Predicted Singers Vocal Fold Lengths and Voice

ClassificationA Study of X-Ray Morphological

Measures
*Friederike Roers, *,Dirk Murbe, and Johan Sundberg, *yDresden, Germany and zStockholm, Sweden

Summary: Students admitted to the solo singing education at the University of Music Dresden, Germany have been
submitted to a detailed physical examination of a variety of factors with relevance to voice function since 1959. In the
years 19591991, this scheme of examinations included X-ray profiles of the singers vocal tracts. This material of 132
X-rays of voice professionals was used to investigate different laryngeal morphological measures and their relation to
vocal fold length. Further, the study aimed to investigate if there are consistent anatomical differences between singers
of different voice classifications. The study design used was a retrospective analysis. Vocal fold length could be mea-
sured in 29 of these singer subjects directly. These data showed a strong correlation with the anterior-posterior diameter
of the subglottis and the trachea as well as with the distance from the anterior contour of the thyroid cartilage to the
anterior contour of the spine. These relations were used in an attempt to predict the 132 singers vocal fold lengths.
The results revealed a clear covariation between predicted vocal fold length and voice classification. Anterior-posterior
subglottic-tracheal diameter yielded mean vocal fold lengths of 14.9, 16.0, 16.6, 18.4, 19.5, and 20.9 mm for sopranos,
mezzo-sopranos, altos, tenors, baritones, and basses, respectively. The data support the assumption that there are con-
sistent anatomical laryngeal differences between singers of different voice classifications, which are of relevance to
pitch range and timbre of the voice.
Key Words: Vocal fold lengthSingingVoice classificationX-ray measures.

INTRODUCTION frequency in phonation, because vocal fold mass per unit of

Vocal fold length is a parameter of great significance to the
pitch range and the timbre of a voice.1,2 Several attempts
have been made in the past to measure vocal fold length. For ex-
ample, Kahane,3 Hirano et al,4 and Friedrich and Kainz5 mea-
sured vocal fold length in cadavers. However, this method
seems rather risky because of the loss of tissue tension and
the drying, and injection of formalin entails shriveling of the
tissues. Eckel and Sittel6 analyzed fresh cadavers and found
an average length of the pars membranacea of 13.2 mm
(n 12) in male adults and 10.6 mm for female adults (n 8).
Attempts have also been made to measure vocal fold length
in living subjects. Trendelenburg7 tried a stereoscopic method,
which, however Flach et al8 later criticized. Photos of the larynx
were presented by Hollien,9 and Sawashima et al,10 and Harries
et al11 used ultrasound, whereas Schuberth et al12 and Schuster
et al13 used a laser technique. Su et al14 measured vocal fold
length in 165 subjects by means of microlaryngoscopy during
anesthesia and found a significant difference between genders.
Also, X-rays have been used for measurement of vocal fold
length.1517 Their data refer to carefully defined landmarks
but were collected during phonation. As pointed out by several
authors,10,12,1820 vocal fold length varies with fundamental
Accepted for publication December 6, 2007.
A preliminary version of this investigation was presented at the PanEuropean Voice Con-
ference, Stockholm, August 2001.
From the *Voice Research Laboratory, University of Music Carl Maria von Weber,
Dresden, Germany; yDepartment of Otorhinolaryngology, Technical University of Dres-
den, Dresden, Germany; and the zDepartment of Speech, Music and Hearing, KTH, Stock-
holm, Sweden.
Address correspondence and reprint requests to Dirk Murbe, MD, MA, PhD, Department
of Otorhinolaryngology, Technical University of Dresden, Fetscherstrae 74, D-01307
Dresden, Germany. E-mail: dirk.muerbe@mailbox.tu-dresden.de
Journal of Voice, Vol. 23, No. 4, pp. 408-413
0892-1997/$36.00
2009 The Voice Foundation
doi:10.1016/j.jvoice.2007.12.003
length is the decisive factor. This factor is controlled by stretch-
ing the vocal folds, that is, by varying their length. Thus, high-
fundamental frequencies are produced with long vocal folds.
Therefore, it is often difficult to compare data on vocal fold
length assembled during phonation.
Phonation and respiration conditions are associated with dif-
ferent positioning of the various laryngeal cartilages, such that
the target vocal fold length and tension is achieved. According
to Sawashima et al10 the vocal folds are longer in the abducted
position, as in breathing than during phonation. The degree of
abduction is influenced by the intensity of the breathing.
Thus, in tidal breathing only the pars intercartilaginea of the
glottis is fully open. For breathing with larger lung volumes,
also the pars intermembranacea opens, and during quite force-
ful breathing the glottis is also maximally widened by contrac-
tion of the posterior cricoarytenoid muscle.21
Vocal fold length is particularly relevant to singers, because it
can be assumed to be one of the main factors behind voice clas-
sification. Data on the vocal fold lengths of singers with differ-
ent classifications are rare in the literature. Pfau22 presented
a large material of measurement of the length of the membra-
nous part of the vocal folds during respiration in singers of dif-
ferent classifications. He used laryngoscope equipment,
a simplified version of that used by Trendelenburg.7
Since 1959, all students admitted to the solo singing educa-
tion at the University of Music Carl Maria von Weber Dresden
were submitted to a detailed examination of a variety of factors
that could of obvious or potential relevance to voice function.23
This examination, made at the beginning of the conservatory
education, focused on different aspects of the singing voice
and included assessment by a phoniatrician, a logoped, and
a group of singing teachers. Although the examination protocol
changed slightly over this period, a great amount of data, based
Friederike Roers, et al Singers Vocal Fold Lengths 409

