CTLF Corpus de textes linguistiques fondamentaux • IMPRIMER • RETOUR ÉCRAN
CTLF - Menu général - Textes

Jones, Daniel. An Outline of English Phonetics – T03

Appendix A
Types of Phonetic Transcription

1. It has long been known that different types of phonetic
transcription are needed for different purposes. Henry Sweet
pointed this out in his Handbook of Phonetics (1877), in which he
published, in their original forms, the systems which he called
‘Narrow Romic’ and ‘Broad Romic.’ Narrow Romic was ‘scientific,’
while the various forms of Broad Romic were ‘practical.’

2. In Narrow Romic, Sweet invented means of symbolizing all
the speech-sounds and shades of speech-sounds he could think of.
In its original form it was composed of Roman lower-case letters
supplemented by capitals and italics with special meanings, digraphs,
inverted letters and letters with diacritics attached. (In a later
revised form he improved it by introducing some new letters to
take the place of some of these.) There were also marks denoting
degrees of length and stress and certain intonations. From Narrow
Romic were derived the ‘broad’ or ‘practical’ systems for particular
languages. Each ‘broad’ system was intended to contain only as
many symbols 11 as were necessary to represent the particular
language without ambiguity, and in selecting the symbols, Sweet
took into consideration their familiarity and the convenience of
their designs from the point of view of handwriting and the legibility
of connected texts.

3. To ensure that his ‘broad’ texts should be unambiguous,
Sweet laid down the principle (now known as the ‘phonemic’
principle) that only those distinctions of sound should be symbolized
which are capable of distinguishing one word from another in the
particular language transcribed. 22 To ensure satisfactory letter
331shapes he restricted his letters as far as possible to those of the
Roman lower case (including æ and œ), making as few additions
as possible. 33

Broad and Narrow Transcription

4. The terms ‘broad’ and ‘narrow’ are convenient, and it is
useful to retain them, giving them the same general meanings that
Sweet did. It is necessary, however, in view of modern developments
in the theory of transcription to introduce some additional
terms to identify types of transcription embodying special features
which at first were but vaguely recognized. Some useful terms
have been proposed by David Abercrombie, Head of the Department
of Phonetics in the University of Edinburgh. 44 I use these and
others in what follows.

5. A ‘broad’ transcription may be defined precisely as one which
represents only the phonemes of a language, using for this purpose
the minimum number of letter shapes of simplest Romanic form
(consistently with the avoidance of undesirable digraphs for ‘single
sounds’ 55) together with such prosodic 66 marks as may be necessary
for the avoidance of lexical ambiguity. 77 This kind of transcription
has been called by Abercrombie ‘simple phonemic’ (For ‘simple’
see § 13.)

6. A ‘narrow transcription’ differs from a ‘broad transcription’
in one or both of two ways. (1) A transcription is ‘narrow’ if it
includes special symbols to denote particular allophones (members
of phonemes, § 197). Abercrombie has proposed the term ‘allophonic’
for this style of narrow transcription. Such transcriptions
may also be termed ‘linguistically narrow.’ (2) A transcription is
332also called ‘narrow’ when use is made of ‘exotic’ or inconvenient
letters when it would be possible to transcribe the language unambiguously
with familiar or more convenient ones. Recourse is
had to such special letters chiefly when it is desired to show ‘external
comparisons’ by means of separate symbols, i.e. that a sound of
one language differs from an analogous sound of another language
or from some ‘cardinal’ sound. Abercrombie has called transcriptions
embodying special letters for this purpose ‘comparative
transcriptions,’ and he has pointed out that ‘a comparative transcription
uses symbols some of which, considered in isolation, are
more specific in their reference than those of a simple transcription.’
Such transcriptions may also be termed ‘typographically narrow.’

7. Recourse may also be had to special letters in order to show
‘internal’ comparisons between sounds in a single language, e.g. to
represent particular allophones or to show that the beginning part
of a diphthong is not identical in sound with any ‘pure vowel’ of
the language transcribed.

8. A transcription must also be called ‘narrow’ if it gives indications
of non-significant degrees of length or of any other ‘prosodic’
distinctions which do not serve to differentiate words.

9. Examples of ‘allophonic’ transcription would be the use of ɫ
to denote the Southern English ‘dark l,’ or to denote the dental t
in eighth eitθ, or (§ 176) to denote the French ‘devoiced l.’
Examples of ‘comparative’ transcription would be the use of ɹ in
place of r in transcriptions of English in order to remind readers
that the sound is not a rolled one, or of ʀ or ʁ (§§ 746, 762, 763)
in transcriptions of French to remind English learners not to use
an English r. An example of a transcription which is neither
‘allophonic’ nor ‘comparative’ (as defined above), but which calls
attention to an ‘internal’ comparison, is the use of a in writing
the English diphthong ai 88 in order to show that the beginning of
this diphthong differs both from æ and from ɑː.

