Neal Goldfarb (LAWnLinguistics blog) has filed this amicus brief in State of Utah v. Planned Parenthood Association of Utah (the Utah Supreme Court case considering the status of abortion under Utah state law). It principally responds to the corpus linguistics-oriented brief filed by Pro-Life Utah (noted here). From the summary of argument (footnotes omitted):
1. In this brief, Amicus undertakes a critical examination of the corpus analysis set out in the amicus brief filed by Pro-Life Utah (“PL Utah”). As he will show, that analysis is seriously flawed in multiple respects, and it should therefore play no part in the Court’s consideration and decision of this case.
a. PL Utah’s brief involves the use of corpus-linguistic methodology in a way that differs strikingly from how corpus linguistics has generally been used in the context of legal interpretation. Rather than using corpus data as evidence as to the meaning of a word or phrase in a legal provision, PL Utah treats it as evidence of public attitudes toward abortion, primarily during the 1890s. That is to say, it tries to use corpus data as a proxy for a public opinion survey targeting Utahns of the 1890s—a demographic group that no longer exists as such and that, Amicus assumes, has no surviving members.
In order for PL Utah’s data to be considered reliable evidence supporting the conclusion PL Utah wishes the Court to reach, it would have to be shown that the attitudes expressed in the texts in the corpus, whatever they might be, are representative of the relevant attitudes of the overall population of 1890s Utah. And that would require that the authors of those texts be shown to have comprised a representative sample of that population.
PL Utah has made no such showing; indeed, it has not tried to do so. And beyond that, it is undeniable that the authors of the newspaper do not constitute a representative sample of Utah’s population. To begin with, some of the articles originated out of state, and therefore were not the work of Utahns at all. Moreover, census data from 1890 and 1900 shows that Utah’s small population of journalists was predominantly male. So to the extent the articles were written by Utahns, women are likely to have been underrepresented in that group of authors.
The unrepresentative nature of the newspaper evidence becomes especially clear when considering the fact that during the 1890s, Utah newspapers published more than 2,000 advertisements for what were euphemistically called “female pills”: concoctions that were reputed to be effective in inducing miscarriages and that were used for that purpose.This is evidence that, contrary to what PL Utah contends, Utahns in the 1890s were not united in opposition to abortion.
b. Serious flaws are also found in PL Utah’s collocation data. First, the data as presented by the COHA collocation display (and as reported by PL Utah) data consists of what seem to be 33 uses of abortion(s) or abortionist(s). But 22 of those apparent uses reflect multiple counting, in that they come from only five sources, and are therefore attributable to only five authors. When these two flaws are taken into account, the apparent number of relevant uses turns out to have been exaggerated by more than 300%: Rather than 33, there are only 10.
2. With the Court’s indulgence, this brief concludes with a short discussion of several issues that relate generally to the use of corpus linguistics in legal interpretation, and that Amicus thinks it is important for this Court to be aware of.
UPDATE: Link fixed now.
Posted at 6:35 AM