more information – www.cambridge.

org/9780521808828
0521808820cfm_pi_viii.qxd 1/27/05 3:09 PM Page i Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Introducing Phonetic Science

This accessible new textbook provides a clear and practical introduction
to phonetics, the study of speech. Assuming no prior knowledge of the
topic, it introduces students to the fundamental concepts in phonetic
science, and equips them with the essential skills needed for
recognising, describing and transcribing a range of speech sounds.
Numerous graded exercises enable students to put these skills into
practice, and the sounds introduced are clearly illustrated with examples
from a variety of English accents and other languages. As well as looking
at traditional articulatory description, the book introduces acoustic and
other instrumental techniques for analysing speech, and covers topics
such as speech and writing, the nature of transcription, hearing and
speech perception, linguistic universals, and the basic concepts of
phonology. Providing a solid foundation in phonetics, Introducing Phonetic
Science will be invaluable to all students beginning courses in linguistics,
speech sciences, language pathology and language therapy.

MICHAEL ASHBY is Senior Lecturer in Phonetics at University College London,

and an experienced teacher of phonetics at both undergraduate and
graduate level. He is the Phonetics editor of the Oxford Advanced Learner’s
Dictionary and other ELT dictionaries, and has published various papers
on both English and general phonetics.

JOHN MAIDMENT is Lecturer in Phonetics at University College London, and

an experienced teacher of phonetics at both undergraduate and graduate
level. He is co-author of English Transcription Course (2000), and has been
Director of an HEFCE-funded project on teaching and learning in
phonetics. He has published various papers on both English and general
phonetics.
0521808820cfm_pi_viii.qxd 1/27/05 3:09 PM Page ii Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Cambridge Introductions to Language and Linguistics

This new textbook series provides students and their teachers with accessible introductions to the major
subjects encountered within the study of language and linguistics. Assuming no prior knowledge of the
subject, each book is written and designed for ease of use in the classroom or seminar, and is ideal for
adoption on a modular course as the core recommended textbook. Each book offers the ideal introductory
material for each subject, presenting students with an overview of the main topics encountered in their
course, and features a glossary of useful terms, chapter previews and summaries, suggestions for further
reading, and helpful exercises. Each book is accompanied by a supporting website.

Books published in the series:

Introducing Phonology David Odden
Introducing Speech and Language Processing John Coleman
Introducing Phonetic Science John Maidment and Michael Ashby

Forthcoming:
Introducing Sociolinguistics Miriam Meyerhoff
Introducing Morphology Maggie Tallerman and S. J. Hannahs
Introducing Historical Linguistics Brian Joseph
Introducing Second Language Acquisition Muriel Saville-Troike
Introducing Language Bert Vaux
0521808820cfm_pi_viii.qxd 1/27/05 3:09 PM Page iii Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Introducing
Phonetic
Science
MICHAEL ASHBY
AND
JOHN MAIDMENT
cambridge university press
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, São Paulo,
D elhi, MexicoiCity

Cambridge University Press

The Edinburgh Building, Cambridge CB2 8RU, UK
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
Information on this title: www.cambridge.org/9780521808828

© Michael Ashby and John Maidment 2005

This publication is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without
the written permission of Cambridge University Press.

First published 2005

7th printing 2012

Printed and iboundiin the United Kingdom byithe MP G Books G roup

A catalogue record for this book is available from the British Library

Library of Congress Cataloguing in Publication data

Ashby, Michael.
Introducing phonetic science / Michael Ashby and John Maidment.
p. cm. – (Cambridge introductions to language and linguistics)
Includes bibliographical references and index.
ISBN 0-521-80882-0 (alk. paper) – ISBN 0-521-00496-9 (pb. : alk. paper)
1. Phonetics. I. Maidment, John A. II. Title. III. Series.

P221. A778 2005 2004054642

414'.8–dc22

ISBN 978-0-521-80882-8 hardback

ISBN 978-0-521-00496-1 paperback

Cambridge University Press has no resonsibility for the

persistence or accuracy of URLs for external or third-party
internet websites referred to in this publication, and does
not guarantee that any content on such websites is, or
will remain, accurate or appropriate.
0521808820cfm_pi_viii.qxd 1/27/05 3:09 PM Page v Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Contents
1 Introduction to speech 1
Introduction 2
Sounds and symbols 2
Transcribing sounds 4
Types of transcription 5
Syllables 7
Segments: vowels and consonants 7
Suprasegmentals 9
Speech as an acoustic signal 9
Writing systems 10
The relationship between speech and writing 13
Problems with syllables and segments 15
Chapter summary 16
Exercises 16
Further reading 19

2 Voice 21
Introduction 22
The larynx 22
Vocal folds 22
Voiced and voiceless sounds 22
Observing vocal fold vibration 24
Voicing in languages 26
Whisper 26
Aperiodic waves 29
Chapter summary 30
Exercises 31
Further reading 32

3 Place of articulation 33
Introduction 34
A brief tour of the vocal tract 34
Articulation 36
Place 36
Complex articulations 40
Homorganic sounds 40
Instrumental techniques 41
Places of articulation used in languages of the world 45
Chapter summary 46
Exercises 46
Further reading 49
0521808820cfm_pi_viii.qxd 1/27/05 3:09 PM Page vi Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

vi Contents

4 Manner of articulation 51
Introduction 52
Airflow, pressure, turbulence 52
Degree of stricture: obstruent and sonorant 53
Stops: oral and nasal, plosive and affricate 54
Fricatives: median and lateral 56
Approximants: median and lateral 57
Taps, flaps and trills 59
Voice, place, manner labels 61
Use of manners of articulation in the world’s languages 61
Chapter summary 63
Exercises 63
Further reading 67

5 Vowels 69
Introduction 70
Spectrum and resonance 70
Vowel articulation 72
Long and short vowels 75
Oral and nasalised vowels 75
Monophthong vs diphthong 75
The vowel quadrilateral 76
Cardinal vowels 77
Vowel systems 79
Vowel instability and variability 80
Chapter summary 83
Exercises 84
Further reading 89

6 Voice II 91
Introduction 92
Voice Onset Time 92
Aspiration 92
Devoicing 95
Phonation type 97
Chapter summary 99
Exercises 99
Further reading 101

7 Airstream mechanisms 103

Introduction 104
Airstream generation 104
Ejectives 105
Implosives 108
Clicks 110
Chapter summary 114
0521808820cfm_pi_viii.qxd 1/27/05 3:09 PM Page vii Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Contents vii

Exercises 114
Further reading 115

8 Speech sounds and speech movements 117

Introduction 118
Primary and secondary articulation 118
The linguistic use of secondary articulation 120
Coarticulation 122
Acoustic and physiological data 129
Chapter summary 132
Exercises 132
Further reading 133

9 Basic phonological concepts 135

Contrast 136
Minimal pair 136
Environment and distribution 137
Alternation 140
Phonological processes 140
Features 142
Underlying and surface forms 144
Phonological rules 144
The syllable 145
Phonotactics 146
Chapter summary 148
Exercises 148
Further reading 151

10 Suprasegmentals 153
Introduction 154
Fundamental frequency and pitch 154
Stress 154
Rhythm 160
Lexical tone 162
Intonation 165
Paralinguistic features 171
Chapter summary 172
Exercises 173
Further reading 176

11 Speaker and hearer 177

Introduction 178
The hearing mechanism 178
Visual cues 180
Speech perception tests 181
Cue redundancy 183
0521808820cfm_pi_viii.qxd 1/27/05 3:09 PM Page viii Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

viii Contents

Speech development 184

Hearing impairments 185
Chapter summary 187
Further reading 187

Glossary 189
Solutions to the exercises 201
Index 219
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 1 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

CHAPTER

1 Introduction
to speech

CHAPTER OUTLINE
In this chapter you will learn about: the basic distinction
KEY TERMS between spoken and written language; the ways in which
Amplitude languages of the world are written; the units from which
Consonant speech is composed: syllables, vowels and consonants;
International phonetic symbols as a means of representing speech;
Phonetic speech considered as an acoustic signal; the similarities and
Alphabet (IPA) differences in the speech sounds used in languages of the
Loudness world.
Phonetic symbol
Phonetic
transcription
Segment
Syllable
Vowel
Waveform
Writing system
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 2 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

2 INTRODUCING PHONETIC SCIENCE

Introduction
The way we usually represent and describe speech depends on a powerful
idea that is already known by everyone who is literate in a language with
an alphabetic writing system. Human listeners can hear speech as a
sequence of sounds, and each sound can be represented by a written
mark. In this chapter we look at how this idea can be the basis of a com-
prehensive system of phonetic symbols, suitable for representing reliably
the sounds of any language – and at how this is different from the many
existing writing systems for particular languages.

Sounds and symbols

Although there are estimated to be 5,000 to 8,000 languages in the world,
each with its own particular selection of sounds, the total number of sym-
bols required to represent all the sounds of these languages is not very
large – it is somewhere around two to three hundred. This, of course, is
because many sounds are found again and again in languages. The human
speech apparatus, which produces sounds, and the hearing mechanism,
which perceives them, are exactly the same all over the world. Languages
make their selection from the stock of humanly possible sounds – and
some sounds are so common and basic that they are found in almost all
languages.

From surveys that compare the sounds employed in hundreds of lan-

guages, we know that almost all languages have consonant sounds
like those at the beginnings of the English words tea, key, pea, see, fee,
me and knee, and vowel sounds resembling those heard in seat and sat.
Although English has quite a large system of sounds, none of the
sounds is very unusual when seen in a global perspective. The most
unusual sounds in English are probably the so-called th- sounds heard
at the beginning of think and this. It’s interesting that some varieties
of English don’t use these sounds, but replace them with others that
are more frequent in the world’s languages. Speakers from London,
for instance, often say fink rather than think.

The International Phonetic Alphabet

The International Phonetic Alphabet (IPA) aims to provide a separate sym-
bol for every sound used distinctively in a human language (see Figure 1.1).
Using symbols from the IPA, we should be able to represent the pronunci-
ation of any word or phrase in any human language. The IPA has grown
and evolved over more than a century of international collaboration, with
new symbols being added when new sounds turned up in languages that
had not previously been described.
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 3 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Introduction to speech 3

FIGURE 1.1
The International Phonetic Alphabet.
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 4 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

4 INTRODUCING PHONETIC SCIENCE

The IPA, then, is intended to contain all the sounds of human languages.
But where do the symbols come from? In a sense, any shapes would do, pro-
vided they were distinctive and available in sufficient numbers – and some
early attempts at phonetic alphabets did involve complicated invented
shapes.
The IPA takes the familiar Latin alphabet as its starting point. The twenty-
This is a specimen six letters of the alphabet are all used, mostly with values that seem very
of Alexander natural to us, and further symbols are obtained in a variety of ways:
Melville Bell’s
Visible Speech sys-
tem, one of many
• by using small capital letters with different meanings from the lower
case ones; for example, [
] and [] stand for different sounds, so do [n]
phonetic notation
and []. Notice how square brackets are put around phonetic symbols.
systems invented in
the past. Only the • by turning or inverting existing letter shapes, as in [ə], [ɹ] [ɯ] [ʁ]
IPA enjoyed any • by using diacritics, which are dots, hooks, and other small marks
lasting success, and added to symbols, as in [n ], [ã]
it is the only system • by using some letters from the Greek alphabet, such as [φ] and [γ],
in use today. (though the values attached to the symbols are not necessarily Greek
sounds)
• by inventing new shapes, such as [ ], [ ]. The new shapes mostly look
like existing letters, or have some logical feature in them that high-
lights resemblances among sounds.

It’s useful to consider some devices the IPA doesn’t generally use in form-
ing new symbols. Differences between fonts – the mere appearance of a
letter – don’t matter. In particular, it doesn’t matter whether letters have
serifs (little cross strokes at the ends of lines). So [n] (with serifs) and [n]
(without) stand for the same sound. In the same way, italic or bold are
never used to make different symbols.

Normal speech and pathological speech

It’s easy to use IPA
As you see from Figure 1.1, the chart of the IPA fits on one page. It provides
symbols in word-
processed docu- nearly everything that is required to represent speech in any language –
ments on a comput- at least, in the normal adult pronunciations of those languages. But
er. For details on speakers who have speech disorders sometimes employ types of sound dif-
how to do it, see the ferent from those used by any normal speaker. The IPA provides a supple-
website that accom- mentary chart, and further diacritics, which may be needed to cope with
panies this book. the range of sounds encountered in pathological speech.

Transcribing sounds
After phonetic training, we can listen analytically to words or phrases,
imitate them, and record them with symbols. We can start from zero with
a language we do not know at all, transcribing words with the help of a
co-operative speaker who is willing to repeat words for us and listen to our
imitations. For instance, in the first session with a speaker of the language
Divehi (Maldive Islands), the first word asked for was the one for ‘hand’.
The response was transcribed as [aeʔ]. The two symbols [ae] represent a
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 5 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Introduction to speech 5

particular gliding vowel sound (roughly like that in English eye), and [ʔ]
represents a glottal stop. The native speaker of Divehi would not accept A glottal stop (or
more fully a glottal
[ae] without the glottal stop as a version of ‘hand’, which showed that she
plosive) is the speech
regarded the glottal stop as an essential sound in the word.
sound that often
The next word asked for was that for ‘finger’. This time the response replaces the [t] of a
was noted down as [iŋ
iliʔ]. Here the [i] symbols represent vowels approx- word such as better
imately like those in English beat, and again there is a final glottal stop. in a London pro-
The symbol [ŋ] represents a sound which is encountered commonly in nunciation, or the
English but spelled with n in conjunction with other letters (commonly [t] of words such as
k or g). It is the nasal sound before the [ k] of think, or the final sound of football quite
thing for most speakers. Actually the English word finger contains exactly generally in both
the same sequence [ŋ
] noted in the Divehi word. The next word asked for British and
was ‘head’ and this was noted down as [bɔ]. Here the symbol [ɔ] repre- American pronunci-
ation. As we will
sents a vowel somewhat like that in English saw. There seemed to be no
see, it is formed by
glottal stop at the end, but as the other two words noted so far had fin-
a momentary
ished with a glottal stop, the investigator pointed to his head and tried a closure of the vocal
version [bɔʔ]. This produced laughter from the informant. It turns out folds, located in the
that [bɔʔ] is a separate word meaning ‘frog’. This established that the larynx. Within a
glottal stop was not just an automatic termination for Divehi words, but word such as
a significant sound capable of marking distinctions between words. better, a glottal
If this process of imitation and transcription were continued, we could stop is heard as a
expect to encounter all of the vowels and consonants employed in the brief interval of
language, and select suitable symbols for transcribing them all. It makes silence. A glottal
no difference to this process whether the language under investigation stop which follows
a vowel, as in this
happens already to have a written form. As a matter of fact, Divehi is a
Divehi example, is
written language, with a unique alphabet of its own; and the speaker in
heard as an abrupt
question was literate in both Divehi and English – but these facts have no termination of the
relevance to the process of listening, imitating and transcribing. vowel sound.

Types of transcription
Once a phonetic transcription has been made, it should sound right
when we read it aloud again. But people often have a wrong idea about
how precise phonetic transcription can be. The IPA symbols may look a bit
like mathematical symbols, but they are not used with mathematical
precision. Learning to transcribe is not at all like learning formal logic or
algebra – it is more like learning how to make a recognisable sketch of a
face or an object (one kind of transcription is even called impressionistic).
Different observers can make somewhat different transcriptions of the
same sample of speech, without either of them being necessarily ‘wrong’.
And however detailed we make our transcription, it remains only a rough
approximation compared with a sound recording, or a film of the speak-
er’s actions in producing the speech.

Symbols for particular languages

The symbols needed to represent the sounds of any particular language –
whether it’s English or Divehi, or any other – will be a selection from
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 6 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

6 INTRODUCING PHONETIC SCIENCE

(a subset of) the IPA. As an illustration, symbols for representing one type
of English are given in the table below. This is followed by a specimen of
transcription for you to try reading.

Keyword Keyword
p pie j yes
t tie h hat
k key i pea
b buy  pit
d die e pet

guy  pat
m my ɑ pa
n no  pup
ŋ sing ɒ pot
f fee ɔ paw
v van υ put
θ thigh u too
ð though ə soda
s so  bird
z zoo e pay
ʃ she a pie
 measure ɔ boy
tʃ chip aυ now
d jam əυ no
w wet ə near
r red eə hair
l let

The mark  indicates that the following syllable is stressed, and | indicates
a slight pause at the end of a phrase.

ðə nɔθ wnd ən ðə sn wə dspjutŋ wtʃ wəz ðə strɒŋ
ə 

wen ə trvlə kem əlɒŋ  rpt n ə wɔm kləυk 
ðe ə
rid ðət ðə wn hu fst səksidd n mekŋ ðə trvlə
tek z kləυk ɒf  ʃυd b kənsdəd strɒŋ
ə ðən ð ðə 
ðen ðə nɔθ wnd blu əz hɑd əz i kυd  bət ðə mɔ hi blu 
ðə mɔ kləυsl dd ðə trvlə fəυld z kləυk əraυnd hm  ən ət
lɑst ðə nɔθ wnd 
ev p ð ətempt  ðen ðə sn ʃɒn aυt
wɔml  ən midjətl ðə trvlə tυk ɒf z kləυk  ən səυ ðə nɔθ
wnd wəz əbladd tə kənfes  ðət ðə sn wəz ðə strɒŋ
ər əv
ðə tu 

(The orthographic version of this text is given at the end of the answers
section for this chapter.)
The IPA does not provide fixed transcription systems for particular lan-
guages. It provides a stock of symbols, and principles and conventions for
using them – but there can be perfectly legitimate differences between
transcription systems for one and the same language. As a simple illus-
tration, consider the vowel in a Southern British pronunciation of a word
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 7 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Introduction to speech 7

such as dress. It is somewhere between the sounds represented [e] and [ε]
in the IPA. We could use a diacritic added to one of these symbols to show
an intermediate quality – for instance [e]. But how inconvenient would it
be, all the way through a book on English pronunciation, or in a diction-
ary that shows pronunciations, to add the diacritic each time? Far better
to choose either [e] or [ε] for regular use, and give a once-for-all statement
that the actual quality is something like [e]. In fact, some transcription
systems for English use [e] while others use [ε].

Syllables
Though both speaker and listener may have the impression that speech is a
sequence of sounds, the shortest stretch of speech that a speaker can actual-
ly pronounce in a fairly natural way is not the individual sound, but the syl-
lable. If a person is asked to speak very slowly, splitting words up into sec-
tions (e.g. for dictation), division will usually be into syllables. Thus the word
signal can be spoken as two chunks separated by a pause: sig - nal. This is
because signal is a word of two syllables. By contrast the word sign is a one-syl-
lable word (a monosyllable). It cannot be divided into two individually pro-
nounceable parts. It begins like sigh, but the remainder is the single sound
represented by n, which we can only pronounce in an unnatural and dis-
jointed way. The middle part of sign is also clearly the same as I or eye, but
removing that portion leaves us with two sounds represented by s and n, sep-
arated by a gap. The conclusion is that sign, sigh and eye are all monosyllables.
A syllable is like one pulse of speech. It always contains one loud or promi-
nent part (almost always a vowel sound), and may optionally have consonant
sounds preceding or following the vowel. If we compare the pronunciations
of the three syllables be, eat and beat we can hear that they all contain the
same vowel sound, which we can represent with its phonetic symbol [i]. In be,
the vowel is preceded by a consonant sound [b], but nothing comes after the
vowel, giving [bi]. In eat there is nothing before the vowel, but a consonant
sound [t] follows. Beat [bit] has both preceding and following consonants.

Segments: vowels and consonants

The term segment is another way of referring to the individual speech
sounds that make up syllables. Segments are of two kinds: vowels and
consonants. Typical vowel segments are [i a u]; a few examples of typical
consonants are [m b k f s]. Using V to stand for any vowel and C for any con-
sonant, the structure of a syllable or word can be shown as a string of Vs and
Cs. So, for example, the word book, pronounced [bυk], is CVC. This means a
sequence of one consonant followed by one vowel which in turn is followed
by one consonant. This sort of representation is called a CV-skeleton.
Here are some examples of CV-skeletons for English, together with some
words that conform to each of them. Remember that we are dealing with
pronunciations, not spellings. The double -oo- in the conventional spelling
of book doesn’t mean that the word contains two vowel sounds. To take
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 8 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

8 INTRODUCING PHONETIC SCIENCE

V eye, oh

CV be, my, see, saw, tea, you

VC eat, each, aim

CVC book, chip, thumb, top, win

CCV draw, glue, stay

CVCCCC texts

CCCV straw

CVCVC unit

CVCCVC signal

CVVC going

another example, look at the skeleton for unit. This word is spelt with a
vowel letter at the beginning, but it is pronounced with a consonant seg-
ment in initial position, exactly like the word you. Notice also that in win
and you the letters w and y stand for consonant sounds, whereas in my, saw,
draw, stay, straw they are used as part of the representation of the vowel.
As we see in these examples, when we represent careful pronuncia-
tions of whole words with CV-skeletons, there must be one V element for
each syllable. The three-syllable word banana has the skeleton CVCVCV.
In addition to simple vowels, like those heard in book, bit or cat, there are
also diphthongs, which are vowels of changing or ‘gliding’ quality like
those heard in voice or house. Since voice and house are one-syllable words,
English diphthongs must count as one V element rather than two. Both
voice and house have the structure CVC, rather than CVVC. But going has
two syllables: one is the stem go, which contains a diphthong, the sec-
ond is the ending -ing, which contains a further vowel followed by the
consonant [ŋ].
Some languages also permit certain consonants to be syllabic – that is,
to form a syllable by themselves. In English, [l] and [n] may be syllabic, as in
the second syllables of certain pronunciations of settle [setl!] or sudden
[sdn!]. (The IPA diacritic added to the relevant consonants means ‘syllabic’.)
These words can be represented CVCC!. But notice that there are also alter-
native pronunciations [setəl], [sdən] in which the second syllables of
these words contain a V element followed by a non-syllabic consonant, giv-
ing the structure CVCVC.

Syllables and words

In a sense, spoken words are composed of syllables. The shortest possible
words are words of one syllable. The English words hand, arm, head, eye,
mouth are all monosyllables, as are have, be, go, do, make, eat, die. Of course,
English words may have two, three or more syllables. For example, the
English words mother, husband, river, heaven, berry have two syllables, while
description, musical, prominent have three, applicable has four, characteristic
has five, and so on.
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 9 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Introduction to speech 9

But syllables are units of pronunciation rather than elements of word

structure. Notice that the word dog is a monosyllable, and so is its plural, While English has
plenty of basic
dogs; but the plural clearly consists of two elements of word structure: one
vocabulary items
is the stem, dog-; the other is the ending indicating ‘plural’. There are thus
that are single sylla-
two elements of word-structure within the one syllable. On the other bles, not all lan-
hand, banana has three syllables, but just one element of word structure guages actually per-
(a stem), as the word isn’t made up of separate meaningful parts. mit monosyllabic
words. In most indi-
Suprasegmentals genous Australian
languages, for
Segments aren’t the whole story. We also have to pay attention to features example, words
that are not themselves segments, and that seem to spread across several have to have at least
successive segments (often a whole syllable). Such properties are called two syllables, follow-
ing the formula
suprasegmentals. Stress and tone are in this category. In English, for
CV(C)CV(C). The
example, the noun import and the verb import have exactly the same seg-
brackets show that
ments. But the words are distinguished according to which of the two syl- the consonants in
lables is stressed: the noun is IMport but the verb is imPORT. Here we’ve those positions may
used capitalisation to give an indication of which syllable is stressed, but be present, but are
as you will see from the chart, the IPA provides symbols for suprasegmen- not obligatory. So
tals too, and we will return to these in a later chapter. the formula covers
CVCV, CVCCVC,
Speech as an acoustic signal CVCCV and
CVCVC.
Like any sound, speech can be picked up with a microphone, recorded
and analysed. Sound is a rapid variation of pressure travelling through
some physical medium (such as air). The velocity of sound in air is about
330 metres per second (about 740 miles per hour). When variations in
pressure arrive at the eardrum, or at a microphone, they cause vibration
(tiny to-and-fro movements) of the eardrum or the diaphragm of the
microphone. The human listener experiences hearing a sound, and the
microphone produces an electrical signal which can be measured.
A graph showing variation of pressure (or equivalently, movements of
the eardrum) as a function of time gives the waveform of a sound. In a
perfectly quiet place, a microphone picks up no sound, and the resulting
waveform will be flat, showing no up and down movement at all. In most
locations there will generally be background or ambient noise (traffic or
aircraft noise from outside, wind, sounds from appliances, and so on) and
this will show as constantly fluctuating energy on a waveform. Speaking
reasonably close to a microphone results in a waveform that is much big-
ger than the ambient noise level.
Because the pressure variations in sound are very rapid, the amount of
detail we can see in a waveform depends on the scale on which it is plot-
ted. If a lot of time is shown in a small space, the details of the waveform
are lost, and we just see blocks of activity.
The amplitude of a wave is a measure of the size of the pressure varia-
tions (or eardrum movements). The auditory property that is correlated
with amplitude is loudness. Other things being equal, a wave with larger
variation in air pressure will correspond to a louder sound. In Figure 1.2 we
can see that the vowels have more energy (are louder) than the consonants.
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 10 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

10 INTRODUCING PHONETIC SCIENCE

FIGURE 1.2
The waveform for Peter Piper picked a peck of pickled pepper. The phrase has been
specially chosen so that the twelve syllables of the utterance can be counted. The
consonant sounds that separate the syllables have low amplitude, making the
prominent sounds at the centre of the syllables easy to see.

Zooming in on the waveform, so that only a small fraction of a second is

shown, reveals the detailed pressure variations and can tell us something
about the sound of individual segments. It is generally convenient to
measure time not in whole seconds, but thousandths of a second. One
millisecond (ms)
1/1000 second.

FIGURE 1.3
A short sample (about 22 milliseconds long) from the first vowel of Peter in the utter-
ance shown above. The time scale shows milliseconds from the start of the recording.
At this point, the waveform has a regular repeating pattern. Listening to this sample, we
hear a very short but recognisable vowel [i].

Writing systems
The writing systems used by the languages of the world are many and var-
ious and it needs a whole book considerably larger than this one to deal
with them in detail. We will give a simple account here, which should give
you an idea of the amount of diversity in writing systems. Ways of writing
fall into three basic categories: (1) alphabetic systems, (2) syllabaries and
(3) logographic systems.

Alphabets
If you are reading this book, you are of course familiar with at least one
alphabetic writing system. The writing system of English is a development
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 11 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Introduction to speech 11

of the Latin alphabet, as are many writing systems used for languages of
Europe and in many other parts of the world. Each of these languages has
adapted the alphabet used in classical Rome to suit its own needs. Some
languages have needed to add letters to the alphabet to suit their needs –
English, for instance, has added w.
Languages often differ in the value, or range of values, given to a par-
ticular letter. So, for instance, the English letter c represents the sounds
[s] as in cinema or [k] as in cow, whereas in the writing system of Zulu
the same letter represents a click sound. In English, x commonly rep-
resents a sequence of consonants [ks], as in six or tax, but in some lan-
guages (Maltese is an example) the same letter is used to represent the
sound [ʃ ], which is the consonant at the beginning of the English word
shop.
There are alphabetic writing systems that have not come directly from
the Latin alphabet. You may be familiar with the Greek alphabet:

 or with the Cyrillic alphabet, which is a
development of it, used to write Russian and some other Slavonic languages:
aбrдeёЖ3ИЙKлMHOпpCTYфXцчшщъыьэюя. Some of the letters in both of these
resemble Latin letters. However, there are alphabets used around the world
where the letters look very different from the Latin ones. Here is part of the
alphabet used to write modern Hebrew: The direction
of writing can vary, too: the writing system of Hebrew and some other lan-
guages such as Arabic runs from right to left.
Despite their very different appearances, all alphabets are thought to be
developments of just one. And all alphabetic writing systems have in
common the fact that by and large each letter is an attempt to represent Writings in an
a segment. Clearly, this is exactly the same principle that is used for pho- unknown script
netic transcription, and the IPA can be thought of as a development of the were found on clay
roman alphabet, adapted and extended to represent the sounds of all tablets during an
languages. archaeological exca-
vation on the island
Syllabaries of Crete around
If spoken language consists of syllables, one way to write a language is 1900, and termed
‘Linear B’. No-one
to provide one symbol for each possible syllable. This is what is done in
could decipher
the type of writing system called a syllabary. Many different syllabaries
them, or even be
have evolved at different times in different parts of the world. A portion sure what language
of the Linear B syllabary, used to write an early form of Greek about 1500 they were in, until
BC, is shown in Figure 1.4. You can see that most of the symbols repre- the 1950s, when an
sent a CV syllable. But the symbols that represent syllables sharing the Englishman called
same vowel do not appear to resemble each other. The same is true for Michael Ventris
symbols that represent syllables that share a consonant. So the symbol showed that attach-
representing [di] looks nothing like that representing [ki] and the sym- ing certain syllabic
bol for [ka] is completely different to that for [ku]. Each symbol therefore values to the marks
represents a complete syllable, which is not analysed into its component led to readings that
were overwhelming-
segments.
ly likely to be an
A syllabary works well if the total number of different syllables that
archaic form of
need to be represented is reasonably small – say of the order of 100. Some Greek.
languages do have highly constrained syllable inventories and hence lend
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 12 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

12 INTRODUCING PHONETIC SCIENCE

FIGURE 1.4
Part of the Linear B syllabary, used to write an archaic form of Greek about 1500 BC
(from G. Sampson, Writing Systems, London: Hutchinson, 1985).

themselves to representation in this way. But the number of possible

English syllables is in tens of thousands; and the number of different syl-
lables that a human being can in principle make or recognise is perhaps
of the order of a million. So syllabic writing is not suitable for all lan-
guages, and would certainly not form a satisfactory basis for a general sys-
tem of phonetic transcription.

Logographic writing
A logographic writing system does not attempt to represent the pronun-
ciation of words at all. The best known example is the one used in Chinese.
Each logogram (or character) represents a complete word and basically it
is up to the reader to know how the word is pronounced. There is just no
way of working out the pronunciation from the logogram. To illustrate
this here are two Chinese characters:

tree, pronounced [mu]
root, pronounced [bən]
It is easy to see the connection between the meanings of the two words and
why the characters resemble each other, but the pronunciations have
absolutely nothing in common. This means that there really is no equivalent
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 13 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Introduction to speech 13

of spelling in this type of system. As a further illustration, here is the char-

acter meaning good. It is pronounced [hau].

The left-hand side of this exists as a character in its own right. It means
woman and is pronounced in a way that can be transcribed [ny] (the sym-
bol [y] means a vowel like that in French lune meaning ‘moon’). The right-
hand side of the character can also be used independently. It means child
and is pronounced [dz]. So there is no way of working out that good

[hau] in Chinese from the character used to represent the word.
Finally, there are writing systems that are mixtures of the above types.
Ancient Egyptian hieroglyphics are a mixture of alphabetic writing and
logograms. Modern Japanese is written with a mixture of Chinese logograms
and two different syllabaries.

The relationship between speech and writing

If we think of a language as a system for communication, then speech and
writing are the two main channels through which communication can
pass. Speech is the original channel for which human language evolved,
and all written languages have (or once had) a spoken form. Writing, the
preservation of language messages with visible marks, is a much later
development (from about the Bronze Age) and many languages today still
have no written form.
Speech and writing are obviously good for different things. Speech is fast
and interactive, and we can talk and listen while at the same time getting
on with something else (such as working or eating). Talking is a social activ-
ity, and when we talk to other people we are doing much more than com-
municate linguistic messages. But like all sounds reaching our ears, speech
is fleeting. Writing is permanent, and gives human beings the power to
keep records. It also helps us organise ideas that would be too complex to
keep in our heads at once. Writing is in fact the most fundamental inven-
tion in the field we now call information technology.
People who do not make use of writing can pass stories, songs and
poems from one generation to the next by oral tradition – that is, by mem-
orisation and repetition. But in our society oral tradition has just about
died out, because it is assumed that everyone will learn to read and write.
Writing is used for law, religion, literature and science, and as a result has
become the basis of almost all education.
Within limits, we can swap messages between speech and writing. For
instance, we can write down the words somebody used when they were
speaking. This may preserve the ‘literal meaning’ of what was said, but we
will lose the intonation, loudness, voice quality, local accent – and the
information that enables us to tell whether the speaker is a man, woman
or child, and whether the speaker is someone we know. Equally, we can
make an attempt at speaking aloud anything that is written – but here we
have to supply all the characteristics that are not specified in writing. And
there certainly isn’t a spoken equivalent for everything that’s written.
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 14 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

14 INTRODUCING PHONETIC SCIENCE

Taking away the visual organisation of a written document may make it

more or less useless: think of the difficulties we would face in using a
purely spoken version of something like a telephone directory.
Writing is not simply a representation of speech. Admittedly, speech
was probably an important factor in the original development of writing,
and it remains true that in most writing systems, the way to write a word
has some connection with its pronunciation. But once writing has
become established, it tends to develop all kinds of peculiarities that do
not directly reflect speech, and which in fact have very little to do with
the language system either. The insistence on standardised spelling, punc-
tuation and the use of capital letters in the writing of English are exam-
ples. A schoolchild who writes beleeve will have the spelling crossed out
and corrected, even though it indicates the intended word quite as clearly
as the conventional spelling believe (and was in fact used in the first print-
ed version of Shakespeare’s plays). The insistence on one spelling rather
than the other, and on a single spelling rather than two or more spellings
used interchangeably, is simply arbitrary. Conventions like this seem very
important to most educated people (perhaps because they have had to
work very hard to learn them) but they are entirely artificial.
We have just shown that writing is not simply a representation of
speech. Much more obviously, speech is not in any sense a version of writ-
ing. People sometimes have quite powerful prejudices, believing that
speech should really be modelled on written language, or that spelling
somehow shows us how we should pronounce. We need to keep spoken
language and written language separate in our minds, and study each on
its own terms.

Homophones
A good example of the independence of speech and writing in the repre-
sentation of language comes from looking at homophones (different
words that are pronounced the same way). Consider, for example, the
English words that are conventionally written see and sea. They are pro-
nounced in exactly the same way [si]. Without the assistance of a mean-
ingful context the spoken words are completely indistinguishable. We can
go further, because there are at least two English words spelt see. One is the
verb meaning ‘to perceive with the eyes’, the other a noun meaning ‘the
Homophones are of area of authority of a bishop’. (These two words have entirely separate ori-
course not just gins.) The same pronunciation is also used for the name of the letter C,
found in English. A which is also used as a word (I tried hard in my exam, but I only got a C). So the
simple example language has four words: see (verb), see (noun), sea and C. The four words are
from French is vin absolutely identical in speech. The writing system makes a three-way dis-
meaning ‘wine’ and tinction, which may be useful, but has nothing to do with pronunciation.
vin
t meaning Homophones are found scattered randomly – but fairly commonly – in
‘twenty’ both are the vocabularies of languages. They are certainly not restricted to languages
pronounced [vε "]
" that happen to have a written form. If we were learning any unwritten lan-
(the diacritic [ ]
guage we would almost certainly come across some examples before very
means a nasalised
vowel). long. Of course, young children meet and cope with homophones such as
see/sea or vin/vingt long before they learn to read and write.
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 15 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Introduction to speech 15

Sign languages
We have described speech and writing as the two channels for language.
Don’t the sign languages used by the deaf show us a third channel?
Certainly signing represents another mode of communication, which is
neither speech nor writing. But sign languages (such as British Sign
Language, BSL) don’t operate by giving a word-for-word version of an oth-
erwise written or spoken language (such as English). The system underly-
ing the signs is a language in its own right. A message that has been orig-
inated in BSL has to be translated into English, not merely transposed into
speech or writing, to make it accessible to a person who doesn’t know BSL.
So signing isn’t an extra channel for the thousands of conventional lan-
guages like English, which have spoken (and possibly written) forms.

Problems with syllables and segments

Although segments and syllables are important in the description of
speech, it is only fair to point out that there are some problems in defin-
ing and making use of these notions.
Sometimes the number of syllables in a word is uncertain or variable.
For example, the English word science may seem to be two syllables or one.
It can be pronounced in a way that clearly has two syllables, or in another
that seems to have only one. There are also various intermediate pronun-
ciations over which people would not agree. The problem in words like
this is caused by adjacent vowel sounds without a consonant to separate
them. Adding the consonant [l] between the two vowels (and keeping the
other sounds the same) gives the word silence, and we have no doubt that
this has two syllables. Sometimes there are alternative pronunciations of
a word, which differ in the number of syllables. For example camera can be
pronounced as three syllables or two. The word prison has two syllables,
but prisoner can be pronounced as three or as two. When we hear particu-
lar examples of words such as camera or prisoner we can’t always decide
whether we hear two syllables or three.
Our perception of syllables is partly governed by our language experi-
ence. To a Japanese listener, the English syllable [bi] (be /bee) seems to con-
sist of two units rather than one, because it contains a vowel that is long
in duration. The first unit is a syllable [bi] with a short example of vowel
[i], the second unit is another [i], needed to account for the length of the
vowel. So the Japanese perception is [bi – i]. Add a consonant to give beat
and a Japanese listener now perceives three units: [bi – i – tu], because syl-
lables in Japanese cannot end with a consonant such as [t].
Some caution is needed with segments, too. Although they are essential
for phonetic description and transcription, we should beware of thinking
that segments are ‘real’ in any simple sense, or that they necessarily form
the basis of the production and perception of speech. Although recordings
of speech can be cut up reasonably well into syllables, the same is not true
for segments. And the movements we make in producing speech are smooth
and continuous, without abrupt joins between successive segments. Some
people even go so far as to claim that segments are complete illusions
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 16 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

16 INTRODUCING PHONETIC SCIENCE

which we believe in only because of the alphabet. It is certainly true that

when we hear a stretch of speech in an unfamiliar language we may have
real doubts about what segments it contains. The same is true if we hear
words of a familiar language (such as English) spoken by a young child, or
by an adult who has suffered a stroke. And if short snippets of casual
speech are taken out of context and played to listeners, the listeners make
wild attempts to make sense of what they hear and differ greatly from each
other in how many segments they seem to hear and what they are.

Chapter summary
Phonetic science is concerned with the objective description and analy-
sis of all aspects of speech. The representation of speech depends upon
treating it as a succession of sounds. The International Phonetic
Alphabet (IPA) aims to provide a symbol for any sound in any language.
The sounds in speech are of two kinds: vowels and consonants. Vowels
may be simple or diphthongal. Segments are grouped into syllables,
which are the smallest units that can be pronounced in a natural way.
Recording speech with a microphone enables us to display a waveform,
on which certain aspects of speech can be observed and measured.
Speech and writing are separate channels through which language
messages can be passed. The two channels have different uses and dif-
ferent characteristics. Writing is important in our culture but we must
avoid the mistake of believing that written language can somehow
show us what speech is really like (or how it ought to be).

Exercises
Exercise 1.1
How many segments have the following words? Represent each word as a
string of Xs, one X for each segment. (Remember you are looking for sounds,
not letters.)
1 box
2 sorry
3 possess
4 knees
5 quickly
6 rhyme
7 climb
8 gnaw
9 mushroom
10 elephants

Exercise 1.2
Represent the following English words using the symbols C and V to make
CV-skeletons.
1 damp
2 fox
0521808820c01_p01_20.qxd 1/27/05 2:55 PM Page 17 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Introduction to speech 17

3 friend
4 unit
5 talker
6 physics
7 knowledge
8 spaghetti
9 columnist
10 ghastliness

Exercise 1.3
For each of the CV-skeletons you have found in 1.2, find at least one other
English word that will fit it.

Exercise 1.4
Examine the short English passage below and then list all the words you can
find in it that conform to each of the CV-skeletons given. Treat the sound [tʃ]
as at the beginning of chew, and the sound [d] as at the beginning of jazz,
as single C elements rather than CC.
He tried to alter the height of the ladder, but it still would not reach the win-
dow. ‘You aren’t having a lot of success, are you?’ said the old man. Colin
shook his head and smiled. He patted his pockets again, but he knew he had
lost the keys. He would have to try and break in, but he wanted to avoid
causing too much damage.
1 CV
2 CVC
3 CVCC
4 VC
5 VCC
6 CVCV
7 CCVC
8 CCVCC
9 CVCVC
10 CVCCVC

Exercise 1.5
Here is a short passage of English. Determine the number of segments in
each word and count the number of words with one segment, with two
segments and so on. Then draw a bar chart with number of segments on the
horizontal axis and number of words on the vertical axis. Keep a record of
any words where the number of segments is problematic.
The first time I drove a car abroad was in Italy. We picked up a car at the airport
and drove to where we were staying. It took us about five and a half hours. I
had a bit of trouble with the gears and spent the first couple of miles stuck in
third, but after I got the hang of it, we got along fine. The next day was a differ-
ent matter. I suppose it was overconfidence or something, but we nearly had a
terrible disaster. We set off to see some sight or other, I forget now where we
were going. Suddenly my wife screamed out that I was driving on the wrong
side of the road. I managed to get back to the correct side just in time to miss
a large lorry which appeared round the bend in front of us. It would have been
a very nasty accident and I was extremely careful for the rest of the holiday.
0521808820c01_p01_20.qxd 1/27/05 2:56 PM Page 18 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

18 INTRODUCING PHONETIC SCIENCE

Exercise 1.6
Use the passage in exercise 1.5 (or take 100 consecutive words from any
other text or from a sample of recorded speech). Determine the number of
words pronounced with one syllable, two syllables, etc. Keep a list of problem-
atic cases. Again, draw a bar chart to show the results. Try this with texts and
recordings of different types and styles.
Exercise 1.7
Try to make up a short passage of English using only monosyllables. There is
actually at least one complete novel written this way (Launcelot Hogben,
Whales for the Welsh, London: Rapp and Carroll, 1967). Here is a specimen:
All the same, he was not a bad clerk as clerks go, and as clerks go he went
home in good time each night to the small house where he slept and took
his one meal of the day. There he could cash in on the thoughts which came
to him at his desk . . .
(p. 7)
Exercise 1.8
[Needs access to a computer, with soundcard and microphone.] Using
Windows Sound Recorder, record some samples of speech into the computer
and replay them. Record a stretch of speech that does not have long pauses
in it. Sound Recorder shows you the duration of a recording (in seconds). By
counting the number of syllables in the recording, you can work out the rate
of speaking, expressed in syllables per second. In a similar way, you can arrive
at an estimation of the number of segments produced per second. [Note:
Sound Recorder records and replays sound perfectly well, but the ‘wave-
forms’ it displays are not accurate. Software to display waveforms correctly
can be downloaded free. See website link.]
Exercise 1.9
Two speech waveforms A and B are shown below. One is potato, while the
other is pepper. Say which is which, and give reasons for your choice.

Waveform A

Waveform B
0521808820c01_p01_20.qxd 1/27/05 2:56 PM Page 19 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Introduction to speech 19

Further reading
Most of the topics introduced in this chapter will be considered in greater depth later in the book. If you’re
just beginning the study of phonetics, the best advice on further reading, once you’ve tried the exercises to
this chapter and looked at the accompanying website, is to continue with the other chapters. You may like to
look at the Handbook of the International Phonetic Association, Cambridge: Cambridge University Press,
1999, for a comprehensive guide to the IPA, and examples of its application to around thirty languages.
Some of the background areas we have touched on in this introduction won’t come up again, and we
wouldn’t recommend getting sidetracked onto them at this early stage. But you may at some point in the
future want to know more about them. On writing systems, we’d recommend Florian Coulmas, Writing sys-
tems: An introduction to their linguistic analysis, Cambridge: Cambridge University Press, 2003. You can learn
about the spelling system of English in Edward Carney, English spelling, London: Routledge, 1997. If you are
interested in the decipherment of old scripts like Linear B, a good start is Maurice Pope, The story of
decipherment: From hieroglyphs to Maya script, London: Thames and Hudson, rev. edn 1999.
0521808820c01_p01_20.qxd 1/27/05 2:56 PM Page 20 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:
0521808820c02_p21_32.qxd 1/27/05 2:56 PM Page 21 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

CHAPTER

2 Voice

CHAPTER OUTLINE
In this chapter you will learn about: the structure of the
KEY TERMS larynx; how the larynx is used to produce voiced and
Aperiodic voiceless sounds; how larynx activity can be observed and
Complex tones monitored; the role of voice in the languages of the world;
Cycle the waveforms of voiced and voiceless sounds.
Frequency
Hertz
Larynx
Periodic
Pitch
Sine wave
Voiced
Voiceless
Whisper
0521808820c02_p21_32.qxd 1/27/05 2:56 PM Page 22 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

22 INTRODUCING PHONETIC SCIENCE

Introduction
The basis of all normal speech is a controlled outflow of air from the
lungs. Air flows up the trachea (windpipe) and out of the body through
the mouth or the nose. On the way it must pass through the larynx, a
structure formed of cartilages and visible on the outside of the neck as the
‘Adam’s apple’. The airway from the larynx to the lips, and the side-branch
via nasal cavities to the nostrils, contain all the organs that control the
production of speech sounds, and are known as the vocal tract.

The larynx
We cannot see directly into the larynx, but it can be observed in a mirror
placed right at the back of the mouth (a laryngoscope mirror) – for this,
the subject must keep the mouth wide open. Another way is with a
fibrescope, which can be inserted via the nose and does not prevent the
subject from speaking. Both methods give a top view of the larynx, and
this is how it is usually shown in pictures and diagrams.
The cartilages of the larynx together form a short section of tube through
which air must flow on its way to and from the lungs. The two main carti-
lages are the cricoid (which is a ring-shaped cartilage at the base) and the
thyroid cartilage, which forms the prominence which can be seen and felt
in the neck. The larynxes of males and females are essentially similar,
though the male larynx is generally larger and more prominent.

Vocal folds
Contained within the cartilages are the two vocal folds (vocal cords in
older terminology), one at the left and one at the right. Looking down on
the larynx from above we see the top surfaces of the folds. The folds are
fixed adjacent to each other at the front, but at the back they are attached
to the two moveable arytenoid cartilages. If the back ends of the folds are
held apart, a roughly triangular space opens up between the folds. The
space between the folds is called the glottis. As the back ends of the folds
are brought together, the glottis can be narrowed to a slit, and then closed
completely as the folds are pressed together. In this position, they prevent
any air flowing through the larynx.
For normal breathing, the glottis is open and air passes in and out
silently. For certain speech sounds, too, the glottis is open. For the sound
[f ], for example, air flows out rapidly through the open glottis (but is
obstructed as it leaves the vocal tract through a narrow gap between
upper teeth and lower lip).

Voiced and voiceless sounds

There is another adjustment of the glottis that is of great importance in the
production of speech. If the vocal folds are held gently together and air
0521808820c02_p21_32.qxd 1/27/05 2:56 PM Page 23 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Voice 23

under pressure from the lungs is pushed between them, the folds can be
made to vibrate evenly to produce the tone we call voice. If we sustain a vowel
sound such as [i] or [a], the tone we hear is contributed by the vibrating folds.
A sound of this type is said to be voiced. The rapid vibration of the folds can
be felt with fingertips applied to the neck on the outside of the larynx. It is
not only vowels that are voiced, but many consonants such as [m] [v] [z]. If we
sing a vowel, or one of these voiced consonants, we can change the pitch of
the note we are singing by varying the rate of vocal fold vibration.
Sounds like [f] and [s], which are made without vocal fold vibration, are
said to be voiceless. The airflow is converted into sound energy not at the
larynx, but elsewhere in the vocal tract. There is no tone from the larynx,
and nothing will be felt with fingertips applied there.
All vowels are normally voiced. Consonant sounds come in both voiced
and voiceless kinds, and in fact voicing versus voicelessness is the only dif-
ference between certain pairs of consonants. For example, comparing [f]
and [v] we find that the only difference between them is that [v] is voiced
while [f] is voiceless. The position of the rest of the vocal tract is the same.
With a little practice, it is possible to hold the position for [f] and contin-
ue to produce sound, while switching voicing on and off. This gives a
sequence [fvfv . . . ]. The sounds [s] and [z] are related in the same way and
a similar alternating sequence [szsz] should be tried.

