Phonics and accents of English:
a view from phonetics

John Wells, UCL

This document uses Unicode to encode IPA phonetic symbols. If you cannot see a schwa here [ə] on your screen, Windows 95/98/ME/NT/2000 users please download Lucida Sans Unicode free of charge and install it on your system, and/or install a newer browser.

This talk was given at the LAGB in Leeds, 6 April 2001.

We are here to consider the implications of variations in regional accents for the teaching of literacy. My task is to consider this question from the point of view of a phonetician. (The examples of differences between accents are all discussed in greater detail in Wells 1982. For a thorough discussion of the nature of English orthography, see Carney 1994.)

English spelling, as we know, is a mixture of consistency and inconsistency. The consistencies mean that phonics is worth doing; the inconsistencies give rise to many difficulties for the learner, both in identifying a word from its written form – decoding – and in selecting the appropriate spelling for a word – encoding.

1. Equal difficulties

The first point that needs to be made is that many of the irregularities and inconsistencies of English orthography offer the same degree of difficulty to all speakers of English, no matter what accent they speak it with. For example, the reader faced with the written words river and diver gets no clue from the spelling to know that one is pronounced with short /ɪ/, the other with long /aɪ/. We are all equally helped to some extent by the fact that we have a word /ˈrɪvə/ in our vocabulary and no */ˈraɪvə/, and a /ˈdaɪvə/ but no */ˈdɪvə/. In the case of the spelling wind, which corresponds both to /wɪnd/ and to /waɪnd/, we have to rely on the context and our lexical and syntactic knowledge (noun, verb) to identify them correctly.

The beginning reader trying to figure out the meaning of the spelling ove in move, love and grove, faces the same difficulty. And the same applies to the notorious ough in rough, dough, through and so on, and to many other cases.

This is one of the principal reasons why people find it so difficult to spell correctly. That friend is spelt friend rather than the expected frend is an arbitrary oddity that is equally arbitrary and odd for everyone. The second syllable in driver and many other words leads us to expect the spelling er for the agent suffix; but in beggar it is ar, in sailor or and in the somewhat comparable martyr yr. That helps no one.

To be successful spellers we need to be aware of the rule "i before e except after c". That will help us with receive and deceive, related words as we see from reception and deception. But it is utterly arbitrary that we have to learn to put a p in receipt though not in deceit.

The pair lose and loose are distinct for everyone in speech. They are supposed to be distinct in writing, too; but the spelling difference between them is arbitrary. and must be learnt by rote.

We all know the difficulties even literate undergraduates often have with its and it's: problematic, because they are homophonous. Examples of other pairs or triplets that are homophones for everyone, but spelt differently, are bear and bare, right, write and rite.

2. Phonetic (realizational) differences

The second important point is that phonetic differences between accents, as opposed to phonological ones, give rise to no differencial advantage or disadvantage.

Here I am following essentially Trubetzkoy's (1931) classification of accent differences. By 'phonetic' differences I refer to differences in the phonetic quality of sounds (vowels or consonants, or even for that matter suprasegmental characteristics), differences which however do not affect the underlying system of oppositions nor the way the underlying units (phonemes or whatever) are distributed in the lexicon.

Take, for example, the minimal pair bed and bad. All native speakers of English consistently distinguish this pair and similar pairs; we all have an opposition in the vowel system between the vowel in the word meaning a place to sleep and the word meaning the opposite of good. Some of us may pronounce [bed] and [bɛd], some may pronounce [bed] and [będ], some may pronounce [bɛd] and [bad]. This does not matter. We can all equally learn to identify the short value of the letter e with whatever we do in the place to sleep, and the short value of the letter a with whatever we do in not good. The actual vowel qualities – the vowel heights, or other phonetic, realizational characteristics of the two sounds, are irrelevant.

So, among the consonants, with the glottal stop as a pronunciation of the /t/ phoneme. Pronouncing better as [ˈbeʔə] rather than [ˈbetə] does not make the word any harder or easier to spell. Nor does pronouncing it [ˈbetsə] or [ˈbeɾə]. It seems that some people find this point counterintuitive – but I am sure it is correct. It has been shown that even quite small London children know that a glottal stop and an alveolar stop in this intervocalic position mean the same thing, i.e. are phonologically identical even though we know they are phonetically different.

