Jones, Daniel. 1997. English pronouncing dictionary. 15th edition, edited by Peter Roach & James Hartman. Cambridge: Cambridge University Press, pp.iv-xv
Introduction
It is strongly recommended that users of this dictionary read the introduction,
since a full understanding of the information in it will ensure the most
effective use of the dictionary.
(PART I) 'What is the English Pronouncing Dictionary?': The intended use of the dictionary, the principles of its design and the accents of English represented in it.
(PART 2) 'Principles of Transcription':The main characteristics of the British and American accents.
(PART 3) 'Explanatory Notes': How
to interpret the information provided with the individual words in the
dictionary.
Part 1: Introduction to the English Pronouncing Dictionary
1.1 What is the English Pronouncing
Dictionary?
This dictionary is designed to provide information on the current pronunciation of approximately 80,000 English words and phrases. For each entry, a British and an American pronunciation is shown (see Section I .2 below). The pronunciation is given in modified phonemic transcription, and you need to understand the principles of phonemic transcription in order to be able to make proper use of this information (see Section 2.I below).
The Pronouncing Dictionary provides much essential information that
is not available in a general dictionary, such as the pronunciation of
proper names, the pronunciation of all inflected forms of each word, and
a larger amount of detail about variant pronunciations than is usual in
a general dictionary.
1.2 Whose pronunciation is represented?
A pronouncing dictionary must base its recommendations on one or more models. A pronunciation model is a carefully chosen and defined accent of a language. In the first edition of this dictionary (1917), Daniel Jones described the type of pronunciation recorded as "that most usually heard in everyday speech in the families of Southern English persons whose menfolk have been educated at the great public boarding-schools". Accordingly, he felt able to refer to his model as "Public School Pronunciation" (PSP). In later editions, e.g. that of 1937, he added the remark that boys in boarding-schools tend to lose their markedly local peculiarities, whereas this is not the case for those in day-schools. He had by 1926, however, abandoned the term PSP in favour of "Received Pronunciation" (RP). The type of speech he had in mind had for centuries been regarded as a kind of standard, having its base in the educated pronunciation of London and the Home Counties (the counties surrounding London). Its use was not restricted to this region, however, being characteristic by the nineteenth century of upper-class speech throughout the country. The Editor of the 14th Edition of this dictionary, A. C. Gimson, commented in 1977 "Such a definition of RP is hardly tenable today", and went on "If I have retained the traditional, though imprecise, term 'received pronunciation', it is because the label has such wide currency in books on present-day English and because it is a convenient name for an accent which remains generally acceptable and intelligible within Britain".
For this edition a more broadly-based and accessible model accent for British English is represented, and pronunciations for one broadly-conceived accent of American English have been added. The time has come to abandon the archaic name Received Pronunciation. The model used for British English is what is referred to as BBC English; this is the pronunciation of professional speakers employed by the BBC as newsreaders and announcers on BBCI and BBC2 television, the World Service and BBC Radio 3 and 4, as well as many commercial broadcasting organisations such as ITN. Of course, one finds differences between such speakers, but there is still a reasonable consensus on pronunciation in this group of professionals, and their speech does not carry for most people the connotations of high social class and privilege that PSP and RP have had in the past. An additional advantage in concentrating on the accent of broadcasters is that it is easy to gain access to examples, and the sound quality is usually of a very high standard.
For American English, the selection also follows what is frequently
heard from professional voices on national network news and information
programmes. It is similar to what has been termed "General American", which
refers to a geographically (largely non-coastal) and socially based set
of pronunciation features. It is important to note that no single dialect
- regional or social has been singled out as an American standard. Even
national media (radio, television, movies, CD-ROM, etc.), with professionally
trained voices have speakers with regionally mixed features. However, "Network
English", in its most colourless form, can be described as a relatively
homogeneous dialect that reflects the ongoing development of progressive
American dialects (Canadian English has several notable differences). This
"dialect" itself contains some variant forms. The variants included within
this targeted accent involve vowels before /r/, possible differences
in words like 'cot' and 'caught' and some vowels before /l/. It is fully
rhotic. These differences largely pass unnoticed by the audiences for Network
English, and are also reflective of age differences. What are thought to
be the more progressive (used by educated, socially mobile, and younger
speakers) variants are listed first in each entry. The intent is to list
the variety of pronunciations with the least amount of regional or social
marking, while still being sensitive to the traits of the individual word.