10. The extra symbols needed for narrowing a transcription can
always be dispensed with by assuming conventions. When a
333transcription is allophonic, the conventions to be stated are the
phonetic environments determining the occurrences of each allophone.
Thus in transcribing Southern English the narrow symbol ɫ
can be dispensed with and replaced by l, if the conditions under
which ‘dark l’ is used are described once for all, as they are in
§ 659. When, on the other hand, a transcription is comparative,
a single symbol can be employed to denote analogous (or occasionally
non-analogous) sounds in two or more languages by specifying once
for all the value to be attached to it in each. Thus the letter r
can be, and generally is, used in transcriptions of English, French
and Italian with conventions as to its values in each of these

11. An allophonic transcription may be comparative or non-comparative.
For instance, it is allophonic and comparative to
denote the fricative and napped varieties of Southern English r
by ɹ (§ 746) and ɾ (§§ 746, 750, 753, 754) respectively, since ɹ is
fairly ‘specific in its reference’ and ɾ is still more so. It is, however,
allophonic but not comparative to use o and ou to distinguish
the Southern English monophthongal o-sound in November from
the ordinary diphthongal sound of below, home, etc., as is done in
EPD transcription of English. 99 This representation is not comparative,
since it does not suggest that the o differs from ‘cardinal’ o
or from the o-sounds of French and German or any other language.
The value of o in EPD transcription has to be understood once
for all.

12. Conversely a comparative transcription may be used in a
phonemic manner or in an allophonic manner. For instance, the
comparative letter ɹ may be employed either phonemically to denote
the Southern English r-phoneme, or allophonically to mean fricative
r as distinguished from the flapped sound denoted by ɾ.

Simple and Complex Transcription

13. A form of transcription which comprises only ordinary
Roman letters, or Roman letters supplemented by the smallest
possible number of new letters is called by Abercrombie ‘simple’;
it might also be termed ‘romanically simple’ or ‘conservative’ or
334‘old-lettered’ or ‘typographically broad.’ The ‘simplified transcription’
of Southern English described in §§ 44-49 of this Appendix
is ‘simple’ in this sense. It is likewise ‘simple’ to denote the Spanish
b-phoneme by the letter b, or the Japanese vowel ɯ (§§ 145, 351,
358) by u, or the Polish vowel ɨ (§ 146) by y.

14. A ‘simple’ mode of transcribing on Romanic basis sometimes
involves giving to a Roman letter a value differing greatly from
that commonly associated with it. For instance, it is ‘simple’ to
use c to denote the dental click in Zulu (as is done in the current
orthography of that language); this procedure is justifiable on the
ground that the letter c is not needed for any other purpose in
Zulu. Likewise, Sweet wrote ‘simply’ when in his original Romic
systems he took q to mean ŋ in English and to mean the nasalization
of vowels in French. Such uses of Roman letters may be objected
to on international grounds, but a transcription employing them
in such ways is none the less ‘simple,’ and may be adequate for
the transcriber's purpose.

15. ‘Simple transcriptions’ are generally phonemic, but they
are not necessarily so. It would in rare cases be possible — though
I do not say desirable — to arrange an allophonic transcription on
a ‘simple’ basis. Thus it would be ‘simple’ and allophonic to
write the French fronted k of qui, caisse, etc., with c, while retaining
k in other situations. Likewise it would be ‘simple’ and allophonic
in transcribing Spanish to write b for the plosive b-sound (used
after m) and v for the non-plosive allophone used in other situations.
Similarly, it would be ‘simple’ to denote the two corresponding
members of the Spanish g-phoneme by g and q, or to use x to
denote the Southern English dark l. It is not likely that anyone
would seriously contemplate using q and x in this manner. It
would doubtless be generally agreed that their associations render
them unsuitable letters for these purposes.

16. When a transcriber wishes to write narrowly (whether
allophonically or comparatively), he is however generally obliged
to introduce exotic letters. 1010 A form of transcription introducing
335exotic letters when it would be possible by conventions to avoid
doing so may be termed a ‘complex’ or ‘new-lettered’ one. A
complex transcription is generally comparative; it may be phonemic
or allophonic. A comparative transcription, however, is not
necessarily complex (see § 15).

17. It is complex and phonemic to use ɹ to denote the English
r-phoneme, or to write the Japanese close back vowel with in
or the Polish close central vowel with ɨ. 1111

18. On the other hand, it is complex (as well as comparative)
and allophonic to write the English fricative r with r (or ɹ) and to
introduce the symbol ɾ to represent the flapped allophone (§ 11).
It is likewise complex (as well as comparative) and allophonic to
write the Spanish intervocalic b with the letter β) (§ 692) while
reserving the letter b for the plosive b occurring after m, as has
been done in narrow transcriptions of Spanish.

