The Undesirability of length marks in EFL phonemic transcription (1975)

(This is a very slight adaptation of the article of this title which appeared in the Journal of the International Phonetic Association 1975 pp 64-71)

1. In 'Another note on RP notation' (1974) Gordon Walsh of the Longman Materials Development Unit made some comments on the 1972 JIPA article of mine 'The notation of the British English segmental phonemes' in which I had offered an explanation of the symbol choices embodied in my Concise Pronouncing Dictionary (1972) and also used by me in the Oxford Advanced Learner's Dictionary of Current English (1974). Apart from certain apparent minor disagreements – for example I consider that use of 'simplified' notations has certain disadvantages in the teaching of the theory of the English segments and that 'comparative' notations are rather unsuitable for use in dictionaries – our main disagreement was clearly over his 'strong plea ... for the retention of the length mark' in transcriptions. He based his plea on three considerations. One of these, I expected he would agree, was essentially trivial. It was the desirability of avoiding occasional potential confusion between the values of three symbols (i, u, ɔ) which indicate different phonemes in the CPD and [then current] EPD types of transcription. The great public for lexical transcriptions, users of reference books like the Advanced Learner's Dictionary, have immediately accessible end-papers with check-lists on the values of all phonetic symbols. The relatively tiny numbers of students of the theoretical subject of English phonetics will no doubt be considering differences between transcriptional systems as a topic of study and thus become fully aware of such problems. If on other grounds it can be agreed that the colons were undesirable it could only be seen as short-sightedly timid to abandon long-term improvement for the sake of avoiding a slight short-term inconvenience to very few people. Walsh's other 'practical' reason for urging the retention of colons is that it is 'often difficult, at first glance, to distinguish between the printed forms /i/ and /ɪ/, or – to a lesser extent – between /u/ and /ʊ/'. I find this not a very urgent argument. That /iː/ is more legible than /i/ is of course undeniable but the essential question is whether there is any problem in distinguishing /i/ from /ɪ/ etc, and I should say there is in fact very little indeed for the user who knows in advance that he has to take account of the distinction in the context of phonetic transcription.

2. The real problem involved I can only accept as being the 'phonetic' one. This he identifies by saying that the five monophthongal pairs / i-ɪ / etc 'differ in both quantity and quality, and for EFL purposes it is not sufficient to make one feature "implicit" in another'. I considered this rather an over-simplification of the facts and moreover in the EFL context a relatively irrelevant one. The question of English vowel lengths is undoubtedly complex. Many authorities have contributed to the discussion of the topic, some like the pioneers Sweet, Grandgent and H. J. Uldall offering only their auditory impressions, while others – among them Meyer, Tuttle, Heffner, Parmenter & Treviño, Rositske, Jones, Wiik, Chen and Cochrane – have made extensive instrumental measurements. There are many problems involved in the interpretation of the data provided by these writers, but it is clear at least that for most forms of educated English there is a marked difference of lengths between / ɔ / and / ɒ / before lenis-category stops and fricatives uttered in pre-pausal tonically-stressed syllables. There are also in varying degrees confirmations of traces of albeit weaker length contrasts in syllables closed by other consonants and with less than full stress and/or not in word-final position. The other main point which is very clearly brought out by the instrumental data is that the intra-phonemic length contrasts between pre-fortis and pre-lenis vowels (as in pairs like port and poured ), which are ignored by blanket-colon transcriptions, are at least as great as the traditional-pairings contrasts (as between eg port and pot). With the four pairs of traditionally coupled vowel phonemes other than /ɔ - ɒ/ there are complicating factors to be noticed. In General British English the phonemes / i / and / u / are characteristically diphthongal so that the CPD use of these symbols for them is as much as anything a typographical economy. Comparison of the lengths of / ɜ / and / ə / is difficult because of the very low incidence of the latter in GB in stressed syllables: no word exists in which tonic / ə / has no alternative and only one (the word threepence which from 1971 underwent a very sharp drop in incidence) has emerged from my surveys as predominantly spoken with stressed / ə / by GB speakers. In the case of / ɑ / in relation to / ӕ / the traditional pairing is highly questionable. The Gimson (1945-49) investigation showed that nine out of ten of his twenty-six London-area informants identified the shortened vowel of part with / ʌ /. English spoken with a slightly raised central open vowel long for / ɑ / and short for / ʌ / would sound perfectly authentic. With a rather centralised open back vowel long and short it would sound perhaps somewhat socially conspicuous but again quite authentic. Neither of these types seems to be common but speakers no doubt do exist who exhibit them. If there were an expression *huff-past which could be used in contexts where it interchanged with half-past the short vowels of *huff and half would often be practically indistinguishable for most GB speakers. The Gimson 1962 decision to continue representing / ɑ / with colon but / ӕ / without seems to have been a traditionalism if one agrees that the contexts in which 'the length component now a feature of RP /ӕ/' (§ 7.12 5) appears are more or less those in which / ɑ / etc are not shortened. Because symbolisations such as / iː / and / ɪ / were quite familiar from the joint and individual works of Armstrong and Ward before Gimson 1962 it is often not realized that that book struck out in a very new direction in employing colons as components of composite phoneme symbols. The only precedent for this indiscriminate 'blanket' as opposed to allophonic use of colons lay in one untidy corner of the EPD transcription. This was where / ӕ / stands in contrast with / ɑː / involving a difference of vowel letter as well as length mark unlike the other four pairs we are considering. Incidentally one of a number of mystifying residues of Jones's occasionally perhaps slightly confused early thinking about phonemics seems to be his persistence in showing on the badly cluttered EPD frontispiece diagram the position of the / ɑ / phoneme without any colon [unlike the representations of all the other four which included the colons found on all five of the dictionary text items] in all twelve editions of the EPD from 1917 to 1963 ie those for which he was responsible. Contrast the revised diagram of Jones-Gimson 1967 which showed colons on all five.