Vibration of the vocal folds

The way in which the vocal folds vibrate for voice is quite similar to the
way that the lips can be made to vibrate by blowing air between them (the
lips are made to vibrate this way by the player of a brass instrument such
as a trumpet or trombone, or by anyone making the sound known as a

Thyroid cartilage
Cricothyroid
muscle

FRONT
Vocal folds

Cricoid Glottis
cartilage

Trachea

FIGURE 2.1
The cartilages of the larynx, seen from the FIGURE 2.2
right-hand side. View of larynx from above.
0521808820c02_p21_32.qxd 1/27/05 2:56 PM Page 24 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

24 INTRODUCING PHONETIC SCIENCE

‘raspberry’). The series of events for one complete cycle of vibration of the
vocal folds is as follows:

• First the vocal folds are together, and stop the airflow.
• Air from beneath pushes up between the folds, forcing them apart
near the middle.
• A burst of air flows through, but this begins to be cut off as the folds
recoil back to the closed position.
• As the opening gets smaller, the rapid airflow through the narrowing
gap leads to suction which helps to complete the closure rapidly and
effectively.
• Once the folds are closed completely, the cycle of vibration begins
again, and the folds are once again forced open.

It is important to understand that the openings which appear as part of the

cycle of vibration at not at all like the stationary open glottis position that
is obtained by moving the arytenoid cartilages apart. During voicing, the
arytenoids remain still and close together. The process of vibration will
repeat for as long as air under pressure is supplied from the lungs, and as
long as the tension and position of the folds is suitable.
This is a simplified account of vocal fold vibration. Because the folds have
vertical thickness, they do not necessarily close face-to-face all at once.
Commonly they close at the bottom first, and the closure then rolls up
through the folds. And the ‘closure’ doesn’t have to be totally complete –
vibration can take the form of alternating more-open and more-closed
positions.

Observing vocal fold vibration

It is very useful for many purposes to observe what actually happens when
the vocal folds are vibrating. For instance, if a speaker is having trouble
with producing vocal fold vibration normally, observation of the vibratory
pattern may give a clue to what is wrong with the larynx. Also, as we shall
see in Chapter 10, some languages use different types of vocal fold vibra-
tion to signal the difference between one word and another. It is obviously
a good idea to see how these differences are actually produced. We will
now have a look at two techniques for observing vocal fold vibration.

The Laryngograph
Vocal fold vibration can be detected and measured with a device called the
laryngograph. It works by passing a tiny high frequency electric current
across the larynx from one fold to the other. The degree of contact
between the folds controls the current, which can be measured and
turned into a waveform showing the vibration. The device is completely
non-invasive and harmless: the electrodes are simply placed comfortably
on the outside of the neck, either side of the thyroid cartilage.
0521808820c02_p21_32.qxd 1/27/05 2:56 PM Page 25 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Voice 25

FIGURE 2.3
Laryngograph (lower) and microphone waveforms.

In Figure 2.3, the lower waveform, obtained with a laryngograph, shows

five closures of the vocal folds from a stretch of voicing. Up corresponds
to contact, down to opening. If heard, this wave is just a buzzing sound.
The upper simultaneous waveform is from a microphone and shows the
result of modification of the voicing tone by the rest of the vocal tract.
This sounds like the vowel of bird, which is the voiced sound that the
speaker was producing. Notice that the two waveforms have the same rep-
etition rate. This rate, as we shall see below, is what determines the pitch
of the sound we hear. As the repetition rate of the speech waveform (the
output of the vocal tract) is identical with that of the laryngograph wave-
form (the contact pattern of the vibrating vocal folds and therefore relat-
ed to the input to the vocal tract), we must conclude that the rate of vocal
fold vibration determines the pitch of voiced sounds.

Stroboscope
The vibrating vocal folds can be filmed, but a very high film speed is
needed to show the details of vibration (ordinary film runs at 24 pictures
0521808820c02_p21_32.qxd 1/27/05 2:56 PM Page 26 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

26 INTRODUCING PHONETIC SCIENCE

per second, and even a low-pitched male speaker will have completed 2 or
3 vibrations in 1/24 of a second). An alternative is to use a stroboscope – a
flashing light of the sort used to slow down or freeze the appearance of
rotating machinery – together with a suitable camera. A refinement is to
trigger a stroboscope from the laryngograph waveform, which allows us
to see the appearance of the folds at precisely determined points in the
waveform (see Figure 2.4).

Voicing in languages
Voiced and voiceless sounds are to be heard in all languages, and for the
most part languages make use of differences between voiced and voiceless
consonants like that already pointed out for English. There are, however,
certain languages where the voicing or voicelessness of consonants can’t
make a difference to meaning, because it is largely an automatic conse-
quence of the position of a consonant within a word. Many indigenous
languages of Australia are like this. In the language Dyirbal, for example,
the word for ‘stone’ is generally [tiban] with a voiceless consonant at the
beginning, and a voiced one within the word. But the initial sound can be
made voiced with no change of meaning, giving [diban], and, though it is
less usual, the consonant within the word can be voiceless, giving two fur-
ther versions [tipan] and [dipan]. It is not unusual for children acquiring
English speech to go through a stage of context-sensitive voicing which
resembles this.

Whisper
When we whisper a sound, there is no vocal fold vibration, and so no
musical pitch. But whisper is not the same as voicelessness. To whisper a
sound, we make a narrow opening between the folds (the glottis becomes
a small opening) and air flows noisily through that opening. The noise
from airflow takes the place of the tone produced by voicing, and is
applied to just those speech sounds that would normally be voiced. If we
whisper a whole utterance, what we do is to use the whisper position for
the voiced sounds, and the voiceless position for the voiceless ones, so we
are still switching back and forth between two adjustments of the larynx.
To prove that whisper and voicelessness are different, try comparing the
sequences [afa] and [ava]. In the ordinary way, they are distinct because [f]
is voiceless and [v] is voiced. But are they still different when whispered?
There seems to be a clearly audible difference, so the attempt at [v] which
turns up in whispered speech cannot be identical with the voiceless sound
[f]. In the whispered [v], there is a narrowed glottis, and airflow noise is
being generated there as well as at the lip-teeth constriction. But for [f],
whether it occurs in ordinary speech or in a whispered context, the glot-
tis is open allowing air to flow freely.
0521808820c02_p21_32.qxd 1/27/05 2:56 PM Page 27 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Voice 27

contact phase

FIGURE 2.4
Stroboscopic photographs
showing two instants
open phase
during the vocal fold
vibratory cycle. The front
of the larynx is at the
bottom of the pictures.
Compare the
laryngograph waveform
with that shown in
Figure 2.2 (courtesy of
Laryngograph Ltd).
0521808820c02_p21_32.qxd 1/27/05 2:56 PM Page 28 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

28 INTRODUCING PHONETIC SCIENCE

Periodic waves
The waveforms of voiced and voiceless sounds are quite different. A wave
that has a pattern that repeats regularly in time is called periodic. A
period runs from one clearly identifiable point on the wave (e.g. an
upwards-going zero-crossing) to the next place where this point occurs. So
one period of a simple (sine) wave contains one upwards-and-over excur-
sion, and one downwards-and-up-again excursion, returning to the zero
line. The length of one period is the periodic time, T.

FIGURE 2.5
Part of a periodic wave
obtained by sounding a
tuning fork close to a
microphone.

The source of periodic waveforms in speech is voicing. So whenever we see

a speech waveform with a regularly repeating pattern, we know it must
have been produced with vocal fold vibration.
The number of repetitions (or cycles) per second is termed the frequency.
This was formerly expressed in cycles per second (c.p.s or c/s), but the unit
now used is the Hertz (Hz): 1 Hz
1 c.p.s.
Frequency and period are clearly related. The shorter the period, the
more periods will be completed in one second. The relationship is f
1/T,
where f is in Hz and T in seconds. Similarly, T
1/f.
Human hearing covers a range of about 20 Hz to 20,000 Hz. Frequency
is related to the auditory sensation of pitch: the higher the frequency,
the higher the pitch. The A which is sounded as the orchestra tunes up
has a frequency of 440 Hz; middle C has a frequency of 264 Hz. An octave
corresponds to a doubling of frequency. So the C below middle C has a
frequency of 132 Hz, the one above it 528 Hz. Alternating current mains
supply in the UK has a frequency of 50 Hz; when this is heard over a loud-
speaker, the effect is of a deep hum. The normal speaking voice of an
adult male has an average fundamental frequency of about 120 Hz.
One thousand Hertz
1 kilohertz (kHz). The frequencies corresponding
to the pitch of the voice are of the order of a few hundred Hz. A soprano
may be able to reach the second C above middle C, which has a fre-
quency a little over 1 kHz. But everyone’s speech contains energy at
frequencies higher than this. Speech is intelligible (though it doesn’t
sound very natural) if only frequencies up to about 2.5 kHz are repro-
duced (as they are in a communications radio link); the telephone passes
frequencies up to about 3 kHz. A ‘hi-fi’ system will typically pass frequen-
cies up to 20 kHz.
0521808820c02_p21_32.qxd 1/27/05 2:56 PM Page 29 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Voice 29

Digitising speech
Nowadays sound waveforms are mainly handled in digitised form. The
tracks on a CD, or wav files in a computer, are simply long strings of num-
bers representing waveforms sampled at regular intervals. The sampling
rate controls what frequencies will be preserved when the wave is recon-
structed. Basically you have to sample at a rate that is at least twice the
highest frequency you need to show. A CD works at a sample rate of more
than 40 kHz, enabling it to provide ‘hi-fi’ sound to 20 kHz or so. For many
of the waves shown in this book we were able to use slower sampling rates
of 16 kHz or even 10 kHz, giving us much smaller wav files.
Music files in computers and on the Internet aren’t usually wav files,
but MP3s. This is a clever way of compressing the information in a wav
file, so that it takes up less storage. The compressed version can then be
expanded again, with little loss of quality. It works for speech too (but isn’t
necessarily the best way to process speech). Actually, methods of data com-
pression and expansion were originally developed for speech – and the
need to fit more and more telephone conversations into a channel with
limited capacity was a spur to the development of speech technology.

Aperiodic waves
Not all sounds have periodic waveforms. A waveform that does not repeat
along the time axis is termed aperiodic; sounds with aperiodic waveforms
strike us as ‘noises’ rather than tones. Hisses, rumbles and roars are exam-
ples. An aperiodic sound results from random, irregular vibration.
Voiceless sounds in speech have aperiodic waveforms; the source of ran-
dom energy is the turbulent flow of air through the articulatory constric-
tion. Aperiodic waves can be sampled and digitised just like periodic ones,
of course.

FIGURE 2.6
A section about 13 ms in
duration cut from an
aperiodic wave (actually a
sustained [s] sound).

Certain speech sounds (for example, voiced fricatives such as [v]) employ
voice and noise together. The waveforms of such sounds are corre-
spondingly a mixture of periodic and aperiodic. The waveform for a
fully voiced [v] shows a basically periodic appearance, with added noise.
One type of periodic waveform has a very simple wavelike appear-
ance. They are sine waves, such as are produced by the simplest vibrat-
ing systems (e.g. a tuning fork, or the electronic tone generator used
0521808820c02_p21_32.qxd 1/27/05 2:56 PM Page 30 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

30 INTRODUCING PHONETIC SCIENCE

FIGURE 2.7
The waveform for a sustained [v] sound, showing a mixture of periodic and aperiodic
energy.

when a person’s hearing is being tested). This waveshape is of great

mathematical and physical importance, as any more complex wave can
be regarded as being composed of a number of sine waves of different
frequencies and amplitudes added together. Human speakers don’t and
can’t produce sine waves (but the wave produced by whistling is fairly
sine-like).
In practice, however, the periodic tones encountered either in speech or
music are not sine waves (or pure tones), but complex periodic tones. The
waveform of a complex tone is periodic, and the period and correspon-
ding frequency can be calculated. But the waveshape within each period
shows a pattern – sometimes very complicated – which results from the
simultaneous presence of many component frequencies.

FIGURE 2.8
A sustained [a] vowel, showing a complex periodic wave.

Chapter summary
We normally speak while breathing out. Air passing out of the lungs
must go through the larynx, where its flow is controlled by the vocal
folds. They may be open (as for breathing, and for voiceless sounds),
closed ( for a glottal stop) or vibrating (producing voice). Voicing makes
a distinction between consonants such as [v] (voiced) and [f] (voiceless),
and this is utilised in many languages of the world. Voiced sounds
have complex periodic waveforms, and voicing is the only source of
periodic energy in speech.
0521808820c02_p21_32.qxd 1/27/05 2:56 PM Page 31 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Voice 31

Exercises
Exercise 2.1
Take the days of the week (Monday, Tuesday . . .) and the months of the year
(January, February . . .). Identify the first segment of each as a voiced conso-
nant, a voiceless consonant or a vowel. Do the same for the initial segments
of the names of the digits one to nine.

Exercise 2.2
Decide whether the final sound of the following words is voiced or voiceless:
1 bus
2 use (noun)
3 as
4 breathe
5 has
6 off
7 buzz
8 use (verb)
9 teeth
10 rule
11 of
12 booth
Exercise 2.3
Every English word contains at least one voiced segment, but not every word
contains a voiceless segment. In fact, whole phrases and sentences can be
concocted that do not contain any voiceless segments (e.g. I’m worried by
Lilian’s wandering around Venezuela alone). What is the longest all-voiced
sequence you can devise?

Exercise 2.4
At a certain stage of development, many children acquiring English have a
rule of context-sensitive voicing. In one form of this, the consonants
[b d
v ð z  d ] may only occur in the child’s speech when followed directly
by a vowel; in all other positions they are replaced by their voiceless counter-
parts. Similarly, all voiceless consonants must be replaced with their voiced
counterparts when followed directly by a vowel. Consonants other than those
listed are not affected. So for example pad is pronounced [bt] pan
becomes [bn] and so on. A number of words are given below with a
transcription of the normal adult pronunciation. Give the pronunciation
expected from a child with context-sensitive voicing. Assume that the child’s
speech has no other differences from the adult target.
adult pronunciation
bag [b
]
boot [but]
beaker [bikə]
can [kn]
coach [kəυtʃ]
pad [pd]
Paddy [pdi]
jug [d
]
0521808820c02_p21_32.qxd 1/27/05 2:56 PM Page 32 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

32 INTRODUCING PHONETIC SCIENCE

father [fɑðə]
footpath [fυtpɑθ]
Jack [dk]
motor [məυtə]
rabbit [rbt]
song [sɒŋ]
shop [ʃɒp]
teatime [titam]

Further reading
For more on sound waves, and on periodic and aperiodic sounds in speech, see Dennis B. Fry, The physics of
speech, Cambridge: Cambridge University Press, 1978. Detailed information on the functioning and use of the
laryngograph can be found in E. R. M. Abberton, D. M. Howard and A. J. Fourcin, Laryngographic assessment of
normal voice: A tutorial, Clinical Linguistics & Phonetics, 3 (1989), 281–96.
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 33 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

CHAPTER

3 Place of
articulation

CHAPTER OUTLINE
In this chapter you will learn about: active and passive
KEY TERMS articulators; complex articulation; instrumental techniques
Active articulator for investigating place of articulation; the anatomy of the
Articulation vocal tract; the concept of articulation; the main places of
Double articulation for consonant sounds.
articulation
Passive
articulator
Place of
articulation
Vocal tract
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 34 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

34 INTRODUCING PHONETIC SCIENCE

Introduction
In this chapter we shall look at one of the features used to describe the pro-
duction of consonant sounds. In Chapter 2 we saw that consonants can differ
in voicing. So, for example, [p] is voiceless and [b] is voiced, [s] is voiceless
and [z] is voiced, and so on. However, it is fairly easy to see that [p] and [ b],
for instance, have quite a lot in common. If you say the English words pat
and bat while looking in a mirror you will see that the first thing you do
for each word is close your lips. It isn’t so easy with [s] and [z], but you may
be able to feel that for both these sounds the tip of your tongue gets very
close to the roof of your mouth not far behind your upper front teeth. Try
saying Sue and zoo, making the first consonant very long. The topic of this
chapter is where in the vocal tract consonants are produced.
Before we look at the details, a word of warning. Speech is really a mat-
ter of movement. Try saying the first sentence of this paragraph out loud
and concentrate on the movements you are making with your lips, tongue
and jaw. You should notice that there are no pauses between sounds.
Things are moving all the time. However, it is impossible to capture this
continual movement on paper and we need to have some points of refer-
ence, some way of categorising the similarities and differences between
speech sounds. So our warning is: when we say, for example, that the con-
sonant at the beginning of the English word thing is made by the tip of
the tongue and the back of the upper front teeth, we may seem to be say-
ing that the consonant is produced by the vocal tract taking up a static
‘This is the sort of position, but this leaves out a lot of the story. How did it get there? How
advice one receives: long does it stay there? How does it move away again? Some of these
‘Press your tonsils aspects of speech production form the topic for Chapter 6, where we will
against the under-
see that the precise articulation of a sound is influenced by the sounds
side of your larynx.
around it. First, we need to establish the basic terminology and concepts
Try, with the convex
part of the septum for the classification of speech sounds before we consider these more
curved upwards so detailed aspects of the production of speech. This chapter and the next
as almost – but not will introduce the basics. We start with the places in the vocal tract
quite – to touch the where speech sounds can be made.
uvula, try with the
tip of your tongue
to reach your thy- A brief tour of the vocal tract
roid. Take a deep
breath, and com- We will now begin a brief guided tour of the vocal tract. It might be help-
press your glottis. ful if you have a small pocket mirror and a small torch. We will start at the
Now, without mov- front of the vocal tract and work our way inside. You can use the vocal
ing your lips, say tract outline in Figure 3.1 to check where we are at any stage.
‘Garoo’.’ And when We begin with the lips. Notice that you can move both your upper and
you have done your lower lip. You can also round your lips and protrude them. Try look-
it they are still not ing in the mirror and saying a long [u] sound as in the English word boot.
satisfied.’ From
Your lower lip can also quite easily contact your upper front teeth. Try say-
Three Men on
ing the English word four while looking in the mirror. Some people can
the Bummel by
Jerome K. Jerome. also make the top lip contact the bottom front teeth, but this is a lot trick-
ier and this position is not normally used to produce speech sounds.
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 35 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Place of articulation 35

FIGURE 3.1
The vocal tract.

For the next stage of our tour we will be concerned with the roof of the
oral cavity. You can get an idea of the shape of this by moving the tip of
your tongue slowly back. Start with the tongue tip on the edge of your
upper front teeth. As you move it backwards, it will travel along a bony
ridge called the alveolar ridge. Then the roof of the mouth domes
upwards. The roof is still hard at this point because there is bone under-
lying the surface. As you move your tongue tip further back, you may be
able to feel that the roof of the mouth becomes soft. Not everyone can curl
the tongue far enough back to feel this, so don’t try too hard. This area is
called the soft palate or velum. That’s about as far as we can go with the
tongue tip.
For the next bit you really need a mirror and a torch. If you open your
mouth as wide as you can and shine the torch into it, you should be able
to adjust the angle of your mirror so that you can see to the very back of
your oral cavity. In the centre, there is a small fleshy part hanging from
the end of the soft palate. This is called the uvula (the name means ‘little
grape’ in Latin).
Beyond the uvula there is a cavity leading down to the larynx and the
lungs. This cavity is known as the pharynx.
We dealt with the structure of the larynx in Chapter 2, where we were
concerned with its function in producing voice. As we will see, the larynx
is also important for the production of consonant sounds. Finally, we
must not forget that above the roof of the mouth is a large space – the
nasal cavity. There are no moveable parts in this, so it is not possible to
perform an articulation here. The nasal cavity is important for speech
sound production, however, as we shall see in Chapter 4.
Next we shall turn our attention to the tongue. For the purposes of
describing speech sounds we need to recognise five areas of the tongue.
These are:

• the tip at the very front end

• the blade, which is the part which lies below the alveolar ridge when
the tongue is at rest
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 36 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

36 INTRODUCING PHONETIC SCIENCE

• the front, which lies below the hard palate when the tongue is at rest
• the back, which lies below the soft palate when the tongue is at rest
• the root, which faces backwards towards the rear wall of the pharynx.
You can see all of the above marked on the vocal tract outline in Figure
3.2.

FIGURE 3.2
Areas of the tongue.

Articulation
The term articulation is used to refer to the narrowing or constriction of
the vocal tract during the production of a speech sound. Two articulators
are involved, a passive articulator, which does not move, and an active
articulator, which moves toward the passive articulator. The passive artic-
ulator is usually some part of the upper surface of the vocal tract and the
active articulator is most often some part of the tongue. For some speech
sounds both articulators move, so the distinction between passive and
active cannot be maintained.
There are many different kinds of articulation. For example, in produc-
This picture comes ing the sound [p] the vocal tract is completely closed because the two artic-
from a book pub- ulators are in firm contact. In a sound like [w], on the other hand, the
lished in 1667. The articulators do not actually touch each other. For all of the descriptions of
author wanted to articulations below, we must remember that the lungs are pushing a
prove that the let- stream of air out of the vocal tract and that the articulation interferes
ters of the Hebrew with this airstream in some way. If the two articulators are in firm con-
alphabet reflect the tact, then the airflow will be blocked. If there is a gap between the artic-
tongue shape for the ulators, then air will escape. This aspect of an articulation is one of the
relevant sound. The topics for Chapter 4. In this chapter we shall concentrate on where in the
right-hand symbol
vocal tract articulations take place when we are producing consonant
in the picture looks
sounds.
like the tongue in
the figure.
Unfortunately, the
sound is [p] and the
Place
tongue is not
The location in the vocal tract where an articulation occurs is called the
involved in the pro-
place of articulation. For most articulations the term used to describe
duction of bilabial
sounds! the place of articulation is based on the name of the passive articulator
concerned. Here is a list of places of articulation:
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 37 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Place of articulation 37

Bilabial: both the upper and lower lips are active articulators for this
place. Sounds like [p b m] are bilabial. For all of these, the two lips are
in contact and the vocal tract is completely closed. For a sound like [φ]
the lips are close together and air is pushed between them. This
sound occurs in Japanese. If you know a Japanese speaker, ask them to
pronounce the word Fuji and watch their lips closely.
Labiodental: the active articulator is the lower lip and the passive articu-
lator is the edge of the upper front teeth. [f v] are labiodental sounds.
For these sounds, the lip does not contact the teeth firmly and air can
escape between the two articulators. Another labiodental sound
which is used sometimes in English is []. Here the contact is firm
and there is no escape of air. This sound is found for the m in words
like comfort and triumph.
Dental: the active articulator is the tongue tip and the passive articula-
tor is the back or the edge of the upper front teeth. The sounds at the
beginning of the words think and though are dental in our pronunciation.
The symbols are [θ] and [ð] respectively. Another example of a dental
sound is the t sound in eighth. The symbol for this is [t ]. (Not all kinds
of English use [θ] and [ð].)
Alveolar: the active articulator is the tip or the blade of the tongue and
the passive articulator is the alveolar ridge. [t d n l s z] are all alveolar
sounds used in English. For the first four of these there is firm con-
tact between the articulators, but for the last two there is a gap
through which air can escape.
Postalveolar: there are two sorts of postalveolar sound. Both have the
rear part of the alveolar ridge as the passive articulator. One type has
the tip of the tongue as the active articulator. An example is the
sound at the beginning of the word red in most varieties of English.
Here the tongue tip is curled back slightly and gets close to, but does
not actually touch, the back of the alveolar ridge. The symbol is [ɹ].
This type of sound is sometimes called apical postalveolar. The sec-
ond type has the tongue blade as the active articulator. The sound at
the end of the word fish is an example. The symbol is for this is [ʃ].
The tongue blade is raised so that it approaches the back part of the
alveolar ridge. This type of sound is sometimes called laminal
postalveolar.
Retroflex: the active articulator is the tongue tip and the passive articu-
lator is the hard palate. The tongue is curled back so that it approaches
the roof of the mouth behind the alveolar ridge. This type of sound
does not occur in most accents of English, but does occur in many
languages of India and in the English of speakers from those areas, as
well as many other languages. [  ] are examples of retroflex sounds.
The Tamil (Southern India and Sri Lanka) word [ma e] ‘skull’ contains
two retroflex consonants. Notice that the symbols all have a descending
hook to the right.
Palatal: the active articulator is the front of the tongue and the passive
articulator is the hard palate. An example of a palatal sound regularly
used in English is the consonant at the beginning of the word yes. The
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 38 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

38 INTRODUCING PHONETIC SCIENCE

symbol for this is [ ]. Another example, from French this time, is the
consonant at the end of the word montagne (‘mountain’). The symbol
for this is [ ]. This sound also occurs in Spanish, for example the sec-
ond consonant in mañana [ma ana] (‘tomorrow’) and in Italian, for
example the third consonant in lasagna [lasa a].
Velar: the active articulator is the back of the tongue and the passive
articulator is the soft palate. [k], as at the beginning of cow and [ ] as
a the end of dog are examples of velar consonants used in English.
Uvular: the active articulator is the back of the tongue and the passive
articulator is the very back part of the velum, where the uvula is
located. Uvular sounds do not occur in English, but can be found in
some varieties of Arabic. They sound like very far back velar sounds.
For example, [q ] is the uvular equivalent of [k] and [] is the uvular
equivalent of [ ].
Pharyngeal: the active articulator is the root of the tongue and the pas-
sive articulator is the rear wall of the pharynx. Again we do not use
this place of articulation for English, but Arabic has pharyngeal
sounds. For instance, [ ] sounds like a very intense and noisy h sound.
Glottal: the articulators are the vocal folds, both of which are active. For
[ʔ], known as glottal stop, the vocal folds are in firm contact, blocking
the airflow from the lungs. Glottal stop is well known as a feature of
the accent of English spoken in London (and in many other places too).
Listen to a Londoner saying butter or water. It is very likely that the [t]
in the middle will be replaced by [ʔ].

Here is a table listing the active and passive articulators for each place
of articulation:

Table 3.1 Summary of place of articulation labels.

Place of articulation Active articulator Passive articulator

Bilabial upper and lower lips none

Labiodental lower lip upper front teeth

Dental tongue tip upper front teeth

Alveolar tongue tip or blade alveolar ridge

Postalveolar tongue tip or blade rear of alveolar ridge

Retroflex tongue tip hard palate

Palatal tongue front hard palate

Velar tongue back soft palate

Uvular tongue back uvula

Pharyngeal tongue root rear wall of pharynx

Glottal vocal folds none

0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 39 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Place of articulation 39

We will now have a look at some examples of vocal tract outlines illus-
trating different articulations. There are four of these in Figure 3.3. Before
reading the rest of this section, have a look at them and see if you can
decide what the place of articulation is in each case.
In outline (a), we can see that the lips are open and that the tongue tip
is touching the alveolar ridge. Notice also that the rear surface of the soft
palate is in contact with the rear wall of the upper part of the pharynx.
The airways through the oral and nasal cavities are completely blocked.
This is an alveolar articulation of the kind used for [t] or [d].
In outline (b), again the lips are open and the soft palate is in more or
less the same position as before. This time, however, the back of the
tongue is bunched up and is in contact with the soft palate. This is a velar
articulation of the kind used for [k] or [ ].

(a) (b)

(c) (d)
FIGURE 3.3
Example articulations.
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 40 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

40 INTRODUCING PHONETIC SCIENCE

In outline (c), the tongue does not seem to be doing anything interest-
ing. The soft palate is in the same position as for the previous two out-
lines. Notice that the lower lip is very close to the upper front teeth and
that the lips are open. This is a labiodental articulation. [ f ] and [v] are
produced like this.
Finally, in outline (d) the tongue is bunched up so that the blade is very
close to the rear part of the alveolar ridge, although it does not actually
touch it. The lips are open and the soft palate is as before. This is a (laminal)
postalveolar articulation, as for [ʃ] or [].

Complex articulations
So far, the articulations we have met have been simple, in the sense that
they have only one constriction in the vocal tract and involve only one pair
of articulators. However, some speech sounds are more complex than this.
As an illustration, let’s look at what happens in a sound like [w], as at the
beginning of the English word well. If you look in a mirror while saying this
sound, you should see that your lips are rounded and protruded. So there
is a bilabial articulation going on. What you won’t be able to see and what
is very difficult to feel, is that at the same time the back of the tongue is
raised towards the soft palate. It doesn’t actually touch the soft palate, but
it gets reasonably close. So there is a velar articulation going on. Sounds
like this are called double articulations and [w] is known as a labial-velar.
Other labial-velar articulations include simultaneous [k] and [p], which
occurs in a number of West African languages, such as Yoruba (Nigeria).
Another example of a double articulation is the first sound in the French
word huit (
‘eight’). The symbol for this is [ɥ]. Again there is lip rounding
and protrusion, but this time it is the front of the tongue which is raised
towards the hard palate. This sound has a labial-palatal articulation.
Double articulations all have in common the fact that the two con-
strictions are of the same type. So in [w], for example, both constrictions
are fairly wide. In the West African simultaneous [kp] sound, both artic-
ulations are complete closure. But there is a second type of complex
articulation where this is not the case. Let’s take the first sound in the
English word twin as an example of this. Again, if you look in a mirror
while saying this sound, you should see that your lips are rounded –
again we have a bilabial articulation. At the same time, your tongue tip
or blade forms a complete closure by contacting the alveolar ridge. But
this is not a double articulation, because one of the constrictions (the
alveolar one in this case) is narrower than the other one. This is a case of
a primary articulation, accompanied by a secondary articulation and
the sound at the beginning of twin is called a labialised alveolar and is
symbolised [tw ].

Homorganic sounds
Sounds that have the same active and passive articulators are said to be
homorganic. For example [t d n l s z] are homorganic since they all have the
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 41 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Place of articulation 41

tip of the tongue as the active articulator and the alveolar ridge as the
passive articulator. But [t] and [] are not homorganic, even though they
both use the tongue tip, because the passive articulator is different (alveo-
lar ridge in one case, hard palate in the other). Similarly [ ] and [ ] are not
homorganic either, because although both are formed on the hard palate
the active articulator is different (tongue tip versus tongue front).
Languages quite often make use of homorganic sequences of consonants.
For example the English words hand, cold, best, jump, think all end with
homorganic sequences.

Instrumental techniques
You may be wondering how we know what articulations are used for
speech sounds. Most of what goes on in speech is very difficult to see and
takes place very rapidly. For a long time phoneticians had to rely mostly on
their own intuitions and feelings. Some more objective techniques were
available, but these tended to be very laborious and time-consuming. More
recently, especially since computer technology became widely available, it
has been possible to record the movements and positions of a speaker’s
articulators with a fair degree of accuracy.

Electropalatography
One technique for doing this is electropalatography. This involves the
speaker wearing a specially made false palate in which a number of elec-
trodes are embedded. The electrodes are sensitive to the contact by the
speaker’s tongue. The outputs of the electrodes are continuously moni-
tored by computer while the speaker is talking and the resulting analysis
is displayed as a series of figures known as palatograms.
A number of different electropalatography schemes are in use, but one
common one involves a false palate with 62 electrodes. These are arranged The earliest known
in groups to cover the alveolar ridge and hard palate. The conventional artificial palate for
way of displaying the output is in a grid of cells like that in Figure 3.4 A. speech research was
The alveolar area is at the top and the border with the soft palate area at used in America in
the bottom. If a cell is empty, then no tongue contact was recorded there. 1880. It was made
of a sheet of thin
Contact is shown as a filled cell, as in B–E in the figure.
black rubber paint-
The examples shown represent static positions of the articulators, but
ed with a mixture
usually electropalatography is used to record the changing state of tongue of chalk dust and
contact during the production of a sequence of speech sounds. alcohol. Tongue con-
Before we look at the palatograms in more detail, we must remember tact wiped away the
that this technique can give us useful information about some, but not all, coating and then
speech sounds. It cannot tell us about the articulation of sounds that do the palate was
not involve some part of the tongue as an articulator – bilabial or glottal removed and
sounds for instance – or sounds where the articulation does not take place photographed.
in the oral cavity, such as pharyngeal sounds.
If we look at palatogram B in the figure, we can see that all the black-
ened cells are at the back of the oral cavity. We can also see that the vocal
tract is completely blocked at this point, and presumably further back
too, beyond the area covered by the palate. This is what one might
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 42 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

42 INTRODUCING PHONETIC SCIENCE

expect to see for a velar sound such as [ k]. In palatogram C the same kind
of blockage is seen, but seems to be even further back.
Palatograms D and E are reasonably similar. They both show active contacts
at the front of the oral cavity. D is a dental sound and E an alveolar. Again
there is a complete blockage in both. Notice also that there are active contacts
all the way down both sides of the oral cavity. We will deal with this aspect of
articulation in greater detail in Chapter 4. For now we will just point out that
A
if the rims of the tongue do not contact the sides of the roof of the mouth,
then air can escape that way. The [d] sound at the beginning of dog would pro-
duce a palatogram like D, whereas the [l] sound at the beginning of log would
produce a palatogram without the active contacts at one or both sides.

X-rays
Another way of investigating the articulation of speech sounds is to use x-
ray photography or even x-ray cinematography. An x-ray photograph show-
B
ing the vocal tract appears as Figure 3.5.
The subject is an adolescent female, and her vocal tract is in a relaxed
breathing position. Her teeth are lightly together, her lips parted slightly,
and the velum down to permit breathing through the nose. The outline of
the alveolar ridge and hard palate can be seen clearly, as well as soft
tissues such as the tongue outline.
Figure 3.6 shows an x-ray photograph of an adult male subject in the
C position for a close central vowel in Russian. The superimposed numbers
on the picture show (1) the centre of the tongue and (2) the nasal cavity
just above the hard palate, and (3) the end of the velum, which is in a
raised position, closed against the rear wall of the pharynx. You can see

E
FIGURE 3.4
Example palatograms.

FIGURE 3.5
An x-ray photograph of
the vocal tract at rest.
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 43 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Place of articulation 43

FIGURE 3.6
X-ray view of the vocal
tract in the position for a
vowel (from G. Fant,
Acoustic Theory of Speech
Production, The Hague:
Mouton, 1960).

that the tongue is domed upwards towards the rear of hard palate, but
that it is not very close to it. In fact the speaker is producing a vowel sound
rather like the [i] in beat, though somewhat further back on the palate.
This articulation is quite similar to that for the consonant sound [j] at the
beginning of yacht.
It is often quite difficult to see the details of articulation on x-rays and
one needs quite a lot of experience to make sense of the pictures. It is
usual to present the results of x-ray investigations by showing tracings
from the original photographs. There are a couple of examples of these,
showing consonant articulations, in Figure 3.7. They are from the same
classic study as Figure 3.6 and show the same adult male subject. You
should be able to work out what places of articulation are involved. Trace
A is a dental sound. In fact it is [ t]. Trace B is a velar [ k].

Magnetic resonance imaging

Apart from the difficulty of interpretation of x-ray pictures, the main
drawback of the technique is the radiation risk to the speaker. A much
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 44 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

44 INTRODUCING PHONETIC SCIENCE

FIGURE 3.7
Tracings from X-ray images. (a) (b)

more recent and, as far is known, completely safe technique is magnetic

resonance imaging (MRI). Put very simply, this involves placing the sub-
ject in an imager which applies a very strong magnetic field to the body.
The field causes the vast majority of protons of hydrogen atoms to line up
in the direction of the subject’s head or feet. Most of the magnetic fields
created by these hydrogen protons are cancelled out by a proton lined up
in the opposite direction. However, a few protons (2 or 3 out of every mil-
lion) do not have their magnetic fields cancelled out in this way. The
machine applies a radio frequency pulse specific to hydrogen and this
causes the uncancelled protons to resonate at a particular frequency.
When the pulse is turned off, these protons return to their natural align-
ment within the magnetic field and give off excess energy as a result. The
imager detects this energy and turns the signal into an image. As you can
see from the two MRI scans in Figure 3.8, the resulting images are very
much clearer and easier to interpret.

FIGURE 3.8
MRI scans (from
S. Narayanan, A. Alwan
and K. Haker, ‘An
Articulatory Study of
Fricative Consonants
using MRI,‘ JASA (1995),
1325–64).
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 45 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Place of articulation 45

Unfortunately, MRI is a much slower technique, at least at present, and

it is not easy to get moving pictures of the vocal tract using MRI. Scan A
shows an alveolar [s]. Notice that the tip of the tongue is quite far forward
and the blade is not very close to the alveolar ridge. Scan B is a postalveo-
lar [ʃ]. This time the tip is a little further back and it is the blade of the
tongue which is articulating with the rear part of the alveolar ridge.

Places of articulation used in languages

of the world
In this section we will take a brief look at how place of articulation is used
in the world’s languages. This is quite a big subject and a complete account
at this stage would involve using many terms we have not yet intro-
duced. So to keep things simple we will confine our account to oral stops –
sounds like [p] or [tʃ ] where there is a complete closure in the vocal tract.
There is more detail on this sound-type in Chapter 4.
Quite a few languages use only three places of articulation for stops and
the most popular places are bilabial, alveolar (or alternatively dental) and
velar. Examples are: Finnish and Maori. English uses four places of articu-
lation for stops: bilabial, alveolar, postalveolar and velar. Other four-place
stop systems may use palatal stops instead of postalveolar. Examples are
Malay and Gugu-Yulanji (Australia) with bilabial, dental, palatal and velar.
Hindi is an example of a language with stops at five places (bilabial, den-
tal, postalveolar, retroflex and velar).
Only a few of the world’s languages use as many as 6 places of articula-
tion for stops. These are almost all Australian languages like Aranda,
which has bilabial, dental, alveolar, postalveolar, retroflex and velar stops
and Diyari with bilabial, dental, alveolar, retroflex, palatal and velar stops.
Places of articulation can be ranked according to their occurrence in
stop systems of the world’s languages. The ranking looks like this:
Alveolar/dental  velar  bilabial  palatal  uvular  retroflex  others
Where  means ‘occurs more frequently than’ and  means ‘much
more frequently than’. In fact, the first three places in the ranking are
almost equal in frequency of occurrence.
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 46 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

46 INTRODUCING PHONETIC SCIENCE

Chapter summary
In this chapter we have introduced the concept of place of articulation.
The main areas of articulation for consonants are the lips, upper front
teeth, alveolar ridge, hard palate, soft palate, uvula, the pharynx and the
larynx. The tongue can be viewed as consisting of five main areas: the
tip, blade, front, back and root. An articulation is a narrowing or con-
striction of the vocal tract caused by an active articulator approaching a
passive articulator, or by two active articulators approaching one
another. Articulations may be simple and consist of only one constriction,
or they may be complex and consist of two equal constrictions (a double
articulation) or of a primary and a secondary articulation.
The chapter includes a brief overview of instrumental techniques for
investigating place of articulation: electropalatography, x-ray photogra-
phy and magnetic resonance imaging and concludes with a look at
how places of articulation for oral stop consonants are used by the
world’s languages.

Exercises
Exercise 3.1
In each of the groups of sounds below there is one that is not produced at
the same place of articulation as the others. Try to identify this odd one out.
You will probably need to look at the IPA chart on page 3 to help you.
(a) [t d s ʃ l] (b) [φ p f b m] (c) [k  ŋ x]

Exercise 3.2
Below are some English words. Each has a suggested change associated with
it. What English words could result if you perform the change in each case?
Example: cap: change any velar consonant into any bilabial consonant.
Answer: pap bap map
(a) fight: change any labiodental consonant into any
velar consonant
(b) take: change any alveolar consonant into any
postalveolar consonant
(c) bathe: change any dental consonant into any alveolar consonant

Exercise 3.3
For which of the following English words is the place of articulation the same
for all of the consonant sounds in the word? (Hint: watch out for silent letters!)
(a) kind (b) sent (c) palm (d) align (e) lazy

Exercise 3.4
For each of the English words listed below we have provided two pronuncia-
tions. Those in the Adult column are what you might expect from an adult
speaker, though they may not be the same as your pronunciation. In the
Child column are pronunciations which could be produced by a young child.
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 47 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Place of articulation 47

Word Adult Child

thumb [θm] [vm]
push [pυʃ] [bυç]
talk [tɔk] [dɔt]
other [ðə] [və]
bag [b ] [bt]
measure [meə] [meə]
thing [θŋ] [vn]
both [bəυθ] [bəυf]
good [ υd] [dυt]
shoe [ ʃu] [ u]
(a) List all the places of articulation for consonants used by the adult speaker
for these words.
(b) List the corresponding places of articulation used by the child.
(c) Sometimes the child uses a voiceless sound in place of the adult’s voiced
sound. Try to work out when this happens.
(d) Conversely, the child also sometimes uses a voiced sound where the
adult uses a voiceless one. When?

Exercise 3.5
Below there are some unlabelled vocal tract outlines. Try to work out what
the place of articulation is in each case.

Exercise 3.6
The vocal tract diagrams below have had the outline of the tongue removed.
Photocopy or trace the outlines and then try to fill in the tongue outline for
the articulations indicated. In each case draw the relevant part of the tongue
in contact with the passive articulator.

Exercise 3.7
On page 3 we mentioned that an articulation involving the upper lip and the
lower front teeth is possible, but that it is not normally used for the produc-
tion of speech sounds. Can you think of any other possible articulations that
we have not mentioned?
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 48 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

48 INTRODUCING PHONETIC SCIENCE

(a) (b)

(c) (d)
Figure for Exercise 3.5

Dental Retroflex Velar

Figure for Exercise 3.6
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 49 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Place of articulation 49

Further reading
An excellent account of place of articulation with many examples from languages around the world can be
found in chapter 2 of Peter Ladefoged and Ian Maddieson, The sounds of the world’s languages, Oxford:
Blackwell, 1996. Another very detailed account you might like to look at is chapter 8 of J. C. Catford,
Fundamental problems in phonetics, Edinburgh: Edinburgh University Press, 1977. There is a survey of the
sound systems of over 300 languages in Ian Maddieson, Patterns of sounds, Cambridge: Cambridge
University Press,1984.
0521808820c03_p33_50.qxd 1/27/05 2:57 PM Page 50 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 51 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

CHAPTER

4 Manner of
articulation

CHAPTER OUTLINE
In this chapter you will learn: how the articulators may
KEY TERMS constrict the airstream in different ways for the production
Affricate of consonant sounds; how sounds may be classified
Approximant according to manner of articulation; how sounds may be
Degree of classed as obstruent or sonorant; how manners of
stricture articulation are used in the world’s languages.
Flap
Fricative
Lateral
Manner of
articulation
Median
Nasal
Obstruent
Plosive
Sonorant
Stop
Tap
Trill
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 52 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

52 INTRODUCING PHONETIC SCIENCE

Introduction
We have now seen that consonant sounds may differ in voicing and in place
of articulation. The final basic distinction between consonants is known as
manner of articulation and that is the subject of this chapter. A manner of
articulation label, such as ‘plosive’ or ‘fricative’, refers to the way in which
the airstream used for a speech sound is modified by the primary and sec-
ondary articulators. Is the airstream completely blocked? Can the air escape
through the nasal cavity? Does the air pressure inside the vocal tract rise?
These are the sorts of questions we will be investigating below.
Once we have learnt to apply manner labels correctly we shall be in a
position to give a three-term label to a consonant, specifying its Voice,
Place, Manner, usually abbreviated VPM. VPM labels are useful shorthand
for identifying most consonants of the world’s languages, although they
are not the complete story by any means.

‘The breath, circulating in the lungs, is forced upwards and impinges

on the head, reaches the speech organs and gives rise to speech
sounds. These are classified in five ways – by tone, by length, by place
of articulation, by process of articulation and by secondary features.
Thus the phoneticians have spoken: take careful heed.’ This quote is
taken from an Indian work thought to have been written sometime
between 500 and 150 BC.

Airflow, pressure, turbulence

In order to describe some aspects of the way air flows through the vocal
tract during the production of a consonant, it is useful to simplify things
a little and view the tract as a straight tube rather like a bicycle pump. In
the case of the vocal tract air is usually set in motion by the lungs, but
in our simpler model we’ll use a piston that fits tightly into the tube as in
Figure 4.1.

FIGURE 4.1
Simple tube and piston
model of the vocal tract. piston

If we were to push the piston to the left, the air in the tube would be
displaced and pushed out of the tube. As the tube is open and without any
obstruction, the air would flow out without making any appreciable
noise. The air particles would all move parallel to one another in what is
known as laminar flow.
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 53 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Manner of articulation 53

FIGURE 4.2
A blocked tube.

In Figure 4.2 we have blocked the end of the tube. Now when we try to
push the piston, the air particles to the left of it cannot escape and they
become more and more crowded together as the space between the piston
and the blockage gets smaller. The air pressure inside the tube rises.
Eventually of course the piston will become very difficult to move and if
we continue with some mechanical means to push the piston, the tube
will burst apart. If we were suddenly to remove the blockage, air would
burst out of the tube with a loud pop.

FIGURE 4.3
A constricted tube.

In Figure 4.3 we have introduced a narrowing into the tube. The air can
escape, but not as freely as in the unconstricted tube. The air pressure
between the piston and the constriction will rise and the air flowing
through the constriction will become turbulent, that is, the air particles
will no longer flow in parallel paths but will collide into one another and
make a noise.
Try this simple experiment. First purse your lips as though you were
going to whistle. Now breathe out as gently as you can. You should not hear
any noise at all. The air is flowing through the constriction too slowly to
become turbulent. Now, with exactly the same lip position, imagine you
are trying to blow out all the candles on a birthday cake. This time you
should hear quite a strong hissing noise caused by the turbulent airstream.
From this we can see that the airflow must be rapid enough and the
constriction narrow enough if there is to be any turbulence and any noise.

Degree of stricture: obstruent and sonorant

We will now leave our simple tube model behind and examine what
happens in the vocal tract itself. First, it is fairly obvious that the two
articulators for a consonant sound may take up a variety of different
positions with respect to one another. For our purposes it is adequate to
recognise three distinct positions or degrees of stricture.
1. Closure: the articulators are in firm contact
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 54 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

54 INTRODUCING PHONETIC SCIENCE

2. Narrowing: the articulators are close together but not touching

3. Approximation: there is a reasonably wide gap between the articulators
Degrees 1 and 2 cause a rise in air pressure inside the vocal tract (unless
the air can escape by another route) and sounds produced like this are
called obstruents. Degree 3 does not cause a pressure rise and sounds
produced in this way are called sonorants.
This obstruent/sonorant distinction is a very important one both pho-
netically and phonologically and we will come across it again and again.
Just to give a very simple example: the rise in air pressure during an obstru-
ent makes it difficult to maintain vocal fold vibration for a long period.
Obstruent sounds quite often lose their voicing in unfavourable environ-
ments – they become devoiced. Sonorants tend not to be affected in the
same way. We will look at devoicing in more detail in a later chapter.

