See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/6346613

Musical Theater and Opera Singing-Why So Different? A Study of Subglottal

Pressure, Voice Source, and Formant Frequency Characteristics

Article in Journal of voice: official journal of the Voice Foundation · June 2007
DOI: 10.1016/j.jvoice.2006.12.007 · Source: PubMed

CITATIONS READS

74 1,298

1 author:

Eva Björkner
Kulturskolan i Kungsbacka
23 PUBLICATIONS 490 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Thesis View project

Linguistic and extra-linguistic parameters for early detection of cognitive impairment View project

All content following this page was uploaded by Eva Björkner on 11 December 2017.

The user has requested enhancement of the downloaded file.

Musical Theater and Opera Singing—Why So
Different? A Study of Subglottal Pressure, Voice
Source, and Formant Frequency Characteristics
Eva Björkner, Stockholm, Sweden

Summary. The considerable voice timbre differences between musical theater (MT) and western operatic singers are
analyzed with respect to voice source and formant frequencies in five representatives of each singer group. Audio, sub-
glottal pressure (Psub), and electroglottograph (EGG) signals were recorded while the subjects sang a sequence of [pae:]
syllables starting at maximal vocal loudness and then gradually decreasing vocal loudness. The task was performed at
each of two fundamental frequencies (F0), approximately one octave apart. Ten equally spaced Psub values were then
selected for each F0. The subsequent vowels were analyzed in terms of flow glottograms derived by inverse filtering the
audio signal, which also yielded formant frequency data. Period time (T0), peak-to-peak pulse amplitude (Up-t-p), and
maximum flow declination rate (MFDR) were measured from the flow glottograms while closed quotient Qclosed (Tcl/T0)
was determined in combination with the differentiated EGG signal. Also the relationship between the first and the sec-
ond harmonic in the spectrum (H1–H2), the amplitude quotient (AQ), that is, the ratio between Up-t-p and MFDR, and
normalized AQ, that is, AQ normalized with respect to period time was calculated as well as the sound pressure level.
The results showed that both the MT and the opera singers varied their Psub systematically, approximately doubling Psub
for a doubling of F0. For a given value of Psub, the MT singers produced higher values of MFDR, Up-t-p, and Qclosed, and
lower values of H1–H2, indicating a weaker fundamental. Further, the MT singers showed higher formant frequencies
and did not show the opera singers’ characteristic clustering of F3, F4, and F5.
Key Words: Musical theater singing–Opera–Subglottal pressure–Inverse filtering–EGG–Voice source–Formant
frequencies–Closed quotient–H1–H2.

