UCL Division of Psychology & Language Sciences

John Wells’s phonetic blog

Archive, December 2006 – January 2007


new email address

To see the phonetic symbols in the text, please ensure that you have installed a Unicode font that includes all the IPA symbols, for example Charis SIL (free download).


Browsers: Some versions of Internet Explorer have bugs that prevent the proper display of certain phonetic symbols. I recommend Firefox (free) or, if you prefer, Opera (also free).


RSS feed for this site RSS feed for this site

John Wells

Wednesday 31 January 2007

The warning triangle

Thanks to Arnold Zwicky of Language Log for this cartoon. It illustrates the point that we consider it wrong to pronounce et cetera as ekcetera (i.e. with /ek's-/ instead of /et's-/) and nuclear as nucular (i.e. with /-kjVl-/ instead of /-kli-/).

Probably most of my readers, sharing my ignorance of American culture, were not able to understand the reference to Hannity and Colmes. Fortunately, these days we can call on Wikipedia to fill some of the gaps in our knowledge. It reveals that H&C is a political debate program shown on Fox TV and that some people consider it excessively right-wing. Presumably the (left-leaning) cartoonist wants to imply that the kind of people it chooses to allow air their views in debate are typically not very educated, as is revealed by their mispronunciations. Those who are sufficiently educated to know the right pronunciation have to dumb themselves down so as to fit themselves to appear.

Dictionaries perform various roles. The main users of pronunciation dictionaries such as LPD are EFL teachers or students who wish to know how to pronounce a word whose written form they perhaps know well but about whose pronunciation they are uncertain. A second group of users are native speakers who want to have their ignorance remedied or their prejudices confirmed or disconfirmed by a lexicographer they consider authoritative.

The lexicographer, though, (me, for instance), if he is trained in linguistics, will always be acutely conscious that there is no absolute standard of correctness in language, particularly in a language with such a wide speakership and inadequate orthography as English.

That is why I have always striven to make LPD play a further role, that of documenting the current actual state of the language. That is why I felt it essential to include such widespread though ‘incorrect’ pronunciations as those targeted by the cartoonist.

However, if I had merely listed these variants along with all the other variants that warrant inclusion, I would be failing in my duty, because I would be ignoring the attitude of careful, educated, literary-inclined, speech-conscious users towards them. That’s why I use a special warning mark: the exclamation mark enclosed in a triangle.

Not to include such variants would be to fail to document the language properly — a charge that I think could reasonably be levelled against certain other pronunciation dictionaries. To include them but without a distinguishing mark would be to fail to indicate the attitude of other speakers towards them.

Tuesday 30 January 2007


There was a programme on the ITV channel yesterday evening about the health problems of British people who retire and move to Spain. Several hundred thousand people have done this, leaving the UK to live in the sun. Very few of them speak much Spanish. If they have become resident in Spain and are not yet 65 years old, and fall ill, they discover to their horror and surprise that they are covered neither by the British National Health Service nor by its Spanish equivalent.

The British people who were interviewed had gone to live in the Murcia region, in the southeast of Spain. The programme provided ample evidence of how native speakers of English actually pronounce the name of the place. And what they say is either /ˈmɜːθiə/ or, less commonly, /ˈmɜːsiə/.

The Spanish pronunciation is /ˈmurθja/, or locally also /ˈmursja/. The /r/ can be either tapped or trilled.

And what do our British pronunciation dictionaries say on the matter? The word is not in LPD, so I’m in no position to crow. But it is in both the Cambridge EPD (17th edition) and in the new Oxford BBC Guide to Pronunciation.

EPD gives it as /ˈmɜːʃiə/, or “as if Spanish” /ˈmʊəθiə/. Right about the vowels, wrong about the medial consonant.

The OBGP gives just /ˈmʊəθiə/. (I’ve converted the book’s tiresome respelling and transcription into my usual system.) Right about the consonants, wrong about the stressed vowel.

We might well think that it would be preferable for English people to anglicize Spanish [ur] to /ʊə/. But they don’t, and it’s wrong to pretend that they do. Why don’t they? Because of the overwhelming influence of English spelling conventions, according to which ur is mapped onto /ɜː/.

Update on apical vowels

I didn’t read Kwan-hin Cheung’s letter carefully enough, and misrepresented his views (blog, 27 Jan.). Apologies. Here’s his letter setting things straight.

