Upload
phamkiet
View
216
Download
1
Embed Size (px)
Citation preview
Dept. for Speech, Music and Hearing
Quarterly Progress andStatus Report
A descriptive model oflaryngeal articulation in
speechLindqvist-Gauffin, J.
journal: STL-QPSRvolume: 13number: 2-3year: 1972pages: 001-009
http://www.speech.kth.se/qpsr
STL-QPSR 2-3/1972
I. SPEECH PRODUCTION
A. A DESCRIPTIVE MODEL OF LARYNGEAL ARTICULATION IN SPEECH
J. Lindqvist
Abstract
In a previous paper(i) we have suggested a two-dimensional model of laryngeal articulation for production of phonatory types used in different languages. A mechanism for controlling downward pitch inflections was also discussed. The model was based on observations made with fiber optics inserted through the nasal passage(2). The philosophy of the model was that laryngeal speech mechanisms a r e closely related to more basic, phylogenetically old, laryngeal mechanisms associated with breathing, phonation, and protection. Essentially the same ideas a r e used in the present paper to describe more explicitly the function of the larynx in t e r m s of a hypothetical model.
Introduction
The laryngeal muscle s a r e organized in functional groups. Each muscle
ac t s in one o r more of these groups. An independent control of a single
muscle i s a difficult task, although not impossible, but can be regarded a s
a l e s s natural activity. In order to develop such activities it i s necessary
to have an accurate control of the resul t of the action by means of, for in-
stance, acoustic feedback. However, in speech communication a very p re -
cise control of the acoustic signal i s neither possible nor necessary. P e r -
ception makes use of the redundancy inherent in speech which allows great
variability in the signal. Coarticulation effects a r e therefore generally not
compensated for(3) and thus the speech apparatus i s f ree within l imits to
produce the speech in the most natural way. In the search for these natu-
ral articulatory gestures which it may be reasonable to look for among
those which a r e innately organized, o r thc basic laryngeal gestures, the
phylogenesis of the larynx may give valuable information. Unfortunately
there a r e divergent opinions about the nature of the basic laryngeal func-
tions. We therefore s t a r t our discussion with some aspects of the evolu-
tion of the human larynx. The conclusions made in this paper a r e p r imar - ily based on our own observations with fiberoptics, but facts found in the
(4) l i terature a r e also considered .
STL-QPSK 2-3/1972
Basic laryngeal functions
An important fcnction of the larynx i s to work a s a sphincter. Phylo-
genetically this sphincter exhibits two functions, one related to a i r -trapping
and one to protection of the subglottic system. We will argue he re that in
thc human larynx these two functions make use of the same laryngeal mech-
ar.ism. The a i r -trapping inlet and outlet valve s have been the t rue and
false vocal fold;, r e spectively. Ah-trapping, which has some influence
on the pectoral girdle function in, for instance, the apes when swjnging
from a branch(5) i s not important for human beings. Instead the t rue vocal i
folds have been specialized for phonation and the retrogression of the false I
vocal folds has made these s t ructures l e s s efficient a s a valve for a i r -
trapping. Effective air-trapping in the human larynx i s in our opinion a l -
ways supplemented by the mechanism of the protective closure. This
c losure i s a t a level above the glottis and i s made between the tubercle of (6) the epiglottis, the cuneiform carti lages, and the arytenoid carti lages .
In Fig. I-A-I a scquence of f r ames f rom a fi lm registration of this type
of laryngeal closure i s shown. Our experimental observations indicate
that breath-holding i s often performed with the r e spiratory muscle s and
not by closing the larynx, but sometimes, a s e. g. if the intrathoracic
p res su re i s ra ised, a s when someone coughc, this protective closurc i s
used.
The larynx i s a lso active during respiration(7). This activity i s found
in the muscles that adduct and abduct the glottis. The adducted state of E
the glottis i s used for phonation. Evolutionary evidence indicate s that
phonation i s not exclusively connected with speech (see ref. 5). Conse-
quently it seems reasonable to conclude that the adducted state of the
glottis, the "voicing position", i s a basic and language independent larynge-
al function.