on more than 340 individuals, was collected regarding a variety

of aspects, such as voice pitch range, mean speaking pitch, neck
circumference, body height, and weight. In addition, between
1959 and 1991 the examination also included X-ray profiles
of the students upper airways, mostly including the entire vocal
tract. The X-ray images were taken during tidal breathing. This
unique database, comprising a great number of X-ray profiles of
singers airways, was stored at the Voice Research Laboratory
at the University of Music Carl Maria von Weber Dresden.
In the present investigation, this material was taken advan-
tage of for examining the relationship between experts classi-
fication of singer voices and laryngeal measures. The specific
question we asked was whether there were consistent anatomi-
cal differences between singers of different voice classifica-
tions.

METHOD
Materials
The Dresden X-ray material was used for the present study;
however, excluding few individuals on the following criteria:1 FIGURE 1. Sagittal X-ray image of the upper airways including
students not unanimously classified by the singing teacher pharynx, larynx, proximal trachea, cervical vertebrae, and parts of
panel,2 students the classification of whom changed during the the skull. Further, the image includes a metal-distance measurement
education, and3 students who interrupted their solo singer edu- calibrator.
cation at this institute. The material (Table 1) then consisted of
132 individuals, 71 females (40 Sopranos, 22 Mezzo-sopranos,
was always 1 m, and the distance between the tube and the pha-
nine Altos) and 61 males (19 Tenors, 23 Baritones, 19 Basses).
ryngeal midline was 25 cm. Six-valve-X-ray-equipment (D800,
The age range was 1631 years, and the mean age 18.8 and 19.7
TUR, Dresden, Germany) was used with a voltage of 4045 kV
years, respectively.
and a 4075 mA current.
The accuracy of distance measurement was about 0.5 mm.
Measurements
When needed contrast improvement was obtained using
All X-ray images, taken under identical technical conditions at
a Cool Brite Illuminator (CB2, RADX, Houston, TX). Often,
the Radiological department of the University Hospital in Dres-
the length of the membranous part of the vocal fold could not
den, showed a sagittal projection of the upper airways (Figure 1).
be accurately determined from the X-ray profiles of these young
The profiles included pharynx, larynx, proximal trachea, cervi-
subjects, because the laryngeal cartilages were not calcified
cal vertebrae, and the base and mostly also the anterior part of
enough to produce an X-ray shadow. Hence, it was often impos-
the skull. Also included was a metal-distance measurement
sible to estimate the length from the thyroid-arytenoid distance.
calibrator, inner diameter 1 3 1 cm, mounted on a stand and
However, in 29 out of the 132 cases calcified structures in the
located in front of the laryngeal midline.
arytenoid cartilages allowed estimation of vocal fold length.
The X-rays were taken with the subject in standing position,
A number of alternative measures were collected from the X-
and with closed mouth. The central ray was about 35 cm dor-
ray material. One alternative was the sagittal distance between
sal to the laryngeal prominence. The tube-photo plate distance
the anterior notch of the thyroid cartilage and the vertebral spine
contour. This distance could be measured in all subjects. A
straight line was drawn between the anterior-most contours of
TABLE 1.
Students Included in the Investigation
the atlas and the upper part of the seventh vertebra, as shown
in the left part of Figure 2. A normal was drawn to this line at
Number of X-Ray the level of the thyroid notch. The distance along this line be-
Singers Available tween the anterior notch of the thyroid cartilage and the anterior
Mean
Classification n % n % Age (yr) contour of the vertebral spine was then determined.
Another alternative was the anterior-posterior diameter of the
Soprano 99 29.1 40 30.3 18.4
subglottic and tracheal contour. This measure, however, varied
Mezzo-soprano 57 16.8 22 16.7 18.9
Alto 21 6.2 9 6.8 20.3 depending on the distance up to the glottis. Therefore, this mea-
Tenor 47 13.8 19 14.4 19.6 sure was determined at several distances, 1 cm apart, along the
Baritone 75 22.1 23 17.4 19.5 length axis of the contour of subglottis and trachea, as illus-
Bass 41 12.1 19 14.4 20.0 trated in the right part of Figure 2.
All measurements were done in the absence of information
Total 340 100 132 100 19.2
about the subjects voice classification.
410 Journal of Voice, Vol. 23, No. 4, 2009

