UCL Division of Psychology & Language Sciences

John Wells’s phonetic blog:
archive 1-15 October 2007


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

Browsers: I recommend Firefox (free) or, if you prefer, Opera (also free).

RSS feed for this site RSS feed

John Wells

Monday 15 October 2007

What do you mean, continuous?

Rather than write a proper blog entry for today I’d just like to refer you to an entry in Language Log.

I drew Geoff Pullum’s attention to a crass instance of faulty logic and faulty grammatical analysis in the Guardian, written by someone who ought to know better: no less than the editor of The Oxford Guide to Effective Speaking and Writing.

Geoff ran with the ball I passed to him. Read what he had to say by clicking on this link.

Saturday 13 April 2007

And can it be?

Yesterday I went to a funeral. The deceased was the father of one of the members of the choir I belong to.

The great thing about West Indian funerals is that you get a congregation of hundreds — many more than the miserable thirty or forty (if you’re lucky) who attend a typical British funeral. Here we had at least three hundred, which makes for good rousing hymn singing.

One of the hymns we sang was Charles Wesley’s And can it be. (Click on the link and you can read the words and hear the music.)

He left His Fatherís throne above
So free, so infinite His grace—
Emptied Himself of all but love,
And bled for Adamís helpless race.

Although in everyday life we pronounce /ˈɪnfɪnɪt/ (with some flexibility about the unstressed vowels), in singing this hymn we always pronounce it /ˈɪnfaɪnaɪt/. This may well have been Wesley’s own pronunciation, but why we still use it today I’m not really sure. The modern pronunciation would fit the scansion just as well.

Given finite (/ˈfaɪnaɪt/), the modern pronunciation of its antonym is actually somewhat mysterious.

The fourth verse of this hymn never fails to move me.

Long my imprisoned spirit lay,
Fast bound in sin and natureís night;
Thine eye diffused a quickening ray—
I woke, the dungeon flamed with light;
My chains fell off, my heart was free,
I rose, went forth, and followed Thee.
My chains fell off, my heart was free,
I rose, went forth, and followed Thee.

It brings tears to my eyes, even when (as at yesterday’s funeral) I hardly knew the deceased. Perhaps I am weeping for my lost faith.

Charles Wesley
Charles Wesley (1707-1788)

Friday 12 October 2007

More cases of phantom r

Yesterday’s posting about words without etymological /r/ but spelt with r evoked a number of comments.

The Scots, Irish, etc., who have been in touch all say they use the spelling pronunciation, i.e. pronounce /r/ in words such as Myanmar, shar pei, sarnie. (But one Scotswoman says “sarnie I hear with an English accent so would avoid, tho I know it well”.) This is the answer I would have expected, given the great influence exerted by spelling on the pronunciation of words we’re not sure about.

It also calls to mind the fact that people in the rhotic west of England tend to pronounce /r/ in words such as khaki, even though there is no r in the spelling. The equivalence “non-rhotic [ɑː] = rhotic [ɑːr]” is a deeply buried part of their phonological knowledge.

By the way, if you don’t know what a west-of-England working-class rhotic accent sounds like, here’s a YouTube clip I came across of a young man from Filton in Bristol. Put it in your archive of English local accents.

Nigel Greenwood comments

The Chinese word for tea (chŠ in Pinyin) happens to be "char" in the Gwoyeu Romatzyh romanization, which uses -r as a rising (2nd) tone marker. This use of silent -r could almost have been invented by a non-rhotic speaker of BrE.
British trade-names such as Polyfilla and the inner-city buses called Hoppa must be either baffling or irritating to rhotic speakers (not to mention non-anglophones).
Of course the /r/ reappears when someone mutters "I'll just put some Polyfilla(r) in it!"

Harry Campbell says

This use of R to indicate a "long" a vowel is very interesting. It's one thing in colloquial terms like scarper but I had never thought of it as normal procedure in standard [= non-rhotic, JCW] English, which after all does have some residual memory of the "silent" non-prevocalic R; surely the normal thing is "ah" rather than "ar". I wonder why they didn't go for Myanmah, conveniently recalling words like mynah, rajah, and indeed the dated spelling Burmah used by the famous Oil Company?

He also mentions two other relevant spelling/pronunciation anomalies:

The recent jocular coinage "lurve", "an emphatic, humorous, or arch pronunciation of LOVE ... sometimes specifically parodying the slow, smooth, crooning pronunciation of love in romantic popular songs" as OED puts it, has now entered the language, but to pronounce the R makes nonsense of the idea.

