Chapter 2
The production of speech:
the physiological aspect
The production of speech 9
2.1 The speech chain
Speech is the result of a highly complicated series of events. The communication in sound of such a simple concept as It's raining involves a number of activities on the part of the speaker. In the first place, the formulation of the coniir-will take place in the brain; this first stage can be said to be psychological or psycholinguistic. The nervous system transmits this message to the organs j speech which will produce a particular pattern of sound; thus the second stace is physiological or, more precisely, articulatory. The movement of the organs of speech will create disturbances in the air; these varying air pressures can le investigated and constitute the third stage in the chain, the physical or, more precisely, acoustic. Since communication generally requires a listener as well as a speaker, these stages will be reversed at the listening end: the sound waves will be received by the hearing apparatus and information transmitted alon t nervous system to the brain, where the linguistic interpretation of the message takes place. Phonetic analysis has often ignored the role of the listener. Bui in, investigation of speech as communication must ultimately be concerned wifl both the production and the reception ends.
However, our first concern is with the second stage, speech production or articulation. For this reason, we now examine how the various organs behave tL produce the sounds of speech. References are made to the videos on the website where the reader can view the organs in action.
2.2 The speech mechanism
Humans possess, in common with many other animals, the ability to product, sounds. Humans differ from other animals in being able to organise the range of sounds which they can emit into a highly efficient system of communication Non-human animals only rarely progress beyond the stage of using the sounds they produce as a reflex of certain basic stimuli to signal fear, hunger, :. \ i. ■ excitement and the like.1 Nevertheless, like other animals, man uses organs for speaking whose original physiological function developed before vocal >> i -munication was acquired; in particular, organs situated in the respiratory tract
j , SoUrces of energy—the lungs
I <;0ui-ce of energy for our vocal activity is provided by an airstream 11,-1 TVf ''i' 'he lunSs- Tnere are 'ang113?68 which possess sounds not requiring '^i1" "" , c) air for their articulation and, indeed, in English we have one 'iii" ■' 1P" .ilgl|jstic sounds, such as that we write as tut-tut and the noise of or ln0 \. 1|lu i-imade to horses, which are produced without the aid of the lungs; Cr"'l'V he ell'ial sounds °f English use lung air for their production. Our ^ '      „,. []ierefore, partly shaped by the physiological limitations imposed
-v ih- n™,r
. ,,f our lungs and by the muscles which control their action. We „ ,,h j-hI '<.■ pause in articulation in order to refill our lungs with air and this """■I u -.1.1111- extent condition the division of speech into intonational phrases ' -.11 r. I ' i In those cases where the airstream is not available for the upper ,". j-^nixeh (as when, after the surgical removal of the larynx, lung air does 1 ^'r'"h the mouth but escapes from an artificial opening in the neck) a new i-J Hi e >.' ay, stomach air, may be employed. A new source of this land -n-u-e-1 i. ■-■ restrictions than those exerted by the lungs and variation of energy jj |,;S ;t iJenrly controlled.
\ ni-i.irrci of techniques are available for the investigation of the activity dur-iru wn.et.1.111 ,ne tags and tneu" controlling muscles. At one time air pressure "■hi | nj [liiijs was observed by the reaction of an air-filled balloon in the stomach On the basis of such evidence from a gastric balloon, it was claimed that *\1 jh's.- were formed by chest pulses.2 Such a primitive procedure was replaced i"i ik -ichrtt'que of electromyography, which demonstrated the electrical activity n| til- i e respiratory muscles most concerned in speech, notably the internal i iieici—IjIs; this technique disproved the relationship between chest pulses and ■iyilables.3 X-ray photography and CT scans can reveal the gross movements nt fK lis and hence by inference of the surrounding muscles, although the wchri |.ie of Magnetic Resonance Imaging (MRI) is now preferred on medical and safetv grounds.4
2.2.2 The larynx and the vocal cords
I he -. ■.in.jL sCiliI :(■=■ i .
vocal 1 i. 'C'cSt n
ire; is
may t | "I .iel
r.iream provided by the lungs undergoes important modifications in the p^rts of the respiratory tract before it acquires the quality of a speech I irst of all, at the top end of the trachea or windpipe, it passes through x containing the vocal folds or, as they are more commonly called, the Cords {see Fig. 1).
larynx is a casing, formed of cartilage and muscle, situated in the upper I f: trachea. Its forward portion is prominent in the neck below the chin enmmonly called the 'Adam's apple'. Housed within this structure from ";ont are the vocal cords, two folds of ligament and elastic tissue which e brought together or parted by the rotation of the arytenoid cartilages led at the posterior end of the cords) through muscular action. The inner
10   Language and speech
The production of speech    11
From the lungs
Figure I Organs of speech.
edge of these cords is typically about 17 to 22 ram long in males and about 11 to 16 mm in females.5 The opening between the cords is known as the glottis. Biologically, the vocal cords act as a valve which is able to prevent the entry into the trachea and lungs of any foreign body, or which may have the effect of enclosing the air within the lungs to assist in muscular effort on the part of the arms or the abdomen. In using the vocal cords for speech, the human being has adapted and elaborated upon this original open-or-shut function in the following ways (see Fig. 2):
(1) The glottis may be held tightly closed, with the lung air pent up below it. A 'glottal stop' [?] is produced when this closure is suddenly released and occurs in English, e.g. as an energetic onset to a vowel as in apple [?apl] or when it reinforces /p,t,k/ as in clock [kro?k] or even replaces them, as in cotton [kD?n]. It may also be heard in defective speech, such as that arising from cleft palate, when [?] may be substituted for the stop consonants [p,t,k], which, because of the nasal air escape associated with cleft palate, cannot be articulated with proper compression in the mouth cavity.
