Pronunciations for World Englishes
Note from Catherine Sangster, Head of Pronunciations, December 2021
“A number of our World English pronunciations lack audio at present. We create our audio with freelance actor-phoneticians in our Oxford recording studio, but since spring 2020 it has been unsafe for us to run these sessions. We are working on addressing this backlog as soon as we can do so safely.”
Pronunciation models and keys
|East African English||key||model|
|Hong Kong English||key||model|
|New Zealand English||key||model|
|Singapore and Malaysian English||key||model|
|South African English||key||model|
|West African English||key||model|
In the model pages above, you will find a discussion of the rationale for each model, along with potential discrepancies and compromises. The major sources used in the development of the model are listed. These are in no way full and comprehensive reference lists; many individuals, websites, articles and books have been consulted and the range of sources is growing as review continues. In particular, variety-specific dictionaries, knowledgeable native speakers, and online evidence of natural language use continue to be used in researching new pronunciations for inclusion in the OED.
For a numbers of the models, a key source of pronunciation information (if not the original starting point) is J.C. Wells’ Accents of English (CUP 1982, three volumes).
Further varieties are under consideration, including Indian English.
Whose pronunciations are these?
As with the British and U.S. pronunciations, all pronunciations given in the OED are based on carefully-devised ‘models’ of the variety in question. As a baseline, the models created for the OED are intended to reflect the speech of educated urban speakers of standard English. The pronunciation of such speakers across the varieties will be largely intelligible to each other, excepting particular groups of words where sound contrasts are made in one variety and not in another. In more homogenous populations, the creation of a model is fairly straightforward (discussed in the next section), but many of the populations covered by the term ‘educated urban speakers’ are increasingly diverse, geographically, ethnically, and culturally, and this in turn increases the linguistic diversity. The Caribbean, for example, spans more than 25 nations and territories, each with their particular dialectal features. It would be impractical to have separate models for each of these within the OED, so rather than single out one as a ‘standard’, the model reflects some of the commonalities across forms of English spoken in the Caribbean. This principle applies across models; no particular dialect is promoted over another, the models simply reflect the key features typically heard in the pronunciations of such speakers.
The process of devising an OED model for each variety broadly follows these steps:
- the intended variety is defined in broad terms, determining the scope of the model
- a phonetician compiles and prioritizes a range of reputable sources, including notes from specialist consultants
- the sources are used to compile an inventory of consonant sounds used in the variety, along with any contextual restrictions
- a list of common key words is cross-referenced to establish the primary vowel qualities and contrasts used in the variety, again with any contextual restrictions
- the variety’s prosodic patterns are reviewed, to assist with the allocation of primary and/or secondary stress
- where multiple subvarieties are to be covered by a single model, the process above is repeated for each subvariety
- commonalities are identified across subvarieties along with areas of greatest difference; if appropriate, the extremes on a continuum may both be adopted, priority given to more populous subvarieties, or priority assigned based on the proportion of OED entries from a given subvariety
- the provisional model is checked against recordings of speakers of the general variety and is reviewed by another phonetician
- model conversion charts are drawn up to assist editors in importing pronunciations from major sources
- still subject to ongoing review, the model is used in reviewing and adding new pronunciations to entries indicated as being associated with that variety.
It is important to consider that dictionary transcriptions are purely a guide, and any given speaker’s pronunciation in any given circumstance will be influenced by a wide range of factors. Linguistic context (words before and after), geographical and cultural background, social context (formality of the situation, urgency, who is listening, and the status of speaker and listener), environmental factors (e.g. noise levels), personal features (e.g. lisps, stammers), and the speaker’s familiarity with the word will all affect how a word is spoken. The OED pronunciations cannot reflect such variation, but where the pronunciation of a word in isolation would be different for a sufficiently high proportion of otherwise similar speakers, multiple pronunciations are given.
All of the OED’s models are essentially phonemic, rather than phonetic. The pronunciations are not intended to represent a precise articulation, but to use a limited set of symbols to reflect the meaningful contrasts in a variety. Substituting one symbol for another within a given variety therefore has the potential to change the meaning of the word. Not all varieties of English have the same contrasts. For example, the words lot and cloth have the same vowel in contemporary British English, but different vowels in Irish English and Caribbean English.
This is one of two reasons why different models are needed for each variety. The other reason is that the symbols used should as accurately as possible reflect the quality of the sounds in those varieties, and even where the same contrast is made in two different varieties, the qualities can be different. For example, both British English and Australian English distinguish between the vowels in dress and trap, but in both cases the Australian English vowels are made with the tongue slightly closer to the roof of the mouth, which contributes to making Australian English sound different from British English. The appropriate contrasts and vowel qualities are established using a common set of vowel keywords, each one an example from a group of similar words (a ‘standard lexical set’). The original list of 24 was devised by John Wells with a focus on standard British and American pronunciations, but was used throughout his landmark Accents of English (1982) volumes and has been widely adopted since with additional keywords added or modified as required.
In some varieties, the range of possible articulations covered by the symbols as used in the OED can be quite broad. Several vowels in South African English, for example, almost seem to overlap. Often this is because of contextual factors, such that a sound may be articulated with subtle differences depending on its position in a word or its surrounding sounds (consider how the /p/ in British English pot and spot are articulated differently but still ‘mean’ the same sound). Other times there may be etymological reasons, such that the ‘same’ vowel in two similar words may be exhibit variation depending on the languages they originated from and the individual speaker’s sensitivity to those languages (such as with words of Afrikaans origin).
However, several more precise (phonetic) details are reflected in the pronunciations. These features are predictable and consistent within the variety, to the extent that they would be notable by their absence. For example, in several varieties, the ‘flapping’ of /t/ between vowels in unstressed syllables is a significant feature, a feature indicated by collapsing the contrast with /d/.
The consonant symbols used in the models are those of the International Phonetic Alphabet (IPA) and are mostly the same across varieties of English. The vowel symbols reflect different perceptual points across the total possible range of vowel qualities, for which common practice in phonetics is to show these on a map known as a vowel quadrilateral (below). Vowels at the top of the quadrilateral are produced with part of the tongue close to the top of the mouth, those at the bottom being more open. Those on the left are made with the focus on the front of the tongue (the part used to make the /j/ in yes). Those on the right are made with the focus on the back of the tongue (the part used to make the /k/ in car). The quadrilateral below shows the approximate regions covered by the various vowel symbols used in across the models. Those in red are usually pronounced with rounded lips, those in blue with the lips neutral or spread. Some symbols are also positioned in a slightly different area than regular users of the IPA may be familiar with. This is to reflect contrastive qualities required for particular varieties; for example, /ɜ/ is the symbol used by one of the Australian English sources to reflect a closer vowel quality than /ə/, even though it is often used equal to or more open than /ə/. /ʌ/ is used much more broadly; it often reflects a central, fairly open vowel quality (per its common use in British English phonemic transcription), but in some of the varieties of English that symbol reflects a quality towards the back of the vowel space nearer its IPA quality. In addition to these symbols, the OED uses the symbols /ᵻ/ and /ᵿ/ to indicate free variation between /ɪ/-/ə/ and /ʊ/-/ə/ respectively.
The OED indicates if some vowels in a model are relatively longer than others in similar contexts, and uses the International Phonetic Alphabet’s length marks /ː/ to indicate this. Diphthongs, which involve gliding the tongue from one position to another in such a way that they are pronounced in one syllable, are represented with two vowel symbols. The approximate movement can be worked out from the constituent symbols, so British English /ɪə/ starts near the /ɪ/ position and moves in the direction of /ə/.