3. Not only would a blanket use of length colons, if carried through, more appropriately contain the symbol / ӕː / than / ӕ /, but if one were to continue the logic of this approach ad absurdum then the theoretical arguments could be said also to indicate the desirability of length colons incorporated in the diphthongs giving /pleːɪ, rəːʊ, kɑːʊ/ etc for play, row, cow etc. Cf Gimson 1962 §704 (e). It can even be argued that relative to other languages English / m, n, ŋ / and / 1 / are long except in such situations as the vocalic phonemes also are shortened. So why not a blanket-colon transcription with / mː, nː, ŋː / and / lː /? The Gimson 1962 decision ''to indicate especially the qualitative opposition, at the same time noting quantity by the length mark'' for vowels is one of essentially the same order.

4. It seems clear that that decision was based on several considerations. One was no doubt the highly justifiable judgement that the types and degrees of allophonic vowel representation in the Armstrong and Ward books were too complicated to be worth following. For example, they may reflect the degrees of grammatical cohesion between groups of words within word-sequences to an extent that can produce large numbers of the doubtful cases which are a great problem of allophonic transcription. The strictly phonemic system – being the most economical – is the best type for most purposes. Another consideration, presumably, was a reluctance to move too far away from the EPD in notation.

5. I suspect that there are more disadvantages to colon identification of phonemes than are realized. I have a strong inclination to believe that the frequent ascriptions of weakforms [with /ɪ/ and /ʊ/] to words like me, he, we, she, the, be and you, who, to and do, is grounded on an at least subconscious resistance to transcribing some of the most frequently heard shortest vowel sounds in the language with symbols containing length marks like / iː / and / uː /. The Gimson 1962 presentations (§ l0.04) such as "she /ʃɪ/ [ʃi]" strike me as indefensible. Jones's different treatment of these words (Outline Chap. XVI) is also unconvincing. Likewise the apparent reluctance of British phoneticians to acknowledge the fact that – whatever criteria may exclude them from the various interpretations put on 'RP' – the two vowel phonemes of a word like easy can be said to be the same for very large numbers of UK speakers with perfectly unexceptionable non-regional accents (a considerable proportion of my 'GB' speakers). Here I think the regional-low-prestige-accent associations of allophones of / i(ː) / of the types heard in stressed syllables are inhibiting arrival at the most appropriate analysis.

6. However, the crucial point in an EFL context in deciding whether EFL texts should contain the blanket colons or not is the question 'What effect do they produce?'. The EPD colons are not just blanket length marks, they are indispensable clues to vowel quality, but in the Gimson 1962 notation colons act as built-in reminders (especially for the non-mother-tongue user?) to make a long vowel. It so happens that in over twenty years of experience of teaching speakers of a quite wide variety of languages, the writer has very rarely felt the need to criticize pupils' vowels because they were too short. In the Daniel Jones Outline, by far the most exhaustive EFL phonetics manual, only French speakers' shortening of / i / and / u / is commented upon. The only others I can recall offhand as requiring such comment have been some Dutch and Chinese or related-language speakers. (One sometimes gets a more acceptable quality by inviting a pupil to lengthen a vowel but as an indirect way of getting him to make adjustments in quality which need not have been acompanied by a length increase.)