INTRODUCTION
Vocal music is a large artistic field where musical theater (MT),
pop, rock, soul, folk music, and jazz occupy an important por-
tion. Voice timbre varies widely between these styles and differs
considerably from the timbres used in western opera. In the
past, most scientific studies of the singing voice have been de-
voted to classically trained singing. In the last two decades,
many authors have examined MT singing. For instance, Miles
and Hollien,1 Estill,2 Sundberg et al,3 Bestebreurtje and
Schutte,4 and Stone and collaborators5 studied female belting
and Björkner and associates6 registered differences. To the
author’s knowledge, however, no studies have been devoted
to male MT singers’ voice production.
Opera singers and MT singers meet quite different working
demands: MT singers use microphones and might be required
to sing seven to eight performances a week, whereas opera
singers sing without electronic amplification and rarely have
more than two or three performances per week. Therefore, a bet-
ter understanding of voice production in MT singing is impor-
tant with regard to both coaching and vocal health.
Differences in voice quality reflect differences in muscular,
aerodynamic, and acoustical conditions in the larynx and in
the vocal tract. The subglottal pressure (Psub) is the essential
Accepted for publication December 19, 2006.
This investigation was first presented at the 3rd Physiology and Acoustics of Singing
Conference (PAS3-06), York, England, May 11–13, 2006 and at the 35th Annual Sympo-
sium: Care of the Professional Voice, Philadelphia, PA, May 31–June 4, 2006.
From the Department of Speech Music and Hearing, KTH, Stockholm, Sweden.
Address correspondence and reprint requests to Eva Björkner, KTH—Royal Institute
of Technology, TMH—Department of Speech, Music and Hearing, Lindstedtsvägen 24,
SE-100 44 Stockholm, Sweden. E-mail: bjorkner@csc.kth.se
Journal of Voice, Vol. 22, No. 5, pp. 533-540
0892-1997/$34.00
Ó 2008 The Voice Foundation
doi:10.1016/j.jvoice.2006.12.007
driving force in vowel production and is also the primary vari-
able for controlling vocal intensity.7,8 Inverse filtering9 is
a widely used technique to eliminate the contribution of vocal
tract resonances during vowel production. The resulting flow
glottogram waveform reflects important voice source proper-
ties, such as the glottal opening and closing, and allows calcu-
lation of a number of other important voice characteristics.
This study compares voice source characteristics and for-
mant frequencies in five male MT and operatic singers, respec-
tively, over a wide range of vocal dynamics at two fundamental
frequencies (F0).
MATERIAL AND METHODS
Subjects and recording
Five classically trained baritones (age range 29–65 years) and
five MT singers (age range 30–44 years), all Swedish and earn-
ing their livelihood from stage singing, volunteered as subjects.
The recording of the classical singers has been described
elsewhere.10 The protocol for the MT group was the same as
for the opera singers and took place in an anechoic chamber
(Linguistic Department at Stockholm University). Audio sig-
nals were captured by means of a Bruel & Kjaer condenser
microphone (B&K 4190 2435611, Naerum, Denmark) and
two-channel electroglottograph (EGG) signals recorded with
a Glottal Enterprises EG-2 equipment (Syracuse, NY, http://
www.glottal.com/). Intra oral pressure, used as an estimate of
the Psub,11 was captured during the [p]-occlusion with a GEL-
TEC (Tokyo, Japan) pressure transducer, mounted on the tip
of a catheter which the singer held in the corner of his mouth.
The microphone distance was set to 50 cm from the subject’s
mouth thus reducing the risk of sound pressure level (SPL) error
resulting from singer’s movement. This distance was repeatedly
534
checked during the recording session. Audio signal was cali-
brated using a B&K sound level calibrator (type 4231). Calibra-
tion of the pressure signal was achieved by lowering the
pressure transducer to the bottom of a water-filled glass cylinder
providing a well-defined reference pressure. For each subject,
separate calibrations were made. All signals were recorded
simultaneously at a sampling rate of 16 kHz using the Sound-
swell Signal Workstation (Hitech Development AB, Stockholm,
Sweden, http://www.hitech.se/development/).
Before the recording session, each singer had 5–10 minutes
for vocal warm-up. Their task was to sing a diminuendo at
a constant pitch while repeating the syllable [pae:], starting
from the maximum degree of vocal loudness with high lung
volume and continuing to the softest possible phonation (Fig-
ure 1). The sequence was repeated three times at each of two
F0, one octave apart, 139 Hz (C#3) and 278 Hz (C#4), respec-
tively. The singers were asked to sing in the chest/modal regis-
ter, also called mechanism 1.12 Nevertheless, some of the
singers’ softest phonations were produced in falsetto register.
Pressure measurements
To improve the accuracy of the pressure measurements, the
singers were instructed to sing the sequence legato using unas-
pirated [p]. The pressure peaks were continuously monitored
during the recording.
By calculating the extremes of the singers’ total Psub range
from the three takes and dividing this range by nine, 10 equally
spaced Psub values were obtained. The measured Psub values ly-
ing closest to these 10 values were then identified and the sub-
sequent vowels were selected for analysis. Because the subjects
continuously decreased vocal loudness while repeating the syl-
lable [pae:], Psub decreased somewhat during each vowel. This
resulted in a marginal overestimation of the Psub value associ-
ated with the subsequent vowel.
Psub values were obtained in terms of the actual values in
cm H2O. They were also expressed in terms of pressure num-
bers (1–10), representing Psub normalized with respect to the
singers’ total Psub range, henceforth referred to as relative Psub.
Inverse filtering and flow glottogram measurements
Inverse filtering eliminates the vocal tract resonances from the
recorded signal and offers a flow glottogram, that is, a visualiza-
FIGURE 1. Audio (top) and pressure (bottom) signals recorded when the singer subject sang a diminuendo, starting at maximum degree of vocal
loudness, at a constant pitch while repeating the syllable [pae:].
Journal of Voice, Vol. 22, No. 5, 2008
tion of the variation of the transglottal airflow with time. A cus-
tom-made inverse filtering program Decap (S. Granqvist) was
used, displaying the waveform and the spectrum before and af-
ter filtering. The formant frequencies were adjusted such that
a ripple-free closed phase was obtained in the waveform with
a smoothly sloping spectral envelope. Thus, the result of inverse
filtering provided information on voice source characteristics
and formant frequencies.
Period time (T0), Qclosed (Tcl/T0), peak-to-peak pulse ampli-
tude (Up-t-p), and maximum flow declination rate (MFDR)
were measured and averaged from four adjacent periods in
the middle of each vowel sample. Then, the ratio between Up-
13,14
t-p and MFDR, that is, amplitude quotient (AQ) and the nor-
15
malized version (NAQ) were calculated. This procedure for
inverse filtering has been found to yield reliable information;10
although for very soft phonation, the flow glottogram wave-
forms and EGG signals were often close to sinusoidal thus pre-
venting accurate waveform analysis.
The closed phases of the opera singers were measured from
the flow glottogram waveforms only, whereas those of the MT
singers were measured both from the differentiated EGG signal
(DEGG) and from the flow glottograms. In most cases, the
closed phase values measured from flow glottograms were iden-
tical to those derived from the DEGG signal. Figure 2 shows an
example. In case of double opening peaks in the DEGG sig-
nal,16 the flow glottogram was used.
The analysis also included some acoustic characteristics,
such as the level difference between the first and the second har-
monic (H1–H2), format frequencies, and the SPL.
RESULTS
The entire material was submitted to two statistical two-way
analyses of variance (ANOVAs). The first concerned differ-
ences between the two groups of opera and MT singers with
style and pitch as factors (Table 1A). The second concerned dif-
ferences within each group of singers, with singer and pitch as
factors (Table 1B).
Pressure
Large individual differences in Psub range were found between
the singers, particularly for the higher F0. As expected both
groups of singers varied Psub quite systematically, using a higher
Eva Björkner Musical Theater and Opera Singing 535

