[Simplified Spelling Society Newsletter, 1986/2. pp5-13. Later designated Journal 3]
[Message from new SSS President, 2004 AGM talk, Newsletter, Media.]
[A recently edited version of this article was published in two parts in J32 2003 and J33.]
[Note: this page is set to display International Phonetic Alphabet characters. If your browser over-rides this setting, by default or by choice, the IPA characters may not show.]

English Accents and their Implications for Spelling Reform.

J C Wells.

Dr John Wells is Reader in Phonetics in the Department of Phonetics & Linguistics, University College London and former Secretary of the International Phonetic Association. His 3-volume study Accents of English (Cambridge University Press) underlies his talk to the Society on 25.1.1986, of which we here publish an edited version with his permission.


1.1 The alphabet.
An ideal spelling system, we all know, will have one symbol for one sound, one grapheme for one phoneme, and it is with the difficulties associated with this principle that we are concerned here. One major difficulty, if we confine ourselves to the Latin alphabet, is that it is an arbitrary list of 26 letters which don't necessarily correspond to the sound-systems of the languages which have to use it; in particular they don't correspond to English.

1.2 Danger of parochialism: any.
However another problem arises from diversity of pronunciation. There is a great danger of insularity or parochialism, of assuming that everybody makes the same particular set of contrasts that we ourselves make and take for granted: unfortunately what is obvious and normal to one speaker may be exotic, unusual, subtle and strange to another. There are thus all sorts of little facts about how English is pronounced round the world by native speakers which may giveus pause in our reforming zeal. We all know the word any for example ought to be written with <e> because it rhymes with penny; we learn any, many as exceptions. But they are typically not exceptions for the southern Irish who say /ani/ rather than /eni/; for them any and penny don't rhyme, and they would see nothing strange about writing any with <a>. Maybe they would want to write <nn>, but that is another matter.


2.1 Morphophonemic past tense <-d>?
Ought mist and missed to be spelt identically because they are pronounced identically, or should we give the past-tense a regular appearance with <d>, even when, as in missed, it is pronounced /t/. In deciding this, we should consider the Nigerians, who either say /misd/, with the difficult sequence /sd/, or more usually assimilate it as /mizd/, with voicing throughout. For them past-tense /d/ is always voiced, and any assimilation changes the stem, so the past of kick tends to be /kigd/. I mention this because one should be aware of it and not because it should necessarily influence spelling reform.

2.2 Spell or omit <r>?
From New Spelling onwards the importance of catering for accents other than Received Pronunciation has been clear from the treatment of historical <r>. In my speech I don't distinguish source:sauce, and if spelling reform proposals do so, the reason is historical: historically source had <r>, and sauce did not. But other varieties of English (Scottish, Irish, west of England, most American, Canadian) make the distinction in speech. Similarly the homophones rotor:rota are distinct in rhotic accents which retain historical <r>, which could justify keeping the distinction in spelling. However people who pronounce intrusive /r/, saying perhaps rotar of duties, will have to remember not to write <r> in some positions where they pronounce it but conversely to write it sometimes when they don't pronounce it as in rotor, where such speakers only pronounce it before a vowel. Faced with such discrepancies, spelling reform has two alternatives: either speakers of non-rhotic accents must learn by rote when to write historical <r> and when not; or those who pronounce historical <r> must ignore every <r> not preceding a vowel; and that of course might be difficult.

2.3 Problem words with <r>.
There are some problem words in this area, too. One is the colour-name khaki, which is written without <r> as it had none in its source-language Hindustani. Nevertheless in the west of England and in Canada people say /ka:rki/. I don't know what the Scots do, but the Americans on the other hand are at least consistent and say /kæki/. Similarly char, tea, another word with no historical /r/ (it is cognate with Russian чай). So one would perhaps have to allow different spellings in such words. Then there is gormless, nowadays written with <r>, which is etymologically wrong, and some people therefore advocate writing <gaumless>.

2.4 Distinguish <ng:ngg>?
A similar problem arises with <ng>. Consider the pair singer:finger, whose endings most speakers distinguish, so that it seems logical to write singer:fingger. There could be other such pairs, like kinglet (a little king): singlet. The trouble is, people in the trapezium linking Birmingham-Manchester-Liverpool rhyme singer:finger, so they would have to learn an extra arbitrary distinction if one were shown in spelling. It is therefore a distinction we might well decide to ignore - so also simplifying the comparison of a word like long, whose irregularly pronounced forms longer, longest we maybe don't need to reflect in writing.


3.1 Greater problems with vowels.
Such variations in pronunciation mean we may have to violate the principle of one sound per letter and one letter per sound in quite obvious ways, ways that probably everyone can accept. Greater difficulties perhaps arise with vowel-sounds and sets of vowel-contrasts, where I think the danger is particularly strong of wrongly assuming everybody makes the same contrasts.

3.2 Sam:psalm contrast.
Thus New Spelling provides specially for the words I shall refer to as the lexical set BATH, like pass, path, etc., by allowing either <a> or <aa>, according to the pronunciations /bæθ, ba:θ/ - a permissive solution which assumes everyone distinguishes the vowels in gather, father, and allows everyone to spell them according to their own pronunciation. The trouble is, not everybody makes this pam:palm, sam:psalm contrast in the first place; even in the west of England and certainly in Scotland and Northern Ireland some people have the same vowel in sam:psalm, saying approximately /sæm/ for both. It would actually have been consistent with this fact for New Spelling to ignore the <a:aa> distinction, because it has very low functional load, in other words there are very few word-pairs that are distinguished as sam:psalm are. That is why Scots and others can ignore the distinction without suffering disaster (or /dizæstər/). We can thus perhaps ignore the contrast RP makes between mass:pass. It is not only northerners who don't distinguish them, but also Scots, who have no choice anyhow between the two possibilities and may say either /mæs, pæs/ or /ma:s, pa:s/. On the one hand we may ignore that distinction, but on the other hand RP speakers must remember to distinguish pass:farce (which I rhyme perfectly), reflecting historical <r> in farce but not in pass. However by thus taking accent variability into account one lies open to the objection of ignoring the principle of one letter per sound: the southern English know that mass:pass differ in sound, while pass:farce do not yet we are proposing the same spelling for the first pair and different spelling for the second. That is quite a well-known problem.