(2) The glottis may be held open as for normal breathing and for voiceless sounds like [s] in sip and [p] in peak.
(3) The most common action of the vocal cords in speech is as a vibrator set in motion by lung air, which produces voice, or, more technically, phonation;
Thyroid cartilage
enoid cartilages
•tfier as for [?]      (b) open fornormal breathing        (c) loosely together and (a) tightly* fi and voiceless sounds. vibrating as for voiced
sounds.
- "Ml- states of the vocal cords as seen from above.
this vocal cord vibration is a normal feature of all vowels or of such a consonant as [z] in zip which makes them voiced as compared with voiceless is I si in Mp- In order to achieve the effect of voice the vocal cords are ■• .müht sufficiently close together that they vibrate when subjected to the .ippfopriate air pressure from the lungs. This vibration is caused by compressed air forcing the opening of the glottis and the resultant reduced air pressure permitting the elastic cords to come together again.6 The vibration may be felt by touching the neck in the region of the larynx or by putting a finger over jjch ear flap when pronouncing a vowel or [z] for instance. In the typical speaking voice of a man, this opening and closing action is likely to be repeated between 100 and 150 times in a second, i.e. there are that number ;,f cycles of vibration (called hertz, which is abbreviated to Hz); in the case nf a woman's voice, this frequency of vibration might well be between 200 and 325 Hz. We are able, within limits, to consciously vary the speed of \ ibration of our vocal cords in order to change the pitch of the voice; the nore rapid the rate of vibration, the higher the pitch (an extremely low rate of vibration being partly responsible for what is usually called creaky voice). Normally the vocal cords come together rapidly and part more slowly, the Mpening phase of each cycle thus being longer than the closing phase. This gives rise to 'modal' (or 'normal') voice which is used for most English speech. Other modes of vibration result in other voice qualities, most notably breathy and creaky voice which are used contiastively in a number of languages and maybe used in English by some individuals and in some styles (see §5.8). Moreover, we are able, by means of variations in pressure from the lungs, lo modify the size of the puff of air which escapes at each vibration of the > ocal cords; in other words, we can alter the amplitude of the vibration, with a. corresponding change of loudness of the sound heard by a listener. The
12   Language and speech
The production of speech 13
normal human being soon learns to manipulate his glottal mechanism so that most delicate changes of pitch and loudness are achieved. Control o this mechanism is, however, very largely exercised by the ear, so that sue] variations are exceedingly difficult to teach to those who are born deaf, anc a derangement of pitch and loudness control is liable to occur among thosi who become totally deaf later in life. (4) One other action of the larynx should be mentioned. A very quiet whisper may result merely from holding the glottis in the voiceless position through-out speech. But the more normal whisper, by means of which we are able tc communicate with some ease, involves energetic articulation and considerable stricture in the glottal region. Such a whisper may in fact be uttered with an almost total closure of the glottis and an escape of air in the region o the arytenoid7 cartilages.
The simplest way of observing the behaviour of the vocal cords is by the use of a laryngoscope, which gives a stationary mirrored image of the glottis. Using stroboscope techniques, it is possible to obtain a moving record and high-speed films have been made of the vocal cords, showing their action in ordinary breathing, producing voice and whisper, and closed as for a glottal stop. The modern technique of observation is to use fibre-optic endoscopy coupled if required with a tiny videocamera.
2.2.3 The resonating cavities
The airstream, having passed through the larynx, is now subject to further modification according to the shape assumed by the upper cavities of the pharynx and mouth, and according to whether the nasal cavities are brought into use or not. These three cavities function as the principal resonators of the voice produced in the larynx.
2.2.3.1 The pharynx
The pharyngeal8 cavity (see Fig. 1) extends from the top of the trachea and oesophagus, past the epiglottis and the root of the tongue, to the region til the rear of the soft palate. It is convenient to identify these sections of the pharynx by naming them: laryngopharynx, oropharynx, nasopharynx. The shape and volume of this long chamber may be considerably modified by the constrictive action of the muscles enclosing the pharynx, by the movement of the back of the tongue, by the position of the soft palate which may, when raised, exclude the nasopharynx, and by the raising of the larynx itself. The position of the tongue in the mouth, whether it is advanced or retracted, will affect the size of the oropharyngeal cavity; the modifications in shape of this cavity should, therefore, be included in the description of any vowel. It is a characteristic of some kinds of English pronunciation that certain vowels, e.g. the [a] vowel in sad, are
1
eel with a strong pharyngeal contraction. Additionally, a constriction may *t'CI!'d ^between the lower rear part of the tongue and the wall of the pharynx 1)6 ma k.vtifiti with or without voice, is produced, such fricative sounds being
jsture of a number of languages, e.g. Arabic.^
d   h  oharvnx may be observed by means of a laryngoscope or fibre-optic endoscopy and its constrictive actions are revealed by MRI. (See video a 14 -21)
[he escape of air from the pharynx may be effected in one of three ways:
, The soft palate may be lowered, as in normal breathing, in which case the ' ' -tir may escape through the nose and the mouth. This is the position taken up by "the .soft palate in articulation of the French nasalised vowels in such a phrase as un bon vin blanc [ds be ve bid], the particular quality of suc[, vowels being achieved through the resonance of the nasopharyngeal cavities. There is no absolute necessity for nasal airflow out of the nose, the most important factor in the production of nasality being the sizes of the posterior oral and nasal openings (some speakers may even make the nasal cavities vibrate through nasopharyngeal mucus or through the soft palate itself).10
(2) The soft palate may be lowered so that a nasal outlet is afforded to the airstream, but a complete obstruction is made at some point in the mouth, with the result that, although air enters all or part of the mouth cavity, no oral escape is possible. A purely nasal escape of this sort occurs in such nasal consonants as [m,n,rj] in the English words ram, ran, rang. In a snore and some kinds of defective speech, this nasal escape may be accompanied by friction or a trill between the rear side of the soft palate and the pharyngeal wall.