TABLE 1.
Results of the Statistical Two-Way ANOVA Tests:
Univariate ANOVA (A) Between the Two Groups of
Opera Singers and MT Singers and (B) Within Each Group
of Singers
(A) Between the Two Groups of Singers

Dependent Variable Style Pitch

Psub (cm H2O) ns s
MFDR (1/s2) s ns
Up-t-p (1/s) s s
AQ ns s
NAQ ns s
Qclosed (%) s ns
SPL (dB) s s
F1 ns ns
F2 s ns
F3 ns ns
F4 s s
F5 s ns
FIGURE 2. Typical example of the agreement between the EGG
(top), DEGG (middle) signals, and flow glottogram (bottom) with re- (B) Within Each Group of Singers
gard to the discontinuities used for identifying the beginning and the
end of the closed phase. MT Singers Opera Singers

Dependent Variable Pitch Singer Pitch Singer

Psub not only for increasing vocal loudness but also for increas-
ing pitch, as can be seen in Figure 3. On average, both singer Psub (cm H2O) s s s s
MFDR (1/s2) s s s s
groups doubled their mean Psub for a doubling of F0, but the var-
Up-t-p (1/s) ns s ns s
iation within the group of MT singers was quite large; one
AQ s ns s ns
singer showed similar values for both F0 values, whereas NAQ s ns s s
another used three times higher Psub for the upper F0. The Qclosed (%) ns s ns ns
ANOVA showed a significant difference for factor pitch. SPL (dB) s s s s
Furthermore, the MT singers tended to use a somewhat Note: Significant with alpha ¼ 0.05.
higher Psub than the opera singers for 139 and 278 Hz. This dif- Abbreviations: s, significant; ns, nonsignificant.
ference was, however, not significant according to the ANOVA
test.
For the very softest phonation (pressure #10), that is, at pho-
nation threshold, the MT singers showed 1.3 cm H2O higher
values than the opera singers for the lower F0. For the higher
F0, the corresponding value was 2.0 cm H2O.