3.3 Cloth:lot:thought.
Less well known is the very similar situation affecting the set of words exemplified by cloth, such as cross, cough etc., where even within RP in this century we had a rival pronunciation which might be spelt clawth, crawss, cawf. RP resolved this by turning them from majority RP forms into very strange minority forms, which people now laugh at. But the problem here is that American pronunciation really corresponds to crawss, cawf, clawth rather than cross, cof, cloth, inasmuch as Americans give cross etc., the vowel of thought rather than of lot. Amongst all these different phonetic realizations, the important thing is to see what the contrasts are. Taking lot as a typical /o/ word, and thought as a typical 'aw' word, my point is that most people in this country speak cloth and other words like it with the same vowel as lot, and we therefore feel they all ought to be spelt with <o>, and not <aw> (unlike thought); but Americans (with very few exceptions) pair cloth with thought, and not with lot. So Americans might logically want to write cloth as clawth, and incidentally long as lawng, since for them it also belongs with thought, not with lot.

3.4 Ignore such contrasts?
If pressed, I would propose the same radical solution here as with TRAP:BATH:PALM, namely to ignore the whole set of contrasts and write them identically, all with <o> or whatever. This would suit Scottish and Northern Irish pronunciation, many Scots having the same vowel for all of lot:thought:cloth, so that a distinction might seem unnecessary. We would then have to forget the cot:caught distinction, which most Scots, Northern Irish, Canadians and many Americans (particularly in the far west, in California) don't make. In fact the lot:thought contrast is declining in the United States, so we who make a sharp difference between lot:thought are perhaps becoming a minority; we may not be a minority yet but there are many people who make no such distinction and who would have to learn an arbitrary distinction between hock:hawk, if reformed spelling distinguished them. On the other hand RP speakers would face the same problem here again with <r>, since there are words with historical <r>, like court, which are presumably going to reflect the distinction still, and logically one should write kot for t.o. cot, caught, but kort for court. Again, some people would have to accept no spelling-distinction for some words they pronounce differently, but a distinction for otherwords they pronounce the same.

3.5 Misconceptions about accents leading to spelling errors?
There might be a danger of misspelling because of mistaken impressions of other accents. RP speakers today have such problems if they try imitating American or Scottish accents - even talented English actors or impressionists attempting American or Scottish speech quite often make mistakes. Peter Sellers, in his Balham gateway to the south, talks lyrically about morning coming and says 'and now at last we see the dorn approaches'. But Americans don't say /dorn/, they say /da:n/. Similarly RP speakers might think words like dawn ought to be spelt with <r> in reformed spelling, because they mistakenly think rhotic speakers pronounce an /r/. Clearly even for highly literate people sound in a sense dominates spelling in their mental picture of words, so when imitating another accent, they set up correspondences between sounds in their own accent and what they imagine are the equivalents in the other accent, rather than by reference to the spelling, even though the <r> wud give them a true indication in this case. English actors playing Scottish parts likewise make many errors with /r/, pronouncing comma and China with final<r>: they feel English shwa ought to correspond to Scottish /r/, as in father, but in many other words there is no such correspondence.

4. <a> SETS AND <o> SETS.

4.1 Merging <a> sets, <o> sets.
I suggested it might be useful to use just <a> in the lexical sets TRAP:BATH:PALM (I use such key-words in my book to represent the many words that have the same vowel, we hope for everybody) and just <o> in LOT:CLOTH:THOUGHT, reflecting the fact that Scots for example may not make those distinctions, while others may send BATH in the direction of either TRAP or PALM, and CLOTH in the direction of either LOT or THOUGHT.

4.2 American <o>-/a/ crossovers.
A difficulty with this simplification is that most Americans pronounce PALM:LOT with the same vowel. Thus father:bother rhyme perfectly in typical American pronunciation, and Americans make puns that don't work for the British: the Swedish car called a Saab sounds like sob, and the Americans put up posters for the painter Salvador Dali saying 'Hello Dali', which to them sounds like 'Hello Dolly'. This means that the Americans, an important influential part of the English-speaking community, are confronted with a further uncertainty here, namely how to spell words in which they use the /a:/ sound: they are going to have to write it <o> like lot in most words, but presumably <a> in father and palm.

4.3 Anglo-American inversions.
Then, although RP says bath with /a:/ and American says it with /æ/, there are other words in which the reverse distinction obtains: the first vowel in pasta is /a:/ in the States, and Americans might therefore want to spell it with <o>. Competing spellings might have to be allowed in this area too, according to pronunciation. But perhaps <pasta> is tolerable because foreign borrowings are always problematic and a source of exceptions.