(3) The soft palate may be held in its raised position, eliminating the action of the nasopharynx, so that the air escape is solely through the mouth. All English sounds, with the exception of the nasal consonants mentioned in (2), usually have this oral escape. Moreover, if for any reason the lowering of the soft palate cannot be effected, or if there is an enlargement of the organs enclosing the nasopharynx or a blockage brought about by mucus, it is often difficult to articulate either nasalised vowels or nasal consonants. In such speech, typical of adenoidal enlargement or the obstruction caused by a cold, the French phrase mentioned above would have its nasalised vowels turned into their oral equivalents and the English word morning would have its nasal consonants replaced by [b,d,g] becoming [boidig]. On the other hand, an inability to make an effective closure by means of the raising of the soft palate—either because the soft palate itself is defective or because :an abnormal opening in the roof of the mouth gives access to the nasal cavities—will result in the general nasalisation of vowels and the failure to articulate such oral stop consonants as [b,d,g]. This excessive nasalisation (or hypernasality) is typical of such a condition as cleft palate.
Y4   Language and speech
The production of speech 15
The action of I he soft palate can be observed by MRI. (See video 6.4ff.) The pressure of the air passing through the nasal cavities may be measured at the nostrils or within the cavities themselves.
2.2.3.2 The mouth
Although all the cavities so far mentioned play an essential part in the production of speech sounds, most attention has traditionally been paid to the behaviour of the cavity formed by the mouth. Indeed, in many languages the word for 'tongue' is used to refer to our speech and language activity. Such a preoccupation with the oral cavity is due to the fact that it is the most readily accessible and easily observed section of the vocal tract. The shape of the mouth determines finally the quality of the majority of our speech sounds. Far more finely controlled variations of shape are possible in the mouth than in any other part of the speech mechanism.
The only boundaries of the mouth which are relatively fixed are, in the front, the teeth; in the upper part, the hard palate; and, in the rear, the pharyngeal wall. The remaining organs are movable: the lips, the various parts of the tongue and the soft palate with its pendant uvula {see Fig. 1). The lower jaw is capable of very considerable movement; its movement will control the gap between the upper and lower teeth and also to a large extent the disposition of the lips. Movement of the lower jaw is also one way of altering the distance between the tongue and the roof of the mouth.
It is convenient for descriptive purposes to divide the roof of the mouth into three parts: moving backwards from the upper teeth, first, the teeth ridge (adjective: alveolar)" which can be clearly felt behind the teeth; second, the bony arch which forms the hard palate (adjective: palatal) and which varies in size and arching from one individual to another; and finally, the soft palate (adjective: velar) which, as we have seen, is capable of being raised or lowered, and at the extremity of which is the uvula (adjective: uvular).12 All these parts can be readily observed by means of a mirror.
(1) Of the movable parts, the lips (adjective: labial), constitute the final obstruction to the airstream when the nasal passage is shut off. The shape which they assume affects very considerably the shape of the total cavity. They may be shut or held apart in various ways. When they are held tightly shut, they form a complete obstruction or occlusion to the airstream, which may either be momentarily prevented from escaping at all, as in the initial sounds of pat and bat, or may be diverted through the nose by the lowering of the soft palate, as in the initial sound of mat. If the lips are held apart, the positions they assume may be summarised under five headings:
(a) held sufficiently close together over all their length that friction occurs between them. Fricative sounds of this sort, with or without voice, occur in many languages and the voiced variety [p] is sometimes wrongly used by foreign speakers of English for the first sound in the words vet or wet;
.. i .sufficiently far apart for no friction to be heard, yet remaining fairly Ji'^t'iosether and energetically spread. This shape is taken up for vowels 1|'-. hat" in see when energetically enunciated and is known as the spread i, position;
held in a relaxed position with a lowering of the lower jaw. This is the ^' ■ o=i. ion 'taken up for the vowel of sat and is known as the neutral position; tiahtly pursed, so that the aperture is small and rounded, as in a precise or ' '-nuiietic enunciation of the vowel of do, or more markedly in the French \,iv. Jl'of drmx. This is the close rounded position; „■ hel> wide apart, but with slight projection and rounding, as in the vowel of , : This is the open rounded position.
. , kn. of these five positions may be encountered, e.g. in the vowel of saw, ■•ir wh eh a type of lip-rounding between open rounded and close rounded is commonly used. It will be seen from the examples given that lip position is particularly significant in the formation of vowel quality. English consonants, on t -t. t.i ier hand, even including [p,b,m,w] whose primary articulation involves lip iK-ion "ill lend to share the lip position of the adjacent vowel. In addition, the \'v.t lif) is an active articulator in the pronunciation of [f,v], a light contact being made between the lower lip and the upper teeth.