F r o m the discussion given above we can identify three basic laryngeal
functions related to breathing, - phonation, and protection. The co r re spond-
ing postures o r positioning of the larynx a r c the breathing position, the (8) voicing position, and the protective closurc . Innately organized muscle
groups a r e used for movements intermediate and superimposed on these
postures. The movement f rom the open position (breathing position) to
the voicing position i s called glottal adduction and the movement f r o m the
voicing position to the open position i s called glottal abduction. The muscles
STL-QPSR 2-3/1972 3.
used for these movements a r e poster ior crico-arytenoids which abduct
the vocal folds, and the inter -arytenoid muscle which adducts the vocal
folds. EMG-measurements indicate that probably no other muscle ac t i -
vitie s a r e nece s sa ry for the se movement n(O) although it i s traditional to
attribute an adducting function to several intrinsic laryngeal muscles.
The observed movements of the arytenoids a r e a rocking o r rotatory
moverr.ent around the cylindrical joint on the crizoid carti lage in a dorso-
la te ra l direction.
Movements f r o m ei ther the open o r the voicing position to the protec-
tive closure a r e made by inhibiting the crico-thyroid muscles and by a c -
tivating al l sphincter muscles of the larynx including the ary-epiglottic
sphincters. The ary-epiglottic sphincters a r e here defined a s the a r y -
epiglottic, the oblique arytenoid, the thyro-epiglottic, and the external
t hyro-arytenoid muscles( lo). In our opinion the ary-epiglottic sphincters
a r e of p r imary importance. The movements of the arytenoids f rom voic-
ing position to closed position take place in a forward tilting direction to-
wards the tubercle of the epiglottis("). The tubercle of the epiglottis is
a t the same time pulled dorsally, by a rotation of the thyroid carti lage
about i t s joint, and approaches the arytenoids.
Evolution of language and speech has a l so made use of a functional
differentiation of the sphincter muscles which has enabled a fine control
of pitch and phonation types. This we will discuss in the following section.
Thc role of the extr insic muscles on laryngeal articulation i s l e s s known;
probably they take an important par t in the regulation of pitch and pos-
sibly a l so in the protective closure(i2). Ilowever, it seems reasonable
to assume that even for pitch regulation thc intrinsic laryngeal muscles
(including crico-thyroidcus) a r e of pr imary importance and we will r e -
s t r ic t our discussion to these muscles.
Laryngeal speech functions
So f a r we have discussed the basic laryngeal mechanisms. Breathing,
protection, and phonation have been considered. Now we will turn to thc
question how these mechanisms may be utilized in speech.
The speech posture of the larynx i s \-?it11 the vocal folds in voicing posi-
tion. In this position the vibrations of tlie vocal folds a r e most easily sus-
tained but if the a i r velocity i s sufficient they can a lso vibrate in a more
STL-QPSR 2-3/1972 4.
open position. The laryngcal speech gestures a r e dynamic in their nature
and they appear as a modulation of the spccch posture.
Three functions of the larynx of communicative importance a r c d i s -
cussed here: to dcvoice, to change the quality of the voice, and to control
the pitch. There a r e two types of devoicing gestures which secm to be
used by speakers of any language. These a r e the glottal abduction gesture,
that i s an articulation towards the open position of the larynx (breathing
position), and the laryngeal closure gcsturc, which i s art iculated towards
thc closed larynx position (protective closure), see Lindqvist, ref. 1.
An example of an almost complete laryngeal c losure is given in Fig. I-A- 1.
The muscles activated for the abduction gesture a r e the poster ior
c r ico -arytenoidei and for the adduction ge s ture the inter -arytenoideus.
The laryngeal closure gesture i s p e r f o r ~ e d by activating a l l sphincter
muscles a s described in the previous section. The degree to which these
gestures a r e invoked in speech i s determined by the acoustic resu l t s
aimed a t , and they ncver reach the breathing or completely closed posi-
tion. In voiced contcxt a more extreme laryngeal closure gesture resu l t s
in a voiceless interval, and i s generally r e fe r red to a s a glottal stop.
A s already pointed out, the vocal folds may vibrate even in a somewhat
open po sition if the a i r -velocity through the glottis i s sufficient. Accord-
ingly, abduction gesture performed to suppress the voicing in voiced con-
text may cause cxccssivc los s of a i r . The abduction gesture i s therefore
efiectivc a s a dcvoiccr only in combination with an o ra l constriction while
%e glottal stop gesture can be used a s a dcvoicer independent of other (13) ar t iculators .