Stops: oral and nasal, plosive and affricate

The narrowest sort of constriction one can produce in the vocal tract is of
course a complete blockage of the airstream. For this the two articulators
are in firm contact for an appreciable amount of time. Try saying a long [n]
sound. You should be able to feel that the tip and/or blade of your tongue
is pressing firmly against your alveolar ridge and the sides of your tongue
are pressed against your upper side teeth and gums. The same is true for a
sound like [t] or [d], but these sounds usually do not last long enough for us
to get a clear impression of what the articulators are doing. Such sounds,
where the vocal tract is completely blocked for a time are called stops, and
they can be produced at many places of articulation, bilabial, labiodental,
dental, alveolar and so on. The diagrams in Figure 4.4 show the position of
the articulators for the two alveolar stops: (a) [t] (or [d]) and (b) [n].

FIGURE 4.4
An alveolar stop
articulation (a) with velic
closure (b) with the
velum open. (a) (b)
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 55 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Manner of articulation 55

Table 4.1. Places of articulation and symbols for nasals.

Bilabial labiodental alveolar palatal retroflex velar uvular

m  n   ŋ 

Although diagrams (a) and (b) in Figure 4.4 are both of alveolar stops,
there is of course a difference between them. Look at the position of the
soft palate. In diagram (a) it is raised and is in contact with the rear wall
of the pharynx. This is called a velic closure or a velopharyngeal closure.
This closure prevents air escaping through the nasal cavity. Stops made
with a velic closure are called oral stops. Oral stops belong to the obstru-
ent class, as there is an appreciable rise in air pressure in the vocal tract.
In diagram (b) there is no velic closure and air can escape freely through
the nasal cavity, although it is prevented from escaping through the oral
cavity by the alveolar stop articulation. Stop articulations without a velic
closure and with nasal airflow are called nasal stops or, more simply,
nasals. Nasals are sonorant sounds, because the airstream produced by
the lungs can escape via the nasal cavity and there is no rise in air pressure
inside the vocal tract.
Try this experiment. Say the word ado and repeat it over and over. After
the first two or three repetitions pinch your nostrils firmly and keep
repeating the word. Then release your nostrils again and repeat the word
some more. You should have had no difficulty saying the word with your
nostrils blocked and the sound of the word should have been exactly the
same when your nostrils were blocked and when they were free. Now
follow the same procedure with the word unknown. No problem with it
when your nostrils were free, was there? But very probably you will have
found that this word is impossible to say normally when pinching your
nostrils. The air cannot flow out of your nose as it should for the nasal
stops in this word and so the vocal tract quickly fills with air. The air
pressure rises, the vocal folds stop vibrating and you get to feel a little
uncomfortable.
Nasals in most languages of the world are voiced and can occur at many
places of articulation. Table 4.1 shows the voiced nasals from the IPA
consonant chart. However, voiceless nasals are perfectly possible and a
small number of languages do use them. In order to symbolise a voiceless
nasal, the voiced nasal symbol is used, but it is accompanied (above or
below) by a small circle. So a voiceless bilabial nasal is represented as [m
].
This sound is used in Burmese, for example in a word like [m
a] meaning
notice. It may also be used in English at the beginning of the interjection
usually written hmm! (
[m
m]).
We now turn to look at oral stops in more detail. Let us take the voiceless
bilabial oral stop [ p] as an example. For this, the lips are completely closed,
blocking the airstream through the oral cavity. There is also a velic clo-
sure, so the air cannot escape via the nose. This configuration of the artic-
ulators is maintained for a short time, around 120 ms. During this phase
the air pressure inside the vocal tract rises rapidly. When the bilabial
closure is released, the compressed air bursts out of the vocal tract with a
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 56 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

56 INTRODUCING PHONETIC SCIENCE

small explosive noise. Such sounds are called plosives. They can be viewed
as consisting of three phases (1) The approach phase, when the articula-
tors are forming the oral closure , (2) the hold or compression phase, when
the air pressure rises and (3) the release or plosion phase when the artic-
ulators part and the compressed air is released. This is sometimes repre-
sented as in Figure 4.5.

approach hold release

FIGURE 4.5
Phases of plosive
production.

Plosives are one kind of oral stop. Another kind involves the same
approach of the articulators, the same kind of hold phase, but a much
slower parting of the articulators during the release phase. During the
slower release, the air rushing between the two articulators makes a hiss-
ing sound. Such consonants are known as affricates. English has two
affricate consonants: a voiceless postalveolar affricate [tʃ], as at the begin-
ning and end of church, and its voiced counterpart [d], as the beginning
and end of judge. There are many other possibilities, however. German, for
instance, has a voiceless alveolar affricate [ts], as at the beginning of the
word Ziel [tsi l], meaning ‘aim, goal’.
The phases of production of an affricate can be represented as in Figure 4.6.

approach hold release

FIGURE 4.6
Phases of affricate
production.

Fricatives: median and lateral

Fricatives are consonant sounds that are produced with a very narrow
opening between the active and passive articulators. As we saw above on
pp. 105–6, when air is forced through a narrow gap, the flow becomes tur-
bulent. It is this turbulence that creates the acoustic energy for fricative
sounds.
Fricatives may be produced at many different places of articulation and
may have all kinds of modifications. The first distinction between fricative
types we will discuss is that of median vs lateral fricatives.
As a good example of a median fricative, let us take [s] – voiceless alveolar
median fricative. If you produce an [s] sound and then, without changing
the position of the articulators, breathe in forcefully, you should feel cold
air on the middle of your alveolar ridge. You should not feel anything at
the sides of your oral cavity. This is because the articulators are arranged
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 57 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Manner of articulation 57

so that air flows only through a narrow channel in the centre of the
cavity. This is why an articulation of this type is termed median. The
sides of the tongue are in firm contact with the upper side teeth and
gums, preventing airflow passing that way. A palatogram for [s] might be
as in Figure 4.7.
A lateral fricative has the articulators in a position where airflow is
blocked on the midline of the oral cavity, but there is a narrow space
between the sides of the tongue and the upper side teeth and gums, so
FIGURE 4.7
the airstream can escape laterally. In order to get an impression of what Palatogram of [s].
this sounds like, first produce a long [l] sound. Then, keeping the articu-
lators in the same position, turn off the vocal fold vibration and blow
hard. The sound involved is [ ], a voiceless alveolar lateral fricative. The
tongue tip contacts the alveolar ridge firmly, preventing air escape. One or
both sides of the tongue are close to, but not actually touching the upper
side teeth. The pattern of palate contact would be something like that in
Figure 4.8.
Table 4.2 shows the symbols for the median fricative consonants from
the IPA chart.
The consonants above come in voiceless (left) and voiced (right) pairs.
The voiced fricatives have two simultaneous sound sources: the friction
noise of the turbulent air produced at the place of articulation and the FIGURE 4.8
sound made by the vibrating vocal folds. We can see the difference Palatogram of [ ].
between the voiceless and voiced fricative sounds in the two speech pres-
sure waveforms in Figure 4.9. The sounds involved are [s] and [z].
The two lateral fricatives on the IPA chart are both alveolar – the voice-
less one we have already met and the voiced one is symbolised [ ]. An
example of the voiceless alveolar lateral fricative in use is the Welsh word
cyllell (‘knife’) [ kə e ]. An example of the voiced alveolar lateral fricative is
the first sound in the Zulu word [ a la] (‘play’).

Approximants: median and lateral

Approximants, like fricatives, have no complete closure in the vocal tract
and also come in two main varieties, median and lateral. Unlike fricatives,
however, approximants are sonorants and do not have a stricture narrow
enough to cause a rise in air pressure and turbulence, at least at the rates
of airflow found in normal speech.

Table 4.2. Median fricative consonants of the IPA.

bilabial φβ palatal ç
labiodental fv velar xγ
dental θð uvular χʁ
alveolar sz pharyngeal ʕ
postalveolar ʃ glottal h
retroflex 
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 58 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

58 INTRODUCING PHONETIC SCIENCE

(a)

(b)

FIGURE 4.9
Waveforms for (a) [s] and (b) [z].

Lateral approximants have a complete closure between the articulators

on the midline of the vocal tract, but one or both sides of the tongue are
not very close to the upper side teeth and gums. When air escapes over
the tongue rims and out of the oral cavity, it does so without becoming
turbulent. Essentially, the position of the articulators is the same as that
for the lateral fricative depicted in Figure 4.8.
Lateral approximants may occur at a number of places of articulation
and are usually voiced. Here are some examples:

voiced alveolar [l]

voiced retroflex []
voiced palatal [ʎ]

Lateral approximants are often simply called laterals. We are familiar with
the voiced alveolar lateral in English. An example of a retroflex lateral in
use is the word [atə] meaning ‘day’ in Aranda, an Australian language.
The palatal lateral occurs in Italian, for instance. The word aglio (‘garlic’)
[aʎʎo] contains a long version of this sound.
Median approximants have no complete closure anywhere on the mid-
line of the vocal tract. Air can pass freely over the centre of the tongue and
exit without turbulence. You may come across two older terms for medi-
an approximants: glide and frictionless continuant.
Some median approximants are very like vowels. They are produced in
the same area of the oral cavity as vowel sounds. They are sometimes called
semivowels. Examples are [w] voiced labial-velar median approximant, as
at the beginning of well and [ j] voiced palatal median approximant, as at
the beginning of yes.
Other median approximants are not so vowel-like. These include r
sounds such as [ɹ], which is a voiced postalveolar median approximant
(shown in Figure 4.10), [] a voiced labiodental median approximant, [] a
voiced velar median approximant, and [ʁ ] a voiced uvular median approxi-
mant. Essentially, at any place where a fricative is possible you can also
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 59 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Manner of articulation 59

FIGURE 4.10
The articulation of [ɹ].

make a median approximant. ‘Median approximant’ is usually abbreviated

to just approximant.

Taps, flaps and trills r sounds and later-

als taken together
The final aspect of manner of articulation that we shall examine has to do are sometimes called
with the timing rather than the nature of the events involved. We will ‘liquids’. Contrary to
start by looking at the difference between a plosive and a tap and take as what you might
our examples the two consonants [d], voiced alveolar plosive, and [ɾ], think, this doesn’t
voiced alveolar tap. The latter sound is quite familiar to speakers of describe their sound
or the way they are
English and is used in many varieties of US English for the consonant in
produced. It results
the middle of words like atom. The sequence of events for both of these
from an inappropri-
sounds is essentially the same, but everything happens much more quick- ate translation of
ly for the tap. Crucially, the amount of time the two articulators stay in the Greek word
contact is very short for the tap (around 20 ms), much shorter than for the hygros which
plosive. This means that there is no time for an appreciable increase in air means ‘variable’ or
pressure in the case of the tap and therefore this sound type belongs to ‘inconstant’ when
the sonorant class. The timing of a plosive and a tap is compared in the applied to the
schematic diagrams in Figure 4.11. changeable behav-
The lower diagram in Figure 4.11 shows what happens during the pro- iour of these sounds
duction of a trill (sometimes called a roll). This sound type consists of in the detailed
analysis of Ancient
a series of very rapid tap-like closures between the active and the pas-
Greek versification.
sive articulator. The mechanism by which this is achieved is very simi-
lar to the way in which the vocal folds vibrate. The rate of vibration is,
however, much slower because the structures involved have a much
greater mass. Figure 4.12 shows the speech pressure waveform for [r]
voiced alveolar trill pronounced between two vowels. The effects of the
individual closures of the tongue tip show up as very brief troughs in
the waveform. There are five such troughs and the trill lasts approxi-
mately 370 ms, so the rate of vibration of the tongue tip is about 23 Hz.
Spanish is an example of a language that uses both an alveolar tap and
an alveolar trill, for instance pero [peɾo] means ‘but’ and perro [pero]
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 60 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

60 INTRODUCING PHONETIC SCIENCE

FIGURE 4.11
Timing comparison of
plosives, taps and trills.

FIGURE 4.12
Speech pressure waveform for [r].

means ‘dog’. Trills at other places of articulation are found in the world’s
languages. A voiced uvular trill [] is reasonably common. It sounds a lit-
tle like a gargle. The uvula itself trills against the back of the tongue. A
much rarer trill, found in only one or two languages, is [] voiced bilabial
trill. This is the sort of noise that is often represented as Brrr and is used
to indicate that you are feeling cold.
The final manner of articulation we shall discuss is the flap. This is
very similar to a tap in that there is a single extremely brief closure
between the articulators. It is difficult to diagram the sequence of events
for a flap in the same manner as that used in Figure 4.11. The diagram
would be identical to that for a tap. However, there is an important dif-
ference between the two articulations. For the tap, the active articulator
moves rapidly towards the passive articulator and rapidly away again, as
we have seen. For a flap, the active articulator strikes the passive articu-
lator as it passes by. The only flap symbol on the IPA chart is [#] voiced
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 61 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Manner of articulation 61

(a) (b) (c)

FIGURE 4.13
The sequence of events in a retroflex flap.

retroflex flap. Figure 4.13 shows the articulation of this sound. First, the
tongue tip curls back behind the alveolar ridge, and then flaps forwards
and downwards, striking the rear part of the alveolar ridge as it does so.
An example of a voiced retroflex flap in use in a word is [sa#i], which
means ‘burnt’ in Punjabi.

Voice, place, manner labels

The manner of articulation labels that we have introduced are: plosive,
affricate, nasal, fricative (lateral and median), approximant (lateral and
median), tap, flap and trill. These labels complete a very frequently used
method of specifying consonants known as Voice, Place, Manner (VPM)
labels. To give a simple example, let us consider the sound [s]. First we need
to determine if the vocal folds are vibrating or not during the production
of the sound. As they are not, we label the sound voiceless, Next we need to
consider which articulators are concerned in the production of the sound.
For [s], the tongue tip or blade articulates with the alveolar ridge, so the
place of articulation is alveolar. Finally we need to specify exactly what
kind of articulation is formed for this sound. The active articulator, the
tongue tip/blade, approaches close to the alveolar ridge and touches at the
sides, but does not form a complete closure. Furthermore, the airstream
produced by the lungs is made turbulent when it passes through the nar-
row constriction. The manner of articulation is therefore fricative. Putting
all this together, [s] has the VPM label ‘voiceless alveolar fricative’.

Use of manners of articulation in the

world’s languages
In Chapter 3 we had a brief look at the use of place of articulation in the
world’s languages by concentrating on oral stops, that is, plosives and
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 62 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

62 INTRODUCING PHONETIC SCIENCE

affricates. Here we will look at how the world’s languages use manner of
articulation.
All known languages have plosives, but that is not true for the other
manners we have described above. The vast majority of languages have
median fricatives, and typically languages have a fricative system that is
as large or larger than the oral stop system of the language. English is
typical in this respect with oral stops at four places: bilabial, alveolar,
postalveolar and velar, and fricatives also at four places: labiodental,
dental, alveolar and postalveolar. Welsh is a language where the fricatives
outnumber the oral stops. The language has labiodental, dental, alveolar
median, alveolar lateral, postalveolar and velar fricatives, but only bilabi-
al, alveolar and velar plosives, and no affricates. A few languages lack
fricatives completely. Most of these are Australian languages.
Most languages have nasals, usually at two or three places of articula-
tion. Again English is typical with bilabial, alveolar and velar nasals.
French is an example of a language with nasals at four places: bilabial, den-
tal, palatal and velar. Languages without nasals are very rare. An example
is Rotokas, spoken on the Pacific island of Bougainville, which has the
smallest number of consonants of any known language. It has only six,
three voiceless plosives, one voiced plosive, one voiced fricative and a tap.
Median approximants of the ‘semivowel’ type occur in most languages,
though it is quite rare to find a language with more than two or three of
them. The commonest are [w] and [ j] and English is again typical in having
just those two. French is an example of a language with three, adding [ɥ]
to those found in English. Nama (Namibia and South Africa) is an example
of a language lacking this sound type, although it has a fairly large con-
sonant system.
Most languages have some sort of r sound. This may be a median approx-
imant like the postalveolar [ɹ] in English or the uvular median approxi-
mant [ʁ ] found in some varieties of French. Other languages use a tap or
trill and some, as we have already seen for Spanish, use both. An example
of a language with none of these sound types is Hawaiian.
Finally we will look at lateral approximants. Again, most languages
have this type of sound, usually at the alveolar or dental place of articu-
lation. If a language has two lateral approximants, they are likely to be
dental/alveolar plus palatal, as in Italian and some varieties of Spanish, or
plus retroflex as in Norwegian and Malayalam (Southern India). A velar
lateral approximant, made by forming a closure between the centre of the
back of the tongue with one or both sides of the tongue lowered, was at
one time thought not to occur in any language, but relatively recently a
number of languages using this sound have been reported, mostly from
New Guinea. In response to this, the IPA adopted the symbol [] to repre-
sent the sound. A few languages, including Japanese and Maori (New
Zealand), have no lateral approximants.
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 63 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Manner of articulation 63

Chapter summary
In this chapter we continued our survey of articulation and have
looked at some of the details of how the articulators constrict the
airstream in various ways. We have seen that manner of articulation is
not a simple feature, but comprises a number of different and largely
independent dimensions. First the articulators may constrict the
airstream to different degrees: either blocking it completely, or con-
stricting it enough to cause air turbulence, or leaving a considerable
gap for the air to pass without turbulence. Secondly, the soft palate
may be raised, preventing nasal airflow, or lowered, allowing such
flow. Thirdly, a closure may be made on the midline of the oral cavity
with or without a lateral opening allowing air to escape over the rims
of the tongue. The rate of release of a closure is the fourth factor dis-
tinguishing speech sounds. A further aspect of manner of articulation
concerns the speed of the whole gesture for the articulation; for taps,
the articulators are in contact for only a very brief period.
A very important distinction among manners of articulation has
also been introduced: the sonorant-obstruent distinction will be
referred to again and again in the coming chapters. Plosives, fricatives
and affricates, the obstruents, are articulated with a constriction that
causes air pressure to rise inside the vocal tract. Other consonants,
together with all vowels, are sonorants and are made with a configura-
tion of the articulators which allows the pressure inside the vocal tract
to be approximately equal to the pressure outside.

Exercises
Exercise 4.1
In the sentence Susie saw five of the shows, she says all of the consonant
sounds have the same manner of articulation. Which? See if you can think of
some similar English sentences.

Exercise 4.2
Which of the following English words contain no sonorant consonants?
soon, clasp, talked, beauty, sky, cows, thorough, quick, captive, chalk

Exercise 4.3
In Welsh words often undergo a process called initial mutation. Each word
has a number of mutated forms, which are used in different contexts. Here
are some examples of two of the mutations.

Basic Form Mutation I Mutation II Meaning

[pen] [ben] [fen] head
[tad] [dad] [θad] father
[karjad] [&arjad] [xarjad] love
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 64 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

64 INTRODUCING PHONETIC SCIENCE

[bas&ed] [vas&ed] not applicable basket

[damwain] [ðamwain] not applicable accident
[&arð] [arð] not applicable garden
(a) What have the initial consonants of the Basic Forms in the first block of
words got in common?
(b) What change does Mutation I cause to these consonants?
(c) What change occurs with Mutation II?
(d) Consider the same three questions for the second block of words.
Exercise 4.4
Look at the following speech pressure waveforms and try to decide the man-
ner of articulation of the consonant sound between the two vertical lines.

Exercise 4.5
D, a young English child, exhibits the following features in her speech.
(a) if a word contains an alveolar lateral approximant, then any other approxi-
mants in the same word will also be alveolar lateral
(b) all alveolar and postalveolar fricatives are produced as alveolar plosives of
the appropriate voicing.
How would D pronounce the following words?
lorry, shoes, yellow, woolly, silly, puzzle, swallow, umbrella
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 65 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Manner of articulation 65

Exercise 4.6
Some speakers have difficulty with alveolar fricatives and produce alveolar
lateral fricatives instead. Explain the articulatory difference between the two
sound types. How would an English speaker with this problem pronounce the
following words?
sixty, amusing, results, cousins, sensible

Exercise 4.7
A language X has three kinds of r sound : a voiced alveolar tap, a voiced
alveolar trill and a voiced uvular fricative. The tap is used between vowels,
the trill at the beginning of a word or immediately following a consonant and
the fricative at the end of a word or immediately before a consonant. Here
are some words in X in their spelling form, where all three r sounds are
represented by the letter r. How would these words be pronounced?
dramar taran garsat rakra aramard

Exercise 4.8
In a language Y, all words of the form C1C2VC3C4, conform to the following
restrictions
(a) C1 must be an obstruent
(b) C2 must be a sonorant
(c) If C3 is a sonorant, C4 must be an obstruent
(d) If C3 is a nasal, C4 must be a homorganic plosive
(e) If C3 is an obstruent, it must be a plosive and C4 must be a fricative
Say whether each of the words below is a possible word of Y and if not, why not.
[slamb] [flant] [&walm] [spats] [twalf]
[kjaŋ&] [blams] [ljalʃ] [zjalt] [kramd]
[dlaps] [snand]

Exercise 4.9
A language Z has a voiceless plosive, a voiced plosive and a voiced nasal at each
of the following places of articulation: bilabial, alveolar, palatal, velar and uvular.
(a) Show the stop system of this language as a table with place of articulation
horizontally and stop type vertically.
(b) Z also has voiceless and voiced fricatives at all of the same places of
articulation. Add the fricative system to your table.
(c) Here are some words of Z. Based on these, say whether the statements
(i–v) about Z are true or false. If you think a statement is false, quote a
word from the data set which proves this.
Data set
[quχa] [cuçi] [saza] ['uu] [&ui]
[dasa] [zaa] [&aʁuŋ] [baφam] [tazas]
[nusaφ] [iax] [ceʁaç] [auχ] [sasan]
Statements:
(i) All consonants in a word are homorganic.
(ii) Stops may not occur between vowels.
(iii) The only obstruent consonants permitted word-finally are voiceless fricatives.
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 66 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

66 INTRODUCING PHONETIC SCIENCE

(iv) Front consonants (articulated on the alveolar ridge or further forward)

may not occur in the same word as back consonants.
(v) Voiced fricatives may not occur word-initially.

Exercise 4.10
The IPA consonant chart has some shaded cells representing combinations of
place and manner that are ‘judged impossible’. One of these combinations is
‘pharyngeal nasal’. Explain why this sound type is impossible.

Exercise 4.11
Here is a list of all the consonant sounds used in one variety of English. Each
is accompanied by a keyword in orthography. Complete the table below. See
how far you can get with this without looking anything up.
Keyword Voice Place Manner
p pie voiceless bilabial plosive
t tie
k key
b buy
d die
& &uy
m my voiced bilabial nasal
n no
ŋ sin&
f fee
v van
θ thi&h
ð thou&h
s so
z zoo
ʃ she
 measure
tʃ chip
d jam
w wet
ɹ red
l let
j yes
h hat
In Chapter 1 we used [r] for the English sound. Now that we have learned
about several r types, we are using the more specific symbol [ɹ].

Exercise 4.12
Hyponasal and hypernasal:
(a) In hyponasal speech, when the nasal cavity is blocked as it is when you
have a heavy cold, nasal stops are replaced by the corresponding oral
stops. Give hyponasal versions of:
singing [sŋŋ] onion [*njən] murmur [m+ mə]
(b) In hypernasal speech, where a velic closure is impossible because the
velum is paralysed or has a hole or split in it, air flows through the nasal
cavity the whole time. Oral stops become nasal stops and all sonorant
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 67 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Manner of articulation 67

sounds have accompanying nasal airflow (they are nasalised). These are
hypernasal versions of English words. Try and work out what they are.
The marks over the vowels indicate nasalisation.
[n
ã̃m]

ĩ n [m
ə̃]

ẽ̃m [mẽ̃ŋ̃ŋ]

Further reading
For a good, but fairly technical account of the aerodynamics of speech production, see chapter 3 of
J. C. Catford, Fundamental problems in phonetics, Edinburgh: Edinburgh University Press, 1977. For this
chapter, as for chapter 3, an excellent source of examples from many languages is Peter Ladefoged and Ian
Maddieson, The sounds of the world’s languages, Oxford: Blackwell, 1966.
0521808820c04_p51_68.qxd 1/27/05 2:59 PM Page 68 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 69 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

CHAPTER

5 Vowels

CHAPTER OUTLINE
In this chapter you will learn about the basic concepts nec-
KEY TERMS essary for the description and classification of vowels,
Cardinal vowel including: acoustic resonance, excitation spectrum and filter-
Diphthong ing by resonances, vowel spectra and formants, articulation
Excitation of vowels, the relationship between vowel articulation and
spectrum acoustic properties, the differences between monophthongs
Filtering and diphthongs, oral and nasalised vowels, long and short
Formant vowels. The chapter also deals with: vowel systems in the
Height world’s languages, vowel instability and variability.

Lip position
Location
Long and short
vowels
Monophthong
Oral and nasalised
vowels
Resonance
Spectrogram
Spectrum
Vowel
quadrilateral
Vowel system
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 70 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

70 INTRODUCING PHONETIC SCIENCE

Introduction
In the vast majority of languages, nearly all syllables include a vowel as
their central portion or nucleus. Vowels are sonorant sounds and are pro-
duced with a relatively open vocal tract. Vowels are nearly always voiced
sounds. This chapter will introduce the basic concepts necessary for the
description and classification of vowels, and will begin with a brief
description of the main acoustic properties of vowels, including the con-
cepts resonance, spectrum and formant. Also covered are the vowel
quadrilateral, cardinal vowels, vowel symbols and diacritics and the dis-
tinctions between long and short vowels and between monophthongs and
diphthongs. The chapter ends with a brief survey of the types of vowel sys-
tem found in languages around the world.

Spectrum and resonance

As water is poured into a glass, you can hear a note that gradually rises in
apparent pitch as the liquid approaches the top. Why is this? The volume
of air above the surface of the liquid in the glass has a resonant frequency.
This is the frequency at which it naturally vibrates. As the glass fills, this
volume gets smaller and the resonant frequency rises. We hear this rise in
frequency as a rise in pitch. Any enclosed body of air has a resonant fre-
quency which depends on both its volume and on the size of the opening
to the outside. Try this experiment for yourself: tap on the bottom of an
empty mug and then half cover the top with your other hand and tap
again. You should be able to hear a difference in the notes that are pro-
duced with the two different sizes of opening.
The mouth cavity is an enclosed volume of air. Its size and the size of
the opening at the lips can be altered, changing the resonant frequency of
the mouth. The equivalent of tapping on the bottom of the mug is the
vibration of the vocal folds in the larynx. No cavity will produce a sound
unless it is supplied with acoustic energy from an external source. This
energy constitutes the input or excitation. The excitation produced by the
larynx has many component frequencies. Only some of these will be
picked out and reinforced by the resonant characteristics of the vocal
tract.
You may be unfamiliar with the notion of a sound in which several dif-
ferent components are present at the same time. Think of this example. If
you pluck a single string on a guitar, a single note is produced. However,
if you pluck two or more strings at the same time, the notes produced are
added together. Moreover, you can pluck one string forcefully and others
more gently, so that the loudness of the individual notes varies. We don’t
generally hear the individual components of complex sounds, but the
components can be separated if such a sound is analysed on a computer.
The usual way of presenting such an analysis is called a spectrum.
Figure 5.1 shows a typical spectrum of larynx excitation. The horizontal
axis shows frequency and the vertical axis shows amplitude. Each of the
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 71 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Vowels 71

FIGURE 5.1
Schematic representation
of larynx spectrum.

vertical bars represents one of the components of the larynx tone before
it has been filtered by the vocal tract.
Notice that the components are at equal steps along the frequency
scale. So the lowest frequency component is at 100 Hz, then next at 200
Hz, the third at 300 Hz and so on. Notice also that the lowest frequency
component has the highest amplitude and that the spectrum shows an
overall downwards slope. These two features, the equal spacing of compo-
nents and the sloping spectrum, are important properties of the acoustic
energy produced by the human larynx. This energy is the input to the
vocal tract for vowels and other voiced speech sounds. The spacing of the
components may be wider or narrower and depends on the rate of vibra-
tion of the vocal folds. This affects our perception of the pitch of the
sounds we hear. The constant downward slope of the larynx spectrum is
modified by the vocal tract above the larynx to shape the spectrum in var-
ious ways and to produce the sounds we recognise as human speech.
There are many familiar examples in everyday life where resonance is
used to pick components from a complex spectrum and suppress unwanted
components. A radio or TV antenna brings in a vast spectrum of signals
and the tuner, which has a variable resonant frequency, works as a filter
to select the wanted signal. A filter need not pick out only one frequency
but can simultaneously alter the shape of the spectrum in a number of
different regions. The graphic equaliser on a sound system is an example
of this. You can boost certain frequencies and reduce others.
The vocal tract is a complex filter rather like a graphic equaliser. The
various resonances of the vocal tract filter out the energy in some fre-
quency regions of the input spectrum and leave other frequency regions
unaffected. This produces an output spectrum which has peaks and
troughs of energy. The peaks are called formants. Because of the sloping
shape of the input spectrum the lowest frequency formant has the high-
est amplitude and the amplitude of the higher frequency formants grad-
ually decreases as can be seen in the examples in Figure 5.2. The most
important property of a formant is its frequency and, as we shall see, the
formant frequencies of the output spectrum are crucial in distinguishing
vowels from one another. In principle there are many formants extending
up through the frequency range. In practice only the lowest two or three
are of importance for speech and they are labelled F1 (lowest), F2 and F3.
Figure 5.2 (a) shows the spectrum of the vowel sound [i] as in keep;
Figure 5.2 (b) shows the spectrum of the vowel [ɑ] as in calm.
For the vowel [i], F1 is low in frequency. In the figure shown, the peak is
at 250 Hz. Then there is a wide trough in the spectrum, indicating that in
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 72 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

72 INTRODUCING PHONETIC SCIENCE

(a)

(b)

FIGURE 5.2
Schematic spectra for (a)
[i] and (b) [ɑ] (after K.
Hayward, Experimental
Phonetics, Harlow:
Longman, 2000).

this frequency region the energy of the excitation has not been reinforced
by the resonances of the vocal tract. F2 is quite high in frequency — the
peak is around 2 kHz and F3 400 Hz higher.
Compare this with the spectrum for the vowel [ɑ]. The peak for F1 is
around 800 Hz. For F2, it is at 1100 Hz and F3 is at 2.6 kHz.
The precise formant frequencies for any vowel are determined by the
overall size of the speaker’s vocal tract — lower for males than females and
higher for children than for adults. This is because smaller cavities have
higher resonant frequencies than larger cavities. However, any speaker’s [i]
vowel will have a low F1 and a high F2 and any speaker’s [ɑ] vowel will have
F1 high and fairly close to F2. It is the pattern of formant frequencies and
their relationship to one another that is important rather than the
absolute values.
The location of formant peaks in the frequency range does not occur
arbitrarily, but is determined in an orderly way by the shape of the mouth
when forming a vowel.

Vowel articulation
Vowels are articulated by raising some part of the tongue body (that is the
front or the back of the tongue, not the tip or blade, see Figure 3.2)
towards the roof of the oral cavity. The shape of the opening formed by the
lips is also important.
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 73 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Vowels 73

Vowel height
The height of a vowel refers to the relationship between the highest point
of the tongue and the roof of the oral cavity. If the tongue is raised so as
to be close to the roof, then a close or high vowel such as [i] is produced. If
the tongue is only slightly raised, so that there is a wide gap between its
highest point and the roof of the oral cavity, then an open or low vowel,
such as [ɑ], results. Vowels produced with tongue positions between high
and low are sometimes called mid, such as the vowel in the English word
dress. Vowels between close and mid are called half-close (or high mid, or
close mid – the terms are interchangeable); vowels between open and mid
are half-open (or low mid, or open mid).
There is a direct connection between the height of a vowel and the fre-
quency of F1. High vowels have a low F1 and low vowels have a high F1.

Vowel location
The location of a vowel refers to the part of the tongue which is highest
in the production of the vowel. Front vowels, such as [i], are produced by
raising the front of the tongue towards the hard palate. Back vowels, such

(a) (b)

(c)

FIGURE 5.3
Lip position: (a) unrounded (spread) in a close vowel, [i]; (b) rounded in [u];
(c) unrounded (neutrally open) vowel [ɑ].
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 74 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

74 INTRODUCING PHONETIC SCIENCE

as [u], are produced by raising the back of the tongue towards the soft
palate. Central vowels, such as the vowel in the Southern British pronun-
ciation of the English word first, are produced by raising the centre part of
the tongue towards the junction of the hard and soft palates. A short mid-
central vowel, termed schwa and symbolised [ə] is in fact the most frequent
vowel in English speech. It is heard, for example, in the first syllable of
about, the middle syllable of photograph, and at the end of soda.
Again this aspect of the articulation of a vowel is reflected in the
spectrum. The frequency of F2 is high for front vowels and low for back
vowels.

Lip position
The position of the lips is the third factor which controls vowel quality.
For most purposes a two-way classification of rounded vs unrounded is
adequate. Figure 5.3 illustrates lip position.
Rounding the lips makes the mouth opening smaller and this, as we
saw with the experiment with the mug, has an effect on the resonant fre-
quencies. Rounding the lips lowers all formant frequencies, the most
important effect being on F2. In Figure 5.4 there are two spectra. (a) is for
the high front unrounded vowel [i] and (b) is for the high front rounded
vowel [y], as in the French word lune [lyn] ‘moon’.
The two spectra are quite similar, except for the position of F2.

[i] [y]

FIGURE 5.4
Spectra of real utterances of [i] and [y]. In the rounded vowel F2 is about 375 Hz lower.
F3 is also lowered, so that it is only about 250 Hz above F2.
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 75 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Vowels 75

Long and short vowels

Compared with many consonants, vowels are loud and long. But vowels
can be sustained for shorter or longer times, and many languages make
use of a distinction between long and short vowels. A long vowel in such
languages can be symbolised with a following length mark, thus [ɑ] is a
short unrounded back open vowel and [ɑ] is its long counterpart. Maori
(New Zealand) is a language that has long and short versions of each of its
five vowel qualities, and some examples are given below:

Long Meaning Short Meaning

[ki] to say [ki] at

[hoko] to buy [hoko] 20 times

[taha] calabash [taha] side

Most accents of English have vowel-length distinctions, but usually the

long and short members of a pair are also distinguished by clear quality
differences as well, for example the two vowels of the words beat and bit.

Oral and nasalised vowels

A further distinction that is utilised by many languages is between oral
vowels and nasalised vowels. An oral vowel is produced with the soft
palate in a raised position, so that there is no airflow through the nasal
cavity. A nasalised vowel on the other hand is produced with a lowered
soft palate, allowing airflow through both oral and nasal cavities simul-
taneously. The resonances of the nasal cavity are added to those of the rest
of the vocal tract and the vowel has a characteristic nasal quality.
Nasalisation is indicated by a diacritic [˜] over the relevant vowel symbol.
Table 5.1 shows pairs of French words that are distinguished by the
absence or presence of nasalisation of the vowel.

Table 5.1 Oral and nasalised vowels in French

fait [fε] ‘fact’ fin [fε̃] ‘end’

faux [fo] ‘false’ font [fõ] ‘(they) do’

Monophthong vs diphthong
In some languages (English is one) the quality of a vowel can change with-
in a single syllable. Such vowel-glides are known as diphthongs. Examples
of English words containing diphthongs are tie, toy, town. Vowels whose
quality remains relatively constant are known as monophthongs.
If the vowel is changing then the formant frequencies will be changing.
This can be seen in a display known as a spectrogram. Figure 5.5 shows the
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 76 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

76 INTRODUCING PHONETIC SCIENCE

FIGURE 5.5
Spectrogram of an utterance of the word ‘boy’.

spectrogram for the English word boy, which contains the diphthong [ɔ].
The horizontal axis shows time. The vertical axis shows frequency. The
darkness of the trace shows the amount of energy present at a given time
and frequency.
The formants show up as dark bands running horizontally. At the very
beginning the formants show brief movements (transitions) connected
with the production of the consonant sound. For the vowel, F1 starts high
and then descends. This is because the diphthong is moving from a low
quality to a high one. F2 starts low and rises, because the tongue is mov-
ing from a back position to a front one and at the same time the lips are
changing from a rounded posture to an unrounded one.

The vowel quadrilateral

The area in the oral cavity within which the highest point of the tongue
must occur for the articulation of a vowel is roughly like the quadrilateral
shown in Figure 5.6. The quality of a particular vowel can be indicated by
placing a dot on the diagram. The meaning of this dot is something like:
the vowel x sounds as if it is produced with the highest part of the tongue
in this position. The dot in the quadrilateral in the figure indicates the
usual quality of the vowel [] in the English word sit. The quality indicated
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 77 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Vowels 77

front central back

close

half close

half open

FIGURE 5.6
The vowel quadrilateral,
open and position of English [].

is around half close, front but not as front as possible. Lip position is not
indicated on the diagram. It used to be thought that the position of the
tongue within the mouth, together with lip position, fully determined the
quality of a vowel. It is now known that the configuration of the entire
vocal tract needs to be taken into account. The vowel quadrilateral has been
retained as a useful tool, but it should be thought of as representing an
auditory space, rather than an accurate articulatory one.

Cardinal vowels
As an alternative to specifying the location, height and lip-posture for a
vowel, the vowel quality can be related to one of a set of language-
independent, reference vowel qualities known as the cardinal vowels.
These can be thought of as landmarks in the auditory space provided by
the vowel quadrilateral. There are in fact two subsets of cardinal vowels:
the primary set [i e ε a ɑ ɔ o u], numbered 1–8, 1–5 being unrounded and
6–8 rounded and the secondary set [y ø   ɒ ɯ], numbered 9–16. The
secondary cardinal vowels are related to the primary ones by reversing the
lip-posture, so for example vowel 1 is close, front, unrounded and vowel 9
is close, front, rounded and vowel 8 is close, back, rounded, so vowel 16 is
close, back, unrounded. The cardinal vowels are auditorily agreed quali-
ties and must be learnt from a teacher who knows how they sound, or
from a recording. However, vowels 1 and 5 can be given an articulatory
specification. Number 1 is produced with the frontest, closest position of
the tongue that does not produce audible friction and number 5 is pro-
duced with the backest, openest possible tongue position, again without
audible friction. Figure 5.7 shows the primary and secondary cardinal
vowels on the vowel quadrilateral.
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 78 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

78 INTRODUCING PHONETIC SCIENCE

i u y ɯ

e o ø

ε ɔ 

a ɑ  ɒ
FIGURE 5.7
The primary cardinal vowels (left) and the secondary cardinal vowels (right).

Although some vowels in certain languages may be very like cardinal vow-
els (for instance the [i] in French is very like cardinal vowel number 1), the
cardinal vowels are independent of particular languages and must be
learnt.

Vowel diacritics
A reasonably accurate specification of vowel quality can be symbolised by
using diacritics attached to an appropriate cardinal vowel symbol. The
three diacritics used are shown attached to cardinal vowel number 2: [ë]
meaning centralised, that is closer to the centre of the quadrilateral, but
still not in the central triangle, [e] meaning closer and [e
] meaning more
The cardinal vowel open. There are a number of vowel symbols in common use that do not
system was invented
represent cardinal qualities, but that have fairly well-defined positions on
and developed by
the quadrilateral. These are shown in Table 5.2.
Daniel Jones
(1881–1967) who
was Professor of Table 5.2 Commonly used vowel symbols and their cardinal vowel equivalents
Phonetics at

ë
University College
London from 
ε
1921–1947. The sys-
υ
ö
tem was first used
in print in the first
edition of Jones’s Most of the phonetic symbols conventionally used to represent the vowels of
English a particular language are also used as cardinal vowel symbols, but this does
Pronouncing
not of course mean that the vowels of that language have exactly cardinal
Dictionary in
qualities. For example the vowel in Southern British cup is generally repre-
1917.
sented [ ], for instance in a pronouncing dictionary or textbook on English
pronunciation. This symbol represents an open-mid back unrounded cardinal
vowel, but the English vowel in question is actually a fairly open central
unrounded vowel. In the same accent, the vowel of dress is usually repre-
sented [e], but is noticeably lower than the cardinal quality represented by
this symbol. For certain applications, it is obviously more convenient to be
able to use a simple vowel symbol, without having to resort to diacritics.
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 79 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Vowels 79

Vowel systems
The vowel systems of the world’s languages differ widely in terms of size
and also in terms of the phonetic features used to differentiate vowels. We
have already seen that languages may make use of length distinctions and
that some languages use both oral and nasalised vowels. In this section we
will confine ourselves to languages that have only monopththongs and
that use only the basic features of height, location and lip position.
Figure 5.8 shows an estimate of the number of languages with systems
of a particular size. The information used to compile this figure was taken
from a database of 317 phonological systems known as UPSID (UCLA
Phonological Segment Inventory Database).

percentage of languages
30

20

10

FIGURE 5.8
1 2 3 4 5 6 7 8 9 10 >10 Size of vowel system in
number of vowel qualities UPSID languages.

The simplest system contains only three vowels – [i a u]. Many languages
of Australia, such as Aranda and Nunggubuyu, have a vowel system of this
type, but it is also found in other parts of the world.
The most frequently occurring vowel system has five vowels [i e/ε a o/ɔ u].
Some five-vowel system languages use close-mid vowels and some use
open-mid vowels. Table 5.3 lists some languages with this type of system.
As we can see, there are five-vowel languages in all parts of the world.
Taking the five-vowel system as a starting point, there are a number of
different ways in which languages may increase the number of vowel
qualities in the system. Some languages use both close-mid and open-mid
vowels, giving a seven-vowel system [i e ε a ɔ o u]. Efik (West Africa) is a
language with this system. Another way to augment a vowel system is to
add central or centralised vowels. Catalan, for instance, has a seven-vowel
system plus [ə]. Kpelle (West Africa) has a system containing [i e a o u] plus
the two centralised vowels [ υ].
All the systems we have mentioned so far have had front unrounded
vowels and back rounded ones. While this is the commonest type of sys-
tem, there are many languages that add front rounded vowels. French,
with the oral vowel system [i y e ø ε  a ɔ o u], is one example. However
there are no known languages that use front rounded vowels and that do
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 80 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

80 INTRODUCING PHONETIC SCIENCE

Table 5.3 Some five-vowel languages

Language Region

Batak Indonesia

Greek Greece, Cyprus

Japanese Japan

Kewa New Guinea

Maidu California

Maori New Zealand

Maung Australia

Spanish Spain, South and Central America

Yulu Sudan

Zulu South Africa

not use front unrounded ones. Back unrounded vowels also occur in quite
a number of languages. Modern Standard Chinese has an unrounded close
mid vowel [ ], as well as a rounded back vowel [o] at this height.

Vowel dispersion
We can view the vowel quadrilateral as a perceptual space in which vow-
els are located. The vast majority of vowel systems conform to the so-called
Vowel Dispersion Principle. This is that vowels tend to be evenly distrib-
uted in the perceptual space and or at least that they are widely distrib-
uted within the limitations of a particular system. This means, for
instance, that a three-vowel system containing [i e u] is very unlikely. Some
known systems appear not to conform to the above principle and these are
known as defective vowel systems. A defective system is one where either
(a) there is no open vowel and/or (b) for mid and high regions there is at
least one vowel that is not matched by a vowel on the opposite side of the
space. The [i e u] system qualifies as defective on both counts, but a system
containing [e o a] only would not count as defective. Approximately 14 per
cent of UPSID’s languages have vowel systems that count as defective.
Missing vowels in defective systems are almost always [e], [o] or [u], the last
being the most frequently missing. Another way of looking at this is to set
up a hierarchy:
i / a  e / o  u,

where  means is ‘more likely to occur in a language than’.

Vowel instability and variability

Try this simple experiment. Think of a short English sentence and tran-
scribe it, but replace every vowel symbol with []. If you were to do this
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 81 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Vowels 81

with the sentence Mary had a little lamb, the transcription would be [mr
hd  ltl lm]. Now practise saying this sentence until you are sure you
have it right and see if your friends can understand the sentence when you
say it to them. You will very probably find that most of them can. If you do
a similar thing with the consonants, replacing them all with [d] say, but
keeping the correct vowel qualities, the sentence will be unintelligible.
It seems then that, at least in some languages, vowels carry less infor-
mation than consonants. It is perhaps because of this that vowels tend to
be more variable and unstable than consonants. One consequence is that
vowels change over time more rapidly than consonants do. If we look at
the system of sounds used in two forms of the same language separated
in time by a few hundred years, it is a good bet that there will have been
more changes in the vowel system than in the consonant system.
We will illustrate the way a vowel system can change by looking briefly
at what is known as the Great Vowel Shift in English. This took place in
the fifteenth century and affected the long vowels of Middle English
(c. 1100–1450, the language Geoffrey Chaucer spoke) and was complete by
the time of Shakespeare, who spoke Early Modern English (c. 1450–1600).
Table 5.4 shows the changes that took place.

Table 5.4 Effects of the Great Vowel Shift in English

Middle English Early Modern English Present-day English

[i] [ei] [ai]

[e] [i] [i]

[ε] [e] [i]

Here is a recon-
[a] [ε] [e] struction in
Elizabethan pronun-
[ɔ] [o] [əυ]
ciation of the begin-
[o] [u] [u] ning of a famous
speech from
[u] [oυ] [aυ]
Shakespeare. Can
you read it?
frεndz,
So, for example, house was [hus] for Chaucer, [hoυs] for Shakespeare, and romənz,
is [haυs] today and moon in Chaucer’s day was pronounced [mon], but was k ntrimεn,
[mun] by Shakespeare’s time and still is. lεnd mi juɹ
Vowel changes such as this do not necessarily affect language as a whole iɹz
in the same way. They may take place in some geographical areas but not əi k m tu bεri
in others. This brings us to the second aspect of vowel variability. Two sezər, nɔt tu
accents of the same language are likely to differ much more in the vowels prεz him
and diphthongs they use than in their consonant systems. For instance, ði ivil ðat mεn
Southern British English is usually analysed as having eight diphthongs du lvz aftəɹ
ðεm
[e a ɔ aυ əυ ə eə υə], whereas one form of Jamaican English as only two
ðə ud iz ɔft
[a ɔυ]. So the two varieties agree in having a diphthong in a word such as
in tarid wið
time, but in a word like day, Jamaican English has a long monophthong [e] ðεɹ bonz
where Southern British has a diphthong [e].
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 82 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

82 INTRODUCING PHONETIC SCIENCE

It is not only the vowel systems that may be different in two accents of
a language. Accents may also differ in the lexical incidence of vowels –
that is, two accents may use a different vowel quality in some words, even
though the two accents have both of the vowels concerned. For instance,
both southern and some northern British English accents have [ɑ] and
both use it in words like palm, father, car. Both accents have [] in words
like cat, stand, gas. However for some words, path, dance, blast are examples,
southern accents have [ɑ] and northern accents have [].
A final point about the difference between the vowels of different
accents concerns the exact phonetic quality used to signal a particular
vowel. A good example of this is the difference in sound between the
vowel used in words like cow, town, out in southern British English and in
Northern Irish English. Both accents use a diphthong in these words and
there are no lexical incidence differences. However, the two diphthongs
sound quite different as is shown in Figure 5.9.

