On July 10, 2019, the Sixth Circuit considered vexing questions of statutory interpretation in an ERISA case. A dispute over whether a transaction bonus plan was an ERISA employee pension benefit plan hinged on the meaning of two terms common in federal statutes: “results in” and “extending to.” While the meaning of the statute was plain to the entire panel, Judges Stranch and Thapar quarreled over the evidence that a court might rightly consider when interpreting a statute—in this case, ERISA. Judge Thapar argues that “[c]ourts should consider adding [corpus linguistics] to their tool belts.”
Judge Thapar explains that corpus linguistics “draws on the common knowledge of the lay person by showing us ordinary uses of words in our common language.” By allowing “lawyers to use a searchable database to find specific examples of how a word was used at any given time,” corpus linguistics might help judges determine how “the public would have understood a statute’s text at the time it was enacted.” And because, according to Judge Thapar, courts are tasked with interpreting laws according to their “ordinary meaning at the time Congress” enacted them, corpus linguistics should be consulted in cases involving statutory interpretation.
Judge Stranch objects to Judge Thapar’s position for two categories of reasons—philosophical and practical. First, she argues that courts are tasked with identifying “what our elected members of Congress meant when they passed [a] statute—even if that is not the meaning we or the public might routinely employ.” With that in mind, she points out that while consulting legislative history has become unpopular in some circles, it “tells us, at a minimum, how some of the statute’s authors understood a term.” Given this fundamental disagreement about whether the law means what Congress meant or what the public at the time of enactment would have thought it meant, it is no surprise that these judges would depart as to the helpfulness of corpus linguistics and legislative history in statutory interpretation.
But perhaps, with respect to the use of these forms of evidence, this disagreement is one of degree rather than kind. Even assuming that Judge Stranch is correct about the meaning of the law, in the absence of other better evidence—which might include legislative history—it seems that corpus linguistics might have something helpful to say to a judge. Since Congress communicates in English and is made up of members of the speaking public, a judge might conclude that corpus linguistics sheds at least some light on the meaning of Congress’s speech. And to her credit, Judge Stranch seems to admit as much in the penultimate paragraph of her concurrence. So, while it is true that judges who think the law means what Judge Thapar thinks it means will often extract more value from corpus linguistics, accepting Judge Stranch’s first criticism does not necessarily lead to the conclusion that corpus linguistics is totally worthless.
Judge Stranch also argues that, as a practical matter, judges should not themselves dip their own toes into corpus linguistics. First, a “keyword search using a corpus linguistics database will likely result in dozens, if not hundreds or thousands, of examples of a term’s usage.” And when interpreting this large dataset, she argues that judges will be tempted to either adopt a meaning based on cherry-picked examples confirming a predisposition or a pure frequency analysis which can lead to various distortions. For this reason, she argues that this kind of analysis should be left to the experts and, it turns out, she argues, that the experts are the authors of dictionaries. So, in most cases, Judge Stranch argues that the best evidence of a word’s meaning at a particular time can be gleaned from then-current dictionaries. Judge Stranch does concede that corpus linguistics can sometimes offer a useful and more in-depth analysis of some terms or phrases. However, she argues that this analysis should be presented to the court by experts in the field.
Judge Thapar responds by arguing that appeals to databases need not be dispositive, but that looking at contemporaneous examples of words and phrases being used in context can help inform a judge’s sense of such a word or phrase’s most probable meaning. He notes that in statutory construction judges have long analogized to common uses of words and phrases in literature, other laws, and in common parlance. So, he says, a database that compiles such examples can only be of help. And judicial proponents of corpus linguistics would probably be happy enough to see this evidence presented and litigated in front of the court by experts.
One point on which neither judge focused was the peculiarity of ERISA—the statute they were tasked with interpreting. ERISA is an extremely complicated statute purporting to regulate an entire sphere of economic conduct. And many of its exhortations are directed at experts of one kind or another. Accordingly, to understand some of the words and phrases in the statute, it might become necessary to understand what a word might mean, for example, to a plan administrator or tax professional. Perhaps even a judge with enthusiasm for corpus linguistics would find its utility somewhat diminished in many ERISA contexts.
Nonetheless, in briefing statutory construction cases, including those where ERISA is at issue, practitioners may wish to consider looking at corpus linguistics because even where, as before the Sixth Circuit in this matter, the parties ignored this evidence, some judges may themselves be engaging with databases before voting and then drafting their opinions. Indeed it is hard to imagine that the inclusion of such analysis could hurt, as even Judge Stranch does not appear offended by its discussion but rather skeptical of its overall utility. And even a skeptic like Judge Stranch appears willing to accept that experts in this field might have something useful to say to courts. So, the presentation of expert opinions through testimony or in amicus briefs, could prove influential across the judiciary.
And for ERISA practitioners specifically, there is one more point worth emphasizing. While, the case discussed here did not deal with the interpretation of plan documents or communications like SPDs, some cases do. And there is an open question as to whether a judge might try to identify the meaning of words and phrases in such documents by reference to corpus linguistics.
There is certainly some logical appeal for this. Indeed, many documents, particularly those that are distributed directly to plan participants, need to be comprehensible to the average plan participant—who in many cases, particularly for large employers, are indistinguishable from the average person. See e.g., ERISA § 102(a). For this reason, as a prospective matter, this is one more of the many reasons to make serious efforts to compose plan documents and communications clearly and with as little jargon or odd usage as possible. When jargon can’t be avoided, the potential judicial use of corpus linguistics in interpreting a document provides even more reason to define clearly such terms or phrases.
 Of course, this argument would apply equally for a judge, skeptical of legislative history, who thinks that the law means what the public would have thought it meant. For all of the failings that critics have identified and for all of the reasons judges might refuse to cite legislative history (e.g., for fear of creating perverse incentives for members of Congress and their staffs to introduce things into legislative history that could not have survived bicameralism and presentment), it is hard to maintain that legislative history provides no evidence about how the public would have understood the words of the statute at the time of enactment.