The STAP distances for the entire group of 132 singers also
showed a systematic relationship with singer classification.
Thus, sopranos tended to show the shortest STAP distances
and the basses the greatest STAP distances as illustrated in Fig-
ure 4. The relationship between classification and the STAP dis-
tance was weaker for the lower than for the higher locations.
Thus, at 6 cm below the glottis the STAP distance differentiated
significantly only between genders.
The strongest relationship was found for the maximum STAP
FIGURE 2. Schematic illustration of TS distance (left) and different distance, as illustrated in Figure 5. Except for mezzo-soprano/
STAP distances (right).
alto all classifications differed significantly at the P < 0.05
level. These results indicate that the maximum STAP distance
varied systematically between singer voice classifications, pre-
Statistical analyses sumably because this distance is related to vocal fold length.
Statistical analyses were carried out by comparison of means by Another possible distance likely to be correlated with vocal
ANOVA function of SPSS 11.5 for Windows (SPSS, Inc., Chi- fold length is the distance from the anterior end of the vocal
cago, IL). Significant differences were identified by applying fold to the spine contour. This distance was determined by
a Student-Newman-Keuls Procedure (a posteriori testing for drawing a straight line from the anterior-most contour of the at-
multiple comparisons of averages). las and the upper anterior contour of the seventh vertebra (the
left part of Figure 2). Then, a line was drawn normal to this ref-
erence line at the level of the vocal fold. The distance was then
RESULTS
measured between the thyroid cartilage and the anterior contour
Because of the lack of calcification of laryngeal structures, vo-
of the nearest cervical vertebra, which includes the postcricoid
cal fold length could be determined in no more in 29 out of the
space. Henceforth, this distance will be referred to as the thy-
132 singers. This is clearly too small for a material to allow any
roid-spine (TS) distance. These measures showed a strong
reliable conclusions regarding the relationship between voice
correlation with vocal fold length for the 29 subjects, for
classification and vocal fold length.
whom reliable vocal fold length data were available (Table 2
It seemed likely that, in some sense, the vocal fold length
and Figure 6).
must be related to the subglottic-tracheal anterior-posterior dis-
The TS distance was measured in all 132 singer subjects. As
tance. The subglottic-tracheal contour often showed a conical
illustrated in the box plot in Figure 7, the mean TS distance dif-
shape, as illustrated in the right part of Figure 2. Therefore,
fered between the singer groups, being shortest for the sopranos
the subglottic-tracheal anterior-posterior distance, henceforth
and longest for the basses. The difference was significant
the STAP distance, was determined at six points, 1 cm apart,
(P < 0.005) between soprano and alto, between alto and tenor,
normal to the length axis of subglottis and trachea, the most cra-
and between tenor and bass.
nial point positioned 1 cm below the glottis level. Further, the
The above results suggest that it would be meaningful to pre-
maximum STAP distance was determined. The relationship be-
dict the actual vocal fold length by means of trendline equations
tween vocal fold length and the STAP distance was measured in
shown in Table 2. Thus, estimates of vocal fold length were cal-
all subjects, for whom reliable vocal fold length data and STAP
culated from the group means of maximum STAP and of TS dis-
distances were available. The result showed that the correlation
tance. The results are listed in Table 3. The differences between
was high (Table 2 and Figure 3).
predictions made on STAP and TS distance were quite small,
0.2 mm, or less. The agreement with the means of the measured
data was less, the greatest differences amounting to 1.8 and
TABLE 2. 1.7 mm for the alto and for the baritone groups, respectively.
Slope, Intercept (Icpt), R2, and Correlation (r) for the However, these means were based on no more than three indi-
Trendlines Approximating the Relationship Between the viduals, and may thus not be representative for these groups.
Indicated Parameters and Vocal Fold Length as
Measured in the 29 Singers, for Whom This was Possible
Slope Icpt
(mm) (mm) R2 r DISCUSSION
STAP distance at 1 cm 1.09 1.25 0.713 0.844 A key issue in our investigation is the classification of the singer
STAP distance at 2 cm 0.86 3.58 0.692 0.832 voices. This classification was carried out by the students sing-
STAP distance at 3 cm 0.73 6.83 0.542 0.736 ing teacher at the University of Music. One might argue that this
STAP distance at 4 cm 0.92 5.30 0.558 0.747 method is likely to contain a good deal of subjectivity. On the
STAP distance at 5 cm 0.90 5.56 0.500 0.707 other hand, all subjects who changed classification during their
STAP distance at 6 cm 0.99 3.48 0.517 0.719 studies were eliminated from the material. This implies that
Max STAP distance 0.98 0.69 0.718 0.848
several singing teachers agreed on the classification of all
TS distance 0.71 4.26 0.802 0.895
singers examined in the present investigation.
Friederike Roers, et al Singers Vocal Fold Lengths 411