Mar Menor beach, Murcia

Monday 29 January 2007

Prefixes ending in -i

The next time I do a revision of LPD I am thinking of changing the entries for words with the unstressed prefixes be-, de-, pre-, re-.

At present we have entries like these:

becalm bɪ ˈkɑːm bə-, §biː-
predict prɪ ˈdɪkt prə-, §priː-
—which I am thinking of changing to:
becalm bi ˈkɑːm bə-
predict pri ˈdɪkt prə-

(You’ll recall that the LPD symbol § denotes ‘BrE non-RP’.)

You’ll also recall that for the last twenty years or so phoneticians have been using the symbol i for the weak close front vowel used at the end of words such as happy, coffee, donkey (‘the happY vowel’).

In conservative Daniel Jones-style RP this vowel is phonetically identical with the ɪ of KIT. But for many other speakers it is more like (or indeed identical with) the of FLEECE.

The environments in which I use i in the current LPD are

  • word-finally: happy, coffee, valley;
  • before a strong vowel: variation, ratio;
  • and before a weak vowel: glorious, convenient.
  • I also use it at the end of a combining form: multilateral, polytechnic
  • and in inflected forms in which d or z is added to a stem ending in i: carried, carries

So it is a very natural extension to use it for the prefixes be-, de-, pre-, re-. It saves a little space, too.

Just as studied was a homophone of studded for Daniel Jones (and is for me), or bandied and banded, so now descent was/is a homophone of dissent. But there are other speakers who make a difference in such pairs.

What do you think about the proposed change?

ships becalmed
Ships becalmed

Saturday 27 January 2007

More on shih tzus

I have received two comments from east Asia about yesterday’s posting.

Masaki Taniguchi writes from Kōchi in Japan, “I did not know shih tzu (Jp 'shii zuu ) comes from 獅子 (Jp 'shishi). Thank you! They are very popular in Japan, too.”

[Update: Tamikazu Date points out that “shi-shi as we use it in Japan is not a real lion, but an imaginary lion-like animal. So we never call lions we see in zoos shishi. They are simply called lai-on.”]

Kwan-hin Cheung of Hong Kong writes, “Your trancription of shizi on your blog today interests me a lot. As far as the dorsum part is concerned, I would prefer a back vowel for zi and a central vowel for shi.

In 1992 I published the following paper in Chinese: “北京話 ‘知’ ‘資’二韻國際音標寫法商榷” [IPA transcription of the so-called 'apical vowels' in Pekinese] , pp 37-46, Research on Chinese Linguistics in Hong Kong (ed. T Lee), HK: Linguistic Society of HK, 1992. The collection itself was published bilingually, with a majority of papers in English. My paper predates John Laver's discussion on apical vowels in Laver (1994:319-320).”

In his article Cheung mentions the symbols [ʅ] and [ɿ], which ‘linguists within Chinese communities ... often regard ... as the IPA symbols for the two sounds’, although in fact ‘not only are they not official IPA symbols but they are hardly internationally known’.

He would represent shi in IPA as [ʂ͡ɨ] and zi as [z͡ɯ]. The slur marks reflect the fact that friction may extend into the vowel.


Friday 26 January 2007

Shi(h) Tzu

A dreadful pun in Wednesday’s freesheet thelondonpaper:

I went to the zoo the other day, and all they had was a dog. It was a shih tzu.

Apologies for the implicit use of a word which LDOCE labels “adj BrE spoken not polite”.

This breed of dog can be spelt either shih tzu or shi tzu. Either way, it is an old (Wade-Giles) romanization of the Chinese word for ‘lion’, which in Hanyu Pinyin would be written shizi, or with tones shīzi. In Chinese characters I think it is 狮子 (simplified), 獅子 (traditional). The Chinese pronunciation is [ʂɯʳ d̥z̥ɨ], said on a high level pitch (the first syllable has tone 1, the second is toneless).

This word is in LPD (but not in EPD). I gave its English pronunciation there as ˌʃɪt ˈtsuː. On reflection, having heard dog fanciers talking about the breed, I think that the second syllable ought rather to be shown as ˈzuː or ˈsuː. (In English we can’t really manage /ts/ in syllable-initial position, so we transfer the plosive into the preceding syllable. Spelling tz normally = /ts/, as in quartz; but the presence of z in the spelling tends to lead people to use /z/ when the plosive and fricative are separated, as here.)