Uniliteral and Multiliteral Transcriptions

19. It is necessary to draw a distinction between the terms ‘letter’
and ‘symbol.’ Any written sign or sequence of signs or accented
letter used for the representation of a single speech-sound may
be called a ‘symbol.’ 1212 Digraphs such as the ʧ and ai used in
ordinary transcriptions of English, 1313 are therefore single symbols,
although they are each composed of two letters. So also are
sequences like hw (often used as a ‘simple’ substitute for ʍ, § 810)
or ɑŋ, ɛŋ, etc. (which have been suggested for use in place of ɑ̃, ɛ̃,
etc., to represent nasalized vowels).

20. A system of phonetic transcription which employs for a
given language not only a minimum number of symbols, but also
336a minimum number of letters, may be termed a ‘uniliteral’ system.
Uniliteral systems embody the principle that digraphs, if any are
needed, are constructed if possible by putting together letters
which are used independently to denote other sounds of the language.
When a uniliteral system is based on Roman letters, it need not
necessarily be a ‘simple’ system (§ 13); it may contain exotic letters
introduced to call attention to differences between one language
and another. For instance, the ‘simplified transcription’ of English
(§§44-49) would remain uniliteral if all the r's were replaced
by ɹ.

21. Systems which employ more than the minimum number of
letters needed to represent a given language effectively and unambiguously
may be called ‘multiliteral.’ A system is multiliteral
(1) if it comprises any special letters to denote particular allophones,
(2) when a letter not otherwise employed is introduced into a
digraph. Multiliteral systems, if basically Romanic, are generally
‘complex,’ since there are seldom enough letters in the Roman
alphabet to provide a multiliteral transcription. It is, for instance,
multiliteral and complex (but not allophonic) to write the English
diphthongs ai and au with a letter that distinguishes their beginnings
from ɑː, as is done in EPD transcription. Uniliteral transcription
would require that ɑː should be written as , or that the diphthongs
should be written ɑi, ɑu. It would be multiliteral and ‘simple’
(though not advisable on other grounds) to represent by q, while
using the customary ai and au for these diphthongs. It is likewise
multiliteral (though convenient) to introduce the letter ʒ into
transcriptions of Italian, as is done by those who transcribe the
voiced affricate in giorno by ʤ.

22. In EPD transcription the following diphthongs are symbolized
uniliterally: ei, ou, ɔi, , ɔə, . In other words each of
these representations is composed of letters which are used independently
for other purposes. On the other hand ɛə, ai, au are
multiliteral representations, 1414 since the letters ɛ and a are not employed
separately for any other purpose. (A uniliteral representation
337of ɛə would be . The notation ɛə would be uniliteral only if the
letter e were everywhere replaced by ɛ, that is to say if the vowels
in get and day were written with ɛ and ɛi.)

23. Similar considerations apply to the representation of affricate
consonants and some other consonant-sounds such as hw and kw.
Such sounds may be, and sometimes are, represented by single
letters, e.g. c for ʧ, ʍ for hw. It is, however, often convenient
to denote them by digraphs. Such digraphs are commonly designed
on a uniliteral basis, that is to say by combining two letters which
are employed separately for other purposes. But the plan of using
special letters in digraphs (i.e. multiliteral representation) has also
been tried occasionally. 1515

Exclusive and Inclusive Transcription

24. Since no two speakers of a given language pronounce exactly
alike in all respects, anyone seeking to transcribe that language
phonetically has to decide what pronunciation to record. When
the language is his own, the transcriber may follow the safe course
of recording his own way of speaking. Henry Sweet, for instance,
did this. 1616 Some have sought to record a kind of norm, basing
their transcripts on their own speech, but making modifications
wherever they have thought their own pronunciation to be unusual.
Others again have represented a style of speech specially selected
so as to facilitate the task of the language learner; when two ways
of pronouncing are possible, they have chosen for representation
338the one that is easier or more effective from the point of view of
the pupils for whom the transcripts are designed. The ‘simplified’
transcription of English described in §§ 44-49 of this Appendix
makes allowance for considerations of this kind.

25. Whatever the basis of a transcription, it is generally found
that some of the symbols can be interpreted by the reader in more
than one way. The symbols may be held to cover (within limits)
certain deviations from what may be considered as the ‘average’
values for the particular language transcribed — ‘diaphonic’ variants
as they may be called (Chap. XI). It has been pointed out, for
instance (in §§271, 388, also in my Pronunciation of English, 1950
and subsequent editions, §§ 89, 160, 161), that the English sounds
of e and ei both admit of diaphonic variants which would be
recognized as coming within the limits of RP; they vary in quality
from speaker to speaker. In such cases the foreign learner need
not restrict himself rigidly to acquiring one particular shade of
sound; it is possible for him to adopt another shade (within limits)
without rendering his pronunciation un-English.