I? i C >l" all the movable organs within the mouth, the tongue is by far the most lle\ible and is capable of assuming a great variety of positions in the articulation o- both vowels and consonants. The tongue is a complex muscular structure ch does not show obvious sections; yet, since its position must often be dL-oribed in considerable detail, certain arbitrary divisions are made. When the lointie is at rest, with its tip lying behind the lower teeth, that part which lies opposite the hard palate is called the front and that which faces the soft palate ■< mailed the back, with the region where the front and back meet known as the centre (adjective: central). These areas together with the root (which forms ili.- front side of the pharynx) are sometimes collectively referred to as the body ot the tongue. The tapering section in front of the body and facing the teeth ridge is called the blade (adjective: laminae)13 and its extremity the tip (adjective: apical); The edges of the tongue are known as the rims.
i renerally, in the articulation of vowels, the tongue tip remains low behind the lower teeth. The body of the tongue may, however, be 'bunched up' in different ■.«.r.s, e.g. the front may be the highest part as when we say the vowel of keen («* video 11.5); or the back may be most prominent as in the case of the vowel in zoo (see video 2.28); or the whole surface may be relatively low and flat as in the case of the vowel m father (see video 10.18). Such changes of shape can be felt if the above words are said in succession. These changes, together with the variations in lip position, have the effect of modifying very considerably the size of the mouth cavity and of dividing this chamber into two parts: that part of the cavity which is in the forward part of the mouth behind the lips and that which, is in the rear in the region of the pharynx.
í 6   Language and speech
The production of speech 17
The various parts of the tongue may also come into contact with the roof of the mouth. Thus, the tip, blade and rims may articulate with the teeth as for the th sounds in English (see video 12.16), or with the upper alveolar ridge as in the case of /t,d,s,z,n/ (see videos 3.15,23, 2.17, 5.16, 6.7) or the apical contact may be only partial as in the case of/I/ (where the tip makes firm contact whilst the rims make none—see the palatooram of III in Figures 48 and 49) or intermittent in a trilled Irl as in some forms of Scottish English. In some languages, notably those of India, Pakistan and Sri Lanka, the tip contact may be retracted to the very back of the teeth ridge or even slightly behind it; the same kind of retroflexion, without the tip contact, is typical of some kinds of English hi, e.g. those used in south-west England and by some speakers in the U.S.
The front of the tongue may articulate against or near to the hard palate. Such a raising of the front of the tongue towards the palate is an essential part of the [£3] (see videos 1.1,11.23) sounds in English words such as show and measure, being additional to an articulation made between the blade and the alveolar ridge: or again, it is the main feature of the [j] sound initially in yield.
The back of the tongue can form a total obstruction by its contact with the front side of the soft palate, the nasopharynx being closed in the case of [k.g], e.g. in guard (see video 12.1) and open for [n] as in hang (see video 10.4); or again, there may merely be a narrowing between the soft palate and the back of the tongue, so that friction of the type occurring finally in the Scottish pronunciation of loch is heard. And finally, the uvula may vibrate against the back of the tongue, or there may be a narrowing in this region which causes uvular friction, as at the beginning of the French word rouge.
It will be seen from these few examples that, whereas for vowels the tongue is generally held in a position which is convex in relation to the roof of the mouth, some consonant articulations, such as the southern British English Irl in ride (see video 4.2) and the /]/ in crawl (see video 2.9), will involve the 'hollowing' of the body of the tongue so that it has, at least partially, a concave relationship with the roof of the mouth.
Moreover, the surface of the tongue, viewed from the front, may take on various forms: there may be a narrow groove running from back to front down the mid line as for the Isl in see, or the grooving may be very much more diffuse as in the case of the /// in ship.
(3) The oral speech mechanism is readily accessible to direct observation as far as the lip movements are concerned as are many of the tongue movements which take place in the forward part of the mouth. Although a lateral view of the shape of the tongue over all its length and its relationship with the palate and the velum can be obtained by MRI, it should not be assumed that pictures of the articulation of, say, the vowel in cat will show an identical tongue position when pronounced by different people. Not only is the sound itself likely to be slightly different from one individual to another, but, even if the sound is for all practical purposes the 'same', the tongue positions may be different, since the shape of the mouth cavity is not identical for any two speakers; and, in any case,
1 mďs judged to be the same may be produced by the same individual with \ hat different articulations. When, therefore, we describe an articulation in "l"'rl' i 1 should be understood that such an articulation is typical for the sound
-=non- but that variations are to be expected. m Palaiogfaphy. showing the extent of the area of contact between the tongue and .. ,,iófof the mouth, has long been a practical and informative way of recording "    11 ■ 1 idvements. At one time the palate was coated with a powdery substance, !i!.~ľ filiation was made, and the 'wipe-off' subsequently photographed. But the idem method uses electropalatography, whereby electrodes on a false palate 'spur■ I io am tongue contact, the contact points being simultaneously registered t v i -uaľ display. This has the advantage of showing a series of representations f the charming contacts between the tongue and the top of the mouth during -■l.ii Electi'opalatograms of this sort are used to illustrate the articulations of Jopm'I .nits in Chapter 9.
Articulators used in speech
Relevance
Airstream Vocal folds So*-. osiK'. Tongue; :: -I'"" ť
Usually pulmonic   ' :; :
Closed, wide apart, or vibrating
Lowered (giving nasality) or raised (excluding nasality)
Back, centre.front, blade, tip and/01* rims raised
Neutral, spread, open-rounded, close-rounded'
Notes
1 Rut see, for example, Fouts & Mills (1999), for the training of chimpanzees to use a rudimentary language (but signed rather than spoken).
2 Mason (1951).
i I. .U'oged (1967).
h -I is chapter and in Chapters 4, 8 and 9, videos on the companion website are i,-v -'need as, for example, video 12.4, where 12 is the number of the video and 4 is 1 ■■ p 'levant point an the video. See Cruttenden (2013) for some discussion of the use of these: videos in teaching.
5 Chirk et al. (2007: 178).
6 According to the 'Bernoulli Effect'.
7 Pronounced /a'rrtanoid/ or /ari'timoid/.
8 Pronounced /'fannks/ and /f-f rmd3i-)l/ or /fann'o^ii-jl/.
■'In .v it research has shown that the pharyngeal consonants in Arabic are more correctly
described as epigloltal (Ladefoged & Maddieson, 1996: 167). 10 Layer (1980: 77ff). U Pronounced /alvi'auta/ or /al'vi:-)!-)/. 12 Pronounced /'ju:vju:b/.