4.4 Merging NORTH:FORCE.
Another contrast that New Spelling is inclined to allow for is one which used to be made in England and still is in Scotland and to some extent in the United States, between for example horse:hoarse. Many Scots distinguish them, and a broad Jamaican accent (really down-market creole) has for example me ride an an arss and me vice oorss. Slightly posher Jamaican has harss/hoorss for the two words, reflecting an historical contrast which we have mostly lost in England, but which may persist in living memory (that is why New Spelling allows it), and is still well found in Scotland. Common sense may tell us to ignore this sort of contrast, because many people don't make it. But if we ignore it, Scots would risk misspelling words like story, because for Scots story begins as stow: for them it is logical to write it with the vowel of goat, and it may seem perverse with another vowel; but maybe they will just have to endure that. Not only these two sets horse:hoarse, which in my sets are NORTH:FORCE) have merged: all the <oor> words (my set CURE) have done so too. Many English people now pronounce Shaw, shore, sure as homophones, but for other speakers they may all be distinct. There are people who can make a 3-way contrast in an old-fashioned kind of RP, and certainly many still make a 2-way distinction. Many accents would merge shore:sure, but keep Shaw distinct. So I would suggest all the forms with <r> could merge, and we ignore the Scottish or Irish distinction between war, shore, sure, writing them all perhaps with <or>. When I once made an effort to learn Jamaican, NORTH:FORCE was the contrast I found most difficult: I had to remember to say poork for the meat, but fark for the implement you might eat it with, although pork and fork rhyme for me. It is striking, but historically quite right that the Scottish and Jamaican contrast are the same, and in just the same words. FORCE had a long vowel, while NORTH had a short vowel, which accounts for the t.o. of hoarse and horse.


5.1 Maximalism and minimalist.
Logically there are two extreme positions one can adopt towards these contrasts. The maximalist position would say our orthography must reflect all the contrasts that anyone makes, and the minimalist position would say we should reflect only those that everyone makes.

5.2 Problems of maximal contrast.
The maximalist position would lead to very undesirable consequences, as the following examples show. Many southern English make a longer vowel in bad than in lad and may even have minimal pairs between the name Sally and the verb to sally, or between shandy the long drink and brandy the short drink; obviously spelling should ignore this distinction. Likewise some Scots distinguish tide:tied, the point being not the quality of the vowel itself, but whether contrasts are made between vowels in different sets of words. In fact Scots here reflect a grammatical boundary, <d> in tied being a suffix on tie; in Fife:five on the other hand the unvoiced or voiced final consonant distinguishes the environment; but insofar as the distinction is predictable, it can be ignored because it is allophonic. In Northern Ireland and a few other places they distinguish days:daze; we are obviously going to have to ignore that. Also in that area is the contrast between late:eight made in many parts of England, where the local accent has /le:t, eit/, essentially reflecting the historical fricative <gh>. If the distinctions are allophones, we can ignore them; but for some people they may not be allophones.

5.3 Who needs the distinctions?
Then there is a contrast among words like nurse, where English and Americans make no distinction; but Scots and Irish may distinguish pearl:curl for example, contrasting them in the same way as perry:curry; and Hertz:hurts, fir:fur may similarly differ. That is a kind of justification for present spellings, which accurately reflect this distinction. Even so, t.o. sometimes spells the first vowel <ear> as in pearl, sometimes <er> as in defer. But if, like New Spelling, we abolish that distinction, then we are removing that contrast. English or Americans will typically say these distinctions are subtleties they can't possibly cope with. The only point I would make is that for those who make the distinction, it is one that logically ought to be retained in a reformed spelling, and they are going to have to be convinced otherwise if it is to be abolished.

5.4 Contrasts matter, not sounds.
The point is again, it doesn't matter what the actual phonetic realization is; what is important is the network of contrasts, because that is what the orthography must reflect. If we take, say, the word soap where we all use our 0 vowel, it doesn't matter what precise quality of vowel we use, provided we use the same vowel in rope, goat, coat and all other words in the same set.


6.1 Jamaican <a:o> merger.
The minimalist position, spelling only the contrasts everybody makes, has difficulties too. One might think everybody contrasts pat:pot, but they don't. Jamaicans for example typically say /pat/ for both. Now people may object that there are not many Jamaicans, so they can be ignored; but as a group they have very special problems with English spelling. Whereas other English-speakers mainly follow their pronunciation in spelling the <a:o> contrast (with one or two exceptions like wash), for Jamaicans it is hard to decide which spelling is right and what therefore is also the posh pronunciation. Ordinary Jamaicans say /rat/ both for the animal and for putrefaction, and have to learn which to spell <rat> and which <rot>. A social factor is also involved: to speak educated Jamaican, they have to learn to distinguish rat:rot as other English-speakers do. But if we are aiming to lighten the burden of arbitrary spelling distinctions, no reform project I have ever seen solves this difficulty for West Indians.

6.2 Southern US <a:o> merger.
Farm:form are similar, but not only Jamaicans say /fa:(r)m/ for both: many Americans speaking popular accents in the south also do. They have a test phrase about being born in a barn, and it is well known that some people reverse the two: country bumpkins in the southern states are ridiculed as being barn in a born, with the typical confusion of people trying to introduce a contrast into their speech that they don't natively have. Again, I think on balance we must retain the <a:o> distinction in those sets of words. But it is difficult.

6.3 Southern US <i:e> merger.
Likewise many American Southerners, of all social classes, don't distinguish the vowels in KIT:DRESS before nasals, and make no difference between pin:pen (hence the terms writing-pen and stick-pin, to make the distinction clear). Listen to Jim Reeves singing Lord give me stringth (like string), not distinguishing /i, e/ before the nasal <ng>. And when these American Southerners look at the spelling <England>, they think we ought to say it with /e/ as it is written. It would be rather drastic, I think, to abolish this contrast and write pin:pen identically, but unless we did, it would cause many American Southerners a big spelling problem. When I say Southerners, I should include California, whose speech is a mixture because it is in many respects northern rather than southern, in United States terms, but has certain southern features, such as not distinguishing /i, e/ before nasals.