[Elvis Presley’s STRUT vowel sounds virtually identical to the RP etc. NURSE vowel. I mentioned this 25 years ago in Accents of English, p. 536, though I wrote lerve rather than lurve.]

Harry continues

Under marm, OED comments: Variant of MA'AM n.1 In U.S. usage, it is not always clear whether the spelling with -r- is merely a graphic device to indicate lengthening of the vowel (esp. in representations of the non-rhotic dialects of New England), or else represents a genuine intrusive /r/: rhotic /mɑrm/ can easily develop from /mɑm/ by analogy with e.g. rhotic /hɑrm/ HARM n. corresponding to non-rhotic /hɑm/. Cf. mars, marse s.v. MAS' n.1

(This must be the third edition of the OED. I can find no such discussion in the second edition, which is the one I have on my computer.)

I have commented elsewhere that the Caribbean English Creole progressive negative particle /naː/ — which, not being standard grammar, has no standard spelling — is spelt nah by the semi-rhotic Jamaicans but nar by the firmly non-rhotic Montserratians.

Lastly, Tristan McLeay mentions the Australian arvo ‘afternoon’. He says “It's doubly fun because not only does it use ar=/a:/, but also the Australian and American pronunciations of the first syllable are obviously quite different.”

A cup of char and a sarnie
Photo: Jupiter Images


Nigel Greenwood


The singing group Muldoon’s Picnic.
Harry Campbell is the one in the middle with a cap.

Thursday 11 October 2007

Myanmar, shar peis and sarnies

As we all know, the ruling junta in Burma would prefer that we call their country Myanmar.

In Burmese, this name Myanmar is essentially just a variant of the name Burma. It is transliterated as Myan-ma or Mran-ma, and in the local language pronounced something like [ma(n) ma], as against [ba ma] for the traditional name.

According to Wikipedia,

within the Burmese language, Myanma is the written, literary name of the country, while Bama ... (from which “Burma” derives) is the oral, colloquial name. In spoken Burmese, the distinction is less clear than the English transliteration suggests.

So the situation is comparable to the Greek colonels who insisted on literary (katharevusa) Ellas rather than popular (dhimotiki) Elladha. Right-wing politics becomes associated with literary variants, left-wing with popular.

What interests me now, however, is the question of how Americans and other rhotic speakers are supposed to pronounce this name. In both Myanmar and Burma the English spellings assume a non-rhotic variety of English, in which the letter r before a consonant or finally serves merely to indicate a long vowel: [ˈmjænmɑː, ˈbɜːmə].

So any American who says the last syllable of Myanmar as [mɑːr] or pronounces Burma as [bɝːmə] is using a spelling pronunciation based on British, non-rhotic, spelling conventions.

The Burmese (Myanmar) flag


In compiling LPD I faced a somewhat similar problem with the words scarper and sarnie. Americans don’t generally know or use these words, but how would they pronounce them if they did? The first is believed to be derived from Italian scappa! or by rhyming slang from Scapa (Flow) (= go); the second obviously comes from sandwich. In neither case is there an /r/ in the source. Would rhotic speakers use an /r/ purely on the basis of the spelling?

For that matter, what do the Scots, the Irish and other rhotic speakers on this side of the Atlantic do in these words? (Please tell me, but not all at once.)

Another example is the breed of dog called shar pei. This name is Chinese, shā pí (or the Cantonese equivalent): again, no etymological /r/.

Have a cup of char! (Chinese chá)

A shar pei

Wednesday 10 October 2007

Pogonophoria among phoneticians

John Maidment (yesterday’s blog) complained that fewer academics have beards these days. Paul Carter leaps in to point out that where he works — the School of Linguistics and English Language at Bangor University, or rather Ysgol Ieithyddiaeth ac Iaith Saesneg Prifysgol Bangor — they are doing their bit for facial hair.

Five out of the six males listed on our staff page are bearded, though I must confess that my phonetic beard (ie trimmed around the mouth so students can still get a good view of my lips and can therefore hear with their eyes as well as their ears) pales into insignificance when compared with Dirk Bury's syntactician's beard.


Left to right: David Crystal, Dirk Bury, Frank Gooding, and Eddie Williams.

The Welsh for ‘beard’ is barf [barv], borrowed from the Latin barba centuries ago, during the Roman occupation of Britain. There’s a fascinating little book (for those who can read Welsh and know Latin) called Yr elfen Ladin yn yr iaith Gymraeg (The Latin element in the Welsh language). It’s by Henry Lewis and was published by the U. of Wales Press in 1943. Paragraph 71 (slightly edited) reads

Pan ddêl b rhwng r a llafariad, ceir y cyfnewidiad b > f. (When b comes between r and a vowel, we get the change b > f.)

