Dept. for Speech, Music and Hearing

Quarterly Progress andStatus Report

Raised and lowered larynx -the effect on vowel formant

frequenciesSundberg, J. and Nordstrom, P-E.

journal: STL-QPSRvolume: 17number: 2-3year: 1976pages: 035-039

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

STL-CPSR 2-3/1976 3 5.

B. RAISED AND LOWERED LARYNX - THE EFFECT ON VOWEL FORMANT FREQUENCIES

J. Sundberg and P. -E. Nordstrom

Abstract

Raised and lowered larynx positions have been studied on two subjects and in a model experiment. Good qualitative agreement was found and i t was concluded that the main effect s tems f rom the shortening o r length- ening of the pharyngeal cavity.

Introduction

Singing teachers generally agree that the larynx should be kept in a

comfortably low position in singing, and veice troubles a r e often associa-

ted with a habitually ra i sed position. Thus, the position of the larynx

seems to be an important parameter in the control of the voice organ.

In this paper we shall repor t an investigation of how vowel formant

frequencies a r e affected when the larynx is ra i sed o r lowered. Model

experiments have been ca r r i ed out i n which a se t of a r e a functions cor-

responding to the Russian vowel system has been perturbed. These r e -

sults a r e compared with formant frequencies observed in two subjects

phonating with a high and a low position of the larynx. I I

Model experiments

Recently, one of the authors studied the effects of sex-related vocal-

t r ac t length variations on vowel formant frequencies ( ~ o r d s t r a m , forth-

coming). The computer program used in that study calculates the for -

mant frequencies up to around 4000 Hz f rom a given a r e a function (based

on the program described in Liljencrants and Fant , 1975). The program

also facil i tates for the user to implement perturbations of a lmost any

conceivable type. Hence, the acoustic effect on the vocal-tract resonator

of a raised/lowered larynx can easi ly be studied by means of this com-

puter program, provided that we know what perturbation a shift in la rynx

height causes.

Obviously, a raised/lowered larynx will shorten/lengthen the

p h r y n x . Moreover, a s ~ u m i n g a pharyr.x wall volume, we can post-

ulate a narrowing/expansion of the lower pa r t of the pharynx when the

larynx i s pulled u p ~ a r d s / ~ r e s s e d downwards.

VELUM

P H A R Y N X MOUTH

Fig . 111- B- I, Schematized vocal- t ract a r e a functions used in the model exper iments ; a: unperturbed configuration, b and c : perturbat ions of the pharynx cavity length, d and e: per turbat ions of the pharynx cavity length

and the c ross -sec t iona l a r e a s above the la rynx tube (constant wall volume cri terion). See text.

MODEL MODEL

0 UNPERTURBED LENGTH CRITERION

V LENGTH A N D CONSTANT VOLUME CRITERION

I I B I I I 1 1 1 .2 .4 .6 .8

FIRST FORMANT FREQUENCV C kHz 1

[ I UNPERTURBED LENGTH CRITERION

V LENGTH A N D C O N S T A N l V O L U M E CRITERION

2.7

2.0 . 2.5 3.0 3.5

THIRD FORMANT FREQUENCY CkHz 1

Fig. 111-B- 2 . Formant frequency values f rom the model experiments. The fourth formant frequencies of [i ] with "ra ised larynx" exceed the calculating range of the computer program.

STL-QPSR 2-3/1976

S U B J E C T P S U B J E C T S

F 1 F2 F 3 5-4 F i F 2 = 3 F 4

(HZ) (HZ) (Hz) (HZ) (Hz) (Hz) (Hz) (Hz)

300 830 2700 3250 360 750 2870 3660 300 780 2650 3350 340 760 2100 3250 300 740 2450 2900 350 700 2 5 5 0 3100

TABLE 111-B-I. F o r m a n t f requencies of the subjects observed in high (H), normal (N), a n d low (L) larynx position.

SUBJECT P SUBJECT P

D RAISED I LARYNX

0 LOWERED

FIRST FORMANT FREQUENCY CkHz)

THIRD FORMANT FREQUENCY CkHz)

. . Fig. 111-B- 3.. Formant frequency values for raised and depressed

larynx observed in subject P. The vowel IPA syrnbofe show the values associated with normal larynx position.

