Friday 9 December 2011

the BBC pronunciation unit

If you’re employed by the BBC, or work for an independent programme maker producing BBC programmes, then you are entitled to consult the BBC Pronunciation Unit for “professional advice about pronunciations in all languages”.

The staff of the Unit are all multilingual trained phoneticians, and do an excellent job.

Their advice covers anything from what, in a British context, might be called recherché and exotic…...through the names you’d think most educated BrE NSs would be familiar with (though presumably someone had asked). (That should, of course, be Oxford, not Cambridge.)

As you can see, their indications of pronunciation do not include IPA transcriptions, but do offer a choice of two respelling systems.

One, “BBC Modified Spelling”, has been in use for many years. Some of the symbols it uses have diacritics: ī for the vowel of PRICE, ō for the vowel of GOAT, and oo with a breve (not available in Unicode) for the vowel of FOOT; breves are also used on ă ĕ ĭ ŏ ŭ to represent schwa. The voiced dental fricative is shown by underlining, th, and underlining is also used for the digraphs zh (IPA ʒ) and hl (IPA ɬ). Stress is shown by a superscript acute mark.

The other scheme is a newer one. It avoids diacritics, but at the expense of being perhaps less transparent. The letter y is used in two different senses, representing in some cases the PRICE vowel and in others the palatal glide. PRICE can also be written igh. Schwa is written uh or uhr. Stress is shown by capitalization.

Symbols that might be open to interpretation are accompanied by a brief explanation.
Respelling systems based on orthographic conventions have one great advantage over IPA or IPA-style transcription systems. They are less phonetically explicit, more abstract. Instead of worrying about whether GOAT has a diphthong with a rounded first element (), a diphthong with an unrounded first element (əʊ) or a monophthong (), we just agree that ō (or oh) stands for whatever vowel you use in GOAT words.

However, respelling systems for English face particular difficulty in finding satisfactory symbols for
• the PRICE vowel, for which neither y nor igh is unambiguous, while ī has a diacritic
• the MOUTH vowel, for which both ou and ow are ambiguous (cf. soul, show)
• the GOAT vowel, for which oh may wrongly suggest a short vowel and oa, ou, ow are ambiguous (cf. broad, loud, now)
• schwa. If oh represents a long vowel, how can we make it clear that uh represents a short weak one?

_ _ _

This blog will now be suspended for a week. (I may see some of you in Łódź.) Next posting: 19 December.


  1. Maybe they do an excellent job, but they do nothing to try to teach presenters in Cornwall how to pronounce Cornish place-names.

  2. I had to think through all of these when posting on the now defunct BBC Word of Mouth Message Board. Many of the regular posters were unfamiliar with transcription — and in any case the software wouldn't display the symbols. A solution that worked for me was:

    • IGH
    • OW
    • OH
    • uh

  3. Some of the blog followers may be interested in the fact that there is a similar pronunciation database maintained by the German Public Broadcasting Corporation (= ARD). Here's a link:

  4. To represent oo with breve, use o͝o to get o͝o.

  5. Non-IPA double o with a breve can be represented in Unicode using U+035D (combining double breve): o͝o.

  6. I am mystified as to why phonetic respelling systems often use "h" to indicate a short vowel --e.g. "ih" for KIT and "eh" for DRESS.

    The only English words in which a vowel is actually followed by "h" are "ah", which has the PALM vowel, and "oh", which has the GOAT vowel. So where did this convention come from? And is there any evidence that the uninitiated understand it?

  7. VP - yes you're right. In the (probably now defunct) American encyclopaedia The World Book, they used a ludicrous respelling system that added a silent H after most vowels. Austria would be respelt AH-stree-uh.

  8. Pete

    AH-stree-uh seems pretty clear to me. In my accent it's OSS-tri-yuh.

    Yes, I know more accurate ways of representing pronunciation to people I know to share certain conventions. But writing for strangers I don't see a better alternative.

  9. vp

    I presume uh is the model on which the others are based.

  10. vp

    On second thoughts, I think I see the logic...

    1. TRAP, DRESS, KIT, LOT, STRUT — the 'short vowels' — are characterised by a closing single consonant letter (when syllable-final).

    2. H is a consonant letter which can't represent a sound when word-final.

    3. Therefore AH, EH, IH, OH, UH represent the 'short vowel' sounds —followed by nothing.

    [This contrast with A, E I, O, U without a closing consonant letter — yielding FACE, FLEECE, PRICE, GOAT, /j/+GOOSE.]