Bearing in mind that Welsh f = [v], we can see that this is the same change from [b] to [v] that we noted in the development of Modern Greek from Classical Greek (blog, 18 September).

The Classical Greek for ‘beard’ (see today’s headline) is πώγων pōgōn.

P.S. David Crystal, whose own blog is here, responds

My beard has in fact reached a higher plane of cosmic consciousness. I learned a few weeks ago that there is a David Crystal Appreciation Society on Facebook in which, from time to time, student faces appear with a DC-like beard superimposed. At least they haven't started Elvis-lookalike competitions yet.

Paul Carter

Tuesday 9 October 2007

John Maidment

My UCL departmental colleague John Maidment’s leaving do was held yesterday evening.

Here you see him receiving a leaving present from Michael Ashby (left), while the departmental administrator Molly Bennett is in the background.

There were written tributes from former students and colleagues, particularly from Japan and from Spain. The trio of Spaniards who achieved their PhDs under John’s guidance (Maria Lecumberri, Mercedes Cabrera, and Eva Estebas) could not be present, but sent witty greetings.

In his own speech, John bemoaned the decline in the prevalence of facial hair in academia, and pointed out that after his retirement there would be only one John left in the department, where we used to have half a dozen or more (me among them, of course).

Internationally, he is best known for his book jointly with Michael Ashby, Introducing Phonetic Science, and for his English Transcription Course written jointly with Maria Lecumberri.

His presence on the web remains, notably in the Speech Internet Dictionary and the English Pronunciation Tip of the Day site.

He was the person I turned to first if I wanted to know something about Chinese or Irish.

And I’m delighted to be able to tell you that although he is retiring to Cornwall he will continue to organize the biennial Phonetics Teaching and Learning Conference.

Monday 8 October 2007


I drove to Lancashire and back at the weekend to attend a family function.

I came back on the M1. It is quite a few years since I last drove down this motorway. One change that I noticed was that the service area that I know as Rothersthorpe Services has been renamed Northampton Services.

(According to Wikipedia, this is ‘so that travellers will have a clearer idea of where they are’ — Northampton being a substantial county town while Rothersthorpe is a tiny village.)

This got me idly thinking: how is it that we know immediately how to interpret these three th spellings? Obviously the pronunciations are /ˈrɒðəzθɔːp/ and /nɔːˈθæmptən/ (or minor variants thereof). So /ð, θ, θ/: but given that the spelling th is used indifferently for both sounds, how do we know which one to use where in an unfamiliar placename?

The spelling-to-sound rules given in LPD can be summed up as saying that word-initial th spells /θ/ except in function words such as articles and determiners. Word-medial th is /ð/ in Germanic words but /θ/ in Greek or Latin words. In the present discussion we can ignore word-final position.

At first it looks as if getting these names right is crucially dependent on recognizing morpheme boundaries. We have to be able to see that Rothersthorpe — an obviously Germanic name — consists of Rothers plus thorpe. The initial th in -thorpe therefore counts as being initial and hence voiceless. Compare Scunthorpe and indeed the surname Thorpe. The medial th in Rothers- follows the rule by being voiced. This is no doubt reinforced by the well-known place name Rotherham.

Northampton is slightly trickier. Although we say /θ/ in north, we say /ð/ in northern, northerly, and place names such as Northall, Northam, Northenden, Northiam, i.e. when th is medial — yet /θ/ in Northumberland. The distinguishing feature of Northampton and Northumberland seems to be that the vowel after the th is strong, thereby attracting the dental fricative into (syllable)-initial position and ensuring that it is voiceless. Perhaps that applies to Rothersthorpe, too, where the strong vowel attracts the second dental fricative into the last syllable. So it is not the morphology we depend on, but the syllabification.

Rothersthorpe church

Friday 5 October 2007


Campbell is a pretty common surname throughout the English-speaking world.

It has always puzzled me why there is a mismatch between its pronunciation and its spelling. We pronounce it ˈkæmbl̩, yet the spelling has an irrelevant (‘silent’) p before the b. Why?

The origin of the name Campbell lies in a Scottish clan name. The Campbells occupied an extensive area of Kintyre and the western Highlands. According to the Oxford Names Companion (OUP 2002), which incorporates A Dictionary of Surnames (1988) by Patrick Hanks and Flavia Hodges, its likely etymology is Scottish Gaelic, in which cam means ‘crooked’ and beul ‘mouth’. So Cambeul or Caimbeul was a nickname, ‘wrymouth’, apparently borne by the clan founder Gillespie Ó Duibhne, who lived at the beginning of the thirteenth century.