26. A transcription which makes allowance for more than one
way of pronouncing may be termed an ‘inclusive’ transcription.
One which definitely excludes or provides no means of representing
certain possible ways of pronouncing may be termed an ‘exclusive’
form of transcription. All transcriptions are to a certain extent
both inclusive and exclusive. It is advisable therefore to
confine the use of these terms to cases where the pronunciations
included or excluded are particularly frequent or otherwise

27. A special case of inclusive transcription is one where alternative
ways of pronouncing can be shown by means of the symbols
employed, and are not merely implicit in the transcription. EPD
transcription of English is in some respects adaptable in this way,
and this is to a certain extent an advantage. In particular, the
following alternative pronunciations in Southern English can be
shown by means of it:

(a) the lengthening of the traditionally short sounds of e, a
and u, as in red, bad, run; this pronunciation may be
symbolized by writing reːd, bæːd, rʌːn (§§ 874-878),339

(b) the reduction of aiə to a diphthong or to a monophthong
(distinct from ɑː) (§ 414), 1717

(c) the reduction of eiə and ouə to diphthongs , (distinct
from ɛə, ɔə) (§§ 392a, 403),

(d) the reduction of oui to a diphthong oi (distinct from oi)

(e) the reduction of ou to a monophthongal o-sound (distinct
from ɔ) in positions of very weak stress, as in the first syllables
of November, obey, molest (§403).

Characteristics of EPD Transcription

28. EPD transcription of English is exclusive in so far as it
does not provide representation for the pronunciation of those
Southern English people who lengthen the traditionally short sounds
of i, o and u, as in this, hot, full (§ 877, also The Pronunciation of
, 1950 and subsequent editions, § 429). To render it
inclusive of these variants would involve altering the transcription
by the introduction of three extra symbols such as ɩ, ʋ, ɷ, to
denote the traditionally short i, ɔ and u, and to use them throughout.
Such a modified system would be narrow (‘allochronic’) if
the length-marks are retained, and would therefore be multiliteral
to those who do not lengthen these vowels.

29. EPD transcription of English comprises twenty-two consonant
letters (p b t d k g m n ŋ l r f v θ ð s z ʃ ʒ h j w),
eleven vowel letters (i e ɛ æ a ɑ ɔ o u ʌ ə), the length-mark (ː)
and the marks for primary and secondary stress (ˈ ˌ). It is allophonic
in one respect only, namely in the use of the letter o to
denote the monophthongal o-sound referred to in (e) above. That
sound, when consistently used, 1818 is a member of the ou-phoneme,
and can properly be written with ou with the convention that it
is sounded as the monophthongal allophone in syllables with very
weak stress.

30. The multiliteral features of EPD transcription, namely the
use of ɛ, æ and ɑ, are introduced for comparative purposes, and
340are thus narrow in one sense; but these letters do not represent
allophones (particular members) of any phonemes. £ and a are
included as being suggestive of the cardinal categories to which
the sounds belong — an object which some teachers consider unnecessary. 1919
Incidentally, they provide a means by which the
transcription is made more inclusive than it would be without
them. As to æ, it has been customary to use this letter in order
to call the attention of foreign learners 2020 to the fact that the sound
is an unusual one and likely to give them trouble — again a plan
which some consider unnecessary. The use of this sign has the
advantage of rendering the transcription inclusive of the speech
of the numerous English people who lengthen the vowel.

31. The characteristics of EPD transcription may then be
summarized as follows:

(1) it is phonemic except in one point,

(2) it is multiliteral in that three letters are employed which
could be dispensed with if the transcription were restricted
to the representation of ‘common pronunciation,’ 2121

(3) it is inclusive in that some of the symbols can by convention
be held to represent diaphonic variants, and special arrangements
of the letters can be made which indicate the variant
pronunciations enumerated in § 27 (a)-(d).

Narrow Transcription of Southern English

32. In the practical teaching of ‘common pronunciation’ of
Southern English 2122 I find the traditional system of regarding iː i,
341ɔː ɔ, uː u, əː ə as pairs of corresponding long and short vowels easy
and convenient to work with. This manner of transcribing is
based on the view that in each pair the length constitutes the
fundamental difference, and that the accompanying quality
difference is incidental. It is however possible, at any rate in
the case of the first three pairs, to take the contrary view — that
the quality difference is the fundamental one and that the differences
of length are incidental. To those holding this view the qualities
of each pair of sounds constitute separate phonemes, requiring
therefore separate symbols in transcription, while the vowel lengths,
having no significance from the semantic standpoint, do not need
to be indicated in broad transcriptions. (See my book The Phoneme,
§§ 510-516.)