FIGURE 5.9
The diphthong of cow or mouth as spoken (left) in Southern British and (right) in
Northern Irish English. Notice how F2 falls in one but rises in the other.
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 83 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Vowels 83

Chapter summary
Vowels are produced when vocal fold vibration excites the resonances
of the vocal tract. The excitation spectrum is filtered by the vocal tract,
giving rise to peaks in the output spectrum which are known as for-
mants. The patterning of three lowest frequency formants essentially
determines vowel quality. This patterning can be related to the articu-
latory features of vowels (height, location, lip position). Vowels may be
long or short, and may be monophthongs or diphthongs. The auditory
properties of vowels can be related to a diagram known as the vowel
quadrilateral, and fixed landmark qualities in the quadrilateral are
known as cardinal vowels. Symbols for these, together with additional
symbols and diacritics permit the transcription of vowel qualities.
Languages make use of the basic vowel features in a wide variety of
ways. Vowel systems range from the very small with only three vowels
to the quite large with 10 or more vowels. Some languages increase the
size of their vowel inventory by the use of modifications such as nasali-
sation, length distinctions and the addition of diphthongs.
Vowels change over time more rapidly than consonants do and vow-
els can differ quite significantly from one accent of a language to
another. Two accents my have different vowel systems. They may differ
in the lexical incidence of vowels and the precise phonetic qualities of
equivalent vowels may differ.
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 84 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

84 INTRODUCING PHONETIC SCIENCE

Exercises
Exercise 5.1
Each of the following sets of cardinal vowels has an ‘odd one out’. Try to
identify this and say why it is unlike the others.

(a) [a ɑ i ɔ e ε]
(b) [e i a ø y u]
(c) [ɯ i y u ]

Exercise 5.2
In the speech of many New Zealanders, certain vowels appear to have
shifted in quality when compared to vowels used in other accents of English.
Here are some correspondences, represented in terms of the cardinal vowel
system with diacritics.

British English NZ English

pat [pε
t] [pet]
pet [pet] [pe
t]
pit [pe
t] [pət]
(a) Describe all the vowel qualities in these words in terms of height, location
and lip position.
(b) Plot the approximate position of the vowels on a vowel quadrilateral.

Exercise 5.3
Here are some words from Kirghiz, a Turkic language.

Meaning Pronunciation
apple [ɑlmɑ]
forest [tokoj]
house [yj]
lake [køl]
meat [et]
salt [tuz]
work [iʃ]
year [ɯl]

Kirghiz has a suffix, meaning with or by, which is pronounced in eight

different ways: [ten], [tøn], [den], [døn], [ton], [don], [tɑn], [dɑn]
When the suffix is added to the end of a word, the pronunciation is
chosen as follows:

(a) the first consonant of the suffix must agree in voicing with the last sound
of the word to which it is added
(b) the suffix vowel must agree in rounding with the vowels in the word
(c) the suffix vowel must agree in frontness or backness with the vowels in
the word

Which form of the suffix should be added to each of the words above?
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 85 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Vowels 85

Spectrograms for questions 5.4 and 5.5

The spectrograms and waveforms which follow are for questions 5.4 and
5.5. To assist you, the first is the English word ‘shout’ with lines added to
show division into three segments.
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 86 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

86 INTRODUCING PHONETIC SCIENCE

0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 87 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Vowels 87
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 88 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

88 INTRODUCING PHONETIC SCIENCE

Exercise 5.4
Look at the spectrograms (a)–(d) which represent English words of one sylla-
ble. Decide which words contain a monophthong and which a diphthong and
explain the reasons for your decisions.

Exercise 5.5
Words (a) and (b) end in a voiceless consonant and (c) and (d) end in a
voiced consonant. Measure the duration of the vowel in each word. For each
pair of words (a)–(c) and (b)–(d) calculate the ratio of the longer vowel to
the shorter.

Exercise 5.6
The spectrograms below are of the Turkish words [kil] ‘clay’ and [kyl] ‘ashes’.
Decide which is which and explain how you have reached this conclusion.
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 89 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Vowels 89

Exercise 5.7
The vowel quadrilaterals (a–d) below show the vowel systems of four real
languages.
(a) Which of the four languages has a mid central vowel [ə]?
(b) Which of the four languages has a fully open back vowel?
(c) Which languages have a vowel that could be symbolised []?
(d) Which language does not have a fully back close vowel?
(e) Which language has a vowel very similar to primary cardinal vowel
number 3?

(a) (b)

(c) (d)

Further reading
On vowel systems and vowel dispersion, see Ian Maddieson, Patterns of sounds, Cambridge: Cambridge
University Press, 1984, chapters 8 and 9.
A wide-ranging account of the varieties of English pronunciation found around the world, very useful for
comparing the vowel systems of different varieties of English, is J. C. Wells, Accents of English, 3 vols.
Cambridge: Cambridge University Press, 1982.
0521808820c05_p69_90.qxd 1/27/05 3:00 PM Page 90 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:
0521808820c06_p91_102.qxd 1/27/05 3:01 PM Page 91 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

CHAPTER

6 Voice II

CHAPTER OUTLINE
In this chapter you will learn about: aspiration; aspiration
KEY TERMS and devoicing in English; the use of aspiration in languages
Aspiration of the world; larynx waveforms for normal voice and other
Breathy voice phonation types; voice onset time and how to measure it.
Devoicing
Fortis
Lenis
Modal voice
Phonation type
VOT (voice onset
time)
0521808820c06_p91_102.qxd 1/27/05 3:01 PM Page 92 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

92 INTRODUCING PHONETIC SCIENCE

Introduction
As we saw in Chapter 2, one important way in which consonant sounds
may differ from one another is by voicing. However, it is important to
be aware that the terms ‘voiced’ and ‘voiceless’ do not simply refer to
the presence or absence of vocal fold vibration. The articulatory and
acoustic events that give rise to the perception of voicing and voiceless-
ness are somewhat more complex than the simple story we told in
Chapter 2. What counts as voiced in one language may, in some
instances, be perceived as voiceless in another language. In this chapter
we look at the topic of voicing in more detail and consider the impor-
tance of the timing of the onset of vocal fold vibration in relation to
other articulatory events. We also look in more detail at vocal fold vibra-
tion itself, and distinguish normal voice from other phonation types.

Voice Onset Time

For voiceless plosive consonants, vocal fold vibration is stopped for a
period that is a little longer than the hold phase (the time when the
mouth is blocked, see Chapter 4), so that there is still no vocal fold vibra-
tion around the moment of release (when the articulators part and the
plosion is heard) and possibly for a further brief time afterwards. This
delay, measured from the start of the explosion to the point where vocal
fold vibration begins, is called the Voice Onset Time (VOT). Though so
short that it is best expressed in milliseconds (1 ms
one thousandth of
a second) it is very important to the listener. We know from experiments
on speech perception that listeners are very sensitive to VOT and use it to
categorise the plosive they are hearing as voiceless or voiced. In Figure 6.1
below there is a schematic representation of the three phases of the pro-
duction of a voiceless velar plosive [k]. Also included in the figure are the
waveform and spectrogram of a real utterance of [aka].

Aspiration
If the VOT is longer than about 30 ms, a plosive doesn’t just sound voiceless;
the VOT can actually be heard as a brief [h]-like segment following the explo-
sion and the plosive is said to be aspirated. Not all languages have voiceless
plosives of this type, but English does, and the initial [p] [t] [k] sounds of
words such as pound, time, keen are all aspirated. The symbol for aspiration
is a superscript [h] following the sound affected – so [ph th kh] and so on.

Aspiration in languages
The commonest pattern in languages around the world is for the voiceless
plosives to have a short VOT (something like 10 ms on average); they are
therefore voiceless unaspirated plosives. The voiced plosives with which
they contrast tend to be fully voiced – that is, to have vocal fold vibration
started well before the explosion, and possibly running right through the
hold phase of the articulation. Among languages which do this are
0521808820c06_p91_102.qxd 1/27/05 3:01 PM Page 93 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Voice II 93

FIGURE 6.1
Schematic parametric representation of the three phases of an intervocalic plosive, and
diagrammatic representation of larynx activity, with aligned waveform and spectrogram
of a real utterance.

French, Spanish, Italian, Greek, Dutch, Russian, Arabic. There are hun-
dreds more.
English does it a different way. The voiceless plosives have a longer VOT
and are thus aspirated. The voiced plosives do not need to be fully voiced,
and, depending on the context in which they appear, they may lack vocal
fold vibration for some or all of the hold. It is enough if vocal fold vibra-
tion begins at about the time of the release – in fact initial plosives with
vocal fold vibration beginning much before this do not sound quite right
for English. Other languages that do it much the same way as English
include German, Japanese and Chinese (both Modern Standard and
Cantonese).
0521808820c06_p91_102.qxd 1/27/05 3:01 PM Page 94 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

94 INTRODUCING PHONETIC SCIENCE

Figure 6.2 shows the waveforms for voiced and voiceless plosives in
French (a, b) and for English (c, d). The waveforms are aligned at the
moment of release of the plosive.

FIGURE 6.2
Waveforms showing
intervocalic plosives in
French and English,
aligned by moment of
release. The words are
French (a) apart,
(b) abeille, and English
(c) apart, (d) obey. In
this context the voiced
plosives show voicing
throughout the hold
phase in both languages.

Aspiration is a matter of degree, and plosives can easily be given more

aspiration (i.e. longer VOT) than they customarily get in English. For
instance, the aspirated plosives of Modern Standard Chinese sound some-
what more aspirated than English ones. Figure 6.3 is a schematic repre-
sentation of the full range of possibilities for VOT.
Some languages (an example is Thai) actually distinguish three sets of plo-
sives by VOT. These are likely to be (1) fully voiced (2) short VOT (10 ms or so
– i.e. voiceless unaspirated) (3) very long VOT (50 ms or more – i.e. strongly
aspirated). The examples in Table 6.1 illustrate this three-way contrast.
The remarks so far apply to plosive consonants in initial position
directly before a vowel. If a language distinguishes voiced and voiceless
plosives at all it will do so in that position; in other positions, the sounds
may be modified. For instance, voiced plosives may become fricatives or
approximants when flanked by voiced sounds, as they do in Spanish. An
example is the word beber (‘to drink’), which is pronounced [beer]. In final
position, it is common for the voiced/voiceless contrast not to operate (it
is said to be neutralised). This is the case in German, for instance, so while
0521808820c06_p91_102.qxd 1/27/05 3:01 PM Page 95 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Voice II 95

Table 6.1 Voicing and aspiration in Thai plosives

voiced voiceless voiceless
unaspirated aspirated

[ba] crazy [pa] aunt [pha] cloth

[d a ] to curse [ta] eye [tha] landing place

the word Bundes (‘of the union’) is [bυndəs], Bund (‘union’) is [bυnt]. In
English the contrast is maintained in all positions except directly after [s].
The plosives in words such as spin, stop, school have approximately zero VOT
and are more-or-less ambiguous between voiced and voiceless categories
for speakers of English. Children often invent spellings such as sdop, sgool,
showing their accurate perception of the neutralisation.

approach hold release

Relative
position of
articulators
e.g. lips

Fully voiced

Pre-voiced

Zero voicing

Short VOT

Long VOT
FIGURE 6.3
Schematic parametric rep-
resentation of the three
VOT = time
from release of phases of plosive produc-
plosive to onset tion, with five different
of voicing VOT possibilities.

Devoicing
English [p t k] are aspirated when at the beginning of a stressed syllable.
If an approximant rather than a vowel follows (e.g. price, play, twin, cute)
the aspiration takes the form of devoicing the approximant, so: [pɹas
ple twn kjut]. If [s] precedes, there is no aspiration and, correspond-
ingly, little or no devoicing of any approximant, so for example, sprain
[spɹen], splash [splʃ ], squeak [skwik], studio [stjudiəυ].
Just as the ‘voiced’ plosives of English are not necessarily voiced all the
way through, so the other ‘voiced’ obstruents (fricatives and affricates)
don’t always have vocal fold vibration throughout. The contrast is main-
tained for the listener by a number of other factors (duration is the most
important). To avoid the awkwardness of talking about ‘voiced’ sounds that
are not voiced, English consonants are sometimes called lenis (
weak i.e.
short, quiet, unaspirated, sometimes voiced) and fortis (
strong i.e. long,
loud, aspirated, always voiceless).
0521808820c06_p91_102.qxd 1/27/05 3:01 PM Page 96 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

96 INTRODUCING PHONETIC SCIENCE

Figure 6.4 shows spectrograms of the English words ice [as] and eyes
[az]. While we conventionally represent the difference between these
words as a difference in the voicing of the final consonant, vocal fold

FIGURE 6.4
Spectrograms of the English words ice (top) and eyes. The diphthong of ice is short-
ened by pre-fortis clipping, while the fortis fricative is considerably longer than the lenis
fricative of eyes.
0521808820c06_p91_102.qxd 1/27/05 3:01 PM Page 97 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Voice II 97

vibration is not present throughout the fricative [z]. The most important
difference for the listener is in the duration of the [a] diphthong. You can
see that the [a] in ice is considerably shorter than that in eyes. The vowel
has been shortened by the presence of a following fortis consonant and
this is a regular feature of many accents of English and also of other lan-
guages. It is known as pre-fortis clipping.

Phonation type
The switching on and off of vocal fold vibration, and the timing of these
events, are not the only ways in which the larynx contributes to sound dis-
tinctions in languages. The nature or type of vibration may also be signifi-
cant. Sound generation in the larynx is termed phonation, and phonation
types (sometimes called voice qualities) are recognisably different kinds of
vocal fold vibration. All languages use a phonation type called normal or
modal voice as the basis for all speech. Many languages (for example
English) make no systematic deviations from modal voice. In modal voice,
the vocal folds vibrate regularly with a rapid closure in each cycle as shown
in the laryngograph waveform in Figure 6.5. Remember that the output of
the laryngograph shows vocal fold contact as a function of time.

one cycle

opening

closing
open phase

FIGURE 6.5
Laryngograph waveform
0 ms 5 10 15 20 25 for modal voice.

In the figure, the main events in one cycle are labelled. The rapid clo-
sure of the vocal folds causes a very efficient excitation of vocal tract res-
onances. Modal voice is a tone in which the individual cycles of vibration
are not separately audible and in which there is little or no noise (aperi-
odic energy) from breath flow, and it can be produced at a wide range of
frequencies.
Departure from the laryngeal adjustments required for normal voice
may result in such phonation types as creaky voice and breathy voice.
Most widely used in languages is breathy voice, sometimes called mur-
mur, where audible breath noise is heard at the same time as vibration.
A laryngograph waveform for one type of breathy voice is shown in
Figure 6.6(a). The closed phase in each cycle is short, leaving a long open
0521808820c06_p91_102.qxd 1/27/05 3:01 PM Page 98 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

98 INTRODUCING PHONETIC SCIENCE

(a) breathy

short closure long open phase

(b) creaky irregular double pulsing

FIGURE 6.6
Laryngograph waveform
for (a) one type of
breathy voice, and
(b) one type of creaky
voice. 0 ms 5 10 15 20 25

phase during which breath escapes. In creaky voice the vocal folds
vibrate in a less regular manner with alternate cycles of vibration dif-
fering in duration. This can be seen in the laryngograph waveform in
Figure 6.6(b).
An example of a language that makes use of the distinction between
normal voice and breathy voice is Gujarati. Some pairs of Gujarati words
are given in Table 6.2. The mark  below the vowel symbol indicates breathy
voice.
A language that makes use of the distinction between normal voice and
creaky voice is Jalapa Mazatec, spoken in Mexico. In the examples in Table 6.3
the mark is used to indicate creaky voice.
~
Table 6.2 Modal and breathy voiced vowels in Gujarati
modal breathy
voice voice

[baɾ] twelve [baɾ] outside

[poɾ] last year [poɾ] early morning

Table 6.3 Modal and creaky voiced vowels in Jalapa Mazatec

modal voice creaky voice

[ ja] tree [ ja] he wears

˜
h
[nt ] seed [nd
] buttocks
˜
0521808820c06_p91_102.qxd 1/27/05 3:01 PM Page 99 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Voice II 99

We may also convey attitude or mood by altering the way our vocal folds
vibrate. Breathy voice is often associated with passionate, intimate utter-
ances. It may be more convincing to say I love you with breathy voice than
with modal voice. This sort of phonation also is associated with gushing
overenthusiasm. Creaky voice is often used when the speaker wishes to
signal boredom and is also found in exasperated utterances like Oh no, not
again!

Chapter summary
In this chapter we have looked at aspects of the timing of vocal fold
vibration. VOT is the interval between the release of a plosive and
the onset of vocal vibration and is an important cue for the listener
in determining whether the plosive belongs in the voiced or voice-
less category. Most languages make use of a two-way distinction
according to VOT. The chapter also introduced the notion of aspira-
tion, a feature of plosives that have a long VOT. Looking more close-
ly at the details of voicing in English we considered the topic of
devoicing and looked at the conditions under which so-called voiced
consonants may be without vocal fold vibration for at least part of
their duration. This led to the introduction of the terms fortis and
lenis as possible alternatives to the terms voiceless and voiced.
Finally, we looked at phonation types, identifying modal (normal)
voice and a number of audibly different kinds of vocal fold vibration
which may be used distinctively by certain languages and in many
languages as signals of the attitude of the speaker. The most impor-
tant conclusion to be drawn from this chapter is that the voicing of
consonants is not simply a matter of the presence or absence of
vocal vibration.

Exercises
Exercise 6.1
Sketch a schematic diagram showing vocal fold activity for an aspirated voice-
less plosive between two vowels, and indicate the following events: hold
phase, moment of release, onset of voicing.

Exercise 6.2
In most accents of English, voiced obstruents are only fully voiced when they
are in between two voiced sonorant sounds (for example the [b] in abbey
[ bi]). Elsewhere they are devoiced – that is, the vocal folds do not vibrate
for the whole duration of the sound . For example this book will contain a
devoiced bilabial plosive [b
]. Add the devoicing diacritic to the relevant
symbols in the transcriptions below:
(a) [ jɔ nju kredt kɑd ʃυd ərav sun]
(b) [ ʃɒpŋ n selz z ə υd we tə et bɑnz]
0521808820c06_p91_102.qxd 1/27/05 3:01 PM Page 100 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

100 INTRODUCING PHONETIC SCIENCE

Exercise 6.3
Shown below are the waveform and spectrogram of an utterance of the
English word apart. Measure the VOT of the intervocalic plosive (in millisec-
onds, to the nearest 5 milliseconds).
Exercise 6.4
Consider the voiceless velar plosives in your pronunciation of the following
words: kite sky occur baker basket. Which do you think have a long VOT,
and which a short one? If possible, use computer software to display
waveforms of recordings of your versions, and measure VOT for each velar
plosive.
Exercise 6.5
Shown below are short specimens of the larynx waveform (obtained from a
laryngograph). Which of the examples (a) to (f) look as if they are likely to
be examples of normal or modal voice?

Figure for Exercise 6.3

Exercise 6.6
Remember that pre-fortis clipping is a reduction in the duration of a vowel or
diphthong when it is in a syllable that has one or more fortis consonants at
the end. In which of the following monosyllabic English words are the vowels
or diphthongs clipped?
(a) fee (b) feet (c) feast (d) fade (e) fail (f ) felt
0521808820c06_p91_102.qxd 1/27/05 3:01 PM Page 101 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Voice II 101

Exercise 6.7
Which of the following English words are likely to contain a devoiced
approximant?
(a) please (b) blow (c) quick (d) crown (e) grow
(f ) pure (g) yours (h) cool (i) string ( j) what

(a)

(b)

(c)

(d)

(e)

(f)

Figure for Exercise 6.5

Further reading
John Laver, The phonetic description of voice quality, Cambridge: Cambridge University Press, 1980, contains
an account of many different phonation types. For more detail on how languages differ in using VOT and
aspiration you could look at L. Lisker and A. S. Abramson, A cross-language study of voicing in initial stops:
Acoustical measurements, Word, 20 (1964), 384–442.
0521808820c06_p91_102.qxd 1/27/05 3:01 PM Page 102 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:
0521808820c07_p103_116.qxd 1/27/05 3:01 PM Page 103 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

CHAPTER

7 Airstream
mechanisms

CHAPTER OUTLINE
In this chapter you will learn about: airstream generation;
KEY TERMS egressive and ingressive flow; ejectives; implosives; clicks;
Click non-pulmonic sounds in the world’s languages.
Compression
Ejective
Egressive
Glottalic
Implosive
Ingressive
Pressure
Pulmonic/non-
pulmonic
Rarefaction
Velaric
0521808820c07_p103_116.qxd 1/27/05 3:01 PM Page 104 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

104 INTRODUCING PHONETIC SCIENCE

Introduction
We have seen how consonant sounds can be described in terms of voice,
place and manner of articulation. We now have to consider a further impor-
tant way in which consonant sounds can differ from one another: the use
of different airstream mechanisms. All the sounds we have dealt with up
to now have used the same airstream mechanism, known as the pulmonic
airstream mechanism because it uses air expelled from the lungs. In the
pulmonic airstream, the lungs supply a large volume of air under pressure,
enough to power the production of one or more phrase-length stretches of
speech between pauses for breath. This is the basis of all normal speech in
all languages, and many languages (e.g. English) make no systematic use of
any other airstream. But some languages additionally make use of other
airstream mechanisms for a proportion of their consonant sounds. Small
volumes of air can be pushed or pulled by muscular action in the mouth or
pharynx independently of the lungs, and the resulting short-term pressure
differences and airflows are enough to power the production of single con-
sonant segments, which sound recognisably different from pulmonic
sounds. There are three types of consonants produced this way, known as
ejectives, implosives and clicks. Because they are not driven by air pressure
from the lungs they are termed non-pulmonic sounds.
This chapter first looks more closely at the generation of air-pressure
differences and airflows, first introduced in Chapter 4. It then describes
the way in which the various non-pulmonic sounds are produced, how
they are symbolised, and gives some examples of how they are used in the
consonant systems of the world’s languages.

Airstream generation
In order to make any speech sound, there must be a flow of air through
the relevant part of the vocal tract. Flow through the larynx is needed
whenever we require to vibrate the vocal folds; flow though a narrow
articulatory constriction is needed to generate the noise typical of frica-
tives; and an interrupted flow, which is suddenly re-started, gives us the
transient sound of plosives.
Air movement within the vocal tract depends on the creation of pres-
sure differences. Air always flows from a region of high pressure towards
a region of lower pressure. Air is a gas, or more accurately a mixture of
gases. A gas may be thought of as particles spread out in space. Pressure is
a measure of how many particles there are in a given volume of space. If
there are few particles, the gas is at low pressure. Making the space smaller
without letting any of the particles escape is called compression, and rais-
es the pressure of the gas. The opposite of this, rarefaction, is obtained
when the space is enlarged without letting any more particles in. The con-
centration of particles is thus lowered and the pressure is reduced. In
speech, the necessary compressions and rarefactions are achieved by
using muscles to move some part of the vocal tract or respiratory system.
0521808820c07_p103_116.qxd 1/27/05 3:01 PM Page 105 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Airstream mechanisms 105

The moving part acts rather like the piston in a pump, and is called the
initiator of the airstream. For example, in the pulmonic airstream the ini-
tiator is the lungs, moved by the respiratory muscles.
When the pulmonic airstream mechanism is in action, there is a supply
of air at high pressure in the lungs and it flows through the vocal tract
towards the lower atmospheric pressure in the air around the speaker.
Much of the muscular work to power the airstream is actually done when
we pause to breathe in. We use our inspiratory muscles to inflate the lungs,
which then tend to recoil back, compressing the air within them. The air
doesn’t come out all at once, of course, because its flow is regulated by the
adjustment at the glottis, and the resistance of the various articulatory
constrictions we are employing as we speak. As we continue speaking, and
the lungs begin to empty, we begin to use expiratory muscles to squeeze
more air out of the lungs before we finally have to pause for breath.

Egressive and ingressive flow

The flow of air we have just described is outwards from the body, a direc-
tion of flow that is called egressive. So the full name for the airstream
mechanism that forms the basis of all normal speech is the egressive pul-
monic airstream mechanism. You will find that it is also possible to
reverse this airstream and speak while breathing in, using the ingressive
pulmonic mechanism. Try saying a short sentence such as ‘Good morning,
how are you?’ this way. You will probably find that you can make a recog-
nisable attempt at this, but the speech is likely to sound strange, and is dif-
ficult for most people to control. Notice that both voiceless and voiced
sounds can be made this way, proving that the vocal folds can be made to
vibrate when air passes between them in a reversed direction. So far, no
language has been found that makes any use of the ingressive pulmonic
airstream. But the distinction between egressive and ingressive is
important in understanding the non-pulmonic mechanism we are going
to consider next.

Ejectives
The first of the non-pulmonic airstream mechanisms we will look at is the
one that is used to produce ejectives, which are the commonest non-
pulmonic consonants among the languages of the world. Remember that
languages that make use of ejectives will still make use of the pulmonic
egressive airstream as the ordinary basis of speech. Each time an ejective
consonant occurs, the non-pulmonic mechanism is powered-up briefly,
but there will generally be a return to the pulmonic mechanism for the
following segment.
In the production of an ejective, the glottis is closed, and the air that is
compressed (and eventually pushed out) is that between the closed glottis
and the articulation in the oral cavity. For this reason, the name of the
airstream is egressive glottalic. Ejectives are symbolised by adding an
apostrophe to the symbol for any voiceless obstruent. This gives symbols
such as [p’ t’ k’ tʃ’].
0521808820c07_p103_116.qxd 1/27/05 3:01 PM Page 106 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

106 INTRODUCING PHONETIC SCIENCE

1 Back of tongue
rises to form
velar closure 4 Back of tongue is
lowered, releasing
Glottis closes compressed air in the pharynx

FIGURE 7.1
3 Body of air in
The sequence of events pharynx is
in the production of an compressed
ejective velar stop (after 2 Closed glottis
P. Ladefoged, A Course in is raised
Phonetics, Fort Worth, Tx: 5 Glottal closure
Harcourt, 2001). is released

The sequence of events in the production of a typical ejective [k’] is as

follows. As shown in Figure 7.1, (1) the back of the tongue rises to form a
closure on the velum, just as for any velar stop. At about the same time,
the glottis closes. The velum is also in its raised position, closed against
the rear wall of the pharynx, so that no air can escape through the nose.
(2) The closed glottis is raised (the whole larynx moves up) reducing the
volume of the pharynx cavity and thus (3) increasing the pressure of
the air in the pharynx. (4) The back of the tongue is lowered, just as for the
release of an ordinary velar stop, thus releasing the compressed pharynx
air and producing an audible explosion. The glottis remains closed during
the release and for a short time afterwards. Finally (5) the glottal closure
is released – for instance, the vocal folds may begin to vibrate for a fol-
lowing vowel. This description is of an unaffricated ejective stop, and this
is the commonest type of ejective found in languages. Other types are pos-
sible, however, and both ejective affricates and less commonly ejective
fricatives are found in languages.

Acoustic and auditory properties of ejectives

Ejectives sound different from pulmonic egressive consonants. One of the
most noticeable differences is that the burst for an ejective stop tends to
be much more intense. This is because a higher pressure difference can be
produced with the glottalic airstream mechanism. During the burst of an
ejective and for a short time afterwards, the glottis remains closed. This
means that the burst is followed by a period of silence rather than by aspi-
ration or the immediate onset of voicing. Languages differ in the length
of time for which the glottis remains closed, and if the time is very short
the ejective auditorily resembles an unaspirated pulmonic egressive plo-
sive quite closely. You can see the differences between a typical ejective
and the corresponding pulmonic consonant in Figure 7.2.

Ejectives in languages
Ejectives are the commonest type of non-pulmonic sound. They occur in
many languages of America and Africa and in languages of the Caucasus.
They occur alongside the corresponding pulmonic sounds, and the
0521808820c07_p103_116.qxd 1/27/05 3:01 PM Page 107 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Airstream mechanisms 107

FIGURE 7.2
Waveforms and spectrograms for [akha] (left) and [ak’a] (right).

difference in airstream mechanism can be the only difference between

contrasting words. For example in Hausa, which is spoken in Nigeria,
[kaa] means ‘to screen off’, whereas [k’aa] means ‘to increase’. In
Chechen, one of the Caucasian languages, [tsa] means ‘not’, but [ts’a] with
an ejective affricate means ‘house’.
Here are examples showing a range of contrasting ejectives and pulmonic
sounds from Chontal (Mexico):
[p’os] ‘sweepings’ [pos] ‘pale’
[t’ub] ‘gourd’ [tub] ‘spittle’
[k’uʃ] ‘it hurts’ [kuʃ] ‘go’
[ts’ilen] ‘split’ [tsilen] ‘rip’
[tʃ’uju] ‘lift it up’ [tʃuju] ‘sew it’

Other uses of ejectives

Even in languages where the difference between ejectives and pulmonic
consonants is not used to signal a difference between words, speakers may
occasionally use ejectives. For example the English phrase Definitely not
may be said with an ejective at the end, giving the impression that the
speaker is being decisive or insistent. Most speakers of English use ejec-
tives only occasionally, and only when a silent pause follows. We don’t
0521808820c07_p103_116.qxd 1/27/05 3:02 PM Page 108 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

108 INTRODUCING PHONETIC SCIENCE

hear normal speakers of English using ejectives in initial position before

a vowel.
Ejectives may sometimes be used in place of pulmonic consonants in
certain types of disordered speech, so that those who are training to be
speech and language therapists must learn to recognise them.

Implosives
Implosives use the ingressive glottalic airstream; instead of moving up to
compress the air as in ejectives, the glottis is moved down so as to enlarge
the pharynx and produce suction. So implosives are essentially reverse
ejectives, but straightforward voiceless implosives are actually very rare in
languages. The implosives that are commonly encountered have a further
small complication in that they are voiced. Instead of being completely
closed as it is pulled down, the glottis is vibrating to produce voice.
Obviously, to produce voice, air from the lungs must be flowing up
through the glottis. Voiced implosives thus employ two airstream mecha-
nisms simultaneously. The symbols for implosives are based on those for
voiced plosives, and always have a hook attached to the top right: [  ].
The sequence of events in a typical voiced implosive, [] is shown in
Figure 7.3. (1) The lips close, as for any bilabial stop. (2) As the vocal folds
continue vibrating in a flow of lung air, the whole larynx is moved down,
thus enlarging the pharynx and tending to reduce the pressure of air
inside it. (3) The reduction in air pressure may not be very great, because
it is offset by further air from the lungs flowing through the glottis to
keep vocal fold vibration going. When (4) the lips are opened, air pressure
in the mouth is still a little below atmospheric pressure, so airflow
(briefly) is inwards.

1 Lips close
4 Lips come
apart
3 Pharynx enlarged
air pressure is slightly
reduced. (This is offset
by the air from the
FIGURE 7.3 lungs.)
The sequence of events
2 Downward movement
in the production of a
of vibrating glottis.
bilabial implosive (after P. Air from the lungs
Ladefoged, A Course in continues to flow
Phonetics, Fort Worth, Tx: through the
Harcourt, 2001). glottis.
0521808820c07_p103_116.qxd 1/27/05 3:02 PM Page 109 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Airstream mechanisms 109

Auditory and acoustic characteristics of implosives

Implosives sound like rather peculiar fully voiced plosives. From the lis-
tener’s point of view, an implosive is not recognised from the ingressive
airflow itself, but from the characteristic sound which results from the
complex change of shape during the production of the consonant. There
are probably also cues for the listener in the type and frequency of vocal
fold vibration that occurs during the implosive manoeuvre.
Figure 7.4 contains spectrograms to illustrate the difference between a
typical voiced pulmonic egressive plosive and a typical implosive.

FIGURE 7.4
Waveforms and spectrograms for a voiced pulmonic plosive (left) and a voiced implo-
sive (right). The fundamental frequency curve (bottom) shows that the two manoeu-
vres can have different effects on the rate of vocal fold vibration.

Implosives in languages
Implosives are the next most common type of non-pulmonic sound after
ejectives. They occur in numerous languages of America, Africa and Asia.
In certain languages, implosives contrast with voiced pulmonic egressive
plosives. In others (Swahili is an example) implosives occur, but as variants
of voiced plosives.
0521808820c07_p103_116.qxd 1/27/05 3:02 PM Page 110 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

110 INTRODUCING PHONETIC SCIENCE

Sindhi, a language of Pakistan, has both voiced pulmonic egressive plo-

sives and implosives at various places of articulation. Some of these are
illustrated below.

bilabial alveolar velar

pulmonic [buυ] ‘shoes’ [dυnυ] ‘navel’ [ ano] ‘song’
implosive [aɾυ] ‘child’ [aɾυ] ‘crevice’ [əɾo] ‘heavy’

Unlike ejectives, implosives are not used in English as stylistic variants

of pulmonic plosives. They are, however, sometimes heard in disordered
speech, especially from speakers with severe hearing impairment.

There seem to be universal tendencies in the places of articulation

that are favoured for ejective and implosive consonants. Looking at
cross-language generalisations, for languages that do not have com-
plete sets of implosives and ejectives, the commonest ejective is [k’]
while the commonest implosive is []. Hausa, for example, makes use
of both implosives and ejectives, but the implosives are found only at
front places of articulation, and there is no ejective at the bilabial
place. There may be a natural explanation of these preferences in the
way that the airstreams are generated. For example [k’] has a back
place of articulation, and thus a small initial volume of air trapped
above the glottis. A fixed amount of larynx-raising will therefore pro-
duce a greater increase in pressure than would be available with a
more forward place of articulation.

Clicks
Clicks use the ingressive velaric airstream. Figure 7.5 shows the sequence
of events in a dental click. (1) The tongue tip forms a dental closure, just
as it would for a dental plosive. At the same time, the back of the tongue
is raised to touch the velum, and the sides of the tongue are in contact

1 Tongue tip up 1 Back of tongue raised

to form dental to form velar closure
closure 2 Body of the tongue moves down while
velar and dental closures are maintained
3 Pressure of air in this part of
4 Tongue tip the mouth is decreased
lowered so that
air rushes into
FIGURE 7.5 the mouth
The sequence of events
in the production of a
dental click (after P.
Ladefoged, A Course in
Phonetics, Fort Worth, Tx:
Harcourt, 2001).
0521808820c07_p103_116.qxd 1/27/05 3:02 PM Page 111 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Airstream mechanisms 111

with the palate and teeth, so that a small volume of air is trapped between
the tongue and the upper surface of the mouth. (2) While the two closures
are maintained, the body of the tongue moves down so as to enlarge the
volume and (3) the pressure of the enclosed air decreases. (4) The dental
closure is released and a characteristic click sound is heard as air rushes
in. The action of the tongue against the upper surface of the oral cavity is
rather like that of a rubber suction cup being pressed against a surface
and then pulled away.
For all clicks the back of the tongue is making a closure against the
velum, but the forward closure may occur at various places of articula-
tion. Those included in the IPA are shown in the table below. Notice
that the IPA symbols for clicks are not based on letters of the Roman
alphabet.

bilabial dental alveolar palatoalveolar alveolar lateral

[◎] [] [!] [] []

The symbol shown here for an alveolar click is also used to represent clicks
made with the tongue tip further back, and whose place of articulation is
described as postalveolar or retroflex. The characteristic sounds of the dif-
ferent clicks result not only from differences in place of articulation, but
also from the type of release. For example the dental click may character-
istically have a noisy, affricated release, while the palatoalveolar click
tends to have a quick, sharp release. For the alveolar lateral click, the clo-
sure is released at the side of the tongue rather than in the centre.

Click modifications
Clicks are produced by a manoeuvre that takes place entirely within the
mouth, leaving the lungs, larynx, pharynx and nasal cavities free to act
independently and so add various modifications to clicks, including voic-
ing and nasality. The basic clicks themselves are of course unvoiced, but if
during the production of a click the vocal folds are made to vibrate on a
flow of egressive pulmonic air, and if at the same time the velum is down,
allowing the voiced airstream to escape via the nose, the click is said to be
voiced and nasalised, although it might be more accurate to say that the
click is accompanied by voice and nasality. This is symbolised with [ŋ]
before the click symbol, because the pulmonic air flowing up through the
vibrating larynx and into the nasal cavity encounters a vocal-tract posi-
tion just like that for a regular velar nasal consonant (recall that clicks
have the back of the tongue closed against the velum). So a voiced
nasalised dental click is represented as [ŋ ]. If the vocal folds are made to
vibrate, but the velum is held shut, the click will be accompanied by voice,
but not by nasality, and the result is described as a voiced click. This is rep-
resented with symbol [ ] along with the click symbol. So for example [ ◎]
is the symbol for a voiced bilabial click. It is worth pointing out that when
clicks are voiced, or voiced-and-nasalised, they use two airstream mecha-
nisms simultaneously to produce sound, because the egressive pulmonic
airstream mechanism is being employed at the same time as the ingres-
sive velaric airstream mechanism.
0521808820c07_p103_116.qxd 1/27/05 3:02 PM Page 112 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

112 INTRODUCING PHONETIC SCIENCE

Strictly, plain voiceless clicks should also be thought of as being always

accompanied by [k], an unaspirated voiceless velar plosive, resulting from
the closure between the back of the tongue and the velum, which is made
for all clicks. So [k ] is another way of symbolising a voiceless dental click.
Variations in the way that this accompanying [k] is released give rise to
further click modifications. If it is released with aspiration, for example,
the whole click is described as being voiceless and aspirated.

Acoustic and auditory characteristics of clicks

Clicks are very different from all other sounds regularly heard in speech.
The noise as a click is released is generally very intense. When we hear an
utterance containing a number of clicks, it may seem as if a series of
abrupt sounds from a percussion instrument are mingled with a much
more familiar sequence of vowels and consonants.
Figure 7.6 shows spectrograms for a dental click and an alveolar lateral
click in intervocalic position.

FIGURE 7.6
Spectrograms of a voiceless dental click (left) and voiceless alveolar lateral click (right).

Clicks in languages
Languages that use clicks as regular speech sounds are found only in the
southern part of Africa. They are all of the Khoisan languages (this group
includes the various Bushman languages) and also certain neighbouring
0521808820c07_p103_116.qxd 1/27/05 3:02 PM Page 113 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Airstream mechanisms 113

Southern Bantu languages (e.g. Zulu and Xhosa). Among languages that
Although the lan-
have clicks, it is common to find not only several different places of artic-
guages that use
ulation, but also at each place several different series of clicks with dif-
clicks are all in the
ferent modifications. southern part of
Examples of voiceless, voiced and (voiced) nasalised clicks in Zulu are Africa, there are
given below. The transcriptions have been simplified to remove some two distinct and
details that are not relevant at this stage. unrelated language
families involved:
dental postalveolar alveolar
the Khoisan lan-
lateral
guages and the
voiceless [ a a] ‘be [!a!a] ‘undo’ [oa] ‘relate’
southern Bantu lan-
clear’
guages. Speakers of
voiced [ a a] ‘marry’ [ !a !a] ‘scatter’ [ u uma] ‘be
the two types of
nervous’
languages are cul-
nasalised [ŋ eŋ eza] ‘jingle’ [ŋ!eŋ!eza] ‘ring’ [ŋeŋeza] ‘urge horse’
turally distinct and
have different styles
Other uses of clicks of life. As other
Although clicks as speech sounds are restricted to relatively few lan- Bantu languages
guages, some of the click sounds themselves are in fact familiar to most from elsewhere in
Africa do not have
people, regardless of what language they speak. A bilabial click [◎] is a kiss-
clicks, it is reason-
ing noise, a dental click [ ] is the ‘tut-tut’ of disapproval, the postalveolar
able to suppose
or retroflex [!] is made in imitation of horses’ hooves, and the alveolar lat- that the clicks in
eral [] is the ‘gee-up’ sound used to encourage horses. Zulu and Xhosa
must somehow
result from the
influence of
Khoisan. There are
other instances of
phonetic and
phonological prop-
erties being shared
by languages that
are spoken in the
same area though
not closely related.
For example, the
languages of the
Indian sub-conti-
nent belong to two
distinct stocks
(Indo-European and
Dravidian) but the
retroflex place of
articulation for
consonants is a fea-
ture of Indian lan-
guages from both
families.
0521808820c07_p103_116.qxd 1/27/05 3:02 PM Page 114 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

114 INTRODUCING PHONETIC SCIENCE

Chapter summary
In this chapter we have looked at the generation of the air pressures
and airflows needed for the production of speech sounds, and then at
those sound-types made with an airstream that is not supplied by the
lungs. These are called non-pulmonic sounds. Sound generation in
speech relies either on the compression of air, giving an outward
(egressive) flow or the rarefaction of air, resulting in inward (ingres-
sive) flow. Three types of non-pulmonic consonants are used in lan-
guages. Two of these types use a glottalic airstream mechanism.
Ejectives are produced with an egressive airflow generated by larynx
raising. Implosives are made with an ingressive flow generated by low-
ering the larynx. The third type, clicks, use the velaric ingressive
airstream mechanism, produced by rarefying a small volume of air
enclosed in the mouth. We have seen that there are sounds that make
simultaneous use of pulmonic and non-pulmonic airstreams and have
considered various modifications that can be applied to clicks.

Exercises
Exercise 7.1
Look at the following descriptions of the production of non-pulmonic
consonants. In each case, work out exactly what sound is being described
and give the appropriate symbol.
(a) The back of the tongue rises and makes firm contact with the velum
which is in a raised position sealing off the nasal cavity. The tip and blade
rise and make contact with the alveolar ridge. The sides of the tongue make
a complete seal with the sides of the palate. The vocal folds are open and
not vibrating. The centre of the tongue is drawn down, reducing the air pres-
sure in the space enclosed between the tongue and palate. The seal at one
side of the palate is broken by lowering the tongue rim. Air rushes in and a
brief burst of noise is generated.
(b) The back of the tongue rises and makes firm contact with the velum
which is in a raised position sealing off the nasal cavity. Air is being expelled
by the lungs making the vocal folds vibrate. The whole larynx is lowered,
enlarging the space between the velar closure and the vocal folds. The air
pressure in this space decreases. The back of the tongue moves away from
the velum and a burst of noise is heard as air flows in. The vocal folds
continue to vibrate.

Exercise 7.2
Look at the following description of the production of [p’ ]. Two important
points, essential for the successful generation of the airstream, have been left
out of the description. Identify what these are.
The lips come together and form a complete closure . The glottis is closed.
Air pressure inside the vocal tract rises. The lips part and a burst of noise is
produced as air rushes out.
0521808820c07_p103_116.qxd 1/27/05 3:02 PM Page 115 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Airstream mechanisms 115

Exercise 7.3
For each of the following sounds draw a diagram of the vocal tract and write
a short paragraph describing the production of the sound between two vow-
els, paying particular attention to the generation of the airstream:
(a) voiceless alveolar ejective stop (unaffricated)
(b) voiceless bilabial click
(c) voiced nasalised dental click

Exercise 7.4
The sound system of Zulu uses four airstream mechanisms: pulmonic egres-
sive, glottalic egressive, glottalic ingressive and velaric ingressive. For each of
the Zulu consonants listed below, give an appropriate phonetic symbol and
say which of the airstream mechanisms are used to produce the sound.
Remember that certain sounds make simultaneous use of more than one
airstream mechanism.
(a) voiceless alveolar lateral fricative
(b) voiced bilabial implosive
(c) voiceless dental click
(d) voiceless alveolar ejective affricate
(e) voiced nasalised postalveolar click
(f) voiceless velar ejective affricate
(g) voiceless aspirated velar plosive
(h) voiced alveolar lateral click
(i) voiceless velar ejective stop (unaffricated)
( j) voiced bilabial nasal

Exercise 7.5
The speech of Mr Y, a speaker of English, shows the following abnormal
features:

• wherever other speakers have voiceless obstruents Mr Y replaces these

with their ejective equivalents
• wherever other speakers have a voiced plosive or a nasal stop Mr Y uses
an implosive at the corresponding place of articulation.

Alter the transcriptions of the following sentences to show how they would
be pronounced by Mr Y.
(a) Watch what you’re doing [wɒtʃ wɒt jɔ duŋ]
(b) Can I get a glass of milk? [kn a et ə las əv mlk]
(c) Is it five past nine? [z t fav pɑst nan]

Further reading
There is a lot of information about non-pulmonic sounds in Peter Ladefoged and Ian Maddieson, Sounds of
the world’s languages, Oxford: Blackwell, 1996. See section 3.2 on airstream mechanisms, and chapter 8 on
clicks. For advice on understanding non-pulmonic sounds, and how to make them yourself, see J. C. Catford,
A practical introduction to phonetics, Oxford: Oxford Uuniversity Press, 2nd edn, 2001.
0521808820c07_p103_116.qxd 1/27/05 3:02 PM Page 116 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:
0521808820c08_p117_134.qxd 1/27/05 3:02 PM Page 117 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

CHAPTER

8 Speech sounds
and speech
movements

CHAPTER OUTLINE
In this chapter you will learn about: sounds involving more
KEY TERMS than one articulation; how sounds are influenced by
Advanced neighbouring sounds; overlapping articulations; variation in
Coarticulation segment duration; how to draw diagrams of estimated
Labialisation articulator movements.

Palatalisation
Palatography
Parametric
diagrams
Pharyngealisation
Pre-fortis
clipping
Pre-pausal
lengthening
Primary and
secondary
articulation
Retracted
velarisation
0521808820c08_p117_134.qxd 1/27/05 3:02 PM Page 118 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

118 INTRODUCING PHONETIC SCIENCE

Introduction
In previous chapters we have gradually built up the descriptive frame-
work that enables us to understand most of the ways in which speech
sounds can differ from each other. For example, in the case of consonant
sounds, we have looked at the airstream mechanism, at the various func-
tions of the larynx, and at places and manners of articulation. There is
still more to be done, though, because so far we have been assuming for
the sake of simplicity that segments occur one-at-a-time in a sequence,
and that within each segment there is one articulation at a time. Both of
these assumptions need to be modified, because by themselves they lead
to a very static view of what speech is like. The articulators and the places
of articulation we have identified, for example, are nothing like as tidy
and distinct in reality as our labels suggest. We have only to look inside
our own mouths with a mirror, or at an X-ray picture of the vocal tract,
to be reminded that the tongue is not divided neatly into the sections we
have labelled blade, front and back, or that it is very hard to say where
the alveolar ridge ends and the hard palate begins. In the same way, the
assumptions that the vocal tract does one thing at a time, and that the
resulting sound is effectively determined by that one activity, is immedi-
ately called into question if we observe our vocal tract movements in pro-
nouncing a word or a phrase. What we observe is complex, co-ordinated
movement. From an acoustic point of view, of course, it will always be the
whole configuration of the vocal tract that shapes its response, not just
the activities of the articulators we’re focusing on when we give simple
phonetic labels.
We will turn to speech movements in the later part of this chapter, but
we begin by looking at sounds that, even from the static segmental view-
point, have two simultaneous articulations.

Primary and secondary articulation

When we were looking at the basic properties of consonants in terms of
their place of articulation in Chapter 3, we briefly mentioned the concept
of secondary articulation. Apart from a constriction at the primary place
of articulation a consonant may have another less extreme constriction at
a different place. In Chapter 3 we gave the example of [tw ] with lip round-
ing added to an alveolar plosive. This is called labialisation and can be
applied to consonants at just about any place of articulation. Figure 8.1
shows the position of the articulators during the hold phase of [tw ]. The
diacritic [w] shows that the sound in question is labialised. Though the dia-
critic is placed after the relevant symbol, it does not mean that the labi-
alisation follows the sound to which it is added. In this example, the hold
phase of the alveolar plosive can be affected just as much as the release.
Figure 8.1 actually shows the position that would be expected during the
0521808820c08_p117_134.qxd 1/27/05 3:02 PM Page 119 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Speech sounds and speech movements 119

FIGURE 8.1
Position of the articulators
in a labialised alveolar
stop. From the side,
labialisation appears as lip
protrusion.

hold phase. Notice that lip rounding shows up from the side as lip
protrusion.
Labialisation is not the only possible type of secondary articulation. The
additional constriction can alternatively be at the hard palate, at the soft
palate or in the pharynx.
A secondary constriction between the front of the tongue and the hard
palate is known as palatalisation, symbolised by means of the diacritic [ j].
Figure 8.2(a) shows the position of the articulators in the hold phase of a
palatalised alveolar plosive [tj]. A raising of the back of the tongue towards
the velum gives velarisation, illustrated again on an alveolar plosive in
Figure 8.2(b). The general diacritic for velarisation is [ γ ], but an alternative
symbolisation, a wavy dash through the middle of a symbol, is commonly
used for certain velarised sounds, such as [] which indicates a voiced
velarised alveolar lateral approximant.