7. On the other hand, I find myself constantly asking all nationalities to make the long vowels shorter in syllables closed by one of the sharp ('fortis') consonants /p, t, k, ʧ, f, θ, s, ʃ / or in syllables preceded or followed by others which are unstressed and with which they are rhythmically in close cohesion. In all cases such as these the presence of a colon length mark is not merely gratuitous but potentially misleading. Jones commented quite emphatically in more than one place on how Germans 'almost invariably make the vowels ... far too long in such words as park, use (n), fruit' etc. He also mentioned that it was sometimes a problem for French speakers. I cannot recall a nationality to which it does not apply, especially when the demands of expressive and emotive speech highlight it. Lengthenings of pre-fortis and pre-enclitic syllables constantly create impressions of unintended word-boundaries making gruesome sound disagreeably like grew some, fleecy like fleas see, seeking like sea-king, parking like Pa King, sheets like she eats, we're renewing like weary new wing etc. (Cf Windsor Lewis 1969 pp. 50-51, Gimson 1962 §11.12 and Jones 1956 Chapter XXXII.)

8. This problem is important enough to be one of the only two ''foreign readers’ '', difficulties (along with the 'qualitative distinction between / iː /- / i / and / uː /-/ u /') mentioned in the Jones-Gimson 1967 Editor's Preface (§7). It was the one point made in Armstrong 1923 about vowel length when she referred to 'foreign students, who do not understand that our so-called long vowels when unstressed or when followed by voiceless sounds may be half-long or quite short'. Abercrombie 1964 remarked ''I have often found it worth while drawing the attention of foreign learners of English to ... the common mispronunciation of eg ceasing ... often due to no more than taking "long" as opposed to "short" vowel too literally (a misunderstanding which may be reinforced by transcription).'' [my italics]. In the first edition of his Outline of English Phonetics in 1918 Jones himself remarked that ''the custom of regarding certain vowels as long and certain others as short is, to say the least of it, unsatisfactory. The length of the long vowels is very variable, and depends on a variety of circumstances; the so-called 'short' vowels on the other hand are sometimes quite long .... It is much to be desired that all writers on English phonetics should come to an agreement to adopt a system of transcription for English independent of length marks''.

9. His later volte-face on transcription he attributed, in the 1956 Appendix A to the eighth edition of the Outline, to the fact that a ''two years' trial of narrow transcription with foreign learners did not by any means give the favourable results I had looked for. For instance, one would expect the use of a special letter such as ɪ ... to help French pupils to remember that the sound it represents is a difficult one for them and that it differs considerably in quality both from French i and from English long iː''. On the contrary it is surprising to me that he could ever have been so sanguine. I make no such claims for colon-free notations, but I suggest to Mr Walsh and others who may be inclined to espouse colons in EFL pedagogy that, it may well be that if we can once rid ourselves of the colons, and preferably also persuade teachers to be more guarded about talking about long and short vowels, we can help many learners to a more satisfactory pronunciation. Jones (1956) with fairness acknowledged at §43 that some teachers of phonetics used colon-free transcriptions ''with foreign learners of English, and obtained good results. Some indeed became enthusiastic for it – not sharing my view as to the advisability of symbol economy. There still are today [1955] many teachers who advocate the use of this narrow transcription in the teaching of English pronunciation to foreign people''.

10. Before finally leaving the question of the colon in EFL transcriptions I should like to make one point clear. Though I have fears that the use of colon length-marks fosters misconceptions about vowel length, I consider the Daniel Jones 'Simplified' transcription – as used eg by Kingdon in his revision of Palmer's Grammar of Spoken English (1969) and his pronunciations to the West International Reader's Dictionary (1965) – in every other respect an entirely admirable transcription and, what is more, definitely more desirable as an EFL lexical transcription than the Gimson 1962 set of symbols. My experience of the non-specialist learner convinces me that, unless forced not to, he will tend to ignore, as if capricious negligible minor typographical variants, the alternation of i and ɪ and take full note only of the length-marking colons as significantly distinguishing the phoneme symbolisations, thus rendering their application for him purely a superfluous exercise on the part of the publisher who goes to the expense of employing them.

