High-Speed Imaging and Electroglottography

Measurements of the Open Quotient in Untrained

Male Voices’ Register Transitions
*Matthias Echternach, †Sebastian Dippold, ‡Johan Sundberg, §Susan Arndt, *Mark F. Zander,
and *Bernhard Richter, *xFreiburg, yLeipzig, Germany, and zStockholm, Sweden

Summary: Vocal fold oscillation patterns in vocal register transitions are still unclarified. The vocal fold oscillations
and the open quotient were analyzed with high-speed digital imaging (HSDI) and electroglottography (EGG) in 18 male
untrained subjects singing a glissando from modal to the falsetto register. Results reveal that the open quotient changed
with register in both HSDI and EGG. The in-class correlations for different HSDI and EGG determinations of the open
quotient were high. However, we found only weak interclass correlations between both methods. In 10 subjects, irreg-
ularities of vocal fold vibration occurred during the register transition. Our results confirm previous observations that
falsetto register is associated with a higher open quotient compared with modal register. These data suggest furthermore
that irregularities typically observed in audio and electroglottographic signals during register transitions are caused by
irregularities in vocal fold vibration.
Key Words: Vocal registers–High-speed imaging–Electroglottography–Falsetto.

INTRODUCTION In their pioneer studies at the Bell Laboratories, Pressman,6

In 1840, Manuel Garcia postulated that each vocal register is
produced by the same mechanical principle.1 Since that time
there has been an extended discussion of how many registers
there are and how they should be called. Different methods
have been used for understanding vocal registers. Thus, strobo-
scopy,2 videokymography,3 inverse filtering,4 and electroglot-
tography (EGG)5 have been used to analyze change of
laryngeal function accompanying register shifts. However, all
these techniques are associated with limitations concerning
the analysis of glottal vibration. Stroboscopy has the disadvan-
tage that the visualization of the glottal cycle is derived from
a number of consecutive images of different glottal cycles.
This prevents a detailed analysis of glottal cycles, especially ir-
regularities and in fast glissandos. Videokymography has the
advantage of high-speed resolution. On the other hand, this
method is limited to one single line along the entire glottis.
Both EGG and inverse filtering represent the glottal cycles
without laryngoscopic control. The EGG signal presents
changes of voltage that reflect vocal fold contact. Thus, leak-
ages caused by glottal chinks, commonly expected to occur dur-
ing falsetto registers, cannot be ascertained with this technique.
Also, the waveform of the EGG signal is not yet understood in
detail. Inverse filtering is a tedious procedure and has several
limitations. For example, it is difficult to use for high funda-
mental frequencies.
Accepted for publication May 20, 2009.
This paper has not been published in this or in substantially similar form.
No financial or other conflict exists.
From the *Department of Musicians’ Medicine, Freiburg University Medical Center,
Freiburg, Germany; yDepartment of Oto-Rhino-Laryngology, Head and Neck Surgery, Leip-
zig University Medical Center, Leipzig, Germany; zDepartment of Speech, Music and Hear-
ing, KTH, Stockholm, Sweden; and the xDepartment of Oto-Rhino-Laryngology, Head and
Neck Surgery, Freiburg University Medical Center, Freiburg, Germany.
Address correspondence and reprint requests to Matthias Echternach, MD, Institute of
Musicians’ Medicine, Freiburg University Medical Center, Breisacherstr 60, 79106
Freiburg, Germany. E-mail: matthias.echternach@uniklinik-freiburg.de
Journal of Voice, Vol. 24, No. 6, pp. 644-650
0892-1997/$36.00
Ó 2010 The Voice Foundation
doi:10.1016/j.jvoice.2009.05.003
Timcke et al,7 and Rubin and Hirt8 all examined vocal registers
by means of high-speed imaging. All of these investigations
confirmed the previous stroboscopic observation that vocal
fold oscillation changes with register. Timcke et al7 found
a higher open quotient in falsetto register. Running the high
speed at a rate of about 6000–8000 frames per second Rubin
and Hirt8 observed three different types of vocal fold vibration
characterized, at maximum adduction, by (1) a persisting glottal
chink, (2) a partial glottal chink, or (3) a complete glottal
closure, the last alternative being rarely observed among their
trained singer subjects. Also, they noted irregular vibrations
at the moment of register transitions.
Recently high-speed digital imaging (HSDI) color cameras
are commercially available for endoscopic laryngoscopic
evaluation of vocal fold oscillations. To our knowledge, no
systematic examination of changes from modal to falsetto reg-
ister in male subjects have been performed with this technique.
The aim of this study was to evaluate how the open quotient
changes between the modal and falsetto registers and during the
register transition in untrained voices. Also, we wanted to find
out whether both techniques, HSDI and EGG, show comparable
results concerning the detection of the open quotient in these
registers and in the transitions.
MATERIAL AND METHODS
Eighteen male subjects (21–38 years), all with untrained voice,
were examined. At the time of the experiment, none of them
was suffering from voice problems. Laryngeal pathologies
were excluded laryngoscopically. The subjects were asked to
sing on the vowel /i/ a glissando starting in modal register, con-
tinuing across the register transition and ending in falsetto. The
examiner demonstrated the experiment starting at around A2
(110 Hz) and ending at around A4 (440 Hz). After each
sequence, both the examiner and the subject confirmed that
the register shift was performed as instructed. In unclear cases,
the experiment was repeated until both agreed that the sequence
included the register shift from modal to falsetto register.
Matthias Echternach, et al High-Speed Imaging and EGG in Register Transitions
FIGURE 1. Indirect laryngoscopic image of the glottis. The lines
normal to the glottal length axis show the locations of the
anterior (A), middle (M), and in posterior (P) trajectories used for
the measurement of the open quotient.
Laryngoscopic HSDI was performed using an HRES-
Endocam 5562 system (Fa. Wolf, Knittlingen, Germany),
which also includes software for digital storing of the recording.
This system allows recording of 2-second long sequences at
a rate of 4000 frames per second. The audio signal was recorded
with the standard microphone offered with the same system. Its
pickup microphone is mounted on the laryngoscope at a dis-
tance of 10 cm from the lips, approximately.
