Dictionaries which try to show the pronunciation of words can basically use one of two methods: either they can use a respelling system (this was the only possibility for dictionaries compiled up to the middle of the 19th century), or they can use a phonetic alphabet. To all intents and purposes nowadays, this means the IPA. Is one of these better than the other?

Writing about the presentation of pronunciation seems to be flavour of the month among linguists and phoneticians at the moment, and the latest edition of Language and History (the journal of the Henry Sweet Society) is devoted to it. One of the articles claims that IPA is gaining the upper hand, and that this is the system used by a majority of current dictionaries. I dispute this: even the Oxford family of dictionaries uses a respelling system in the Little, Pocket and Paperback versions, the IPA transcription not going ‘below’ (in size) the Concise. Similarly, Chambers Dictionary still uses a respelling, as do the Penguin English Dictionary, Collins’ Dictionary and Thesaurus of the English Language, and Cassell’s English Dictionary, all reputable dictionaries. US dictionaries have never favoured IPA.

These are all dictionaries intended for native speakers, as opposed to the ever-proliferating dictionaries aimed at non-native learners (the Oxford Advanced Learners’ being the best known). One justification given for using the IPA in learners’ dictionaries is that while we British are unaccustomed to seeing the IPA, and so easily confused by it, the rest of the world has been using it happily for many years. Another reason is that the IPA gives a consistent and accurate representation of the pronunciation, while a respelling does not. Is that true?

All the dictionaries that use IPA still transcribe the words phonemically, or phonologically, rather than phonetically, so that the letters used are an approximation to the actual sound value, and may differ quite a lot from the value given to the same letter in the transcription of other languages. It is notorious, for instance, that almost all the phonemes of French are articulatorily different from their apparent equivalent in English, so that where a French dictionary will write, for instance, /də/ for orthographic ‘de’, English will use the same schwa symbol in /ðə/ for ‘the’; but the two schwas are articulatorily, auditorily and acoustically quite distinct. The German schwa is different again. Also, IPA transcriptions tend to become as fixed as any traditional orthography, even when the phonetic detail changes: Danish, for instance, is still written with symbols that may have been appropriate in the 1940s, but are not so today. When I visited Denmark in 1982, an eminent professor of phonetics, now deceased, told me that had he spoken Danish as a child like the current (i.e. in 1982) television newsreaders, his father would have beaten him to within an inch of his life, and yet this 1982 version of Danish had become the norm. It may be argued that the introductory material can deal with this, but how many people actually read the introduction?

Respelling is often justified because it is easier to interpret by reference to traditional orthography, and also because it can be read by speakers of different accents, and so is less prescriptive than IPA. (It is ironic that phoneticians, who may advocate IPA, and are certainly descriptivists, should be accused in this way of being prescriptive.) However, in many ways, respelling is just as prescriptive: most systems do not allow for the accents of that triangle of England in which there is no velar nasal phoneme (all occurrences of ‘ng’ are pronounced /ŋg/ – finger and singer rhyme). Likewise, whatever vowel letter or combination of letters is chosen to represent /ɑː/ is generally used in John Wells’ BATH set, even though in the northern half of England, and also in many parts of the US, it is the TRAP vowel that is used in bath and its like. At one stage in its history, the Chambers 20th Century Dictionary had a symbol that allowed for this: ‘â’ could be pronounced ‘long’ or ‘short’ in bath, while ‘a’ was always ‘short’ (as in hat), and ‘ä’ was ‘long’ (as in calm), but unfortunately, in later editions this innovation was dropped, and the dictionary became more prescriptive again. No dictionary that I know of deals satisafctorily with the complex of STRUT, FOOT and GOOSE: northern England never carried out the STRUT~FOOT split, and Scots has a different distribution of FOOT and GOOSE from England. Some words even in southern England may vary in their FOOT~GOOSE realisation: room is a well-known example.

Both systems have their advocates, and both systems have their advantages and disadvantages. I don’t see the demise of respelling coming any time soon.