13 Pronounced /Taminal/ and /'eipiksl/.
Chapter 3
The sounds of speech: the acoustic § and auditory aspects
The sounds of speech 19
3.1 Sound quality \
To complete an act of communication, it is not sufficient that our speech -mechanism should produce sounds; these have to be heard and interpreted, JB after transmission through a medium, normally the air, which is capable of conveying sounds. We now examine the nature of the sounds which we hear, ihe characteristics of the transmission phase of these sounds and the way in which 1 these sounds are perceived by a listener. T
When we listen to a continuous utterance, we perceive an ever-changing pattern * of sound. When it is a question of our own language, we are not conscious of alii the complexities of pattern which reach our ears: we tend consciously to perceive and interpret only those sound features which are relevant to the intelligibility •■ of our language. Nevertheless, despite this linguistic selection which we ultimately make, we are aware that this changing pattern consists of variations of different ■' kinds: of sound quality—we hear a variety of vowels and consonants; of pitch— '. we appreciate the melody, or intonation, of the utterance; of amplitude—some sounds or syllables sound 'louder' than others; and of length—some sounds will Jk be longer to our ears than others. These are judgements made by a listener about : a sound continuum from a speaker and, if the sound from the speaker and the response from the listener are made in the same linguistic system, then the utterance will be meaningful for speaker and listener alike. The transmission phase links the listener's impressions of changes of quality, pitch, amplitude and length with the articulatory activity on the part of the speaker. But an exact ' correlation between the production, transmission and reception phases of speech is not always easy to establish. ;
The production of sound requires some kind of energy and frequently this f energy makes something vibrate. In the case of human speech the thing vibrating is usually the vocal cords which are energised by air pressure from the lungs. Any such sound produced in the larynx is then modified by the resonating chambers of the pharynx, the mouth and, in some cases, the nasal cavities. The listener's impression of sound quality will be determined by the way in which the speaker's ; vibrator and resonators function together. f
h blinds- like other sounds, are conveyed to our ears by means of waves "*lw'1^ ,^ nnand rarefaction of the air particles (the commonest medium cf CUIT,"i'[1K,ljj0T))i These variations in pressure, initiated by the action of the 0f oil"' ■■ ^ pacil|e(j m an directions from the source, the air particles fhem-^''"'"-'''ibrjiiria at Ihe same rate (or frequency) as the original vibrator. Ln speech, V" inon.i may be of a complex but regular pattern, producing 'tone' ,"uS<". in i» be heard in a vowel sound; or they may be of an irregular kind, '""''j i~ ii'" i'.iisC. such as in the consonant Is/ or there may be both regular and P '' 1"||ir\,biations present, i.e. a combination of tone and noise, as in Izl. In 'he production of normal vowels, the vibrator is provided by the vocal cords; in of many consonant articulations, however, a source of air disturbance i ided by constriction at a point above the larynx, with or without accom-''■ ",|t■■vocal cord vibrations.
1 i-'-nite the fact that the basis of all normal vowels is the glottal tone, we are all capable of distinguishing a large number of vowel qualities. Yet the glottal vibrations in the case of [a:] are not very different from those for [i:], when L-\. wwels arc said with the same pitch. The modifications in quality which >u i!ei<.:tve are due to the action of the supraglottal resonators which we have p-e.!l'U-ly described. To understand this action, it is necessary to consider a little m.iir closely the nature of the glottal vibrations.
1.1„; already been mentioned that the glottal tone is the result of a complex, IV niuiiily regular, vibratory motion. In fact, the vocal cords vibrate in such a ■m> t- to produce, in addition to a basic vibration over their whole length (the FUNDAMfl wtal frequency), a number of overtones or harmonics having frequencies which are simple multiples of the fundamental or first harmonic. Thus, if there ... j iLxdamerital frequency of vibration of 100 Hz, the upper harmonics will b-j i'f iheorder of 200, 300, 400 Hz. Indeed, there may be no energy at the iu.i.hrrufital frequency, but merely the harmonics of higher frequency such as 200. 300, 400 Hz. Nevertheless, we still perceive a pitch which is appropriate m j iLiidamental frequency of 100 Hz, i.e. the fundamental frequency is the Liahew common factor of all the frequencies present, whether or not it is present inJI -
' he number and strength of the component frequencies of this complex glottal tune differ from one individual to another and this accounts at least in part for the differences of voice quality by which we are able to recognise speakers. H.if we can all modify the glottal tone so as to produce vowels as different as fi:l and [a: |, so that despite our divergences of voice quality we can convey the ci-tinction between two words such as key and car. This variation of quality, or timbre, of the glottal tone is achieved by the shapes which we give the resonators .i iove the larynx—the pharynx, the mouth and the nasal cavities. These chambers j 'i capable of assuming a very large number of shapes, each of which will have i vharacteristic vibrating resonance of its own. Those harmonics of the glottal lurie which coincide with the chamber's own resonance are very considerably amplified: Thus, certain bands of strongly reinforced harmonics are characteristic
20   Language and speech
of a particular arrangement of the resonating chambers which produces, f0 instance, a certain vowel sound. Moreover, these bands of frequencies wiil bi reinforced whatever the fundamental frequency, in other words, whatever thi pitch on which we say, for instance, the vowel [a:], the shaping of the resonator and their resonances will be very much the same, so that it is still possible except on extremely high or low pitches, to recognise the quality intended ]( is found that, for male speakers, the vowel [v.] has one such characteristic band of strong components in the region of 280 Hz and another at about 2,200 Hz: while for female speakers these bands of energy are at about 300 Hz and 2,700 Hz (see §8.6). ■
3.2 The acoustic spectrum
This complex range of frequencies of varying intensity which go to make up the quality of a sound is known as the acoustic spectrum, those bands of energy; which are characteristic of a particular sound are known as the sound's formants Thus, formants of [a;] are said to occur, for female speakers, in the regions around 700 and 1,300 Hz.
