Articulatory interpretation of the "singing formant"

Johan Sundberg
Department of SpeechCommunication,Royal Institute of Technology(KTH). S-100 44 Stockholm 70, Sweden
(Received 3 December 1973; revised10 January 1974)

The "singingformant" is a high spectrumenvelopepeak near 2.8 kHz

characteristicof vowel soundsproducedin male Westernopera and concert
singing.An acousticalmodel of the vocal tract is capableof generatingsuch a
peak provided that three conditionsare met: (i) The cross-sectionalarea in the
pharynx must be at least six times wider than that of the larynx tube opening.If
so, the larynx tube is acousticallymismatchedwith the rest of the vocal tract,
and an extra formant is added to the vocal tract transfer function. (2) The sinus
Morgagni must be wide in relation to the rest of the larynx tube. This may tune
the frequencyof the extra formant to a value betweenthe frequenciesof the third
and fourth formantsin normal speech.(3) The sinuspiriformesmust be wide.
This reducesthe frequencyof the fifth formant to about 3 kHz. X-ray studiesof
a raised and lowered larynx showedthat these three conditionsmay be fulfilled
when the larynx is lowered.Thus, the larynx lowering,typical of male
professionalsinging,seemsto explain the "singingformant" and other formant
frequencydifferencesbetweennormal speechand male professionalsinging.
Subject Classification: 70.20; 75.70.

INTRODUCTION "singing formant" in terms of formants only we have to

assume an extraordinary density of the higher formants.
The "singing formant" or the "2.8 kHz" belongs to •he
Spectrographic measurements on various vowels in four
acoustical characteristics of professional male singing
in Western opera and concert performances. Rather in- professional bass singers yielded the average formant
dependent of vowel quality and pitch, it appears as a re- frequencyvaluesgiven in Fig. 2. 4 For comparison,
mean values observed by Fant et al. in normal male
gionof highspectralenergynear 3 kHz. ' A typicalex-
ample is given in Fig. 1, displaying a spectrum level speechare givenin the samefigure.s Analysesof later-
al x-ray pictures and photos of the lip opening taken dur-
difference of approximately 20 dB between a spoken and
ing vowel phonation in speech and singing suggested ar-
a sung[u]. The spectrumenvelopepeakis frequently
ticulatory interpretations of the formant frequency dif-
accompanied by an envelope minimum slightly higher in
ferences as regards the three lowest formants: In sing-
frequency and often clearly discernible in the spectro-
gram.
ing as compared with speech the larynx was lowered, the
jaw opening increased, the tongue tip was advanced in
How can such a "singing formant" be generated acous- the back vowels, and the lips were excessively protruded
tically ? If •vo or more formants approach each other in in front vowels. The purpose of the present paper is to
frequencythe formantlevels will rise. 2 In our previous suggest an articulatory explanation for the frequencies
research, attempts were made to match spectra of vowels of the fourth and fifth formants in singing.
sungby male opera singerson a formant synthesizer.a
These experiments showed that five formants below 4
I. ANATOMY
kHz or even 3 kHz have to be postulated in order to gen-
erate a "singing formant" with formants only. In normal It can be assumed that an acoustic feature such as the
male speech the frequency of the fifth formant lies "singing formant," which is rather independent of vowel
•round 4.5 kHz. Therefore, when we interpret the quality, stems from a region of the vocal tract that is

VOWEL [u]

SPEECH SING NG

0
FIG. 1. Spectra of the vowel [u]
- 10
spoken (tell) and sung (right) by a
-20
professional bass singer.
-30 -30

-/,0 -/-0

-50 -50 '

0 I 2 3 k•tz

838 J. AcoustSoc.Am., Vol. 55, No. 4, April 1974 Copyright¸ 1974 by the Acoustical
Societyof America 838

ownloaded 28 Mar 2011 to 130.237.67.192. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms.js
 839 J. Sundherg:Singingformant 839