According to Wikipedia, there is a separate surname Campbell, of Irish origin, from the Irish Mac Cathmhaoil meaning ‘son of the battle chieftain’ (though personally I’d have thought that an Irish name of that spelling would yield the English form McCall).

But what about the spelling? As so often, it turns out that Norman scribes and/or lawyers writing in Latin were to blame. They came up with an elegant Latin equivalent of the Gaelic name, turning it into de Campo Bello ‘of the fair field’.

Just as the b in debt and doubt is down to Latin etymologizing, so is the p in Campbell.

Clan Campbell crest
(‘Lest you forget’)

Thursday 4 October 2007

London diphthongs

Yesterday afternoon I received a phone call from the BBC R4 programme You and Yours asking whether I would be willing to be a studio guest for today’s edition. The topic they wanted to discuss was English dialects, and whether they are disappearing.

I have often been interviewed on this sort of topic before, though I do sometimes feel that what they really want is a sociolinguist or a social psychologist rather than a descriptive phonetician like me.

So on this occasion I suggested that they turn instead to Sue Fox of Queen Mary University of London, who has been doing interesting sociolinguistic work in the East End of London (blog, 16 Nov. 2006). She interviewed speakers (i) in Hackney, which is a traditional ‘Cockney’ area, and (ii) in Havering, an outer-London borough halfway to Southend.

Put crudely, what she has found is that traditional Cockney, with its shifted diphthongs, has moved out to places like Havering; the inner-city areas such as Hackney have reversed the diphthong shift and reverted to much more standard-sounding FACE and GOAT vowels, presumably under the influence of the Bangladeshis and other incomers who now make up a fair proportion of the local population.

The diagrams you see here are from the typescript (kindly made available to me by Paul Kerswill) of an article ‘Reversing “drift”: Innovation and diffusion in the London diphthong system’ by Kerswill, Torgersen and Fox, which is due to appear in issue 20(3) of the journal Language Variation and Change. It should be out next year.

Wednesday 3 October 2007

Talking faces

At the Saarbrücken congress (blog, 13-17 August) one of the hot topics was talking faces: animated cartoons to accompany synthesized speech or other sound-only material.

I’m glad to say that my own Department is involved in this research, too. Our contribution is called the SYNFACE project, led by Andy Faulkner, and you can read about it by following the link. This project does not involve synthesized speech, but rather aims to assist hearing-impaired listeners to understand speech — telephone conversations — by giving them the supplementary sensory input of an appropriate synthetic animated face to lip-read.

The animated model is called Kattis. She can cope with English and Dutch, but no other languages at present. You can also see a longer video clip.

What is intriguing about this scheme is that the lip movements are generated entirely by analysing the natural speech signal in real time, plus knowledge of the language involved. Other talking-face projects usually seem to present the face as a back-up to stretches of speech generated entirely by a speech synthesizer.

Anyhow, this kind of research is bringing phonetics and video animation together in what promises to be a very fruitful way.

Kattis Click image to animate

Andy Faulkner

Another talking face, from Barry-John Theobald

Tuesday 2 October 2007

The spoken and the speaker

Petr Rösel (see last two blog entries) writes further:

I couldn't stop thinking about the auditory effects of the sentence and manipulated it:
  1. I increased the loudness of the fricative portion, first by 5 dB and then by 10 dB. The comprehensibility increases dramatically.
  2. I increased the length of the fricative first by 100msecs and then by 200 msecs. The effect was much less dramatic.
Listen for yourself!

But Michael Ashby, whose recorded voice it is, has some views on the matter, too:

May I correct one thing, and add a couple of observations? The text on page 109 is not Do you object to dogs? but D’you object to dogs?. I specifically recall thinking that the D’you spelling prevented me from placing the onset accent on Do (which I might otherwise have preferred to do).
For what it’s worth, far from being unusually reduced or sloppy, the speech in the clip strikes me as obviously scripted, rather than spontaneous, and actually a bit hyperarticulated. The absence of yod-coalescence in d’you, the hint of [u] in the schwa of to, and the fairly firm consonant closures and crisp release of [t] are all giveaways.
I agree with David Marjanovic in hearing the vowel of dogs as an upgliding diphthong, but I think it goes considerably higher than his transcription suggests. Probably this is the raising effect of the following velar. Devoicing of the final /gz/ cluster is, of course, completely regular.
As for his comment “the first syllable is really at the limit of the (untrained) time resolution of my ears” — well, this is exactly the kind of thing we work on all the time in phonetics ear-training. Extreme reduction of preheads (anacrusis) is quite usual in my kind of English. Even very advanced learners of English undergoing phonetic training will often admit that they hear a vague blur at the beginning of an English phrase before the first accented syllable is reached.