33. Broad transcription of ‘common pronunciation’ on the
latter supposition would involve the adoption of some such system
of vowel representation as the following (which should be compared
with the chart in § 236). This system is uniliteral (§ 20) and is
phonemic, as long as no indications of length are given ; 2223 it excludes
representation of the monophthongal allophone of ou and the
reduced forms of eiə, ouə and oui. It presupposes a value of
differing in quality from all the values of ə (see § 342, also § 355, etc.).

i ɩ e æ a ʋ o ɷ u ʌ ɜ ə
əɩ oɷ aɩ aɷ oɩ
ɩə eə oə ɷə

34. A similar system of vowel representation also suits well
the pronunciation of those who lengthen the traditionally short
vowels (§§ 874-879). Here again the system would be phonemic,
as long as the lengths are not marked. 2224

35. As far as I know, no author has yet employed a uniliteral
system of the above type for transcribing ‘common pronunciation.’
342Transcriptions involving special symbols to denote the qualities of
the traditionally short vowels have, however, often been used.
But they have been both narrow (in both senses) and multiliteral.
They are of value especially in comparative work, e.g. when a
teacher or author desires to show in his transcripts differences
between different varieties of English pronunciation or between
English and foreign languages.

36. Notable among such systems is one in which the vowels of
RP are denoted by a system of the following type, with length-marking
of i ɑ ɔ u and ɜ, and with o to represent the monophthongal
variant of ou. 2325

i ɩ ɛ æ ɑ ɑ ɔ ɷ u ʌ ɜ ə
eɩ oɷ aɩ aɷ ɔɩ
ɩə ɛə ɔə ɷə

37. This system has generally been called ‘Narrow Transcription.’
It is, however, not merely narrow (i.e. allophonic and comparative)
but also multiliteral and complex. It could be made uniliteral and
simple by substituting e for ɛ and a for ɑ. It would then remain
allophonic (in one respect, namely in its use of o) but would cease
to be comparative. It could be made uniliteral but complex by
substituting ɛ for e and/or ɑ for a.

38. At one time (round about 1918) I made considerable use
of a narrow transcription of this type. My experience with it
originated in the following way. In teaching the pronunciation
of foreign languages to English pupils it has generally been my
custom to get the pupils to make the foreign sounds and their
combinations without teaching them anything about English
sounds. I use my own knowledge of phonetics for the purpose,
and give the pupils only a minimum of theory. I use a phonemic
343or nearly phonemic transcription of the foreign language, and cause
the pupils to associate each sound they learn with the appropriate
symbol of that transcription; I do not as a rule give them phonetic
representations of any English sounds. I still find this to be the
most effective way of teaching the pronunciation of a foreign
language for the average learner, i.e. the learner who does not
desire to become a specialist in phonetics. 2426

39. However, for several years (from about 1916) I experimented
with the plan of showing students phonetic transcriptions illustrating
English words in order to demonstrate such differences between the
foreign languages and English as can be shown by this means.
For this purpose I used a fairly narrow transcription (of the type
described in § 36) of the English words. I did not find it needful
to transcribe any connected English texts, but only isolated words
and occasionally short sentences. In this way I became accustomed
to making a limited use of a narrow transcription for English with
some students.

40. To test further the value of this kind of transcription, I
subsequently prepared some continuous texts in this type of transcription,
and used them both with English students of phonetics
and with foreign students desiring to improve their knowledge of
English. With English students of phonetics the narrow transcription
gave reasonably good results, since with its use various
differences between different types of English could be well demonstrated,
as also could certain differences between English and foreign

41. With foreign learners of English, on the other hand, I did
not find the narrow transcription a success. It did not by any
means give the favourable results I had looked for and which
might be expected on theoretical grounds. For instance, one
would expect the use of a special letter such as ɩ (or ɪ as it was
at that time) 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 the English long . My
344experience was, however, that they persisted in pronouncing it as
French i, and they had to be drilled in the use of the opener sound
just as much as when the broad transcription (iː i) was used. So
after two years' trial of narrow transcription with foreign learners
I abandoned it and reverted to broader forms. I continued,
however, to use narrow transcription for many years with English-speaking
students of phonetics.

42. The first reader to be published employing full narrow transcription
of Southern English was P. W. Drew and C. F. Mackenzie's
Phonetic Reader for Junior Classes (Manchester University Press,
1919). This made a good beginning, but soon proved inadequate
for comprehensive courses of instruction. So I asked my colleague,
Miss L. E. Armstrong, to prepare a larger English Phonetic Reader
containing literary texts, so that there should be plenty of material
to work with. This she readily consented to do, and the work
was published (by the University of London Press) in 1923. I also
started contributing English texts in narrow transcription to Le
Maître Phonétique
when its publication was resumed in 1923, and
I continued to do so for some years. I did this for two reasons:
firstly to make generally known the fact that a narrow method of
transcribing English suitable for exact comparative work was
available within the framework of the IPA recommendations, and
secondly to provide some additional material for experiments in
teaching and research with its aid. Some other good books
employing narrow transcription were published in the course of
the next few years; they included Armstrong and Ward's Handbook
of English Intonation
(Teubner, Leipzig and Heffer, Cambridge,
1926) and the original edition of I. C. Ward's English Phonetics
(Heffer, 1929).