FORMANTFREOUENCIES
IN SINGING AND NORMAL SPEECH II. ACOUSTIC INTERPRETATION

The acoustical effects of the sinus Morgagni and piri-

formes have been studied in several investigations. On
the average good voices have been found to possess
larger sinusMorgagnithanpoorvoices.e Studying
sound
generated by excised larynges with and without cotton in
the sinus Morgagni, Beckmann found reason to doubt that
the sinus Morgagni can have any major effects on the
voice timbre in singers. His spectral analyses, however,
were notextended
beyond2 kHz.9 vandenBerg con-
cludes, on the basis of theoretical considerations, that
the sinus Morgagni acts as a low-pass filter, thus sup-
pressinghigherharmonics.
•0 Fantdemonstrated
by ex-
periments on an electrical analog of the vocal tract that
the sinus Morgagni is of importance to the fourth for-
0.5 F•
mant frequency, and that the sinus piriformes contribute
0 to the effective length of the pharynx. Also, he found a
[u:] (o4 (o:] (•:] [e:] [1:] [y:] (u:] (oQ
cutoff frequency near 5 kHz associated with the piriform
pockets. His observations were made for a single area
VOWEL (IPA SY•OOLS) function corresponding to the articulation of a neutral
vowel (see Ref. 6, pp. 102 fl. ). Later on, Flach and
FIG. 2. Formant fr•uencies • long •ish vowels in norm•
Schwickardie performed some experiments filling
m•e speech(dashedlines) andin p•fessio•l male singing
(solid lines). the sinuspiriformeswith cottonin living subjects.•'
Their method and conclusions were criticized by Mer-
reelstein, who found theoretical suppert for expecting
that closing off these pockets would not have any major
comparatively unaffected by the articulatory maneuvers
acoustic effect below 4 kHz. n
required to keep vowels acoustically distinct. Such a
region is the pharyngeal cavity system below the epi- In view of these diverging results a systematical in-
glottis. The anatomy of this system can be described as vestigation appears to be needed. In the present investi-
follows. The vocal chords constitute the bottom of a gation experimental data were collected from an acous-
small tube with a length of about 2 cm in male subjects. tical model of the vocal tract.
This tube is called the larynx tube. At the bottom of this
tube there is a small cavity, the sinus Morgagni, or the III. MODEL
laryngeal ventricle. The larynx tube is vertically and
The model consisted of a cylindrical brass tube with a
excentrically inserted into the larger pl•'ynx tuSe, which diameter of 3. ! cm. The tube was closed off at one end
is closed off in its low end by the esophagus mouth at the
by a plug into which holes were drilled to simulate the
level of the vocal chords approximately. The lowest end
larynx tube and the sinus piriformes. A constriction of
of the pharynx tube is thus surrounding the larynx tube
and is divided into two pockets, one left and one right.
6-cm length, 1-cm2 cross-sectionalarea, andcontinu-
These pockets are called the sinus pirif•rmes.
LARYNX
It can be assumed that the geometry of these cavities
is affected by the larynx height. Therefore lateral and HIGH LO•

frontal tomograms were taken when a singer phonated the

vowels[G]and[i] with raisedandloweredlarynx, Trac-
ings of the iomograms that showed the largest dimen-
sions of the sinus Morgagni and the sinus piriformis are
given in Fig. 3. A larynx lowering seems to be associ-
ated with two major effects: the sinus Morgagni and the
sinus plriformes are both expanded. An additional effect
not seen in the figure is of course an increase in the total
length of the vocal tract. It appears that these observa-
tions agree with what has previously been observed to
characterize "covered"singing.8 It is also observed
that the size of the sinus piriformes seems to be smaller
in [a] thanin [i] in the caseof raised larynx. This sug-
gests that the size of the sinus.piriformes may be de-
pendent upon the vowel in the case of nontowered larynx.
FIG. 3. Tracings of frontal and lateral tomograms (left and
This would support the assumption proposed by Fant to
right in each pair. respectively) of the larynx in high (left pairs)
explain the low value of the first formant frequency in an and low (right pairs) positionduring phonationof the vowel [•]
[a] whichwas observedin the case of an area function (upper series) and [i] (lower series). [The tomograms were
that includedthe sinuspitiformes. 7 taken by Dr. M. Hayerling. Karolinska Sjukhuset. Sœockholm.]