1.3 How are the pronunciations chosen?
It is important to remember that the pronunciation of English words
is not governed by a strict set of rules; most words have more than one
pronunciation, and the speaker's choice of which to use depends on a wide
range of factors. These include the degree of formality, the amount of
background noise, the speed of utterance, the speaker's perception of the
listener and the frequency with which the speaker uses the word. For example,
the two words 'virtuous' and 'virtuoso' are closely similar in spelling
and share a common origin. However, the former is more common than the
latter, and for British English 
is
given as the first pronunciation of the former but 
for the latter (which in general is typical of more careful speech). If
such variation did not exist, most of the work of compiling a pronouncing
dictionary could be done easily by means of one of the available computer
programs that convert English spelling into a phonemic transcription. Ultimately,
however, the decisions about which pronunciation to recommend, which pronunciations
have dropped out of use, and so on, have been based on the editors' intuitions
as professional phoneticians and observers of the pronunciation of English
(particularly broadcast English) over many years. The opinion of many colleagues
and acquaintances has also been a valuable source of advice.
In general, a pronunciation typical of a more casual, informal style
of speaking is given for common words, and a more careful pronunciation
for uncommon words. In real life, speakers tend to articulate most carefully
when listeners are likely to have difficulty in recognising the words they
hear. When more than one pronunciation of a word is given, the order of
the alternatives is important. The first pronunciation given is believed
to be the most usual one although the distance between the alternatives
may vary, with some alternant forms rivalling the first-given in perceived
frequency while others may be a more distant second.
1.4 Regional Accents
P pronouncing dictionary that systematically presented the pronunciations
of a range of regional accents would be very valuable, but it would be
very much bigger than the present volume and the job of ensuring an adequate
coverage which treated all accents as equally important would have taken
many years. In the case of place-names, information about local pronunciations
has been retained or added as well as "official" broadcasting ones, but
the other words are given only in the standard accents chosen for British
and American English.
1.5 Pronunciation of foreign
words
Many of the words in an English dictionary are of foreign origin, and
in previous editions of this dictionary many such words have been given
both in an Anglicised pronunciation used by most English speakers, and
in a broad phonetic transcription of the "authentic" pronunciation in the
original language. This edition does not give phonetic transcriptions of
all the foreign words; the primary aim of this dictionary is to list pronunciations
likely to be used by educated speakers of English, and an authentic pronunciation
would in some circumstances be quite inappropriate (pronouncing 'Paris'
as 
, for example).
In some cases the information is unnecessary (very few English speakers
would attempt, or even recognise, an authentic pronunciation of a word
from a non-European language), while in other cases it is difficult to
establish the authentic original (many African place-names, for example,
have reached us after being adapted by British, French or Portuguese colonists;
place-names in Spain may be pronounced in different ways according to their
regional affiliation, so that the name of Barcelona might be given a Catalan
or a Castilian Spanish pronunciation, while other Spanish names are different
according to whether they originate in Spain or South America). Words and
names of foreign origin are therefore given in what is felt to be the pronunciation
most likely to be used among educated speakers of English.
In some cases it is possible to identify an alternative pronunciation
which represents an attempt to pronounce in a manner closer to the supposed
original. This is marked by first indicating the language which the speaker
would be aiming at, then giving the pronunciation, using where necessary
additional phonetic symbols not required for the phonemic transcription
of English. For example, the word 'bolognese' is widely used to refer to
a sauce served with pasta. This is given as 
for British English and as 
for American; for speakers of both groups, a pronunciation aimed at being
nearer to the Italian original would be 
(though this would still be different from the pronunciation that would
be produced by an Italian speaker). To indicate that this last pronunciation
is aimed at sounding Italian, it is marked in the entry as: as if Italian:
. In a few cases it has been necessary to mark separate British and American
pronunciations within this field, as the degree of Anglicisation of any
given word may vary between British and American English.
1.6 Usage notes
Usage notes are included with some words. In some cases these are needed
so that users of the dictionary can understand how alternative pronunciations
are to be used. In some cases the rules needed for correct pronunciation
are quite complex, most noticeably in the case of the so-called "weak-form
words" such as 'there', 'her'. Explanations with examples are given in
such cases.
1.7 Syllable divisions
Earlier editions of this dictionary regularly marked the division between
syllables. This practice was largely abandoned in the 14th Edition, but
this new edition gives syllable divisions in all cases, since it is felt
that foreign learners will find the information useful. Syllable division
is marked with the symbol . recommended by the International Phonetic Association.