FIGURE 2. Schematic illustration of the analysis of one 2-second sequence with Henrich’s MatLab Interface. Shown is the spectrogram, the
intensity (I), the open quotient (Oq), and the frequency (F0). The register shift, indicated by the line, occurred at 1270 milliseconds.
645
To facilitate detection of the free edges of the vocal folds, the
sequences were converted into black/white mode and the
contrast was adjusted. The longitudinal axis of the glottis was
manually determined. The HRES software (Fa. Wolf, Knittlin-
gen, Germany) then automatically constructed three lines (tra-
jectories), one in the anterior part, one in the middle, and one in
the posterior part of the glottis. These three lines were normal to
this axis and in each image they extended to the contours of the
vocal folds (Figure 1). When both endpoints meet, the glottis is
considered closed. The open and closed phases were automati-
cally detected and the open quotient was computed for each of
the three trajectories. In addition, the area of the rima glottidis
was automatically calculated.
During the HSDI measurement, an EGG registration (Elec-
troglottograph E90, Fa F-J Electronics, Ellebuen, Denmark)
was made, simultaneously. The EGG open quotient was deter-
mined using two definitions. One was to apply the derivative
EGG correlation-based open quotient measurement (DECOM)
criterion.9 This criterion defines the start and endpoint of the
open phase based on peaks of the derivative EGG signal.
The other was to define the open phase as the time interval
between the instants at which the EGG waveform crossed
a reference line placed at 35% of the EGG amplitude. For
both these calculations, Henrich’s MatLab Interface was used
(http://voiceresearch.free.fr/egg/).
The moment of register transition was identified by a panel of
two experts, both singers and physicians. The experts listened to
the audio sequence and were able to watch the EGG and the audio
646
FIGURE 3. Schematic illustration of the modal, transition, and fal-
setto time windows.
signals with spectrogram on the computer screen (Figure 2). All
transitions were unanimously identified by the experts.
For the analysis of the HSDI as well as of the EGG recordings,
three time intervals were defined, as illustrated in Figure 3.
A 200-millisecond wide time window was constructed symmetri-
FIGURE 4. (A, B) Digital kymographic images and EGG signals of modal register (left), register transitions (middle), and falsetto register (right)
in a subject with and without (panels a and b, respectively) vocal fold oscillation irregularities at the register transition. All kymographic images were
obtained in the region of the middle trajectory (compare Figure 1).
Journal of Voice, Vol. 24, No. 6, 2010
cally around the register transition (t2). For the modal register, an-
other 200-millisecond window was constructed that started 400
milliseconds before the point of transition (t1). Likewise, another
200-millisecond window was constructed for the falsetto starting
100 milliseconds after the end of the transition window (t3).
In each of these windows, open quotient measures were deter-
mined both from the HSDI and from the EGG signal. In the HSDI,
the (1) anterior, (2) middle, (3) posterior glottis, and (4) the glottal
area were determined. In the EGG recording, the open quotient
was measured in accordance with the (5) 35% above baseline cri-
terion, and (6) in accordance with the DECOM criterion. In each
of the three time windows, averages were calculated of the open
quotient as defined according to each of these six alternatives.
A nonparametric analysis (SPSS 14.0, Fa. SPSS Inc., Chicago,
IL) was performed. For the pair comparisons of the three mea-
surement points the Wilcoxon test and for the correlation the Pear-
son test were used. The general level of significance was set at
a ¼ 0.05.
RESULTS
In 10 of the 18 subjects, register transitions were accompanied
by irregularities of vocal fold oscillations. Figure 4A shows an
Matthias Echternach, et al High-Speed Imaging and EGG in Register Transitions
FIGURE 5. Mean values across all subjects of the open quotient
measured from the glottal area in the modal, transition, and falsetto
time windows in the high-speed material. The bars represent ±one
SD. ns ¼ nonsignificant, *P < 0.05, **P < 0.01.
example in terms of a digital kymogram and the EGG signal.
The remaining eight subjects, by contrast, showed no clear
irregularity events during the transition, as illustrated in
Figure 4B.
The open quotient determined from glottal area differed
between modal and falsetto, whereas there was no difference be-
tween the modal and the transition phase. As shown in Figure 5,
this open quotient tended to show a higher value in the falsetto
register as compared with both the modal register and the regis-
ter transition (P ¼ 0.004 for t3 vs t1; P ¼ 0.034 for t3 vs t2).
In eight subjects, we found a persisting chink in the falsetto
register, whereas the remaining 10 subjects had a complete
closure. Interestingly, a persisting glottal chink was also found
in the modal register in four subjects. This implies that their
glottal area open quotient would equal to 1.0. However, the
open quotient can also be defined on the basis of the three tra-
jectories mentioned earlier.
FIGURE 6. Mean values across all subjects of the open quotient measured at the anterior (,), middle (C), and posterior (:) trajectory in the
modal, transition, and falsetto time windows in the high-speed material, respectively. The bars represent ±one SD. ns ¼ nonsignificant, *P < 0.05,
**P < 0.01.
647
FIGURE 7. Mean values across all subjects of the open quotient
measured from the EGG data in the modal, transition, and falsetto
time windows applying the 35% over baseline and the DECOM crite-
rion. The bars represent ±one SD. ns ¼ nonsignificant, *P < 0.05,
***P < 0.001.
At the posterior trajectory, a persistent chink was found in
only four subjects in falsetto register. In all three trajectories,
the modal register was associated with a lower open quotient
as compared with both the register transition phase and falsetto
register (Figure 6). All trajectories revealed a continuous
increase of the open quotient from modal over transition to
falsetto.
Also, the EGG open quotient tended to be lower in modal
than in falsetto register, both according to the 35% above
baseline and the DECOM criteria (Figure 7).
As has been shown above, the three types of open quotient
are by no means equal. This is illustrated in Figures 8A–D.
The in-class correlations, that is, the correlations between the
open quotient data derived from the HDSI and from the EGG,
respectively, were substantially higher than the interclass
648 Journal of Voice, Vol. 24, No. 6, 2010