Given that, the pun reduces to ˌʃɪt ˈzuː vs. —what? Well, actually ˈʃɪt ˌzuː. What stops the pun being perfect is intonation: the fact that in the latent version zoo has already been mentioned, so that on the second mention it will be deaccented, throwing the nucleus onto shit. The dog is a shih tzu, but the rubbishy zoo is in this context a shit zoo.

shih tzu
A shih tzu

Thursday 25 January 2007

Casual-speech reduction

A colleague from another department at UCL was enquiring about the phonetic transcription of real-life colloquial speech. He started by asking me how a phonetician would transcribe There you go!, so I wrote down for him

ˈðeə ju ˈgəʊ

“OK,” he said, “but what if it was said like this?” — and he produced something like


“Has anyone attempted to produce systematic transcriptions of this type of speech?”

I referred him to various publications that attempt to do just that, or something like it — Linda Shockey’s Sound Patterns of Spoken English (Blackwell, 2002), and Gillian Brown’s Listening to Spoken English (Longman, 2nd ed. 1990). I also mentioned Jack Windsor Lewis’s EFL-oriented People Speaking (OUP 1977).

I also suggested that such reduced versions could (in principle) always be derived by rule from careful versions. We may not be sure of all the rules, but in this particular case there appears to be no difficulty in deriving the reduced version by rules of quite general applicability:

input ˈðeə ju ˈgəʊ
1. reduce you to its weakest form, → ˈðeə jə ˈgəʊ
2. drop the second element of diphthongs → ˈðe jə ˈgə
3. drop initial [ð] → ˈe jə ˈgə
4. drop all stresses except the nuclear accent → e jə ˈgə
5. drop phrase-internal schwa → e j ˈgə
= output

QED, or, as young people say these days, ta-dah!

Shockey book cover

Brown book cover

Wednesday 24 January 2007

The accenting of phrasal verbs

If you manoeuvre an English phrasal verb into a position where it bears the intonation nucleus, the nucleus will be placed either on the verb word or on the particle. Generally speaking, the nucleus goes (i) on the verb word if the particle is prepositional, but (ii) on the particle if it is adverbial. Usually, prepositional (= pseudo) phrasal verbs are lexically single-stressed, but adverbial (= true) phrasal verbs are lexically double-stressed.

(i), prepositional 'What are you 'looking at?
'Tell me what you’re 'looking for.
(ii), adverbial 'Don’t (')look 'back!
'Look a'way!
They 'all (')look 'up to him.

Accents shown in the examples as (') are likely to be downgraded for rhythmical reasons (‘Rule of Three’, English Intonation page 229).

There are a few phrasal verbs which rather rarely bear the nucleus. With a transitive verb, where there is an object following, the nucleus naturally tends to go on the object. In the resultant string Verb + Particle + Object, it can be difficult to tell whether the lack of an accent on the particle is because the particle is lexically unstressed (type (i)), or because it is rhythmically downgraded by the Rule of Three (type (ii)).

'Look up the 'answer.

You can usually resolve the uncertainty

  • by replacing the object NP with a pronoun (and extraposing the particle if appropriate),
  • by switching to a wh question,
  • or by switching to a passive construction.
This forces the nucleus onto the phrasal verb.
'Look it 'up.
'What did you (')look 'up?
The answer | was 'duly (')looked 'up.

Hence look up ‘find in a list’ is type (ii), adverbial.

For most phrasal verbs I have no hesitation with this classification. My native-speaker intuition, supported by observation, readily tells me whether they are type (i) or type (ii). However, there are cases where I find myself hesitating. One such is to look after. Applying the techniques mentioned resolves the uncertainty. (This is a verb in which the particle cannot be extraposed.)

What about Mary? | We must 'look 'after her.
'Who do you want me to (')look 'after?
She’s 'being (')looked 'after.

Hence look after, despite seeming to be prepositional, is actually double-stressed.

Michael Ashby thinks that these are fixed focus patterns rather than matters of lexical stress. For EFL purposes it comes to the same thing.

look back

Tuesday 23 January 2007

Anomalous g = /dʒ/

We all know about the spelling-to-sound oddity of

  • gaol /dʒeɪl/, alternatively (and more logically) spelt jail, and
  • margarine /ˌmɑːdʒəˈriːn/.