The decision on where to place a syllable division is not always easy,
and the rules used in this work are explained later in this Introduction
(Section 2.6).
Part 2: Principles of transcription
2.1 The phoneme principle
The basic principle of the transcription used is, as in all previous
editions, phonemic. This means that a small set of symbols is used
to represent the sounds that can be shown to be distinctive in English,
so that replacing one phoneme by another can change the identity of a word.
We do not usually add phonetic detail such as the presence of glottal stops,
aspiration or vowel devoicing. It is usual to put slant brackets before
and after symbols representing phonemes (e.g. the word 'cat' would be represented
phonemically as 
).
When non-phonemic symbols are used, the convention is to use square brackets
(e.g. the glottal stop will be represented as 
).
In entries in the dictionary itself, however, we do not use these brackets,
in order to keep the information simple; only in explanatory notes do we
use slant or square brackets. For an explanation of the principle of the
phoneme and some of the problems associated with it, see Roach (I99I),
Chapters 5 and I3. The use of
phonemic transcription in works on pronunciation (including this one) has
remained in the "realist" tradition established by Jones, while approaches
to the phoneme by theoretical phonologists have changed radically during
recent decades and become much more abstract. There are a few exceptions
to our general use of the phoneme principle that should be mentioned here,
however. One is the use, in American pronunciations, of the 
diacritic to indicate the "flapping" of /t/ in words such as 'getting' 
,
and 'better' 
.
This is an important feature of American pronunciation, but speakers of
British English find it difficult to apply the rule which determines when
phonemes are flapped. Another is the use of the symbols [i] and [u], the
use of which is explained below (Section 2.9). Finally,
it is necessary to use a number of special symbols which are not normally
used for English phonemes. This set includes some nasalised vowels used
particularly in some words taken from French, the [x] sound found in Scottish
words such as 'loch', and some non-linguistic sounds used in certain exclamations
and interjections (see Section 2.4).
2.2 Vowels and diphthongs
It is standard practice in phonetics to represent the quality of vowels
and diphthongs by placing them on a four-sided figure usually known as
the Cardinal Vowel quadrilateral (see Roach (I99I),
pp II-I4). This device is used in the vowel descriptions
in the following section.
(a) British English
British English (BBC accent) is generally described as having short
vowels, long vowels and diphthongs. There are said to be seven short vowels,
five long ones and eight diphthongs. At the end of this section some attention
is also given to triphthongs.
- Short vowels:
- Long vowels:
- Diphthongs:
>> Fig. I BBC English pure vowels <<
>> Fig. 2 BBC English closing diphthongs <<
>> Fig. 3 BBC English centring diphthongs <<
A few comments on individual vowels and vowel symbols are needed. The
pronunciation of any language is constantly changing, and a dictionary
such as this one should reflect such changes. However, there is a general
reluctance among users of phonemic transcription to change the symbols
used too frequently, as this causes existing teaching materials and textbooks
to become out of date. The following remarks apply chiefly to BBC pronunciation.
(a) The length of long vowels and diphthongs is very much reduced when they occur in syllables closed by the consonants(b) American English. Thus
in 'beat' has only about half the length of
in 'place' is much reduced in length compared with
(b) The vowel
, classified as a short vowel, is nevertheless generally lengthened before
. Thus
might one day be considered preferable. We have retained the
(c) The vowel
used to be a back vowel, and the symbol was chosen for this reason. This is no longer a back vowel, but a central one. Alternative symbols could be considered in the future.
(d) Among younger speakers, the
vowel has moved to a more front quality, with less lip-rounding, particularly when preceded by
as in 'use'.
(e) Among the diphthongs, there seems to be a progressive decline in the use of
, with
taking its place (e.g. the pronunciation of the word 'poor' as
is increasingly common).
(f) Triphthongs create some problems. These three-vowel sequences are generally held to be composed of one of the diphthongs
plus a schwa (e.g.
). In British English many of these triphthongs are pronounced with such slight movement in vowel quality that it is difficult for foreign learners to recognise them; for example, the name 'Ireland', which is generally transcribed
, frequently has an initial syllable which sounds virtually indistinguishable from
. It seems reasonable in this case to treat these sounds as being monosyllabic (e.g. the word 'fire' is a single syllable), but in other words and names transcribed with the same symbols it seems necessary to insert a syllable division. This is usually done (i) when there is a morpheme boundary (e.g.
) and (ii) when the word is felt to be foreign (this includes many Biblical names originating from Hebrew, e.g. 'Messiah').