Flow glottogram data

Mean ranges and maximum values of all parameters are pre-
sented in Table 2. Clear differences can be observed between
the groups. Table 3 lists the correlations (best linear fit) between
mean Psub and the means of MFDR and Up-t-p, respectively, in
terms of the squared correlation (R2), the slope, and intercept
for the high and the low F0. For mean SPL, the values represent
the best fit with the log mean Psub. The correlations were quite
strong throughout.
MFDR reflects vocal intensity in a form that it is unaffected
by harmonics and formants as opposed to SPL. In Figure 4, the
MT and opera singers’ mean MFDR values are plotted as a func-
tion of mean Psub. MFDR increased with increasing Psub. The
MT singers’ slightly higher Psub values were associated with
clearly higher MFDR values, especially at 278 Hz. The FIGURE 3. Mean pressures for the 10 selected Psub values observed
ANOVA (Table 1A) confirmed this observation; MFDR was the at the indicated F0 values as produced by the MT and opera (Op)
only glottal parameter that showed significance for both factors’ singers.
536 Journal of Voice, Vol. 22, No. 5, 2008

TABLE 2.
Variation Range of the Means and the Highest Individual Value of the Indicated Parameters at the Two F0 Values for the MT
and the Op Singers
MT 139 Hz MT 278 Hz Op 139 Hz Op 278 Hz

Mean Max Mean Max Mean Max Mean Max

Psub (cm H2O) 5–22 31 7–39 62 4–18 21 5–36 53
Up-t-p (1/s) 0.37–1.3 1.7 0.19–1.2 1.9 0.18–0.77 1.2 0.11–0.81 1.5
MFDR (1/s2) 325–2447 3872 213–3518 7860 141–1472 2321 168–1908 3556
Qclosed (%) 29–59 73 25–62 65 22–47 58 22–47 54
SPL (dB) 62–90 97 67–94 101 57–83 89 62–91 97
H1–H2 (dB) 4–15 24 4–18 26 5–21 25 5–23 23
Abbreviation: Op, opera singers.

style (P < 0.000) and pitch (P < 0.007). This suggests that
MFDR might be a particularly informative parameter when an-
alyzing glottal characteristics.
The relationship between Up-t-p and Psub is illustrated in Fig-
ure 5, showing mean Up-t-p values as a function of mean Psub.
Like MFDR, the Up-t-p increased with increasing Psub and the
MT singers showed clearly higher Up-t-p values than the opera
singers for both F0 values. Accordingly, factor style (Table
1A) was statistically significant for Up-t-p. Both groups showed
higher Up-t-p values for the low F0, especially for soft phonation.
This is not surprising because at high F0 the vocal folds are
stiffer, thus allowing less air through the glottis.
Mean Qclosed as a function of mean Psub is shown in Figure 6,
and the constants of power function approximations of this re-
lationship are listed in Table 4. The power function approxima-
tion was of type
Qclosed ¼ A  eða½Psub ÞB (1)
where A, a, and B are constants. Qclosed increased with increas-
ing Psub and seemed to saturate for the loudest phonations.
Within each group, Qclosed was similar for the two F0 values,
but large differences could be observed between the groups.
Compared to the opera singers, the MT singers showed higher
Qclosed values (Table 2), and a slightly higher asymptote value,
particularly for the upper F0. The highest individual value of
Qclosed found within the MT group was 73% and 65%, and in
the opera group 58% and 54%, for 139 and 278 Hz, respec-
tively.
The level difference between the first and the second har-
monic in the spectrum (H1–H2) has been found to be linearly re-
lated to Qclosed, a high value of Qclosed being associated with
a weaker fundamental.17 Figure 7 shows the relationships be-
tween mean Qclosed and mean H1–H2 (left panel), and the rela-
tionship between mean relative Psub and mean H1–H2 (right
panel). For a given Qclosed, the MT singers’ H1–H2 tended to
be greater than for the opera singers, particularly at the higher
F0. The opera singers, however, did not reach as high values
of Qclosed as the MT singers. This means that, for a given rela-
tive Psub value, the opera singers produced a stronger funda-
mental than the MT singers (Figure 7, right panel).
In Figure 8, the five lowest formant frequencies are given, av-
eraged across subjects. The MT singers showed higher formant
frequencies than the opera singers, who also tended to keep for-
mant frequencies unaffected by F0, tuning F3, F4, and F5
closely together in a cluster. The MT singers, on the other
hand, showed no signs of such strategy. Rather they showed
clear differences between the two F0 values, especially for
F2, F4, and F5. According to the ANOVA (Table 1), factor style
was significant for F2, F4, and F5, and both style and pitch were
significant for F4.
In Figure 9, mean SPL values are shown as a function of
mean relative Psub. For a given mean relative Psub, the MT
singers showed higher SPL values for both F0 values compared
to the opera singers. The ANOVA tests (Table 1A and B)
showed style and pitch as highly significant for SPL.
Figure 10A and B shows mean NAQ and AQ values, respec-
tively, for the two groups as a function of mean Psub. Both NAQ