Their oddity consists in the fact that the letter g normally corresponds to /dʒ/ (rather than to /g/) only when followed by e, i, or y. Even that applies mostly to words of Romance rather than Germanic origin: compare gem and get. In gaol and margarine, though, we have g = /dʒ/ even when followed by something else.

Ted Carney, in his comprehensive Survey of English Spelling (Routledge, 1994), page 327, also mentions

  • veg /vedʒ/, the colloquial abbreviation of vegetable — compare the abbreviation of refrigerator, usually now spelt fridge; and
  • (rather strangely) syringa, which nevertheless for me and all other reference books I can lay my hands on has /g/, not /dʒ/.

But now I have a number of further candidates for this category — words that I have noticed native speakers pronouncing with /dʒ/ even though the spelling would suggest /g/.

  • meningococcal. Here it is clearly the influence of meningitis with regular /dʒ/ that leads people to say /meˌnɪndʒəˈkɒkl̩/. The word frequently occurs in the phrase meningococcal meningitis.
  • digoxin. Similarly, it must be the influence of digitalis and digit that leads people to call this heart drug /dɪˈdʒɒksɪn/. (That’s what everyone at my cardiac support group calls it.)
  • purgative, which I heard as /ˈpɜːdʒətɪv/ rather than the usual /ˈpɜːgətɪv/ on the radio the other day. It must be the influence of purge;
  • analogous, which I have quite often heard pronounced with /dʒ/, no doubt because of analogy with analogy.

What we have in each case here is morphological regularization. The morpheme in question (mening-, dig-, purg-, analog-) keeps the same phonetic shape (with /dʒ/) despite suffixation. Chomsky and Halle’s rule of Velar Softening clearly does not work quite as regularly as they would have us believe.



Carney's book

Monday 22 January 2007


As you see above, UCL Phonetics and Linguistics is discontinuing its email server facilities at the end of this month, so all of us with or addresses have to change to addresses on the central UCL server, In my case this actually leads to a rather simpler address.

UCL logo

I have also taken the opportunity to change my email client from the Eudora that I have been using till now. I had become very fond of Eudora, which presents a much more friendly and intuitive interface than Microsoft’s offerings (Outlook, Outlook Express). But a major drawback of Eudora is that it is not Unicode-compliant, and for a phonetician in today’s world that just will not do. So now I have installed Thunderbird and started to use it. It has a built-in trainable spam detector and anti-phishing protection, and I find it very flexible and easily customized. It comes from the same stable as the Firefox browser, and is free. And it is Unicode-compliant, allowing you to input, display and read any Unicode characters for which you have a font — including all the IPA symbols.

Thunderbird comes in 32 different language varieties, and for Windows, MacOS X and Linux.

Eudora logo

Thunderbird logo

Talking of Unicode, a large parcel awaiting me on my return last week turned out to contain the printed version of The Unicode Standard 5.0. This is the sort of book I can spend many happy geekish hours with (as I have already done with its predecessors, versions 1, 2, 3 and 4).

As Geoff Pullum puts it in his review in Language Log,

This is the most spectacularly nerdy book I have ever seen. All the details about how all the writing systems in all the world are to be encoded in a standard way for computer systems.

Among the phonetic symbols, the one for the labial flap was too late to be included. What we do find for the first time are the old IPA symbols for consonants with an attached palatalization hook and for vowels with a retroflex hook (for which the IPA now recommends free-standing diacritics).

There are several intriguing new scripts provided for: Balinese, N’Ko, Phags-pa, Phoenician, and Sumero-Akkadian Cuneiform.

And if you want an explanation of line breaks in Tibetan, Khmer, or Ugaritic, this is the place to look.

Modesty nearly prevents me from mentioning that among the messages of praise printed at the beginning of the book is one from me:

Unicode is marvellous. It makes it possible for phoneticians throughout the world to use all manner of phonetic symbols in their work and display them on computer screens in the certainty that they will not now be garbled or turned into wingdings (as once used to happen all too often). All alphabetic phonetic symbols officially recognized by the International Phonetic Association are now included in the Unicode Standard.