7.1 Fool:goose merger.
New Spelling was criticized for distinguishing the vowels of foot:goose, and this distinction, I think, could be dropped. Consider the pair good:mood, which English people don't rhyme (we could spend hours discussing whether we pronounce room with the vowel of good or the vowel of mood). However in Scotland good:mood rhyme perfectly, and also in Ulster. Forcing a contrast of spelling with <oo:uu>, as New Spelling does, is therefore an arbitrary extra distinction. As there are very few word-pairs distinguished in this way (pull:pool, full:fool, and a few pairs including inflected forms, such as wood:wooed, could:cooed), I think we won't suffer too seriously if we ignore the distinction; we would soon learn to cope with the homophony of you fool, you've pulled me into the pool and it's full, because after all the words are homophones in Scottish and some Irish speech.

7.2 Functional load of <sh:zh>.
What is involved here is the important question of what is known technically as functional load, that is, the number of words that are distinguished by a given contrast. When functional load is low, then a contrast can be ignored, whereas when functional load is rather high, then presumably it ought to be reflected in the spelling. Using this criterion, one might well decide not to distinguish <sh:zh> as consonants, because there are very few word-pairs in which they actually represent a minimal pair. There are non-rhyming pairs like pressure:measure, mission:vision, but in no everyday words at least does the <sh:zh> contrast carry the power to distinguish two words. (We can safely overlook such rarities as Aleutian:allusion, Confucian:confusion.) Rather than impose the rebarbative <zh>, we might write <sh> for both.

There is another contrast with very low functional load which nobody has proposed abolishing as far as I know: that exemplified by my lexical set STRUT:FOOT, which contains pairs like cut:put. However we are really considering three sets here, STRUT:FOOT:GOOSE, because Scots, as we have seen, don't contrast foot:goose, while northerners don't contrast strut:foot, though they do distinguish these from goose. So we face a kind of chain, on which everybody contrasts the extreme points, but not everybody contrasts both extremes and the middle. The situation with the BATH words is parallel to this. I suppose the logical solution is again to allow the intermediate group, in this case FOOT (just as BATH and CLOTH represented intermediate groups), to be written with either vowel. So words like put, foot could be spelt either like strut, cut, which would seem logical for people in northern England, or like goose, which would seem logical for people probably in southern England and certainly for Scots and Irish.

7.4 A fourth contrast?
The Scots do have an extra possibility of contrast here that RP doesn't between for example /brud, bru:d/, for brood and brewed, which are identical for me but tend to be distinct in Scottish pronunciation. Distinctions like this may just have to be ignored: brood:brewed is a grammatical distinction like tide:tied that reflects a suffix in one case and not in the other, and again there are very few pairs of words involved.

7.5 Small lexical set FOOT.
A factor to bear in mind here is that the lexical set FOOT is very small, containing only about 40 words (see p. 132 in my book), and although some are quite common words like good, should, foot, put and so on, there are not many of them, and so it is only a small number of words we might decide to allow to be spelt in two competing ways. If diacritics were used, a solution might be to spell all the vowels in the STRUT:FOOT:GOOSE chain with <u>, but to minimize the visual differences by perhaps optional use of diacritics according to pronunciation.


8.1 Accents as social labels.
A complication with the northern pronunciation which merges cut:put is that it bears considerable sociolinguistic value, in that it is well known to distinguish social class, or educated versus uneducated speech, in northern England. Even that is perhaps not quite exact because there are educated northerners particularly in the Northumberland area who don't make this distinction at all.

8.2 Hypercorrection.
Even people elsewhere may be uncertain and reverse the vowel sounds - you hear northerners pronouncing sugar like RP rugger, and pudding like RP budding, in a mistaken attempt to make the distinction. Or they may stumble over the vowels in good luck: if you grow up as a northerner giving luck the same vowel as good, and you then try to speak RP, no longer giving cut and love the vowel of put, then the temptation is to say /gʌd lʎk/ (despite the <oo>), or if you know that's a danger, you may even get it the wrong way round and say /gʌk lʋk/. The pair gas:mask gives rise to the same problem: if you first learn to say /gas mask/, and you later discover it's posh to say /gra:s/ instead of /gras/, you might say /ga:s ma:sk/ or even /ga:s mask/.

8.3 Accent prejudice and spelling reform.
Spelling reformers have to confront sociolinguistic facts of the following kind. Most of the features of Scottish pronunciation as heard in the speech of somebody like David Steel are regarded as entirely educated and beautiful and just different. But some other divergences from RP are associated with particular regions, or with lower social class, and they tend to be condemned as ugly and nasty and not to be encouraged. If the first type is catered for in a spelling reform, it is not going to raise any contrary voices, but the second type may well do so, because the reform may be seen as bolstering undesirable pronunciations. The objective, scientific observer of course discounts these social views and refuses to make such value judgments, but a reforming movement does have to take such prejudices into account. And I think these prejudices might well say that we must retain the difference between the strut and foot spellings, and therefore we must spell the FOOT words with <oo> or whatever, rather than with <u> or whatever. In other words we must retain the contrasts that everybody makes except northerners. This is hard on the northerners, but maybe that's life as it is - unless we can reform these prejudices about accents.