Th- glottal abduction gesture and thc laryngeal closure gesture can a lso
be used to change the quality of the voice. According to our hypothesis
these gestures a r e independent a t a motor command level and different
combinations of them may be used a s phonatory types of laryngeal ar t icula-
tions in different languages. A small degrec of glottal abduction tends to
decrease the closcd phase of the vocal fold vibrations ( increase the open
quotient). T h e acoustic consequence of this i s a reduction of the amplitude
of the higher and that the voice obtains a more o r l e s s noisy
character. In phonetic t e r m s this type of voicing i s often re fer red to a s
lax voice o r breathy voice(i5). A laryngcal closure articulation has a -
STL-QPSR 2-3/1972 5.
quite different effect on the quality of the voice. The increase of the a c -
tivity in the ary-epiglottic sphincters resu l t s in a shortening and thickening
of the vocal folds and if the activity in the vocalis muscles was not in-
c reased , this would resu l t in a lowering of the pitch. However, a s defined
ea r l i e r a laryngeal closure gesture i s accompanied by an increased activi-
ty in the vocalis muscles. This activity counteracts the effect the short-
ening of the vocal folds has on the pitch. The vocal folds a r e also pressed
together which cause them to vibrate with a decreased open quotient. A
more extreme articulation resu l t s in i r regular periodicity of the vibrations
o r in a complete suppression of the vibrations, depending of the p res su re
drop a c r o s s the glottis.
The consequence on the acoustic signal i s increased intensity of the
higher harmonics , s imilar to the effect of an increase in the subglottic
pressure . This type of voice i s often re fer red to a s tense voice. The a -
periodicity gives the voice a rough charactcr and this i s often called
creaky voice o r creak (see Ladefoged, 1967, and Catford, 1964, ref. 15),
depending on the degree of aperiodicity.
The same acoustic features in combination with consonant articulation
a r e often described a s laryngealized sounds and sometimes a l so a s glot-
talized sounds. We choose to use the t e r m laryngealization to define a
laryngeal articulation towards the laryngeal closure. A glottal stop i s ( 16) then an extreme case of laryngealization .
The evolution of language and speech has made use of an independent
pitch control mechanism by a functional differentiation among the sphincter
muscles of the larynx. F o r regulation of the pitch i t i s essent ial to have
control of the activity of the vocalis muscles , independent on a motor
command level, of abduction and laryngealization ge sture s. The hypo-
thesis has been put forward that laryngealization in combination with low
vocalis activity i s used a s a mechanism for producing a low pitch voice
a s in the Swedish word accent ( see ref. 1). More recent data collected
with fibcroptics seem to verify this These data a l so in-.
dicate that a laryngeal closure can be performed without the assis tance
of activity in the vocalis muscle. IVe suggest that this type of laryngeal
c losure, which i s s imilar to laryngealization but with no activity in the
vocalis muscles , should be called laryngeal constriction. According to
Fig. I-A-2. Selected f r a m e s f r o m a f iberscope f i lm r e g i s t r a - tion of connected speech, showing the larynx for low (a) and high (c) pitch. Glottal abduction a t low (b) and high (d) pitch i s a l so shown.
Whispering
adduction *t-- Glottis -* abduction
All languages (but not phonemic)
Ladefoged' s phonatory types in relat ion to laryngealization and glottal abduction
All languages
3nvoiced
adduction 4- Glottis + abduction
p Gujarnti Breathy voice
Indone s ian 4
Lax voice
r : . 0 .A C,
fd N
.PI 4 rd Q, bO c 2- rd 14
*
Catford' s phonatory s t r i c tu re - types in re la - tion to laryngealization and glottal abduction.
Korean
All language s
Voice
Fig. I-A-3. Tentative two-dimensional descr ipt ions of laryngeal ;irticulation of phonatory types used in different lan- guages. According to m o r e recent f ind ings t he vocalis activity may be i l lus t ra ted separated f rom laryngeali- zation in a th i rd dimension (Lindqvist, 1969).
Hausa
Glottal stop
f Kumam
Tense voice Murmur
J
Unvoiced
(vo ice less )
Voice
Norma l glottal
(Voice)
Creak
Kumam
Creaky voice
Tense voice
Ligamental
(Voiced
Creak
Ventricular
(Creak)
Lax voice Breathy voice
(Breathy voice)
?
(Stop) Glottal stop
Murmur
Arytenoidal ?
creak)
Creaky voice
(Whispery voice)
Whispering
(Whisper?
e-) Whispery c r eak
STL-QPSR 2-3/1972
4. T h e g l o t t a l s t o p g e s t u r e i s i n t e r p r c t e d a s a g e s t u r e towards the protective closure and laryngealization i s defined with reference to this basic laryngeal mechanism.