Saturday 20 January 2007

Latin stress in English

The stress rules of Latin still have a kind of ghostly presence in modern English. A Latin noun or adjective, if it has three or more syllables, is stressed on the antepenultimate if the penultimate vowel is short and followed by at most one consonant ('facĭlĭs); otherwise it is stressed on the penultimate (impe'rātor, tre'mendus). We have taken this principle into English and not only apply it to words derived from Latin, but extend it to words from Greek and sometimes other languages. So we have antepenultimate stress in 'deficit, 'miracle, 'stamina, 'memory, 'Jupiter (Latin) and a'nemone, ge'ography, a'natomy (Greek). All of these meet the criterion of having a short penultimate vowel followed by a single consonant sound. But we get penultimate stress in cre'ator, Oc'tober, bron'chitis, a'roma, with their long penultimate vowels, and equally in tre'mendous, De'cember, a'orta, with more than one (historical) consonant between the penultimate and final vowels.

That is the etymological reason that we stress po'lygamous (Greek -γαμ- -gam-, with short a) differently from Poly'phemus (-φημ- -phēm-, with long e) and poly'morphous (-μορφ- -morph-, with two consonants after the vowel).

A nice pair is Di'ogĕnes vs. Dio'mēdes.

Faced with an unfamiliar word ending -VCV(C), we crucially need to know whether the penultimate V is long or short. If it is long, it will be stressed; if it is short, it won’t be, and indeed will undergo vowel reduction.

Everybody in higher education pronounces diploma with penultimate stress, because the o is long, /dɪˈpləʊmə/. (It comes from Latin diplōma, Greek δίπλωμα.) But I knew a member of the general public who called it a /ˈdɪpləmə/, perhaps under the influence of diplomat /ˈdɪpləmæt/ (which originated as a French back-formation from diplomatique, and does not follow the Latin stress rule).

There’s a kind of dinosaur called a diplodocus. I’ve always called it a /dɪˈplɒdəkəs/, which may or may not reflect the fact that I know that its Greek etymon consists of διπλό- dipló- ‘double’ plus δοκός dokós (short o) ‘beam’. I was initially taken aback to find that my colleague Jill House called it a /ˌdɪpləˈdəʊkəs/. But she’s not the only one. Our ordinary spelling, of course, doesn’t indicate vowel length in -VCV(C) words. It’s one of the biggest shortcomings of our spelling system.

And then there is the word diaspora. It has an etymologically short penultimate o (Greek διασπορά diasporá, ‘sowing around, scattering’), and a corresponding traditional English pronunciation /daɪˈæspərə/. But I recently heard someone pronounce it /ˌdaɪəˈspɔːrə/. The spelling doesn’t tell you whether the o is long or short: and that’s the factor that determines the stressing. Because of the Latin stress rule.

Acknowledgement: this is all based on Chomsky and Hallé’s SPE, of course.


Friday 19 January 2007


Apropos of wronger, Kwan-hin Cheung writes to remind me that the matter was discussed in Language Log just under three years ago. On that occasion Geoff Pullum bemoaned the lack of evidence about how this comparative adjective is pronounced; but his intuition (like mine) was that it is regular, i.e. pronounced with no velar plosive. Kwan-hin says that as a non-native but very fluent speaker he has the same intuition: and he finds further support in the on-line Merriam-Webster, which even has a sound file of the word, duly pronounced /ˈrɑŋɚ/.

Apart from its questionable gradability, another reason for the rareness of this form of the adjective is that we can always avoid it by saying more wrong. A Google search showed 207k hits for wronger (which no doubt included some for the agent noun wronger ‘one who wrongs’), but 980k — over four times as many — for more wrong (which no doubt includes phrases such as a few more wrong numbers).

Volcano. T. Lalibus writes to say that many Italians use [æ] for words that should properly have [ɑː]. (Despite this claim, though, I can’t say that I’ve heard Italians pronouncing car park as [ˈkæ pæk] and the like. I don’t think it can extend to words spelt with r.)

T. Lalibus continues, “I have a lady friend whose L1 is Italian. She is posh, mature, educated, cultured, intelligent, quite refined, self-aware and quite ready to criticise other people's mistakes. But when she speaks English - although her syntax is flawless - she has /ɒ/ (the LOT vowel) in London and money. These are words she must have heard pronounced correctly with the STRUT vowel a million times! How can she make that mistake? Other than some form of crystallisation, why do you think it could be?”

Why indeed? I blame our crazy English spelling.

Kwan-hin Cheung
Kwan-hin Cheung

Thursday 18 January 2007


While I was away I had to rely for television news on either CNN or BBC World. One item on the latter attracted my attention. Interviewing the Scottish Nationalist Alex Salmond on the show Hardtalk on 14 December, Stephen Sackur said:

You’re placing undue reliance on a price of oil which is already wrong and which may well end up wronger.

