Monday, 21 March 2011

shattered brains

Yesterday I heard an announcement that so-and-so had died of səˈriːbrəl cancer. Personally, I would have said ˈserəbrəl.

Cerebral is one of those medical-anatomical adjectives in which we don’t all agree where the main stress should fall. Other similar cases are skeletal, cervical and our own palatal.

What characterizes these derivatives is a clash between two psychophonological principles. One principle is that of maintaining the shape of the stem to which the suffix -al is attached. Just as person gives us personal and option optional etc, so we may expect to treat the suffix as stress-neutral, giving ˈserəbrəl, ˈskelɪtl̩, ˈsɜːvɪkl̩, ˈpælətl̩, all with initial stress. Compare the initial stress of skeleton, cervix, palate.

(Where the stem has three or more syllables, Chomsky and Halle’s ‘Alternating Stess Rule’ moves the initial stress of the stem to penultimate stress in the adjective, as in ˈuniverse – uniˈversal, ˈdialect – diaˈlectal etc. That is irrelevant here.)

The other principle is the Latin stress rule, largely inherited in English, which says that a long vowel in the penultimate, or any vowel followed by two or more consonants, attracts the main stress to that syllable. So ˈhormone gives us horˈmonal, ˈsepulchre seˈpulchral, and so on. But if the penultimate vowel is part of a ‘weak cluster’, i.e. light — having a short vowel, followed by one consonant or none — then the stress goes not on the penultimate but on the antepenultimate, as ˈpyramidpyˈramidal.

In English as in Latin, to operate the Latin rule it is crucial to know the quantity (length) of the penultimate vowel. But very few people know Latin these days, and even those of us who do may be uncertain about the classical quantity of this or that vowel.

Dentists and anatomists (in BrE at any rate) “know” in some sense that Latin pălātum has a penultimate long vowel, and accordingly pronounce palatal as pəˈleɪtl̩. Phoneticians don’t have this “knowledge”, or “choose” to ignore it, so we pronounce the word as ˈpælətl̩.

The Greek form of skeleton has -λετ-, with a short vowel, which would regularly yield, via Latin, ˈskelɪtl̩; but those who say skɪˈliːtl̩ do not “know” this.

Latin cervix has the stem cervīc-, with a long vowel, and the surgical tradition is to pronounce cervical accordingly as səˈvaɪkl̩; but the pressure of all the hundreds of other adjectives in -ical leads most lay people to prefer ˈsɜːvɪkl̩.

In the case of cerebral, the relevant vowel of Latin cĕrĕbrum is short. The issue here is whether the consonant cluster br is sufficient to ‘make position’, i.e. make the syllable heavy. Clusters with liquids are characteristically uncertain in this regard (which is incidentally the origin of the term ‘liquid’, i.e. ambiguous or uncertain). If we make the syllable heavy we get ceˈrebral, if light ˈcerebral.

You may recognize this Latin word as part of one of the standard examples of tmesis, in the famous half-line from the poet Ennius,
saxō cĕrĕ- commĭnŭit -brum
‘he shattered his br- -ain with a stone’.

25 comments:

  1. Sometimes I think doctors are trying to make a phonetic point of difference between the two human parts called cervix: saying sˈvaɪkl̩ cancer but ˈsɜːvɪkl̩ vertebrae. But then I hear one who reverses this. Doctors, I've found, enjoy being contrary in their pronunciations: my pet peeve is ʌmbɪlˈaɪkl̩.

    ReplyDelete
  2. 38 or so seasons of Law and Order votes for səˈriːbrəl (as in 'hemorrhage', see 'died of'). Originally it struck me as a wee bit strange, but I just assumed it was an American thing and then got used to it.

    ReplyDelete
  3. Clusters with liquids are characteristically uncertain in this regard
    Is there any cluster which consistently doesn't make the previous syllable heavy?
    one of the standard examples of tmesis
    Abso-f***ing-lutely awesome. So much for “the only reason you can't split infinitives in Latin is that they are one word”. :-)

    ReplyDelete
  4. bulbul: Ce'rebral hemorrhage, certainly, but 'cerebral palsy.

    army1987: The Romance languages have come down firmly on the side of muta cum liquida not being magic, which is why Latin TENEBRAS gives Spanish tinieblas, not *tiéneblas. English, however, still hasn't got with the program; the Tenebrae service of the Roman and Anglican churches is 'Tenebrae.