TABLE 3.
Correlation Squared (R2), Slope and Intercept of the Best Linear Fit for Mean Values of MFDR, Up-t-p as Function of Mean
Psub, and of the Best Log Fit of the Relationship Between SPL at 0.5 m and Mean Psub at the Two F0 Values for the MT and the
Op Singers
MT 139 Hz (C#3) MT 278 Hz (C#4) Op 139 Hz (C#3) Op 278 Hz (C#4)

R2 Slope Intercept R2 Slope Intercept R2 Slope Intercept R2 Slope Intercept

2
MFDR (1/s ) 0.934 151.1 499.7 0.961 123.2 688.4 0.979 94.86 208.12 0.955 62.33 120.69
Up-t-p (1/s) 0.962 0.0586 0.099 0.949 0.0341 0.02 0.971 0.041 0.034 0.939 0.022 0.057
SPL at 0.5 m 0.965 21.394 25.97 0.974 7.44 31.96 0.973 17.39 35.12 0.993 16.41 34.32
Abbreviation: Op, opera singers.
Eva Björkner Musical Theater and Opera Singing 537

FIGURE 4. Mean MFDR as a function of mean Psub for the MT and FIGURE 6. Mean Qclosed as a function of mean Psub for the MT and
the opera (Op) singers at the indicated F0 values. The bars represent the opera (Op) singers at the indicated F0 values. The bars represent
±one standard error. ±one standard error.

and AQ decreased with Psub. The NAQ values were surprisingly
similar between the groups but differed with the two F0 values
(Table 1A). This difference was clearly smaller for the AQ
values (Figure 10B).
DISCUSSION AND CONCLUSION
The aim of this study was to investigate voice production fac-
tors underlying the timbral differences between MT and oper-
atic singing. The singers appeared to produce typical samples
of their respective singing style and our measurements revealed
clear differences, likely to be characteristic for the two groups.
We asked our singers to repeatedly sing the syllable [pae:]
from the loudest to the softest, starting at high lung volume.
Thus all soft phonations were sung with a low lung volume.
Iwarsson et al18 found that phonation at low lung volumes
tended to increase glottal adduction in untrained singers. On
the other hand, Thomasson19 observed that this effect did not
exist in trained operatic singers. Thus, there are no reasons to
assume that our opera singers increased glottal adduction
with decreasing lung volume. Their NAQ values were actually
lower than those of the MT singers. This suggests that the lung
volume decrease did not affect the phonation type in either of
our singer groups.
In the previous investigation of our group of opera singers,
the reproducibility of flow glottogram measurements was found
to be quite high.11 For the MT singers also, an EGG signal was
included in the present investigation. The good agreement be-
tween the DEGG signal and corresponding flow glottogram
for the closed phase measurements (Figure 2) was encouraging,
showing that identification of the closed phase from flow glotto-
gram yielded reliable results.
The MT singers used slightly higher Psub throughout their dy-
namic ranges both for 139 and for 278 Hz, and on average their
threshold pressures, that is, the lowest pressures needed for vo-
cal fold vibration, were higher than those of the opera singers.