5 . A mechanism for controlling low pitch and downward pitch inflection i s described which makes use of the ary-epiglottic sphincters.
There a r e , of cour sc , some uncertainties regarding the validity of the
suggested model due to laclc of EMG-data. This i s especially t rue for the
ary-epiglottic sphincters which have not been discussed previously in con-
nection with speech production. Also the role of the extr insic laryngeal
muscles i s not considered in the present model ( l9 ) ,
However, our model seems to account for the most essent ial functions
of the larynx, The described physiological and acoustical relations be - tween low tone, laryngealizatioli, and glottal stop may give a better under -
standing of dialectal variations and his tor ical change s in language s using
low tone. One exam* of such variation is the different realizations of the
word intonation in Danish, f rom low tone, extfcmely low tone to glottal
stop. Another example i s the historical change that seems to have taken
place in the Peking dialect of Chinese . In this case a final laryngealized
s top consonant has developed into a glottal stop which in turn h s s developed
into a low falling tone,
Two examples of how this model can be used for linguistic purposes a r e
shown in Fig. I-A-3. This figure shows tentative two-dimensional de scr ip-
t ions of laryngeal articulation of phonatory types used in different languages,
suggested by Ladcfoged and by Catford (see ref. 15). The two dimensions
used a r e the glottal abduction and laryngcalization. A better description
can be made with the vocalis activity in 3 third dimension, although the
present simplified description seems to capture the most essential linguist-
ic facts.
References
( 1) Lindqvist, J. : "Laryngeal mechanisms in speech", STL-QPSR 2-3/1969, pp. 26-31.
(2 ) Sawashima, M. and Hiro se , H. : "New laryngoscopic technique by use of fiber opticsf ' , J.Acoust. Soc.Am. - 43 (1968), pp. 168-169.
( 3 ) Intrinsic allophones occur by definition due to coarticulation with the context. An interesting question is i f language ever make use of corn- pensatory articulation in specific context in o rde r to maintain cer tain acoustic features. J. Lindqvist, Intern repor t 31. 1. 1969 (in Swedish). It can, for instance, be shown that the interaction f rom articulatory
STL-QPSR 2-3/1972 8.
cont. re f . (3)
and pulmonary gestures on the pitch in speech a r e not compensated for. See Ohman, S. and Lindqvist, J. : ' 'Prosodic pitch contours", Z. f. Phonetik usw. - 2 1, Heft 1/2 (1968), pp. 163-170.
(4) Laryngeal features a r c not discussed since in our opinion the ab- stractions made in feature systems a r c not capable of giving a full understanding of the speech proce ssc s . The theories of distinctive features have a lso resulted in a conc.cntration of too much effort on the descriptive needs of the linguist instead of on finding the l imita- tions and intrinsic character is t ics in the speech mechanisms which in it self may have explanatory powcr . Lindblom, B. : "Phonetics and the dc scription of language", to bc publ. in the Proc . of the VIIth International Congress of Phonetic Science s , Montreal, by Mouton & Co. (forthcoming).
(5) F o r a discussion; see pp. 98 -1 12 in Ncgus, V. E . : The Comparative Anatomy and Physiology of the Larynx o on don 1949).
( 6 ) The epiglottis, which i s a rudiment, does not function a s a lid to close the larynx a s sometimes i s stated. Only the lower pa r t , the tubercle of the epiglottis, takes par t in the closure. However, the laryngeal closure ge s turc has a tendency to obscure the view of the larynx by a backward. t i l t of thc epiglottis.
(7) Faaborg-Andersen, I<. : Electromyographic Investigation of Intrinsic Laryngeal Muscles in Humans (Copenhagen 1957) p. 18.