FIGURE 3. Relationship between vocal fold length and the STAP distances at 1 (left panel) and 6 cm (right panel) below glottis level for all sub-
jects, for whom reliable vocal fold length data and STAP distances were available.

Considering the anatomy of the larynx the strong correlation
between measured vocal fold length and the STAP distance is
not surprising, because the vocal folds constitute the upper
end of subglottis and trachea. Somewhat more surprising is
the correlation between vocal fold length and the TS distance,
because the larynx can move anteriorly. However, this case is
not likely to occur under the conditions prevailing during the
X-ray session.
The method used for measuring vocal fold length was also
used in previous studies.8,1517 This measurement could be car-
ried out in the 29 out of 132 X-ray images (22%), because in
these images a shadow could be observed that safely could be
assumed to originate from the arytenoid cartilages.
Although we used the same landmarks as other previous
studies for measuring vocal fold length, it is difficult to compare
our results with those gained in these investigations. The reason
is that they refer to vocal fold length during phonation. Also,
they did not specify the subjects voice classifications.
Hollien and Moore18 measured the lengths of the membra-
nous parts of the vocal fold from photos of the larynges of six
male nonsinger subjects whom they classified as tenors, bari-
tones, or basses. They found that these groups had an average
vocal fold length of 19.7 mm for three tenor voices, 19.8 mm
for two baritone voices, and 23 mm in a single bass voice.
Our corresponding values were 18.5, 19.4, and 21.1 mm, re-
spectively. This agreement is somewhat surprising because
our values refer to an abducted position of the vocal folds.
Benninghoff21 reported a vocal fold length range of 1421
and 1827 mm for female and male singers, respectively. He
also found that, on average, alto singers had 3 mm longer vocal