J. Acoust.Soc.Am., Vol. 55, No. 4, April 1974

wnloaded 28 Mar 2011 to 130.237.67.192. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms.
 840 J. Sundberg:
Singingformant 840

ously variable position could be introduced into the large theresonators

areacoustically
uncoupled•5:
tube. The transfer function of the model was measured
using the STL-Ionophone as a sound source. It was Alo•-0.48 ß(Ao)t/a [1- 1.25 (Aa/A•,)tla]. (2)
adapted at the bottom of the cavity simulating the larynx This equation was found to give values that agreed within
tube. This sound source behaves like a point source pro- 1% with the resonancefrequencymeasuredin the case of
viding constant volume velocity and possessing an almost a small tube centricaily inserted into a larger tube.
infinitely high inner acoustic impedance. It consists of When applied to excentric tubes the calculated end cor-
a discharge between two fine electrodes loaded with a rection is slightly too long, as stated by Ingard. The
high dc voltage. By modulating this voltage sound is gen- length correction for the twin-tube model of the larynx
erated that is analogous to the modulating signal. The tube mentioned above is 0. 2 cm if Ap is assumed to be
frequency characteristic of the STL-Ionophone is linear 4 cm2. The resonancefrequencywas calculatedto be
from the infra- to the ultrasoundregion.t• The fre- 2.5 kHz approximately by applying Eqs. 1 and 2 to the
quency response of the model was recorded by a level twin-tube model of the larynx tube. This value is slight-
recorder with a filter frequency synchronized with the ly too low since the estimated end correction is some-
modulation frequency. what too long because of the excentered position of the
larynx tube in the pharynx. However, is seems reason-
able to assume that the larynx tube considered is capable
IV. LARYNX TUBE
of resonating at a frequency below 3 kHz. Therefore,
From the tomograms traced in Fig. 3 estLmates were the larynx tube can be expected to add an extra formant
made of the dimensions of the larynx tube in the lower tuned to a frequency located between the frequencies of
position. On the assumption of elliptical shapes in all the formants that in normal speech are referred to as
the third and the fourth.
directions the following estimates were obtained. The
sinusMorgagnicontainedan air volumeof 0. 56 cm• and
The area of the larynx tube opening has been observed
had a vertical length of 0.4 cm. The rest of the larynx
to increasewith risingvoicepitch.is The area of the
tube was 2.2 cm long and had a cross-sectional area of
larynx tube opening is important for two reasons. First,
0.52 cmz on the average. In a first approximation
we
it determines whether or not the larynx tube will behave
may regard this larynx tube as a twin-tube resona•r.
as a separate resonator. Second, an increase in this
The upper tube in that system would have a length of 2.2
cm and a cross-sectional area of 0.52 cm•. The corre- area will tend to raise the resonance frequency of the
larynxtube. An opening of 1.5 cm• requiresa pharyn-
sponding measures for the lower tube would be 0.4 cm
geal cross-sectionalarea of at least 9 cm2 if the larynx
and1.4 cm2, respectively. tube is to act as a separate. resonator. It seems likely
A small tube such as the larynx tube inserted into a that this condition can be met if the larynx is lowered.
larger tube such as the pharynx tube will act as a sepa- If so, a larynx lowering would be particularly important
rate resonator provided that the ratio between the cross- during high-pitched singing. The effect on the resonance
sectional area of the larger tube and the opening area of frequency of an expansion of the larynx tube opening may
the smaller tube exceeds 6:1. •4 This condition would be be compensated for by an increase in the cross-sectional
met by the larynx tube under consideration if the pharyn- area near the closed end, i.e., by an expansion of the
geal cross-sectional area at the level of the larynx tube sinus Morgagni. The volume required in order to keep
openingis greater than6x 0.52 = 3.1 cm2. It seemsper- the resonance frequency at 2.8 kHz for different values
fectiy safe to assume that this is the case, particularly of the opening area was measured in the model. The
since the pharynx appears to be widened durir• larynx- larynx tube was simulated by a twin-tube model excen-
lowering, as indicated in Fig. 3. Consequently, in sing- trically inserted into a larger tube (diameter: 3.1 cm)
ing with the larynx lowered, the larynx tube can be ex- that simulated the pharynx and mouth cavities. The
pected to act as a separate resonator capable of adding length of the lower tube was varied so as to give the res-
an extra formant to the transfer function of the vocal onance at 2.8 kHz. The results are given in Fig. 4. The
tract. calculated values were obtained from Eqs. 1 and 2. It
is seen from the figure that the calculated volumes are
The resonance frequency of a twin-tube system can be
slightly smaller than the observed. Presumably, this is
calculatedusingan equationgivenby Fant (seeRef. 7, an effect of the overestimation of the end correction of
p. 65). H A• andAo are the cross-sectionalareas, and
the larynx tube opening owing to the excentricity. Ac-
ls• and l 0 are the lengihs of the lower (wider) andupper
cordingto the measureddata, a volumeof 0.9 cms would
(narrower) tubes, respectively, the resonancewill occur
be requiredfor a larynxtubeopeningarea (A0)of 1.1
at a frequency f which satisfies the condition
cmz combinedwith a sinusMorgagniwith a cross-sec-
As•.
•I 2•f(lø
+•o)f• 2nfl• =1, (1) tional area(A•) of 2.01 cm2. This seemsto be a rath-
A0 c c er moderate increase of the volume of 0. 56 cmSestimated
in the tomograms. It seems likely that some in-
where c is the speed of sound propagation.
crease of the sinus Morgagni volume is achieved auto-
The effective length of the upper tube is I o+ Al 0. The matically when pitch is raised because of the length in-
end correction Alo depends on the termination of the crease in the stretehed vocal folds. It is also likely that
narrow tube. If the twin-tube is inserted into a wide tube an extra lowering of the larynx is capable of stretching
of cross-sectional area A•, and A•, >6Ao, Alo can be cal- the larynx somewhat in the vertical direction. The •_at.a
culated using a formula given by Ingard, provided that given in Fig. 4 seem to support the assumption that the