The same applies to different varieties of l-sound, including the vocalized variety that we now usually get in London. Speakers are able to identify [fɪɫ] and [fɪo] as repetitions of the same word, fill, and they know that this is the same item as the first part of [ˈfɪlɪŋ] or [ˈfɪlɪn] filling. That is to say, the existence of allophonic differences – dark l, vocalized l, clear l – causes no special problem.

In Scouse, the working-class accent of Liverpool, there tend to be fricatives where the rest of us pronounce final voiceless plosives, as in [teɪx ɪt̞ bax]. These fricatives are just positional/social allophones or variants of the plosives, and it certainly wouldn't help little Scousers if we were to use special spellings with kh and so on for them.

3. Phonological differences

Phonological differences between accents, on the other hand, may sometimes confer a relative advantage or disadvantage on speakers of a particular accent.

I say 'sometimes', because this is by no means always the case. For the numeral we agree in spelling one, some speakers say [wʌn], rhyming with run (or its northern equivalent [wʊn]), while others say [wɒn], rhyming with Ron. Both lots of speakers face an equally arbitrary spelling for this word, which is spelt as if pronounced to rhyme with cone (which is not a pronunciation used by anyone).

Conversely, the spelling of zebra is equally easy or difficult for those who say [ˈzebrə] and those who say [ˈziːbrə]. It merely exemplifies the long-short indeterminacy as between short and long of vowel letters in this sort of position, as we have already seen in river/diver.

I would like to subdivide phonological differences into four groups, again following Trubetzkoy, though with modifications. I shall consider in order first differences in lexical incidence, then those in structure (phonotactics), then those in system, and lastly the differential signalling of word or morpheme boundaries.

4. Lexical incidence

Virtually everywhere where English is spoken there is a social-stylistic variability in the pronunciation of the suffix and word ending that we spell ing, as in running, talking, eating. We have a 'high' variant with a velar nasal, [ˈrʌnɪŋ, ˈtɔːkɪŋ, ˈiːtɪŋ], and a 'low' variant with an alveolar nasal, [ˈrʌnɪn, ˈtɔːkɪn, ˈiːtɪn ~ ˈiːtn̩]. Although we can simulate the low variant in spelling by showing the so-called dropped g by an apostrophe (runnin' etc.) we do not on the whole do so (though the lyrics of pop songs and contact ads on the internet may constitute an exception). More to the point, this variability applies not only to the participial/gerundive suffix of verbs but also to various other items such as pudding, morning, ceiling (though not, of course, words such as king, string, sing). And for those who often use the alveolar variant it may be less than clear how far this group extends. We hear this when we find people hypercorrecting into [ˈmaʊntɪŋ] for mountain or [ˈkɪtʃɪŋ] for kitchen, and obviously people who do this have an extra problem in learning to spell such words correctly, with no g, as compared with those of us who have a consistent native distinction between the velar and alveolar word endings.

One of the best-known examples of regional difference in accent concerns the BATH words. We have a group of some fifty or sixty words, some of them very common, which have a long vowel in the south and in RP, but a short vowel in the north: bath, laugh, pass, example. Southerners are thus faced with pairs such as cancel–chancel, cancer–answer where the same vowel letter corresponds to two different vowel sounds (a breakdown in regular grapheme-phoneme correspondence), while northerners face no such irregularity. Southerners might find it confusing that bath and hearth, perfect rhymes, are spelt so differently; whereas for northerners, who pronounce them differently, it is obvious that they should be spelt differently.

The standard way of pronouncing the noun derived from to pronounce is pron[ʌ]nciation. Many people, however, use a form reflecting a kind of morphological regularization: pron[aʊ]nciation. These people are going to have trouble with the spelling, too, and will tend to write pronounciation. Those of us who have always made the standard vowel alternation will have no such trouble. (Compare profound/profundity, and also south/southern, where however the spoken vowel alternation is not reflected in the spelling.)

