What’s in a pronunciation? British and U.S. transcription models in the OED

The questions which could not be addressed during the webinar session were addressed by the panelists and answer are available to view below.


How many phonemes are there in the British English?

Usually we refer to 44 phonemes in English. There are 24 consonants, with 9 ‘fricatives’ /f, v, θ, ð, s, z, ʃ, ʒ, h/, 6 ‘plosives’ /p, b, t, d, k, ɡ/, 4 ‘approximants’ /l, r, w, j/, 3 ‘nasals’ /m, n, ŋ/, and 2 ‘affricates’ /tʃ, dʒ/. OED’s base consonant set for British English includes 26 consonants though, since we also recognise /x/ (as in ‘loch’) and /ɬ/ (the Welsh ‘ll’ sound). The 20 vowels are the 6 ‘short’ vowels /ɪ, ɛ, a, ʌ, ɒ, ʊ/, 6 ‘long’ vowels /iː, ɛː, ɜː, ɑː, ɔː, uː/, 7 ‘diphthongs’ /eɪ, ʌɪ, ɔɪ, ɪə, ʊə, əʊ, aʊ/, plus schwa /ə/. OED’s base vowel set for British English also includes the ‘nasalized’ vowels /ã/ and /ɒ̃/, and the barred-/ɪ/ and barred-/ʊ/ symbols (sometimes called ‘schwi’ and ‘schwu’), but these are not separate ‘phonemes’.

To L2 speakers of English, like those in Nigeria, which pronunciation do you recommend should be taught?

It really depends on the purpose. OED is not prescribing a variety or advocating one pronunciation over another, we just describe pronunciations in the least regionally-marked accents of the varieties we cover. For an accessible, straightforward guide to British and American pronunciations, see the Oxford Advanced Learner’s Dictionary (it would be up to an individual’s intended use whether they focus on British or American pronunciations). But varieties of English around the world are evolving and becoming more established in their own patterns, which is why we now have pronunciations for words from e.g. Nigeria (see the West African English model).

One of the problems with the “British English Model” is that there are many rhotic accents in European English (both in Britain and in Ireland). The phonemic representation of these is the one bias toward Southern English dialects which is rather problematic. In preparing a major new English-Cornish dictionary I chose to give IPA transcriptions for heteronymous headwords. Since Cornu-English is a rhotic dialect, such as “converse”. OED gives [kənˈvəːs] v. and [ˈkɒnvəs] n. We ended up choosing [kənˈvɝs] and [ˈkɒnvɚs] following John Wells. More specific guidance from the OED itself for British rhotic forms would be welcome.

We’d love to be able to represent words with other accents, especially where they are dialectal forms. But at the moment we’re focusing on completing the revisions with the British accent being RP. Although it’s certainly the case that the version of RP we represent is more common in southeast England, it remains the one British accent which cannot immediately associate the speaker with a geographical region. Whereas Scouse speakers will always have a Liverpool connection, or Cockney speakers a London connection, the same cannot be said of the British variety we currently represent. The Cornu-English dictionary sounds fascinating though, and if we have the opportunity to add a Cornish English variety in future, we’d love to tap your expertise on this! The rhoticity diacritic is a good alternative to the ‘plus-r’ approach we take with the U.S., Scottish, and Irish varieties (amongst others), but the ‘plus-r’ encompasses a slightly broader range of possibilities for our purposes and helps with limiting the symbol set required to cover our varieties.

I don’t think that the schwu and schwi in “beautiful” and “because” are formally sanctioned by the IPA. (They ought to be.)

Schwi and schwu don’t appear on the International Phonetic Alphabet, but IPA transcriptions have the ability to show narrower detail such as raising, lip-rounding, or lip-spreading through the use of diacritics. The OED transcription system is based on the IPA, but it’s important to remember that the transcriptions are phonological with a few phonetic details, so the symbols represent meaningful categories rather than precise articulations. In deciding what information is most useful to OED users, we are not limited to IPA symbols or conventions and can adapt them as needed.

Frequently the OED Word of the Day editors choose words that don’t have pronounciations listed in the OED, like “ambigue” on July 4 and “brigue” on July 5. What does that happen and how do we pronounce those words?