Another problem with triphthongs is that before an /r/ consonant at the beginning of a following syllable, the distinction between
and
seems to be neutralised - it seems to make no difference whether one represents 'Irish', 'irate' as
or as
, since there is no regular distinction made in pronunciation. In general, the practice of this edition is to transcribe such cases as
.
American English is commonly described as having lax vowels, tense vowels,
and wide diphthongs. Generally speaking, lax vowels are lower and made
with less oral tension; they do not usually end syllables. Vowel length
in American English is generally considered to be conditioned by phonological
environment, so the long/short distinction described for BBC English is
not usually present, though we have retained the length mark on the tense
vowels 
in order to mark their relationship to the English long vowels. Since the
diphthongal movement in 
and 
is
small in American pronunciation, these are treated as tense vowels. Vowels
preceding /r/ are notably influenced by rhotic colouring. Word spellings
such as 'bird', 'word', 'curd', 'earth', 'jerk', which now rhyme with 
in American English, at one time in history had differing vowels. The retroflexed
vowels and 
,
stressed and unstressed, are among those features that noticeably distinguish
American English from BBC English. All vowels occurring before /r/ within
a syllable are likely to become "r-coloured" to some extent.
There is an issue in the symbolization of the diphthong in the word "home". This has for many years been represented as
but in earlier editions of this and others of Jones' works the symbolization 
indicated
a rounded initial vowel. This is still the preferred transcription for
the American English diphthong. In order to preserve compatibility with
other works, we have chosen to use
for BBC English and
for American, though it can be argued that the latter symbolization would
be suitable for both.
The American
vowel is somewhat closer than BBC ,and
seems to be evolving into an even closer vowel in many speakers. It is
used in the same words as BBC
and also in most of the words which in BBC have 
when there is no letter r in the spelling, e.g. 'pass', 'ask'. The quality
of American
is similar to the BBC
vowel; it is used in some of the words which have
in BBC when there is no letter r in the spelling (e.g. 'father', 'calm').
It also replaces the BBC short 
vowel in many words (e.g. 'hot', 'top', 'bother'): 'bother' rhymes with
'father'. American 
is more open in quality than BBC .
It is used where BBC has
(e.g. 'cause', 'walk'), and also replaces BBC short
in many words, e.g. 'long', 'dog'. American 
is similar to BBC ,
but is also used where BBC has /ju:/ after alveolar consonants (e.g. 'new',
'duty').
2.3. Consonants
(a) British English (BBC)
These consonants can be arranged in table form as shown below. The
layout of the symbols follows the principle that, where there are two consonants
which differ only in voicing, they are placed side by side with the voiceless
one to the left.
(a) Certain types of consonant have a distinction such as that between /t/ and /d/; this is commonly classed as a distinction between voiceless and voiced consonants, but the distinction is in fact much more complex. Consonants usually classed as voiceless are
,
with voiced partners 
.
Since the presence or absence of voicing is often less important than some
other phonetic features, it has been suggested that instead the terms fortis
(equivalent
to voiceless) and lenis (equivalent to voiced) should be used. These
terms imply that the main distinguishing factor is the amount of energy
used in the articulation (fortis consonants being made with greater energy
than lenis). These terms are not used in this dictionary, since the usefulness
of this terminology is uncertain. Some of the characteristics of the two
types of consonant are set out below.
(b) /p, t, k/ are typically accompanied by aspiration (i.e. an interval of breath before the following vowel onset), especially when initial in a stressed syllable. Thus, ‘pin’ is distinguished from 'bin' very largely by the aspiration accompanying /p/. However, in the syllable-initial sequences /sp-, st-, sk- /, /p, t, k/ lack such aspiration. When /1/ or /r/ immediately follow /p,t,k/, they are devoiced and are pronounced as fricatives. Another characteristic of /p,t,k/ that is not marked in transcriptions is glottalization; when one of these consonants is followed by another consonant it is now usual to find that a glottal closure precedes the /p/, /t/ or /k/, particularly if the syllable in which they occur is stressed. Thus the pronunciation of 'captain', 'rightful', 'Yorkshire', which are phonemically, could be shown (using the symbol
. Similarly, in American English 'mountain' has one pronunciation that could be represented as
.
(c) Voiceless consonants have a shortening effect on sounds preceding them within a syllable. Thus in the words 'right' and 'ride'
the diphthong is noticeably shorter in the first word than in the second; in the words 'bent' and 'bend'
, both the vowel /e/ and the nasal consonant /n/ are shorter in the first word. This length difference is not always easy to observe in connected speech.
