UCL Division of Psychology & Language Sciences

John Wells’s phonetic blog
archive for April 2007


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 and Safari 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

John Wells

Monday 30 April 2007


Saturday’s minor earthquake in Kent meant that the newsreaders made several references to the Richter scale. As usual, they mostly pronounced it either [ˈrɪxtə] or [ˈrɪçtə]. But Charles Richter, the creator of the earthquale magnitude scale, was not from a German-speaking country, and the newsreaders’ otherwise admirable familiarity with German pronunciation is here misapplied. He was an American, born in Ohio. Being American, he naturally pronounced his name /ˈrɪktɚ/, and we should do the same (or, in the case of us non-rhotic speakers, as /ˈrɪktə/).

This information is readily available from the BBC Pronunciation Unit and in the Oxford BBC Guide to Pronunciation. It is explained in some detail in LPD. But the newsreaders persist in their ignorance. (Sorry, I’m beginning to sound like a reader of the Daily Telegraph.)

Charles Richter

Friday 27 April 2007

Television newsreaders’ RP (iii)

The remaining two variables investigated by Hannisdal in her dissertation (blog, yesterday and the day before) were these.

  • Smoothing

    Pronunciation of /aɪə/ and /aʊə/ as [aə, aː] proved to be “very frequent in broadcast RP”. My native-English-speaking undergraduate students often used to refuse to believe that any such thing was possible: how wrong they were! Among the newsreaders, there was a highly significant statistical difference between the sexes, with men smoothing much more than women. As you might expect, smoothing was less common across a morpheme boundary (e.g. higher as compared with its possible homophone hire), and less common before a pause.
  • R sandhi

    There was a tendency for men to use more linking/intrusive /r/ than women, but the difference was not significant. The /r/ was realized as [ɾ] only in foreign names (Igor Ivanov); otherwise it was an ordinary [ɹ]. (Compare Daniel Jones’s pronunciation, where the tap was the norm for all linking r.) The factors favouring r sandhi are exemplified, in descending order, by for a, four o’clock, former economic. It was least likely in proper names (Doctor Austen, Sir Alex). The main factor disfavouring it was another /r/ adjacent (e.g. terror alert). Overall, the link was used in 60% of possible cases. Interestingly, intrusive /r/ was more prevalent after /ɑː, ɔː/ than after /ə/. That is, it was more likely that there would be an /r/ in Jack Straw_ is than in Nelson Mandela_ is. However the raw number of cases of intrusive /r/ after /ə/ was higher because of the greater frequency of schwa in word-final position.

ə ˈlaən
Ron Kimball

Thursday 26 April 2007

Television newsreaders’ RP (ii)

More findings from Hannisdal’s survey of newsreaders’ pronunciation (yesterday’s blog):

  • T Voicing

    An incredible 35% of final /t/s, prevocalic across a word boundary, were voiced, i.e. pronounced American-style as [ɾ]. Unlike in American English, this is not found in content words: Brits don’t make atom sound like Adam. (There were one or two exceptions, including British and pre-adjectival pretty.) In Hannisdal’s corpus it is mainly found only in function words, across a word boundary. It was particularly prevalent in such cases as it is, that it’s, but if, at a, what about, a lot of, a bit of, not exactly, get elected. Yes, we do need to update our descriptions.
  • Yod Coalescence

    Words with traditional /tj, dj/ before a stressed vowel (tube, Tuesday, student, due, reduce, duty, during) were pronounced instead with /tʃ, dʒ/ in a massive 46% of cases. In LPD I labelled these variants “non-RP”. Clearly I was wrong to do so (even if it’s true for people of my own advanced age).

The University of Bergen

Wednesday 25 April 2007

Television newsreaders’ RP

Besides the visit to China, my other overseas engagement during the last two weeks was a doctoral defence (or “disputation”) at the University of Bergen in Norway. I was one of the two “opponents” (examiners) of a PhD thesis by Bente Hannisdal entitled Variability and change in Received Pronunciation.