8.4 The nurse:square contrast.
Thus there are contrasts which some people don't make, though there may be a widespread feeling that they ought to make them and therefore that they ought to be catered for in a spelling reform. Another example is the nurse:square vowel contrast which many Liverpudlians for example don't make, so that fur:fair are not distinguished, and the name Mary is pronounced like furry, or conversely you have cairtains hanging in the room rather than curtains. It is going to be difficult for speakers of such accents to remember the spellings if we keep the distinction in a reformed orthography. But these accents are widely condemned as ugly, etc., and that may be a reason for ignoring this problem of the Liverpudlians.

8.5 H-dropping.
The case of <h> is obviously similar. Millions of English people do not pronounce it, so it might seem logical to write harm:arm identically, in line with that pronunciation. But that would go against the social attitude that it's incorrect to drop /h/ and therefore the spelling ought to reflect its presence. Obviously we shan't go on writing <h> in honest and hour, but no doubt people would say we ought to go on writing it in harm to reflect the prestige pronunciation that does distinguish harm:arm.

8.6 The <ing> ending.
Similarly with the ending <-ing>: almost wherever English is spoken there is a rivalry between a relatively high-status pronunciation with <-ing> as in sing, thus running, wailing, eating, and a relatively low-status pronunciation with final <n> as in tin, thus runnin or run'n, eatn, walk'n. I suppose again the feeling would be that we have to keep the <ng> spelling to reflect the prestige pronunciation. Nevertheless it is a source of uncertainty for many people, who may produce hypercorrections like a brazing hussy rather than a brazen hussy, or chicking, kitching instead of chicken, kitchen.


9.1 Declining status of RP.
I have the impression that reform proposals this century have been very firmly based upon RP, together with some nods in the direction of archaizing tendencies (which is why historical <r> is reflected). It's clear that in the last quarter of a century in England the position of RP has been very seriously eroded, in that RP no longer enjoys the unquestioned status that it previously did. There are now many people who not only don't speak it - that was always so - but who don't aspire to it any more, and who would regard it as quite unrealistic to aspire towards it. In fact I think what has changed is the perceived model of beautiful or ideal speech, which is for many people no longer RP. This can be seen in all sorts of ways. Teachers of English as a foreign language for example get increasingly dissatisfied with the transcriptions the phoneticians offer them as the models for foreigners to imitate - one thing that has aroused particular complaint is pronouncing happy with a final /i/, because the many people who say /hæpi:/ with a final long vowel feel that it's unreasonable that they should be expected to change; they don't feel the lack of prestige that once attached to saying /hæpi:, siti:/ and so on, and don't feel it in any way reprehensible to use the formerly non-RP pronunciation /i:/ in such words.

9.2 Allophonic variation.
I have yet to discuss various technical phonological questions like the phonemic principle. It's clear that where we have allophonic differences, that is, realizational differences, we can ignore them. This means that essentially where any two sounds are used in such a way that we can predict from the surrounding sounds which will be used, then we can ignore any such difference. This is why we can ignore the difference between /t/ and glottal stop: many people in a word like department have a glottal stop before the <m> ([dipa:?mənt] rather than (dipa:tmənt]). To say [dipa:tmənt] is a really careful, special pronunciation for most people nowadays. But we can ignore that distinction because we can set up a simple rule that <t> before nasals is pronounced as glottal stop. Atmosphere is another example ([æ?məsfi:ə]), and indeed the fact that its a variable rule is another reason why it needn't affect spelling. Or again many people feel there is really a rather sharp difference between the O they use in most cases and the O they use before /l/, so that goal:go may be [gol, gəυ]. Students often complain when I expect them to use the same transcription symbol for these two different vowels, as they feel them to be; but as long as we can set up a rule that <o> before <l> has a special pronunciation, there is no problem: they are just allophonic variants of the same phoneme. They may sound a bit different but the difference is predictable, and so it may be ignored in an orthography. The problem with goal:go is that the /l/ is now being lost and once people drop the final /l/ in goal, then it's a non-rhyme of go, although both of them end in vowels (that is virtually the situation for some speakers already). Here I think in our spelling reform we had better be very carefully archaizing or conservative and retain written <l> wherever an /l/ occurred and RP still pronounces it.

9.3 American intervocalic <t>.
American intervocalic <t> is an interesting case, because it's moving from being allophonic to involving a neutralization and therefore becoming phonemic. Once the minimal pair Adam:atom becomes identical, then <t> in atom is no longer allophonic. But they are not yet necessarily identical: Americans may say atom with voiced <t>, which goes with /əta:mik/, but /ædəm/ the first man, maybe a subtle distinction, but many Americans insist they are different and felt to be so. Another pair is latter:ladder. But my impression is that increasingly Americans not only make no difference, but don't even claim to make one. Up to about 1960 many Americans may have made no difference, but they could make one if they tried, and they would certainly claim to do so if they were naive and phonetically untrained; whereas now some people don't even claim to make a difference. In 1961 Webster's Third International was the first American dictionary to transcribe these words with /d/, and came into considerable criticism for this: it was said to be slovenly speech which shouldn't appear in the dictionary. But I think that criticism is passing now. I have even encountered reverse spellings: I read an American novel in which somebody gave an 'involuntary shutter', shutter:shudder for the author clearly not being distinct. So maybe Americans will have to learn by rote which words are written with <t> and which with <d>. That would accord with their prejudices in many cases anyhow, so it's not yet a problem, but it may be so in the next century, particularly if this sound-change spreads, as seems likely, to all other accents of English. Already it occurs in Australia, South Africa and England, even in RP. RP speakers talking fast will often use expressions like /ged of/, flapping or tapping and voicing the <t>. The <t, d> contrast may still survive, but its disappearance is the logical next step, now we've started on the same path as the Americans. Americans keep the <t, d> distinction in ten:den, even if they lose it in atom:Adam. Word-finally they keep wait:wade apart, but the <-ing> form is ambiguous: 'they were /weidiŋ/ in the river' - did they wait or did they wade? We don't know in American speech, but one can always determine from the base form of the word whether it should be spelt with <t> or <d>. Only a few words like atom lack a base form for reference, and even there the adjective atomic reminds Americans to spell with <t>. I don't see a major problem. A similar neutralization occurs after /s/ in words like spot, stop, Scot. Here the voiced:unvoiced contrast, as between <p, t, k> and <b, d, g>, is neutralized. For spelling this is no great problem: we just have to decide arbitrarily to write them all either with <p, t, k> or with <b, d, g>.