5. Structure (phonotactics)

Certain sequences of sounds are permitted in some accents but not in others. The most obvious instance of this concerns the phoneme /r/. In non-rhotic accents such as my own it cannot occur except before a vowel sound; historical /r/ has been lost when not followed by a vowel, that is before a consonant or before a pause. In rhotic accents, found recessively in the west of England and part of Lancashire, and also of course in Scotland and Ireland, no such restriction applies.

Thus for people like me lava and larva are homophones. I had to learn by rote that the stuff that comes from a volcano is spelt without a letter r, but that the caterpillar is spelt with one. The same applies to caught and court, rota and rotor, peninsula and peninsular, and the endings of half and scarf. We English face a spelling difficulty that the Scots, the Irish, and most Americans do not.

I do not believe that the prevalence of intrusive r in pronunciation has any bearing on this. What is important is differential homophony.

In Jamaican and other West Indian creoles there are constraints on final clusters such that pairs such as guest and guess, hold and hole, find and fine are pronounced identically. This is a categorical homophony, not to be confused with the effects of contextual elision in other accents of English. It leads to hypercorrections such as [ˈgestɪn] guessing, well attested in my survey of Jamaicans in London of thirty years ago (Wells 1974). The result is that West Indians face an added burden in learning the spelling of such words. Unlike the rest of us, they can't consult their own pronunciation to decide whether or not to write a t or a d in such words. This extra difficulty appears to carry over into the written English of children of West Indian parentage who have grown up in the UK.

For the same reason, it is easier for the Scots than for the English to spell words written with initial wh or w, e.g. whine–wine, where–wear. A Scot has only to consult his own pronunciation: is it [hw]ine or [w]ine? — and he knows which word to write with wh and which to write with w. But an English or Welsh person, who pronounces whine and wine identically, is faced by an arbitrary decision and must learn the spellings by heart. (I am assuming here that we analyse the first member of each such pair as phonologically involving a consonant cluster, /hw/, so that the issue is one of whether this cluster is permitted. An alternative analysis would make it a unitary phoneme /ʍ/, and this would then be classified as a difference in the consonant system.)

6. System

We turn now to differences of system between accents. A system in this sense is a set of contrasting oppositions. We are concerned with the fact that each point in the system is distinct from each other point, rather than with the question of how they are distinct.

Working-class accents in most parts of England lack a consistent /h/ phoneme: there is no /h/ phoneme in the consonant system, and speakers do not consistently distinguish between [h] and non-[h]. Hence pairs of words such as hedge and edge are homophones. This means that h-droppers are faced with a spelling dilemma that h-pronouncers are not. If we know that the barrier is a [hedʒ] and the rim is an [edʒ], we do not hesitate about their spelling. But if we think of them both as [edʒ], we have to learn by heart which one to write with h and which without. Equally, coming across hedge or edge in a written text, we might have to think which fits the context better, rather than relate the spelling immediately to a familiar difference in sound.

In the accents of the south of England, and in RP, the system of short strong vowels typically has six terms: pit, pet, pat, cut, pot, put. None of these words rhymes with any of the others. But in the popular accents of the north, the system comprises only five terms: pit, pet, pat, pot, put. The reason is that there cut and put rhyme perfectly. So do flood and good, and there is the same vowel in double as in should. Notice that the spellings here are no guide to the pronunciation of those who have the extra vowel. It follows that southerners and RP speakers arguably face an extra difficulty in learning to read the written forms, since they have to try out two variants each time before hitting on the right one. We have to learn by rote that buck is an animal (but compare butcher) while book is something to read (but compare blood). Northerners have a different problem: they know that both buck and book make [bʊk], but have to learn to associate one spelling with the animal and the other with the thing to read. (I am ignoring the fact that a recessive minority of northerners pronounce book as [buːk].)

So we need to distinguish between cases where the variably present term in the system is associated with a reasonably consistent spelling difference, as with /h/, and those where it is not, as with /ʌ/.

As linguists know, phonemic analysis is not without its contradictions and absurdities, which have led many phonologists to abandon the concept of phoneme. But we must not throw out the baby with the bathwater. The basic point remains, that some phonetic differences are distinctive while others are not.