43. During this period a number of teachers of phonetics
expressed themselves as well satisfied with this form of transcription,
especially when working with English-speaking students.
Some also used it 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 to-day (1960) many teachers who advocate the use of this
narrow transcription in the teaching of English pronunciation to
foreign people.345

Simplified Transcription of English

44. About 1930 I began to realize that EPD transcription of
Southern British English, good as it is, is not the simplest possible
transcription that can effectively help the foreign learner towards
ability to pronounce English properly. I endeavoured to look at
the question from the point of view of the very numerous foreign
learners whose sole object is to learn to speak English well, and
who have no need either to become specialized phoneticians or to
concern themselves with more than one variety of English pronunciation.
It became clearer to me than it had been previously
that a transcription for the use of the foreign learner need not
be so much a precise record of the speech of particular English
people as a guide designed to give him a pronunciation recognizable
as ‘good’ English; and that consequently when two ways of pronouncing
are current in Southern England, the transcriber may
quite properly indicate the form likely to prove easier or otherwise
more effective for the pupil. Viewed from this angle EPD transcription
is undoubtedly more ‘inclusive’ than it need be, and
although it is very nearly ‘broad’ (for the pronunciations which
it represents), it is for the average foreign learner unduly ‘multiliteral’
and ‘comparative.’ I then came to the conclusion that a
still simpler system — one in which the number of special letters is
diminished to an irreducible minimum — is what would meet their
needs best. The value of such a system had indeed been demonstrated
long since by Sweet, whose revised Broad Romic as used
in the texts in his Elementarbuch des Gesprochenen Englisch (first
published in 1885) fulfilled nearly all the requisite conditions, and
had long enjoyed much success abroad. 2527346

45. It seemed to me necessary therefore to construct and try
out a system of vowel representation for Southern English which,
while remaining within the framework of the recommendations of
the IPA, should completely fulfil all the conditions necessary to
simplicity. These are (1) that the transcription should be ‘exclusive’
of all varieties of pronunciation needing letters that can be dispensed
with by selecting a special form of Southern pronunciation, (2) that
the selected form of the language should be transcribed broadly
and uniliterally. Further, it was and is in my opinion desirable
that the transcription should be ‘simple’ (§ 13), i.e. that the letters
used should be, as far as practicable, familiar Roman ones (this in
accordance with the provision formulated in The Principles of the
International Phonetic Association
, 1949, §§20, 21).

46. The first of these requirements is satisfied by selecting for
transcription a form of Received British English from which the
following pronunciations are excluded:

(a) lengthening the traditionally short vowels,

(b) reduced pronunciations of eiə, ouə, aiə, auə, oui,

(c) the reduction of ou to a monophthongal o-sound in various
unstressed positions.

The second and third requirements are met by replacing the vowel
letters of EPD transcription as follows:

using a for the æ of EDP transcription
ɑː ″ ″
ɛə ″ ″
oɔ ″ ″
ɔː ″ ″
(or ) for the ɔə of EDP transcription
oi for the ɔi of EDP transcription

47. A chart of the vowel system on the same lines as those of
EPD transcription (§ 236) and ‘narrow transcription’ (§ 36 of this
Appendix) is shown overleaf. 2628347

iː i e a aː o oː u uː ʌ əː ə
ei ou ai au ɔi
iə eə (ɔə) uə

48. Feeling convinced that this system ought to be tried, I
prepared a text in it in 1930 and published it (not without misgivings)
in Le Maître Phonétique, January, 1931, p. 12. Further
specimens by myself and others followed in January, 1932, p. 8,
April, 1932, p. 44, July, 1932, p. 60, January, 1938, p. 10, April,
1938, p. 25, January, 1939, p. 12, April, 1939, p. 32, July, 1939,
p. 53, July, 1940, p. 51, October, 1940, p. 69, and texts in this form
of transcription have been published in most subsequent numbers. 2729
Books using this system began to appear in 1942, starting with N. C.
Scott's English Conversations. 2830 Several others are now available,
the chief ones being P. MacCarthy's English Pronunciation 2831 and
English Pronouncing Vocabulary 2832 and English Conversation Reader, 2933
A. S. Hornby's Oxford Progressive English, 3034 Hornby and Parnwell's
English-Reader's Dictionary, 3035 E. L. Tibbitts' Phonetic Reader for
Foreign Students of English
, 2836 and my book on The Phoneme. 2837
The system has also been used in the periodical English Language
(from 1946 onwards). 3138

49. The simplified transcription of Southern English above
described, being phonemic, uniliteral, simple and very exclusive,
348combines in a remarkable degree neatness of appearance with
effectiveness as an aid to teaching pronunciation. In fact, I do
not think it possible to construct anything simpler which will do
its work adequately. Specialists in phonetics will continue to need
transcriptions of more elaborate types, but those who teach English
to average foreign .earners — learners who wish to pronounce well,
but who have no time or inclination to make a detailed study of
phonetic science — will in my opinion do well to explain the use and
distribution of the English sounds with the aid of this simplified
system of transcription.