What I hope I have done is to highlight the dangers of parochialism in designing a reformed orthography for English, of being unaware of the varying patterns of contrast in different accents. But even with this awareness, it is impossible to satisfy all of the speakers all of the time; the best that can be hoped is that a proposed reform will satisfy most of the speakers most of the time.


11.1 Jamaica.
The linguistic situation in Jamaica is fairly homogenous, though there is sharp social distinction, just as there is in London. At the top of the scale is the local variety of standard English, quite different from RP, but having the same status as say General American. The worry about taking that standard as the basis for spelling is that it will not help poor children struggling to learn reading and writing in school.

11.2 Historical /a/ to /e/ change?
There is no evidence to suggest that the spelling vs. pronunciation discrepancy observed in any, many is linked to the special case of catch pronounced /ketʃ/, nor that there is a general law describing any such regular sound-shift, in fact there are many other frequently used words that have not undergone such a change. Catch is a well-known special case, where the preceding velar may play a part, just as in parts of the west of Scotland and Northern Ireland they don't distinguish ketch:catch, kettle:cattle. Each word really has its own history here, and you can't set up general principles unless they applied pretty widely.

11.3 Functional load of <r>.
A major obstacle to the omission of <r> is that it affects preceding vowels, even when it is not itself pronounced. The pattern day:dare for example has quite a high functional load. Historically it was the same vowel, with or without <r>, as it still is in Scottish English, /de:, de:r/. RP has lost the <r> but very sensibly made a vowel difference beforehand, so we still keep the contrast. Similarly between cat:cart the contrast is a vowel contrast, and if we omitted <r>, we should suffer quite a high loss of functional load. In any case, we should remember that world-wide the speakers with rhotic accents which pronounce this /r/ are probably in a majority. So the <r> has an effect for all speakers (except possibly the Africans and the Trinidadians): one can usually say, either the <r> changes the sound of the preceding vowel, or it is itself pronounced. Furthermore, speakers of non-rhotic accents will frequently sound the <r> if there is a following vowel, even if not otherwise; and even in words where such speakers make no distinction in vowel-quality before <r>, as in the pairs father:farther, calve:carve, rhotic speakers will pronounce the <r> where it is present.

11.4 Changes in RP.
As well as losing its status in Britain in recent years, RP has also changed, as one immediately recognizes if one listens to pre-war newsreels. However there is a problem of definition here. What is RP? Is it forever that particular form that was codified by Daniel Jones? If so, it has disappeared, it is something we associate with the 1930s. Or has it like every other language, and every other way of pronouncing a language, undergone changes? Maybe to define RP today, we merely need to examine the present usage of upper-class English people, and that will give us present-day RP. That seems a more sensible way of defining it. But RP has unquestionably undergone various changes.

11.5 Dictionaries.
A number of dictionaries give phonemic transcriptions. Daniel Jones of course is very thorough and exhaustive in his treatment of derivatives and so on, which general dictionaries tend not to show. Collins English Dictionary has IPA transcriptions. In the past year or so two cheap paper-back dictionaries with IPA notation have appeared: the Penguin Dictionary 85-86, and the Oxford Dictionary of Current English, which have phonetic transcriptions. And there are also dictionaries that have careful attempts to show pronunciation. Let me refer to the dictionary I worked on: the Readers Digest Great Illustrated Dictionary. This is the first dictionary that explicitly in its phonetic notation tries to cater for other accents than RP: for example I made an attempt to reflect the Scottish <er, ur> contrast in the notation we used, and hoarse:horse, so you might findit worth consulting. I am myself working on a new pronouncing dictionary which I hope will displace Jones-Gimson as the standard, but that remains to be seen.

11.6 Needs of teachers and pupils.
I agree that spelling-reform proposals so far tend to have been conceived from the top down, rather than from the point of view of someone trying to learn to read and write, and that we need to look for a common denominator among the accents, which will serve as a standard for tying the lexicon to the spelling system; for this the starting point can't be RP or any dialect: the written word needs to represent a standard spelling-pronunciation in a logical, simplified way. I have been describing the difficulties of establishing such a common denominator, especially in respect of the chains of word-sets.

11.7 Types of orthography.
The alphabetic principle seems a very sound one that has proved successful in many parts of the world, though there are of course other systems like Chinese and the Japanese and Hindi syllabaries; Vietnamese is an example of a far eastern language written alphabetically. Whether the Roman alphabet is necessarily the best for a reformed English orthography is another question; it has the advantage that it is widely used and familiar, but the Shaw alphabet was one attempt to get away from it. Word-boundaries are not normally a problem in writing a language; even illiterate peoples, like Amazonian Indians, are conscious of them and a computer programmed to detect long plosives can register the difference between <gg> as spoken in bigger and in big girl; but there may be curious parts of a language where word-divisions are not so self-evident as in English compound nouns, or French pronouns with verbs.