J. Acou•t.Soc.Am., VoL 55, No. 4, April 1974

ownloaded 28 Mar 2011 to 130.237.67.192. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms.js
 a41 J. Sundbe•g:
Singing
formant s41

OI:•Ne'•G
A,•IEA
(Ao)FO•
A•-2ma•2(cr•) only one instead of beth pockets has a small effect. The
zero appears somewhere between 3 and 4 kHz for a
pocket depth of 2 cm. Recalling that a spectrum envelope
minimum often can be observed in this frequency region
in spectra of sung vowels, we can suggest that this mini-
mum probably stems from the sinus pitiformes.
The sinus piriformes also affect the formant frequen-
cies. Provided that the area raUo between the larynx
tube opening and the pharynx tube is smaller than 1:6,
•0,2 C,lU.
CUC•E
0 the sinus piriformes can be expected to be equivalent to
a length increase of the pharynx. Those formant fre-
quencies will drop in particular which correspond to
FIG. 4. • volume • sinus Morg• r•red for va•ous
standing wave resonances in the pharynx tube, e.g., the
v•ues • the l• tube ope•ng area A 0 in order • keep the second formant in front vowels. Thus we may expect
resonance of •e la• tube at 2.8 kHz. The parameer is that the sinus pitiformes affect the formant frequencies
the c•ss-sectional •ea of the sinusMo•agni A•. •e mea- in a manner that depends on the articulation, i.e., the
sured da• were •n• f•m an acoustical m•el of the vocal location of a constriction of the vocal tract.
tract.
The effect of varying the location of a constriction(see
Sec. IH) was studiedexperimentallyin the model. The
resonance frequency of the larynx tube was tuned to ap-
larynx tube acts as an acoustically mismatched, and proximately 2.8 kHz before the constriction was inserted
hence separate, resonator and is thus capable of adding into the tube. The area function of the larynx tube was
an extra formant at 2.8 kHz to the normal vocal tract then kept constant in all measurements giving the results
transfer function in sh-ging. shown in Fig. 6. The effect of adding the sinus piti-
formes canbe described as follows. The fifth formant fre-
In addition to the effect of generating an extra formant,
quency drops considerably for certain locations of the
an acoustical mismatch between the larynx and the
constriction. The fourth is essentiaily unaffected, as can
pharynx will also tend to raise the frequencies of the
be expected in view of the acoustical mismatch between
lower formants (seeRef. 7). However, as will be shown
its resonator and the pharynx tube. The third formant
below, this effect is counteracted by the influence of the
frequency drops slightly and the second is lowered par-
sinus piriformes and the lengtheningof the vocal tract
which axe two other effects of a lowering of the larynx.
ticularly in cases of fronted constriction. The effect on
the first formant frequency is rather small.
V. SINUS PIRIFORMES A lowering of the larynx not only expands the sinus
piritorraes, but also increases the physical length of the
The dimensions of the sinus piriformes were estimated
from the tomograms traced in Fig. 3 under condi- pharynx tube. This length increase will expectediy rein-
force the effect of introducing the sinus pitiformes. This
tions of lowered larynx. In the tomograms the pockets
is because of the fact that adding the sinus piriformes
were found to possess a length of about 2 cm and a cross-
sectionalarea between1 and2 cm2oSucha pair of cav-
ities will resonate at a frequency of