TABLE 4.
Mean Values of A, a, and B for Best Power Function
Approximations of the Relationship Between Mean
Qclosed and Mean Psub at the Two F0 Values for the MT and
the Op Singers
A a B Sum Sq
MT 139 Hz 0.58 0.10 0.545 0.056
MT 278 Hz 0.65 0.75 0.431 0.034
Op 139 Hz 0.47 0.23 0.755 0.069
Op 278 Hz 0.49 0.06 0.713 0.020
Notes: The approximation is of type Qclosed ¼ A  eða½Psub ÞB . Sum Sq rep-
FIGURE 5. Mean Up-t-p as a function of mean Psub for the MT and resents the average summed squared difference between the data points
the opera (Op) singers at the indicated F0 values. The bars represent and this approximation.
Abbreviation: Op, opera singers.
±one standard error.
538 Journal of Voice, Vol. 22, No. 5, 2008

FIGURE 7. (Left panel) Mean H1–H2 as a function of mean Qclosed for the MT and the opera (Op) singers at the indicated F0 values. The bars
represent ±one standard error. (Right panel) Mean H1–H2 as a function of 10 selected pressure numbers, that is, in terms of percentage of the singers’
total pressure ranges. Low numbers refer to high pressures. The bars represent ±one standard error.

As mentioned in the Introduction, most investigations of
singing have been devoted to male classically trained singers.
Few studies have presented data on MT singing that could be
compared to our findings. However, Cleveland and co-
workers,20 studying country singers, found higher Psub values
than those typically observed in classically trained singers.
We made similar observations. Also, they found that the SPL
for a given Psub was lower than the SPL produced by operatic
singers, whereas we found that our MT singers produced higher
SPL than our opera singers. On the other hand, country singing
and MT singing are two different singing styles. Further, Hen-
rich et al12 found no strong correlation between open quotient
and F0 in classically trained male singers. Our results for the
MT singers are in good agreement with this observation but
not for the opera singers.
The increase of intensity level for a doubling of Psub has been
measured in untrained voices. Fant7 reported an increase of be-
tween 9 and 13 dB, and Holmberg et al21 observed similar
values. In an investigation of operatic tenors, Sundberg and
Högset22 found an increase of about 10 dB. Our values for
the MT singers amounted to 15 dB for the high F0 and 12 dB
for the low F0. The corresponding values for the opera singers
were 11 and 12 dB for the low and the high F0, respectively.
Thus, our results on the increase of SPL for a doubling of
Psub are somewhat higher than those previously reported.
As we analyzed a comparatively high number of pressure
values, our investigation offers a rather detailed view of the re-
lationship between Psub and various voice source parameters.
The MT singers tended to produce higher Up-t-p values, higher
SPL, and higher MFDR than the operatic singers. In accordance
with results reported by Sundberg et al,23,24 for singers and un-
trained voices, respectively, we found that Qclosed reached an
asymptote at high values of Psub, but the MT singers’ asymptote
was significantly higher than that of the opera singers. The MT
singers’ higher asymptote seems hard to explain with their mar-
ginally higher Psub, but may be characteristic of MT singing.
H1–H2 has been found to be inversely proportional to Qclosed,17