11.8 Pronunciation standards?
The difficulty with determining spelling by reference to some sort of standard pronunciation lies in the choice of accent, if it isn't to be RP or General American. Foreign learners regularly learn either RP or American English. There was an experiment in Holland to teach Scottish English, which has certain very definite phonetic advantages, but it foundered essentially on social grounds: people felt they were learning something that wasn't genuine English - it was ultimately a question of the metropolis versus the provinces, most unfair, but that is how life is, and the pupils would have nothing else. As to whether one could reform pronunciation and impose, say, a mid-Atlantic accent there are enormous problems, such as that no one would natively speak it and it would have to be learnt as a kind of foreign accent. In the 3rd edition of his Introduction to the Pronunciation of English Gimson added a Chapter 12, in which he made a proposal of this kind. In reviewing the book, I pointed out that the only people who come anywhere near to pronouncing English in the way he suggested were the Barbadians, who have historical /r/, pronounce face with /e:/ rather than /ei/ (this being in a sense easier than a diphthong), and have /o:/ in coat and court, in fact their pronunciation corresponds in many respects to the existing spelling. The trouble is that the quarter of a million Barbadians are not held in particularly high prestige throughout the English-speaking world, which would be necessary if we were thinking of adopting their pronunciation as standard. The prestige of the speakers is the reason why RP has played the role that it has: as long as the products of English public schools set the tone, their pronunciation was regarded as the ideal. It is because social attitudes have changed that attitudes towards RP have changed too: accents deriving from a public school education may now be regarded as pompous and ridiculously posh, and so no longer any kind of ideal. It used to be widely thought that people with non-standard accents were thereby prevented from discussing any technical, specialist subject in a sophisticated way, which is quite wrong: you can discuss anything in any language, indeed English itself has ransacked the languages of classical antiquity for its scientific terminology, and any language or dialect can do the same. Dialectologists like Ellis back in the nineteenth century had no such prejudices about accents, being fired with an interest not in something thought to be wrong and regrettable, but in something valid and worth studying for its own sake. Bidialectalism, the ability which some people have to speak two different accents of English, does have interesting implications for this question of a standard spelling-pronunciation.

11.9 <z> as standard inflection?
A difficulty with adopting <z> in place of <s> as the standard inflection is that it looks strange. Nevertheless, on the whole it might be best to write both cats and dogs with <z>, as katz, dogz. One reason for doing so is that presumably we are going to write words like hence with <s>, and would need to keep the difference between hence:hens, so the latter word would have to become henz. An alternative would be to write hence as hents, because that is how many people pronounce it. Writing both words with <s> is going to produce homograph clashes between words that are pronounced differently, whereas <z> will not produce any such unfortunate consequences.

11.10 General American.
The nearest we come to an American standard is what is known as General American, sometimes called network English. It is a kind of American accent that is not noticeably eastern (e.g. Boston and New York) and not noticeably southern. It covers quite a range of different phonetic possibilities, but it is what is spoken westwards from Philadelphia or Baltimore right across to California. Webster's Ninth gives it very nicely.

11.11 Foreign borrowings.
What account should spelling reform take of the spelling systems of other languages which impinge on English whenever a word is borrowed, and should perhaps some kind of internationally standard values of letters (such as in the IPA) be adopted in a reformed English orthography to minimize this problem of compatibility? There are essentially two approaches to the spelling of foreign words: one is to retain the orthography of the foreign language, and the other is to adapt it. Swedish and Welsh for example very readily change the spelling of foreign words in accordance with their own conventions. The word emotion in Welsh is written emosiwn, which is regular Welsh spelling for the way they pronounce the word; it would not occur to the Welsh to keep the English spelling. Similarly Swedish spelling for restaurant is restorång with no attempt to reflect the French. The word bureau in German reflects the pronunciation as Büro. This I think is the right approach, because otherwise you have to learn the reading rules for the source language. At present English adopts the other approach however, and normally retains the spelling of the source language. The broad representation of IPA is phonemic and therefore language-specific, so would not seem to lend itself to this purpose. To take the word science as an example: it has international currency, but is variously pronounced, indeed it is even variously written - the Spanish spell it ciencia without <s>. More generally, I think that if one were designing an English orthography from scratch, one would obviously prefer to write team as <tiim>, tame as <teim>, or the like, and so on: the only argument in favour of <ee> in team is the transitional one of familiarity to those brought up on t.o.

11.12 Basic English.
I believe Basic English is a misguided system which reduces the vocabulary but has nothing to do with spelling. The British government bought the copyright of it for £30,000 in 1946, but the British Council didn't feel it could accept it for teaching English as a foreign language, and the idea was dropped.

11.13 Pronunciation in connected speech.
In connected speech, as opposed to the citation forms of words, we observe competing weak and strong forms. We say saint as /seint/, but in St Pauls we say /sənt/; we say I can but I cn do it. There are a number of words in English which are well-known for having two, or sometimes more than two, different pronunciations, and the general feeling is that we should base the spelling on the strong form and, from, was and not on the weak form nd, frm, wz. In connected speech one can then apply a simple rule of obscuring certain vowel-contrasts. Similarly with elision: there are certain sounds we tend to leave out in connected speech, thus we say stand, but in stand back we don't pronounce the <d>. Similarly we typically say wind with /d/, but we pronounce windmill in connected speech without /d/. I think everybody would agree that we should spell them with <d>, which reflects the careful or slow form, though we all understand that when we are speaking faster, in connected speech, we can omit them, just as we can omit the vowels <o, a> in temporary. Likewise we assimilate in connected speech: we say ten with final /n/, but in ten minutes we pronounce it with a final /m/, and we tend to say umpopular rather than unpopular. This is a well-known and well-described process of assimilation of place of articulation, and we can ignore it because we can always recover the unassimilated form: we know that tem is really the word ten, and um - is really the prefix un-. There are certain illogicalities of course, in that we change the spelling in impossible (compare indifferent), although we don't change it in unpopular (compare undecided), though this may or may not correspond to a difference in pronunciation. It is true that we don't have to assimilate in unpopular, whereas we would never try to say /in/ with possible; but it is nevertheless an oddity about the present system. Then there is the matter of linking /r/ we referred to earlier: non-rhotic accents pronounce no /r/ at the end of father by itself, but in the phrase father and mother, they do pronounce /r/; this however leads to the pronunciation of intrusive /r/, as if we were to write Africar and Asia.