This dissertation is a brilliant piece of work, which we had no hesitation in passing. Ms Hannisdal has collected a large number of recordings of British television newsreaders (thirty people; half of them men, half women; from the BBC, ITV and Sky channels; an hour of speech from each individual). This constitutes an extensive corpus of contemporary RP, and has enabled her to discover some interesting facts, not previously established, about this variety of English. Among them:

  • GOAT Allophony

    In words such as sold, roll, where the GOAT vowel is followed by dark /l/, 24 of the 30 newsreaders use the [ɒʊ] allophone categorically, and a further four of them do so variably. Only two do not use it at all. We need to update our descriptions of RP to cover this.
  • CURE Lowering

    Despite reports of its imminent demise, it turns out that /ʊə/ is alive and well, persisting particularly after /j/ and before /r/, as in Euro, Europe, European, security, secure, during. The switch to /ɔː/ is found principally in the complementary phonetic environments, e.g. in poor, sure, tour. The only important exceptions to this generalization are your and you’re, which — despite the /j/ preceding and the possible linking /r/ following — the newsreaders pronounced 100% with /ɔː/.
More tomorrow.

Bente Hannisdal
Bente Hannisdal
Photo: Eli Kristine Korsmo

Tuesday 24 April 2007

Banded or bandied?

Here is what was written in the Financial Times for 14-15 March, European edition. (I came across it on board the aircraft on the way home from Chengdu. This is my first blog entry to be composed, in advance, at a height of 35,000 feet or 10,000 m.)

What the author meant, of course, was bandied about, from the verb bandy about ‘to mention an idea, name, remark etc several times, especially in order to seem impressive’ (LDOCE). There is no verb band about.

Bandy is one of the words that ends with the happY vowel, which we nowadays write /i/, representing the neutralization (non-distinction) of // and /ɪ/ in weak final position, and ranging phonetically between tense [i] and lax [ɪ].

The conditions for bandied and banded to be homophonous, as they presumably are for John Gapper, the author of this piece, are twofold. First, the happY vowel must be realized with a quality and quantity identical to those of /ɪ/, and secondly the past tense/past participle suffix -ed must be pronounced as /ɪd/, not as /əd/.

Since my own pronunciation satisfies these two criteria, the words are homophonous for me, [ˈbændɪd]. I therefore have to be careful to spell each one appropriately. But Tony Blair, for instance, with his pervasive use of /ə/ in weak syllables, would presumably make a difference between them. So would anyone who has an [i]-like quality for the presuffixal happY vowel of bandied. Some speakers have both these characteristics. The trend is towards distinguishing the two words as [ˈbændid] bandied vs. [ˈbændəd] banded.

Other relevant pairs are candied – candid, taxis – taxes, studied – studded, and even Billy’s – Bill is. In the first of these it is a matter not of an inflectional ending, but of the adjectival suffix -id. But this usually seems to agree with -ed.

Sichuan University, Chengdu:
Jiang Hongliu, JCW, Gao Jun

Monday 23 April 2007

Chengdu conference

The week before last I was in Chengdu in China to attend a conference on English Phonetics Teaching and Research. This was apparently the first such nationwide conference, and was attended by some sixty phoneticians and EFL teachers from all over China. They had been kind enough to invite me as a keynote speaker.

I was impressed by the enthusiasm displayed on all sides. Those present felt strongly that it was urgent to improve the generally rather poor level of pronunciation among Chinese learners of EFL. They eagerly discussed teaching techniques and reported the various successful strategies they had used. My own view on this is that the key thing for the teacher of pronunciation is a combination of knowledge (i.e. phonetic understanding of what is required) and skill (i.e. the actor-like qualities that enable the teacher, through his/her performance, to enthuse the learners and bring out their best).

Chengdu, a city of eleven million people, is the capital of Sichuan province. I found it bustling with economic vigour. Apartment blocks and motorways were being constructed on all sides just as much as in Hong Kong or Seoul. Mobile phones were everywhere.

Chengdu is also a tourist centre. Two of the attractions I was able to visit were the Panda Breeding Research Centre, where I saw young pandas frolicking around and adults engaged in the more serious business of eating and sleeping, and the 70-metre high Buddha figure carved in the rock at nearly Leshan, surviving as the largest such figure in the world after the Taliban destroyed the ones at Bamiyan in Afghanistan.