(d) The consonant /l/ has two different allophones in BBC English, the so-called "clear" and "dark" allophones. The "clear" one (which has an /iù /-like quality) occurs before vowels, the "dark" one (which has an /uù /-like quality) before consonants or before a pause.
-
The consonants
are usually accompanied by lip-rounding.
(a) American English
The consonants of the American English model, at the phonemic level, may be represented by the safe broad scheme used for British English above. Similarly, many of the distinguishing phonetic traits discussed for British English hold for
American English as well: initial /p, t, k/ are normally aspirated except
when immediately preceded by /s/. Glottalization preceding and, at times,
replacing the plosives occurs often in rapid speech. There are, of course,
numerous phonetic and phonological differences between British and American
English, as there are within regional and social varieties within the two
political entities. Two differences receive sufficient attention and have
attained sufficient generality within the two varieties that they are represented
here. One is phonetic: the "flapped" medial /t/ (as in 'butter')
is transcribed as 
(see
Section 2.1 above); the other is phonological:
the presence (in American English) of postvocalic /r/ (as in 'farmer' 
).
It should also be noted that the difference between "clear" and "dark"
/l/ is much less marked in American than in the BBC accent, so that
even prevocalic /l/ in American pronunciation sounds dark to English ears.
2.4 Non-English sounds
In addition to the phonemes of English described above, most English
speakers are aware of, and often attempt to pronounce, some sounds of languages
other than English. The number of such sounds is small, since most foreign
words and names are Anglicised so that they are pronounced with English
phonemes. We find the voiceless velar fricative [x]
in the Gaelic languages of Scotland and Ireland in words such as 'loch'
and names such as 'Strachan'. The same sound is often used by English speakers
for the German sound which is written 'ch' (e.g. 'Bach' 
)and
the Spanish sound spelt 'j' (e.g. 'Badajoz' 
),
though these sounds are often not really velar fricatives. The voiceless
lateral fricative 
is found (always represented in spelling with 'll') in Welsh words and
names such as 'Llanberis'; we give the pronunciation of this sound as hl
to indicate that it may be pronounced as a voiceless
(as many British English speakers do), but alternatively as a voiced one:
thus 
.
The dictionary lists a few names with more than one of these sounds (e.g.
Llanelli). Most non-Welsh speakers are unlikely to pronounce more than
one sound
in a word, so we give the pronunciation as 
for 'll' sounds after the initial one.
The other case which needs special attention is the pronunciation of
French nasalised vowels. Many English speakers attempt to produce something
similar to the French vowels
in
words such as 'vin rouge', 'restaurant', ' bon marché',
'Verdun'.
Although many speakers do not get close to the French vowels, the principle
adopted here is to use symbols for English vowels, with added nasalisation.
The equivalents are:
2.5 Stress
Stress patterns present one of the most difficult problems in a pronouncing
dictionary. One reason for this is that many polysyllabic words have more
than one possible stress pattern, and one must consider carefully which
should be recommended. Secondly, the stress of many words changes in different
contexts, and it is necessary to indicate how this happens. Thirdly, there
is no straightforward way to decide on how many different levels of stress
are recognisable.
(a) Where more than one stress pattern is possible, the preferred pronunciation
is given first and then alternatives are listed. Many dictionaries use
the convention of representing stress patterns using dashes to represent
syllables: thus the two possible patterns for 'cigarette' 
can be shown as 
.
This convention, which is sometimes referred to (incorrectly) as "Morse
Code", is used in this work for short words, since it is economical on
space. However, in longer words users are likely to find it difficult to
interpret. In the planning of this edition, an experiment was carried out
to test this, and it was found that readers (both native speakers and non-native
speakers of English) do indeed take less time to read word stress patterns
when the whole word is given, rather than just a "dashes and dots" pattern
(Stromberg and Roach, 1993). Consequently, words of more than three syllables
are given in full when alternative stress patterns are being given.
b) The most common case of variable stress placement caused by
context is what is usually nowadays known as "stress-shift". As a general
rule, when a word of several syllables has a stress near the end of the
word, and is followed by another word with stress near its beginning, there
is a tendency for the stress in the first word to move nearer the beginning
if it contains a syllable that is capable of receiving stress. For example,
the word 'academic' in isolation usually has the stress on the penultimate
syllable /-dem-/. However, when the word 'year' follows, the stress is
often found to move to the first syllable 
.