Such complex waveforms can be analysed and displayed as a spectrogram; (see Fig. 3). Originally this display required a special instrument, a spectrograph, but nowadays it is generally done by computer. The spectrogram consists of a three-dimensional display: frequency is shown on the vertical axis, time on the horizontal axis and the energy at any frequency level either by the density of blackness in a black and white display, or by colours in a colour display. Thus the concentrations of energy at particular frequency bands (the formants) stand out very clearly. Fig. 3 shows, in the spectrogram of Manchester music shops, the extent to which utterances are not neatly segmented into a succession of sounds but, on the contrary, there is considerable overlap. Such spectrographs analysis provides a great deal of acoustic information in a convenient form.
The sounds of speech 21
			f	11
MM 11 m' i/	s		'#1 |t	
vf< r < \ '\	4 :i		k	
'Hi i				
0    100  200  300 400  500 600		700	aoo wo	1,000 1,100 1,200 1,300 1,400 msec
/ m k n f est		j u:	z   i k	J       D         p S/
4,500 Hz
3,600 2,700 1,800 900
figure 3 Spectrogram of the phrase Manchester music shops as said by a male speaker of GB.
'■: {the information given on a spectrogram is, in fact, irrelevant to our lnr'bf speech and we need to establish which elements of the spectrum jnJfr,r^   (|   -speech communication. For instance, two, or at most three, for-- near to be- sufficient for the correct identification of vowels. As far as "iiin.1" ^|?■ jj v(jwels are concerned, the first three formants are all included in the , |ivc o - 4.000 Hz, so that the spectrum above 4,000 Hz would appear "'^behufi^y irrelevant to the recognition of our vowels. In both older (analogue) ■° " "| ..nig systems with a frequency range of approximately 300-3,000 Hz and ■''   -^r-f (di"ifai) systems with a range up to around 4,000 Hz, we have little (" |   in identifying the sound patterns of speakers and even in recognising " hiv ill ial voice qualities. Indeed, when we are dealing with a complete utterance '.\ i uteri'context, where there is a multiplicity of cues to help our under--tiii'ii"1"' a high degree of intelligibility may be retained even when there are no frequencies above 1,500 Hz.
\^ o:ie would suspect, there appear to be certain relationships between the formants of vowels and the cavities of the vocal tract (i.e. the shapes taken on - i i -sonators, notably the relation of the oral and pharyngeal cavities). Thus, the first formant appears to be low when the tongue is high in the mouth: e.g.
1 jmI [us] have high tongue positions and have first formants for both men and «imui around 280-330 Hz, whereas [a:] and [n] have their first formants in ■ie rci;i jn 600-800 Hz, their tongue positions being relatively low. On the other ^ -i,! rhe Second formant seems to be inversely related to the length of the front z*\ i.' thus [it], where the tongue is raised high in the front of the mouth, .i-s i -jcond formant around 2,200-2,700 Hz, whereas [ui], where the tongue is raised at the back of the mouth (and Hps are rounded), has a relatively low second formant around 1,200-1,400 Hz. (For averages for the first and second formants of all GB vowels for male and female speakers see Table 3 and Figs 10 and II in-Chapter 8.)
It is also confirmed from spectrographs analysis that a diphthong, such as that i.i , n is indeed a glide between two vowel elements (reflecting a perceptible .uliuiLlory movement), since the formants bend from those positions typical of .in. i-iwel to those characteristic of another (see Table 4 and Fig. 9 in Chapter 8).
For many consonant articulations (e.g. the initial sounds in pin, tin, kin, thin, Jm, sin, shin, in which the glottal vibrations play no part) there is an essential noise, component, deriving from an obstruction or constriction within the mouth, •('proximately within the range 2,000-8,000 Hz (see Chapter 9, Fig. 32). This noise-component is also present in the corresponding articulations in which \ncal fold excitation is present, as hi the final sounds of ruse and rouge, where •'-.i are dealing with sounds which consist of a combination of glottal tone and .wise: Detailed acoustic data concerning vowel and consonant articulations in HI3 is given in Chapters 8 and 9.