The blog took a two-week Easter break

Young pandas

Adult pandas

The Grand Buddha
Photos: John Wells

Friday 6 April 2007

Apical vowels again

Since I shall be in China next week, I think it’s appropriate to return to the question of the Chinese (Mandarin) apical vowels and their symbolization.

You will remember that Cheung’s proposal for the narrow IPA representation of the vowel in the sequences written in Hanyu Pinyin as si ci zi was [zɯ] and for that in the sequences written shi chi zhi [ʐɨ], with various other possibilities for those who pronounce them in slightly different ways: [ɯ, z͡ɯ, z̞͡ɯ, ɯz] (or with [ɹ instead of ]) and [ɻɨ, ʐ͡ɨ, ɻ͡ɨ, ɨɻ] respectively. (There’s an important lesson here: the non-uniqueness of IPA representation of a particular speech ‘sound’.)

The complexity of these phonetic representations can be interpreted as a shortcoming in the IPA alphabet. In fact sinologists — experts in Chinese — have long used the symbols ɿ for the alveolar, z-type vowel, and ʅ for the retroflex, ʐ-type one.

Many sinologists, and many Chinese phoneticians, believe that these are official IPA symbols. But they’re not.

However it strikes me that an easy way out of this problem would be for someone to propose to the new IPA Council (now in the course of being elected) that the symbols ɿ and ʅ be added to the Association’s alphabet. On the other hand we don’t really need new symbols: as phonemic symbols, // and /ʐ̩/ would seem to be entirely adequate (or even /z/ and /ʐ/ without the syllabicity mark).

If you would like to listen to the sounds in question, here are sound clips of si and shi (thanks to the University of Vermont’s Chinese program).

ɿ = z̩

ʅ = ʐ̩

Thursday 5 April 2007

That’s funny

Quite a common idiomatic intonation pattern in English consists of a fall on a the word That with the verb to be, followed by a rise on the complement.

\That’s | /funny.
\That was | /good.
\That’s | a re/lief.
\That’s | a /pity.
\That’ll be | the /day!

A: Your 'keys aren’t \here.
B: Well \that’s | rather /odd. ‖
(I could have \sworn | I 'left them here \/somewhere.)

A: I’m 'going to start going to the \gym.
B: \That’ll be | the /day!

This pattern is rather difficult to explain. The word that is merely anaphoric — it refers to some situation or thing already mentioned, rather as a pronoun does. But if we replace that by it, we cannot use the same tone pattern.

* \It’s | /funny.
* \It was | /good.

Obviously, this is connected to the fact that the default for demonstratives is to be accented, whereas the default for pronouns is to be unaccented. But why is that so important that it is not merely accented, but carries a nuclear tone?

We might wonder whether this pattern is not a fall plus rise but a fall-rise.

? \/That’s funny.
? \/That was good.

But in this analysis, if the single nucleus is on the demonstrative, the words funny and good are deaccented, postnuclear — which would normally imply that they must be given (old), or repetitious, or predictable in the context, none of which is the case here. So that can’t be right. And in any case we would still be left with a difficult-to-explain nucleus on the demonstrative.

Compare contrastive focus on that, with a genuine fall-rise:

A: They’re 'both pretty \good.
B: Well \/that one is | but \/this one isn’t.

Unfortunately, I don’t think I discussed this intonational idiom in my English Intonation book. You may answer

\That’s | a /pity.

Wednesday 4 April 2007

Unaspirated /p t k/

Every now and again I get emails from Chinese correspondents worried about the statement found in textbooks to the effect that English voiceless plosives, usually aspirated, are nevertheless unaspirated after /s/, as in spin, steam, skin.

The statement is correct. There is an audible difference between the bilabial plosive in pin and that in spin. In the first it is aspirated, in the second it is unaspirated. This means that in the first there is a delay after the release of the bilabial closure before the onset of voicing for the vowel (the duration of this delay being known as ‘voice onset time’): [pʰɪn]. In the second there is no such delay: [sp⁼ɪn]. (The IPA offers no diacritic to mean ‘unaspirated’; here I shall show it by using a raised equals sign, though unless you’ve got Text Size set to pretty large you may not be able to see more than a raised hyphen.)

