Tuesday, 1 January 2013

on the edge

And a happy new year to all of you too!

It’s many months now since we last discussed the Latin stress rule and its consequences in English. Chomsky and Halle, in their seminal The Sound Pattern of English (1968) claimed that this rule is part of English phonology, i.e. part of what native speakers know about English. Nothing demonstrates the falsity of this claim so much as the persistent failure of educated native speakers to observe it in their pronunciation of Latin names and other words.

The rule is: a Latin word in which the penultimate vowel is short and not followed by a consonant cluster has its stress on the antepenultimate. (For a longer explanation, see my blog for 27 Oct 2010.) Native speakers are supposed to be able to have inferred this from such examples as deficit, stimulus, nebula, monitor, omnibus, which meet the structural formula and are stressed on the antepenultimate (here =first) syllable, as against such examples as aroma, gladiolus, dictator, hiatus (with a long penultimate vowel) and agenda, propaganda, protector, hibiscus, professor, bacillus, colossus (with a consonant cluster after the penultimate vowel; note that in English it is consonant letters that count, not consonant sounds) — which do not meet the structural formula and accordingly take penultimate stress. (The account in SPE is actually considerably more complicated. In English, nouns are subject, it is claimed, to a further rule that says we ignore the final syllable if weak and apply the Latin rule to the remaining stem; longer words of all word classes are then likely to be subject to other rules that shift the main stress elsewhere. We’ll ignore all that.)

No one ever explicitly taught me this rule when I studied Latin as a schoolboy, but I certainly absorbed it unconsciously, since I left school knowing where to stress any given Latin word appropriately (if I knew the quantity of the penultimate vowel). So to this extent Chomsky and Halle’s claim is perhaps justified.

Not so for Jeremy Paxman, as was evident from his hosting of Christmas University Challenge on BBC2 TV yesterday evening. Although he has a degree in English from Oxford, he doesn’t know where to stress Latin words. He asked the teams a question about the meaning of the Latin mottos engraved on the rims of our £1 coins.


We’ll forgive ˈtuːtəmen for what should be tuːˈtɑːmen , since the quantity of the middle vowel is not indicated in the spelling (though gravamen and stamen might have given him a hint). But the double ss in the spelling of lacessit indicates beyond doubt that it must be stressed ləˈkesɪt, not (as he had it) ˈlækesɪt.

What a good thing he didn’t have to attempt to pronounce the third of the £1 coin edge inscriptions — it’s in Welsh, about which I’m sure he knows even less than he does about Latin.


(That’s ˈpləjdjol ujv im ɡwlaːd, or anglicized ˈplaɪdiɒl ˈuːɪv ɪm ˈɡlɑːd.)


  1. '[...] the Latin stress rule and its consequences in English. Chomsky and Halle, [...] claimed that this rule is part of English phonology, i.e. part of what native speakers know about English. Nothing demonstrates the falsity of this claim so much as the persistent failure of educated native speakers to observe it in their pronunciation of Latin names and other words.'

    What astonishes me is not that 'educated native speakers [of English should fail] to observe [the rule] in their pronunciation of Latin names and other words', because I have heard too many examples of such a failure, but, much rather, that Chomsky and Halle should ever have thought that the rule was 'part of what native speakers know about English'. Have you, gentlemen, any (linguistic, psychological, whatever) explanation for this obvious error?

    It's certainly pedantic to add that according to the rule the penultimate still did not attract stress when it had a short vowel and a muta-cum-liquida (br, cl, etc.) consonant cluster, especially that I can't think of any example of this having repercussions for stress in English (can you?) --- perhaps apart from 'integer', should it be derived from 'integrum' (Acc.) or 'integro' (Abl.), rather than from 'integer' (Nom.), where, however, the second-but-last vowel [e] became long (or so say the Dictionaries...)

    Full name --- see Profile

    A happy and prosperous year to everyone

  2. Chomsky and Halle's excessive degree of education, leading to the assumption that their English is everyone's English, would I think be the explanation. It's well-known that Chomsky is a Cartesian rationalist who believes that introspection rather than empirical research is the royal road to psychological truth.