(a) (b)

FIGURE 8.2
Position of the articulators
in (a) palatalised, and (b)
velarised alveolar stop.
0521808820c08_p117_134.qxd 1/27/05 3:02 PM Page 120 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

120 INTRODUCING PHONETIC SCIENCE

Pharyngealisation involves a narrowing of the pharynx. One way of

achieving this is to pull the root of the tongue backwards into the phar-
ynx. It is symbolised with the diacritic [ ʕ]. Figure 8.3 shows a pharyn-
gealised alveolar plosive.

FIGURE 8.3
Estimated position of
articulators in the produc-
tion of a pharyngealised
alveolar stop.

Acoustic effects of secondary articulations

Sounds with secondary articulations are audibly different from plain
sounds that lack them. The effects of secondary articulations are quite
complex, because they alter the overall shape of the vocal tract, not only
during the relatively steady portion at the middle of a consonant articu-
lation, but also in the transitional movements into and away from the
combined articulation. For instance, rounding the lips, as in labialisa-
tion, has the effect of lowering all the resonant frequencies of the vocal
tract. This has an audible effect on the burst of a plosive (try producing
[t] and [k] releases with and without rounded lips). Palatalisation, velari-
sation and pharyngealisation involve movements of the whole body of
the tongue, very like those used in vowels. These adjustments take time
to set up and remove, and as a result vowels that are next to consonants
with secondary articulations are often affected in recognisable ways as
the tongue manoeuvres into position. Figure 8.4 shows spectrograms of
the English words bit and built. Comparing the two vowels, which are
both supposedly examples of [], you can see big differences. This is
because the vowel in built is affected by the tongue getting ready for the
velarisation of the lateral consonant. The difference is audible, too, espe-
cially if you cut off the remainder of each word and listen only to the two
instances of [b-].

The linguistic use of secondary articulation

There are many languages that use the difference between consonants
with secondary articulation and consonants without secondary articula-
tion to distinguish one word from another. Irish is a good example of such
0521808820c08_p117_134.qxd 1/27/05 3:02 PM Page 121 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Speech sounds and speech movements 121

(a)

(b)
FIGURE 8.4
Spectrograms of bit and built, showing the effect of the velarised lateral consonant on
F2 of the preceding vowel.
0521808820c08_p117_134.qxd 1/27/05 3:02 PM Page 122 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

122 INTRODUCING PHONETIC SCIENCE

a language and there are examples of pairs of Irish words distinguished in

this way in the table below.

Non-velarised Velarised
[le] ‘range’ [lγe] ‘of the day’
[ni] ‘washing’ [nγi] ‘nine’

Non-palatalised Palatalised
[bɑd] ‘boat’ [bɑdj] ‘boats’
[sul] ‘of eyes’ [sulj] ‘eye’

Russian is a language that makes extensive use of the difference between

velarised consonants and palatalised consonants. [uəlγ] ‘corner’ and
[uəlj] ‘coal’ is a pair of Russian words exemplifying this. Some languages
use labialised and non-labialised consonants in the same way. In Kwakiutl,
a language spoken in Canada, [q’asa] means ‘sea otter’, but [qw’asa] means
‘crying’. Arabic is well known for having what are usually called ‘emphatic’
versions of certain consonants in contrast with plain versions. The
emphatic sounds are pharyngealised (or velarised, or both). For example,
in the Arabic spoken in Qatar [sad] means ‘to prevail’, whereas [s ad] is the
name of a letter of the alphabet.

Velarisation in English
In many accents of English (such as standard southern British) there is a
pair of sounds distinguished by presence or absence of a secondary artic-
ulation. Both are voiced alveolar lateral approximants, one velarised and
one not. The non-velarised lateral, informally called ‘clear l’, occurs when
the next sound is a vowel or [ j]. The velarised lateral, informally called
‘dark l’ and usually symbolised [], occurs elsewhere. So for many accents
words like lead, low, follow, lure have a non-velarised [l], whereas words like
bell and belt have the velarised version. However, the status of these
sounds in English is different from the examples given above for Irish. In
English, [l] and [] are tied to mutually exclusive environments and the
difference between them can never be used to distinguish one word from
another. We will look at this difference in status in more detail in
Chapter 9.

Coarticulation
We now turn to look at some aspects of speech movements. Look at the
spectrogram and waveform of the English sentence We were away a year in
Figure 8.5. Notice that there are no gaps in either the spectrogram or the
waveform. The vowels and consonants that we hear as separate units are
not in fact separate at all. If we were to take a moving X-ray film of a vocal
tract producing this sentence we would see no pauses between segments.
The vocal tract would be in continual motion.
During the production of speech any particular articulator is required
to go through a rapid sequence of varied gestures. The position that an
0521808820c08_p117_134.qxd 1/27/05 3:02 PM Page 123 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Speech sounds and speech movements 123

FIGURE 8.5
Spectrogram and waveform of We were away a year.

articulator achieves for a sound is likely to be influenced by its position

during the previous sound, and the position it will be required to move
towards for the following sound. Another way of thinking of this is that
properties of one sound tend to spread onto neighbouring sounds. This
phenomenon is known as coarticulation. The following sections look at
some of the different forms coarticulation may take.

Fine adjustment of place of articulation

In the English phrase keep calm, for example, for the first [k], as the body
of the tongue will have to take up a high front position for the following
vowel, the point of contact of the tongue with the roof of the mouth is
likely to be quite far forward on the velum, whereas the reverse is
observed for the second [k], where the following back vowel has the effect
of retracting the point of contact. Thus the frontness of the vowel in keep
spreads onto the first [k], and the backness of the vowel of calm onto the
second [k]. If you repeat the phrase a number of times you should be able
to feel the difference in the two [k] sounds.
A very similar adjustment of point of contact may affect alveolar
sounds. As an example from English, consider the two plosive sounds in
daydream. You should be able to feel that the point of contact for the
0521808820c08_p117_134.qxd 1/27/05 3:02 PM Page 124 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

124 INTRODUCING PHONETIC SCIENCE

second is somewhat further back on the alveolar ridge than for the first.
This is caused by the post-alveolar position required for the immediately
following consonant [ɹ], which like the [d] is formed with the tongue tip
and blade.
A sound that is shifted forwards in place of articulation may be
described as advanced and this is symbolised by means of a subscript plus
sign, so the word keep can be represented as [k
hip]. A backward shift may
be termed retraction and retracted sounds are very logically shown with
a subscript minus sign, so that calm may be transcribed [khɑm]. The sub-
script minus sign applies to alveolar sounds too, so that [d ] indicates a plo-
sive retracted from the alveolar place so that it is post-alveolar. The sub-
script plus sign, when applied to an alveolar symbol, essentially indicates
a dental articulation and we have already met a diacritic for dental, so
that [d+ ] is equivalent to [d ]. A voiced dental plosive may be heard in the
phrase good thing for example.

Nasalisation caused by coarticulation

We have already seen that vowels can be produced with a lowered soft
palate and that the resulting nasalised vowels function as separate sounds
in certain languages. In other languages, such as English, nasalised vow-
els result from coarticulation. The lowered velum position required for a
nasal consonant may spread onto an adjacent vowel. So the vowel in words
like ban and bang is likely to be somewhat nasalised, [b ˜n, b ˜ŋ], whereas
we have an oral vowel in words like bad, bat, back, bag and so on. The exam-
ples we have given involve anticipation of the nasality of the following
consonant, but it is also possible for nasality to persist into a vowel that
follows a nasal consonant: compare the vowels in mad and bad. If a vowel
is both preceded and followed by nasals, it will be affected from both
directions: compare the vowels of moon and boot.
It is not only vowels that can be affected by the spread of nasality, but
also sonorant consonants such as English [w ɹ l j]. You may find examples
in your pronunciations of words like film, news, Henry, runway.

Diagrams of velum action

A good way of understanding the timing of velum movements relative
to other speech events is by means of diagrams of the type shown in
Figure 8.6. Diagrams that show the positions of different articulators as
a function of time are called parametric diagrams. A parameter can be
thought of in this context as one of the important time-varying proper-
ties that describe relevant aspects of speech production.

b i h Ωe
m b aI n aI n
FIGURE 8.6
Parametric diagram of
velum action in Be home
by nine.
0521808820c08_p117_134.qxd 1/27/05 3:02 PM Page 125 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Speech sounds and speech movements 125

The velum is in the down position before the speech starts, to allow the
speaker to breathe quietly in and out through the nose. The first segment
of the utterance is an obstruent, and so the velum rises to prevent the
escape of air through the nose. All obstruent consonants require the velum
to take up this closed position. The first sound that requires the velum to
be in the down position is the final consonant in the word home, but notice
that the movement towards this position takes place during the preceding
diphthong, which is somewhat nasalised as a result. The velum must be
closed for the hold phase of the following plosive, and so the closing move-
ment takes place during the last portion of the [m]. The diphthong of by is
similarly nasalised, this time by the first consonant of nine, and the diph-
thong of nine itself is affected by both preceding and following nasals. At
the end of the utterance the velum remains down as the speaker returns
to quiet breathing. Notice that there are no silent gaps between words, and
that coarticulation of nasality occurs both within words and across word
boundaries.

Drawing velum diagrams

1 transcribe the word or phrase phonetically
2 divide into segments, with one column for each
3 find any obstruent consonants, and indicate that the velum is up
for these
4 find any nasal consonants, and indicate that the velum is down for
them
5 fill in remaining segments, showing coarticulation where it occurs
6 remember to start and finish with the breathing position (which
does not have the effect of nasalising adjacent segments).

Overlapping articulations
The examples of coarticulation we have looked at so far involve a single
moving articulator, but it is also possible for the movements of quite sep-
arate articulators to overlap in time, giving new combinations of proper-
ties. For example in the word bread, it is very likely that during the pro-
duction of the first consonant, the tongue tip will take up the position
required for the second consonant. This is possible because the bilabial
plosive has no particular specification for what the tongue tip must do.
Because of this coarticulation, the [b] in this context is quite different
from the one in bed for example.

Coarticulation of plosives
In Chapter 4 the production of plosive consonants was described in terms
of approach, hold and release phases. When plosives occur in sequence,
these phases of production can be affected by coarticulation in various
ways. Consider the English word act, where there is a sequence of a velar
and an alveolar plosive. It is very likely that the timing of the production
of these plosives will be as in Figure 8.7.
0521808820c08_p117_134.qxd 1/27/05 3:02 PM Page 126 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

126 INTRODUCING PHONETIC SCIENCE

FIGURE 8.7
Overlapping [kt] phases.

Notice that the release phase of the first plosive takes place during the
hold phase of the second plosive. This means that although the back of
the tongue breaks its contact with the velum, no audible plosion will be
heard at this stage, because the air is stopped by the alveolar closure. A
transcription of act in the pronunciation represented in Figure 8.7 is
[ k t] where the diacritic indicates inaudible release. In general in
English the release of the first plosive in a sequence is inaudible. The
sequence may be within a word, or may result from putting two words
together, as in black tie. It is worth noticing that languages may handle
this situation in different ways. For instance both French and German
regularly have audible release for the first of two plosives in a sequence.
Examples like this show us that languages can differ in their coarticula-
Find a native tory patterns.
speaker of French or When two successive plosives are at the same place of articulation, coar-
German and explore ticulation takes a different form. In a word like background, for example,
the differences in the velar closure is formed for the [k] and retained for the []. The first
pronunciation of plosive lacks any release phase, and the second lacks an approach phase,
plosive sequences, as shown in Figure 8.8. The same diacritic is used for this, so that back-
using French acte ground would be transcribed [b kraυnd].
or German Akt.

FIGURE 8.8
Phases of [kg] sequence.

Coarticulation and secondary articulation

One property that commonly spreads onto adjacent sounds is lip
rounding. In the word sweet, for example, it is very likely that the lip
rounding required for the [w] will be anticipated during the [s]. This has
an obvious auditory effect, as you can hear by comparing it with the
consonant at the beginning of the word seat. The [s] of sweet has the sec-
ondary articulation of labialisation, symbolised [sw]. In this instance
the secondary articulation results from coarticulation. You should be
careful not to confuse the two concepts. A sound may have an inherent
secondary articulation that cannot be explained as resulting from coar-
ticulation. For instance, the English consonant [ʃ] is always somewhat
rounded for many speakers, regardless of the context in which it
occurs.
0521808820c08_p117_134.qxd 1/27/05 3:02 PM Page 127 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Speech sounds and speech movements 127

Duration variation
We have pointed out previously that it is not always easy to identify the
beginnings and ends of segments in the acoustic signal. Nevertheless, use-
ful segmentations can often be made and the duration of speech sounds
can be estimated. Figure 8.9 shows a spectrogram of the English word
sounds. Plausible segmentation points are indicated, enabling us to deter-
mine that the [s] has a duration of approximately 130 milliseconds, the
diphthong [aυ] a duration of about 250 milliseconds, the nasal [n] about 125
milliseconds, the plosive [d] about 80 milliseconds, and the final fricative
[z] around 165 milliseconds.

FIGURE 8.9
Spectrogram of sounds with segmentation.

Speech sounds measured in this way generally have durations some-

where in a range of about 30 milliseconds to 200 or 300 milliseconds, but
the duration of a particular sort of segment is not fixed. It will vary
depending on the context in which the segment appears. For instance,
the last sound or two at the end of an utterance are generally prolonged,
an effect known as pre-pausal lengthening. As another example, a
syllable-initial consonant is generally longer when alone but consider-
ably shorter when preceding another consonant in a cluster. This can be
seen in Figure 8.10, where we have spectrograms of the two words sigh
and sky.
In Chapter 6 we pointed out that the duration of a vowel may be influ-
enced by whether the following consonant is voiced or voiceless. This effect
is very prominent in some accents of English and is known as pre-fortis
clipping.
0521808820c08_p117_134.qxd 1/27/05 3:02 PM Page 128 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

128 INTRODUCING PHONETIC SCIENCE

(a)

(b)

FIGURE 8.10
Spectrograms of (a) sigh and (b) sky.
0521808820c08_p117_134.qxd 1/27/05 3:02 PM Page 129 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Speech sounds and speech movements 129

Other factors that influence segment duration include overall speech

rate, and the degree of stress placed on syllables.

Acoustic and physiological data

There are a number of different ways of investigating coarticulation and
showing how the vocal tract changes shape over time. Spectrograms show
the changing acoustic patterns, or direct measurements can be made of
articulator positions.

Spectrograms
A simple illustration of the use of spectrograms to reveal coarticulatory
effects is shown in Figure 8.11. Spectrogram (a) is of the word key and (b) is
of the word car. Following the release of the two plosives the formants can
be seen moving towards the frequencies characteristic of the two different
vowels, but it is clear that the formants do not start at the same frequen-
cies. This difference is a reflection of the coarticulatory variation in pre-
cise point of contact between the articulators.

Electropalatography
The technique of electropalatography is explained in Chapter 3 and is
ideal for investigating some aspects of coarticulation. As an illustration of
this look at Figure 8.12. Here we have electropalatograms showing the
hold phase of [k], produced by a speaker of Catalan, in three different envi-
ronments. Notice that the closure extends further forward in the oral cav-
ity for the [k] produced in the environment of a high front vowel (the left-
most picture).

X-ray studies
Although X-ray studies are now very rarely carried out because of the
danger of irradiation, some classic studies show coarticulation very
clearly. Figure 8.13 consists of tracings from X-ray pictures of the vocal
tract. The lefthand column shows the positions for steady vowels [y ɑ u]
and the righthand column shows the hold phase of [] pronounced
between vowels of each type. One can clearly see the relationship
between the position of the tongue for the vowel and the location of the
closure for the plosive.
0521808820c08_p117_134.qxd 1/27/05 3:02 PM Page 130 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

130 INTRODUCING PHONETIC SCIENCE

(a)

(b)

FIGURE 8.11
Spectrograms of (a) key and (b) car.
0521808820c08_p117_134.qxd 1/27/05 3:03 PM Page 131 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Speech sounds and speech movements 131

FIGURE 8.12
Electropalatograms of
intervocalic [k] in three
contexts (after W. J.
Hardcastle and N.
Hewlett, Coarticulation,
Cambridge: Cambridge
University Press, 1999).

y yy

a aa

FIGURE 8.13
X-ray tracings showing
coarticulatory effects of
surrounding vowels on a
velar stop (after S.
Öhman, ‘Numerical
u uu Model of Co-articulation’,
JASA 41 (1967)).
0521808820c08_p117_134.qxd 1/27/05 3:03 PM Page 132 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

132 INTRODUCING PHONETIC SCIENCE

Chapter summary
In this chapter we have looked at the articulation of speech sounds in
more detail. We have seen that sounds may have more than one con-
striction. Common secondary articulations are labialisation, palatalisa-
tion, velarisation and pharyngealisation. The articulation of all sounds
is likely to be influenced by the context in which they appear. Examples
of such influence are: the addition of a secondary articulation to a
sound, the fine adjustment of the place of articulation, the addition of
nasalisation, or variations in duration. The chapter also introduced
diagrams showing the activity of articulators as a good way of
indicating the co-ordination of speech movements.

Exercises
Exercise 8.1
Give a description of the movements of the articulators during the pronuncia-
tion of the word swim [sw ̃m]. Pay particular attention to coarticulation.

Exercise 8.2
Draw parametric diagrams, showing velum activity and vocal fold vibration, for
the following words.
(a) time
(b) pounds
(c) training
(d) screen
(e) important
You may need to refresh your memory of the details of aspiration and
devoicing as explained in Chapter 6.

Exercise 8.3

1 2 3 4

This diagram shows velum movement in a word with four segments labelled
1, 2, 3, 4. Which of the following English words would be produced with
velum movement matching the diagram. Explain your decisions and suggest
other words that would match.
(a) camp (b) cold (c) mission (d) elms (e) film

Exercise 8.4
In the following passage, find all the places where overlapping of plosives
may occur. Remember that the two plosives do not have to be in the same
word.
0521808820c08_p117_134.qxd 1/27/05 3:03 PM Page 133 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Speech sounds and speech movements 133

The chap grinned. He picked himself up off the floor and rubbed his
knee. Up till that moment the atmosphere had been quite relaxed.
Everyone started to look tense. Maybe he wouldn’t be able to do it.
Exercise 8.5
Look at the spectrograms of sigh and sky in Figure 8.10 and estimate in
milliseconds the durations of the two examples of [s].

Further reading
For much more detail on the descriptive framework for segments, see J. Laver, Principles of phonetics,
Cambridge: Cambridge University Press, 1994. Chapter 11 is concerned with multiple articulations of all
kinds, including secondary articulation. For an up-to-date survey of coarticulation, see W. J. Hardcastle and
N. Hewlett, Coarticulation: Theory, data and techniques, Cambridge: Cambridge University Press, 1999.
Part IV of the same book is devoted to ways of measuring and imaging speech.
0521808820c08_p117_134.qxd 1/27/05 3:03 PM Page 134 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 135 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

CHAPTER

9 Basic
phonological
concepts

CHAPTER OUTLINE
In this chapter you will learn: the way that sound differ-
KEY TERMS ences are used to distinguish words; how similar, non-
Allophone contrastive sounds can be grouped into phonemes; how
Alternation sounds form patterns in a language; how to look for pat-
Assimilation terns in samples of phonetic data; about the grouping of
Coda sounds into syllables; differences between languages in per-
Contrast mitted sound groupings; about widespread phonological
processes in languages; about phonological features and
Distribution
phonological rules.
Elision
Environment
Feature
Lenition
Minimal pair
Onset
Phoneme
Phonological
process
Rhyme
Rule
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 136 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

136 INTRODUCING PHONETIC SCIENCE

Contrast
Imagine a language where all word-initial consonants are voiceless and all
consonants elsewhere in a word are voiced. This would mean that in this
language a word like [pida
] is possible, but one like [bitak] is impossible.
The voicing of a consonant in this language is completely determined by
some aspect of the environment of the consonant, in this case the posi-
tion within a word. It is impossible for this language to use the phonetic
distinction between voiced consonants and voiceless ones to distinguish
one word from another, even though the language has voiced and voice-
less pairs of consonants. We can say that in this language there is no voic-
ing contrast or, equivalently, that voicing of consonants is not contrastive
in this language.
In English and in many other languages, voicing is contrastive for con-
sonants. We can demonstrate this with a pair of words like fussy [fsi] and
fuzzy [fzi] in English or a pair in French like choux [ʃu] ‘cabbage’ and joue
[u] ‘play’. In the imaginary language we have just considered such pairs
of words could not occur.

Most languages in the world make some use of the voicing contrast,
but it is not universal. For example in almost all indigenous lan-
guages of Australia, voiced and voiceless plosive consonants are used
non-contrastively. The language Dyirbal is very similar to the imagi-
nary example used here. In Dyirbal, plosives are generally voiced
when they occur between vowels, and voiceless when they occur at
the beginning of a word, as in [tiban] meaning ‘stone’. But Dyirbal
additionally allows a free variation between voiced and voiceless in
initial position, and will also accept the same variation within words
(though native speakers don’t generally do it). So ‘stone’ can be any of
[diban] [tiban] [tipan] [dipan].

An example of a phonetic distinction that is not contrastive in English is

velarisation. Many accents of English have a velarised lateral approximant
[], which appears before a consonant or before a pause, as in milk or cool,
alongside an ordinary non-velarised [ l ] which appears when there is a vowel
or [ j] following as in look or familiar. It is easy to see that if the velarised
sound occurs automatically according to differences in the surrounding
sounds, then the presence or absence of velarisation can never be found
on its own as the only factor that distinguishes one word from another.

Minimal pair
In order to prove conclusively that a phonetic distinction is contrastive in
a particular language it is necessary to find a pair of words in the language
that differ in only one segment. The differing segments must be in the
same position in the two words, and be different in only their values for
the phonetic distinction in question. Finally, the two must mean different
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 137 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Basic phonological concepts 137

Table 9.1 Minimal pairs in various languages

LANGUAGE EXAMPLE OF MINIMAL PAIR

Burmese [ma] healthy [m

 a] order

Ewe [efa] he was cold [eφa] he polished

German [laidn] suffer [laitn] to lead

Italian [fato] fate [fato] done

Korean [pul] fire [phul] grass

Spanish [pero] dog [peɾo] but

Welsh [
win] wine [
win] white

things – that is, they must not simply be alternative pronunciations of the
same word. Such a pair of words is called a minimal pair. The words fussy
and fuzzy constitute a minimal pair in English, as do choux and joue in
French. The presence or absence of a segment can also make a minimal
pair. For example tack and tax form a minimal pair in English.
Languages make contrastive use of a wide variety of phonetic differ-
ences, and a small number of examples appear in Table 9.1. The sound
differences that are contrastive in a language seem obvious and clear to
speakers of that language, but may seem strange and difficult to a person
who does not speak the language.

Environment and distribution

The environment of a segment is simply its position relative to other
sounds. An environment can be symbolised A ¬B. The underscore (¬) rep-
resents the position of the segment in question, while A means the seg-
ment(s) occurring before it and B means the segment(s) that follows it. To
give an example from English: one of the environments for the segment
[t] in English is [s] ¬[ j]. This tells us that [ t] can be found in between [s] and
[j] in English words, such as student and stew. An example of an environ-
ment where [t] cannot be found in English is #¬[
]. The symbol # means
‘word boundary’, so the environment statement means: at the beginning
of a word before a [
]. The environment V¬# means: at the end of a word
after a vowel. Of course, [t] in English can be found in this environment,
but [w], for example, cannot.
If we were to draw up a long list of all the environments in which [t] can
be found in English then we would in a sense have completely specified
the distribution of [t] in English, but the list would be cumbersome and
unrevealing. By searching for patterns in the list, we can hope to sum-
marise it with general statements. For instance, in the imaginary lan-
guage we introduced above where voicing is not contrastive, a summary of
the distribution of [t] in that language would be that it occurs at the begin-
ning of a word regardless of what follows. We can symbolise this as #¬.
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 138 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

138 INTRODUCING PHONETIC SCIENCE

Remember that distributions only apply in the language for which they
Don’t forget the dis- are worked out, and any particular type of sound is quite likely to have dif-
tinction between
ferent distributions in different languages. For instance, the word-final
spellings and
environment ¬# is part of the distribution of the sound [d] in English, as
sounds. Plenty of
German words are we can see from words such as red and cold, but not of the sound [d] in
written with a German. No matter how long we search, we will never find an example of
final d but the pro- a German word ending in the sound [d] because in German all word-final
nunciation is obstruents have to be voiceless.
always with [t]. For
instance the noun Phoneme and allophone
which is written We saw above how [l] and [] are not contrastive in English. We can find fur-
Tod ‘death’ is pro- ther sounds in this non-contrastive set. For instance, in words like health,
nounced [tot], where an l-sound is followed by a dental fricative, a common pronunciation
identically with the
involves [], which is a voiced velarised dental (rather than alveolar) lateral
adjective written
approximant. A speaker cannot use the difference between dental place and
tot meaning ‘dead’.
alveolar place for lateral approximants in English to signal a difference in
word identity, so the difference between the two laterals is non-contrastive.
Another member of this set is [ ], which is a voiced nasalised and velarised
alveolar lateral approximant. This is found immediately before [n] and [m]
in words like film and kiln or in phrases like all men, and again cannot con-
trast with any of the other lateral approximants in English.
The consonant sounds we have identified here are similar phonetically
in that they are all voiced lateral approximants. They differ in some
aspects of their production and the differences are perfectly audible if one
listens carefully. However, these phonetic differences are irrelevant for dis-
tinguishing between one word and another in English. We sum this up by
saying that the set of lateral sounds we have just identified belong together
in one of the phonemes of English, a phoneme being one of the basic
phonological units of which all words in the language are composed.
We represent a phoneme with a symbol placed in slant brackets, as
opposed to square brackets. The various sounds that make up a phoneme
are known as its allophones. The sounds [l] [] [] [ ] are allophones of an
English phoneme that we can represent as /l/. Using the brackets in the
same way, we can say for instance that the aspirated bilabial plosive [p ] is
one of the allophones of the /p/ phoneme in English (of course, there are
other allophones of /p/ too, which are found in different environments).
The grouping of sounds into phonemes is different in different lan-
guages, or even in different varieties within one language. The two sounds
[d] and [ð] belong to different phonemes in English, but are allophones of
a single phoneme in Castilian Spanish, where [ð] is an allophone of the /d/
phoneme. As this example shows, allophones of a phoneme are often suf-
ficiently different that entirely different phonetic symbols are needed for
them. To give another example, the two sounds [f] and [v] belong to dif-
ferent phonemes in present-day English, but in Old English, spoken about
1000 years ago, the two sounds were allophones of a single /f/ phoneme.
See Table 9.2 for further examples of phonemes and some of their allo-
phones in various languages. We saw earlier that the sound differences
that are contrastive in a particular language may seem surprising to an
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 139 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Basic phonological concepts 139

Table 9.2 Phonemes and allophones in various languages

language phoneme allophones

Burmese /j/ [j  . . .]

Canadian English /a/ [ae

 əe . . .]
British English /t/ [t t . . .]

/əυ/ [əo  . . .]
 ɒo
/r/ [ɹ 
ɹ . . .]
Castilian Spanish /b/ [b
. . .]

French /r/ [ʁ χ . . .]

General American /t/ [t ɾ . . .]

German /x/ [x ç . . .]

Japanese /t/ [t ts . . .]

Modern Standard Chinese /i/ [i ɯ . . .]

Old English /h/ [h x . . .]

South American Spanish /s/ [s h . . .]

Welsh /m/ [m m . . .]

observer who speaks a different language. In a corresponding way, the

extent of the phonetic differences among allophones of one phoneme can
seem surprisingly large, yet speakers of the language in question accept
the allophones as versions of ‘the same sound’ and may find it hard to
hear the differences even when they are pointed out.

Complementary distribution
When we compare the distributions of two sounds in a language, we
may find that they are partially similar – that is, the two sounds share
some environments. For example [ph] may appear in much the same
environments as [th] in English. However, if we compare the environ-
ments of two sounds that are allophones of the same phoneme in the
language in question we will often find they can never appear in the
same environments – their distributions are completely different from
one another. This is called complementary distribution. Consider the
Canadian English example in Table 9.2. The phoneme /a/ has amongst
its allophones a sound [əe], which occurs only before voiceless conso-
nants, in words such as white or rice. Another allophone is [ae], which
occurs elsewhere, for example in words such as why, wide or rise. These
two sounds are in complementary distribution in this accent of
English. However, the fact that two sounds are in complementary dis-
tribution is no guarantee that they should be classed as allophones of
the same phoneme. Allophones of the same phoneme must also have
some shared phonetic characteristics. To take an extreme example, if
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 140 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

140 INTRODUCING PHONETIC SCIENCE

one were investigating a new language and found that [ ] and [m] hap-
pen to be in complementary distribution, we would not be tempted to
regard the two sounds as being allophones of one and the same
phoneme, because they are just too dissimilar phonetically.

Alternation
Phonology is not only concerned with the establishment of the phonemes
of a language and the distribution of sounds. It is concerned with sound
patterns in general. Sound patterns are of two basic types:

• alternations, where a word turns up in two or more forms, and there

are systematic sound-differences between the forms.
• distributional regularities concerning sounds, such as restrictions on
possible sound sequences within words, syllables etc. We deal with
this topic later in this chapter.

As an example of an alternation, consider the word fan. It is general-

ly pronounced with an alveolar nasal at the end [fn], but in fan club it
can be pronounced with a velar nasal [fŋ], making it indistinguishable
from fang. In fan belt, there is a third alternant [fm]. The only satisfac-
tory way to account for the fact that many words or components of
words have more than one pronunciation is to suppose that some pro-
nunciations are derived by rules from other possible pronunciations, or
that all pronunciations are derived by rules from some kind of abstract
form.

Phonological processes
Phonological processes operate upon natural groupings of sounds, and
give rise to alternations in the forms of words. Phonological processes
belong to a small number of frequently occurring types.

• assimilation, where a sound becomes more like some other sound in

its immediate context. Here are some data showing a simple assimila-
tory process from Lithuanian:
[ samburis] ‘assembly’ [ sampilas] ‘stock’
[ sandora] ‘covenant’ [ santaka] ‘confluence’
[ saŋkaba] ‘connection’
The first syllable of all these words is a prefix that is the equivalent of con-
in English. Notice that the place of articulation of the final nasal in this
prefix is adjusted to agree with the place of articulation of the following
consonant.
These examples show assimilation affecting place of articulation, but it
can affect other properties of sounds equally well. Look at the data from
Sudanese Arabic below:
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 141 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Basic phonological concepts 141

[ kitab] ‘book’
[ kitaf fathi] ‘Fathi’s book’
[ bit] ‘daughter’
[ bis samir] ‘Samir’s daughter’
[ biʃ ʃarif] ‘Sharif ’s daughter’
[ samak] ‘fish’
[ samax xalid] ‘Khalid’s fish’

Notice that the voiced consonant at the end of [ kitab] is made to agree in
voicing with the following sound, and that the final consonant in the first
word of all these phrases agrees in manner of articulation with the first
consonant of the following word.

• deletion (also called elision and truncation) where a sound is lost

from a sequence. A good example of this is found in the speech of
young children acquiring English. Look at the words below:

word adult form typical child form

play [ ple] [ pe]
green [
ɹin] [
in]
twin [ twn] [ tn]

The pattern shown here, in which approximant consonants are deleted

when they are the second member of a word-initial cluster, is a common
one in normal acquisition.

• the opposite process, insertion, puts in a sound where none was pres-
ent before. A very common feature of many accents of English is
called r-linking and involves inserting an r-sound of some sort
between two vowels as in the words and phrases below.

floor [ flɔ] flooring [ flɔɹŋ]

mother [ mðə] mother-in-law [ mðəɹ n lɔ]
vanilla [ vənlə] vanilla ice-cream [ vənləɹ askɹim]

• lenition (also called weakening): consonants can be arranged on

scales of strength:

plosive  fricative  approximant  zero

aspirated  voiceless  voiced

Thus a voiceless aspirated plosive is the strongest type of consonant, and

a voiced approximant is the weakest. The scales can be summed-up by
saying that a consonant is stronger the more it differs from vowels; a
consonant becomes weaker the more it comes to resemble a vowel.
A lenition process affects voiced plosives between vowels in Spanish.
The conventional orthography gives no indication of the changes, but the
alternations are perfectly regular.

orthography pronunciation meaning

donde [ donde] where
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 142 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

142 INTRODUCING PHONETIC SCIENCE

dado [ daðo] dice

todo [ toðo] all
bomba [ bomba] bomb
beber [ be
er] to drink
bodega [ boðeγa] wine cellar
gas [
as] gas
digame [ diγame] hello/can I help you?

In initial position or following a consonant we find voiced plosives, but

between vowels we find the equivalent voiced fricatives. It seems reason-
able to regard these fricatives as resulting from lenition of the plosives,
and reasonable too that the weakening of the closure gesture for the plo-
sive should occur in the position between two vowels.

• dissimilation: this process is the opposite of assimilation, and has

the effect of preventing two similar segments occurring close
together. It is much less common than assimilation, and the best
examples we can give are historical ones showing its operation at
some stage in the past. For example the Latin word for ‘tree’ [arbor]
developed into Spanish [arbol] where the final [l] results from the
avoidance of two r-sounds close together. The same sort of change
can be seen in the development of the French word [peleʁε̃] ‘pilgrim’
from Latin [pere
rinus].
• fortition: this is the opposite of lenition and involves sounds moving
towards the stronger end of the scales mentioned above. An example
of this comes from German, which allows both voiced and voiceless
obstruents within words, but only voiceless obstruents word-finally.
Comparison of alternations shows us that obstruents are getting
devoiced in final position.

Rates [ʁatəs] ‘of advice’ Rat [ʁat] ‘advice’

Rades [ʁadəs] ‘of wheel’ Rad [ʁat] ‘wheel’

Here it is the final sound of the word for ‘wheel’ that has undergone
fortition.

Features
We can think of sounds as being specified by certain properties, called
phonological features. In a sense these features summarise and for-
malise the phonetic description of sounds by providing a checklist of
properties that can be used to classify all of the sounds in a language. In
one influential version of this idea, the features are assumed to be binary –
that is, they can only have the values plus () or minus ().

Some important binary features

It is beyond the scope of this book to discuss a complete list of features,
but a few important ones will give an idea of how binary features are
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 143 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Basic phonological concepts 143

defined and used. Some are fairly obvious, such as [/ voice], [/ nasal]
[/ syllabic] and [/ lateral]. So for instance vowels are [syllabic,
 voice], nasalised vowels will be [syllabic, voice, nasal], while the
specification [syllabic, voice, nasal] identifies specifically the class
of non-nasalised vowels. Any lateral consonant will be [lateral], while
one which is also syllabic (as in one common pronunciation of the English
word bottle) will be [lateral, syllabic].
For certain other features we must attend to the precise definition. The
feature [/ continuant] helps us define the category stop. Sounds count
as stops (that is, they are [continuant]) if they have a complete closure
on the midline of the vocal tract. So plosives are [continuant], but so
are nasals (because they have a complete closure in the mouth). Plosives
are thus oral stops, nasal consonants are nasal stops. Notice that ‘com-
plete closure on the midline’ needn’t mean that airflow is actually
stopped, even in the oral cavity, so that a lateral consonant can also count
as [continuant].
The important distinction between sonorants and obstruents is handled
with the feature [/ sonorant]. All obstruents (that is, plosives, fricatives
and affricates) are [sonorant], while everything else, including vowels, is
[sonorant].
The treatment of place of articulation with binary features doesn’t
make use of familiar place-labels such as alveolar, velar and so on. Much
of the work is done by two important features [/ coronal] and [/
anterior]. A sound is [coronal] if it has an articulation in which the
tongue blade is raised. So dentals, alveolars, post-alveolars, retroflexes and
palatals are all [coronal] while sounds at all other places of articulation,
such as labials and velars, are [coronal]. A sound is [anterior] if it is
produced with a constriction at, or in front of, the alveolar ridge. So labi-
als, labiodentals, dentals and alveolars are all [anterior], while all
places of articulation further back are all [anterior]. Note that post-
alveolars, including the sounds sometimes called palato-alveolar, are in
this [anterior] class.
The examples given so far show features as applied to consonant sounds,
but of course vowel sounds can also be categorised using binary features.
Consonants as a class are [consonantal], while vowels are [consonantal,
syllabic]. The main features employed to deal with vowel quality are
[/ hi
h], [/ low], [/ back] and [/ round]. There is obviously some
interaction between the features [hi
h] and [low], since a vowel could not
be both [hi
h] and [low], though [hi
h, low] is a reasonable way of
indicating a vowel of mid height. To give an example of feature specifica-
tion applied to vowels, a vowel of type [ i] is [hi
h, low, back, round].
The semivowel type of approximant will share the feature specification of
the corresponding vowel, but will be [consonantal, syllabic]. So [ j] has
the same feature specification as [i] except for the value of [syllabic].

Natural classes
We said above that the phonological processes operate on natural group-
ings of sounds, but what exactly are these groupings, and how do they
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 144 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

144 INTRODUCING PHONETIC SCIENCE

arise? Phonological features provide a way to answer this question. Each

particular sound will have its own unique list of feature-specifications,
but two or more sounds that are similar in some respects will have partly
overlapping feature specifications. A natural class is thus a group of
sounds that is easily specified with relatively few features. So, for instance,
the class of all obstruents is captured with the simple specification
[sonorant], the class of all fricatives is [sonorant, continuant] and so
on. On the other hand, if we take a random assortment of sounds, such as
[s] [e] [m] [k] we find that it is not a natural class of sounds, and there is no
simple way of specifying the group using features.

Underlying and surface forms

Pronunciations of words (the ‘surface’ forms) can be thought of as the out-
put of a system of rules and representations, called the phonological com-
ponent, which is internalised by speakers of a language. This starts from
fairly abstract ‘underlying’ forms of words, in which each sound is repre-
sented as a list of feature-values, and then processes them in appropriate
ways. The underlying form of a word can be different from any actual spo-
ken form. Looking at the German example given above one can see that it
makes sense to assume that the underlying form of the word ‘wheel’ is
[ʁad], very much as the spelling Rad suggests, even though the word is
never spoken with a [d] at the end. If we don’t make that assumption, we
can’t explain why there is a [d] rather than a [t] in the inflected form
[ʁadəs] ‘of wheel’. The actual pronunciation [ʁat] results from putting
the underlying form through a rule that forces every word-final obstruent
to be voiceless in actual spoken output.
To take another example, aspiration or non-aspiration of voiceless plo-
sives need not be shown in underlying forms in English, and need not be
stored mentally for every word in which it is found. A process in the
phonological component can check every word as it is produced to see if
it contains any voiceless plosives in the right position (before a vowel with-
in a stressed syllable). When it finds one, the process will add a specifica-
tion of the appropriate aspiration by altering the feature representation.
When the word leaves the phonological component and is spoken aloud
by vocal tract movements, aspiration will appear in the correct place.

Phonological rules
Obviously, unvoicing final obstruents in German, or adding aspiration to
voiceless plosives in English, are just examples of many regularities that the
phonology will have to deal with, since all other predictable characteristics
will also have been removed from underlying forms. The phonological com-
ponent may be visualised as a long list of rules, each one scanning for the
conditions in which it must operate, and changing feature specifications
where appropriate. If a certain rule doesn’t apply, it passes on the input
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 145 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Basic phonological concepts 145

without changing it. The output of each rule becomes the input to the next,
until all the rules have had a chance to operate.
There is a whole system of formalism for showing rules, and we started
to use some of this earlier in this chapter when we were discussing the
notion of environment. Phonological rules have the general form

X S Y/A ¬B

Here X is the input, Y the output, and A ¬B indicates the context. A, B,

X and Y stand for sets of feature specifications, each one identifying a cer-
tain type of sound (or more usually a class of sounds). It means that if the
specification X is encountered between an example of A and an example
of B, then X must be changed into specification Y. Either A or B may be
missing if the preceding or following context is not relevant. If both A and
B are missing the rule applies regardless of context and is said to be ‘con-
text-free’. Sounds can be added or deleted as well as modified, and the
symbol Ø (‘zero’) is used for this, and A or B can be boundaries rather than
sounds; so for instance XSØ/A ¬# would mean that any example of X
should be deleted if it is at the end of a word and preceded by A.
The rule to devoice a final obstruent in German would thus be

[sonorant] S [voice]/¬#

It is often useful to employ this type of notation informally, with phonetic

symbols or verbal descriptions instead of rigorous feature-specifications,
and there are examples of this in the exercises to this chapter. The symbols
C and V are used in rules with the respective meanings [consonantal,
syllabic] and [consonantal, syllabic]. So

C
[nasal]

means any consonant except nasals, while

V
[hi
h]

means any non-high vowel.

The syllable
In Chapter 1, we attempted a rough definition of a phonetic syllable; there
are plenty of clear examples, but the concept is hard to pin down exactly.
The syllable can also be thought of as a phonological unit – that is, as a
grouping of sounds in accordance with certain rules, which differ from
language to language. The essential component in any syllable is a vowel,
which provides the nucleus of the syllable. This may be preceded or fol-
lowed by one or more consonants. Consonants before the vowel form the
onset of the syllable; consonants after the vowel form the coda. There are
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 146 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

146 INTRODUCING PHONETIC SCIENCE

syllable reasons for believing that the nucleus and coda together form another
unit within the syllable. In English for example, two syllables are said to
rhyme with each other (in verse) if they have the same nucleus and the
rhyme
same coda (or absence of coda). So day rhymes with bay, and date rhymes
with bait. But languages seem not to make use of a unit consisting of the
onset nucleus coda onset and nucleus together. The nucleus-plus-coda unit of the syllable is
FIGURE 9.1 called the rhyme. The structure of the syllable can be shown as in Figure 9.1.
The structure of the These terms can be very useful in describing the distributions of
syllable. sounds. In English the velar nasal sound may appear only in the coda of
syllables, never in the onset. It need not be the very last sound in a sylla-
ble, though it may be, as we see from words such as bang, bank, banks.
Without the term coda, we would have to say that the velar nasal occurs
only in the part of the syllable that follows the vowel. Similarly, the
English consonants /w h j/ occur only in onsets. If the onset or the coda is
missing from a particular syllable, we say that the position is ‘empty’. So
for instance the English word eat is an example of a syllable with an empty
onset. (Though if we pronounce the word in isolation we may add a glot-
tal stop at the beginning, supplying an onset that is not present in the
underlying form.) A syllable with no coda is called an open syllable,
whereas a syllable with at least one coda consonant is said to be closed.
These terms, too, are very useful. For instance, certain English vowels
(examples are [, e, ɒ]) have the property that they cannot occur in
stressed open syllables, and this is why we know there just couldn’t be any
English words [b] or [bɒ]. With closed syllables, there’s no problem: bat
[bt] and box [bɒks] are perfectly good English words.

Phonotactics
Languages differ considerably in the types of syllables they use, and speak-
ers of the languages have an intuitive understanding of what are, and
what are not, permitted syllables in their language. To use a famous exam-
ple, speakers of English will generally agree that blick is a ‘possible’
English word, while bnick is not. It certainly isn’t the case that bnick is
‘unpronounceable’ in any general sense. It’s just that [bn-] isn’t a permit-
ted syllable onset in English. In fact, no onsets of plosive consonant fol-
lowed by nasal consonant are allowed in English.
A statement of the permitted combinations of phonemes into struc-
tures (often the syllable) is a statement of the phonotactics of the lan-
guage. So it is quite common to find that different ranges of consonants
can be used in onset and coda positions (a common pattern is for there to
be a wide range of choice in the onset position, but just a few possibilities
for the coda). And if sequences of consonants are permitted in the onset
or coda, their length and composition is likely to be restricted. A sequence
of consonants within the onset, or within the coda, is called a consonant
cluster.
Onsets in English will provide an illustration of phonotactic patterning.
For a single consonant, we find that any English consonant phoneme can
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 147 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Basic phonological concepts 147

occur, with the exception of /ŋ/ the velar nasal, which as explained above
is restricted to coda position. (In plenty of other languages, of course, the
velar nasal is found as an onset – for example, in Vietnamese.) When we
consider two-consonant onset clusters in English, we find that it is far
from being the case that any two-consonant combination will do; in fact,
if you make a list and try out the hundreds of logically possible two-
consonant combinations, you find that most of them are not allowed.
There’s no /bd-, b
-, bm-, mk-/, and so on. Those that do occur turn out to
belong to one of two types. Either the first consonant is /s/ and the second
is one of a group that includes chiefly /p t k m n/, or the second consonant
is one of the approximants /w r l j/ and the first is drawn from a group that
includes plosives, nasals and fricatives. The first type can be represented s
 C, the second as C  approximant. (When searching for examples, we
will sometimes find isolated instances of other onset clusters in foreign
words borrowed in English. For example, English doesn’t really have an
onset cluster /pt-/, though some people pronounce it at the beginning of
the word pterodactyl which is taken from Greek.)
For three-consonant onsets, we find even less choice. The first conso-
nant is always /s/, the second generally one of /p t k/, and the third is an
approximant. This gives combinations such as /str-/ in string, /spl-/ as in
splash, and /skw-/ as in squash. Notice that these strings of three conso-
nants always include two of the possible two-consonant onsets: /str-/
includes /st-/ and /tr-/. So the possible three-consonant onsets seem to be
like two of the possible two-consonant onsets stuck together.
We have searched for onset clusters at the beginnings of English words,
making the assumption that the beginning of a word must also be the
beginning of a syllable. Of course onsets may be within words that have
two or more syllables. For example in construct there is a sequence of con-
sonants /nstr/. The first syllable has /n/ as its coda, and the second syllable
appears to have the onset /str-/. If we look at the ends of words we can find
0, 1, 2, 3 or 4 consonants, as in go, egg, friend, milked, texts. So an English
syllable can have up to three consonants in the onset, and up to four in
the coda.
The situation is usually summed up in terms of a template:

C03 V C04

Other languages will have different templates. Maori is an example of a

language that has very severe constraints on clusters. Initial clusters must
consist of 0 or 1 consonant and final clusters must consist of no conso-
nants at all. So the Maori template is:

C01 V

Another language with a very restrictive template is Modern Standard

Chinese:

C01 V C01
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 148 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

148 INTRODUCING PHONETIC SCIENCE

Certain languages allow very much more complex clusters than English,
both in the length of clusters that are allowed and also in the permitted
combinations of sound types. In Polish, for example, a word may begin
with up to five consonants. And whereas it is the case in English that the
affricates /tʃ ð/ do not occur in onset clusters, and that no consonant ever
clusters with itself, one permitted type of onset in Polish consists of two
occurrences of the same affricate. Try pronouncing the examples of inter-
esting onsets in the table below, taken from Polish and Russian.

