The 'Lawyer v. Human' Problem in Corpus Linguistics

Abstract

New Prime Inc. v. Oliviera, which the Supreme Court will decide this term, exemplifies both the virtues and potential pitfalls of corpus linguistics — that is, the enterprise of resolving lexical ambiguity by reference to popular usage patterns. The main question in New Prime is whether the phrase "contracts of employment," in the section of the Federal Arbitration Act that exempts "contracts of employment of seaman, railroad employees, or [other transportation workers]" from its reach, includes agreements with independent contractors. In other words, when the FAA refers to "employment," is it talking about the act of working writ large, or is it describing the more technical idea of being an employee?

Evidence of usage at the time of the Act's passage (in the 1920's) points different ways. Some sources bolster the broader, less formal construction — i.e., that "employment" is a synonym for "work" — while others support the more technical construction. One might think, in light of this ambiguity, that New Prime would be an ideal candidate for examination of the corpora.

In this paper, we use New Prime to sound two cautionary notes — in a skeptical but friendly spirit — about corpus linguistics. First, whether examining usage patterns is a useful way of resolving a case like New Prime depends, as it would in any case, on whether the phrase “contracts of employment” should be understood as written-by-lawyers ("WBL") or written-by-humans ("WBH"). The thrust of New Prime's argument, at bottom, is that the phase "contract of employment" is WBL, and that most lawyers who use the phrase (or would have used the phrase in the 1920's) would have meant something more technical and specific than "work." If New Prime is right about this, it would almost certainly render large swaths of usage evidence, as invoked by Mr. Olivier, irrelevant — and likewise the inverse. The trouble is, nothing in the usage patterns themselves can resolve (or perhaps even address) the WBL v. WBH question; which gives rise to a significant danger that when all is said and done, debate about usage patterns will end up turning on a first-order conceptual dispute about which sources are relevant, not a second-order empirical dispute about what those sources convey. Scholars and judges must be careful, in other words, to develop a theory of usage — as between WBL and WBH — before moving onto to the inductive question that corpus linguistic is best-suited, in principle, to resolve.

Second, even assuming we could resolve the WBL v. WBH question to satisfaction (and assuming further that one’s use of corpora could be tailored accordingly), the meaning of the phrase may depend on what the writer — lawyer or human — was trying, as J.L. Austin might say, to do with it. When this is true, meaning will depend not only on the usage patterns of phrase itself, but also on usage patterns of functionally similar phrases, since the plausibility that a speaker used Phrase X to bring about Effect Y will depend, in part, on how useful Phrase X is for bringing about Effect Y (measured across cases) but also in part on how useful other phrases would be for performing the same linguistic work. In New Prime, for example, it is far from obvious whether a different phrase, such as "labor contracts," might have conveyed the broader meaning urged by respondent more efficiently or felicitously than the phrase actually used. Perhaps yes, perhaps no; the point is that it would be strange, and maybe even misleading, to limit the empirical question to how the phrase "contract or employment" (or more simply, the word "employment") has historically been used, given the potential availability of other options. But this, in turn, would give rise — in at least some cases — to serious conceptual complexities, some of which might threaten the utility of looking to usage patterns at all.

JEL Classification: K10