!= c/4- l,, (3)

where f is the resonance frequency, c is the speed of
soundpropagation, and la is the effective length of the
tube. At the resonance frequency the sound fed into the
vocal tract will be absorbedby the sinus pitiformes
which will give a zero in the transfer function and hence
a minimum in the spectral envelope. Zero frequencies
were calculated for cress-sectionai areas (A) and lengths
(l) of the sinuspiriformes, using

l• =l +O.7 (A/•r)z/•-. (4)

Experimental data were provided by the model. The 0,5 IO t,,S 2,0

sinus piriformes were simulated by one or two cylindri- DEPTH OF SINUS PIRIFORNIIS(cm)
cal tubes of systematically varied len•d•s and diameters
inserted in parailel to the larynx tube into the closed end FIG. 5. Frequency of the transfer function zero stemming
of a wider tube simulating the pharynx anci the mouth. from various types of cavities simulating the sinus pitiformes
In Fig. 5 the calculateddata and measuredvalues can be in the acoustical model.. The cross-sectional area of the sinus
pirfformes was 1.2 and 3.5 cm2 (crosses and dots, respective-
compared. Note that the main determinant for the zero ly). The circled crosses show values obtained with two
frequency is the depth of the pockets whilst their cross- tubesof 1.2-cm2 cross-sectionalarea. The solid lines give the
sectional area is less import•tnt. Moreover, simulating calculated
values
[/= c/41e,is=
1+0.7(A/•r)
112].

J. Acoust Soc. Am., Vol. 55, No. 4. Alxil 1974

wnloaded 28 Mar 2011 to 130.237.67.192. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms.
 842 J. Sundberg:Singingformant 842

of the sinus pitiformes is counteracted when the sinus

Morgagni is expanded and [-he frequency of the third for-
mant is raised. This rise is typical for sungback vowels,
andit canbe ascribedto the more frontedtonguetip.
The maximum gain in spectrum amplitude around 3 kHz
is about 20 dB. This value agrees with the initially ob-
served differencebetweena spokenanda sung[u]. The
conclusion therefore seems justified that a singer is
capable of generating a "singing formant" by adjusting
his sinus Morgagni, sinus pitiformes, and in the back
vowels, the location of his tongue tip. Except for the
tongue tip movement these effects are presumably
secondary consequences of a lowering of [he larynx.