Spectrogiaphic analysis also reveals the way in which there tends, on the .■cmistic. level, to be a merging of features of units which, linguistically, we treat separately. Thus, our discrimination of [f ] and [6] sounds depends only partly
22   Language and speech
The sounds of speech 23
on the frequency and duration of the noise component but also upon a characteristic bending of the formants of the adjacent vowel. Indeed, in the case of such consonants as [p,t,k], which involve a complete obstruction of the airstream v and whose release is characterised acoustically by only a very brief burst of noise, the transition between the noise of the consonant and the formant structure of the vowel appears to be of prime importance for our recognition of the consonant (see §9.2.2(3) and Fig. 32). '
3.2.1 Fundamental frequency: pitch
Our perception of the pitch of a speech sound depends directly upon the frequency of vibration of the vocal cords. Thus, we are normally conscious of the pitch caused by the 'voiced' sounds, especially vowels; pitch judgements made on voiceless or whispered sounds, without the glottal tone, are limited in comparison with those made on voiced sounds, but may be induced by producing friction in the larynx or pharynx and varying the shape of the resonating cavities in the usual way. *
The higher the glottal fundamental frequency, the higher our impression of ; pitch. A male voice may have an average pitch level of about 120 Hz and a female voice a level in the region of 220 Hz.1 The pitch level of voices, how- w ever, will vary a great deal between individuals and also within the speech of one speaker, the total range of one male speaking voice being liable to have a range of anything between 80 and 350 Hz. Following adolescence men's voices become lower until the forties after which they rise continuously into old age; " women's voices generally become lower up to the time of the menopause, remain 3 much the same for twenty years or so and then possibly rise slightly.2 Yet our : .-perception of frequency extends further than the limits of glottal fundamental frequency, since our recognition of quality depends upon frequencies of a much higher order. In fact, the human ear perceives frequencies from as low as 16 11/, to about 20,000 Hz and in some cases even higher. As one becomes older, this upper limit may fall considerably, so that at the age of 50 it may extend no higher '■' than about 10,000 Hz. As we have seen, such a reduced range is no impediment to perfect understanding of speech, since a high percentage of acoustic cues for 8 speech recognition fall within the range 0-4,000 Hz. »
Our perception of pitch is not, however, solely dependent upon fundamental : frequency. Variations of intensity on the same frequency may induce impressions of a change of pitch; and conversely, tones of very high or very low frequency, ^ if they are to be audible at all, require greater intensity than those in a middle range of frequencies.3
Instrumental measurement of fundamental frequency based on signals received through a microphone employs two general methods. The first is to count the r number of times that a particular pattern is repeated within a selected segment of a waveform such as that provided on an oscillogram. The second is to track : the progress of the fundamental frequency on a spectral display like that provided „
, i spectrogram, or, alternatively, to track the progress of a particular harmonic !,.. divide by the relevant number. Nowadays various computer programs are *.\ i I ihle1 which average the results from a range of measurements based on the
.L. ■■incral methods noted above. But even with such sophisticated programs ■ li'ii1 .ire still likely to be occasional mistakes like octave jumps (each doubling 'nl|. ^presenting an octave).
■A" third method of fundamental frequency extraction involves direct measurement of the vibration of the vocal cords either by glottal illumination or by electroglottography. One well-known technique in the latter class involves using
e^bykgograph.s With this technique electrodes are attached to the outside of the throat and the varying electrical impedance is monitored and projected onto a visual display. The signal generated by the variation in impedance can also be stored, enabling this technique to be used outside the laboratory.
Measures of fundamental frequency do not always correspond to our auditory perception of pitch. Besides the dependence on intensity mentioned above, dif--erent segments affect the fundamental frequency in different ways: for example, other things being equal, an [i] will have a higher fundamental frequency than .,n [a|. and a [p] will produce a higher frequency on a following vowel than a [b]. Such (slight) changes in frequency will generally be undetectable by the ear. As in many other cases of instrumental measurement, we still have to use our ..uditory perception to interpret what instruments tell us.
3.2.2 intensity: loudness
Our sensation of the relative loudness of sounds may depend on several factors, i .fcsound or syllable may appear to stand out from its neighbours—be 'louder' —because a marked pitch change is associated with it or because it is longer than its neighbours. It is better to use a term such as prominence to cover these listener impressions of variations in the perceptibility of sounds. More Mr c lj. what is 'loudness' at the receiving end should be related to intensity at 'Iv pi i duction stage, which in turn is related to the size or amplitude of the i bration. An increase in amplitude of vibration, with its resultant impression of ■reater loudness, is brought about by an increase in air pressure from the lungs. U:we shall see (§10.2), this greater intensity is not in itself usually the most uiporumt factor in rendering a sound prominent in English. Moreover, all other "I'ings being equal, some sounds appear by their nature to be more prominent or ■■"inorous than others, e.g, the vowel in barn has more carrying power than that J bean and vowels generally are more powerful than consonants.
The judgements we make concerning loudness are not as fine as those made "•>r either quality or pitch. We may judge which of two sounds is the louder, but "v" find it difficult to express the extent of the difference. Indeed we generally perceive and interpret only gross differences of overall loudness, despite the uvct that we recognise different vowels by reacting to characteristic regions of 'ii.:ensity in the spectrum.
24   Language and speech
3.2.3 Duration; length
In addition to affording different auditory impressions of quality, pitch and loudness, sounds may appear to a listener to be of different length. Clearly, whenever it is possible to establish the boundaries of sounds or syllables, it will be possible to measure their duration on traces provided by oscillograms or spectrograms. Such delimitation of units, in both the articulatory and acoustic sense, may be difficult, as we shall see when we deal with the segmentation of the utterance. But, even when it can be done, variations of duration in acoustic terms may not correspond to our linguistic judgements of length. We shall, for instance, refer later to the ' long' vowels of English such as those of bean and barn, as compared with the 'short' vowel in bin. But, in making such statements, we shall not be referring to absolute duration values, since the duration of all vowels will vary considerably from utterance to utterance, according to factors like whether the utterance is spoken rapidly or slowly, whether the syllable containing the vowel is accented or not and whether the vowel is followed by a voiced or voiceless consonant. In the English system, however, we know that no more than two degrees of length are ever linguistically significant and all absolute durations will be interpreted in terms of this relationship. This distinction between measurable duration and linguistic length provides another example of the way in which our linguistic sense interprets from the acoustic material only that which is significant.
The sounds comprising any utterance will have varying durations and we will have the impression that some syllables are longer than others. Such variations of length within the utterance constitute one manifestation of the rhythmic delivery which is characteristic of English and so is fundamentally different from the flow of other languages, such as French, where syllables tend to be of much more even length.