FIGURE 4. Mean STAP distances and standard deviations for sopranos, mezzo-sopranos, altos, tenors, baritones, and basses (S, M, A, T, BR, and
B, respectively) for the different levels of STAP distance.
412
FIGURE 5. Comparison between singers of different voice classifi-
cations (sopranos, mezzo-sopranos, altos, tenors, baritones, and basses)
in terms of a box plot diagram showing the distribution of maximum
STAP distance data (median, 1st and 3rd quartiles, and data range).
folds than sopranos and basses had 5 mm longer folds than
tenors. We found smaller differences, about 1.8 mm between
sopranos and altos and 2.6 mm between tenors and basses.
However, Benninghoff did not describe his method in detail,
so it is difficult to comment upon the difference between his
and our results.
Comparison with the results presented by Pfau22 is easier,
particularly as his data concern singers during respiration,
just as ours. On the other hand, he used laryngoscope technique
rather than X-ray and did not describe his statistical methods. In
any event, he found average vocal fold lengths for 24 sopranos,
13 altos, 11 tenors, seven baritones, and nine basses of 12, 12,
13, 14, and 16 mm, respectively. These values are between 3
FIGURE 6. Relationship between vocal fold length and the TS dis-
tance for all subjects, for whom reliable vocal fold length data and TS
distances were available.
Journal of Voice, Vol. 23, No. 4, 2009
FIGURE 7. Comparison between singers of different voice classifi-
cations (sopranos, mezzo-sopranos, altos, tenors, baritones, and basses)
in terms of a box plot diagram showing the distribution of TS distance
data (median, 1st and 3rd quartiles, and data range).
and 5 mm shorter than ours. The reason would be that Pfaus
data include merely the membranous part, whereas our data
are also likely to include part of the arytenoid cartilage. There-
fore, it seems reasonable to conclude that our data are represen-
tative for the singer classification concerned.
Even though vocal fold length varies systematically with
singer voice classification, it would be difficult to use as the
sole or even as the main criterion for such classification. First,
it is difficult to measure in a reliable manner. Second, formant
frequencies, that is, vocal tract shape factors, would also be of
relevance. Third, vocal fold length varies depending on laryn-
geal conditions, particularly the pitch during phonation.
Nevertheless, our study has revealed that rather accurate es-
timates of vocal fold length can be derived from both the TS and
maximum STAP distances. The value of the former is limited
by the same dependence on pitch as the absolute vocal fold
length. The latter, by contrast, may be more useful, because it
does not vary with phonatory conditions. Future studies, based
on high-resolution magnetic resonance imaging will allow
TABLE 3.
Observed and Estimated Mean Vocal Fold Lengths (mm)
for Sopranos, Mezzo-Sopranos, Altos, Tenors, Baritones,
and Basses (S, M, A, T, BR, and B, Respectively)
Calculated From
Classification Measured n STAP Max TS
S 14.3 7 14.9 14.7
M 15.9 8 16.0 15.8
A 18.4 3 16.6 16.6
T 19.5 5 18.4 18.5
BR 21.1 3 19.5 19.4
B 21.2 3 20.9 21.1
The estimations were based on the trendlines relating the STAP max and
TS distances to the observed vocal fold lengths for the 29 singers for
whom this measure could be determined from the X-ray images.
Friederike Roers, et al Singers Vocal Fold Lengths
assessing morphological criteria in different phonatory condi-
tions. Although absolute values of vocal fold length cannot
serve as the main criterion for voice classification, they may
be useful as a contributing factor.
Vocal fold length should have acoustical consequences. For
a given subglottal pressure, a long glottis should allow a greater
amount of airflow when it opens than a shorter glottis. Large
amplitudes of the transglottal airpulses generate a stronger
voice source fundamental.24 Hence, at a given fundamental fre-
quency, a singer with longer vocal folds will generate stronger
fundamentals than a singer with shorter vocal folds. This should
contribute the timbre differences between singer voice classifi-
cations.
CONCLUSIONS
Our data show that the X-ray-morphological measures of TS
and STAP distances are correlated with vocal fold length. Fur-
ther, the results revealed a clear covariation between predicted
vocal fold length and voice classification. This supports the