The whole phrase 'academic year' will have its primary stress on the word
'year', so the resulting stress pattern will be 
(where 
represents
secondary stress and 
represents primary stress). To make this process easier to understand,
this dictionary now gives specific examples in each case where stress-shift
is possible except where certain prefixes such as 'un-' produce hundreds
of such cases. In general, this shift is not obligatory: it would not be
a mispronunciation to say 
.
However, it is undoubtedly widespread and in some cases is used almost
without exception: for example, although the adjective 'compact' on its
own is pronounced with the stress pattern 
,
in the phrase 'compact disc' it is virtually always pronounced with stress
on the first syllable.
(c) It is necessary to decide how many levels of stress to mark. The
minimum possible range is two: stressed and unstressed. This is inadequate
for representing English words in a pronouncing dictionary: a word such
as 'controversial' clearly has stress(es) on the first and third syllables,
and equally clearly has stronger stress on the third syllable than on the
first. It is therefore necessary to recognise an intermediate level of
stress ("secondary"). The transcription of this word, therefore, is 
.
An argument can be made for recognising yet another level (tertiary stress):
in a word such as 'indivisibility', for example, it can be claimed that
the level of stress on the third syllable 
is weaker than that on the first syllable 
,
which has secondary stress (primary stress being placed on the penultimate
syllable 
).
However, introducing this extra level creates a degree of complexity that
it is better to avoid. In EPDI4 some long polysyllabic
words were transcribed with two primary stress marks (e.g. 'cross-examination'
was given as 
):
for the present edition only one primary stress may occur in a word or
compound.
(d) Secondary stresses have only limited occurrence after a primary
stress: such a secondary stress is only marked in closed or hyphenated
compound words where the second element is polysyllabic (e.g. 
).
(e) Stress assignment on prefixes:
(i) In words containing a prefix such as, for example, con-, de-, im-, in-, secondary stress is not applied to the prefix where the following (i.e., second) syllable is stressed. Examples include 'intoxicate'2.6 Syllable divisions.
(ii) Where the prefix is separable, however, as in impossible, a variant showing secondary stress on the prefix is listed, as follows:
.
(iii) In all other cases, primary or secondary stress is applied to the prefix where appropriate.
The I4th Edition of EPD marked syllable division
(using hyphens) only when it was important to distinguish between the affricate 
and the phonemes 
and 
at a syllable juncture (e.g. 'satchel' 
and 'nutshell' 
).
However, although native speakers may well find no difficulty in dividing
words into syllables, it seems that learners of English have trouble in
doing so, and the divisions are therefore marked. Descriptions of stress
and rhythm are usually expressed in terms of syllables, and so it is helpful
to have polysyllabic words clearly broken up into their constituent syllables.
The syllabified transcription of a polysyllabic word is easier to read
and interpret than an undivided one. In addition, the dictionary is likely
to be of interest to the field of speech and language technology, where
syllable divisions can be useful in developing automatic speech and language
analysis systems.
A dot . is used to divide syllables, in accordance with the current
recommendations of the International Phonetic Association (these may be
read in the Journal of the International Phonetic Association, Vol.19.2,
(1989),
pp.67-80, and in the forthcoming IPA Handbook).
However, this is not used where a stress mark or
occurs, as these are effectively also syllable division markers.
No completely satisfactory scheme of syllable division can be produced - all sets of rules will throw up some cases which cannot be dealt with properly. The principles used in this edition are set out below. This requires some discussion of phonotactics, the study of permissible phoneme sequences.
(a) As far as possible, syllables should not be divided in a way that
violates what is known of English syllable structure. The 'Maximal Onsets
Principle', which is widely recognised in contemporary phonology, is followed
as far as possible. This means that, where possible, syllables should be
divided in such a way that as many consonants as possible are assigned
to the beginning of the syllable to the right (if one thinks in terms of
how they are written in transcription), rather than to the end of the syllable
to the left. However, when this would result in a syllable ending with
a stressed 
,
it is considered that this would constitute a violation of English phonotactics,
and the first (or only) intervocalic consonant is assigned to the preceding
syllable; thus the word 'better' is divided 
,
whereas 'beater' is divided 
.
In the case of unstressed short vowels, 
are
also prevented from appearing in syllable-final position; however, unstressed 
are allowed the same "privilege of occurrence" as 
when a consonant begins a following syllable, and may therefore occur in
final position in unstressed syllables except pre-pausally. Thus in a word
such as 'develop', the syllable division is 
.