Polish [ spstrɔw̃!εm] ‘with (the) trout’

[vz
lɔnt] ‘respect’ (noun)
[ ddst] ‘rainy’
Russian [ zdrastvuj] ‘hello/how do you do?’
[ mtʃatsə] ‘to rush’
[ vzmorjm] ‘sea shore’

Chapter summary
Contrast is the use of phonetic differences to signal the distinction
between words. A pair of distinct words that differ only in one segment
is a minimal pair. Two sounds that cannot mark a contrast between
words may be allophones of the same basic unit (members of the same
phoneme) especially if each has its own specific environments, mean-
ing that the two are in complementary distribution. The aim of
phonology is not only to establish the set of phonemes in a language,
but also the patterning of these sounds, both dynamic and static.
Patterns of changes in the pronunciation of words (alternations) often
result from the operation of phonological processes, the same general
types of process being found widely in languages. Phonological pat-
terning can be described formally by specifying sounds (e.g. with rigor-
ously defined features) and then writing rules that alter sound specifi-
cations. The internal or mental representation of a word (the underly-
ing form) can be different from the observable (surface) form. The
phonological syllable is a structure that specifies how sounds of differ-
ent types can be combined. The rules for combinations (phonotactics)
are different in different languages.

Exercises
Exercise 9.1
Look at the examples of minimal pairs in various languages given in Table
9.1. Try to read the words from the transcriptions given. Give phonetic labels
for the two sounds that are responsible for the contrast, and say in what ways
the sounds are similar and exactly how they are different.

Exercise 9.2
Use your own pronunciation of English to search for minimal pairs among the
words below. Pronounce the first word given at the left, then try each of the
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 149 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Basic phonological concepts 149

others in turn to see if they make a minimal pair with the first word. When
you find an example that does not a make a minimal pair, say why not.
dot dock dog dots debt doll got
can cat cone cane cape man mat
hat flat that had hate at heart
green bean groan gain grease glean grain
shoe show shoot shrew Sue blue who

Exercise 9.3
The following forms are typical of the speech of C, an English child about
three years old. The child-like patterns of pronunciation exemplified here are
all frequent in normal acquisition. You may assume that other words would
be similarly affected, as if the child were applying ‘rules’ to change the adult
target form into her own version.
brush [bws] shop [sɒp]
change [tendz] smack [m mk]
class [klɑs] smudge [m mdz]
crunch [kwnts] sneeze [n
niz]
dish [ds] sorry [sɒwi]
James [demz] Spot [pɒt]
judge [ddz] spring [pwŋ]
road [wəυd] stretch [twets]
school [kul] watch [wɒts]
(a) Use a minimal pair to prove that normal adult English has a phonemic
contrast between /w/ and /r/ (the latter is usually realised as [ɹ]). What
is the situation in this child’s speech, and what will happen to minimal
pairs that depend on this contrast?
(b) Find a minimal pair to show that /s/ and /ʃ/ are in contrast in adult
speech. Again, what is the situation in this child’s speech, and what will
happen to minimal pairs that depend on this contrast?
(c) Find all the words that in their adult versions would contain the affricates
/tʃ d/, and work out the pattern in the child’s treatment of them. Show
that the child seems to be following a rule that is ‘context-sensitive’.
(d) Find all the words in the data that begin with clusters of two or three
consonants in their normal adult versions (don’t include affricates) and
group them according to type. Explain what the child does with the
different types of cluster.
(e) Write a formal rule that describes the child’s treatment of /s/ in onset
clusters that do not contain a nasal consonant.
(f) How is C likely to pronounce the following words? Transcribe your
answers phonetically.
Charles choose fridge juice lunch prince scratch snake stamp touch

Exercise 9.4
Make a list of the symbols for consonant segments that are used in the
child’s forms in Exercise 9.3 above. Then from the list select the consonant,
or group of consonants, that fit the feature specifications given.
Example [nasal] [m n ŋ m ]
 n
(a) [nasal, voiced]
(b) [anterior, coronal, continuant]
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 150 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

150 INTRODUCING PHONETIC SCIENCE

(c) [lateral, voiced]

(d) [sonorant, continuant, nasal]
(e) [anterior, continuant]

Exercise 9.5
Look at the following transcribed words of Japanese which contain examples
of all of the vowel qualities employed by the language. Give statements of
the distributions of the three fricatives [h ç φ] paying particular attention to
the following vowel.

[hoŋ] book [çikiφɯne] tugboat

[hako] box [heta] unskilful
[kooçii] coffee [φɯ] urban prefecture
[çi] day [heŋ] vicinity
[çira
ana] Hiragana (writing) [saiφɯ] wallet
[nihoŋ] Japan [çiroi] wide
[heja] room [φɯde] writing-brush
[ha] tooth [hai] yes

Do you think that the three fricatives in question are in complementary

distribution? Can they be regarded as allophones of one phoneme? What
indicates that [s] and [ h] must be allophones of distinct phonemes?

Exercise 9.6
The transcribed words below are from the language Campa (Peru). Notice
that some items are given in two forms. This indicates that two pronuncia-
tions are used interchangeably by speakers of the language. (When it makes
no difference which of two sounds is used, the sounds are said to be in free
variation.) The sequence [ts] represents a voiceless alveolar affricate.
a kind of bird [εto]
a rodent [ʃawo]
ant [katstoɾi]
blind [ʃetεɾi]
butterfly [thampi] [thambi]
corn [s#ŋki] [s#ŋ
i]
dead limb [oεmpiki] [oεmbiki]
egg [ithoki]
fish [s#wa] [s#wə]
I doubt [nokihotats#]
macaw [hawawo]
my paddle [nomaɾε]
nothing [tikats#]
on his head [itoki]
salt [ts#
i]
small gourd [iʃko]
spider monkey [tho
εɾo]
termite [kaεɾo]
vine or rope [ʃi
itha] [ʃi
ithə]
we chase [akεntiɾi] [akεndiri
your daughter [piʃinto] [piʃindo]
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 151 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Basic phonological concepts 151

(a) The data include both voiced and voiceless plosives. List them and
describe them. Are voiced plosives in contrast with voiceless plosives in
this language? Give evidence and discussion.
(b) Is the voiceless aspirated alveolar plosive in contrast with the voiceless
unaspirated alveolar plosive in this language? Give evidence.
(c) Examine the distributions of [ p b
w]. Might all of them be allophones
of a single phoneme?
(d) On the evidence given here, how many different nasal consonant
phonemes must be set up for this language, and what are they?
(e) List the vowel sounds that occur in the data, and give a description for
each (specifying tongue and lip position). Draw a vowel quadrilateral and
plot on it the approximate positions of these vowel sounds.
(f) What vowel phonemes need to be set up for this language? Explain why
the number of vowel phonemes is smaller than the number of different
vowel sounds that occur in the data.
Exercise 9.7
Explain in words what effect the following rules will have:
(a) VS [+nasal]/ [+nasal]
(b) [sonorant]S [voiced]/¬#
(c) Ø S [ɹ]/ V —– V
[hi
h]

Further reading
We have given only the most basic introduction to phonological concepts. Modern phonology is a field where
there are many competing approaches, and constant innovation in theories. For some more recent ideas applied
to English, see H. Giegerich, English phonology: An introduction, Cambridge: Cambridge University Press, 1992.
A book that explains the later development of the ideas explained in this chapter is C. Gussenhoven and
H. Jacobs, Understanding phonology, London: Arnold, 1998.
0521808820c09_p135_152.qxd 1/27/05 3:03 PM Page 152 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 153 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

CHAPTER

10 Suprasegmentals

CHAPTER OUTLINE
In this chapter you will learn about: the use of fundamental
KEY TERMS frequency; the difference between lexical stress and
Accent rhythmic stress; lexical stress types, including fixed stress
Contour tone and variable stress; lexical tone languages, including contour
Fixed stress tone and register tone languages; intonation and its role in
Fundamental determining the interpretation of utterances; paralinguistic
frequency features and their use in signalling speaker attitude.
Intonation
Key
Lexical tone
Nuclear tone
Nucleus
Paralinguistic
features
Pitch
Register tone
Rhythm
Stress, primary
and secondary
Variable stress
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 154 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

154 INTRODUCING PHONETIC SCIENCE

Introduction
We have looked in some detail at how speech sounds are produced, how
they may influence each other and how they are organised into syllables.
However, speech is not just a string of speech sounds. There are features
of speech that span a number of speech sounds or a number of syllables
or even whole utterances. Such features are called suprasegmental or
prosodic features. They include lexical and rhythmic stress, lexical tone
and intonation. In this chapter we will deal with variation in fundamen-
tal frequency and pitch and see how pitch variation is used in language to
signal the difference between the meanings of words in lexical tone lan-
guages and how pitch is connected with lexical stress. Pitch variation is
also very important at the level of the utterance. Intonation is used to
signal how a speaker intends his or her utterances to be interpreted.

Fundamental frequency and pitch

We saw in Chapter 2 that the vocal folds can vibrate at different rates. The
typical range of rate of vibration for an adult male speaker is from about
80 Hz (80 complete cycles of vibration) to about 210 Hz. The typical range
for adult female speakers is from about 150 Hz to about 320 Hz. The rate of
vibration of the vocal folds determines the fundamental frequency (F0) of
the resulting speech sound. This is the rate at which the speech pressure
waveform repeats. Figure 10.1 shows three short sections of the vowel [ɑ],
pronounced by an adult male. The waveforms and a plot of the F0 are given
for each. Vowel (a) has a fairly constant F0 of around 200 Hz. Vowel (b) has a
constant F0 of approximately 100 Hz. Notice that the repetition rate visible
in the waveform is much lower in the second vowel. Each cycle of vibration
is longer. Vowel (c) has a falling F0 pattern, starting at about the same value
as the first vowel and finishing a little lower than that of the second vowel.
The fundamental frequency of a sound is intimately connected with the
pitch we perceive. The higher the F0, the higher the pitch. So vowel (a) would
sound higher in pitch than vowel (b) and vowel (c) would be heard as having
a fall in pitch from high to low.
It is important to remember that it is not the absolute Hz values of a fun-
damental frequency contour that matter, but the relative values. So, for
example, if a female speaker were to produce the vowel [ɑ] at high pitch, at
low pitch and with a falling pitch, it is very likely that the Hz values would
all be greater than those for the male speaker above. This is because typi-
cally women have a smaller larynx and shorter vocal folds than men.

Stress
Stress affects whole syllables rather than single segments. The effect of
stressing a syllable is to make it more prominent, more audible than
neighbouring syllables. Thus, stress is a relational feature, unlike features
such as place and manner features or vowel quality features. So while it
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 155 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

Suprasegmentals 155

FIGURE 10.1
(a) 25 ms sample of
[ɑ] at a high pitch, and its
measured fundamental
frequency of 200 Hz;
(b) 25 ms sample of [ɑ]
at low pitch, fundamental
100 Hz; (c) a fall from
200 Hz to 100 Hz in
about 650 ms.
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 156 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

156 INTRODUCING PHONETIC SCIENCE

makes sense to ask whether an isolated consonant is, for example, alveo-
lar, or whether an isolated vowel is back, it can never be sensible to ask
whether an isolated syllable is stressed. The way in which a stressed sylla-
ble is made more prominent than other syllables in the utterance is
usually by a combination of factors: length, loudness and pitch. Thus, usu-
ally a stressed syllable is longer and louder than its neighbours and may be
marked by some pitch movement or new level in pitch. Look at Figure 10.2.
This is a display of the speech pressure waveforms of the utterance of two
English words by a male speaker. The words are written [ɹtən] and return
[ɹtn]. Also shown is the fundamental frequency variation for these
utterances. Notice that the two syllables of each of these words are clearly
discernible because the consonant that separates them is a voiceless

FIGURE 10.2
Waveforms and
fundamental frequency
traces for English written
and return.
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 157 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

Suprasegmentals 157

Table 10.1 Fixed penultimate stress in Welsh

[mənð] ‘mountain’ [mənəðɔð] ‘mountains’

[əs ɔ] ‘school’ [əs ɔljɔn] ‘schools’

[bυrjad] ‘intention’ [bυrjadi] ‘to intend’

[ə ɔd] ‘mice’ [ə ɔden] ‘mouse’

plosive and there is no waveform activity during its hold phase. The first
syllable of written is longer than the second and the excursion of the wave-
form above and below the centre line is greater for the first syllable than
the second. The extent of this excursion is correlated with loudness. For
return it is the second syllable that shows greater length and loudness.
There is also a pitch difference between the utterances of these two words.
The pitch of the first syllable of written is high and falling, while this pattern
occurs on the second syllable of return.

Primary and secondary stress

In words of two or three syllables it is likely that there will be only one
stressed syllable. However, for longer words two or more syllables may
bear stress. One of the stresses will be more prominent than the others
and is called the primary stress. This is marked, as we have been doing
already, with a raised vertical tick at the beginning of the syllable as
in: [pləυ] and [bləυ]. Other stresses in a word are called secondary
stresses and are marked with a lowered tick as in: [ kɒnsəntreʃn] and
[ nd vzbləti]. In some books, especially those published in the USA,
you may see an alternative method of stress marking using diacritics over
the vowel of the relevant syllables, as in [ ̀ndv̀zb́ləti].

Fixed stress
Some languages have fixed stress in the sense that most words in the lan-
guage bear stress on the same syllable. Favoured places for stress in fixed
stress languages are the first syllable of the word, the last syllable of the
word and the penultimate syllable of the word. Look at the Welsh data in
Table 10.1. In the left-hand column are words that all bear stress on the
penultimate syllable. In the right-hand column the same words appear, but
now with a single syllable suffix added. Notice how the stress is shifted to
remain on the penultimate syllable of the new word. The vast majority of
words in Welsh have stress on the penultimate syllable, although there are
a few exceptional words all of which bear stress on the final syllable.
Now look at the Czech data in Table 10.2. Czech is a fixed stress lan-
guage with word-initial stress. The addition of prefixes to a word makes no
difference to this. The stress moves so that it is still word-initial.
One final example of a fixed stress language, this time with word-final
stress, is Turkish. There are a couple of examples in Table 10.3. Again, the
addition of a suffix, or even more than one suffix, does not disturb this
pattern.
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 158 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

158 INTRODUCING PHONETIC SCIENCE

Table 10.2 Fixed initial stress in Czech

[mluvil] ‘he/she spoke’ [nεmluvil] ‘he/she didn’t speak’

[maso] ‘meat, flesh’ [bεzmasi] ‘fleshless’

[pɔrυtʃit] ‘to commend’ [ɔtpɔrutʃit] ‘to recommend’

[stali] ‘they stood’ [nastali] ‘they arose’

Table 10.3 Fixed final stress in Turkish

[otel] ‘hotel’ [oteller] ‘hotels’ [otellerden] ‘from the hotels’

[adam] ‘man’ [adamlar] ‘men’ [adamlarɯn] ‘of the men’

Table 10.4 Variable stress in Catalan

[əstel] ‘star’ [bensəs] ‘you defeat’ [ɔmniβus] ‘omnibus’

[kəlaʃ] ‘drawer’ [bərəaβəm] ‘we mixed’ [demɔkrətə] ‘democrat’

Variable stress
Many languages have variable stress (sometimes called free stress). This
does not mean that a particular word sometimes has one stress pattern
and sometimes a different one. It means that one cannot make a simple,
general statement as to which syllable of all (or most) words will be
stressed. Look at the Catalan data in Table 10.4.
The words in the first column all have final stress. Those in the third
column all have penultimate stress and those in the fifth column have
antepenultimate stress. These are the only possibilities for Catalan. So
although Catalan is a variable stress language, one can still make a gen-
eralisation about its stress placement: stress never falls more than three
syllables from the end of a word. This and similar restrictions are common
in variable stress languages.

Syllable weight and stress

Many variable stress languages favour syllables of a particular type to
bear stress. A heavy syllable is one that contains a long vowel or diph-
thong, or a short vowel followed by at least two consonants. A light
syllable is one that contains a short vowel followed by one consonant or
no consonants. Examples of heavy syllables in English are the first syl-
lables of the words beater, super, moaning, mighty, under, sixty. Examples
of light syllables are the final syllables in horrid, hiccup, city, water. In
many variable stress languages heavy syllables in a word attract stress.
Look at the English verbs in Table 10.5. If the final syllable of a verb is
heavy then that syllable is stressed. If the final syllable of a verb is light
then the penultimate syllable is stressed, no matter whether it itself is
heavy or light.
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 159 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

Suprasegmentals 159

Table 10.5 Stress and syllable weight in English verbs

Final heavy syllable: believe contend depart explain resign

Final light syllable: enter edit envelop imagine listen

Table 10.6 Stress in English Nouns

Final heavy syllable: attitude porcupine universe

Final light syllable: aroma asbestos computer

No heavy syllables: America insulin jealousy

You should remember however that the stress system of English is very
complex and it is not difficult to find exceptions to the above and to most
general statements about stress in English.

Stress and word class

Stress placement in variable stress languages may be sensitive to the lexical
class of a word, that is, whether it is a noun, a verb, an adjective and so on.
English is an example of such a language. Look at the English nouns in
Table 10.6 and compare them with the verbs we have just looked at. The
first thing to notice is that none of them have stress on the final syllable,
unlike the verbs. Secondly, heavy syllables in these nouns attract stress if
they are penultimate, not if they are final. Another thing to notice is that
if there are no heavy syllables in a word then the antepenultimate sylla-
ble is stressed. Again, though, it is not difficult to find exceptions to this
generalisation in English.

Stress and suffixes

In the fixed stress languages we examined earlier the addition of a suffix
to create a new word resulted in an adjustment of the place of the stress.
In some variable stress languages, some suffixes also cause stress to shift,
while other suffixes leave the stress exactly where it was before the suffix
was added. The type of suffix that leaves the stress alone we will call stress-
neutral suffixes and those that cause a change in the place of stress we
will call stress-imposing suffixes. We will illustrate the difference using
English. Look at the examples in Table 10.7. All of the suffixes in the first
group (there are many more) are simply added to the word without any
adjustment of the stress pattern. The suffixes in the second group (again
there are more than presented here) cause the stress to shift to the sylla-
ble immediately before the suffix.

Stress and vowel quality

In many languages the placement of stress has implications for the iden-
tity of the vowels used in the pronunciation of a word. For instance, some
vowels in a language may be unable to occur in a stressed syllable. A good
example of this is [ə] in many accents of English. If the addition of a suffix
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 160 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

160 INTRODUCING PHONETIC SCIENCE

Table 10.7 Examples of English stress-neutral and stress-imposing suffixes

Stress-neutral

-ed edit [edt] edited [edtəd]

-er pretty [prti] prettier [prtiə]

-est lovely [lvli] loveliest [lvliəst]

-ing examine [ zmn] examining [ zmnŋ]

-ive suggest [sədest] suggestive [sədestv]

-ly rapid [rpd] rapidly [rpdli]

Stress-imposing

-al universe [junivs] universal [junivsl]

-ic photograph [fəυtə rf] photographic [fəυtə rfk]

-ical hypothesis [hapɒθəss] hypothetical [hapəθetkl]

-ity complex [kɒmpleks] complexity [kəmpleksəti]

Table 10.8 Stress minimal pairs

Albanian [atə] ‘father‘ [atə] ‘him, her‘

Arabic [wasafa] ‘they described‘ [wasafa] ‘and it cleared up‘

Greek [politikɔs] ‘of Constantinople‘ [politikɔs] ‘political‘

Italian [pɔrtɔ] ‘I carry‘ [pɔrtɔ] ‘he/she carried‘

Russian [ue] ‘narrower‘ [ue] ‘already‘

Spanish [entre] ‘between‘ [entre] ‘I entered‘

to a word imposes stress on a syllable that contains [ə] in the unsuffixed

form, then [ə] will change to another vowel. An example is [ə] → [ɒ] in the
second syllable of the words photograph [fəυtə rf] and photography
[fətɒ rəfi]. An alternative way of looking at this is to say that the shift of
stress away from a syllable often causes a vowel to be weakened to [ə].

Stress contrasts
In variable stress languages it is often possible to demonstrate that the
place of stress is contrastive. That is, it is possible to find minimal pairs
that differ only in the placement of stress. The two English words import
(verb) [mpɔt] and import (noun) [mpɔt] constitute such a pair. Look at
Table 10.8 where there are some more examples of stress minimal pairs
from various languages.

Rhythm
So far we have looked at stress as a property of words. This is sometimes
called lexical stress. However, when words are strung together in phrases
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 161 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

Suprasegmentals 161

Table 10.9 Foot duration measurements

Foot Syllable count Duration (ms)

John 1 406

can’t have for- 3 542

-gotten 2 427

Sally’s 2 500

birthday 2 676

or sentences some of the stressed syllables remain stressed and others

may lose their stress. Let’s take a simple example from English. The word
over is stressed on the first syllable – [əυvə]. However, if we put this into a
sentence like The dog jumped over the fence, we may hear that this syllable
has no stress. The stress pattern of the sentence could be [ðə dɒ dmpt
əυvə ðə fens], with no noticeable prominence on the first syllable of the
word over. Of course it is possible to say the sentence in a number of dif-
ferent ways and some of these may have may have a stress on over, for
instance, [ðə dɒ dmpt əυvə ðə fens]. So lexical stress and stress in
sentences, while obviously connected, are not exactly the same thing.
Here is another example of the difference between lexical stress and what
is often called sentence stress or rhythmic stress. The word afternoon is
[ ɑftənun], with secondary lexical stress on the first syllable and primary
lexical stress on the last. However, a phrase like afternoon tea is likely to be
pronounced as [ɑftənun ti] rather than [ ɑftənun ti]. The primary
stress has moved to the first syllable of the word.
Stress in many languages is what defines the rhythm of speech. Rhythm
can be defined as the pattern of occurrence in time of relatively ‘strong’
and relatively ‘weak’ events. In a language like English, the strong events
are stressed syllables and the weak events are the unstressed ones. There
is a tendency, and perhaps it is no more than that, for stressed syllables to
occur at roughly equal intervals in time in English and in other languages.
If we listen to a sentence like:
John can’t have forgotten Sallys birthday

the strong beats that fall on the stressed syllables appear to be roughly
equally spaced in time, athough as we can see, the number of unstressed
syllables between each pair of beats varies: 0 between the first pair, 2
between the second pair, and 1 between the third and fourth pairs. In
Table 10.9 are some very rough measurements in milliseconds of the
durations of each stressed syllable and any following unstressed ones,
taken from a recording of the sentence.
Each stressed syllable and any following syllables constitute a unit
known as a foot. Compare the duration of the first and second feet.
Certainly the second is longer than the first, but it is nowhere near three
times as long, although there are three times as many syllables. This
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 162 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

162 INTRODUCING PHONETIC SCIENCE

means that the syllable rate in the second foot must be faster than that of
the first foot. The constant alteration of syllable rate to maintain a rough-
ly equal foot duration is characteristic of many accents of English and also
of other languages. These languages are sometimes called stress-timed.
Not all languages, and not even all accents of English, are like this. French,
for example, tends to have, or at least to sound as if it has, most syllables
equal in duration. Languages like this are called syllable-timed.
This distinction between stress-timed and syllable-timed languages is
very probably an oversimplification. It is probably more accurate to say
that some languages make greater use of one kind of rhythm, but that
both types can be found in most languages. It is also true that different
accents of the same language may have different rhythmic characteristics.

Rhythmic stress and word type

You may have noticed in the examples above that the word the was not
marked as bearing rhythmic stress. This word is an example of what is
often called a grammatical word. Other examples from English are pro-
nouns like he, she, him, her, us, prepositions like at, for, from, modal verbs
like can, may, must, possessives like my, his, their and conjunctions like and,
but, as. Grammatical words in English and in most languages are less often
rhythmically stressed than words like nouns (book, John, honesty), adjectives
(red, silly, international), main verbs (look, contemplate, understand) and
adverbs (often, slowly, intentionally). These sorts of words, called content
words, very often do bear rhythmic stress. Here is a short passage of Welsh
in phonetic symbols with a word for word translation. Notice how nearly
all the content words are stressed and the grammatical words unstressed.
You might like to try reading the passage out loud.
[pan oið ʃon wiljam a i dəili n buita i kinjau
when was John William and his family in eating their dinner
daiθ karre i laur o bεn ər at a θrui r to
came rock in down from top the slope and through the roof
a dis ənoð ar drəinʃur ʃon wiljam
and fell on plate John William
a divεθθoð i dattus a i  i ]
and destroyed his potatoes and his meat
Of course, grammatical words may be stressed in certain circumstances.
Take for example the English sentence Hey those are my socks, not yours! This
is likely to be pronounced with the words my and yours stressed to indicate
the contrast between the two words.

Lexical tone
The Modern Standard Chinese sentence for I haven’t got a pen is
[wo mei jou bi]. However, if you say this, you have to be careful about the
pitch pattern you use, especially on the final word [bi], which means ‘pen’.
The pitch pattern should be a low, slightly falling then slightly rising tone.
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 163 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

Suprasegmentals 163

Table 10.10 Tonal contrasts in Modern Standard Chinese

Tone 1 Tone2 Tone 3 Tone 4

[fu] husband to help rotten father

[ma] mother hemp horse to scold

h
[t an] to try deep level to sigh

The corresponding fundamental frequency trace is shown in Figure 10.3(a).

If you make a mistake and use a high rising pitch pattern instead then you
would be saying I haven’t got a nose! The fundamental frequency for [bi]
meaning ‘nose’ is given in Figure 10.3(b).

FIGURE 10.3
Fundamental frequency
for MSC ‘pen’ and ‘nose’.

Modern Standard Chinese, like nearly half the world’s languages, uses
Examples of tone
pitch patterns to distinguish between one word and another. It is a lexical
languages from
tone language. Lexical tone languages are found throughout Asia, in
Asia, Africa and
many parts of Africa and in the Americas. Some lexical tone systems are Americas
very simple, consisting of just a high level tone and a low level tone. Some Asia: Chinese, Thai,
are more complex. Cantonese, for example, has six distinct tone patterns. Vietnamese
Africa: Hausa,
Contour tone Yoruba, Xhosa, Zulu
There are two kinds of tone language. The first type, of which Modern Americas: Mixtec,
Standard Chinese is an example, have tones where the pitch movement is Navajo
important and it is the difference in the pattern of pitch movement –
falling, rising, falling-rising and so on – that distinguishes one word from
another. There are four tones in Modern Standard Chinese: (1) high level,
(2) high rising, (3) low falling-rising, (4) high falling. Table 10.10 shows
some four-way tonal contrasts.
Languages such as this, where at least some of the tones have moving
pitch patterns are called contour tone languages. Many lexical tone lan-
guages of Asia are of this sort.

Register tone
In contrast to contour tone languages, some languages have tone invento-
ries that consist entirely of level tones. Table 10.11 shows some examples
of two-way and three-way contrasts, involving level tones, from various
languages.
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 164 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

164 INTRODUCING PHONETIC SCIENCE

Table 10.11 Tone contrasts in register tone languages

High Mid Low

Bafang [lox] sleep — stone

(Cameroon)

Bole [lo] meat — who

(Nigeria)

Karen [tə] one spoon ant

(Burma)

Margi [ʃu] tail — to dry up

(Nigeria)

Mixtec [ juku] (2nd yoke mountain brush

(Mexico) syllable)
Nupe [ba] to be sour to cut to count
(Nigeria)

Yoruba [wa] comes looked existed

(Nigeria)

Table 10.12 IPA tone transcription

Extra Extra High Low Rising
high High Mid Low low Falling Rising rising rising falling

bã bá bā bà ba
`` bâ bǎ bă bă bã

ba ba ba ba ba ba ba ba ba ba

A language that uses level tones only is known as a register tone language.
Sample text in Many African tonal languages are this sort, although register tones
Yoruba
languages are found in other parts of the world too.
Gbogbo ènìyàn ni a
bí ní òmìnira Transcription and representation of tones
Yoruba is a West The chart of the IPA provides diacritics for marking tone, and these are
African register illustrated in Table 10.12 using the syllable [ba]. Also in the table you will
tone language with find what the IPA calls tone letters, which are a simple graphical way of
high, mid and low indicating the shape of a tone.
tones. High tone is Yet one more way to represent the shape and pitch height of a tone is by
marked in the the use of superscript numbers, with 5
extra high, 4
high, 3
mid,
orthography by an 2
low and 1
extra low. Usually two superscript numbers are used for
acute accent, low each tone, the first indicating the height at the beginning of the tone and
tone by a grave the second that at the end. So, for example, [ba44] indicates a high level
accent and mid tone
tone, while [ba42] indicates a tone that falls from high to low. More com-
is left unmarked.
plex tone shapes need three numbers, so [ba434] is a way of representing a
high falling-rising tone.
Many of the above ways of transcribing tones suggest that contour tones
are best thought of as being in some sense sequences of level tone targets,
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 165 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

Suprasegmentals 165

and in fact much recent work on the phonology of lexical tone has been
carried out under this assumption. Modern phonological treatments of
tone make use of what is known as an autosegmental representation,
where the vowels and consonants of a syllable are present on one tier of
the representation and the tone on an independent tier. The tiers are
connected by association lines as in Figure 10.4, where H means high, M
means mid and L means low.

ba ba ba

(a) (b) (c) FIGURE 10.4

The autosegmental
H L L M representation of tone.

In Figure 10.4 (a) represents a high level tone, (b) a low level tone and (c) a
tone that rises from low to mid. The difference between what we have
called register tone languages and contour tone languages can now be
thought of in a slightly different way. Register tone languages allow only
one tonal target to be associated to each syllable, whereas contour tone
languages allow two or more tonal targets to be associated to a syllable.

Tones in connected speech

We have seen that vowels and consonants may turn up in different guises
in different environments. This is true of tones too. If we look at Figure 10.5
which is a schematic representation of the pitch contour for the Modern
Standard Chinese phrase [hau bu hau], which means ‘OK?’ (literally FIGURE 10.5
‘good not good’), we can see that the pattern for the first occurrence of Schematic representation
[hau] is not the same as that for the second. They both have Tone 3, low of the pitch contour for
falling-rising, but the non-final occurrence has lost its rising part. So we the Modern Standard
Chinese phrase
can view Tone 3 as a toneme, in the same way as /p/ is a phoneme. Modern
[hau bu hau].
Standard Chinese Tone 3 has at least two variants or allotones, one that is
used in final position before a pause, and the other in non-final position.
The two variants are in complementary distribution.
It is also possible to find examples of the tonal equivalent of assimila-
tion, where the actual pitch height of a tone is affected by that of a neigh-
bouring tone. The example in Figure 10.6 is a schematic representation of
the pitch heights of the syllables in the word [ód ì bó] (‘banana’) from
Urhobo (Nigeria).
FIGURE 10.6
The first and third syllables of the word both have high tone, but the
Tonal assimilation in
actual pitch height of the third syllable is lower than that of the first Urhobo.
because of the low-toned syllable that precedes.

Intonation
Languages like English, French, German, Spanish, Italian are not lexical tone
languages. There are no pairs of words in these languages that are distin-
guished solely by the pitch pattern used to say them. However, all languages,
as far as is known, use pitch variation to communicate meaning.
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 166 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

166 INTRODUCING PHONETIC SCIENCE

Let’s take a simple example from Italian. The two utterances in Figure 10.7
One of the earliest use the same words, but (a) is a statement meaning ‘It’s cold’ and (b) is a
accounts of English
question meaning ‘Is it cold?’ As you can see, (a) ends with a falling fun-
intonation was
damental frequency contour, while that for (b) rises at the end.
published in 1775.
It was written by This use of pitch variation that does not affect the meanings of individ-
Joshua Steele. He ual words, but that does affect the way in which utterances are interpreted,
attempted to use a is known as intonation. In the rest of this section we will look at some of
kind of musical the main characteristics of the intonation system of English and draw a
notation to repre- few comparisons with what goes on in other languages.
sent the rhythm
and pitch move-
ment in a speech by
the actor David
Garrick.

FIGURE 10.7
Fundamental frequency
traces for Italian Fa freddo
vs Fa freddo? (‘It’s
cold’/’Is it cold?’).

Intonational phrases
One of the functions of intonation is to divide what we say into manageable-
sized chunks known as intonational phrases or IPs. For any given bit of
speech a speaker has a considerable freedom of choice about how many
IPs to use and where they start and end. The two examples below are taken
from recordings of the same passage by two different speakers. The sym-
bol | indicates the boundary between two IPs and || indicates an IP bound-
ary accompanied by a pause.
Speaker 1: Work is scheduled to continue on both carriageways | to
the south of junction seventeen | the M45 intersection ||
and there are more northbound lane closures | about five
miles north of junction twenty | the turnoff | to
Lutterworth.
Speaker 2: Work is scheduled to continue | on both carriageways | to
the south of junction seventeen | the M45 intersection ||
and there are more northbound lane closures | about five
miles north of junction twenty | the turnoff to Lutterworth.
We shall see later on exactly how a speaker signals the end of an IP and
the start of a new one, but for the time being, notice that IP boundaries
are not always accompanied by a pause.
In English and other languages it is possible to use IP boundary placement
to signal which of two possible meanings an utterance has. For example, the
sentence Those who spoke quickly got an angry response is ambiguous. Is it the
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 167 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

Suprasegmentals 167

speaking that is quick or the response? If it is the first, then the sentence is
likely to be uttered as:
Those who spoke quickly | got an angry response
but the second meaning is much more likely to be signalled by
Those who spoke | quickly got an angry response
In general, the placement of IP boundaries is used by the speaker to signal
his or her judgement of what belongs with what. Here is another pair of
examples:
Old men and women
Old men | and women
In the first, the hearer is likely to conclude that both the men and the
women are old. In the second, the men are definitely old, but what about
the women?

Accent and nucleus

In Figure 10.8 (a ) and (b) are schematic representations of the pitch con-
tour of two utterances of the sentence I don’t remember his telephone number.
Each is produced as one IP. The large circles represent stressed syllables
and the smaller ones unstressed syllables. The tail attached to a circle
indicates that there is a very noticeable glide in pitch during the produc-
tion of that syllable. The upper and lower lines represent the limits of the
speaker’s pitch range.

(a)

(b)

FIGURE 10.8
Stress, accent and nucleus.

As you can see both (a) and (b) contain four stressed syllables, exactly the
same ones in each case. In (a) the stressed syllables -mem- and num- are at
the same pitch level as the preceding and following syllables. The syllables
don’t and tel-, however, are different. They both initiate a new trend in the
pitch contour. They are not only stressed, but also accented. In (b) there
are also two accented syllables – don’t again and -mem-. The final accented
syllable in an IP is known as the nucleus and its placement is very impor-
tant for the interpretation of an utterance. The nucleus in (a) is tel- and in
(b) it is -mem-.
Again the speaker, at least in English, has a good deal of choice where
to put the nucleus within an IP. For our sentence above, the nucleus could
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 168 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

168 INTRODUCING PHONETIC SCIENCE

be any of the following in addition to the two places we have already men-
tioned: I, don’t, his. You might like to try sketching the pitch contours of
the utterances with the nucleus in those positions and to try saying the
utterances.
The main reason for choosing a particular position for the nucleus is
to focus the hearer’s attention on important or new information in the
IP. If the speaker considers that all the information is new to the hearer
(this is called broad focus), then the nucleus will be placed in a ‘default’
position. For most IPs in English the default position is the primary
stressed syllable of the last content word. If the speaker considers that
some of the information is old or unimportant because it is inferable
from the context in which the conversation is taking place, then narrow
focus will be used and the nucleus is likely to be placed in a position
other than the default.
Let’s try and make all that clearer with a couple of examples. The position
of accented syllables will be indicated by using capital letters and under-
lining will be used to mark the nucleus. First a broad focus utterance:
Fred: WHY are you looking so anNOYed?
Bill: I’ve LOST my WALLet.
Fred’s question doesn’t mention losing things or wallets, so Bill’s reply has
broad focus and the nucleus falls on wallet, which is the last content word
in the IP. Now a narrow focus utterance:
Sue: LET’S go and see a MOVie.
Sally: I CAN’T afFORD to go to the cinema.
Here Sue’s utterance mentions going somewhere, so Sally avoids placing
the nucleus on go. Sue’s words see a movie imply going to the cinema, so
that is old information too and Sally avoids placing the nucleus on cinema.
The last piece of new information in Sally’s reply is afford, so that is where
she chooses to place the nucleus.
Accounting for nucleus placement in English, and in other languages,
is a fairly complex problem and one can easily find examples of utterances
where the nucleus placement is rather puzzling. Take this little conversa-
tion for example:
Mary: Could you pass me the German dictionary, please?
Jim: Here you are.
Mary: Ah, I wanted the German dictionary | You’ve GIVen me the
wrong ONE.
At least in some accents of English Mary’s last IP is perfectly usual, but it
is difficult to see why the nucleus should be the word one. It isn’t a con-
tent word and it certainly isn’t very informative.
Not all languages are as flexible as English when it comes to moving the
nucleus around within an IP. Take for example the following utterance in
Spanish (given in orthography) and compare it with the equivalent
English utterance:
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 169 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

Suprasegmentals 169

Con LECHe | o sin LECHe

WITH milk | or withOUT milk
English places the nuclei on the contrasting items in the two IPs, but
Spanish sticks to the last content word in both, even though it is exactly
the same word.

Intonational tone
Both the examples in Figure 10.8 above show the nucleus as having a high
falling tone. However, this is not always the case. In Figure 10.9 we have
the same sentence, again produced as one IP, with the nucleus on the syl-
lable tel-. However, this time the pitch rises from the nucleus to the end of
the IP.

FIGURE 10.9
A rising nuclear tone.

Notice also that the pitch pattern of the part of the IP before the nucleus is
slightly different too. We have introduced an additional accent by making
the pitch step down slightly at the syllable -mem-.
A speaker of English may choose from a small number of pitch patterns
known as nuclear tones which begin at the nucleus and continue to the
end of the IP. Exactly how many nuclear tones there are and what types
there are is a matter of some debate amongst intonation analysts. Also dif-
ferent accents of English certainly make use of a different set of nuclear
tones.
We will not go into the fine details of the nuclear tones of English here,
but will concentrate on the sorts of meaning difference that nuclear tone
choice can signal. Before that we need to clear up a matter that we left
unresolved earlier. How does a speaker signal the end of an IP, even when
there is no pause associated with the IP boundary? The simple answer to
that question is: by using a nuclear tone. Hearers know that an IP has
come to an end because they know how IPs can end, rather like the way
in which we can usually tell when a musical tune has ended. If the tune
stops before its end, then we can usually tell it is incomplete, even if we
have never heard the tune before. In a similar way, we can tell that an
intonation pattern is incomplete if we haven’t heard one of the patterns
of pitch that count as a nuclear tone in the accent or language we are
familiar with.

Functions of intonational tone

One very important use of intonational tone, at least in English, is to sig-
nal the speaker’s assessment of how the information in an IP fits into
the conversation she or he is taking part in. The use of a falling nuclear
tone is the speaker’s signal that the information is being added to the
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 170 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

170 INTRODUCING PHONETIC SCIENCE

conversation, that it is new. A speaker can also signal that the informa-
tion in an IP is already part of the background to the conversation and is
being reactivated. This is done by choosing a falling-rising nuclear tone.
Here is an example with just the IP boundary and the nuclear tones
marked. The mark \ means falling and \/ means falling and then rising.
When you go to the \/supermarket | you can call in at the \bookshop.
Here the speaker signals that (in his or her opinion) going to the super-
market is already an established fact and already part of the knowledge
that the participants in the conversation share. The new contribution is
‘calling in at the bookshop’.
You can call in at the \/bookshop | when you go to the \supermarket.
In this second example, it is ‘calling in at the bookshop’ that is the estab-
lished fact. The new bit is when this should take place.
Another important use of nuclear tone choice is to signal the attitude
that the speaker wishes to convey towards the hearer, the subject matter,
or sometimes life in general. As a simple example of the sort of attitudi-
nal difference intonation can signal, let’s imagine you have a weepy child
with you. You might say Don’t cry or It’s all right. If you use a pattern where
the words before the nucleus are high and level in pitch and the nucleus
itself (cry or right) falls from mid pitch to low pitch, you would sound
decidedly grumpy and unsympathetic. However, if you used a low rising
nuclear tone, again with a high level pitch preceding it, you would sound
much more soothing and sympathetic.
It isn’t only nuclear tone choices that can signal attitudinal differ-
ences. Take the sentence I’m NOT going to the SUPermarket. With a high
level pitch from NOT to the and then a high falling nuclear tone on SUP,
this is simply a statement of fact. With a low level pitch on NOT which
then rises steadily through the following syllables as far as the and again
with a high falling nuclear tone on SUP, the utterance is not a neutral
sounding statement. The attitude conveyed is something like ‘I’ve told
you before. I am surprised you haven’t accepted the fact. I want to hear
no more about it.’
Another use of intonational tone is to signal the completeness or non-
completeness of an utterance. Some tones are used as a signal that there
is more to come. A simple example of this is a list of items like:
/carrots | /peas | /celery | po\tatoes
The rising nuclear tones on the first three items, shown with /, indicate
that the list is not yet complete and the falling tone on the last item says
that we have come to the end.

Key
We will briefly look at another aspect of intonation known as key. Key
affects IPs as a whole and raises or lowers the pitch of all the tones in
the IP. Just as we saw that intonational tone may be used to signal non-
completeness or completeness of a list, key may be used to signal the
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 171 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

Suprasegmentals 171

completeness or non-completeness of longer structures, rather like para-

graphs in written language. Each of these structures deals with what the
speaker considers to be a particular topic. The start of the topic is
marked by a high key IP. This is then followed by a number of mid key
IPs (possibly none at all). The end of the topic is marked by an IP in low
key. This is sometimes very evident when newsreaders are changing
from one news item to the next. See if you can spot it when you next
listen to the news.
Another use of key is to mark an IP as being a parenthetical comment,
rather like a bit of writing in brackets. In a sentence like:
I saw Fred Smith | you know the guy I’m talking about don’t you | in
the pub yesterday
the second IP is likely to be uttered with low key.

Paralinguistic features
When we speak, we are continually giving out signals about our emotional
state, our mood, our attitude towards the speakers in the conversation
and towards the topic of the conversation itself. Some of these signals are
non-verbal. Facial expression, posture and gesture can all tell our conver-
sational partners how we are feeling. Some of the signals, however, are
connected with the way we are actually speaking. These include the
tempo of our speech, the overall loudness, the overall pitch range, the fre-
quency of pauses and the type of phonation that we use. Such aspects of
our speech are known as paralinguistic features.
If we are excited about a topic we are likely to speak more rapidly, with
fewer pauses, with a wide pitch range and possibly quite loudly. If we are
bored the tempo may be slower and quieter and the pitch range narrowed.
Anger is often conveyed in our speech by increased loudness and wide
pitch range.
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 172 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

172 INTRODUCING PHONETIC SCIENCE

Chapter summary
In this chapter we have looked at some aspects of speech that affect
units larger than a single speech sound. Such suprasegmental features
include lexical stress and we distinguished between primary and sec-
ondary stress and, in variable stress languages, investigated the con-
nection between stress and word class and syllable weight. In fixed
stress languages these considerations are not an issue because stress
regularly falls on the same syllable within a word.
We also looked at lexical tone which occurs in a large number of the
world’s languages. Words in these languages may be distinguished by
their pitch pattern alone. We saw that there are two different types of
tone language: contour tone languages where some of the tone pat-
terns are falls in pitch, or rises in pitch, or sometimes more complex
pitch movements. Register tone languages on the other hand use only
tones that have a level pitch contour. We saw that tones, just like vow-
els and consonants, may have contextual variants.
Intonation is the use of pitch variation to aid the interpretation of
utterances rather than to signal word meaning. We saw that speech
can be divided into intonational phrases and that the choices available
to the speaker in the placement of IP boundaries may affect how an
utterance is interpreted. Within each IP, the speaker may choose to
make certain syllables accented and the final accent, called the nucleus,
is a very important way of highlighting information that the speaker
considers new or significant. Accented syllables may be associated with
pitch contours of different types and the choice of intonational tone is
used both to signal speaker attitude and to indicate the speaker’s
assessment of whether the information in an IP is new or whether it is
already shared knowledge. We also briefly looked at key and its role in
organising speech into topics.
Finally, we introduced the concept of paralinguistic features such as
tempo, pitch range, loudness, and their role in indicating the speaker’s
attitude or mood.
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 173 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

Suprasegmentals 173

Exercises
Exercise 10.1
Sort the following English words into three lists. List 1 should contain all the
words that have primary stress on the first syllable. List 2 should contain
words with primary stress on the second syllable. List 3 is for words with
primary stress on the last syllable.

kangaroo, mystery, fantasy, fantastic, recover, intention, concentrate, elephant,

millionaire, understand, excitement, troublesome, crocodile, afternoon, con-
nection, imagine, horizon, attitude, universe, puppeteer

Exercise 10.2
The placement of stress in Maori is governed by the following principles:
(a) Stress never falls more than four syllables from the end of the word
(b) Long vowels attract stress
(c) Diphthongs attract stress except in the final syllable
(d) Long vowels take precedence over diphthongs for stress placement
purposes
(e) Stress always falls as near to the beginning of the word as possible with-
out contravening (a)–(d).
Copy out the transcriptions of the Maori words below, placing a stress mark at
the beginning of the stressed syllable. Note: all vowel sequences count as
diphthongs and Maori does not allow consonants to follow the vowel within
the syllable. When you have done this and checked your answers, try saying
the words with the correct stress.

[atapo] ‘early dawn’ [papaɾakauta] ‘hotel’

[kihiɾimete] ‘Christmas’ [kawanataŋa] ‘government’
[paepae] ‘beam’ [kaumatua] ‘male elder’
[tauɾekaɾeka] ‘slave’ [papakaiŋa] ‘village’
[maɾae] ‘courtyard’ [kaimoana] ‘seafood’
[fafai] ‘war’ [mokopuna] ‘grandchild’
Exercise 10.3
Below are some words from Manam, a language of Papua New Guinea.

[alaŋa] ‘reef’ [motu] ‘island’

[boazŋa] ‘hole’ [wari e] ‘rope’
[malaboŋ] ‘flying fox’ [uzem] ‘I chewed’
[atabala] ‘up’ [moarepi] ‘rice’
[siŋabaʔ] ‘bush’ [urapun] ‘I waited’
[iʔint] ‘he pinched’ [ʔipoasa ena] ‘we are tired’
In some words the final syllable is stressed, in others it is the penultimate syl-
lable. What appears to determine which syllable is stressed? Try saying the
above words.