At the beginning of a word, only some of us distinguish the two possibilities [ɪ] and [ə]. Those of us who do distinguish them have no difficulty making the spelling distinction between effect and affect, because we pronounce the words differently. Those who pronounce them identically will inevitably find them tricky to spell.

In a Scottish accent there is usually a distinction made between pairs such as hoarse and horse, four and for, which are homophonous in England. It follows that it is easier for the Scots than for the English to spell and identify members of such pairs. (The spelling or is ambiguous as between the Scottish sequences /or/ and /ɔr/, as seen in the non-rhyming pair sport and short; but oar and our never correspond to Scottish /ɔr/.)

Likewise there are those who consistently distinguish such pairs as mane vs. main (Swansea Valley) or wait vs. weight (parts of the North of England). They must find these words easier to spell and identify than those who do not distinguish them.

On the other hand many Londoners, and increasingly younger people elsewhere, do not distinguish thin from fin, pronouncing both [fɪn], and likewise make mother rhyme with lover. This th-fronting may be variable or perhaps sometimes categorical; if the latter, it clearly gives rise to an added difficulty in spelling and recognizing words containing /θ, ð, f, v/: if you say [fɪŋk], you would clearly expect think to be spelt fink. The same applies to another widespread pronunciation trend, that of yod coalescence, which leads to the loss of distinction between /tjuː/ (typically spelt tu) and /tʃuː/ (typically spelt choo, chew), /djuː/ (du) and /dʒuː/ (ju). If the first syllable of Tuesday sounds just like choose, you would expect it to be spelt in the same way.

7. Morpheme and word boundary effects

In Cockney, feel and fill are typically homophonous, [fɪo]. However, feeling and filling are distinct, [ˈfiːlɪn] vs. [ˈfɪlɪn]. The vowel-initial suffix triggers a preceding clear /l/, which in turn inhibits the vowel neutralization. Thus the -ing forms, whose spelling corresponds to their pronunciation, are a guide to how to spell the base forms. Similarly, the Cockney [pɔo] may be either pool or pull: the difference in their spelling and identity is recoverable from pooling [ˈpɔolɪn] vs. pulling [ˈpʊlɪn]. In Scotland and Northern Ireland, on the other hand, pooling and pulling are just as homophonous as pool and pull, and their spellings must be learnt by rote.

In American English, the correct spelling of t vs. d in the homophones waited and waded [ˈweɪɾɪd] is recoverable from the non-homophonous base forms wait [weɪt] vs. wade [weɪd]. But the medial consonants in Betty [ˈbeɾi] and ready [ˈreɾi] cannot be recovered in this way, and American children accordingly are liable to produce spelling errors such as Beddy, retty. Compare children in the North of England, who ought readily to be able to recover the underlying /t/ from get off [geɾ ˈɒf], shut up [ʃʊɾ ˈʊp].

In Scottish English, overt signalling of the morpheme boundary in tie#d [taˑed] keeps it distinct from tide [tʌɪd], whereas in England these words are homophonous. So they are easier to spell for the Scots than for the English. In Northern Ireland, a tenner [ˈtɛːnɚ] sounds different from a tenor [ˈtɛnɚ]: so the locals have a spelling advantage over the rest of us.

Since we are in Yorkshire, the last process I shall discuss is Yorkshire Assimilation (Wells 1982: 366-367). Even though wide spaces and white spaces are locally pronounced identically, the difference between wide and white can be recovered from their pronunciation in other phonetic environments, and the assimilation causes no extra spelling problems.

8. Final note

Throughout this discussion I have made the assumption that English is the learner's first language. That is to say, s/he knows the language before starting to read and write. I make a sharp distinction between teaching literacy and teaching the standard language or an approved way of pronouncing it. Teaching literacy within the context of EFL is a rather different matter, and in EFL it may well be necessary to teach sounds and sound contrasts as well as how to read and write them.

The other final point I must emphasize is that I know of no research that might demonstrate to what extent predictions such as I have made above are supported by facts. My comments are therefore speculative.


Valid XHTML 1.0!

Placed on the web by JCW 2001 04 12
My home page