(b) Notwithstanding the above, words in compounds should not be re-divided
syllabically in a way that does not agree with perceived word boundaries.
For example, 'hardware' could in theory be divided 
,
but most readers would find this counter-intuitive and would prefer 
.
This principle applies to open, closed and hyphenated compounds.
2.7 Assimilation
Assimilation is a process found in all languages which causes speech
sounds to be modified in a way which makes them more similar to their neighbours.
A well-known example is that of English alveolar consonants such as /t,
d, n/, which, when they are followed by a consonant which does not have
alveolar place of articulation, tend to adopt the place of articulation
of the following consonant. Thus the /t/ at the end of 'foot' 
changes to /p/ when followed by /b/ in the word 'football', giving the
pronunciation 
.
A similar case is the assimilation of /s/ to a following
or /j/, resulting in the pronunciation of 'this ship' as 
and 'this year' as 
.
This assimilation can be considered to be optional.
The assimilation of /n/ is a rather special case: many English words
begin with the prefixes 'in-' and 'un-', and in a number of cases the /n/
of these prefixes is followed by a consonant which is not alveolar. In
some cases it seems to be normal that the /n/ is regularly assimilated
to the place of articulation of the following consonant (e.g. 'inquest' 
),
while in others this assimilation is optional (e.g. 'incautious' may be 
).
Where it is clear that the prefix is attached to a word that exists independently,
so that prefix and stem are easily separable, the assimilation is normally
treated as optional. When it seems more like an integral part of the word,
the assimilation is shown as obligatory. The occurrence of assimilation
in British and American English may differ.
2.8 Treatment of /r/
The accent used for British English is classed as non-rhotic - the phoneme /r/ is not usually pronounced except when a vowel follows it. The American pronunciations, on the other hand, do show a rhotic accent, and in general in the accent described, /r/ is pronounced where the letter r is found in the spelling.
It is necessary to show, in British English entries, cases of potential
pronunciation
of /r/, mainly in word-final position; in other words, it is necessary
to indicate, in a word such as 'car', that though the word when said in
isolation does not have /r/ in the pronunciation (
),
there is a potential
/r/ which is realised if a vowel follows
(e.g. in 'car owner'). This is indicated by giving the transcription as
,
where the superscript /r/ indicates the potential for pronunciation.
This is traditionally known as 'linking r'. A controversial question is
that of so-called 'intrusive r', where the phoneme /r/ is pronounced
when no 'r' is seen in the spelling. For example, the phrase 'china and
glass' will often be pronounced with /r/ at the end of the word
'china'; although this type of pronunciation is widespread in the speech
of native speakers of the accent described, it is still safer not to recommend
it to foreign learners, and it is therefore avoided in this dictionary.
2.9 Use of /i/ and /u/
There are many places in present-day British and American English where
the distinction between 
and 
is neutralised.
For example, the final vowel of 'city' and 'seedy' seems to belong neither
to the phoneme
nor to . The
symbol /i/ is used in this case (though it is not, strictly speaking,
a phoneme symbol; there is no obvious way to choose suitable brackets for
this symbol, but phoneme brackets // will be used for simplicity). A parallel
argument can be made for the distinction between 
and (with
a corresponding 'neutralised' symbol 
),
though this is needed much less frequently. This issue, and the issues
which follow, are discussed in detail in Roach (I99I),
pp. 77-78.
(a) In word-final position,
and do not
occur. Word-final, close vowels are transcribed with 
and
if unstressed. Word-final
and are
possible both with stress ('grandee', 'bamboo') and without ('Hindi', 'argue'),
although in the unstressed case it is often not possible to draw a clear
line between
and ,or
between
and .
(b) In compounds such as 'busybody' and names such as 'Merryweather', is
permitted to occur word-medially, e.g. 'busybody' is transcribed 
and 'Merryweather' as 
.
In all other cases word-medially,
is used when the vowel is unstressed, unless a vowel follows (see below).
(c) The vowel symbols
and only occur
in front of another vowel symbol if they form part of a composite (diphthong
or triphthong) phoneme symbol (e.g. 
).
Otherwise or
is used (e.g. 'scurrying' 
,
'influenza' 
.
(d) A matter related to this decision concerns words ending in '-ier',
'-eer', '-ia'. The usual transcription in the I4th
Edition of the EPD was 
.
However, 'reindeer' and 'windier' (comparative form of 'windy') do not
have identical pronunciations in their final syllables in British English
(BBC). In this edition, the alternative
previously given for the latter type of word has been dropped; 'reindeer'
is transcribed as 
and 'windier' as 
.