(8) The opinion that thc human larynx i s a three -tied sphincter (P res sman , 1954) i s according to our observations wrong. The third level of l a r - yngeal closure should, according to that hypothesis, be a t the level of the false vocal folds. The size of the false vocal folds var ics bc - tween different individuals and i t i s not unusual to find subjects that can put their false vocal folds into contact and even phonate with them. However, this i s not very common and cannot be regarded a s a basic laryngeal function. On the other hand, the false vocal folds nlay play a secondary role in the protective closure and a c t a s a cushim to complete the closure between the tubercle of the epiglottis and the arytenoids. The false vocal folds contain only few muscle f ibers and the positioning of the sc folds i s determined by the surrounding s t ructures like the ary-epiglottic sphincters. (See P r e s s m a n , J . J. : "Sphincters of the. larynx", A. M. A. Arch. of Otolaryngology - 59 ( 1954), pp. 22 1-236. )
(9) Hirose, H. : "An electromyographic study of laryngeal adjustments during specch articulation", Haskins Laborator ies SR-25/26 ( 197 I ) , pp. 107-116; Hirose, H. and Gay, T. : "The activity of the intrinsic laryngeal muscles in voicing control: An electromyographic study", Hrrskins Laborator ies SR-28 (1971), pp. 115-142.
(10) Unfortunately speech r e s e a r c h e r s have paid no attention to these muscles and no EMG-data on their functions a r e available.
( 1 I ) Sone sson, B. : "Die funktionelle Anatomie de s Cricoarytaneoidgelenke s t ' , Z . f. Anatornie und Entwicklungsgc schichtc 12 1 (1959), pp. 292 -303; - Kaplan, H. M. : Anatomy and Physiology of Speech (New York 1960), p. 154. van Leden, H. and Moore, P. : "Tlle mechanism of the crico-arytenoid joint", Arch. Otolaryng. (Chic) 73 (1961), pp. 541 -550.
STL-QPSR 2-3/1972 5.
Nasser Kotby, M. and Haugen, L. I<. : "The mechanism of laryngeal functionti, Acta Otolaryng. 70 (1970), pp. 203-211; Faaborg-Andersen, K. and Sonninen, A. : "The function of the ex- t r ins ic laryngeal muscles a t different pitcht' , Acta Oto-Laryng. - 51 (1959), pp. 89-93; Sonninen, A. : "The external f rame function in the control of pitch in the human voicei' , Ann. of the New York Academy of Sciences - 155 (1968), pp. 68-90.
The hypothesis has been suggested that a stiffening of the vocal folds in combination with glottal abduction i s used a s a mean to stop vocal fold vibrations for the production of voiceless consonants (Halle and Stevens, 197 1). Even if such an articulation should give the expected resu l t it i s not likely that it occurs due to the functional organization of the laryngeal muscles. Fur thermore EMG-measurements have shown that the activity in thc vocalis muscles i s generally dccreased for consonant articulations (Hirose, rcf. 9). Halle, M. and Stevens, K.N.: "A note on laryngeal features", MIT, QPR No. 10 1 ( 197 i ) , pp. 198 -2 11.
Lindqvist, J. : "The voice source studicd by means of inverse filtering", STL-QPSR 1/1970, pp. 3-9.
Ladefoged, P. : "Linguistic phonetics1', Working P a p e r s in Phonetics 6 (UCLA), 1967, pp. 1-98; Catford, J. C. : "Phonation types: Thc classification of some laryngeal components of speech productionr', in In Honour of Daniel Jones (London 1964)) pp. 26-37.
Both the glottal abduction gesture and the laryngealization gesture a r e covered by the linguistic feature "glottal s t r ic ture" (Ladefoged, P. : "The h r e e glottal feature s t ' , Working P a p e r s in Phonetics 2 (UCLA), 1972, pp. 95-101). This description of what i s going on may satisfy some linguistic purposes, but for a complete theory about speech and language, a more detailed description of the use of the larynx i s needed.
In another feature system used by Halle and Stevens ( 1967, ref. 13) the t e r m "constrictcd glottisf ' i s uscd to represent a n adjustment which
'kauses the vocal cords to b e pressed together andthe glottis to narrow o r to close". In our opinion the mechanism behind this type of ar t icula- tion i s laryngealization a s defined in this paper. However, we prefer the t e r m laryngealization, because i t does not emphasize what happens with the glottis and the general use of the word seems to be in accord- ance with our definition.
Lindqvist, J. : !'Laryngeal articulation studicd on Swedish subjects", I
this issue of the STL-QPSR.
A glottal stop gesture i s generally accompanied by a creaky onset and offset of voicing. Thus extreme lo\-/ pitch and glottal stop have both acoustical and physiological similarit ie s , which may explain, for in- stance, the dialectal variation in thc realization of the Danish "stbd".
Ohala, J. and Hirose, H. : "The function of the sternohyoid muscle in speech1', Ann. Bull. , Re search Institute of Logopedic s and Phoniatrics 4 (1970), Univ. of Tokyo, pp. 41 -44.