He pronounced the last word /ˈrɒŋə/.

This was a rare sighting in the wild of this comparative. Wrong is the only monosyllabic adjective ending in [ŋ] other than long, strong, young. The comparative and superlative forms of these latter three are apparently irregular inasmuch as they are pronounced with [-g-]: [ˈlɒŋgə, ˈjʌŋgɪst] etc. Is this because of some minor rule concerning the pronunciation of comparatives and superlatives of all adjectives ending in [ŋ], or because long, strong, young are lexical exceptions? Wrong provides a possible test case to answer this question. If it too has [-g-], then there seems to be some general rule in operation. But if it doesn’t, the other three are exceptions.

The difficulty is that wronger is rarely encountered, because wrong is not a readily gradable adjective: things are either right or wrong, rather than having degrees of wrongness. But as the quote shows, this is not always the case. Stephen Sackur’s pronunciation of wronger, which is also my own, shows that — for him and me at least — wrong follows the ordinary rule for comparatives, which means that long, strong, young must be exceptions.

BBC World logo

Wednesday 17 January 2007

Compression in hymnody (ii)

Another hymn in which glorious is to be sung compressed (= as two syllables) is

Glorious things of thee are spoken, ('tum-ti 'tum-ti...)
 Zion, city of our God;
He, whose word cannot be broken,
 Formed thee for his own abode...

Compression can also operate across a word boundary. The relevant vowels, weak [i] and [u], arise from the weakening of function words such as me, we, she, the and you, to. Compression means that [i, u] lose syllabicity, becoming [j, w] (or if you prefer [i̯, u̯]).

Both within a word and across a word boundary, compression can operate only if the following vowel is weak — one of [i, u, ə, ɪ]. Thus within a word we can have compression in victorious /vɪkˈtɔːri‿əs/ → [vɪkˈtɔːrjəs], where the following vowel is the weak schwa, but not for example in Victoriana /vɪkˌtɔːriˈɑːnə/, where it is the strong /ɑː/.

Mount Zion
Mount Zion

Eighteenth-century hymnodists follow this principle pretty faithfully. Look for example at the last verse of Charles Wesley’s And can it be. The hymn is in meter. Each verse is an iambic tetrameter: it consists of four feet, the first syllable of each foot being unstressed and the second stressed (except that the first foot may be trochaic, stressed-unstressed, instead of iambic — producing an ‘initial choriamb’):

No condemnation now I dread
 Jesus, and all in him, is mine!
Alive in him, my living head,
 And clothed in righteousness divine,
Bold I approach the eternal throne,
 And claim the crown, through Christ, my own. 

In the penultimate line the phrase the eternal throne is to be scanned as four syllables, [ðjɪˈtɜːnəl ˈθrəʊn], as the undergoes compression before the weak initial vowel of eternal.

The same author’s Let earth and heaven combine offers further examples. It is in meter, so has four iambic trimeters followed by two iambic tetrameters.

 Let heaven and earth combine,
  Angels and men agree,
 To praise in songs divine
  The incarnate deity.
Our God contracted to a span,
Incomprehensibly made man.

 He laid his glory by,
  He wrapped him in our clay;
 Unmarked by human eye,
  The latent Godhead lay;
Infant of days he here became,
And bore the mild Immanuel’s name.

 He deigns in flesh to appear,
  Widest extremes to join;
 To bring our vileness near,
  And make us all divine;
And we the life of God shall know,
For God is manifest below.

Here the first verse has the incarnate deity as six syllables, with the undergoing compression before the weak initial vowel of incarnate. The second verse has Immanuel’s as a trisyllable [ɪˈmænjwəlz], with the [u] compressed to [w] before the schwa (compare genuine optionally pronounced as two syllables, [ˈdʒenjwɪn]). The last verse quoted has to appear as two syllables, with to (prevocalic weak form [tu]) undergoing compression to [tw] before the initial weak schwa of appear.

Wesley would never (for example) have scanned the only or to answer as disyllables. Compression is allowed only before an unstressed vowel. Normally, this vowel must also be weak, as set out above. Exceptionally, though, the hymnodists allow compression before an unstressed but strong vowel, as in the Christmas hymn Hark! the herald angels sing:

Joyful, all ye nations rise,
Join the triumph of the skies;
With the angelic host proclaim,
Christ is born in Bethlehem.