The latter transcription, which indicates a different (closer) vowel quality
in the second syllable of 'windier', and implies a pronunciation with three
rather than two syllables, is felt to be accurate in terms of contemporary
pronunciation.
The long vowels
and may
also occur before other vowels, but only when in a stressed syllable (e.g.
'skiing' 
,
'canoeing' 
.
2.10 Syllabic consonants
Syllabic consonants are frequently found in English pronunciation: these
are cases where instead of an expected vowel-plus-consonant sequence, the
consonant alone (usually one of 
)
is pronounced with the rhythmical value of a syllable (See Roach, I99I,
pp.78-82). In EPDI4, syllabic
consonants were only marked where there is ambiguity in the pronunciation
of a word; for example, in a word such as 'bottle', the transcription 
is said to imply unambiguously that the /l/ is syllabic, whereas in the
derived form 'bottling' there may be two pronunciations, one with and one
without a syllabic /l/. In this instance the EPDI4
preferred pronunciation was
,
with 
given
as an alternative.
(a) The main problem here is how to deal with optional and obligatory
syllabicity and the permissibility of vowels. The most frequently found
case is where an item may have (i) a schwa vowel followed by a non-syllabic
consonant, (ii) a syllabic consonant not preceded by schwa or (iii) a nonsyllabic
consonant not preceded by schwa. For example, 'lightening' may be (i) 
,
(ii) 
or
(iii) 
. Such
items are transcribed as 
and 
, the
first representing cases (i) and (ii), in which there are three syllables,
and the second representing only the disyllabic pronunciation, (iii). The
use of superscript schwa in words such as
should be interpreted as meaning that the schwa may be pronounced, or may
be
omitted while giving its syllabic character to the following consonant.
(b) The problem remaining is that of /l/ corresponding to the
'-le' spelling form, preceded by any plosive or homorganic fricative, as
in 'bottle', 'wrestle'. It is not felt to be acceptable in BBC pronunciation
to pronounce this with a vowel in the second syllable, and therefore the
superscript schwa convention is not used in such cases: rather, the /l/
is marked as syllabic, i.e. 
. The entry for 'bottle' is 
,
and for 'cycle', 
.
(c) Where a word such as the above carries a suffix with an initial
vowel, as in 'bottling', 'cycling', speakers' intuitions about the number
of syllables are very divergent, so no single recommendation will be adequate.
The entry for 'bottle' gives the syllabic (three-syllable)
version of 'bottling' as the first recommendation, and it is therefore
necessary to add for the ring form the two-syllable alternative 
.
(d) Syllabic nasals are not usual where they would result in a nasal-plosive-syllabic
consonant sequence (e.g. 'London', 'abandon' must contain a schwa vowel
in the final syllable).
2.11 Optional sounds
The convention used in EPDI4 of printing phoneme
symbols in italics to indicate that they may be omitted is retained, though
used more sparingly. It is not necessary to give alternative pronunciations
that simply follow general rules of simplification that apply in rapid
speech. For example, pointing out the possibility of omitting the [d] sound
in 'engine' seems unnecessary, whereas it does seem worth recording the
fact that some speakers pronounce words such as 'lunch' and 'French' with
a final 
while others have final 
. There is a difference between the two cases: the former is a straightforward
example of elision, and needs no special explanation that refers to a specific
word or class of word, while the latter is a particular case of an insertion
or deletion that is restricted to a particular phonological environment;
speakers are usually consistent in using one or other of the alternative
pronunciations in the latter case.
2.12 Elision
As mentioned in the preceding section, there are many cases where sounds which are produced in words pronounced on their own, or in slow, careful speech, are not found in a different style of speech. This is known as elision, and this dictionary normally does not show elisions in order to avoid adding a large number of additional pronunciations that are typical of casual speech. It is usual to explain elision in terms of the Principle of Least Effort - we try to avoid doing more work than is necessary. We find elision most commonly in the simplification of consonant clusters. A common example is the loss of /t/ and /d/in combination with other consonants. Examples are:
The fricative 
is also frequently lost in clusters in rapid speech. Examples are:
Elision of vowels is also found, and again this seems to be characteristic
of rapid or casual speech. Examples are:
Jones, Daniel. 1997. English pronouncing dictionary. 15th edition, edited by Peter Roach & James Hartman. Cambridge: Cambridge University Press, pp.iv-xv