It is difficult to hear the difference between unaspirated [p⁼ t⁼ k⁼] and devoiced [b̥ d̥ ɡ̊]. Theoretically it depends on the difference between fortis articulation, for [p⁼ t⁼ k⁼], and lenis articulation, for [b̥ d̥ ɡ̊]. But fortis-lenis is much easier to hear in fricatives and affricates, where it affects the volume of air flow and therefore the intensity of the turbulence, than it is in plosives.

Whereas we like to think of the English plosives as being basically voiceless vs. voiced (despite positional devoicing), it is clear that the Chinese ones are aspirated vs. unaspirated. Their representation in Hanyu Pinyin may be p, t, k vs. b, d, g, but their phonetic representation is usually reckoned to be /pʰ tʰ kʰ/ vs. /p t k/. Mapping these onto English /p t k/ vs. /b d g/ respectively works well, except for the position after /s/. In that position Chinese EFL learners should be encouraged to use the sound that they usually map onto the English voiced plosive.

It may be interesting to note that when English words are borrowed into Welsh (which, like Chinese, has basically aspirated vs. unaspirated plosives), spite becomes sbeit, spell becomes sbel, (di)scourse becomes sgwrs, and screw becomes sgriw. But Welsh is inconsistent: for some reason, studio is stiwdio, not *sdiwdio, and station is stesion, not *sdesion.




Tuesday 3 April 2007

Starting an ewe

Look carefully at this display, photographed in my local supermarket.

Do you notice anything odd about the language? OK, there are some unfamiliar French proper names, the use of capitals is erratic, there’s a missing definite article and it says frm instead of from — but what else?

It’s the indefinite article. It says an ewe’s cheese. But no one says an ewe. We say a ewe.

(Welsh English is a possible exception, for reasons we’ll come to.)

The rule for the a ~ an alternation in the indefinite article is of course that we use a before a consonant, an before a vowel. But this is a phonetic rule, not an orthographic one. What matters is the sound, not the spelling. Ewe /juː/ begins with a semivowel, and semivowels in English count as consonants. That’s why we say, and write, a yard, a yob, a useful idea, a unit, a European, a ewe.

On the other hand we say, and write, an heiress, an hour, an honest opinion, an MP. And argue about a(n) hotel and a(n) historic event.

Tesco’s signwriters almost certainly would never say /ən ˈjuː/. But they must have been dazzled by the spelling into convincing themselves that ewe begins with a vowel sound.

In Welsh there is a falling (diminuendo) diphthong /ɪu/, spelt iw, uw, yw, as in Rhiwbina and Troed-y-Rhiw, both of which are places in south Wales. Those with strong Welsh influence in their pronunciation use this diphthong in place of English /juː/. So that’s why I’ll let the signwriters off if they’re Welsh.

Monday 2 April 2007

Becky Dauer

I was very sad to hear that Rebecca (Becky) Dauer died last Thursday.

Rebecca Dauer
15 March 1947 – 29 March 2007
Photo: Steven Dauer

She will be remembered particularly for her book Accurate English (Prentice Hall, 1993), which I consider to be the best EFL pronunciation course for those who want to learn American English (see reviews). She was also the pronunciation editor of the Longman Advanced American Dictionary and of the Longman Dictionary of American English.

Becky had an MA from the University of Michigan and a PhD from the University of Edinburgh. I got to know her in 1983 at the ICPhS in Utrecht. From that year until her death she was on the staff (AmE: faculty) of the University of Massachusetts at Amherst, where she taught ESL.

She acted as a consultant on AmE for the Longman Pronunciation Dictionary, and was always ready to answer my questions and to offer sensible advice. She wrote to me less than a month ago saying she was being treated for cancer, but I didn’t realize how serious it was.

I shall miss her. Sincere condolences to her widower, Steve.

Archived from previous months:

Current blog

my home page