Dept. for Speech, Music and Hearing

Quarterly Progress andStatus Report

Larynx height and voicesource. A relationship?

Sundberg, J. and Askenfelt, A.

journal: STL-QPSRvolume: 22number: 2-3year: 1981pages: 023-036

http://www.speech.kth.se/qpsr

INTRODUCTION

I n normal speech the v e r t i c a l p s i L i o n o f t h e larynx o f t en rises a s t h e

fundamental frequency of phonation is raised. I ~ I sin(j inc~, O I I the o the r

hand, phonation frequency dependent changes i n 'larynx 'heighthavebeen

found to be small or even o f reversed s ign , so t-hat t h e larynx drops

s l ight ly with r is ing frequency of phonation ( s h i p & Izdehski 1975). Moreo-

ver, most sirrging teachers agree that a raising of the larynx with r is ing

pitch is harmful to the voice.

A r a i s e d larynx w i l l a f f e c t t h e vowel qual i ty . One reason f o r t h i s is t h e

fact tha t a larynx rise resul t s in raised formant freq~~encies, because of

t h e shortening of t h e vocal t r a c t ( see e.g. Sundberg r; Nordstrom 1976).

Fbwever, there may be other effects of a raised larynx as w e l l . The larynx

is a highly complicated structure the exact physiolqy c?E which we do not

y e t understand thoroughly. The 1 n u s c l . e ~ 11sec1 f o r r a i s i n g t h e larynx may

a f fec t the way in which the vocal folds vibrate. In other wnnqs, there may

be a relationship between larynx height and the voice source.

In a previous investigation (Sundberg & Gauffin 1979) such a relationship

was observed in a male sirrger phonating with a deliberately varied 1 arynx

pos i t ion . Under condi t ions of ra ised larynx, t h e waveform oE h i s voice

source t y p i c a l l y changed i n t h a t i ts ampl i t wle decreased and t h e closed

phase was extended, see Fig. 111-B-1.

This particular type of voice source waveform w a s fomd t_o characterize a

spec ia l type of phonation, which w e have c a l l e d "press" phonaticm in

previous investigations. It invol ves a relat ively high subqlottic pressure

and, a s may be in fe r red from t h e associated, long closed phase, a high

adduction ac t iv i ty i n the larynx. I f the suhglottic pressure and the atlduc-

t ion act lv l ty are relt~tx3, phoplatjc~n changes townnls ar13t ller extreme, which

we have called "flow" phonation. Its glottal waveform characteristics are a

s h o r t e r closed phase and a r a t h e r high amplitude, .+s can also be seen in

Fig. 111-B-1. I I

Normal phonation 71 dB SPL

1 10 msec ----------)

Pressed 68 dB SPL

Flov "L .C.

Raised larynx 70 dB SPL

Lowered larynx

Fig. 1 1 1 1 . Glottal v o l w velocity waveform (i .e . , glottograms) during sus- tained -1 phonation in a singer subject uho deliberately va- ried his type of phonation. P and SPL give the subglottic pres- sure (as determined from the oral pressure during the occlusion for a p-sound) and the sound pressure level .5 m in front of the muth. Adapted from Sundberg & Gauffin, 1979.

These waveform characteristics shaw a high correlation with certain acuxls-

t ic properties of the voice source spectrum, a s i l lus t ra ted i n Fig. 111-B-2. There is a high mrrelat ion between tl-rr cjlottogram amplitude an11 the ampli-

tude of the voice source fundamental. Moreover, tl~t.? mpli tude of the diffe-

r e n t i a t e d glottogram waveform shows a high c o r r e l a t i o n with t h e sound

pressure l e v e l of t h e vowel produced, provided t h a t the inf luence of t h e

formant frequencies on t h i s level is compensated for, see &&fin & Sund-

berg (1980).

Given this information, we can conclude tha t press phonation is manifested

a c o u s t i c a l l y by a weak amplitude o f the source spectrum fundamental, as

shown in Fig. 111-B-3. Note also tha t the professional sinqer increases the

ampl i to ie of t h e source spectrum Fundamental a t almost exact ly L l ~ e same

r a t e a s the sound pressure level of the vowel. The mn-singer, on the other

hand increases h i s furdamental amp1 i t d e less than h i s sourvl pressure level

under t h e same condit ions, which means t h a t he changes h i s phonation

towards the press extreme, as he increases h i s vocal effort .

I.t does not. s s e m unlikely tha t a larynx rise typically i iffects the male of

pbnat ion i n the way i l lustrated i n ~ i ~ . 111-B-1. The n u m s e of the ~ r e s e n t

inves t iga t ion was t o f ind ou t i f a ra ised larynx p s i t i o n is t y p i c a l l y

associated with a "press" type of phonation.