One last little point: although the words were scripted, the intonation was not. (This sentence merely provides the context for the reply that follows.) Michael chose exactly the same pattern as I would have chosen myself: a high head (high-pitched accent) on -ject and then a drop down to a low rise nuclear tone on dogs. For us, a high head plus a (low) rise is the default for a yes-no question.

This pattern exemplifies the one major intonation difference between British (RP) and American English. The default for Americans has not a high but a low head. See p. 227 in the book.


Michael Ashby

Monday 1 October 2007

/What was that, | a/gain?

Several correspondents have commented on the sound clip (blog, Friday 28 September).

Lance Eccles writes

I'm a native English speaker (Australian), but I heard almost the same as what your Japanese correspondent heard.
I thought maybe the speaker was using one of those dialects in which "t" before a noun = "the". So I heard: "Do you get the dogs?", which admittedly didn't make much sense.
I certainly heard the verb as "get", but after reading what it really was, I could heard "object".

I think I ought to point out that in the north-of-England traditional-dialect where the is pronounced /t/, this plosive always (I think) has its release masked by a following plosive, and is often glottal anyhow. As far as I know, there is no north-of-England dialect in which the article is //.

Petr Rösel writes

When I listen to it I can easily switch from one version to the other. In my opinion and for my ears the ambiguity arises mainly from the affricate in object. The voiced fricative is fairly 'thin-lipped', in other words, it lacks a certain degree of fullness in sound.

And lastly this month’s Local-Kelly award for detailed auditory phonetic analysis goes to David Marjanovic for this comment:

I have had an interesting experience with this sound clip. The first time I listened to it, I wanted to hear "do you object to dogs", and that's what I did (barely). The next few times, I wanted to hear "do you get to dogs", and indeed I heard precisely that! Then I listened perhaps 10 times more, trying to (and succeeding in) hearing the correct version, and then I wasn't able to ignore the [ʒ] anymore even when I wanted to hear [g]. Still, the whole [d͡ʒ] is articulated farther back than I would have expected, making it shorter and thus more plosive-like.

On another note, I can't find a [j] in the clip (even though this didn't derail your Japanese correspondent), and I note with great (if parochial) delight that the [g] is partially and the supposed [z] entirely voiceless.

In sum, I hear something like [dʉu̯b̚ˌd͡ʒɛ̝k̚təˈdɒɔ̯g͡g̊ˁs] (or maybe rather [-z̥]), but the first syllable is really at the limit of the (untrained) time resolution of my ears.

Lance Eccles

(Correspondents get their picture here if I can find one on the web, or if you send me one. Sorry, nothing to hand for the other two correspondents quoted.)

John Local

Friday 28 September 2007

/What was that?

Listen to this sound clip from the CD-ROM that accompanies my book English Intonation.

What do you hear the speaker say?

Since the publication of the book, several readers have kindly written to draw my attention to various misprints and so on in the text or errors on the CD. I am grateful to them.

Sometimes, though, the supposed error is not an error at all. In the sound clip (p. 109 in the book) Michael Ashby is actually saying

Do you object to dogs?

and that is what I hear if I listen.

But a Japanese correspondent heard

Do you get to dogs?

and listed this for me as an error on the CD.

You can see why the misapprehension might arise. The first three syllables of Do you object are spoken very fast. I can still hear (or think I hear) the segments [dəjuəbˈdʒek-]. But I can see how a Japanese listener might miss the second [ə] and the non-audibly released [b], then mishear [] as [g]. The usual contextual elision of the final [t] of object and the equally usual non-audible release of [k] accounts for the rest of the mishearing.

The moral I draw is that EFL leaners do need to know about contextual effects such as weakening, elision, assimilation, and plosive releases. Otherwise they are in danger of failing to understand ordinary spoken English.

And this wasn’t even sloppy colloquial speech with accompanying extraneous noise. It was scripted and then read aloud in soundproof studio conditions.

From [dəjuəbˈdʒektəˈdɒgz] (the whole thing lasting less than a second) you have to be able to recover Do you object to dogs?.


Current blog

Archived from previous months:

To search my web pages, use this Google search.

Google Custom Search

my home page