Exercise 10.4
Below are the waveforms for utterances of the two English words [mpɔt]
and [mpɔt]. Can you tell which is which?
What is the duration of the initial syllable in each case?
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 174 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

174 INTRODUCING PHONETIC SCIENCE

Figure for Exercise 10.4

0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 175 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

Suprasegmentals 175

Exercise 10.5
The variety of Chinese spoken in Chengdu in Sichuan has four tonemes:

I High Rising II Low falling III High Falling IV Low Falling-rising

tʃin45 tʃin21 tʃin41 tʃin213
‘clear’ ‘fine (weather)’ ‘to invite’ ‘to celebrate’

Look at the following transcriptions of two-syllable sequences. In each case

the syllables are given with their original tone and then the tone pattern for
the sequence is given.
oŋ45  fu45 → oŋ45 fu33 ‘time’
ue21  dia45 → ue21 dia33 ‘nation’
ʃiaŋ45  aŋ41 → ʃiaŋ45 aŋ41 ‘Hongkong’
fa21  dia21 → fa21 dia21 ‘hair clip’
li21  ba45 → li21 ba33 ‘fence’
fei21  dzau213 → fei21 dzauʔ211 ‘soap’
ən45  bən41 → ən45 bən41 ‘origin’
poŋ21  iəu41 → poŋ21 iəu41 ‘friend’
ʃiau41  tʃəu41 → ʃiau55 tʃəu41 ‘clown’
fu213  mu41 → fu213 mu41 ‘parents’
tʃin41  i45 → tʃin55 i33 ‘pyjamas’
ʃiau41  tʃi21 → ʃiau55 tʃi21 ‘small flags’
ʃiau41  tʃi213 → ʃiau55 tʃiʔ211 ‘stingy’
bau41  bau41 → bau55 bau41 ‘baby’
tʃiŋ45  dɔŋ213 → tʃiŋ45 dɔŋ211 ‘weight’
ʃi21  uan213 → ʃi21 uan211 ‘habit’
koŋ41  pa213 → koŋ55 paʔ211 ‘perhaps’
jin213  tʃi213 → jin213 tʃiʔ211 ‘luck’
di213  faŋ45 → di213 faŋ45 ‘place’
ʃiaŋ213  pi21 → ʃiaŋ213 pi21 ‘rubber’
ʃaŋ213  hai41 → ʃaŋ213 hai41 ‘Shanghai’
di213  di213 → di213 di33 ‘younger brother’
(a) Draw schematic pitch patterns (like that in Figure 10.5) for the words for
‘clown’ and ‘pyjamas’.
(b) Comment on the occurrence of [ʔ].
(c) Retranscribe the word for ‘time’ using IPA tone letters.
(d) List the variants of each toneme and state where each occurs.

Exercise 10.6
Each of the following sentences is ambiguous. Try to work out what
meanings each can have and how a particular meaning could be signalled
using intonation.
(a) My sister who lives in Liverpool has just changed her job.
(b) She played and I sang the national anthem.
(c) He didn’t leave because he was afraid.

Exercise 10.7
Here is a short dialogue that has been divided into IPs for you. The place of
the nucleus has been underlined in Sid’s utterances. Your task is to decide
where the nucleus falls in each IP of Clare’s utterances.
0521808820c10_p153_176.qxd 3/2/05 3:32 PM Page 176 Quark01 SUBHAKANT:CU1937:Chapters:Chapter-10:

176 INTRODUCING PHONETIC SCIENCE

Sid: I’ve got an idea | Let’s go out for a meal.

Clare: All right | Which restaurant do you want to go to?
Sid: There’s a good Chinese place | near the library.
Clare: I’m not terribly fond of Chinese food.
Sid: Maybe that’s because you’ve never had any good Chinese food.
Clare: That may be true | but I think I’d prefer to try that place | some
other time.

Further reading
A very good general introduction to the study of intonation is Alan Cruttenden, Intonation, Cambridge:
Cambridge University Press, 2nd edn, 1997. D. R. Ladd, Intonational phonology, Cambridge: Cambridge
University Press, 1996 is an advanced treatment of up-to-date ideas in the study of intonational form. Paul
Tench, The intonation systems of English, London: Cassell, 1996 is an excellent source of examples of
intonational features and their role in signalling utterance meaning. Moira Yip, Tone, Cambridge: Cambridge
University Press, 2002 is a wide-ranging treatment of lexical tone systems, largely in the framework of the
Optimality Theory model of phonology. We have not dealt with this theory in our book, but Moira Yip’s book
introduces the basic concepts of the theory.
0521808820c11_p177_188.qxd 1/27/05 3:08 PM Page 177 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

CHAPTER

11 Speaker and
hearer

CHAPTER OUTLINE
In this chapter you will learn about: the human hearing
KEY TERMS mechanism; the use of visual clues in speech perception;
Acoustic cue speech perception tests – labelling and discrimination;
Audiogram speech development in children; hearing impairment.
Categorical
perception
Cochlea
Cue redundancy
Discrimination
Labelling
0521808820c11_p177_188.qxd 1/27/05 3:08 PM Page 178 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

178 INTRODUCING PHONETIC SCIENCE

Introduction
This book has concentrated largely on how human beings produce speech
and how speech sounds are organised in languages of the world. However,
we usually speak for the purpose of communicating with others who
must receive the acoustic signal created by movements of the speech
organs. This signal must be decoded and then understood by the hearer.
We begin the chapter by looking at the hearing mechanism and the sup-
plementary use of visual cues to aid speech perception. We then describe
the methods used to investigate which aspects of the acoustic signal are
important for the perception of phonetic distinctions. Phoneticians are
not only interested in how speech is produced by the speech organs and
the acoustic properties of speech sounds, but also in how human beings
receive and process the speech signal. The study of speech perception is
one of the major applications of phonetics. Speech perception tests involv-
ing labelling and discrimination tasks with synthetic speech stimuli have
been used in many areas of research and we look briefly at some of these,
including cross-language comparison of acoustic cue perception. Next we
look at the development of speech perception and production in children.
The chapter concludes with an account of the ways in which hearing
impairment may interfere with the perception of speech.

The hearing mechanism

The expression ‘the organs of speech’ naturally calls to mind the tongue,
lips, larynx and so on, which are used to produce speech. But our ears, and
the hearing mechanism generally, have an equal claim to be considered
organs of speech, because speech must be detected and understood, as
well as produced. The characteristics and capabilities of human hearing
must have shaped the evolution of human speech quite as much as the
articulatory capabilities of our vocal tracts.
Hearing involves peripheral and central processing. Peripheral processing
is carried out in the ear itself, central processing in the cortex. The periph-
eral auditory mechanism consists of three parts as shown in Figure 11.1. The

FIGURE 11.1
The hearing mechanism.
0521808820c11_p177_188.qxd 1/27/05 3:08 PM Page 179 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Speaker and hearer 179

outer ear comprises the pinna (ear flap) and auditory canal (also called
meatus). These collect and channel sound. The eardrum (tympanic mem-
brane) divides the outer ear from the middle ear. In the middle ear three
tiny bones (the ossicles) act as a mechanical linkage passing on vibration
through another membrane (the oval window) to the inner ear. The organ
of hearing proper is a spiral structure, the cochlea. It is here that conver-
sion of vibration into electrical nerve impulses takes place, and prelimi-
nary analysis of sound according to frequency and loudness is carried out.
A coded representation of the sound passes along the auditory nerve as a
series of pulses.
There are considerable similarities between the human hearing mecha-
nism and that of other mammals, the biggest difference between species
being in the range of frequencies that can be heard. The lowest frequen-
cies that we hear as sounds are in the region of 30 Hz, the highest (for a
young person) around 20 kHz. Frequencies that are below our normal
range of hearing are referred to as infrasound and frequencies above our
normal hearing range are ultrasound, and some creatures are adapted
to respond to sound in those regions. Dogs, for example, can detect fre-
quencies higher than 20 kHz, and bats use ultrasound as a navigation
system.
One can test listeners’ sensitivity to sounds by presenting pure tones (sine
waves) one at a time, and varying the intensity to determine when each one
is just detectable. The testing must be done in a very quiet place and with
carefully calibrated equipment. Young adults with good hearing tend to
have a characteristic sensitivity curve like that shown in Figure 11.2(a). The
horizontal axis of this graph shows the frequency of sine waves used in
testing, and the vertical axis shows the intensity of sound that must be pre-
sented at each frequency for the sound to be just detectable. What the figure
shows is that although we are able to detect sounds over a range of frequen-
cies from 20 Hz to 20 kHz, very low frequencies and very high frequencies
must be made very intense for us to detect them. In the middle range of fre-
quencies, especially in the region of 1–3 kHz, our sensitivity is very much
higher.

(a) (b) (c)

120

100
Sound pressure level (dB)

80

60

40

20

–20
31 62.5 125 250 500 1 k 2 k 4 k 8 k 16 k

FIGURE 11.2
(a) The threshold of hearing (b) audiogram of a child aged 10 with normal hearing
(X left ear, O right ear) (c) audiogram of child of similar age with high frequency loss.
0521808820c11_p177_188.qxd 1/27/05 3:08 PM Page 180 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

180 INTRODUCING PHONETIC SCIENCE

If the normal threshold curve is taken as a baseline, one can determine

how far an individual’s hearing differs from normal by using a device
called an audiometer. The results can be plotted as an audiogram. A nor-
mally hearing person will show an audiogram close to the zero line across
the frequency range as in Figure 11.2(b). As people get older, their sensi-
tivity to sound generally decreases, especially at higher frequencies, while
people with seriously impaired hearing often have audiograms that look
very different indeed from the normal curve. You can see an example of
this in Figure 11.2(c) which shows an audiogram for a child with high fre-
quency hearing loss in both ears.
The identification of speech sounds, syllables and words (a process gener-
ally called speech perception) takes place in the cortex of the listener’s brain,
and is influenced not only by the incoming sound but also by the listener’s
experience and linguistic knowledge. The final stage, of understanding
speech and arriving at the intended interpretation, is further guided by the
listener’s general knowledge and expectations in the context.

Visual cues
Although sound is the main input to the process of speech perception, it
is not the only one. If we can see the speaker, we will also see some of the
speech movements that he or she is making, and visual information can
also influence our perception. We are accustomed to the idea that the
deaf can learn to ‘lip-read’, but in fact even those of us with normal hear-
ing make considerable use of visual cues. It is not at all surprising that
humans make use of visual cues to aid them in the perception of speech
when we remember that through most of its history spoken language has
evolved for face-to-face communication.
A simple illustration of our sensitivity to visual information is how
quickly we notice the small failures that sometimes occur in the synchro-
nisation between picture and sound in a film. If we were not attending to
visual cues about speech, and able to integrate them rapidly with auditory
information, we would not notice the mismatch.
It has been known for a long time that visual cues help us to follow
speech in a noisy situation. In one classic experiment, masking noise is
mixed with a recording of speech at such a high level that subjects who
hear only the noisy speech can make little sense of it. Subjects who can see
the talker, but hear nothing, have to rely on lip-reading and are also unable
to follow the message. So neither the sound nor the visual information is
intelligible on its own. But subjects who hear the noisy speech and see the
talker at the same time are able to follow the meaning quite well.
There are obvious limitations on what can be gathered about speech
sounds from visual information alone. The lips are the most visible articu-
lators. One can see if they are closed or open, and whether they are round-
ed or spread. If the lips are open, one can see a limited amount of what the
tongue is doing in the dental/alveolar region. But back articulations such
as velars are virtually impossible to see. The degree of jaw opening can also
0521808820c11_p177_188.qxd 1/27/05 3:08 PM Page 181 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Speaker and hearer 181

be followed visually, and this has implications for tongue height and pos-
ture and can help us make judgements about both vowels and consonants.
Even expert lip-readers can’t see things that are invisible. One can’t see
the difference between voiced and voiceless consonants which are other-
wise alike, because the vocal folds are out of sight within the larynx.
Similarly, moving the soft palate up and down has no visible effect, so
there’s no reliable way of lip-reading the difference between a nasal con-
sonant such as [m] and the corresponding plosives [b] or [p]. However, expe-
rienced lip-readers are often able to use the cues that are visible together
with information that can be inferred from the context of the conversa-
tion to achieve very good understanding of speech.
Another illustration of our sensitivity to visual information in speech
comes from what is generally called the McGurk effect, after the psychol-
ogist who first reported it. In this, experimental subjects watch a film of
a talker producing a certain syllable, which has been dubbed with the
wrong soundtrack. For example, the film might show the production of
[
a] while the soundtrack contains the syllable [ba]. Under these condi-
tions, many subjects report hearing the syllable [da]. The conflicting visu-
al and auditory cues lead them to ‘hear’ a consonant somewhere between
[ba] and [
a]. Their perception changes, however, if they close their eyes so
that the visual information does not influence their perception.

Speech perception tests

How do we know what acoustic events give rise to the perception of partic-
ular speech sounds or phonetic distinctions? There has been a long tradi-
tion of using synthetic speech, that is, an electronic simulation of human
speech, to present carefully controlled speech-like stimuli to listeners in
order to obtain simple judgements on their perception. Synthetic speech is
used because it would be extremely difficult for a human speaker to control
the acoustic properties of speech with the required degree of accuracy.
Suppose we are interested in how voice onset time (VOT) is related to the
perception of voicing in English plosives. We need a series of stimuli (known
as a continuum) where the VOT of a plosive is increased in equal steps. It
would be no use asking a human being to say a word like bay and then to
repeat it with the VOT for the initial plosive exactly 10 ms longer and then
again but with the VOT 10ms longer and so on. We do not have conscious
control of our speech organs to that degree of accuracy. Using synthetic
speech, our first step is to make measurements of spectrograms of a real
utterance, [be] say, and to use these measurements to produce a synthetic
version of this syllable. From this we could produce a series of test stimuli
where VOT varies in steps of 10 ms from 0 to 50 ms. The stimuli would then
be used to make up a perception test, where many repetitions of the stim-
uli are presented in random order to a group of subjects. Such a test is called
a labelling experiment. The subjects are asked to make the simple judge-
ment of whether they hear [be] or [pe] each time they hear a test item. The
aim of such a test is to discover the VOT value at which the subjects’
0521808820c11_p177_188.qxd 1/27/05 3:08 PM Page 182 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

182 INTRODUCING PHONETIC SCIENCE

perception switches from [be] to [pe]. This is done by plotting a graph of

the percentage of [be] responses for each VOT value. Such a graph is known
as a labelling curve (see Figure 11.3). For this (imaginary) group of subjects
[ b] gives way to [ p] at about 30 ms, which is where the curve crosses the 50%
level. Stimuli with a VOT below 30 ms are all judged to be instances of [be],
while those with a VOT above 30 ms are all perceived as [pe].

FIGURE 11.3
Labelling curve for
synthetic [beı] versus [peı]
stimuli.

Another sort of test we could do with these stimuli is to present pairs of

adjacent VOT values, 0 and 10 ms, 10 and 20 ms and so on, and ask for
same/different judgements from our subjects. We would then obtain what
is known as a discrimination curve, which records the percentage of ‘dif-
ferent’ responses for each stimulus pair. The discrimination curve for our
imaginary experiment is in Figure 11.4. There is a peak in the curve at the
30/40 ms pair of stimuli.

FIGURE 11.4
Discrimination function
(two-step) for synthetic
[beı] versus [peı]
continuum.

This result together with the sharp cross over from one percept to the
other in the labelling test suggests that our subjects are perceiving VOT
categorically. This means that there are two perceptually well-defined
areas in the VOT continuum without any intermediate percepts and that
within each area subjects are unable to tell the difference between
0521808820c11_p177_188.qxd 1/27/05 3:08 PM Page 183 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Speaker and hearer 183

acoustically different events. Many other acoustic cues have been investi-
gated in this way, including: burst frequency for place of articulation in
plosives, direction of formant transitions for place of articulation of con-
sonants and the relationship between F1 and F2 frequencies for vowel
perception.

Cross-language comparison of cue perception

If we extended our VOT continuum to include negative values, say from

30 ms to 50 ms, we could investigate the voiced/voiceless threshold for
plosives in a language like French. French voiceless plosives are not aspi-
rated and have a short or zero VOT value. Negative VOT values mean
that the vocal folds start vibrating before the release of the plosive. From
this we can conclude that different languages, while using the same
acoustic cue to signal a phonetic distinction, may use the cue in differ-
ent ways. So, for instance, plosives with a VOT in the region of 0–20 ms
are categorised as voiced by English hearers, but as voiceless by French
hearers.

Cue redundancy
It is rare for a phonetic distinction to be signalled by a single acoustic cue.
For instance, place of articulation for plosives is signalled simultaneously
by the second formant transition (the shape of the second formant at the
beginning of a following vowel) and also by the frequency of the plosive
burst itself. In Figure 11.5 we can see these patterns in the forms of
schematic spectrograms for the syllables [ ba da
a].

FIGURE 11.5
Schematic spectrogram
for [ba da
a].

For [ ba] the burst is relatively low in frequency and the F2 transition is ris-
ing. For [da] the burst is high in frequency and the F2 transition is almost
flat. [
a] has a falling F2 transition and a burst frequency between those
for [ ba] and [da].
This situation is known as cue redundancy – the presence of more infor-
mation than is strictly necessary to distinguish phonetic events. Given per-
fect signal transmission, and perfect decoding, burst frequency alone, or F2
0521808820c11_p177_188.qxd 1/27/05 3:08 PM Page 184 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

184 INTRODUCING PHONETIC SCIENCE

transition alone, should be enough. However, we rarely hear speech in per-

fect surroundings, and the redundancy of cues provides a safeguard against
the effects of environmental noise and errors in processing. For example,
noise in the situation might interfere with the perception of the bursts, but
leave the formant transitions still audible. Speech has built into it a degree
of resistance to interference by changes in listening conditions.

Speech development
In the normal course of events, a child acquires much of its first language
in the first three or four years of life and is able to communicate with
other children, and with unfamiliar adults, by the time she or he goes to
school. Of course, it is not uncommon for a child to be brought up in an
environment where two (or even more) languages are being used, but for
our purposes it will be simpler to consider the case of a child with normal
cognitive and sensory abilities growing up in a monolingual environ-
ment. Obviously the child has to acquire the vocabulary and grammar of
the language in question, but along with this, even before the child’s first
words can be uttered, must come development of the ability to identify
and discriminate the sounds of the language, and to produce them.
It is usual for perception (in the sense of sound identification and dis-
crimination) to be ahead of production in speech development. This
means that children can often hear sound differences that they are not
yet able to reproduce. A child learning English, for example, has to cope
with a large system of fricatives, and the child’s own output will certainly
not at first contain all of [f θ s ʃ h] (to say nothing of the voicing difference
which adds [v ð z ]). The child’s own speech will at first contain only one
or two of these fricatives (often [f ] and [h] are found early), and the dis-
tinction between [s] and [ ʃ ] may not appear until relatively late. But there
is plenty of evidence that the child can hear and identify at least some of
the sounds that are not yet present in her/his output. In a famous exam-
ple, a child at a certain stage of development produced [s] for both [s] and
[ ʃ ], saying [fs] for fish. However, when an adult spoke the form [fs] back
to the child, the child rejected the pronunciation. This indicates clearly
that the child could perceive the difference between the forms [fs] and [fʃ ]
but could not yet manage the distinction in production.
In general then, childlike speech is probably not the direct result of limit-
ed perceptual abilities. Whatever language they are acquiring, children
begin with a simpler sound system than the adult’s, adding sounds gradual-
ly as more and more difficult distinctions are mastered. It is not only a ques-
tion of smaller inventories of contrasting sounds. Children also commonly
simplify the phonological structures (syllables and words) they produce. For
example, children acquiring English generally go through a stage of weak
syllable deletion, in which words that are meant to have two, three or
more syllables are simplified so as to contain fewer syllables. So the adult
three-syllable word banana may become [nɑnə]. The components of syllables
can also be simplified. Consonant clusters, which occur in adult English as
0521808820c11_p177_188.qxd 1/27/05 3:08 PM Page 185 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Speaker and hearer 185

onsets and codas to syllables, are commonly reduced. So for instance a word
such as play, which begins with the cluster [ pl] for an adult, may be just [pe]
for the child, and therefore indistinguishable from the same child’s version
of pay.

Labelling experiments with infants

Labelling experiments of the type described earlier have been performed
with very young children in order to investigate the development of per-
ceptual abilities. Of course, it is impossible to ask a baby only a few
months old to tell you what they hear, so an alternative strategy has to be
used. One such strategy involves playing a stimulus repeatedly to a child
and observing the child’s movement. After a few repetitions of a stimulus
the baby will lose interest. If the stimulus is then altered and the child
does not react, by for instance turning the head, one can conclude that
the difference between the two stimuli is not perceptible. If, however, the
child does react when the stimulus is changed, the conclusion is that the
two stimuli lie on either side of a perceptual threshold for the child.

Hearing impairments
The discussion of speech perception, and of speech development, has so
far concerned those with normal hearing. Hearing impairments can occur
at various levels in the hearing mechanism. The outer ear may be tem-
porarily blocked by wax. The middle ear can be affected by infection, or
the ossicles damaged (often surgical repair is possible). But if the cochlea
itself is not functioning properly the result is a permanent impairment of
the ability to detect and discriminate sounds.
Different kinds of hearing loss will affect speech perception in different
ways. Quite commonly, a person may have a high-frequency loss, in which
higher frequencies (such as those found in fricatives and plosive bursts)
are difficult to detect, while lower frequencies (such as found in the lower
formants of voiced sounds), are detected quite well. A person who can hear
only the very lowest frequencies may be able to respond to pitch informa-
tion (tone and intonation), but hear only the first formants of vowels, so
that [i] and [u], which have similar first formant frequencies, will be con-
fused (obviously lip reading will help, if the speaker is visible).
The popular conception of deafness as loss of sensitivity to sound is only
partly correct. One function of most hearing aids is indeed to amplify sounds
so that they are easier to detect. But hearing impairment generally also
involves some loss of the ability to discriminate sounds (to tell apart sounds
of differing time and frequency composition), and hence affects speech per-
ception directly, even if amplification makes the speech loud enough to be
detected. Most hearing aids amplify and process speech so that a person can
make best use of their residual hearing, but one development in the last thirty
years or so has been the cochlea implant, in which one or more electrodes
are introduced into a non-functioning cochlea, and electrical stimulation of
the electrodes gives the otherwise entirely deaf person a degree of hearing.
0521808820c11_p177_188.qxd 1/27/05 3:08 PM Page 186 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

186 INTRODUCING PHONETIC SCIENCE

The effect of hearing impairment on a person’s own speech depends on

when the hearing loss begins and on its severity. Those who become deaf
after speech is established generally retain the ability to speak very intelli-
gibly, whereas a child who is congenitally deaf will have severe difficulty
acquiring speech, and in monitoring his or her own attempts at speaking.

Labelling experiments with hearing-impaired subjects

Labelling experiments can also be used to investigate how people with
hearing impairment perceive (or fail to perceive) speech. As an example of
such an experiment, let’s look at the results for a profoundly hearing-
impaired fourteen-year-old subject in response to stimuli that range from
an extreme rise in pitch on the syllable [əυ] to an extreme fall in pitch on
the same syllable. The subject was asked ‘Do you hear a question or not?’
The continuum consists of nine steps and the fifth point on the continu-
um is a stimulus with level pitch, as can be seen in Figure 11.6a. The
results are shown in Figure 11.6b and c.
From a normally hearing subject one would expect a pretty high rate of
question judgements for stimuli 1–4, all of which have rising pitch move-
ments. Stimulus 5 should produce a response rate of about 50% and stimuli
6–9 should be judged as non-questions. However, we can see that for this
subject most of the stimuli produce responses around 50%, indicating that
subject’s hearing loss is such that the responses are essentially random.

FIGURE 11.6
(a) Stimulus continuum (rising versus falling fundamental frequency), (b) Question/
Statement labelling curve for a typical normal subject, (c) labelling curve for a child with
severe hearing loss. After A. J. Fourcin, ‘Speech pattern tests for deaf children’, Speech
and Hearing: Work in Progress, 1976, 46–62.
0521808820c11_p177_188.qxd 1/27/05 3:08 PM Page 187 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Speaker and hearer 187

Chapter summary
The main aim of this chapter has been to provide a survey of the way
in which the phonetic concepts introduced throughout the book can
be applied to the investigation of speech and the areas in which
phonetic knowledge is useful and important. Much of our book has
been concerned with how speech is produced from the articulatory
point of view and with the acoustic consequences of the movements of
the speech organs. However, human speech is used for communicating
information to a listener, so in this chapter we have looked at speech
from the receiving end and have outlined ways in which human
perception of speech may be investigated. We have pointed out the
main characteristics of the hearing mechanism, shown how auditory
sensitivity can be measured, and looked at some types of hearing
impairment and their effect on speech perception We have seen that
visual cues can be used to aid speech perception and understanding.
We have also looked briefly at the way in which the ability to produce
and perceive speech develops in children.

Further reading
A very useful book covering hearing, speech acoustics and speech perception is Peter Denes and Elliot
Pinson, The speech chain, New York: W.H. Freeman, 1993. If you wish to know about how children develop
speech and what can go wrong with this process, see Pamela Grunwell, Clinical phonology, London: Croom
Helm, 2nd edn, 1987 and David Ingram, Phonological disability in children, London: Arnold, 1976. For all
aspects of the measurement of speech, see Katrina Hayward, Experimental phonetics, Harlow: Longman,
2000. For a quantitative introduction to the interrelationships between production, acoustics and perception,
see Keith Johnson, Acoustic and auditory phonetics, Oxford: Blackwell, 2nd edn, 2003. On hearing and
deafness, see John Graham and Mike Martin, Ballantyne’s deafness, London: Whurr 6th edn, 2001. For more
detail on the human hearing mechanism and its role in speech perception you should look at Brian Moore,
An introduction to the psychology of hearing, London: Academic Press, 3rd edn, 2003.
0521808820c11_p177_188.qxd 1/27/05 3:08 PM Page 188 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:
0521808820cglos_p189_200.qxd 1/27/05 3:09 PM Page 189 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Glossary
Accented syllable A syllable made prominent by pitch (commonly a pitch peak). In
English intonation, a stressed syllable made prominent by being
accompanied by the start of a new trend in the pitch contour of the
utterance.
Active articulator The articulator that moves to form a constriction in the vocal tract.
For example, in palatal sounds the active articulator is the front of
the tongue which moves towards the hard palate.
Advanced Produced slightly further forward in the vocal tract. For instance, the
first consonant in the English word keen is an advanced velar and is
produced with a closure towards the front of the velum.
Affricate A consonant sound produced with a complete closure between two
articulators and with velic closure preventing air escaping via the
nasal cavity. The articulators part relatively slowly and the com-
pressed air escaping between them becomes turbulent resulting in
audible friction noise. [ts d
kx] are examples of affricates.
Airstream One of a number of ways in which airflow can be created in the vocal
mechanism tract for the purpose of producing speech sounds.
Allophone A speech sound considered as a positional variant of a phoneme. So,
for instance, in Spanish the two sounds [b] and [β] are allophones of
the phoneme /b/. The bilabial fricative [β] occurs word-internally
between vowels and the plosive [b] occurs in other environments.
Allotone A pitch configuration considered as a positional variant of a toneme.
For example, Modern Standard Chinese Tone 3 has a terminal rise in
pitch when found before a pause, but lacks this terminal rise when
not before a pause. These two pitch configurations can be regarded as
allotones of the Tone 3 toneme.
Alphabetic writing A system of writing, like that used for English, where the symbols used
system represent the individual vowels and consonants of the language.
Alternation The situation where a word or word-part turns up in different phono-
logical forms in different environments. For example, the English
plural ending s is /s/ if the preceding sound is voiceless, as in books
/bυks/, but it is /z/ if the preceding sound is voiced, as in dogs /dɑz/.
Alveolar The name of a place of articulation. The active articulator is the tip or
blade of the tongue. The passive articulator is the alveolar ridge. [t s n]
are examples of alveolar sounds.
Alveolar ridge The front part of the roof of the mouth, just behind the upper front teeth.
Ambient noise Noise in the environment which may interfere with the recording of
sounds.
Amplitude A measure of the up-and-down extent of a waveform (as distinct from
its frequency). For a sound wave, the extent of pressure variation
above or below atmospheric pressure.
Aperiodic Of a waveform, one that does not have a regular repeating pattern.
0521808820cglos_p189_200.qxd 1/27/05 3:09 PM Page 190 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

190 Glossary

Approximant A consonant sound made with a constriction between two articulators

which is not narrow enough to cause air turbulence. [w l j] are
examples of approximants.
Articulation A constriction of the vocal tract.
Articulators The parts of the vocal tract involved in the production of speech
sounds: the lips, teeth, tongue, alveolar ridge, hard palate, soft palate,
uvula, pharyngeal wall and vocal folds.
Arytenoid cartilages Two small cone-shaped cartilages that sit on the upper surface of the
cricoid cartilage. The vocal folds are attached at the back to the ary-
tenoids and the positioning of the cartilages is largely responsible for
the position of and tension of the folds.
Aspiration Weak friction noise heard on the release of a plosive sound. The noise is
caused by air moving at high speed through the vocal tract. Aspiration
is associated with plosives that have a long Voice Onset Time.
Assimilation A phonological process that involves a change in a speech sound to
make it more similar to some sound in its environment. An example
from English: the word ten which is /ten/ in most environments may
become /tem/ if the following word begins with a bilabial consonant,
as for example in the phrase ten books, which may be pronounced
/tem bυks/.
Audiogram A graph showing the sensitivity of a human subject to sounds at various
frequencies.
Audiometer A device capable of producing tones of precise frequency and intensity
which is used to test human sensitivity to sound at various frequencies.
Auditory canal The channel that leads from the outer part of the ear to the eardrum.
Back (of tongue) The part of the tongue that lies below the soft palate when the
tongue is at rest.
Bilabial The name of a place of articulation. The articulators are the upper
and lower lips. [p b m] are examples of bilabial sounds.
Binary Of phonological features, taking one of two opposite values. So for
example, speech sounds may be classified as [
nasal] if they are made
with the velum lowered or [nasal] if not.
Blade (of tongue) The part of the tongue that lies below the alveolar ridge when the
tongue is at rest.
Breathy voice A type of phonation in which the pattern of vocal fold vibration
allows the escape of relatively large amounts of air in each cycle of
vibration, producing audible noise along with voicing.
Cardinal vowels A set of agreed vowel qualities used as a reference for the purposes of
describing vowels encountered in speech.
Categorical A characteristic of the perception of certain speech sounds. Sounds
perception are said to be perceived categorically if there is a sharp cross-over
from one perceptual category to another and if, in addition, human
listeners are unable to distinguish between acoustically different
sounds that fall in the same category.
Click A sound produced with an ingressive velaric airstream mechanism.
Coarticulation The adjustments made to the articulation of a speech sound under the
influence of neighbouring sounds. For instance, the first consonant in
the English word queen [kwin] is likely to be produced with rounded
lips because it is followed immediately by a lip-rounded sound.
0521808820cglos_p189_200.qxd 1/27/05 3:09 PM Page 191 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Glossary 191

Cochlea The organ of hearing. A spiral structure in the inner ear where
mechanical vibrations are converted to nerve impulses.
Coda A syllable constituent consisting of any consonant sounds following
the syllable nucleus.
Complementary Two sounds are said to be in complementary distribution in a particular
distribution language if they can never appear in the same environment as each
other. For example, in English [h] can only appear in the onset of a
syllable and never in the syllable coda, whereas [ŋ] can only appear in
the syllable coda and never in the syllable onset. So in English [h] and
[ŋ] are in complementary distribution.
Complex periodic A sound, such as a vowel sound, whose waveform can be analysed as the
tone sum of two or more sine waves.
Compression The rise in air pressure in an enclosed space caused by a decrease in
the size of the space without outflow of air.
Consonant Sounds made with a relatively close constriction or complete closure in
the vocal tract and that occur singly or in clusters at the edges of syllables.
Context-sensitive A phenomenon where the voicing of consonant sounds is determined
voicing by the context in which they appear. For example, young children
often go through a stage in the development of their speech when all
obstruent consonants are always voiced if they are immediately
followed by a vowel and are always voiceless in other contexts.
Continuum A series of synthetic speech tokens used as stimuli in a speech
perception experiment. Each member of the continuum is acoustically
identical to the others except in the value of the particular acoustic
feature under investigation.
Contour tone A lexical tone that changes in pitch.
Contrast The situation where a phonetic difference is capable of signalling a
difference between words in a particular language. So, for example,
vowel nasalisation or lack of it in French can change the identity
of a word, as can be seen with the pair of words mot /mo/ ‘word’ and
mon /mo ~/ ‘my’. So there is a contrast between oral and nasalised vow-
els in French. English, while it has oral and nasalised vowels, never
uses the difference to signal the difference between words.
Creaky voice A type of phonation in which the vocal folds vibrate at a low frequen-
cy (and usually somewhat irregularly) with a very low rate of airflow
through the glottis.
Cricoid cartilage A ring-shaped cartilage at the top of the windpipe. It is attached to
the thyroid cartilage; the arytenoid cartilages sit on its upper surface
at the rear.
Cue redundancy The presence of more acoustic cues than are logically necessary to
signal a perceptual distinction between speech sounds.
CV-skeleton A representation of the structure of a word or phrase in terms of the
sequence of consonant and vowel sounds it contains. C is used to
represent consonants and V to represent vowels. For instance, CVCV is
the CV-skeleton for English words such as below, city, data.
Defective vowel A type of vowel system, found in a small number of languages, that does
system not conform to the Vowel Dispersion Principle, either because it lacks
open vowels or because, for mid and high vowels, there is an asymmetry
of front and back vowels.
0521808820cglos_p189_200.qxd 1/27/05 3:09 PM Page 192 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

192 Glossary

Degree of stricture The narrowness of the space between two articulators during the
production of a sound.
Dental The name of a place of articulation. The active articulator is the tip or
blade of the tongue. The passive articulator is the upper front teeth.
[θ ð] are examples of dental sounds.
Devoiced Of a sound normally voiced, produced without vocal fold vibration for
part or all of its duration.
Diacritic A small mark placed near or attached to a phonetic symbol which is
used to modify the usual meaning of the symbol or to supply further
phonetic detail. For instance, [m] represents a voiceless bilabial nasal
whereas the symbol alone without the diacritic represents a voiced
bilabial nasal.
Diphthong A vowel that changes quality within a single syllable. An example of a
diphthong in English is [a], in such words as fine, time, sight.
Discrimination A graph showing how well human listeners can distinguish between
curve adjacent stimuli on an acoustic continuum.
Dissimilation The opposite of assimilation, that is a change in a sound to make it
more dissimilar from some sound in its environment.
Distribution The environments in which a speech sound may occur in a particular
language.
Double articulation An articulation in which there are two simultaneous constrictions of
the vocal tract. The two constrictions are of equal narrowness. [w] is
an example of a double articulation.
Eardrum A membrane stretched across the auditory canal and dividing the outer
ear from the middle ear. The eardrum converts pressure variation into
mechanical movement.
Egressive Of an airstream mechanism, one that pushes air out of the vocal tract.
Ejective A sound produced with an egressive glottalic airstream mechanism.
Electropalatography An instrumental technique for investigating the pattern of contact of
the tongue with the roof of the mouth during speech sounds. Tongue
contact is recorded by electrodes embedded in an artificial palate
worn by the speaker.
Elision The omission of a sound. For example, the voiceless alveolar plosive
phoneme may be elided in a phrase like last month, giving the
pronunciation /lɑs m nθ/ rather than /lɑst m nθ/.
Ending Another term for suffix. Sounds (or letters in the written form)
added to the end of a word to indicate verb tense, the plural of
nouns and the like. For example, the regular plural ending for
English nouns is one of /s z z/, the choice depending on the last
sound of the noun concerned.
Environment The environment of a sound is the preceding and following sound or
sounds. Statements of environments may also include the symbol #,
which means a word boundary. For example, in a word like bring
[bɹŋ], the environment of [ɹ] is [b] [] and that of [ŋ] is [] #.
Excitation See input.
Feature See phonological feature.
Fibrescope A flexible bundle of light-transmitting fibres used to collect an image
from inaccessible areas inside the human body.
0521808820cglos_p189_200.qxd 1/27/05 3:09 PM Page 193 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Glossary 193

Fixed stress A language where the position of the primary stress is the same for the
language vast majority of words. For example, Polish is a fixed stress language
because nearly all words have primary stress on the penultimate syllable.
Flap A consonant produced by making the active articulator strike the
passive articulator in passing. [ ] is an example of a flap
Formant A peak in the spectrum of a speech sound such as a vowel.
Formant frequency The frequency at which a peak in energy occurs in the spectrum of a
speech sound.
Fortis Produced with increased muscular tension. Fortis speech sounds are
normally voiceless.
Fortition A phonological process, the opposite of lenition, which involves a
change from a weaker sound to a stronger one that has a more radical
obstruction to airflow.
Free stress The same as variable stress.
Frequency The rate of repetition of a cycle of vibration. Frequency is measured in
Hertz (Hz). 1 Hz is one cycle per second, so, for example, if a periodic
waveform repeats 100 times in a second, its frequency is 100 Hz.
Fricative A consonant sound produced with a narrow constriction of the vocal
tract which causes the airstream to become turbulent, resulting in
audible friction noise. [s
γ] are examples of fricatives.
Frictionless continuant An older term for median approximant.
Front (of tongue) The part of the tongue that lies below the hard palate when the
tongue is at rest.
Glide An older term for median approximant.
Glottal The name of a place of articulation. The articulators are the vocal
folds. [ ʔ h] are examples of glottal sounds.
Glottal stop The sound symbolised [ ʔ]. It is produced by closing the vocal folds
tightly, blocking the airstream from the lungs.
Glottalic The name of an airstream mechanism in which the closed vocal folds
inititate the airflow. Ejective consonants are produced with an egressive
glottalic airstream and implosives with an ingressive one.
Glottis The space between the vocal folds.
Heavy syllable A syllable that contains a long vowel or diphthong, or alternatively a
short vowel followed by more than one consonant.
Height (of vowels) An auditory property of vowels, corresponding with the degree of
raising of the highest point of the tongue during the production of a
vowel.
Homophone One of a pair (or larger set) of words that sound exactly the same
when spoken, but which mean different things. For example, the
words write, right, rite and wright are homophones in English, all being
pronounced [rat].
Homorganic Two sounds are homorganic if they have the same place of articula-
tion. For example, [s] and [l] are both alveolar, so they are homorganic.
Implosive A sound produced with an ingressive glottalic airstream mechanism.
Impressionistic A transcription of speech that represents its superficial auditory effect,
transcription but is not made in accordance with a stable and economical system of
symbols worked out to suit the phonological structure of the language
involved.
0521808820cglos_p189_200.qxd 1/27/05 3:09 PM Page 194 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

194 Glossary

Infrasound Low frequency vibration below the lower frequency limits of normal
human hearing.
Ingressive Of an airstream, one that pulls air into the vocal tract.
Initiator The part of the vocal tract that moves in order to create a rise or fall
in air pressure, resulting in flow out of or into the vocal tract. For
instance, in the pulmonic airstream mechanism the initiator is the
walls of the lungs.
Input The energy introduced into a resonating system. An alternative term
is ‘excitation’. The input to the vocal tract is essentially of two kinds:
(1) periodic energy produced by the vibration of the vocal folds, (2)
aperiodic energy produced by air turbulence caused by a narrow
constriction or complete closure between two articulators. Some
speech sounds, voiced fricatives especially, are produced with a
combination of these two sorts of input.
International A set of internationally agreed symbols used for representing speech
Phonetic Alphabet sounds.
Intonational phrase A group of words accompanied by a complete well-formed intonation
pattern.
IPA The International Phonetic Association or The International Phonetic
Alphabet.
Key An intonational feature that affects the pitch characteristics of a
whole intonational phrase. If an intonational phrase is produced with
high key, then all of the pitch values are higher. Similarly, low key
makes all pitch values in the phrase lower.
Labelling curve A graph showing the results of a labelling experiment. The horizontal
axis shows the points in the acoustic continuum under investigation.
The vertical axis shows the percentage responses for one of the two
response labels for the experiment.
Labelling A speech perception experiment where large numbers of synthetic
experiment speech tokens are presented to human subjects. The subjects’ task is to
respond to each token with one of two labels. For example, the sub-
jects might be asked whether they hear each token as the word coat or
as the word goat. The stimuli for the experiment are taken from an
acoustic continuum of stimuli, but are presented in random order.
Labialisation A secondary articulation involving the rounding of the lips.
Labial-palatal The name of a double articulation with simultaneous articulations at
the lips and at the hard palate. [ɥ] is an example of a labial-palatal sound.
Labial-velar The name of a double articulation with simultaneous articulations at
the lips and at the velum. [w] is an example of a labial-velar sound.
Labiodental The name of a place of articulation. The active articulator is the lower
lip. The passive articulator is the upper front teeth. [f v] are examples
of labiodental sounds.
Laminar flow A characteristic of flow in a fluid such as air or water where the
particles that make up the fluid move in parallel paths and there
are few if any collisions between the particles. Laminar airflow does
not create any sound.
Laryngograph An electronic device that records vocal fold vibration by means of
two electrodes placed externally on a speaker’s neck. A small electric
0521808820cglos_p189_200.qxd 1/27/05 3:09 PM Page 195 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Glossary 195

current is passed through the neck and the laryngograph measures

the changing resistance of the neck to the passage of the current.
When the vocal folds are in contact, the resistance is lower than
when they are apart.
Larynx A structure made of cartilage and connective tissue at the lower end
of the vocal tract and above the windpipe, containing the vocal folds.
Lateral Of an approximant or fricative, produced with a complete closure on
the midline of the vocal tract, but with one or both sides of the
tongue lowered and not contacting the side teeth or gums, so that the
air escapes over the sides of the tongue. [l ] are examples of lateral
sounds.
Lenis Produced with reduced muscular tension. Lenis speech sounds are
usually voiced.
Lenition A phonological process involving the change from a stronger sound to
a weaker one, where stronger sounds are defined as having a more
radical obstruction to airflow than weaker ones. For example, the
change from a plosive to a fricative in certain environments is lenition
and so is the change from a fricative to an approximant. In addition,
voiceless sounds are stronger than voiced ones.
Lexical tone The use of a small number of contrasting pitch patterns to distin-
guish words from one another.
Light syllable A syllable that contains a short vowel followed by a maximum of one
consonant.
Location (of vowels) The part of the tongue (front, centre or back) that is raised highest in
the oral cavity for the production of a vowel sound.
Logographic A writing system, such as that for Chinese, where the symbols used
writing system represent whole words, rather than the sounds or syllables that make
up the word.
Long vowel A vowel of relatively long duration when compared to a vowel of similar
or identical quality in the same vowel system.
Loudness The subjective impression of the magnitude of a sound. Loudness is
connected with the amplitude of the waveform of the sound.
Magnetic resonance A non-hazardous technique for producing images of the interior of the
imaging body. The technique involves applying a very strong magnetic field to
the body, followed by a radio-frequency pulse specific to hydrogen.
Manner of The way in which the articulators interfere with and direct the airstream
articulation for the purposes of producing speech sounds. Manner of articulation
is a complex of features such as degree of stricture, speed of articula-
tor movement, soft palate position and the like.
McGurk effect A perceptual effect demonstrating that visual cues influence speech
perception. A video of the speaker’s face is overdubbed with the
soundtrack of an utterance different from the one that the viewer is
seeing. The visual and auditory cues may be integrated by the hearer,
and the resulting perception can differ from both of the speaker’s
utterances.
Meatus Another name for the auditory canal.
Median Of a fricative or approximant sound, articulated in such a way that
the air escapes down the midline of the vocal tract.
0521808820cglos_p189_200.qxd 1/27/05 3:09 PM Page 196 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

196 Glossary

Minimal pair A pair of words that differ in only one phoneme. An example from
French: peau /po/ ‘skin’ and beau /bo/ ‘beautiful’.
Modal voice A phonation type in which the vocal folds snap shut rapidly and peel
apart relatively slowly. Most speech is produced with modal voice.
Monophthong A vowel sound that does not change in quality.
Monosyllable A word consisting of a single syllable, such as do and snap in English.
Nasal (stop) A consonant sound produced with a complete closure in the mouth
but without velic closure so that the airstream escapes only through
the nasal cavity. [m  ŋ] are examples of nasals.
Nasal cavity The large cavity above the roof of the mouth, connected to the upper
part of the pharynx at the rear and having the nostrils at the front.
Nasalised vowel A vowel sound produced without velic closure so that air escapes
simultaneously through the oral cavity and the nasal cavity.
Neutralisation The suspension, in certain environments, of a phonetic contrast nor-
mally found in a language. For example, voicing is contrastive for
obstruent consonants in German, except at the end of words, where
only voiceless consonants are permitted.
Non-pulmonic Of airstream mechanisms, not involving the lungs as an initiator.
The two non-pulmonic airstream mechanisms used for speech sounds
are glottalic and velaric.
Nuclear tone One of a small number of pitch configurations associated with the
nucleus and any following syllables in the intonational phrase.
Nucleus (1) The syllable that bears the last accent in an intonational phrase.
(2) The only obligatory component of a syllable.
Obstruent A sound that is produced with a constriction narrow enough to cause
an appreciable rise in air pressure inside the vocal tract. Plosives,
affricates and fricatives are all obstruents.
Octave An interval between two musical notes in a scale. If the frequency of
any note is doubled, the result is the same note, but one octave higher.
Onset A syllable constituent consisting of any consonant sounds preceding
the syllable nucleus.
Oral A sound produced with velic closure to prevent nasal escape of air, so
that the airstream escapes through the oral cavity alone.
Oral cavity The mouth. The cavity bounded by the lips at the front and joined to
the pharynx at the rear.
Oral vowel A vowel sound where all of the air escapes via the oral cavity and
where there is no nasal airflow.
Ossicles The three small bones of the middle ear which act as a linkage
between the eardum and the oval window.
Oval window An opening into the cochlea, covered by a membrane, to which
mechanical movement is transmitted from the eardrum via the ossicles.
Palatal The name of a place of articulation. The active articulator is the front
of the tongue. The passive articulator is the hard palate. [j ] are
examples of palatal sounds.
Palatalisation A secondary articulation in which the front of the tongue is raised
towards the hard palate.
Palatogram A record of the pattern of contact between the tongue and the roof of
the mouth during the production of a speech sound.
0521808820cglos_p189_200.qxd 1/27/05 3:09 PM Page 197 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Glossary 197

Paralinguistic A feature of speech, such as loudness or voice quality, used to convey

feature speaker attitude rather than linguistic meaning.
Parametric diagram A diagram showing the estimated movements or activity of various
parts of the vocal tract such as the soft palate and the vocal folds.
Passive articulator The stationary articulator involved in forming a constriction in the
vocal tract. For instance, the hard palate is the passive articulator for
both palatal and retroflex sounds.
Periodic Of a waveform, regularly repeating in time.
Pharyngeal The name of a place of articulation. The active articulator is the root
of the tongue. The passive articulator is the rear wall of the pharynx.
[ ʕ] are examples of pharyngeal sounds.
Pharyngealisation A secondary articulation in which the root of the tongue is retracted
and approaches the rear wall of the pharynx.
Pharynx The part of the vocal tract immediately above the larynx.
Phonation The generation of voice in the larynx, by vibration of the vocal folds.
Phoneme A set of speech sounds that form a basic contrastive unit in a lan-
guage. For example in English, the set which contains the sounds [l  l̃
˜ ], amongst others, is the /l/ phoneme. The members (or allophones) do
not contrast with one another in the language in question.
Phonetic symbol A symbol representing a speech sound.
Phonetic transcription The representation of speech using phonetic symbols.
Phonological feature One of a small set of labels used to specify an aspect of the articulation
(or of the acoustic properties) of a speech sound or a set of speech
sounds. For example, the phonological feature [anterior] is used to
distinguish those sounds produced on or in front of the alveolar ridge
from those sounds produced further back in the vocal tract.
Phonological process A widespread phonological phenomenon such as assimilation, elision
or lenition.
Phonotactics The patterning of permissible phoneme sequences in syllables.
Pinna The outer, visible part of the ear.
Pitch The perceptual attribute of a sound that enables the hearer to locate
the sound on a scale from high to low. The physical correlate of pitch
is the frequency of the sound.
Place of articulation The point within the vocal tract where a constriction is formed to
produce a consonant sound. The place of articulation is defined by
the active and passive articulators involved.
Plosive A consonant sound produced with a complete closure between two
articulators and with velic closure preventing air escaping via the nasal
cavity. The articulators part rapidly allowing rapid release of the
compressed air without any accompanying friction noise. [p d c q] are
examples of plosive sounds.
Postalveolar The name of a place of articulation. The active articulator is the tip or
blade of the tongue. The passive articulator is the rear of the alveolar
ridge. [ɹ ʃ
] are examples of postalveolar sounds.
Pre-fortis clipping The reduction in duration of sonorant sounds when followed in
the same syllable by a fortis consonant. For example, the vowel
in the English word seat [sit] is clipped, whereas that in seed [sid]
is not.
0521808820cglos_p189_200.qxd 1/27/05 3:09 PM Page 198 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