Vl. DISCUSSION
3 6 7 9 11 13
Our results suggest that a singer is able to produce a
DISTANCECONSTRICTION
CEHTER-OPENEND (cm)
vowel with a "singing formant" using his vocal chords
and vocal tract exclusively. Given the explanation sug-
FIG. 6. Formant frequency values observed in an acoustical gested above, there is no motivation to postulate "chest
model of the vocal tract as a function of the location of a con- resonance," "head resonance," "singing in the mask,"
striction. Dotted lines: without simulation of sinus pirfformes. etc., i.e., they appear to lack acoustical relevance for
Squares: with simulation of sinus piriformes. $oli•t lines: the tone. On the other hand, these terms play a predom-
simulation of sinus pirfformes and lengthening of the pharynx inant role in current vocal pedagogy. The reason for
tube by 1.5 cm. this might be that these terms describe patterns of vi-
bration sensaUons that appear only when the tone is prop-
erly produced. Perhaps recognizing the vibrations ac-
can be interpreted as an increased pharynx length. This companying good tones is an efficient way to control vo-
assumption is confirmed by the experimentai data given cal production.
m Fig. 6. Note, however• that the length increase ap-
From the acoustical point of view it is essential for
pears to reduce the variability of the fifth formant fre-
the generation of the "singing formant" that the larynx
quency quite considerably: it keeps a value close to 3
tube act as a separate resonator the resonance of which
kHz in all articulations except one.
is not a/fected by articulatory movements in the rest of
The summed effects on the form•mt frequencies of in- the vocal tract. This is accomplished when the larynx
troducing the sinus piriformes and of increasing the phar- tube opening is less than one-sixth of the cross-sec-
ynx leng• are rather similar to the differences observed tionai area in the pharynx. AFdculatdrily, this effect is
in the formant frequencies in speechand singing (cf. obtainedby loweringthe lary/•x whichwidensthe pharynx
Figs. 2 and 6). In both cases we observe a fourth for- cavity. Particulariy at higher pitches where the larynx
mant frequency essentially unaffected by the articulation, tube opening is wide it would be essential to acqmre a
a fifth formant that is constantly.tow in frequency, and a wide pharynx. It was mentioned that an increase in pitch
second formant that drops considerably under conditions
of fronted constriction. Differences occur in the first
formant frequency, which drops in the model experiments EFFECT ON SPECTRUM ENVELOPE 01: SIHUS PIRIFORMIS
but is slightly higher in singing than in speech. This ARTICULATION:
[u] •
discrepancy is probably explained by the fact that the jaw A 3O
opening tends to be larger in singing than in speech, since
a wide jaw opening tends to raise the first formant fre-
quency.•v Anotherdiscrepancyis the behaviorof the
third formant, which• however, can be explained by the
sensitivity which this formant exhibits to the position of
the tongue tip. Therefore, the experiments seem to
suggest that the formant frequency differences observed - - "-•-.-•-_•:.o---o' ," -.
between spoken and sung vowels can to a large extent be
explained with reference to changes in the pharynx and
larynx tubes associated with a lowering of the larynx.
Shifts in formant frequencies are accompanied by ai-
, ,
ternLions of the spectrum envelope. Measured changes FREQUENCY (kHz)
in the spectrum envelope due to the presence of a wide
FIG. ?. Spect• enve[o• c•es due • the int•d•c•[on o•
sinus Morgagai, the sinus piriformes, and to a rise in sinus p[•Eo;mis (s. p. ), sinus •o•;a• (s.M.), end • •sed
[he third formant frequency can be studied in Fig. 7.
The values were obtained from the model sLmulating an an acoustic m•el • the vocal tract including •o •t•etions
articulationsimilar to that of an [u]. The daxnping
effect so as • sim•a/e the artic•ation of a • [u].

J. Acoust. Soc. Am., Voh 55, No. 4, April 1974 .

ownloaded 28 Mar 2011 to 130.237.67.192. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms.js
 843 J. Sundberg:
Singinfl
formant 843