    ReplyDelete
  5. @ John Cowan: Ce'rebral for both here, but I think I'm a bit younger than you.

    ReplyDelete
  6. Presumably as knowledge of Latin and Greek dies out, and as more and more exceptions chip away at the classical rule, the Latin-influenced forms will lose their advantage and the situation will revert to a straight competition between two alternative pronunciations.

    Which alternative catches on will then depend on the usual factors: fitting into a wider pattern, avoidance of confusing homophones etc.

    Based on these criteria my guess is that the winners will be:
    - ˈskelɪtl̩ (the classical pronunciation)
    - ˈpælətl̩ (the common man's pronunciation)
    - ˈsɜːvɪkl̩ (the classical pronunciation)
    - ˈserəbrəl or sə'riːbrəl (could go either way)

    Let's meet up in 100 years and see if I was right.

    ReplyDelete
  7. In the meantime, let's go back to 1924 and see what Fowler writes s. v. False Quantities.

    ReplyDelete
  8. @Lipman: Quite. Apart from the issue of "false quantity", though səˈvaɪkl̩̩ is consistent with the long i in the Latin stem cervīc-, that property of Latin doesn't always cause English to have a long vowel in the corresponding place, as we see from "radical" (L. radix, radīc-).

    ReplyDelete
  9. The image is shocking.

    ReplyDelete
  10. John W,
    «In the case of cerebral, the relevant vowel of Latin cĕrĕbrum is short.»

    How do you know this? On the strength of the Ennius tmesis? Since you can make the syllable long or short metrically, he may have been able to please himself what length he made the vowel in the tmesis.

    You did Latin to degree level, and no doubt far beyond, so you may be privy to some thinking that never came my way, having only done it to Scholarship level, but the trouble with these liquid clusters is that all the dictionaries seem to mark the vowel long/short. (How do you do a combined macron and caron?) I have always supposed that might be because the vowel itself could be either, as perhaps in the Ennius, or because nobody knew. I do remember that the older primers used to have short vowels in various clusters in which in newer ones they had decided were long, marking them as in īnfāns crēscit.

    So whereas we can say with certainty that ˈskelɪtl̩ has no business becoming skɪˈliːtl̩ if we trace it back to Greek -λετ-, perhaps we cannot be so sure about sɪˈriːbrəl being inappropriate.

    When I have the facetal nerves of the facets of my ˈvɜːtɪbriː cauterized the wielder of the curling tongs calls them fəˈsiːtl and I persist in calling them ˈfæsɪtl. So why doesn’t he call them *vɜːˈtiːbrl? Or his computer program for visualizing them dɪˈdʒaɪtl?

    ReplyDelete
  11. John Cowan says "The Romance languages have come down firmly on the side of muta cum liquida not being magic, which is why Latin TENEBRAS gives Spanish tinieblas, not *tiéneblas." But in the case of cerebrum the Romance languages aren't consistent: Portuguese has cérebro with stress on the first syllable, while Spanish has cerebro with stress on the second syllable. However, it may be significant that this word has been borrowed from Scientific Latin, not inherited via Vulgar Latin.

    Mallamb: We know the second e of cerebrum is short because it's short in the diminutive cerebellum.

    ReplyDelete
  12. Tonio Green: Yes, I was referring to inherited words only.

    ReplyDelete
  13. Tonio,
    I don't see how cerebellum etc help, and neither it seems do the dictionaries: the conditions for the vowel length being in free variance, if indeed it was, no longer obtain in diminutives like that, which might therefore have conformed to different conventions. Or do you believe that the dictionaries really do mark the vowel length as being in free variance merely because the syllable weight can go either way? Even your own Iberian examples show that. Not that I think that the syllable weight was a reliable guide in anything but scansion! My feeling is that that as you say, the second e was indeed short, and that the syllable was too, so that the stress was on the first syllable, and I don’t find John C's appeal to "inherited" status convincing. What about pálpebra and Italian ˈpalpebra, for example?