Or perhaps we are intended to sing this with elision of the vowel of the article, th’angelic. But if that were the case I think Wesley would have spelt it in that way.

Charles Wesley
Charles Wesley

Tuesday 16 January 2007

Compression in hymnody (i)

In Montserrat I became a temporary member of a local church choir. This led me to muse about the phonetics of hymns. One thing I noticed in them (and have noticed before) is the way hymnodists exploit compression.

By ‘compression’, I refer to variability in the number of syllables in a word or phrase. A compressed version of a word or phrase has one syllable fewer than the uncompressed version. In LPD I show sites of possible compression by using the special mark [].

Many words have two pronunciations, differing only in a segment that in one version is syllabic, e.g. [i, u, n̩, l̩] and in the other non-syllabic, e.g. [j, w, n, l]. This variability can be exploited by versifiers, who can require a particular sequence to be pronounced either uncompressed or compressed, to fit the metrical requirements of the poem, hymn or song in which it appears.

Thus the word victorious is pronounced uncompressed, as four syllables, in the dactylic rhythm of the British national anthem, and glorious likewise as three syllables:

Send her victorious ('tum-ti ti-'tum-ti-ti)
Happy and glorious
Long to reign over us
God save the Queen!

But they are to be sung compressed in a well-known hymn by Charles Wesley:

Ye servants of God, (ti-'tum-ti ti-'tum)
 Your master proclaim,
And publish abroad
 His wonderful name;
The name all-victorious (ti-'tum-ti ti-'tum-ti)
 Of Jesus extol;
His kingdom is glorious
 And rules over all.

More on this tomorrow.

Monday 15 January 2007


I’m back in London now, so it’s time to awaken this blog from its hibernation. Happy New Year, everyone!

As you know, I have been in Montserrat, an island in the Caribbean, parts of which are threatened by an active volcano. (Contrary to what some people would have you believe, life still goes on in the ‘safe zone’ in the north of the island, where some five thousand people continue to live happily and safely.) You can see a page of my holiday snaps.

The Montserrat Volcano Observatory staff make regular reports on the current volcanic situation, both written and spoken, the latter on the local radio station ZJB. Not all the volcanologists and other experts are native speakers of English.

One specialist who gave a spoken radio report recently is a Dr D., whose first language is Italian. Listening to him, I was struck by the fact that each time he used the word volcano — a frequent and crucial word in his report, as you might imagine — he pronounced it volc[æ]no.

You would think that as a volcanologist he would have noticed by now that all his English-speaking colleagues, not to mention all the other half-billion or whatever native English speakers in the world, pronounce it volc[]no. He must have made a wrong inference many years ago about the pronunciation of this word, on the basis of the spelling, and somehow become deaf to all the spoken evidence around him showing that he was wrong. (The error must be spelling-based, attributable to the uncertainty over whether the letter a represents [] or [æ]. It is not a matter of negative transfer from his native Italian, which would give volc[ɑː]no.)

Does this sort of thing matter? Well, yes, it does. The effect of his mispronunciation, on me at any rate, was to make me discount the value of what he had to say. If he doesn’t register the abundant evidence about the pronunciation of this everyday word, why should we suppose that he pays proper attention to the evidence on which he bases his scientific findings?

“To seek to change someone’s pronunciation — whether of the L1 or of an L2 — is to tamper with their self-image, and is thus unethical — morally wrong.” (Porter D. and S. Garvin, 1989, Attitudes to pronunciation in EFL, Speak Out! 5:8-15).

What rubbish! If some kind person were to teach Dr D. the correct pronunciation of volcano, they would be doing him a positive and useful service that would enhance his scientific credibility and thus his professional standing.

risk map of Montserrat
click to enlarge
Monday 4 December 2006


I haven’t after all succeeded in establishing an internet connection where we are staying in Montserrat. (I have uploaded this posting at an internet café.)

So I think the best thing is to have a six-week rest from blogging and to put this blog into hibernation. I hope to wake it up again in mid-January.

Meanwhile, a happy Christmas and New Year (or whatever your own cultural equivalents are) to all my readers.

Alongside you see the view from my balcony. In the distance you can just make out the outline of Nevis. That’s /ˈniːvɪs/, remember.

Nevis, seen from Montserrat

Current blog

Archived from previous months:

my home page