11.14 Syllabic consonants.
Cut Spelling uses the concept of post-accentual schwa in words like station, seven, although many people do not pronounce a schwa in that position, going straight to the following consonant instead. This doesn't matter, because the presence or absence of schwa is a realizational question, the option of saying schwa plus one of the consonants concerned, or that consonant on its own with syllabic value. Which choice is made is irrelevant. A complication however arises from the tendency to lose the syllabicity of the final consonant in certain contexts, such as with following <-ing>. With the verb struggle, for instance, the schwa preceding the <l> (or alternatively the syllabicity of the <l> itself) is lost in the form struggling. The word tunneling on the other hand is subtly different because the schwa (or the syllabicity of the <l>) in tunnel is not lost in the <-ing > form. Or compare the <bl> sequence in bleed, table: the <l> in table is syllabic, so the word has two syllables, whereas in bleed it is not syllabic, and the word has only one syllable; but if you say tabling a motion, you lose the syllabicity, and tabling still has only two syllables, rather than three as one might expect.

11.15 A different history of English?
If English had never been written or used for literary purposes until around, say, 1800, then it might have resembled Finnish or Czech or the various kinds of Yugoslav language. If then a great literary figure had arisen, Charles Dickens perhaps, and he had designed his own logical orthography reflecting essentially his own pronunciation, then the power of his example might have established his orthography, and we might feel that it was the only correct form of English, and that we must all imitate it. But further than that: a logical orthography acquires a power of its own, and people feel they must make their pronunciation conform to it; thus standard Finnish or Serbo-Croat is Finnish or Serbo-Croat that is pronounced in accordance with thespelling. Standard German is in many respects a kind of artificial creation, but it had the prestige of Luther behind it, and although there is some variability in the pronunciation, basically it has got to conform to the spelling which is laid down and fixed. In English we have got a standard spelling system, but because it is irregular, there has been no such close association between spelling and pronunciation which could have given a standard for pronunciation. Instead, we have seen a number of important sound changes over the past 500 years which have had very far-reaching effects. If Dickens had been the one to codify English spelling, he would probably not have shown historical <r> at all, in which case we would perhaps have adopted the attitude that Scots and Americans and so on have a provincial pronunciation which must be disregarded and treated as wrong. It is an interesting question whether English spelling could be taught by reference to a standard spelling pronunciation that was in fact independent of any particular accent perhaps an archaizing pronunciation that would distinguish calve:carve and aim to rationalize traditional orthography, rather than represent any one accent.

11.16 The case of Spanish.
Spanish has at least two pronunciation standards which are reflected in the different treatment of the letter <c> before <e, i> and of <z>. In metropolitan Spain the pronunciation /koθer/ for the verb cocer (to cook) is regarded as correct but in South America they say /koser/. On both sides of the Atlantic the other pronunciation exists as a minority one, which is generally regarded as provincial and wrong. There are also many social differences, and in general the upper class pronunciation corresponds to the spelling (and is therefore held in prestige), while the lower class pronunciation can be condemned for leaving out letters that ought to be there. Thus lower class pronunciation tends to omit final <d> in ciudad (city) - which is indeed also how it is pronounced in a lot of South America - and the ending <-ado> would be pronounced /ao/. But these omissions are regarded as wrong: because there is a <d> in the spelling, it is felt it ought to be pronounced. The orthography is therefore a powerful unifying force, but regrettably it doesn't remove the social factors which result in many people's pronunciation being condemned. Spanish spelling may be much more regular than English, but the sound-symbol correspondence isn't one for one, and people do complain about it. There is ambiguity for instance between <ll, y> for the sound /j/, the words rallo (meaning file, rasp) and rayo (meaning ray, beam) being homophones, and there are strange rules for deciding between <qu, c>. Nevertheless, anyone who grows up in a country that has a standard pronunciation generally conforming to the spelling, and learns it right through school, will be able to spell on the whole without much problem.

11.17 Minimizing visual change.
The question is asked whether a spelling reform might mean that those people who can already read and write would have to give up that advantage and learn something new, and whether for that reason the fewer deviations there were from the familiar spelling, the more acceptable the reform would probably be. This consideration is part of the rationale behind Cut Spelling.

11.18 Speech recognition by machine.
The present state of the art is that we can get machines to recognize the speech of one speaker and a limited set of vocabulary, and if you train the machine to recognize one speaker's contrast between double and single consonants (the example big girl comes to mind), it can make the distinction. Anything the human ear can distinguish, in principle the machine can distinguish too. The problem is to move from one speaker with a limited vocabulary to an infinite range of speakers and an unlimited vocabulary. It is not a matter of recognizing phonemes, but of parameters: the machine will need to respond to frequencies in the sound-wave at particular intensities with particular durations. However there are tremendous difficulties. For example imagine you have a printer attached to your word-processor, and you want to tell it start printing; to interpret the phrase correctly, the machine would have to know whether you were English or American, because the English start printing could be phonetically identical with the American stop printing.

Back to the top.