    ReplyDelete
  14. mallamb:

    Do you have some reason to think palpebra is not a learned word in Portuguese and Italian? If so, it may well keep the Latin stress. Spanish has both learned pálpebra and inherited (or perhaps semi-learned) párpado; the latter keeps the Latin stress, but there is no muta cum liquida cluster any more either. Catalan parpella and Romanian pleoapă have penultimate stress, which also underlies French paupière.

    ReplyDelete
  15. The Romance languages have come down firmly on the side of muta cum liquida not being magic, which is why Latin TENEBRAS gives Spanish tinieblas, not *tiéneblas.

    I'm not sure this applies to Italian. Right now I can't think of any example of such a word which was obviously inherited from vulgar Latin (as opposes as possibly a later borrowing from classical Latin); in any event, the (e.g.) T of modern Italian scheletro definitely belongs to the onset of the third syllable, not to the coda of the second one.

    ReplyDelete
  16. John C,
    Do you have some reason to think palpebra is just a learned word in Portuguese and Italian? Are they not the ordinary everyday vernacular words? Or do you have a few more dicey candidates like párpado up your sleeve for those? I see that you correctly identified the Portuguese pálpebra from my all-purpose appellation 'Iberian', crediting me with the realization that "Spanish has both learned pálpebra and inherited (or perhaps semi-learned) párpado", but you will have to enlighten me about what you mean by 'semi-learned'. Catalan parpella is no grist to your mill for the same reason as cerebellum, and Romanian pleoapă is beyond me.

    So army1987,
    It doesn’t look as if it does apply to Italian, does it? I guess you would also say that the b of palpebra belongs to the onset of the third syllable, not to the coda of the second one, and the alternative interpretations of such clusters for syllabification purposes are presumably what is behind this whole carry-on.

    BTW, I wonder if the mysterious lengthening of the e of cresco in my Latin primers was the result of a flash of inspiration from comparative philology. It's a close e in It. crescere, after all.

    ReplyDelete
  17. mallamb: For the purpose of stress (outside poetry), a muta cum liquida in classical Latin always form the onset of a syllable, and so leave the preceding syllable light/short if it has a short vowel. But the poets had the choice to divide the muta and the liquida, so as to make the preceding syllable long by position. For example, Vergil writes

    "ossaque dispersit cerebro permixta cruento"

    with the middle syllable of cerebro short, even though he earlier had written

    "conlapsos artus atque arma cruenta cerebro"

    with the same syllable long.

    The fact that the syllable can be short in poetry proves that the vowel itself is indubitably short; what is variable is the way the syllables are divided. Dictionaries that have "cerē̆brum" with both macron and breve on the e are indeed only concerned about the syllable length (for the purpose of scanning), not the quantity of the vowel proper.

    ReplyDelete
  18. I guess you would also say that the b of palpebra belongs to the onset of the third syllable, not to the coda of the second one
    I deliberately avoid using /b/ in the examples because in most central and southern accents (excluding the Tuscan one on which the standard is based) it is always or almost always geminated after vowels (i.e. /'pal.peb.bra/), so the second syllable *is* closed for them (making such words among the very few ones with an unstressed closed penultimate syllable).

    As for cresco, dunno 'bout Latin, but in Italian, hyphenation rules treat it as cre-sco even though IMO /'kres.ko/ is much more realistic phonetically.

    ReplyDelete
  19. Johan,
    I sometimes despair of the impenetrability of my prose. I had already said everything you have just said except
    1) to mention the nice contrasting examples from Virgil
    2) The fact that the syllable can be short in poetry proves that the vowel itself is indubitably short (though I did say that what is variable is the way the syllables are divided)
    3) Dictionaries that have both macron and breve on the e are indeed only concerned about the syllable length (for the purpose of scanning), not the quantity of the vowel proper.

