lawyers as linguists” alerted me to Facebook v. Duguid, a case now pending before the Supreme Court, which grabbed my attention for several reasons. First, the case presents an interesting linguistic issue. Second, the parties on both sides have framed their linguistic arguments in terms of three of the canons of interpretation in Reading Law: The Interpretation of Legal Texts (2012) the book coauthored Bryan Garner and the late Justice Antonin Scalia, and I’ve previously criticized the canons at issue (e.g., here). Finally, Garner himself is on the legal team representing the plaintiff, Noah Duguid.
An unusual confluence of circumstances.
This post will be the first of several in which I’ll discuss the relevant canons, the issues in Facebook v. Duguid, and (to some extent) the parties’ arguments. In this post, I’ll begin by providing a quick introduction to the canons in question, the legal context in which the linguistic issue arises, and the linguistic issue itself. The next few posts will deal with the canons, and I’ll conclude with a post or two giving my take on the case and commenting on some of the arguments that the parties made.
“CANONS” OF INTERPRETATION are essentially rules of thumb that are intended to guide judges in dealing with issues of legal interpretation. Such canons (originally referred to as “maxims”) have been part of Anglo-American law for centuries, and the core of Reading Law consists of 57 such canons, all of which are described as reflecting well established law. But that description is inaccurate as to the canons that are in play in Facebook v. Duguid, all of which differ in some respect from the law as it existed when Reading Law was published.
The canons in question are the Last Antecedent Canon, the Series-Qualifier Canon, and the Nearest Reasonable Referent Canon (although only the first two are cited or discussed in the briefs by name). They are set out below, as they appear in the book, but without the accompanying commentary:
Last-Antecedent Canon. A pronoun, relative pronoun, or demonstrative adjective generally refers to the nearest reasonable antecedent.
Series-Qualifier Canon. When there is a straightforward, parallel construction that involves all nouns or verbs in a series, a prepositive or postpositive modifier normally applies to the entire series.
Nearest-Reasonable-Referent Canon. When the syntax involves something other than a parallel series of nouns or verbs, a prepositive or postpositive modifier normally applies only to the nearest reasonable referent.
I’ve previously written about these canons at LAWnLinguistics, and in those posts (collected here), I criticize the canons for several reasons. In addition to pointing out their departure from preexisting law, I said in my first post that “they don’t hang together as a coherent set of principles.” In a later post, I said that I’d come to believe “that the framework established by Reading Law is not merely problematic but deeply flawed.” That remains my conclusion, but at this point I’ll be more blunt.
The Last Antecedent Canon, the Series Qualifier Canon, and the Nearest Reasonable Referent Canon collectively add up to a confused and confusing mess. The distinctions they draw are arbitrary, and Series Qualifier and Nearest Reasonable Referent Canons are in conflict with one another: they overlap in their coverage, and within the area of overlap their interpretive prescriptions point in opposite directions.
As between the two coauthors of Reading Law, I assume that it is Garner who bears primary if not exclusive responsibility for these problems. It seems likely to me that the substance of the canons at issue, and of the book’s commentary on them, probably reflects Garner’s ideas much more than Scalia’s. After all, these canons concern grammatical issues as to which Scalia would probably have deferred to Garner’s presumed expertise. So I will proceed on the assumption that the substance of the Last Antecedent, Series Qualifier, and Nearest Reasonable Referent Canons, along with that of the accompanying commentaries, is attributable to Garner alone (with the assistance of staffers at his company, LawProse).
I’LL TURN NOW to the background of Facebook v. Duguid and of the issue the Supreme Court will be called on to decide. Warning: things will get a bit complicated.
The case began when Noah Duguid filed a class-action suit against Facebook for violating the Telephone Consumer Protection Act, a 1991 statute that regulates telemarketing. He accused Facebook of having violated the statute’s restrictions on sending texts by means of what the statute referred to as an “automatic telephone dialing system”—an “ATDS,” for short. Facebook argues that the texts about which Duguid is complaining were not sent using such a system, and at issue is how to interpret the statute’s definition of an ATDS.