FIGURE 8. Means of the five lowest formant frequencies for the in-
dicated F0 values. Filled columns represent the MT and striped bars FIGURE 9. Mean SPL as a function of mean Psub for the MT and the
represent the opera (Op) singers. The bars represent ±one standard opera (Op) singers at the indicated F0 values. The bars represent ±one
error. standard error.
Eva Björkner Musical Theater and Opera Singing
FIGURE 10. Mean NAQ (left panel) and mean AQ (right panel) as a function of mean Psub at the indicated F0 values for the MT and the opera (Op)
singers. The bars represent ±one standard error.
such that a greater Qclosed is associated with a smaller H1–H2
difference. The smaller H1–H2 observed for the MT singers,
therefore, seems related to their higher values of Qclosed. Possi-
bly, it belongs to the voice characteristics of this style of sing-
ing.
The NAQ and AQ, on the other hand, did not differ between
the singer groups. If these measures reflect phonation
mode,14,24 this would imply that the MT singers did not use
a more hyperfunctional phonation mode than the operatic
singers. It would be interesting to find out if less successful
MT singers also show NAQ and AQ values similar to those
found here in well-established professional representatives of
this singer group.
Another clear difference was found in the formant frequen-
cies. All MT singers’ formant frequencies were higher than
those of the opera singers, particularly F2, F4, and F5. Similar
differences between styles of singing were reported by Sund-
berg et al3 and Stone et al.5 Presumably, the formant frequen-
cies, reflecting articulatory characteristics, account for a good
deal of the timbral differences between opera and MT singing
styles.
Above it was argued that the MT singers’ higher Qclosed
values must be produced by some other factor than Psub. Ac-
cording to Hirano and colleagues,25 thick vocal folds are likely
to produce a long closed phase, and a thickening of the folds re-
sults chiefly from contraction of the vocalis muscle. They fur-
ther argued that the vocalis contraction is typically weaker in
falsetto register phonation compared to modal register phona-
tion. As the register transition generally occurs around 300–
400 Hz in a male voice,26 both F0 values examined in the pres-
ent study (139 and 278 Hz) were produced in the same register,
except for some examples of higher F0 in the softest phonation.
Thus, no register effects are likely to have contributed to the
main trends in our data. Taking all this into consideration, it
seems reasonable to hypothesize that the MT singers sang
with a firmer contraction of the vocalis muscle.
Singing is an accurately controlled type of phonation. This
was clearly illustrated by the strong correlation found between
mean Psub and the various flow glottogram parameters (Table
539
3). This suggests that the use of singers as subjects is advanta-
geous in investigations of relationships between various phona-
tory parameters.
According to current views, our MT singers’ high values of
Qclosed would be interpreted as signs of pressed phonation. It
seems reasonable to doubt that the habitual use of pressed pho-
nation is acceptable from a vocal hygiene point of view in pro-
fessional singing. The AQ and NAQ parameters, previously
found to reflect phonation mode in speech,15 showed quite sim-
ilar values between our singer groups. This similarity suggests
that our MT singers used a phonation mode no more pressed
than that used by our opera singers.
Returning to the question raised in the title of this investiga-
tion, our results have shown both differences and similarities
between MT and opera singers. It seems likely that two factors
play a major role in accounting for the timbral differences be-
tween these two groups of singers. One concerns the formant
frequencies, particularly the presence and absence of the sing-
er’s formant cluster. The other is related to the opera singers’
greater H1–H2 difference, indicating that the MT singers have
a weaker voice source fundamental. These hypotheses need to
be tested in the future by means of listening experiments with
synthesized stimuli.
Acknowledgments
The author would like to thank Professor Johan Sundberg for
invaluable supervision and help with the manuscript, the singers
for their kind participation, Svante Granqvist for software and
technical support, Mattias Heldner for performing the statistical
analysis, and Peter Branderud and Hassan Djamshidpey at the
Linguistic Department at Stockholm University for profes-