STL-QPSR 2-3/1976 38.

tongue slightly in phonating with normal larynx position the low values

of the third formant frequency can be explained. When he r a i s e s h i s

larynx, the tongue is lifted upwards and is very likely to reduce the

volume in the same pa r t of the mouth. Hence, the third formant f r e -

quency can be expected to be high. YJhen the l a rynx i s lowered this sub-

ject is likely to r e s o r t to his singing habits including not only a low la-

rynx position but a lso a fronted position of the tongue tip. The l a s t men-

tioned gesture effectively closes the cavity behind the lower inc isors and

r a i s e s the third formant frequency which in turn adds to the amplitude

of the "singing formant" (cf. Sundberg, 1974). Thus, i t seems plausible

that the normal values a r c not fully comparable with the values obtained

I when the larynx was raised and lowered in subject S. Rather , we have

reasons to assume that the values pertaining to high and low position of

the larynx a r e more comparable. Therefore, in Fig. 111-B-3b the values

of the normal larynx height have been omitted. The trend, common to

both subjects and evident in Figs. 111-B- 3a and 111-B-3b, is that both the

I

third and fourth formant frequencies tend to r i s e a s the larynx is raised.

Long- t ime-average- spectra (L TAS) have been found to provide in-

formation related to voice quality ( see e. g. Jansson and Sundberg, 1975).

Vowels phonated with raised and lowered larynx differ greatly in quality,

and above we have found them to differ with respec t to formant frequen-

cies. As the peaks in a long-term-average-spectrum of the yoice a r e

dependent on the t ime average of the formant frequencies we w o ~ l d ex-

pect that these peaks differ depending on the position of the larynx. This

assumption is confirmed by Fig. 111-B-4, which is a long-time-average-

spectrum of the singer (subject S) singing the same song twice, once with

raised and once with lowered larynx. The f i r s t peak r i s e s f rom 400 Hz

to 650 Hz, the second peak f rom 900 Hz to 1500 Hz, and the third peak

f rom 3000 Hz to approximately 3500 Hz. Thus i t i s possible .that such

types of differences between two LTAS:s s tem f rom differences in larynx

position.

I t must be pointed out, however, that i t i s asking too much to have

the subjects a l ter the larynx height without changing the positions of any

other articulator since the major ar t iculators (lip, tongue, jiiw and larynx)

a r e connected by t issue and muscles. Nevertheless, the pat terns seen

for the subjects a r e to a la rge degree reflected by the model experiments.

This suggests that the main acoustic effect of a shift i n the larynx height

can be explained by the resulting pharynx length change.

RELATIVE AMPLITUDE

STL-QPSR 2-3/1976 39.

A minor point regarding the model experiments i s that the r e su l t s

given he re a r e not corrected for the los ses due to the impedance of the

walls of the vocal t ract . This factor i s speaker-dependent and has not

been fully examined yet. The effect is small , however (Fant , personal

communication), and does not significantly influence our resul ts .

Conclusions

Above we have found a qualitative agreement between the behavior

of a vocal t rac t model and t w subjects. This agreement suggests that

the f i r s t o rder effect on the formant f requencies of a shift i n larynx posi- I

tion s tems f rom the shortening/lengthening of the pharynx cavity. The

effects on the formant frequencies of a r a i sed larynx a re : 1) a sub-

stantial rismi-n the second formant frequency in high front vowels, 2 ) a

r i s e in both the f i r s t and the second formant frequency in open vowels, 1 3j a combined r i s e in several vowels of the third and the fourth formant

frequencies. These effects on the formants can be ident ified in long-

t ime-average-spectra of phonations with r a i sed and lowered larynx. The

question of how the positioning of the larynx influences the glottal source

character is t ics is open to future investigations.

Acknowledpments

This work was supported by the Bank of Sweden Tercentenary Founda- I

tion.

References

FANT, G. (1960): Acoustic Theory of Speech Production, The Hague (second edition, 1970).

FANT, G. and PAULI, S. (1974): "Spatial character is t ics of vocal t rac t resonance modes", pp. 121- 132 i n P roc . of SCS-74, Speech Com- munication, Vol. 2 (ed. G. Fant) , Stockholm.

JANSSON, E. and SUNDBERG, J. (1975): "Long-time-average-spectra applied to analysis of music. P a r t I: Method and general applications", Acustica 34, pp. 15- 19. - LILJENCRANTS, J . and FANT, G. (1975) : "Computer program for VT-resonance frequency calculations", STL-QPSR 4/1975, pp. 20-33.

NORDSTR~M, F. -E. (forthcoming): "Female and infant vocal t r a c t s simulated f r o m male a r e a functions", accepted for publication in J . of Phonetics - 5: 1 ( 1977).

SUNDBERG, J. (1974): "Articulatory interpretation of the ' singing formant"', J. Acoust. Soc. Am. 55, pp. 838-844.