is associated with an increase of the larynx tube open- suggests. The reasonfor Beckmann'snegativeresults
ing. This increase has to be compensatedfor by a wid- as regards the acoustical effect of the laryngeal ventri-
ening of the sinus Morgagni if the larynx tube resonance cle seems to be that his spectrum analysis was not ex-
is to remain at 2.8 kHz. A contribution to the widening tendedbeyond2 kHz.9 As hasbeendemonstrated
above,
required is provided by the stretching of the vocai folds effectscanbe expectedaround2.5 kHz. Also, filling a
constituting the bottom of the sinus Morgagni. But prob- cavity with cotton does not mean that it is acoustically
ably additional compensation is required as well. If so, "ausgeschaltet." This fact has already been pointed out
the singer will have to lower his larynx more and more by Mermelstein in his criticism of Finch and Schwickar-
the higher the tones he sings. This agrees with the com- die's work. t2 Mermelstein's conclusionswith respect
moniy stated need for "covering" high notes, since cov- to the acoustical effect of the sinus piriformes are not
ering has been shown to be associated with phenomena supported by our findings. The reason for this dis-
similar to the effects observed in larynx lowering. agreement
is thatweregard
these
pockets
asa partOf
the pharynx cavity. This is necessary as soon as there
There are great individual differences as regards the
is an acoustical mismatch between the pharynx and the
dimensions of the vocal chords, the sinus Morgagm, and
larynx. Apparently, Mermelstein did not count on the
the sinus piriformes. We may concinde from the pres-
possibility of this mismatch. His conclusions are ap-
ent investigations that a good singer needs a large sinus
plicable when the pharynx and the larynx are not mis-
Morgagni and a wide pharynx. This conclusion may ex- matched.
plain the predominance of large sinus Morgagni in
singersobservedby Finch.s VII. CONCLUSIONS

According to Finch, the sinus Morgagni in females is The acoustically most essential articulatory difference
smaller than in males. Also, accordingto Finch, the between maie speech and singing appears to be the wid-
differences betweenpoor and goodvoices are accom- ening of the pharynx at the level of the larynx tube open-
paniedby a larger difference in the sinus Morgagni in ing. Such a widening of the pharynx seems to be accom-
males than in females. Bartholome•r t has observed that panying a lowering of the larynx. Acoustically, the
soprano voices frequently lack a "singing formant" in pharynx widening has the effect of isolating the larynx
their higher range. In view of our results we may as- tube from the rest of the vocal tract so that its reso-
sume that the acoustical mismatch between the pharynx nance frequency is not altered by articulatory move-
and larynx tubes cannotbe maintained in the upper range ments outside it. Its resonance can be tuned to a fre-
of a soprano. Under these conditions, no "singing for- quency between the third and fourth formant in normal
mant" is likely to be produced. speech by adjusting the volume contained in the sinus
Morgagni to the opening area of the larynx tube. An ad-
Professional singers seem to adopt a special vowel
ditional effect of a larynx lowering is that it increases
articulation in singing with the result that a "singing for-
the dimensions of the sinus pitiformes and the length of
mant" is generated. This suggests that this spectrum
the pharynx tube. This lowers the fifth formant frequen-
envetope peak is desirable. At the same time this can-
cy and in front vowels also the second. In this way a
not be the case in high-pitehed female singin_g. The rea-
great deal of the acoustical differences between spoken
son for this dissimilarity probably lies in a perceptual
and sung vowels in male voices can be explained with
function of the "singing formant." A long-time-average
reference to the larynx lowering, the major articulatory
spectrum of a symphony orchestra has a peak around
gesture associated with the production of a "singing
450 Hz and a slope of about - 10 dB/octave abovethis formant."
frequency.Thus, the "singingformant"is locatedin'a
frequency region where the average soundpressure level VIII. ACKNOWLEDGMENTS
of an orchestral soundhas dropped more than 20 dB be-
low its maximum vaiue. This reduces the risk that the The author gratefully acknowledgesvainable discus-
singer's voice is maskedby an orchestral accompani- sions with J. Lindqvist-Gauffin, M. Sondhi, and B.
ment. The risk of masking is probably smaller in high- Lindblom. He is also indebted to Mrs. S. Felicetti for
pitched female singing, as all partials are higher in fre- editing the manuscript. The work was supported by the
quency than the strongest sounds from the accompani- Bank of Sweden Tercentenary Fund.
ment. Rs