sional technical support during the recording sessions. This in-
vestigation is part of Eva Björkner’s doctoral dissertation work,
which is financially partly supported by the European Commu-
nity’s Human Potential Programme under contract HPRN-CT-
2002-00276 [HOARSE-network], and an Institutional Grant
IG 2002-2049 from STINT, the Swedish Foundation for Inter-
national Cooperation in Research and Higher Education.
540
REFERENCES
1. Miles B, Hollien H. Whither belting? J Voice. 1990;4:64-70.
2. Estill J. Belting and classic voice quality: some physiological differences.
Med Probl Perform Artists. 1988;3:37-43.
3. Sundberg J, Gramming P, Lovetri J. Comparisons of pharynx, source, for-
mant, and pressure characteristics in operatic and musical theatre singing.
J Voice. 1993;7:301-310.
4. Bestebreurtje ME, Schutte HK. Resonant strategies for the belting style:
results of a single female subject. J Voice. 2000;14:194-204.
5. Stone RE Jr, Cleveland T, Sundberg J, Prokop J. Aerodynamic and acous-
tical measures of speech, operatic, and Broadway styles in a professional
female singer. J Voice. 2003;17:283-297.
6. Björkner E, Sundberg J, Cleveland T, Stone RE. Voice source differences
between registers in female musical theatre singers. J Voice. 2006;20:
187-197.
7. Fant G. Preliminaries to analysis of the human voice source. Available at:
STL-QPSR. KTH. 1982;4:1-27 http://www.speech.kth.se/qpsr/qpsr1960-1996.
html, 1982.
8. Titze I. On the relation between subglottal pressure and fundamental fre-
quency in phonation. J Acoust Soc Am. 1989;85:901-906.
9. Miller RL. Nature of the vocal cord wave. J Acoust Soc Am. 1959;3l:
667-679.
10. Björkner E, Sundberg J, Alku P. Subglottal pressure and normalized ampli-
tude quotient variation in classically trained baritone singers. Logop Pho-
niatr Vocol. 2006;31:157–165.
11. Löfqvist A, Carlborg B, Kitzing P. Initial validation of an indirect measure
of subglottal pressure during vowels. J Acoust Soc Am. 1982;72:633-635.
12. Henrich N, d’Alessandro C, Doval B, Castellengo M. Glottal open quotient
in singing: measurements and correlation with laryngeal mechanisms, vo-
cal intensity, and fundamental frequency. J Acoust Soc Am. 2005;17:
1417-1430.
13. Alku P, Vilkman E. Amplitude domain quotient for characterization of the
glottal volume velocity waveform estimated by inverse filtering. Speech
Commun. 1996;18:131-138.
View publication stats
Journal of Voice, Vol. 22, No. 5, 2008
14. Alku P, Vilkman E. A comparison of the glottal voice source quantification
parameters in breathy, normal and pressed phonation of female and male
speakers. Folia Phoniatr. 1996;48:240-254.
15. Alku P, Bäckström T, Vilkman E. Normalized amplitude quotient for pa-
rameterization of the glottal flow. J Acoust Soc Am. 2002;112:701-710.
16. Henrich N, d’Alessandro C, Doval B, Castellengo M. On the use of the de-
rivative of the electroglottographic signals for the characterization of non-
pathological phonation. J Acoust Soc Am. 2004;115:1321-1332.
17. Sundberg J, Andersson M, Hultqvist C. Effects of subglottal pressure var-
iation on professional baritone singers’ voice sources. J Acoust Soc Am.
1999;105:1965-1971.
18. Iwarsson J, Thomasson M, Sundberg J. Effects of lung volume on the glot-
tal voice source. J Voice. 1998;12:424-433.
19. Thomasson M. From air to aria. Relevance of respiratory behavior to voice
function in classical western vocal art [doctoral dissertation]. KTH, Stock-
holm; 2003.
20. Cleveland TF, Stone RE, Sundberg J, Iwarsson J. Estimated subglottal pres-
sure in six professional country singers. J Voice. 1997;11:403-409.
21. Holmberg E, Hillman R, Perkell J. Glottal airflow and transglottal air pres-
sure measurements for male and female speakers in soft, normal and loud
voice. J Acoust Soc Am. 1988;84:511-529.
22. Sundberg J, Högset C. Voice source differences between falsetto and modal
registers in counter tenors, tenors and baritones. Logop Phoniatr Vocol.
2001;26:26-36.
23. Sundberg J, Fahlstedt E, Morell A. Effects on the glottal voice source of vo-
cal loudness variation in untrained female and male subjects. J Acoust Soc
Am. 2005;117:879-885.
24. Sundberg J, Thalén M, Alku P, Vilkman E. Estimating perceived phonatory
pressedness in singing from flow glottograms. J Voice. 2004;18:56-62.
25. Hirano M, Vennard W, Ohala J. Regulation of register, pitch and intensity of
voice: an electromyographic investigation of intrinsic laryngeal muscles.
Folia Phoniatr. 1970;22:1-20.
26. Titze I. Principles of Voice Production. Englewood Cliffs, NJ: Prentice-
Hall; 1994.