That definition reads as follows:
The term “automatic telephone dialing system” means equipment which has the capacity—
(A) to store or produce telephone numbers to be called, using a random or sequential number generator; and
(B) to dial such numbers.
The dispute concerns the phrase using a random or sequential number generator. The specific question is whether the phrase should be understood as modifying both store…telephone numbers and produce telephone numbers (Facebook’s position) or as only produce telephone numbers (Duguid’s).
Disputes framed in this way (“which of two words or phrases does phrase X modify?”) usually call for resolving a grammatical ambiguity. For example, in the phrase infringement of copyright or unlawful use of slogans in your advertising, does in your advertising modify everything that precedes it, or just unlawful use of slogans? But the case we’re looking at, I think the disagreement, might be more appropriately seen as concerning whether there exists an ambiguity needing to be resolved.
Facebook argues that as a matter of English grammar, using a random or sequential number generator unambiguously modifies both store…telephone numbers to be called and produce telephone numbers to be called. And as I read Duguid’s argument, his only basis for disagreeing with that conclusion is the seeming incongruity of referring to storing telephone numbers using a number generator. His brief strikes me as waffling somewhat on whether he is contending that the incongruity causes the definition to have a different grammatical structure than would otherwise be the case. On the one hand he argues that “construction of a text often is governed not by the rules of syntax but by the sense of the passage” and that statutory interpretation does not turn “on grammar alone,” but on the other hand he says that “the correct grammatical reading of the definition aligns with the semantic content of the words.”
To add yet another complication, the two sides disagree about whether the apparent incongruity actually exists. As a result, the proper interpretation of the statute might well turn on factual issues having to do with the technology of generating and storing telephone numbers, a kind of situation I don’t think I’ve ever encountered before. But for the purpose of these posts, I’ll assume that the anomaly is real.
Before the case reached the Supreme Court, the specific allegations in Duguid’s complaint (which showed that the texts he received had specifically targeted at his phone number), were held to rule out any claim that the texts were the result of phone numbers that had been produced using a random or sequential number generator.
The viability of his suit therefore depends on persuading the Supreme Court that while the statute requires him to show that the texts were sent by a system capable of storing telephone numbers, he isn’t required to show that the system was capable of doing so by using a random or sequential number generator.
The case will be heard on December 8, with Garner arguing on behalf of Duguid.
Coming up next: The precursors of the Scalia/Garner Canons.
Ken said,
November 30, 2020 @ 10:59 pm
Speaking as neither a linguist nor a lawyer, but as a computer programmer: If someone were to come to me and say, "We require that the software will store numbers using a random or sequential number generator," I would question whether they understood what all the words meant. Maybe you could argue that storing them with a hash algorithm would qualify, but it's a stretch.
Ryan W said,
November 30, 2020 @ 11:02 pm
If you're not already aware, you may be interested to know that Karl Llewellyn published a short article in 1950 called "Remarks on the Theory of Appellate Decision and the Rules or Canons about How Statutes Are to Be Construed." In it, he criticizes the canons and arranges them into two columns of opposing principles. A quick Google search will lead you to a free copy of the publication.
eub said,
December 1, 2020 @ 12:24 am
I'm with Ken. You can dial in sequence, you can dial randomly, you can run down a stored list. The law's wording is poor, but I don't see real ambiguity about what it describes. Facebook should need to demonstrate an example of a dialer that "stores numbers to be called, using a random number generator" that the authors could have had in mind.
GH said,
December 1, 2020 @ 2:12 am
Speaking also as a programmer, I disagree with Ken. I see "store numbers to be called, using a random or sequential number generator" as a perfectly coherent software requirement, which would be fulfilled by a system that generated random telephone numbers and stored them in some database table or list, which is then later on used to dial those numbers.
The apparent redundancy of "store or produce" can easily be explained as an attempt to close a possible loophole so that the law would cover the situation whether the system generates the random telephone numbers ahead of time (and is potentially distinct from the system that actually dials them), or on-demand as needed.
If the law is intended to cover any system that stores telephone numbers and allows them to be dialed, whether or not the numbers are randomly generated, then it encompasses every cell phone on the market as well as every landline with a speed dial function, and it becomes technically illegal to call a hospital room from your cell phone (other than for emergency purposes or with express permission). It is a plainly absurd reading of the law.