    Although not based on reality, the scheme does get support from the spelling of interjection Uh! and some names in -eh: Mannasseh, Tecsumeh.

    Despite the shaky support and naive logic, the system clearly had something going for it, because it was accepted and survived for some time.

  11. @David Crosbie:

    What you say makes a certain amount of sense, except that it contradicts the spelling/sound correspondence of the actual words "ah" and "oh".

    Since the "short vowels" must always be followed by a consonant in stressed syllables, why doesn't a respelling system exploit this by always following them with a representation of the next sound that follows in the word?

    Using "uh" for schwa doesn't work for me since it suggests the STRUT vowel, which for me is distinct from schwa. I actually think that "a", in an unstressed syllable, represents schwa pretty unambiguously in spelling: the only other possibility is TRAP but that is rarely/never found in unstressed syllables.

  12. @John Cowan and SH:

    That solution varies on how it shows. Yesterday on one computer, the breve was fully over both syllables, but a little greyed out.

    Today on a different computer, the breve (now properly dark) is double wide, but centered over the second O.

  13. vp

    I didn't say it was a good rationale. But despite its obvious defects it was favoured by natural selection over more plausible competing systems. I think I see some reasons for this:

    1. As rationales go, it's a very simple one — which is is a comfort to both writer and readers. By contrast, IPA or other symbols feel distinctly uncomfortable when first encountered. Even simple diacritics are a pain.

    2. The rationale is based on an existing sub-system — covering the 'short' values of the five vowel letters — which is already family to (almost) anyone who is literate in English.

    3. It's actually an advantage that the sequence VOWEL+h+CONSONANT is virtually non-existent in English spelling. The more the transcription resembles the orthography, the less likely the reader is to pay attention. Of course, this doesn't matter if the two are obviously equivalent — but in that case, the transcription would be rather pointless.

    Yes the existence of ah spelling for PALM words and oh spelling for GOAT words does tell against the system. However:
    • These words are pretty rare.
    • The sounds are pretty well confined to the end of words — precisely where no 'short vowel' is going to be transcribed.

    I don't think we should expect too much from a mainstream popular dictionary produced for native speakers. A given non-specialist user looking up a given word at a given time is probably after just one piece of information. The problem for the compilers is that they can't predict which: they can't assume, for example, that all users want to know only the spelling or only the pronunciation or only the most unusual of the meanings.

    So the compliers are forced to give pronunciations for every entry. But the typical user will very seldom consult them.

    I stand by uh for schwa. The only thing that woks better for nion-secialists is er — except that it doesn't work at all if the reader has a rhotic accent. Some people on the old Board preferred @ — but this suggest æ to me in the same way as uh suggests ʌ to you.

  14. Ellen K.: Yes, the exact display depends on what browser (and what version of the browser) you use, and what fonts are on your system. This is true of all text, really.

  15. To really do the job properly -- and if we can rely on a sufficiently intelligent readership to make use of any key that's provided, so that the system isn't unduly constrained by a need to keep it simple -- I think we'd have to differentiate between non-English place names, which should probably be pronounced in a uniform way that doesn't depend on the newsreader's own accent of English, and English place names, where for the newsreader to put on an accent other than his/her own might cause offence. A scheme to cope with this should have an arsenal of both phonetic and GOAT-vowel-like symbols.

  16. The BBC Modified Spelling document gives the strange recommendation:

    Gennadiy ROZHDESTVENSKIY.................. gĕnáadi roz̲h̲-d͜yáystvĕnski

    (In Cyrillic Геннадий Рождественский)

    My wife, a native-speaker Russian, was appalled. If you're going for something resembling authentic pronunciation, why mess with the first two vowels of the surname? The combination áy leads the reader to say the English FACE vowel, when e for the DRESS vowel would be closer to the sound and closer to the spelling of the transliteration. She would like to represent the first vowel as the STRUT vowel, which would need u. Perhaps it would be easier on the reader to mark the vowel as reduced: rŏz̲h̲.

    It's not impossible to type this transcription scheme on a computer, but it is a pain.

    PS I'm sure that the note...

    In Cambridge a joke. Even if it was entered by mistake, it must have been left in as a joke.


Note: only a member of this blog may post a comment.