The US Supreme Court recently decided a case in which language was discussed on the basis of corpora. The question was about the words person and personal.

The decision was FCC v. AT&T Inc.
(PDF file), decided on March 1. This is a slip opinion, which means it has not yet been officially published. It has a headnote, which they call a syllabus.

The situation was that AT&T Inc. claimed that as it was a person (all corporations are persons), it could rely on the right of personal privacy.

Language evidence was presented to show that it does not follow from the noun that the related adjective has the same meaning, particularly in compounds.

In fact, “personal” is often used to mean precisely the opposite of business-related: We speak of personal expenses and business expenses, personal life and work life, personal opinion and a company’s view. Dictionary definitions also suggest that “personal” does not ordinarily relate to artificial “persons” like corporations.

I can’t help feeling that the Supreme Court would have come to this conclusion even without the language evidence. It seems pretty obvious to me. But the definition of person has been expanded in recent years, and at all events the Court of Appeals for the Third Circuit found in favour of AT&T.

We disagree. Adjectives typically reflect the meaning of corresponding nouns, but not always. Sometimes they acquire distinct meanings of their own. The noun “crab” refers variously to a crustacean and a type of apple, while the related adjective “crabbed” can refer to handwriting that is “difficult to read,” Webster’s Third New Interna-tional Dictionary 527 (2002); “corny” can mean “using familiar and stereotyped formulas believed to appeal to the unsophisticated,” id., at 509, which has little to do with “corn,” id., at 507 (“the seeds of any of the cereal grasses used for food”); and while “crank” is “a part of anaxis bent at right angles,” “cranky” can mean “given tofretful fussiness,” id., at 530.

To see what linguistic evidence was presented, you can look at Neal Goldfarb’s amicus curiae brief, which ca also be found via his blog.

This amicus brief was filed on behalf of Project On Government Oversight, the Brechner Center for Freedom of Information, and Tax Analysts. The parties have to agree to a filing. The brief lists the dictionaries and other works cited. A partial quote:

The following are the pairings in each corpus that occurred at least ten times, listed in order of
their frequency:
COHA: personal life, personal income, personal property, personal interest, personal experience,
personal relationship, personal problem, personal reason, personal injury, personal thing,
personal appearance, personal contact, personal matter, personal friend, personal power, personal
opinion, personal fortune, personal gain, personal history, personal letter, personal use, personal
view, personal question, personal tragedy, personal physician, personal attack, personal affair…

The brief relied on three corpora: the Corpus of Historical American English (COHA), the Corpus of Contemporary American English (COCA), and the TIME Magazine Corpus, all of which are the handiwork of Prof. Mark Davies at Brigham Young University. What we did was to search for the string personal [NOUN], in order to find out what words most frequently filled the NOUN slot.

This decision seems correct and well-founded, but I can’t help wondering whether all judges can be relied on to interpret corpus evidence properly.

Via Mark Liberman on Language Log, who links to other weblogs on the topic.