Gregory Kusnick said,
December 1, 2020 @ 2:26 am
If you used your cell phone for unattended bulk dialing of numbers stored in your contacts list, then it would be an ATDS within the meaning of the statute.
Facebook's interpretation that a system qualifies as an ATDS only if it dials numbers it has generated internally (sequentially or randomly, on demand or at some previous time) opens the door for spammers to robodial stored lists of harvested numbers with impunity, on grounds that they're not technically using an ATDS. It seems unlikely that that's what the legislators intended.
GH said,
December 1, 2020 @ 2:29 am
Actually, let me revise my comment slightly.
On reflection, I believe the intended meaning of "store numbers to be called, using a random or sequential number generator" is most probably "store numbers to be called [that have been produced] using a random or sequential number generator." This has the same intended effect as my initial interpretation: closing the loophole where you could simply separate the system that generates the random/sequential telephone numbers from the system that dials them.
It is badly written and conceived because of course any system that can store a list of telephone numbers can store an externally provided list of random or sequential ones, and because having the "capacity" to do so is not the same as actually doing so. This would imply that the literal meaning of the law's definition would in fact encompass pretty much every telephone, surely unintentionally.
GH said,
December 1, 2020 @ 2:49 am
@Greg Kusnick
But the definition does not speak of how I actually use my phone. It speaks of a system having the "capacity" to do it. And since my phone has the capacity to store numbers and dial them, it is an ATDS by your interpretation, and any and all of the strictures of the law and regulations apply—not just rules against bulk dialing. This would make it pretty much impossible for any company to call any potential customers with commercial intent.
I find it easier to believe that the drafters did not consider (or chose not to address) harvested lists of telephone numbers than to believe that they intended to outlaw conducting business via telephone altogether.
Peter Taylor said,
December 1, 2020 @ 9:31 am
Another possible construction, which I don't think is the same as any described in the blog post or preceding comments, is that the antecedent of "using" is "called". This is the most recent antecedent, and I argue that it's a reasonable construction. It would select any device which can either retrieve or generate numbers and then use a deterministic or non-deterministic ordering on those numbers to feed them to a dialling module.
Neal Goldfarb said,
December 1, 2020 @ 11:00 am
(Responding to Peter Taylor)
The comma immediately after “to be called” weighs against that interpretation, as a matter of both legal interpretation and comprehension.
Gregory Kusnick said,
December 1, 2020 @ 11:42 am
GH: If the Museum of Telephone Technology has amongst their exhibits an old-school ATDS of the kind envisioned by the statute, they are presumably not in violation unless they actually put it to use for soliciting donations.
The word "automatic" has to count for something. A system that dials a single number in response to human input is not, in my view, an automatic TDS; automatic implies unattended. So cell phones as ordinarily used would be excluded on this view.
GH said,
December 2, 2020 @ 8:47 am
@Gregory Kusnick
We have to separate the definition of what counts as an ATDS from the rules governing their use. The law does not ban the existence of ATDSes, it only prohibits making certain calls from them, so it would obviously not affect a museum exhibit. But the definition of what counts as an ATDS does not depend on how the system is actually used, only its "capacity."
But all these arguments were hashed out in the comments to the earlier post, and having read through them I don't think I have much to add. I don' believe there is a plausible reading of the text of the law that prohibits Facebook's actions without also potentially prohibiting a huge number of routine, legitimate telephone calls and messages. Nor am I at all persuaded that this kind of telephone messages were what the law's drafters intended to prohibit.
And taking a step back from what the law actually says, I don't think that as a matter of good policy the messages in question should be illegal simply because they were automated. It wouldn't have been any better if they had been sent out manually by some office admin or agent working off an incorrect entry in the contact list that was never corrected, while on the other hand they would have been fine had they gone to the right person, whether or not they were sent automatically.
From reading only the summary of the case, the actual issue is rather one of (unintentional) harassment that the company failed to adequately address. That's the sort of thing they should probably be liable for regardless of the technology involved.