One way to inves t iga te t h i s problem apparent ly would be to simp1 y ask

subjects to sing with deliberately varied larynx height aid then to analyze

their formant frequencies and voice source properties. However, w e do not

seem to have a very eff ic ient control (we]- the voice organ when we use i t .

i n unfamil iar ways. For instance, t h e e f f e c t on g l o t t a l volume v e l o c i t y

waveform of a lowering of the larynx was fo1u~1 tn be quite different in a

t r a i n e d s inger sub jec t and a less t ra ined subject . Thus, t h e phonatory

behavior of a subject seems to be reasonably m l s i s t e n t only 4s lory as the

subject is performing tasks, to which he i.s acquainted. Hence, data collec-

La3 f r a n subjects deliberately raising the i r larynx woul~l probably repre-

sent r-ather unrel iable infoxmation.

I I I 1 , I I 1 I

VOICE E, SINGER / - VOICE F

A 240 Hz

SPL AT 0.5 m (dB) SPL AT 0.5 m (dB)

Fig. 11143-3. Peak-to-peak glottogram amplitude versus sound pressure level (SPL) of a vow1 produced a t va- rious degrees of vocal effort in a singer and i n an untrained voice. The thin solid line shws the case of increase a t equal rate. The dashed line in the right plot shws the case that a 10 dB increase in SPL is accampanied by a 4 dB increase in the level of the fundamental, which is a value typical for untrained voices, according to Fant ( 1 960) .

t STL-QPSR 2-3/1981 29,

The above reasoning suggests that an approach involving experts "diaqno-

sing" systematically varied and well-defined synthetic s t imuli m l d lead

to more reliable results. For this reason a series of stimuli were prepared

using the MUSSE vowel synthesizer (Larsson 1977). The s t i m u l i consisted of

an asc~diny, ocle octave scale on the vowel < . ?he formant frequencies

were cMsen so as to represent a barytone singer. Technical. details on the

synthesis are shown in Table 111-B-I .

The contributions to the impression of a person singing with a raised

larynx position of different acais t ic charclc:t-er istics were tried. As jus t

mentioned, two cmlidates could be expected tn Iw involved i n a larynx rise

gesture according to previous measurement-ti, 11amely (1 ) car) increase i n

formant frqlencies (~undberg & N o n l s t r 5 m 1976); and ( 2 ) a decrease of the

amplitule of the voice source spectrum f d a n e n t a l . Tn addition to I-lwse

two candidates a third one was tried, namely ( 3 ) a decrease in the extent

(i.e. amplitude) of the vibrato undulatiotls OF I-he fundamental frequency.

I n the asceding scales these three parameters started to change as : x m as

the furdamental frequency went higher than a limit value, which was set to

262 Hz. The parameter values used are listed i n Table 111-B-I.

difgerent listenin? tests were made.

Table 111-B-I. Synthesis data for the simulation of larynx rise. The fun- k n t a l frequency range was 175-350 Hz. The parameters started to change logarithmically with fundamental frequen- cy as soon as this frequency %nt higher than 262 IIz.

Formant frequencies

a t low pitches:

a t tap pitch :

Amplitude of source spectrun £undanental

a t l o w pitches : O dR

at top pitch: -5 dR

ILnplitule of vibrato

a t low pitches: +/- 50 cent a t top pit&: +/- 20 cent

Table 111-B-11. Responses indicating a greater similarity to a h q n x elevation i n a pair comparison test; "no" and "yes" means that the parameter did not and did vary with fundamental frequency, respectively. Numbers under "Responses" show the percentage of "correct" votes (i.e., for that scale i n the pair, which had the greater number of pitch dependent parameters.)

FO-DEPENDEXE IN FIRST/SECCND SCALE W I T H RM;ARD TO:

formant amplitude of amp1 it ude Responses

frequency? fundmental? of vibrato? %

pated as jurors. The test was run in the same way as the precedina one.

The results are shown in sam Table. They suqqest that mre m r t a n c e

should he attributed to the vibrato mlitude than to the fundamental under these conditions. Thus, i t seems justified to conclude that a dec-

rease i n the amplitude of the source spectrum fundamental is not n-essari-

ly accompanying a r ise of the larynx. On the other hand, a larynx rise is

~c~likely to take place without a reluction of the vibrato anpl. itude.

To summarize our results so far, a larynx rise is associated with a n

increase of the formant frequencies. In addition to this, khe amplitude of

the vibrato is typically reduced, i f the larynx is raised, but such a

rduction lnay tctke place for other reasons as well, which are not related

to larynx height. A s regards the amplillwle of the voice source spectrum

funZamenta1, it is typically reduced when the larynx rises, but the larynx

may also rise without this effect.

Given these findings it is also interesting to find out the combined effect

o f a l l these three parameters. This was tested i n a less formal way. A

y-wp of 30 voice experts, including singing teachers, spech therapists,

logopedists, phoniatricians and other people joining a class on voice

theory, listened to a p i r of scales, presented three tiines i n a secpence

by means of a loudspeaker. In this pair of scales a l l three parameters

remained unchanged i n the f irst scale and varied with fundamental frequency

i n the second scale. Thus, in the second scale, the formant frequencies

started to increase, while the ampliturle of t h l voice source fundarnerital CIS

well as that of the vibrato urrdulations startal to clecreast?, as so011 as I t~c .

fundamental frequency passed 262 Hz. ?he subjects were asked to describe i n

their own terminology the difference between the firs1 c\nI the secord way

OF singing the scale, and also to t.t.11 which way they f e l t to he more

healthy. The results are listed i n Table 111-P-111.