As already mentioned, the absolute duration of sounds or syllables will depend, among other things, upon the speed of utterance. An average rate of delivery might contain anything from about 6 to 20 sounds per second, but lower and much higher speeds are frequently used without loss of intelligibility. The time required for the recognition of a sound will depend upon the nature of the sound and its pitch, vowels and consonants differing considerably in this respect, but it seems that a vowel lasting only about 4 msecs may have a good chance of being recognised.
3.2.4 'Stress'
We have purposely avoided the use of the word 'stress' in this chapter because this word has been used in different and ambiguous ways in phonetics and linguistics. It has sometimes been used as simply equivalent to loudness, sometimes as meaning 'made prominent by means other than pitch' (i.e. by loudness or length) and sometimes as referring just to syllables in words in the lexicon
The sounds of speech 25
. .-anihfl something like 'having the potential for accent on utterances'. "l , 1"hout"this book we will avoid use of the term 'stress' altogether, using '"-nee as the general term referring to segments or syllables, sonority as ' -l'.rricular term referring to the carrying power of individual sounds and 'L "       referring to those syllables which stand out above others, either in
ACCFM as
jui.lual words or in longer utterances.
3 3 Hearing
nur hearing mechanism must be thought of in two ways: the physiological mechanism which reacts to the acoustic stimuli—the varying pressures in the air ihich constitute sound; and the psychological activity which, at the level of the brain, selects from the gross acoustic information that which is relevant in terms of the linguistic system involved. In this way, measurably different acoustic stimuli inav be interpreted as being the 'same' sound unit. As we have seen, only nart of the total acoustic information seems to be necessary for the perception j "pail iidar sound values. One of the tasks which confront the phonetician is il.e ,-i=wiitanglement of these relevant features from the mass of acoustic material that modem methods of sound analysis make available. The most fruitful tech-i itic for discovering the significant acoustic cues is that of speech synthesis, controlled by listeners'judgements. After all, the sounds [a:] and [s] are [cu] and fs| only if listeners recognise them as such. Thus, it has been established that iii.l) two formants are necessary for the recognition of vowels, because machines which generate sound of the appropriate frequency bands and intensity produce \p«els which are correctly identified by listeners.
Listeners without any phonetic training can, therefore, frequently give valuable guidíince by their judgements of synthetic qualities. But it is important to be ■iv are of the limitations of such listeners, so as to be able to make a proper ji,[Hialion of their judgements. A listener's reactions are normally conditioned by h r- experience of handling his own language. Thus, if there are only five significant vowel units in his language, he is liable to allow a great deal more latitude in his assessment of what is the 'same' vowel sound than if he has twenty. An 'Englishman, for instance, having a complex vowel system and being accustomed to distinguishing such subtle distinctions as those in sit, set, sat, will be fairly precise in his judgement of vowel qualities. A Spaniard, however, whose vowel system is made up of fewer significant units, is likely for this reason to be more tolerant of variation of quality. Or again, if a listener is presented with a system of synthetic vowels which is numerically the same as his own, he is able to make allowance for considerable variations of quality between his and the synthetic system and still identify the vowels correctly—by their 'place' in a system rather than by their precise quality; this is partly what he does when he listens to and understands his language as used by a speaker of a different dialect.
Our hearing mechanism also plays an important part in monitoring our own : speech; it places a control upon our speech production which is complementary
26   Language and speech
to our motor, articulatory, habits. If this feedback control is disturbed, e.g. bj the imposition of an artificial delay upon our reception of our own speech, dfe turbance in the production of our utterance is likely to result. Those who are born deaf or who become deaf before the acquisition of speech habits are rarely able to learn normal speech completely; similarly, a severe hearing loss later hi life is likely to lead eventually to a deterioration of speech, although not down to the same level as those born deaf.
Chapt_er_i__ _-
fhe description and classification ©f speech sounds
Notes
1 Fant(1956).
2 Hollien & Shipp (1972); Russell et at. (1995).
3 Denes & Pinson (1993: lOlff).
4 For example, Praat (Boersma & Weenink, 2011).
5 Abberton & Fourcin (1984).
4 1 phonetic description
\\i h i,c;considered briefly both the mechanism which produces speech sounds =n i iir some of the acoustic and auditory characteristics of the sounds them-4 j We have seen that a speech sound has at least three stages available for 'i»^«-.imtions—the production, transmission and reception stages. The most (vriifnijilt arid brief descriptive classification of speech sounds relies either on articulatory criteria or on auditory judgements, or on a combination of both. Those sounds which are commonly known as consonants are most easily described mainly in terms of their articulation, whereas the description of vowels depends more on auditory impressions.
4.1 Vowel and consonant
pio types of meaning are associated with the terms vowel and consonant. In one type of definition consonants are those segments which, in a particular hrguage, occur at the edges of syllables, while vowels are those which occur at Lie centre of syllables. So, in red, wed, dead, lead, said, the sounds represented bv <r,w,d,l,s> are consonants, while in beat, bit, bet, but, bought, the sounds represented by <ea,i,e,u,ou> are vowels. This reference to the functioning of sounds in syllables in a particular language is a phonological definition. But once any attempt is made to define what sorts of sounds generally occur in these different syllable-positions, then we are moving to a phonetic definition. This type of definition might define vowels as median (air must escape over the ■middle of the tongue, thus excluding the lateral [1]), oral (air must escape through the mouth, thus excluding nasals like [n]), frictionless (thus excluding fricatives like [s]), and continuant (thus excluding plosives like [p]); all sounds excluded from this definition would be consonants. But difficulties arise in English with this definition (and with others of this sort) because English /j,w,r/, which are consonants phonologically (functioning at the edges of syllables), are vowels phonetically. Because of this these sounds are often called semi-vowels. The reverse type of difficulty is encountered in words like sudden and little where