11. Regarding Walsh's comment on the CPD use of /ɑɪ/ for Gimson 1962 /aɪ/ that 'few phoneticians' will feel happy at CPD's implication of a back-like quality in /aɪ/, I would point out that the CPD notation was not, I'm afraid, dedicated to the satisfaction of the phonetic specialist. That the first elements of the CPD diphthongs /ɑɪ/ and /ɑʊ/ are not the same value as either /ӕ/ or /ɑ/ is less than fully true as regards the latter. On the contrary I should regard a back-vowel quality not much behind central and fully or rather less open an ideal model for the learner of British English for both his /ɑ/ and his /ɑʊ/ phoneme's beginning. As for /ɑɪ/, undoubtedly the average value in GB is front of central so that the symbol element [ɑ] is being used in two slightly different values as is, with Mr Walsh's approval I'm gratified to note, e in /eɪ/ and /eə/. The other alternative which temptingly presented itself was to use /ʌɪ/ for /ɑɪ/ but, among other considerations, its unfamiliarity made me reluctant to employ it. After all, I offer no objection to any printer's substituting /a, aɪ, aʊ/ for /ɑ, ɑɪ, ɑʊ/: my essential wish was to operate with a minimum of important letter shape contrasts which the reader is required to take account of.

12. As to the matter of the so-called diphthongs, the Jones-Gimson 1967 Editor's Introduction indicated that we were to prepare ourselves to see in the next edition of the EPD the first pronunciation of such words as tire and power 'shown with a diphthong or monophthong rather than with a sequence' of phonemes [a threat not carried out: JWL 2007]. Although there is a very large overlap between Jonesian RP, Gimsonian RP and the GB of my CPD and the ALD, in matters of estimating frequencies of particular features I have chosen to be guided chiefly by my observations of the usages of a hundred or so British national newsreaders (and ex-newsreaders) in coming to decisions in matters of statistical doubtfulness that relate to my recommendations to the EFL user of British English. Those who have different aims and base their descriptions on a much wider range of speakers will naturally be inclined at times to present a different picture. The omission of mention in Windsor Lewis 1972b of the vowel-phoneme sequences of fire, power etc, and their omission from my lists of the phonemes in the CPD and ALD endpaper inventories is because reference to them would amount to dealing with allophonic matters rather than phonemic, which I take to be outside the bounds of such works of reference.


Abercrombie, D. 1964. Syllable Quantity and Enclitics in English In Honour of Daniel Jones, ed. Abercrombie et al. London: Longman

Armstrong, L. E. 1923. An English Phonetic Reader. London: University of London Press

Gimson, A. C. 1962, 1970. An Introduction to the Pronunciation of English London: Edward Arnold

Gimson, A. C. 1945-49. 'Implications of the phonemic/chronemic grouping of English vowels' reprinted in Phonetics in Linguistics ed. W. E. Jones & J. Laver (1973). Longman: London

Hornby, A. S., Cowie, A. P. & Windsor Lewis, J. (1974). The Oxford Advanced Learner's Dictionary of Current English. London: Oxford University Press

Jones, D. 1917- 63. An English Pronouncing Dictionary ('EPD') London: Dent

Jones, D. 1918;1956. An Outline of English Phonetics. Teubner, Leipzig; Heffer, Cambridge

Jones, D. & A. C. Gimson 1961-72. An English Pronouncing Dictionary ('Jones- Gimson'). London: Dent

Windsor Lewis, J. 1969. A Guide to English Pronunciation. Oslo: Scandinavian Universities Press

Windsor Lewis, J. 1972a. A Concise Pronouncing Dictionary of British and American English ('CPD'). London: Oxford University Press

Windsor Lewis, J. 1972b. The Notation of the General British English Segments Journal of the International Phonetic Association Vol.2 No.2 pp 59-66

Palmer, H. E. 1969 A Grammar of Spoken English re-written by Roger Kingdon. Cambridge: Heffer

Walsh, G. 1974. Another note on RP notation Journal of the International Phonetic Association. ("JIPA") Vol.4 No.1 pp 31-36

West, M. 1965. An International Reader's Dictionary (with pronunciations by Roger Kingdon). London:Longman