Symbols for sounds: our dictionary pronunciations explained
One of the markers of a comprehensive and respectable dictionary is the inclusion of pronunciations, a representation of how the words are spoken and heard. The Oxford English Dictionary uses around 45 symbols to show how words are pronounced, with slightly different symbol systems for different varieties of English (one system for British English, another for US English, another for South African English, and so on). To many people, the symbols we use don’t mean very much, but they’re a lot more straightforward than they seem at first glance.
What’s wrong with letters?
English is such a mixture of different languages that there is very little consistency between the 26 letters of the alphabet and the sounds used when we say them out loud. The classic example (often erroneously attributed to George Bernard Shaw) is the idea that you could pronounce ‘ghoti’ as ‘fish’, since ‘gh’ is pronounced as ‘f’ in laugh, ‘o’ as ‘i’ in women, and ‘ti’ as ‘sh’ in action. There are problems with this example – such as the fact that ‘gh’ is never ‘f’ at the beginning of a word – but there are plenty of others we can find.
Take the vowel in mt. It is the same vowel as in b, polce, pce, recve, pce, k, q, and pple. ‘s’ is pronounced differently in it, raie, expanion, and leiure, while letters can be ‘silent’ in some words (such as the ‘p’ in sychology and cuboard). ‘u’ occurs in combinations with other letters to be pronounced differently in bsiness, bry, cgh, tgh, fll, q, se, cd, lgh, f, gge, g, bant, hse, shlder, musem, pport, and pre (though for some speakers this rhymes with four). A big part of the problem for English is that we have just five vowel letters (a, e, i, o, u) but around 20 different vowel sounds. Consonants are a little different, in that we have 21 consonant letters and around 25 consonant sounds, but it’s not as simple as just ‘adding’ four more.
How do we show that all those instances of ‘u’ are pronounced differently? Some dictionaries use a ‘re-spell’ system. They use particular combinations of letters to reflect particular sounds, so the vowel in guy might be written as /ahy/. But we still need to learn the system, as there isn’t a single word in the OED where the letters ‘ahy’ would be pronounced like the vowel in guy. It is often easier to think of pronunciation as completely separate from the spelling.
So what can we use instead?
A group of French language teachers called the Association Phonétique Internationale (the International Phonetic Association) published the first International Phonetic Alphabet (IPA) back in the 1880s. They wanted a way of transcribing speech sounds that could be used for any language. The principle is essentially that of ‘one sound, one symbol’. Not all of the possible speech sounds are used in English (for example, some African languages have ‘click’ sounds), so for our dictionaries we only need to focus on those that are.
Of course, some sounds in the IPA are found in English but are not contrastive. In other words, you can swap them with each other without altering the meaning of a word. Our dictionaries do not show glottal stops, for example, because they always mean the same thing as another sound (usually ‘t’). It doesn’t change the literal meaning if I say bottle with a ‘t’ or with a glottal stop (transcribed as [ʔ] on the IPA) – it still means bottle.
The aim of dictionary pronunciations is not to display all the possible precise articulations and details, but to show the distinctive sounds in each word. These transcriptions are more technically called ‘phonemic’. To understand how this works, imagine a colour wheel. There is an infinite number of possible colours, but generally we divide the possibilities into broader categories like green, blue, and red. We can be more precise if we need to, but most of the time it isn’t necessary. A similar principle applies to speech sounds. There is an infinite number of possible speech sounds, but similar sounds that ‘mean’ the same thing in our language are in groups called ‘phonemes’ and we have one symbol for each phoneme. To show that we are using sound symbols instead of letters, we put them between slash brackets, so a word such as cheese is shown as /tʃiːz/ (made up of the three phonemes /tʃ/, /iː/ and /z/), while yacht is shown as /jɒt/.
For British English, there are around 25 consonant groups (phonemes) and 20 vowel groups. Many of the symbols are English letters, but there are some additional ones worth noting. In particular, the sound at the end of sing (/ŋ/) looks like an ‘n’ with a tail that goes in the same direction as ‘ɡ’. As a phonetician, this symbol has become so common in my day-job that I habitually handwrite ‘ŋ’ instead of ‘nɡ’! It’s also helpful to remember that the symbol /θ/ is the Greek letter theta, and is used to represent the voiceless th sound at the start of that word (so theta is /ˈθiːtə/).
We also have symbols which indicate the one or two most prominent (‘stressed’) syllables in a word. The most prominent is preceded by /ˈ/ and the next most prominent by /ˌ/, so the word paranormal is shown as /ˌparəˈnɔːml/ as the ‘nor’ part is the most prominent and the ‘pa’ the next most.
What about other varieties of English?
One of the benefits of using IPA-based transcription instead of respelling is that it is easier for us to show some of the subtle differences between different varieties, such as the fact that the American vowel in goat is slightly different from the British equivalent, so we can use different symbols (/ɡəʊt/ in British but /ɡoʊt/ in US). With 10 more varieties now reflected in the OED, each has its own set of symbols to show what is distinctive in that variety. For example, unlike most other varieties of English, Philippine English has a similar vowel in trap as it does in lot.
Why use the transcriptions when there are audio files?
Over the past few years, we’ve been adding lots of audio files to our online dictionary entries so you can hear someone speaking the words aloud. We hope visitors find these useful, but they could never replace the transcriptions. The transcriptions are deliberately broad, covering a range of possibilities. Our speakers have their own particular speech features and it may not be clear how far you could deviate from his or her pronunciation and still be recognizably saying the same word. Also, not everyone can access the audio files all of the time (because of hearing difficulties, or they have no headphones/speakers) and once you’re familiar with the symbols, it can be quicker to read them than listen!