    I have said that 2 is my own supposition, but it hardly attains to the status of "proof", and doesn’t account for tinieblas etc.
    I have said that 3 (apart from the screw-up caused by mistaking ˇ 02C7 for ˘ 02D8 and thinking 'caron' was some newfangled name for 'breve') has always been my suspicion, but it seems such a bizarre and unnecessary practice that I would still like to see some evidence for it such as I gave for the long vowel in L crēscō, and the nearest I can get to that is that Lewis & Short distinguish the vowel with both macron and breve from the vowel with macron in the case of labrum 'lip' v lābrum 'basin'.

    ReplyDelete
  20. Army,
    Yes of course, I should have thought about the rampant raddoppiamento fonosintattico, but that is after all a funny place for a geminate.

    I agree that /'kres.ko/ is synchronically much more realistic phonetically, but when I said it's a close e in It. crescere I was talking about the close quality of the vowel, not implying that the syllable was closed.

    ReplyDelete
  21. Johan,

    « Dictionaries that have "cerē̆brum" with both macron and breve on the e are indeed only concerned about the syllable length (for the purpose of scanning), not the quantity of the vowel proper.»

    Again I'm sure you're right, but their rampant inconsistencies don’t inspire much confidence. Thus I do realize there are good etymological reasons for my above examples ‹lăbrum› and ‹lābrum›, to give them with a no-nonsense marking of the presumable length implications of that etymology, and for words like ‹muliěbris›, which L&S do mark with a breve, but what then of their putting only an etymologically implausible breve on ‹anclăbris›, from ‹anclo/anculo, -are›? They also mark ‹ephĕdra› and ‹synědrus› with the breve, but if those are because of Greek ἐφέδρα and σύνεδρος, why do they mark ‹cathedra› (καθέδρα) and ‹cedrus› (κέδρος) with both macron and breve? Sometimes they don't mark the ambiguity at all, as in ‹fellebris›, and sometimes they're just baffling: ‹hȳdra›, but ‹hydrus›, both water-serpents, and ‹tetrametrus› with macron and breve on the first e and only the breve on the second.

    ReplyDelete
  22. I suspect that when they mark a vowel with both macron and breve before muta+liquida, what they're trying to say is "This word is attested in poetry with both a heavy syllable and a light syllable here". That would account for the inconsistencies, because not all words are attested with both weights. I don't know how often Hydra is attested in poetry; maybe it just happens to only be attested in lines where the y has to scan as long, despite the fact that the vowel is for etymological reasons almost certainly short by nature.

    ReplyDelete
  23. Yes, Tonio, that's what it was beginning to look like to me too. Don’t you agree it's a bizarre practice? It's a simple metrical rule, and if when both forms are not attested they only give one of them, it's totally misleading, as with the unadorned macron on this hȳdra, marking it as long by nature and not by position, when as you say it's unlikely for etymological reasons that it was ever pronounced that way. The unadorned breve for lemmata in which for etymological reasons the relevant vowel is almost certainly short by nature is even more misleading, and implies some sort of robotic application of the rule to make a long vowel short for metrical purposes when no fancy manipulation of syllabification can possibly justify that.

    ReplyDelete
  24. Mallamb: A semi-learned word, or semicultismo in Spanish, is a borrowing from Latin that has picked up some features of inherited/popular words. For example, regla 'rule' is a borrowing of Latin REGULA, but with the usual loss of penult vowel in words with antepenultimate stress that's typical of Spanish; we can contrast it with the learned word (cultismo) regular 'regular'. The inherited descendant of REGULA is reja 'grid', which has not only lost its penultimate vowel, but has also undergone the regular change /gl/ > /x/.

    For some reason, the Spanish Wikipedia does not have an article on semicultismos, but there's a good short article at the Aragonese Biquipedia that gives some more examples: virgen/*verzen, angel/*anlo, reinar/*reñar, reino/*reño, siglo/*siejo, apóstol/abocho, obispo/besbo, milagro/mirajo, peligro/perijo, cabildo/cabejo. As you can see, they tend to be political or ecclesiastical.

    ReplyDelete
  25. I see that I said "The unadorned breve for lemmata in which for etymological reasons the relevant vowel is almost certainly short by nature is even more misleading" when I obviously meant "The unadorned breve for lemmata in which for etymological reasons the relevant vowel is almost certainly LONG by nature is even more misleading". Huh.

    ReplyDelete