Our results agreewith thosepresentedby Fant.• They

iSee, e.g., N. T. Bartholomew,"The Role of Imageryin Voice
do not contradict van den Berg's interpretation of the
Teaching," Nat. Assoc. of Teachers of Singing, USA, Re-
laryngeal ventricle, since he considers this cavity in search Committee Release No. 1 (May 1951).
isolation.to On the otherhand,the laryngeaiventricle 2G. Fant. "Onthe Predictabilityof FormantLevelsandSpec-
belongs to a resonant system: the entire larynx tube. trum Envelopesfrom Formant Frequencies," in Fo• Roman
As demonstrated above, this tube may act as a separate Jal•obson(Mouton, The Hague, 1956), pp. 109-120.
resonator. From the point of view of the acoustical ef- 3jo Sunalbert."The SourceSpectrumin Prof•aional Singing,"
Fol. Phoniat. 25, 71--90 (1973).
fect of the ventricle, it seems better to regard it as a
4j. Sundberg,"FormantStructureandArticulationof Spoken
part of the larynx tube. Since it contributes to the prop- and SungVowels," Fol. Phoniat. 22, 28-48 (1970).
erties of the sound transfer system, it appears motivated SG.Fant, G. Henningsson,
andU. St•lhammar,"FormantFre-
to consider it as belonging to that system instead of re- quencies of Swedish Vowels," STL-QPSR, Speech Transmis-
garding it as a part of the voice source, as van den Berg sion Lab., Quarterly Progress and Status Report, KTH,

J. Acoust. Soc. Am., VoL 55, No. 4, April 1974

wnloaded 28 Mar 2011 to 130.237.67.192. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms.
 844 J. Sundberg:Singingformant 844

Stockholm, No. 4 (1964) pp. 26-31. Quarterly Progress and Status Report, Stockholm, No. 2-3
•See, e.g., the review over this field offeredby J. Large, "To- (1971), pp. 43--52.
wards an Integrated Physiologie--Acoustie Theory of Vocal laU. Ingard,"OntheTheoryandDesignof Acoustic
Resonators,"
Registers," Nat. Assoc. of Teachers of Singing, USA, Bul- J. Aeonst. Soe. Am. 25, 1037-1067 (1953).
let;tu (Feb. -March 1972), p. 18 if. 15Intheequationquoted(Eq. 4 in Ingard'sarticle) an indexhas
?G. Font, AcousticTheoryof SpeechP•roductiou
(Mouten,The been dropped out. In his previous work on the acoustics of
Hague, 1970), 2rid ed. the larynx tube [STL--QPSR No. 1 (1972), pp. 45--53], the
øM. Flach, "'lJberdie unterschiedliche
Gr6sseder Morgagn- author of the present paper did not observe this and this led
ischen Ventrikel bei S•ngern," Fol. Phoniat. 16, 67-74 (1964). him to erroneously claim that the larynx tube acts as a Helm-
sG. Beckmann,"DosVerhaltender Ober•ne bei der akustis- holz resonator.
chen Ausschaltungder Kehlkopfventrikel," Fol. Phoniat. 10, I•J. Lindqvist,"LaryngealArticulationStudiedonSwedishSub-
149--153 (1958). jects," STL-QPSR, Speech Transmission Lab., Quarterly
1øJw.vandenBerg, "Onthe Roleof theLaryngealVentricle ProgressandStatusReport, Stockholm,No. •-3 (1972), pp.
in Voice Production," Fol. Phoniat. 7, 57-69 (1955). 10-27.
IiM. FinchandH. Schwickardie,
"Die RecessusPitiformes •?B.LindbiotaandJ. Sundberg,"AcousticalConsequences
of
unter phoniatrischer Sicht," Fol. Phoniat. 18, 153-167 (1966). Lip, Tongue, Jaw, and Larynx Movement," J. Acoust. SOc.
i2p. Mermelstein, "On the Pitiform Recessusandtheir Acous- Am. 50, 1166-1179 (1971).
tic Effects," Fol. Phoniat. 19, 388--389 (1967). 18j. Sundberg,"ProductionandFunctionof the Singing
13F.FranssonandE. Jansson,"Propertiesof theSTL-Iono- Formant," Proc. of the Eleventh Congr. of the International
phone transducer," STL-QPSR, Speech Transmission Lab., Musicological Society Copenhagen, 1972 (forthcoming).

J. Acoust. Soe. Am., Vol. 55, No. 4, April 1974

ownloaded 28 Mar 2011 to 130.237.67.192. Redistribution subject to ASA license or copyright; see http://asadl.org/journals/doc/ASALIB-home/info/terms.js