Systematic and Impressionistic Transcription

50. The various types of transcription described in the foregoing
paragraphs have one feature in common. They are all designed
for the representation of languages and forms of speech which have
already been analysed phonetically.

51. It is useful to have a general term to express the fact that
a transcription has been, like all these, constructed to suit the
phonetic structure of a particular language. The term ‘systematic’
proposed by Abercrombie suits this purpose well. It is to be
observed that systematic types of transcription always have to be
accompanied by sets of conventions, an understanding of which
is necessary to their correct interpretation.

52. Systematic transcriptions have to be distinguished from
transcriptions made on a general phonetic basis, without reference
to the needs of any particular language. The latter may be
described as ‘non-systematic’ or ‘impressionistic’ (to use another
term proposed by Abercrombie). An example of impressionistic
transcription is the kind of phonetic writing a research worker
has to use when he begins taking down a language new to him
and about which he has no advance information. Such transcriptions
‘are made by drawing on a theoretically unlimited number of
symbols, which are defined with reference to the total range of
human speech sounds…. No conventions accompany them, for
they are made on the same basis for every language.’ 3139349

53. Ear-training exercises (§§ 21-24, Chap. XIII and Appendix B)
have to be written largely impressionistically, except when the
teacher states expressly that an exercise contains the sounds of a
certain language arranged in sequences that are possible in that

11 For the special use of the term ‘symbol’ as distinguished from ‘letter,’
see § 19 of this Appendix.

22 ‘In giving passages of any length in phonetic writing, and especially
in dealing with a limited number of sounds, as in treating of a single language,
it is necessary to have an alphabet which indicates only those broader distinctions
of sound which actually correspond to distinctions of meaning’ (Handbook,
p. 103). ‘We may lay down as a general rule that only those distinctions of
sound require to be symbolized in any one language which are independently
’ (Handbook, p. 104). The italics are Sweet's; by ‘independent’ he
meant ‘not linked with length or stress’.

33 Sweet did not always follow his stated principles with complete consistency,
but this was pardonable enough in those early days and did not
invalidate the principles themselves.

44 See his article Phonetic Transcriptions in Le Maître Phonétique, July-December,

55 Such as the sh (for ʃ) and ao (for ɔː) of Sweet's first version of Broad

66 See footnote 2 to § 1 and footnote 23 to § 39.

77 This concise wording of the definition was suggested to me by J. L. M.
Trim in November, 1954.

88 I.e. the variety of ai which begins with cardinal a (§ 407). There exist
English speakers who start their diphthong with æ and others who start
theirs with ɑ (see The Pronunciation of English, 1950 and subsequent editions,
§§ 175, 177).

99 By EPD transcription is meant the type of transcription of English
used in my English Pronouncing Dictionary (and in this book).

1010 Unless he elects to resort to the unsatisfactory device of using arbitrary
digraphs on a large scale, or unless he adopts the unusual course of assigning
specialized meanings to superfluous Roman letters after the manner suggested
in §§ 14, 15.

1111 As is done in Arend-Choińsky's Polish Phonetic Reader (University of
London Press).

1212 This special meaning of the word ‘symbol’ has often been implied but,
as far as I remember, it was not definitely formulated until R. T. Butlin did
so in an article on the phonetics of Malayalam in the Bulletin of the School of
Oriental Studies
, 1936, p. 437. P. A. D. MacCarthy also drew attention to
it in Appendix C of his English Pronunciation (1944).

1313 It is proper in my opinion to consider affricates and diphthongs as ‘single
sounds.’ Transcribers have at times written them with single letters. This
was done for instance in the Pitman-Ellis phonetic alphabet, which was
used by Alexander Ellis in his Essentials of Phonetics (1848) and by Sir Isaac
Pitman in numerous works.

1414 As long as EPD transcription is used in its full form, i.e. when the
sequences eiə, aiə, auə are written in full. When the reduced form of eiə
is written by the reduced symbol , and when the length-mark is added
to a to show reduced forms of aiə and auə, the representations are uniliteral.