These words are obsolete, and OED policy is not to give current pronunciation transcriptions or audio for obsolete words. In some cases (though not for the two you mention) NED did give a pronunciation, in which case there will be a note in the etymology like the one demonstrated for ZOOPHILE in the talk.

Dr. Sangster read a quote of Daniel Jones a few slides ago about the purpose of transcription. Can you let us know where we can find the quote?

All references referred to in the talk can be found under the ‘Key Sources’ heading of the newly-published models for British English and U.S. English.

The choice to represent American flaps as /d/ seems very oriented to a non-American audience, since American dictionaries generall use /t/. Can you say a bit more about the role of the imagined audience in making decisions about how to represent pronunciations as well as which pronunciations are included?

The OED has a wide range of users, so we do not explicitly target any one demographic. As indicated in the talk, the choice to represent the U.S. tapped/flapped /t/ as /d/ was a convention adopted by Upton & Kretzschmar for the Oxford Dictionary of Pronunciation for Current English (‘ODP’), and by extension for OED. Kretzschmar’s 2008 chapter explains why, but it is not to do with the audience demographics: ‘Intervocalic t is most often realized as a tap or flap, frequently with voicing, so that latter/ladder are homonyms for educated Americans, as [læɾɚ]; this pronunciation is transcribed as [lædər] in ODP, because the dictionary uses a restricted symbol set that does not include the ɾ or ɚ.’ (2008, p.48). So this approach is justified on both homonymy and lexicographical/publishing grounds. As shown at the end of the webinar though, through our OPTRA system we will have the ability in future to alter this easily if required.

On what principles do you transcribe words of foreign origin?

Decisions for inclusion (of loan words, etc.) are made by other parts of OED’s editorial team. Once that decision is made, we transcribe for British and US English, and any appropriate World English. We do not supply original-language transcriptions, although we may use them for our own research.

Why show an optinal schwa for US in fear (ɪ(ə)) (but not in fearing) if it’s allophonic? Aspiration is not shown (and that’s allophonic too).

It’s an interesting point; there are lots of allophonic details we could show that we don’t happen to in OED. Unlike aspiration patterns though, the schwa is not compulsory. There is an element of editorial judgement as to what we show, and it’s often down to the salience and impact of the feature in question, as well as how much choice the speaker has. Sometimes it can come down to whether the difference could/should be represented by an additional phonemic-level symbol, but in this case, it also potentially makes a one-syllable word become two-syllables, so that variation would make quite a big difference and would be something we would want to represent with two pronunciations (condensed with the brackets).

You yourself pronounced NEAR as a long monophthong as do many (most?) speakers of SSBE. Why not transcribe it as such?

Although a long [ɪː] quality is increasingly found in British English, it has not yet become as clearly established (or recognized) as the dominant allophone for NEAR as [ɛː] has for SQUARE. It’s looking like that may simply be a matter of time though!

I appreciate that you’re not aiming for comprehensive history of the pronunciation of any entry, but when pronunciations are updated to be contemporary are you archiving previous versions?

Yes these can be retrieved from our editorial system’s version history.

Are there any plans to incorporate pronunciation concerns for lyric (singing) diction?

No, that would be outside OED’s remit I think.

Phonetic Phun Facts

Every first-year linguistics or speech and language therapy student in country sits in a lecture where they’re told that the most consonants you can have in an English syllable onset is 3, as in straw. But there’s one, and only one word in the OED which potentially begins with four in British English: spleuchan /ˈspl(j)uːxən/; a tobacco pouch often used as a purse, from Scottish & Irish varieties.

Onset clusters (consonants found at the beginning of a syllable) have always fascinated me. There’s a very broad pattern that if there are 3 consonants, it will be /s/ followed by one of /p t k/ followed by one of /l r w j/ (though not all of those combinations are attested). But OED has some beautiful supplements to this, such as:

The systematic approach to some aspects of variation just occasionally leads to more variants than perhaps we’d like. The words with the most variants: 16 British for endurance, translationally, and transindividual; 16 U.S. for piezoresistivity.

Some other rare onset clusters:

A few years ago, I looked into the longest and shortest words in the OED, so this may have changed a little since then: