March 21, 2005

Axe a stupid question

Safire's "On Language" column in the New York Times Magazine of 3/20/05 (p. 26) has a segment on nucular that strikes me as deeply confused.  It also fails entirely to mention Geoff Nunberg's book Going Nucular and the piece that gave the book its title (not to mention the various Language Log postings that have discussed this pronunciation: here, here, here, here, and here).  I would have thought that the title of Geoff's book would be a pretty clear hint that the pronunciation nucular was going to be discussed somewhere in the volume.

Instead, Safire asked for Steve Pinker's advice, and Pinker came up with a metathesis account -- Safire dutifully defines "metathesis" and indicates its pronunciation -- that can't be the right story.  Here's how Safire starts his analysis:

Many of us replace an unfamiliar sequence of phonemes (the smallest units of speech sounds) with a familiar one.  The only other common word that rhymes with nuclear is the unfamiliar cochlear.  But in our spectacular language, there are dozens of words like secular, vascular, jocular and molecular, and our brains are tempted to make nuclear fit that familiar pattern.

I'm not sure how to share out the blame here -- to Safire, to Pinker, to Safire for consulting Pinker instead of a linguist who's thought about the phonology at issue.  But there are at least four problems with what Safire says, and two of them are serious.

Problem 1, a minor annoyance: "the smallest units of sounds" isn't going to elucidate the notion of "phonemes" to anyone who doesn't already know what phonemes are.  It's just baffling.

Problem 2, more serious: "an unfamiliar sequence of phonemes".  As Geoff points out in his book, the /liǝr/ at the end of "nuclear" isn't at all unfamiliar to or difficult for speakers of English: comparatives like pricklier are unproblematic and show no inclination towards being reshaped.  The problem with nuclear isn't phonological but morphological, and that's why words in -cular /kyǝlǝr/ are relevant; they appear to have some sort of root ending in c /k/, followed by morphological elements ul /yǝl/ and ar /ǝr/, or perhaps an indivisible ular /yǝlǝr/.  (Back on 7/3/04, in fact,  Alison Murie suggested on ADS-L that nucular might be a reanalysis in which the root is the word nuke, and the word nucleus isn't involved at all.  And Geoff entertains a similar idea in his article, noting that this would predict a difference, for some people, between nuclear in things like nuclear family and nucular in things like nucular weapons.)

Problem 3, also serious: getting the metathesis proposal to work.  Metathesis of the /l/ and /i/ of /nukliǝr/ would give /nukilǝr/, with primary accent on the first syllable and secondary accent on the second (as in nuclear).  To get towards nucular, that second syllable would have to lose its accent (this is not particularly unlikely), yielding /nukIlǝr/ or /nukǝlǝr/.  This isn't all the way home, though, because there's still that /y/ to pick up.  It looks like Safire is assuming a metathesis and then a reshaping to match other -cular words, which would supply a /y/.  But direct reshaping is a more parsimonious account of the phenomenon; the metathesis is unnecessary (as well as insufficient).

Problem 4, another mere annoyance.  Safire is being really sloppy when he says that nuclear rhymes with cochlear.  It doesn't, because the accented vowel /u/ of nuclear doesn't match the accented vowel /o/ or /a/ (depending on your pronunciation) of cochlear.  ( If they "rhyme", then so do noodles and models.)  Rhyme involves a matching between accented vowels and everything that follows them.  The pair nuclear/cochlear is a kind of almost-rhyme, in which everything that follows the accented vowels matches. Almost, but definitely no cigar.

[An earlier version of this critique appeared on ADS-L on 3/21/05.  My thanks to Geoff Nunberg for suggesting the title.]

zwicky at-sign csli period stanford period edu Posted by Arnold Zwicky at March 21, 2005 10:46 PM