1515 For instance, Professor C. M. Doke of Johannesburg has used the specially
designed letters ť and đ in his representation of the affricates commonly
written ʧ, ʤ. He has written them ť, ʤ, in order to show in his transcripts
that the initial parts of these affricates have different articulations from the
ordinary t and d in such words as ten, dull. This is a multiliteral and complex
way of symbolizing these sounds. It is presumably to be considered, as a
‘narrow’ representation on account of the ‘internal’ comparison involved. I
am inclined not to consider it an allophonic representation, since the beginnings
of ʧ and ʤ have nothing to do with the t and d phonemes. J. L. M. Trim
has, however, suggested that it should be considered allophonic, on the
ground that the beginning of ʧ may occur by assimilation as an allophone
of t when ʧ follows, as in that cheese ˈðæt ˈʧiːz.

1616 ‘All I can do is to describe that form of London dialect with which I
am sufficiently familiar to enable me to deal with it satisfactorily. The only
real familiarity we can have is with the language we speak ourselves.’ (Sweet,
Primer of Spoken English, 3rd edition, p. vii.)

1717 EPD transcription does not allow for the reductions of auə as I pronounce
them (see § 430). It does cover the speech of those whose reductions of
auə are identical with those of aiə (see § 431).

1818 See my book The Phoneme, §§ 252-254.

1919 I share this view when it is a question of teaching English to the numerous
foreign learners who have no time or inclination to study phonetics. These
learners have no need to concern themselves with cardinal categories. What
they need is (1) to be taught to make the English sounds, and (2) to be taught
to associate each sound with a symbol of a simple phonetic transcription
— the simpler the better in my opinion. See §§ 44-49 of this Appendix.

2020 Other than Scandinavians, for whom the letter æ suggests the English e.

2121 I use the term ‘common pronunciation’ to denote the pronunciation
shown by EPD transcription when (1) the vowel letters are given what may
be considered to be ‘average values’ for Southern English, (2) the sequences
eiə, ouə, aiə, auəand oui are given their full pronunciation (§§ 392a, 403,
414, 430) and (3) the traditionally short vowels, i, e, ɔ, u and ʌ are pronounced
short, but æ is pronounced long when appropriate (§§ 874-878).

22 Voir note 21.

2322 Except possibly in a few words containing æ. It would seem that some
Southern English people distinguish ʤæːm (fruit preserve) from ʤæm
(squeeze) and bæːnd (band) from bænd (banned).

24 Voir note 23.

2523 For the most part using the old symbols ɪ u for the sounds here written
ɩ ɷ The newer symbols date from 1943.

2624 My experience thus does not support the theory held by some that a
phonetic study of the mother tongue is a useful preliminary to acquiring the
pronunciation of a foreign language.

2725 Sweet's use of ö and ü in this work rendered his system not quite phonemic,
and his use of ɔ rendered it not quite uniliteral nor ‘simple.’ He might have
rendered it completely uniliteral and simple by substituting oo for ɔ — a plan
which would not have been out of keeping with the rest of his system.

The modified form of transcription adopted in his Primer of Spoken English
(first published in 1890) and Sounds of English (1908) was narrow in two
respects, namely, that əi and əu were used to denote special varieties of ai
and au occurring in weakly stressed positions. In these books he placed ˘ on
unstressed i, o and u. This was apparently intended as a narrowing, indicating
modified qualities of these vowels when unstressed. Actually it may be
regarded as a stress-mark (denoting lack of stress).

2826 A modification favoured by some (including P. A. D. MacCarthy) is to
show length by doubling the vowel-letters, thus ii, aa, etc. This plan has
much to commend it; its chief advantage lies in the fact that doubled letters
are more readily legible than letters with a length-mark. It involves, however,
introducing a hyphen or other special mark to separate short i from an
adjacent i or ii, as in hurrying ˈhʌri-iŋ, seeing ˈsii-iŋ, and one pronunciation
of rabies ˈreibi-iiz.

2927 Trials were also made with two modified forms of this transcription.
One represented the speech of the numerous Southern English people who
lengthen æ under certain conditions (§ 874). In this transcription æ was
represented by a, lengthened æ by ai, ɑː by ɑː, and ai, au by ɑi, ɑu (see
Le Maître Phonétique, October, 1932, p. 84, January, 1933, p. 14, April, 1933,
pp. 28-33, July, 1933, p. 60, October, 1933, p. 82, April, 1934, p. 56, October,
1934, p. 108, January, 1935, p. 16 and April, 1935, p. 33). In the other æ
was represented by æ and ʌ by a, as in Sweet's revised Broad Romic (see
Le Maître Phonétique, October, 1939, p. 73, January, 1940, p. 18, and April,
1940, p. 36).

3028 Published by Heffer, Cambridge.

31 Voir note 30.

32 Voir note 30.

3329 Published by Longmans, Green & Co.

3430 Published by the Oxford University Press.

35 Voir note 34.

36 Voir note 30.

37 Voir note 30.

3831 Published by the British Council, 65, Davies Street, London, W.1.

3931 Quotation from D. Abercrombie's article Phonetic Transcriptions in Le
Maître Phonétique
, July-December, 1953.