    There are a few obscure, rare, and/or obsolete English derivatives of Latin tenebrae that have initial stress and so illustrate the muta cum liquida rule: tenebrate 'darkened', tenebres 'darkness', tenebrize 'spend time in darkness'. Most common is Tenebrae itself: '[t]he name given to the office of matins and lauds of the following day, usually sung in the afternoon or evening of Wednesday, Thursday, and Friday in Holy Week, at which the candles lighted at the beginning of the service are extinguished one by one after each psalm, in memory of the darkness at the time of the crucifixion.' (OED1)

    1. Thank you John, that is convincing.

      Re 'tenebres': the Dictionaries say that the 'e' in 'ebr' was long in 'tenebrae', so it should be stressed on the penultimate in Classical Latin, yet in Italian too it is stressed on the antepenultimate, perhaps as a result of a hyper-correct application of the muta-cum-liquida rule, when vowel length began to disappear in vulgar Latin or proto-Italian.

      Full true name --- see Profile.

    2. It seems to have been short for Lucretius:

      quid dubitas quin omnis sit haec rationis potestas,
      omnis cum in tenebris praesertim vita laboret?
      nam vel uti pueri trepidant atque omnia caecis
      in tenebris metuunt, sic nos in luce timemus
      inter dum, nihilo quae sunt metuenda magis quam
      quae pueri in tenebris pavitant finguntque futura.
      hunc igitur terrorem animi tenebrasque necessest
      non radii solis neque lucida tela diei
      discutiant, sed naturae species ratioque.

      (De Rerum Natura, ii. 53 ff.)

  3. I too noticed Jeremy Paxman's mispronunciations: as University Challenge presenter he is persistently slapdash in his pronunciation not only of foreign words but even of uncommon English words. He is frankly unprepared for the contests: contrast the care taken by Bamber Gascoigne in "asking the questions".

    1. In the latest programme, he came up with ˌmiːtiəriˈɒlədʒi. Meteor rheology sounds like a rather unusual branch of science, which may not have much to say about anemometers.

      (And don't get me started on his, and the programme's, anti-science bias in general, lest this turn into an off-topic rant...)

  4. I would think that if it's true that "in English it is consonant letters that count, not consonant sounds", it is impossible to have it as part of the phonology, since phonology obviously doesn't care about letters (unless they reflect pronunciation).

  5. I think gladiolus is actually an exception: the Latin is ''glădĭŏlus''. Fowler recommended ɡlæˈdaɪələs or ɡlæˈdaɪələs/, and the OED still has ɡlæˈdaɪələs. I don't think I've ever heard anything but ɡlædiˈoʊləs, though.

    There are other exceptions, such as vertigo, which "should be" vərˈtaɪgoʊ. Of course, none of the exceptions in English really excuse getting the stress wrong when reading an actual Latin text.

  6. Generative Phonology, the guiding principle of Chomsky & Halle, consists of rules that resemble equations.

    • On the right hand side is an abstraction that stands in manifest equivalence to a pronunciation, a phonetic realisation. (Manifest, that is, if you have sufficient familiarity with the accent and idiosyncrasies of the speaker.)

    • In the middle is a specification of the phonological context in which that abstraction is 'generated'

    • On the left is a 'deeper' abstraction, the underlying form which may not stand in a predictable relation to the concrete pronunciation.

    Chomsky&Hallian rules for stress assignment to Latinate words can only operate if the left hand side abstraction — the 'input' — includes a feature which triggers the rules. People like John (and me) who have received a Classical education absorb the knowledge of that feature in many or most or even all of the lexical items that qualify for the feature 'Latin'.

    I say 'more or most or even all' because there's no predictability in what people acquire as knowledge. (I don't mean 'learn' as in 'commit to memory'; I mean 'absorb into intuitive knowledge'.) The underlying abstraction we acquire may depend on conflicting factors, in particular:

    1. the way we hear other speakers pronounce words which incorporate the stem
    2. the spelling of the words incorporating the stem — especially when those words are seldom heard

    (This explains, for example, why my instincts are similar to John's but less consistent.)

    So what Chomsky & Halle claimed is not really so surprising. English speakers who have studied Latin (or have been affected by teachers and elders who have studied Latin) have an instinctive sense of how to stress words of Latin type — even when they are not conscious of the fact.

    So far, so obvious. Where Chomsky & Halle go beyond the obvious is in saying that, for many words in common use, the stress rules have entered the phonological repertoire of English speakers in general, along with the intuitive knowledge as to which words are affected,

    It's not a strong claim. It's far too difficult to disprove by counter-example.

    Still, it could be an explanation as to how people arrive at stress pattern equal or related to Latin. For those who don't arrive at such a pronunciation the absence of the rules and intuitive knowledge would be an explanation.

    But this may be unnecessary. It may be the case that people primarily acquire the concrete phonetic pronunciations, and that the phonological abstraction is secondarily derived from them. It would be enlightening to compare what the same speakers do when confronted with entirely novel stems, for example stems which are not Latin but bear a strong resemblance.