Idany interesting observations can be mde from Table 111-B-111. One obser-

vation i s that the same stimulus i s often diagnosed i n c q l e t e l y onpsite

ways. For instance, with resl3ect to the healthyness of the mice use inferred

from the sounds, 18 experts (64 %) prefer the case ( A ) , i n which a l l three

voice parameters remain independent of fundamental frequency, but no less

than 9 experts (31 %) recammend the opposite case (R), where a l l parameters

show a pitch dependence. This suggests that there are solid reasons for

disagreement between voice experts.

Another interesting finding is the great number of tenns u s e l for descri-

bing exactly the same acoustic characteristics. Morewer, idenl:i.cal terms

are rarely used by many experts as is evident from the gemerally low

numbers within parentheses. Ornsiderable time i s likely t~ b? spri t or1 the

clarification of terminology before an efficient co~nmtir~ioation on central

issues can be established between many voice experts.

It can also be seen in Table 111-B-111, that ,s t i l l , 20 of the terms used to

describe the case of (R) refer to tension, stress, pressed phonation,

explicitely or implicitely: indeed 5 experts diagnosed a raised larynx for

this scale. Tlne correspondirq nurnbers for the case of (A) is only 4 and 1,

repectively. In this respect, then, the responses were rea9onabl.y consis-

tent. I t should be added here, that none of the experts recommending the

(B) alternative for the sake of voice health characterized this alternative

i n terms associated wit% tension.

Results of m i m e n t I1

a: Expressions used in the free cbice test where voice experts were asked to "diagnose" two synthetic scales. In the f irst scale (A) a l l parameters remained the same throughout the scale, while in the second (B) the formant frequencies, the amplitude of the source spectrum fdamental, a d the amplitde of the vibrato undulations decreased as soon as the pitch frequenc:~ went higher than 262 Hz. The numbers within parenthesis after each term stows the number of subjects who used the term.

a: me scales differed w i t h respect to:

A i r flw, lack of control in B Articul ation/~ocal tract shape Canpliance/Elasticity , more in A Covering EQalizaticn, better in A (1 ) Fundmental, amp1 itude of,

too low in B (2), too high in B (1) Glottal closure "Wfklang", missing i n B Larynx rising, i n A ( I ) , i n B (5) Larynx, tense i n B L~udmss, higher i n B Nasalization i n A Openness, too much i n B Overtones, more in A Pharynx width, greater in B (2 )

narrower in B (2 ) Pharynx, tense (1) -nation Pinched timbre, i n A Placement, better i n A Pressed pkronation, in A ( l ) , in B (7) Register Relaxation, lacking i n B Shallw phonation, i n B Subglottic pressure, badly used/too high Support, lacking i n B Throatiness, in A Timbre, more i n B (1) Tongue root, too posterior (1 ) , tensions Training Velun, too relaxed i n B Vowel aolour, changed in B (1)

b: Preference with respect to voice health:

Scale A 18 Scale B 9 No preference 2

-- - -

STL-QPSR 2-3/1981 35.

In n o m l speech, the vertical position of the larynx generally shows a

clear pitch depedence. Moreover, it is well known that the mice aource of

an untrained speaker tends to become less wealthy in high overtones as

pitch is increased. The presetll. fidings are apparently contrary Lo this,

as they show that a low amplitude of the source spectrum Fundamental i s

often associated with a larynx rise. The explanation @ably i s that the

larynx can be raised in different ways. In one way, it is part of a general

increase in muscle activity i n the entire laryngeal region, so that the

ghonation is changed towards the "press" extreme. ?his would be the type of

larynx rise which singing teachers do rnt rwanmerd &an the point of view

of phonatory health. A different type of larynx rise probably occurs i n

normal speech and it i s not associatetl with an increase of the general

tension i n the laryngeal muscles. T h i s type o f larynx rise would then be

harmless. We may assume that those voice experts, w h ~ recmeded the scale

indicating a rising larynx, were referring to this type of larynx rise.

Interestingly, several of these sub j wts were sp~u 9 ) therapists, not sing-

inq teachers.

F'rm the present investigation w e w a n t to draw the following conclusions.

The most striking and typical acoustic sign of a rising larynx is a shift

i n the furmant frequencies. A rise of the larynx, which is associated with j

singing a pitch i n the upper region of a singer's range, is typically

associated with a decrease i n the amplitude of the fundamental frequency

undulations due to the vibrato, and there may also be an associated dec-

rease in the amplitude of the source spectrum fundamental. ?he results dlso

suggest that the larynx can be raiser] i n different ways, and only one of

these i s associated with an increased tension i n the addwary muscles of

the larynx. This type of larynx rise does not seem to be recornmeridable,

according to the reactions of singing teachers.