198 Glossary

Pressure The molecules that make up a gas such as air are constantly moving
and colliding with the surface of objects that contain the gas. The
force created by the sum of these collisions is the pressure of the gas.
Pressure is measured in force per unit area.
Primary articulation The narrowest constriction in the vocal tract during the production of
a speech sound.
Primary stress See stress.
Prosodic The same as suprasegmental.
Pulmonic The name of an airstream mechanism in which the lungs initiate the
airflow. The majority of sounds of human speech are produced with a
pulmonic egressive airstream.
Pure tone The sound associated with a sine wave.
Rarefaction The fall in air pressure in an enclosed space caused by an increase in
the size of the space without inflow of air.
Register tone A lexical tone with a level pitch contour.
Resonant frequency The frequency at which an acoustic system vibrates when excited by
input energy.
Retracted Produced slightly further back in the vocal tract. For instance, the
first consonant in the English word trip is a retracted alveolar and is
produced at the back of the alveolar ridge.
Retroflex The name of a place of articulation. The active articulator is the tip of
the tongue. The passive articulator is the hard palate. [  ] are
examples of retroflex sounds.
Rhyme A syllable constituent consisting of the syllable nucleus plus the coda.
Roll Another term for trill.
Root (of tongue) The very back part of the tongue which faces the rear wall of the
pharynx.
Rounded Of a vowel, produced with rounded lips. [y u o ɔ] are examples of
rounded vowels.
Rule A statement of a phonological phenomenon in terms of the input (the
sounds affected), the output (the changes made), and the context in
which the change occurs. An example is: [son] → [voi] / [voi].
This means: any obstruent (the input) becomes voiceless (the output)
before another voiceless sound (the context).
Sampling rate The frequency that samples are taken from a waveform for the
purposes of digitising speech. For example, a sampling rate of 10 kHz
means that each second of the waveform is represented by 10,000
equally spaced samples.
Secondary articulation A constriction of the vocal tract that takes place at the same time as a
narrower constriction elsewhere. [] is an example of a sound where
the secondary articulation is formed by the tongue back being raised
towards the soft palate while at the same the tongue tip makes the
primary articulation (a complete closure on the alveolar ridge).
Secondary stress See stress.
Segment Another term for speech sound.
Semivowel A type of approximant sound, like [w] or [ j], whose articulation is
very similar to that of a vowel.
Short vowel A vowel of relatively short duration when compared to a vowel of
similar or identical quality in the same vowel system.
0521808820cglos_p189_200.qxd 1/27/05 3:09 PM Page 199 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Glossary 199

Sine wave A waveform with a simple shape and a single constant frequency and
constant amplitude.
Soft palate The rear part of the roof of the mouth.
Sonorant An articulation where the constriction is not so narrow, or is not of
sufficient duration, as to cause an appreciable rise in air pressure
inside the vocal tract. Vowels, nasals, approximants, taps, flaps and
trills are all sonorants.
Spectrum A visual display of the relative amplitudes of the different frequency
components of a sound such as a vowel.
Stop A consonant sound that involves a complete closure in the oral cavity.
Plosives, affricates and nasals are all stops.
Stress The relative prominence of syllables. In many languages one (or more)
syllables in a word are produced louder and longer than others. These
syllables are said to be stressed. Examples from English: butter, teaching
(first syllable stressed) and resign, unsure (second syllable stressed). In
many languages there is more than one degree of stress. So, for exam-
ple, in the English word international there is a primary stress on the
third syllable and a secondary stress on the first syllable:
[ntənʃnəl].
Stress-timed language A language whose rhythm shows perceptually equal time intervals
between stressed syllables.
Suprasegmental A name for features such as stress, lexical tone and intonation, which
affect more than one segment.
Syllabary A system of writing where the symbols used represent whole syllables
rather than the individual consonants and vowels that make up the
syllable.
Syllabic Of a speech sound, capable of forming the core of a syllable. In most
languages vowels are the only syllabic sounds, but some consonants,
mainly sonorants, are also used as syllabic sounds in some languages.
Syllable A phonological structure composed of speech sounds. Words are
made up of syllables. The syllable is the domain of association for
such phenomena as stress and lexical tone.
Syllable-timed A language whose rhythm is based on perceptually equal duration of
language syllables.
Synthetic speech An electronic simulation of human speech.
Tap A consonant sound produced by making the active articulator strike the
passive articulator for a very brief duration. [ɾ] is an example of a tap.
Thyroid cartilage The largest cartilage of the larynx. It is attached to the cricoid
cartilage below and is supported by muscles attached to the bones of
the skull. The front ends of the vocal folds are attached to its inner
surface.
Tip (of tongue) The extreme front end of the tongue.
Tone See lexical tone.
Toneme A group of pitch configurations that are functionally equivalent in a
particular lexical tone language.
Trill A consonant sound where the active articulator vibrates rapidly
striking the passive articulator repeatedly. [r ] are examples of trills.
Turbulent flow A characteristic of flow in a fluid such as air or water where the
particles that make up the fluid move in a chaotic, unpredictable
0521808820cglos_p189_200.qxd 1/27/05 3:09 PM Page 200 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

200 Glossary

fashion. The particles collide with each other and the energy created
by these collisions is heard as noise.
Tympanic membrane Another name for the eardrum.
Ultrasound High frequency vibration above the upper frequency limit of normal
human hearing.
Unrounded Of a vowel, produced without rounding of the lips. [i a ɯ] are examples
of unrounded vowels.
Uvula The fleshy protuberance that hangs from the end of the soft palate.
Uvular The name of a place of articulation. The active articulator is the back
of the tongue. The passive articulator is the uvula. [q  ] are examples
of uvular sounds.
Variable stress A language where the primary stress is not fixed to a particular position
language in a word.
Velar The name of a place of articulation. The active articulator is the back
of the tongue. The passive articulator is the soft palate. [k  ŋ] are
examples of velar sounds.
Velaric The name of an airstream mechanism in which the airflow is intitiated
by a closure of the back of the tongue on the velum. Clicks are produced
with a velaric ingressive airstream.
Velarisation A secondary articulation in which the back of the tongue is raised
towards the velum.
Velic closure The closure formed by raising the soft palate to contact the rear wall
of the pharynx. Oral sounds are produced with velic closure, which
prevents air escaping via the nasal cavity.
Velopharyngeal The same as velic.
Velum Another name for the soft palate.
Vocal cords An older term for the vocal folds.
Vocal folds Two folds of tissue with embedded muscle and ligaments found inside
the larynx. They are attached at the back to the arytenoid cartilages and
at the front to the inner surface of the thyroid cartilage. Their vibration
is the source of the periodic energy for human speech sounds.
Vocal tract The oral cavity, nasal cavity, pharynx and larynx.
Voice Onset Time The time lapse, measured in milliseconds, between the release of a
plosive sound and the onset of voicing for the following sound. Voice
onset time is an important perceptual cue for voicing of plosives.
Voiced Produced with accompanying vocal fold vibration or, for plosives,
produced with a short or zero voice onset time.
Voiceless Produced without accompanying vocal fold vibration or, for plosives,
produced with a relatively long voice onset time.
VOT The abbreviation for Voice Onset Time.
Vowel A sound produced without a close obstruction in the vocal tract and
which forms the centre of a syllable.
Vowel Dispersion The tendency for vowel systems to consist of qualities that are widely
Principle and evenly dispersed in perceptual space.
Vowel system The set of contrastive vowel qualities found in a particular language.
Wav file One common format for storing sounds as computer files.
Waveform A visual representation of the variation of air pressure caused by a
sound. The horizontal axis represents time and the vertical axis
represents pressure variation above and below atmospheric pressure.
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 201 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Solutions to the
exercises
Exercise 1.1
1 box XXXX 6 rhyme XXX
2 sorry XXXX 7 climb XXXX
3 possess XXXXX 8 gnaw XX
4 knees XXX 9 mushroom XXXXXX
5 quickly XXXXXX 10 elephants XXXXXXXX

Exercise 1.2
1 damp CVCC 6 physics CVCVCC
2 fox CVCC 7 knowledge CVCVC
3 friend CCVCC 8 spaghetti CVCVCV
4 unit CVCVC 9 columnist CVCVCVCC
5 talker CVCV 10 ghastliness CVCCCVCVC
5 Is CVCVC in some accents of English where the final r is pronounced.

Exercise 1.4
1 CV he to the you to too
2 CVC height but would not reach lot said man shook his head
had keys have much
3 CVCC lost
4 VC of it in
5 VCC aren’t old and
6 CVCV ladder
7 CCVC tried still break
8 CCVCC smiled
9 CVCVC having Colin patted causing damage
10 CVCCVC success wanted

Exercise 1.5
The number below each word indicates the number of segments. If you get a
different figure, it may be because you pronounce that word differently.

The first time I drove a car in Europe was in Italy. We picked up a car at
2 3 3 14 12 2 5 3 2 5 2 4 2 13 2
the airport and drove to the place we were staying. It took us about five
2 4 3 4 2 2 4 2 2 5 23 2 4 3
and a half hours. I had a bit of trouble with the gears and spent the first
3 13 3 13 13 2 5 3 2 3 3 4 2 4
couple of miles stuck in third, but after I got the hang of it, we got along
4 2 4 4 2 3 3 4 13 2 3 2 2 2 3 4
fine. The next day was a different matter. I suppose it was
3 2 5 2 3 17 4 14 23
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 202 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

202 Solutions to the exercises

overconfidence or something, but we nearly had a terrible disaster. We

11 1 6 3 2 4 3 16 7 2
set off to see some sight or other, I forget now where we were going.
3 2 2 2 3 3 2 3 15 2 2 2 2 4
Suddenly my wife screamed out that I was driving on the wrong side of
6 2 3 6 2 3 13 6 2 2 3 3 2
the road. I managed to get back to the correct side just in time to miss
2 3 16 2 3 3 2 2 5 3 4 2 3 2 3
a large lorry which appeared round the bend in front of us. It would have
13 4 3 4 4 2 4 2 5 2 2 23 3
been a very nasty accident and I was extremely careful for the rest of the
3 14 5 7 3 13 9 4 2 2 4 2 2
holiday.
6

Exercise 1.6
As an example, here is an analysis of the passage in exercise 1.5, which
contains 166 words. The syllable count is as follows

1 syllable 133 words

2 syllables 25 words
3 syllables 7 words
5 syllables 1 word

There are some problematic words: correct, hours, suppose could be viewed
as having 1 syllable or 2 and different can be pronounced with 2 syllables or
with 3.

Exercise 1.9
Waveform A is the word potato. There are three separate sections of high
amplitude waveform, corresponding to the three syllables in the word.
Waveform B is the word pepper. There are only two sections of high
amplitude waveform here, corresponding to the two syllables in this word.
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 203 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Solutions to the exercises 203

Exercise 2.1
The first segment is a:
Monday voiced consonant
Tuesday voiceless consonant
Wednesday voiced consonant
Thursday voiceless consonant
Friday voiceless consonant
Saturday voiceless consonant
Sunday voiceless consonant
January voiced consonant
February voiceless consonant
March voiced consonant
April vowel
May voiced consonant
June voiced consonant
July voiced consonant
August vowel
September voiceless consonant
October vowel
November voiced consonant
December voiced consonant
one voiced consonant
two voiceless consonant
three voiceless consonant
four voiceless consonant
five voiceless consonant
six voiceless consonant
seven voiceless consonant
eight vowel
nine voiced consonant

Exercise 2.2
1 bus voiceless 7 buzz voiced
2 use (noun) voiceless 8 use (verb) voiced
3 as voiced 9 teeth voiceless
4 breathe voiced 10 rule voiced
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 204 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

204 Solutions to the exercises

5 has voiced 11 of voiced

6 off voiceless 12 booth voiceless for some speakers,
voiced for others

Exercise 2.4
adult pronunciation child’s form
bag [b
] [b
k]
boot [but] [but]
beaker [bikə] [biə]
can [k
n] [
n]
coach [kəυtʃ] [əυtʃ]
pad [p
d] [b
t]
Paddy [p
di] [b
di]
jug [d] [dk]
father [fɑðə] [vɑðə]
footpath [fυtpɑθ] [vυtbɑθ]
Jack [d
k] [d
k]
motor [məυtə] [məυdə]
rabbit [r
bt] [r
bt]
song [sɒŋ] [zɒŋ]
shop [ʃɒp] [ɒp]
teatime [titam] [didam]

Exercise 3.1
(a) [ʃ] is postalveolar, the rest are alveolar
(b) [f ] is labiodental, the rest are bilabial
(c) [] is uvular, the rest are velar

Exercise 3.2
(a) fight: kite
(b) take: shake, Jake, rake
(c) bathe: base, baize, bale, bait, bate, bane, bayed

Exercise 3.3
(b) sent: all alveolar
(c) palm: all bilabial, the l is silent
(d) align: all alveolar, the g is silent
(e) lazy: all alveolar, the y represents a vowel

Exercise 3.4
(a, b)

Adult Child

bilabial bilabial

dental labiodental

alveolar alveolar

postalveolar palatal

velar alveolar

(c) This happens at the end of a word.

(d) This happens at the beginning or in the middle of a word.
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 205 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Solutions to the exercises 205

Exercise 3.5
(A) labiodental (B) palatal (C) bilabial (D) uvular

Exercise 3.6

Dental Retroflex Velar

Exercise 3.7
It is possible to form an articulation involving the tongue and the upper lip.
This is actually reported in a few languages and is called a linguolabial articu-
lation. It is just about possible to make the tongue tip articulate with the
velum, but this is not known to occur in any language. The lower lip can be
drawn back to approach or touch the alveolar ridge, but again this type of
articulation does not occur regularly in any language.

Exercise 4.1
The consonants are all fricatives.

Exercise 4.2
talked, sky, cows, captive, chalk

Exercise 4.3
(a) They are all voiceless plosives.
(b) Mutation I changes them to the equivalent voiced plosives.
(c) Mutation II changes them into homorganic voiceless fricatives, or in the
case of [p] into a near-homorganic voiceless fricative.
(d) The initial consonants are all voiced plosives. In the case of the bilabial
and alveolar plosives, Mutation I changes them into near-homorganic
voiced fricatives and in the case of the velar the consonant is deleted.

Exercise 4.4
A is a plosive. B is a fricative. C is a trill.

Exercise 4.5
lorry
[lɒli], shoes
[tud], yellow
[leləυ], woolly
[lυli], silly
[tli],
puzzle
[pdl], swallow
[tlɒləυ], umbrella
[mblelə]
You could write either [i] or [ ] instead of the symbol [i] which we have used
at the ends of words here.

Exercise 4.6
sixty
[kti], amusing
[əmjuŋ], results
[ɹlt],
cousins
[kən], sensible
[enəbl]
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 206 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

206 Solutions to the exercises

Exercise 4.7
dramar
[dramaʁ], taran
[taɾan], garsat
[aʁsat], rakra
[rakra],
aramard
[aɾamaʁd]

Exercise 4.8
[slamb] This is possible.
[flant] This is possible.
[walm] This not possible because [m] is a sonorant, so the
word contravenes (c).
[spats] This is not possible because [p] is an obstruent, so the
word contravenes (b).
[twalf] This is possible.
[kjaŋ] This is possible.
[blams] This is not possible. The word contravenes (d). C3 is a
bilabial nasal so C4 must be a bilabial plosive. [s] is an
alveolar fricative.
[ljalʃ] This word is not possible. [l] is a sonorant, so the word
contravenes (a).
[zjalt] This is possible.
[kramd] This is not possible. The word contravenes (d). C3 is a
bilabial nasal so C4 must be a bilabial plosive. [d] is an
alveolar plosive.
[dlaps] This is possible.
[snand] This is possible.

Exercise 4.9
(a)

bilabial alveolar palatal velar uvular

voiceless plosive p t c k q

voiced plosive b d   

voiced nasal m n  ŋ 

(b)

bilabial alveolar palatal velar uvular

voiceless fricative φ s ç x χ

voiced fricative β z  γ ʁ

(c)
(i) All consonants in a word are hormoganic.
This is false.
[zaβa], [aʁuŋ], [baφam], [nusaφ], [iγax], [ceʁaç] and [auχ] are all
words that disprove the statement.
(ii) Stops may not occur between vowels.
This is true. All the intervocalic consonants are fricatives.
(iii) The only obstruent consonants permitted word-finally are voiceless fricatives.
This is true.
(iv) Front consonants (articulated on the alveolar ridge or further forward)
may not occur in the same word as back consonants.
This is true.
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 207 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Solutions to the exercises 207

(v) Voiced fricatives may not occur word-initially.

This is false. [zaβa] is the single word that disproves the statement.

Exercise 4.10
A pharyngeal nasal is impossible because a nasal consonant requires airflow
through the nasal cavity while the oral cavity is completely blocked. While it
is possible, though not particularly easy, to form a complete closure by
drawing back the root of the tongue to contact the rear wall of the pharynx,
this closure would prevent air flowing through the nasal cavity. The closure
would be below the point where the vocal tract divides into oral and nasal
cavities

Exercise 4.11
Keyword Voice Place Manner
p pie voiceless bilabial plosive
t tie voiceless alveolar plosive
k key voiceless velar plosive
b buy voiced bilabial plosive
d die voiced alveolar plosive
 guy voiced velar plosive
m my voiced bilabial nasal
n no voiced alveolar nasal
ŋ sing voiced velar nasal
f fee voiceless labiodental fricative
v van voiced labiodental fricative
θ thigh voiceless dental fricative
ð though voiced dental fricative
s so voiceless alveolar fricative
z zoo voiced alveolar fricative
ʃ she voiceless postalveolar fricative
 measure voiced postalveolar fricative
tʃ chip voiceless postalveolar affricate
d jam voiced postalveolar affricate
w wet voiced labial-velar median approximant
ɹ red voiced postalveolar median approximant
l let voiced alveolar lateral approximant
j yes voiced palatal median approximant
h hat voiceless glottal fricative

Exercise 4.12
(a) singing
[s], onion
[djəd], murmur
[bbə]
˜n
(b) [ni e m
ã̃m]
teatime, [m  ə̃]
paper, [mẽ̃ŋ̊̃ŋ]
baking or making

Exercise 5.1
(a) [ɔ] is the odd one out, because it is the only rounded vowel
(b) [u] is the odd one out, because it is the only back vowel
(c) ["] is the odd one out, because it is the open-mid and the others are close

Exercise 5.2
[ε$] lowered open-mid front unrounded
[e] close mid front unrounded
[e
%] lowered close mid front unrounded
[e
&% ] raised centralised close-mid front unrounded
[e
%] raised mid central unrounded
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 208 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

208 Solutions to the exercises

Exercise 5.3
[ɑlmɑdɑn] [tokojdon] [yjdøn] [køldøn] [etten] [tuzdon] [iʃten] [ɯldɑn]

Exercise 5.4
Words (a) and (c) contain monophthongs. They show much less change in
formant frequencies than words (b) and (d). The words are: (a) seat (b)
sight (c) seed (d) side.

Exercise 5.5
The durations are approximately as follows:
(a) 130 ms
(b) 180 ms
(c) 330 ms
(d) 335 ms
The ratio of (a) to (c) is 0.39
The ratio of (b) to (d) is 0.54

Exercise 5.6
Spectrogram (a) is [kil] and (b) is [kyl]. The lower F2 frequency in (b) is a
sign that the vowel is rounded.

Exercise 5.7
(a) Hindi
(b) Farsi
(c) Hindi, Igbo
(d) Japanese
(e) Hindi

Exercise 6.1

Exercise 6.2
(a) [jɔ nju kredt kɑd  sun]
 ʃυd ərav
(b) [ʃɒpŋ n selz z ə υd we tə et b ]
ɑnz
Exercise 6.3
The vertical lines on the diagram mark the start and end of the VOT which is
approximately 65 ms.

Exercise 6.4
The velar plosives in the words kite and occur are likely to have a longer VOT
than those in sky, baker and basket.

Exercise 6.5
Waveforms (a), (c) and (f) show examples of normal voice at different fun-
damental frequencies. Waveform (b) shows a low-frequency double-pulsing
excitation which sounds creaky. Waveform (e), with very short closures, is
breathy. In waveform (d) the rate of closing is slower than that of opening. It
was made on an inward flow of air.
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 209 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Solutions to the exercises 209

Exercise 6.6
(b) feet
(c) feast
(f) felt

Exercise 6.7
(a) please
(c) quick
(d) crown
(f) pure

Exercise 7.1
(a) a voiceless alveolar lateral click
(b) a voiced velar implosive

Exercise 7.2
The larynx must rise to cause the increase in air pressure. The soft palate
must be in a raised position throughout, sealing off the nasal cavity, otherwise
there will be no increase in pressure because air will escape via the nose.

Exercise 7.3
(a) The soft palate is raised shutting off nasal airflow. The glottis is closed
tightly. The tongue tip makes a firm closure on the alveolar ridge and the
sides of the tongue contact the upper side teeth and gums. The larynx moves
upwards compressing the air in the mouth and pharynx. The tongue rapidly
breaks its contact with the alveolar ridge and the compressed air escapes.
Shortly afterwards the glottis opens.
(b) The lips are closed. The back of the tongue makes firm contact with the
velum. The centre of the tongue moves down, enlarging the space between
the two closures and the air pressure in the oral cavity decreases. The lips
part and air rushes into the oral cavity.
(c) The tongue tip makes a firm closure with the very front of the alveolar
ridge and the back of the upper front teeth. The rims of the tongue make
firm contact with the upper side teeth and gums. The back of the tongue
makes firm contact with the soft palate. The soft palate itself is lowered,
allowing nasal airflow. The vocal folds are held loosely closed. The lungs push
air up the trachea, causing the vocal folds to vibrate. The tongue centre
moves down, enlarging the enclosed space between the dental and velar clo-
sures and causing a decrease in air pressure. The dental closure is released
and air rushes into the oral cavity.

Exercise 7.4
(a) voiceless alveolar lateral fricative: pulmonic egressive
(b) voiced bilabial implosive: pulmonic egressive  glottalic ingressive
(c) voiceless dental click: velaric ingressive
(d) voiceless alveolar ejective affricate: glottalic egressive
(e) voiced nasalised postalveolar click: velaric ingressive  pulmonic egressive
(f) voiceless velar ejective affricate: glottalic egressive
(g) voiceless aspirated velar plosive: pulmonic egressive
(h) voiced alveolar lateral click: velaric ingressive  pulmonic egressive
(i) voiceless velar ejective stop (unaffricated): glottalic egressive
(j) voiced bilabial nasal: pulmonic egressive
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 210 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

210 Solutions to the exercises

Exercise 7.5
(a) Watch what you’re doing [wɒtʃ( wɒt( jɔ )u*]
(b) Can I get a glass of milk? [k(
) a *et( ə *lɑs( əv +lk(]
(c) Is it five past nine? [z t( f(av p(ɑs(t( )a)]

Exercise 8.1
At the beginning of the utterance the organs of speech are in a position of
rest. The glottis is open for quiet breathing and the soft palate is lowered. The
lips are closed and the tongue fills most of the oral cavity. The word is pro-
duced on a pulmonic egressive airstream. The lips open. The soft palate rises
to form a closure with the back wall of the pharynx preventing air from escap-
ing via the nose. The tongue tip approaches the alveolar ridge and forms a
narrow constriction. The rims of the tongue are in firm contact with the upper
side teeth and gums, preventing lateral escape of air. The lips are rounded in
anticipation of the following rounded sound. As the air passes through the
alveolar constriction, it becomes turbulent and generates friction noise for the
first segment [sw]. The tongue tip moves away from the alveolar ridge and
the back of the tongue rises towards the soft palate to form a wide approxi-
mation. The lips remain rounded. The vocal folds are drawn together loosely
and begin to vibrate soon after the friction noise dies away. They continue to
vibrate for the rest of the utterance. The lips lose their rounded shape and
the back of the tongue moves away from the soft palate. Simultaneously, the
front of the tongue rises towards the hard palate for the production of the
vowel [̃ ]. During the production of the vowel the soft palate begins to fall in
anticipation of the following nasal segment. This allows air to escape through
the nasal cavity and the vowel is nasalised. The lips close, shutting off the oral
airflow, so that the airstream can escape only through the nasal cavity for the
nasal stop [m]. The vocal folds stop vibrating and the organs of speech return
to their rest position.
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 211 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Solutions to the exercises 211

Exercise 8.2
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 212 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

212 Solutions to the exercises

Exercise 8.3
(a) camp fits the diagram. The number of segments is correct and there is a
nasal stop in position 3 as shown in the diagram.
(b) cold does not fit, as it contains no nasal segment.
(c) mission does not fit. Although the word may be pronounced with 4 seg-
ments [mʃn], the first and last of these need the soft palate to be low-
ered. The diagram shows segments 1 and 4 with a raised soft palate.
Furthermore, [ ʃ ] is produced with the soft palate raised but the diagram
shows segment 3 with a lowered soft palate.
(d) elms fits. The [ l] in this word will be nasalised as is shown for segment 2
in the diagram.
(e) film does not fit. The nasal is in the wrong place.

Exercise 8.4
The sequences of plosives where release masking is likely are underlined in
the passage.
The chap grinned, picked himself up off the floor and rubbed his knee. Up till
that moment the atmosphere had been quite relaxed. Everyone started to
look tense. Maybe he wouldn’t be able to do it.

Exercise 8.5
We estimate the duration of the [s] in sigh to be 235 ms and the [s] in sky to
be 165 ms.

Exercise 9.1
Burmese: The two sounds involved are [m] voiced bilabial nasal and
[m ] voiceless bilabial nasal. For the first the vocal folds are
vibrating and for the second the glottis open and there is
no vibration.
Ewe: The sounds are [f] voiceless labiodental fricative and [φ]
voiceless bilabial fricative. For the first of these the articu-
lators are the top front teeth and the lower lip. For the
second the articulators are the upper and lower lips.
German: The sounds are [d] and [t] voiced and voiceless alveolar
plosives. The difference is the presence versus absence of
vocal fold vibration.
Italian: The sounds are both voiceless alveolar plosives. The plo-
sive in the word for done has a much longer hold phase.
Korean: The two sounds are [p] voiceless unaspirated bilabial plo-
sive and [ph] voiceless aspirated bilabial plosive. The sec-
ond has a much longer VOT.
Spanish: The sounds are [r] voiced alveolar trill and [ɾ] voiced alve-
olar tap. For the first the tongue tip vibrates against the
alveolar ridge, making a series of rapid closures. For the
second the tongue tip strikes the alveolar ridge just once.
Welsh: The sounds involved are both close, front, unrounded vow-
els. The vowel in the word for white has a longer duration.
Exercise 9.2
The words that make minimal pairs with the first word are underlined
dot dock dog dots debt doll got
can cat cone cane cape man mat
hat flat that had hate at heart
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 213 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Solutions to the exercises 213

green bean groan gain grease glean grain

shoe show shoot shrew Sue blue who
Can/cape is not a minimal pair because there is more than a single segment
difference between the words /k
n/ and /kep/. The same is true of the
pair can/mat /k
n/ – /m
t/.
Hat/flat is not a minimal pair because the second word has an extra segment
and the initial segments are different.
Green/bean is not a minimal pair because the initial segments are different
and the second word has one fewer segments. Green/gain is not a minimal
pair because there is a mismatch in the number of segments and the vowels
are different too.
Shoe/blue is not a minimal pair because there is a mismatch in the number
of segments and the initial segments are different.

Exercise 9.3
(a) An example of a minimal pair for /w/ versus /r/ is went and rent. The
child produces only [w], so all minimal pairs depending on the contrast will
be homophones in the child’s speech. The child would say, for example,
[went] for both went and rent.
(b) An example of a minimal pair for /s/ versus /ʃ/ is bus and bush. The
child produces only [s], so again all minimal pairs depending on the contrast
will be homophones in the child’s speech. The child would say [bs] for both
bus and bush, for example.
(c) The adult words that contain affricates are: change, crunch, James, judge,
smudge, stretch, watch. Sometimes, the child produces an affricate, but an
alveolar one rather than a postalveolar one like the adult target. This happens
at the end of a word. At the beginning of a word, the child uses an alveolar
plosive instead of an affricate. The voicing of the child’s consonants matches
that of the adult target.
(d)
• Words beginning with a cluster of /s/  a plosive: school, spot. The
child deletes the /s/.
• Words beginning with a cluster of /s/  plosive  approximant:
spring, stretch. Again the child deletes the /s/, but retains the approx-
imant, although as we have seen it is [w] rather than [ɹ].
• Words beginning with a cluster of a plosive  approximant: brush,
class, crunch. Both segments are retained by the child.
• Words beginning with a cluster of /s/  nasal: smack, smudge,
sneeze. Here the first segment is not deleted, but is produced as a
voiceless nasal at the same place of articulation as the following nasal.
(e) /s/ ¡ ø /# C
[-nasal]
(f) Charles [tɑlz], choose [tuz], fridge [fwdz], juice [dus], lunch [lnts],
prince [pwns], scratch [kw
ts], snake [n nek], stamp [t
mp], touch [tts]

Exercise 9.4
The child’s consonants are p b m m
 tdnn
 s z ts dz k ŋ w l
(a) [nasal, voiced]: m n ŋ
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 214 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

214 Solutions to the exercises

(b) [anterior, –coronal, –continuant]: p b m m

(c) [–lateral, voiced]: b m d n z dz ŋ w
(d) [–sonorant, –continuant, –nasal]: p b t d ts dz k
(e) [–anterior, –continuant]: k ŋ

Exercise 9.5
The three fricative sounds are all voiceless [h] glottal, [ç] palatal, [φ] bilabial.
None of these occurs word finally, but all occur both word initially and
between vowels in the middle of a word. [ç] only occurs immediately preced-
ing [i]. [φ] only occurs immediately preceding [ɯ]. [h] occurs immediately
preceding all the other vowels [e a o], but not before [i] or [ɯ]. They are
therefore in complementary distribution because where one can occur the
other two cannot. As they are also phonetically quite similar they can be
regarded as allophones of the same phoneme. As [h] has the widest distribu-
tion this single phoneme is best symbolised as /h/.
[s] and [h] on the other hand are not in complementary distribution. Both
occur immediately before [a], as can be seen in the words for tooth and wal-
let. This indicates that they should be regarded as allophones of two distinct
phonemes.

Exercise 9.6
(a) The plosive sounds in the data are
[p] voiceless bilabial plosive
[b] voiced bilabial plosive
[t] voiceless alveolar plosive
[th] voiceless aspirated alveolar plosive
[k] voiceless velar plosive
[] voiced velar plosive
Both voiced and voiceless plosives occur immediately after a homorganic
nasal, but elsewhere plosives are all voiceless. The occurrence of voiced plo-
sives only in words where they are in free variation with the corresponding
voiceless plosive (butterfly, corn, dead limb, we chase, your daughter) indi-
cates that the voicing distinction is not contrastive for plosives in Campa.

(b) [t] and [th] are in contrast in Campa. The minimal pair [itoki] on his
head and [ithoki] egg proves this.

(c) We have already established that there is no contrast between [p] and
[b]. As these are phonetically similar, we may safely conclude that they are
allophones of the same phoneme. As [p] has the wider distribution, let us
represent the phoneme as /p/. Notice also that [p] does not occur between
two vowels, but only word-initially or in the middle of a word preceded by a
consonant. As we saw [b] occurs in this latter environment. Both [w] and [β]
only occur between two vowels. Moreover, [β] only occurs immediately
before the front vowels [i] and [e], while [w] only occurs before other vowels
[a], [o] and [ə]. These two sounds are therefore in complementary distribu-
tion, both with each other and with [p] and [b]. As the sounds are all labial
([p b β] are bilabial and [w] is labial-velar), it would be reasonable to con-
clude that they are all allophones of the /p/ phoneme.

(d) It is clear that [ n] and [ ŋ] are in complementary distribution. [ŋ] only

occurs before velar plosives, whereas [n] nevers occurs in this environment.
We can then conclude that these two sounds are allophones of a single
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 215 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Solutions to the exercises 215

phoneme /n/ and that the occurrence of [ŋ] be dealt with by an assimilation
rule. Turning to [m], we can see that two instances of this sound are immedi-
ately before a bilabial plosive and we might be tempted to deal with these
using an assimilation rule too. However there is another occurrence of [m] in
the word [no,maɾε] my paddle. This cannot be accounted for by assimila-
tion. We must therefore conclude that there are two nasal phonemes in
Campa: /m/ and /n/.
(e) The vowel sounds are:
[i] front, close, unrounded
[-] central, close, unrounded
[ε] front, open-mid unrounded
[ə] central, between close-mid and open-mid, unrounded
[a] front, open, unrounded
[o] back, close-mid rounded
(f) The vowel phonemes of Campa are /i ε a o/. They are fewer in number
than the vowel sounds for the following reasons:

• [a] and [ə] are in free variation. A word may end in either of these
two vowels and no meaning difference is signalled. [a] has a wider
distribution than [ə], so these two sounds may be assigned to the
/a/ phoneme.
• [i] and [-] are in complementary distribution. [-] only occurs immedi-
ately following [s] or [ts]. [i] never occurs in this environment. They
are both allophones of the /i/ phoneme.

Exercise 9.7
(a) V→ [nasal]/ [nasal]
A vowel is nasalised before an immediately following nasal (or nasalised
sound).
(b) [–sonorant]→ [–voiced]/ #
Any obstruent (plosive, fricative or affricate) is voiceless if it occurs
word-finally.
(c) ø → [ɹ]/ V V
[–high]
[ɹ] is inserted between a mid or low vowel and any following vowel.

Exercise 10.1
List 1 Words with primary stress on the first syllable:
attitude, concentrate, crocodile, elephant, fantasy, mystery, troublesome,
universe
List 2 Words with primary stress on the second syllable:
connection, excitement, fantastic, horizon, imagine, intention, recover
List 3 Words with primary stress on the last syllable:
afternoon, kangaroo, millionaire, puppeteer, understand

Exercise 10.2
Maori stress
[ata,po] ‘early dawn’ [papaɾa,kauta] ‘hotel’
[ki,hiɾimete] ‘Christmas’ [ka,wanataŋa] ‘government’
[,paepae] ‘beam’ [kau,matua] ‘male elder’
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 216 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

216 Solutions to the exercises

[tau,ɾekaɾeka] ‘slave’ [papa,kaiŋa] ‘village’

[,maɾae] ‘courtyard’ [,kaimoana] ‘seafood’
[,fafai] ‘war’ [,mokopuna] ‘grandchild’

Exercise 10.3
Manam stress: If the final syllable ends in a vowel, then the penultimate sylla-
ble is stressed. Otherwise the final syllable is stressed.

Exercise 10.4
Waveform (a) is of an utterance of [m,pɔt] import (verb) and the duration
of the first syllable is 154 ms. Waveform (b) is [,mpɔt] import (noun) and
the initial syllable duration is 191 ms.

Exercise 10.5
Chengdu tones
(a)
ʃiau0055 tʃəu0041

tʃin0550 i0033

(b) Glottal stop appears at the end of the following words:

fei21dzauʔ211
soap(, ʃiau5500 tʃiʔ211
stiny(, jin213 tʃiʔ211
luck(, koŋ5500
paʔ211 ‘perhaps’
For all of these, the citation form of the second element ends in a vowel
sound and has the falling rising toneme.
(c) oŋ45 fu33 ‘time’ oŋ 1fu
(d) I High Rising
Citation: 45 Variant: 33 in second element except after 213
II Low Falling
Citation: 21 No variants
III High Falling
Citation: 41 Variant: 55 in first element
IV Low Falling-rising
Citation: 213 Variant: 33 in second element following another 213
Variant: 211 in second element after other tones

Exercise 10.6
Underlining indicates the place of nuclear syllables
(a) My sister who lives in Liverpool | has just changed her job.
(This would probably be interpreted as meaning that I have more than
one sister and I am telling you which one I am talking about.)
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 217 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Solutions to the exercises 217

My sister | who lives in Liverpool | has just changed her job.

(This is the equivalent of ‘My sister lives in Liverpool and she has just
changed her job’.)
(b) She played | and I sang the national anthem.
(She played something, not even a musical instrument necessarily.)
She played | and I sang the national anthem.
(She played the national anthem and I sang it.)
(c) He didn’t leave because he was afraid.
(He left for some other reason.)
He didn’t leave | because he was afraid.
(Being afraid was the reason for his not leaving.)

Exercise 10.7
Sid: I’ve got an idea | Let’s go out for a meal.
Clare: All right | Which restaurant do you want to go to?
OR
Clare: All right | Which restaurant do you want to go to?
Sid: There’s a good Chinese place | near the library.
Clare: I’m not terribly fond of Chinese food.
Sid: Maybe that’s because you’ve never had any good Chinese
food.
Clare: That may be true | but I think I’d prefer to try that place |
some other time.
OR
Clare: That may be true | but I think I’d prefer to try that place |
some other time.
0521808820csol_p201_218.qxd 1/27/05 3:10 PM Page 218 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:
0521808820cind_p219_222.qxd 1/27/05 3:11 PM Page 219 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Index
abstract form, 140 cardinal vowels, 77–78
accent, 167–69 Catalan, 79, 129, 158
affricate, 56 Chaucer, Geoffrey, 81
airflow Chechen, 107
laminar, 52 Chinese, 12, 80, 93, 138, 147, 163, 165, 175
turbulent, 53 Chontal, 107
airstream mechanisms, 103–115 class, natural, 143
glottalic see ejective, implosive ‘clear I’ 122
non-pulmonic, 104 click, 110–13
pulmonic, 104–5 modifications, 111–12
velaric see clicks clipping, pre-fortis, 127
allophone, 138–39 coarticulation, 122–26
allotone, 165 cochlea, 179
alphabets, 10 cochlear implant, 185
alternation, 140 coda, 145
alveolar ridge, 35 complex periodic tones, 30
amplitude, 9 consonant, 30
aperiodicity, 29 cluster, 146
approximant, 58–59 cluster reduction, 185
lateral, 58 systems, 45, 62
median, 58–59, 62 continuum, 181
Arabic, 38, 93, 122, 140 contrast, 136
Aranda, 45, 58, 79 creaky voice, 98
articulation, 36 cricoid cartilage, 22
advanced, 124 cue redundancy, 183–85
overlapping, 125 CV-skeleton, 7
primary, 118–20 cycle of vibration, 28
retracted, 124 Cyrillic alphabet, 11
secondary, 40, 118–20 Czech, 157
articulators, 120
active and passive, 36 ‘dark l’ 122
arytenoid cartilages, 22 deletion, 141
aspiration, 94 weak syllable, 184
assimilation, 140 devoicing
association line, 165 of approximants, 94
audiogram, 180 of obstruents, 95
audiometer, 180 diacritic, 4
auditory canal, 179 diagrams, parametric, 124
Australian languages, 9, 45, 62, 79 digitising speech, 29
diphthong, 8, 75–76
background, 170 discrimination curve, 182
Bafang, 164 dissimilation, 142
Bell, Alexander Melville, 4 distribution, 137–38
Bole, 164 complementary, 139–40
boundary, intonational, 166 Divehi, 4–5
brackets Diyari, 45
slant, 138 double articulation, 40
square, 4 Dutch, 93
breathy voice, 97–98 Dyirbal, 136
BSL (British Sign Language), 15
Burmese, 55, 137, 138 eardrum, 179
Early Modern English, 81
Campa, 150 Efik, 79
Canadian English, 138–39 egressive flow, 105
Cantonese, 93, 163 Egyptian, 13
0521808820cind_p219_222.qxd 1/27/05 3:11 PM Page 220 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

220 INDEX

ejective, 105–8, 110 homorganic sounds, 40

electropalatography (EPG), 41–42, 129 hypernasal speech, 66–67
elision, 141 hyponasal speech, 66–67
ending, 8, 9
environment, 137–38 implosive, 108–110
Ewe, 137 infrasound, 179
excitation, larynx, 70 ingressive flow, 105
initiator, 105
F0 see frequency, fundamental insertion, 141
features, 142 International Phonetic Alphabet, 2–4
binary, 142 intonation, 165–71
[anterior], 143 signalling attitude, 170
[back], 143 IP see phrase, intonational, 170
[continuant], 143 Irish, 120
[coronal], 143 Italian, 38, 58, 62, 92, 166
[high], 143
[lateral], 143 Jalapa Mazatec, 98
[low], 143 Jamaican English, 81
[nasal], 143 Japanese, 13, 15, 62, 93
[round], 143
[sonorant], 143 Karen, 164
[syllabic], 143 key, 170
[voice], 143 Khoisan languages, 112
Finnish, 45 kilohertz (kHz), 28
[fis] phenomenon, 184 Kirghiz, 84
flap, 60 Korean, 137
focus Kpelle, 79
broad, 168 Kwakiutl, 122
narrow, 168
foot, 161–62 labelling, 122
formant, 71–72 curve, 182
fortis sounds, 95 experiment, 181
fortition, 142 experiments with hearing-impaired subjects, 186
French, 13, 14, 38, 40, 62, 74, 75, 79, 93, 126, 136, 142, 183 experiments with infants, 185
frequency, 28 labialisation, 118
fundamental, 154 laryngograph, 24–25, 97–98, 100
frequency, range of, 154 larynx, 22
fricative, 56–58 lateral
lateral, 57–58 approximant, 58, 62
median, 56–57 fricative, 57
frictionless continuant, 58 Latin, 142
alphabet, 4
General American English, 139 lengthening, pre-pausal, 127
German, 56, 93, 126, 138, 142, 144 lenis sounds, 95
glide, 58 lenition, 141
glottal stop, 4, 38 Linear B, 11
Great Vowel Shift, 81 lip rounding, 74
Greek, 92–93 see also labialisation
alphabet, 4, 11 lip-reading, 180–81
Gugu-Yulanji, 45 lips, 34
Gujarati, 98 liquid, 59
Lithuanian, 140
Hausa, 107, 163 logographic writing, 12
Hawaiian, 62 loudness, 9
hearing, 62
aids, 185 magnetic resonance imaging (MRI), 43–45
impairments, 185–86 Malay, 45
loss, 180, 185 Malayalam, 62
mechanism, 178–79 Maltese, 11
Hebrew, 11 Manam, 173
Hertz (Hz), 28 Maori, 45, 62, 75, 147, 173
hieroglyphs, 13 Margi, 163
homophones, 14 McGurk effect, 181
0521808820cind_p219_222.qxd 1/27/05 3:11 PM Page 221 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

Index 221

meatus, 179 Polish, 148

Middle English, 81 pre-fortis clipping, 97, 101
minimal pair, 136–37 processes, phonological, 140–42
Mixtec, 163 processing, 178
modal voice, 97 central, 178
monophthong, 75 peripheral, 178
monosyllable, 7 prosody, 154
MP3, 29 Punjabi, 61
murmur see breathy voice
release, inaudible, 126
Nama, 62 resonance, 70–72
nasal, 55, 62 rhyme, 146
voiceless, 55 rhythm
nasal cavity, 35 stress-timed, 162
nasalised (vowel), 14, 124 syllable-timed, 162
Navaho, 163 roll see trill, 62
New Zealand English, 84 Rotokas, 62
noise, ambient, 9 rules
Northern Irish English, 82 context-free, 145
Norwegian, 62 formalism for, 145
nucleus, 62 input of, 145
intonational, 167–69 output of, 145
of syllable, 145 phonological, 144–45
Nunggubuyu, 79 Russian, 11, 122, 148
Nupe, 163
sampling rate, 29
obstruent, 54 schwa, 74
octave, 28 segment, 7, 15
Old English, 138 semivowel, 58, 62
onset, 145 sensitivity, auditory, 179
oral cavity, 35 Shakespeare, William, 81
oral tradition, 13 sign languages, 15
organs of speech, 34–36 Sindhi, 110
ossicles, 179 sine wave, 29
oval window, 179 soft palate see velum, 54
sonorant, 54
palatalisation, 119 sound, 9
paralinguistics, 171 Spanish, 38, 59, 62, 137, 138, 141, 142,
pause, 166 168
periodicity, 28 spectrogram, 75
pharyngealisation, 120 spectrum, vowel, 70
pharynx, 35 speech development, 184–85
phonation type, 97–99 speech perception
phoneme, 138–39 tests, 181–184
phonology, 143–51 categorical, 182
autosegmental, 165 cross-language comparison of, 183
phonotactics, 146–48 speech synthesis, 181
phrase, intonational, 166 spelling, English, 7, 14
pinna, 179 Steele, Joshua, 166
pitch, 28, 154 stem, 8, 9
places of articulation, 38 stop, 51–56
alveolar, 37 nasal, 55
bilabial, 37 oral, 55
dental, 37 strength, scales of, 141
glottal, 38 stress, 9, 154
labiodental, 37 and suffixes, 159
palatal, 37 and vowel quality, 159–60
pharyngeal, 38 and word class, 159
postalveolar, 37 contrasts, 160
retroflex, 37 fixed, 157
uvular, 38 free, 158
velar, 38 lexical, 160
plosive, 55–56 primary, 157
0521808820cind_p219_222.qxd 1/27/05 3:11 PM Page 222 Quark07 Quark07:BOOKS:CU1937-Ashby:Chapters:

222 INDEX

stress (cont.) ultrasound, 179

rhythmic, 161–162 underlying form, 144
secondary, 157 UPSID, 79
sentence, 161 Urhobo, 165
variable, 158 uvula, 35
stricture, degree, 53
stroboscope, 26 velarisation, 119
suprasegmentals, 9, 154 velum, 35, 55
surface form, 144 Ventris, Michael, 11
Swahili, 109 Vietnamese, 147, 163
syllabaries, 11–12 Visible Speech, 4
syllabic consonant, 8 visual cues, 180–81
syllable, 7, 8, 15, 145–46 vocal folds, 22
closed, 147 vibration of, 23
heavy, 158 vocal tract, 22
light, 158 voice onset time (VOT), 93
nucleus of, 145 voiced sounds, 23
open, 146 voiceless sounds, 23
template, 147 voicing, 23
weight, 158–159 context-sensitive, 31
in languages, 26
Tamil, 37 vowel, 7
tap, 59–60 dispersion, 80
Thai, 163 height, 73
threshold curve, 180 length, 75
thyroid cartilage, 22 nasalised, 75
tiers, 165 quadrilateral, 76–77
tone, 4 system, 79–80
contour, 163 VPM labels, 52
falling-rising, 170
functions of intonational, 169 wav files, 29
intonational, 169 waveform, 9–10
letters, 164 weakening, 141
lexical, 162 Welsh, 57, 62, 157, 162
nuclear, 169 whisper, 26–27
numbers, 164 word
register, 163 content, 162
transcription and representation grammatical, 162
of, 164 writing, 13–14
toneme, 165 writing system, 10–13
tongue, areas of, 35
Xhosa, 113, 163
transcription, 5–7
X-rays, 42–43, 129
trill, 59
truncation, 141 Yoruba, 40, 163
Turkish, 157
tympanic membrane, 179 Zulu, 11, 57, 113, 163