assumption that there are consistent anatomical differences
between singers of different voice classifications.
Acknowledgments
The database at the Voice Research Laboratory, University of
Music Carl Maria von Weber, Dresden, Germany, is adminis-
tered by Hartmut Zabel. The investigation including analysis
of X-rays and data acquisition was carried out by coauthor
F.R. with support in terms of a Marie Curie scholarship from
the European Commission during her stay at the Department
of Speech, Music, and Hearing, KTH, Stockholm. Eberhard
Kuhlisch performed the ANOVA analysis. The manuscript
was written by coauthors D.M. and J.S.
REFERENCES
1. Titze IR. Physiologic and acoustic differences between male and female
voices. J Acoust Soc Am. 1989;85:1699-1707.
2. Sundberg J. The Science of the Singing Voice. Dekalb, IL: Northern Illinois
Univ. Press; 1987.
3. Kahane JC. A morphological study of the human prepubertal and pubertal
larynx. Am J Anat. 1978;151:11-19.
4. Hirano M, Kurita S, Nakashima T. Growth, development and aging of
human vocal folds. In: Bless DM, Abbs JH, eds. Vocal Fold Physiology.
Contemporary Research and Clinical Issues. San Diego, CA: College-Hill
Press; 1983:22-43.
413
5. Friedrich G, Kainz J. Morphometrie des Kehlkopfes an Horizontalschnit-
ten. Basisdaten fur die quantitative Auswertung moderner bildgebender
Verfahren. Laryngol Rhinol Otol. 1988;67:269-274.
6. Eckel HE, Sittel C. Morphometry of the larynx in horizontal sections. Am J
Otolaryngol. 1995;16:40-48.
7. Trendelenburg W. Ein Apparat zur Vorfuhrung und zur Ausmessung des
Kehlkopfspiegelbildes. Z. Hals-, Nas.- u. Ohrenheilk. 1929;22:159-164.
8. Flach M, Schwickardi H, Steinert R. Zur Frage des Einflusses erblicher
Faktoren auf den Stimmklang (Zwillingsuntersuchungen). Folia Phoniatr.
1968;20:369-378.
9. Hollien H. Vocal pitch variation related to changes in vocal fold length.
J Speech Hear Res. 1960;3:150-156.
10. Sawashima M, Hirose H, Honda K, Yoshioka H, Hibi SR, Kawase N,
Yamada M. Stereoendoscopic measurement of the laryngeal structure. In:
Bless DM, Abbs JH, eds. Vocal Fold Physiology. Contemporary Research
and Clinical Issues. San Diego, CA: College-Hill Press; 1983:264-276.
11. Harries M, Hawkins S, Hacking J, Hughes I. Changes in the male voice at
puberty: vocal fold length and its relationship to the fundamental frequency
of the voice. J Laryngol Otol. 1998;112:451-454.
12. Schuberth S, Hoppe U, Dollinger M, Lohscheller J, Eysholdt U. High-
precision measurement of the vocal fold length and vibratory amplitudes.
Laryngoscope. 2002;112:1043-1049.
13. Schuster M, Lohscheller J, Kummer P, Eysholdt U, Hoppe U. Laser projec-
tion in high-speed glottography for high-precision measurements of laryn-
geal dimensions and dynamics. Eur Arch Otorhinolaryngol. 2005;262:
477-481.
14. Su MC, Yeh TH, Ta CT, Lin CD, Linne OC, Lee SY. Measurement of adult
vocal fold length. J Laryngol Otol. 2002;116:447-449.
15. Sonninen A. Is the length of the vocal cords the same at all different levels
of singing? Acta Otol Laryngol Suppl. 1954;118:219-231.
16. Damste PH, Hollien H, Moore GP, Murry T. An X-ray study of vocal fold
length. Folia Phoniatr. 1968;20:349-359.
17. Hollien H, Damste PH, Murry T. Vocal fold length during vocal fry phona-
tion. Folia Phoniatr. 1969;21:257-265.
18. Hollien H, Moore GP. Measurements of the vocal folds during changes in
pitch. J Speech Hear Res. 1960;3:157-165.
19. Luchsinger R, Pfister K. Die Messung der Stimmlippenverlangerung beim
Steigern der Tonhohe. Folia Phoniatr. 1961;13:1-12.
20. Sonninen A, Hurme P, Laukkanen AM. The external frame function in the
control of pitch, register and singing mode: radiographic observations of
a female singer. J Voice. 1999;13:319-340.
21. Benninghoff A. Makroskopische Anatomie, Histologie, Embryologie,
Zellbiologie. In: Benninghoff A, Drenckhahn D, eds. Anatomie (I).
Munchen, Jena: Urban & Fischer; 2003:553.
22. Pfau W. Klassifizierung der menschlichen Stimme. In: Herrmann A,
Jakobi H, eds. Hals-, Nasen- und Ohrenheilkunde; Band 22. Leipzig:
Johann Ambrosius Barth; 1973.
23. Murbe D, Sundberg J, Iwarsson J, Pabst F, Hofmann G. Longitudinal study
of solo singer education effects on maximum SPL and level in the singers
formant range. Logoped Phoniatr Vocol. 1999;24:178-186.
24. Gauffin J, Sundberg J. Spectral correlates of glottal voice source waveform
characteristics. J Speech Hear Res. 1989;32:556-565.