FIGURE 8. (A–D) Mean open quotient in the modal, transition, and falsetto time windows for the subjects derived from the EGG DECOM
material and from the (A) high-speed glottal area, (B) the anterior, (C) the middle, and (D) the posterior trajectories. The lines and the equations
show the best linear approximations of the data. R2 represents the coefficient of determination.

correlations (Table 1). The interclass correlations are very weak
and the slope of the trendline deviating substantially from 1.0.
DISCUSSION
This prospective study of male untrained voices compares the
open quotient, derived from high-speed imaging and from
EGG, for modal and falsetto register as well as in the transition
between these registers.
In their pioneering studies of vocal registers, Pressman,6
Timcke et al,7 Rubin and Hirt8 applied high-speed imaging.
They found that falsetto register was associated with a shorter
closed phase than modal register. Furthermore, Rubin and
Hirt were able to demonstrate different types of closure in their
trained subjects. They referred to these closures as ‘‘open chink,
closed chink, and stop closure.’’ However, at that time the
high-speed imaging technique was limited to mirror laryngos-
copy and obviously no digital analysis was possible.
In 2003, Hoppe et al published a high-speed examination of
a single subject performing an ascending glissando.10 They
observed a shortened closed phase and sound pressure level
drop at a fundamental frequency of 150 Hz. They interpreted
this as signs of a transition from modal to falsetto. However,
since that register transition typically occurs at a considerably
higher fundamental frequency5,11–13 the observed changes
might have had other causes.
Timcke et al7 found that the open quotient was greater in
falsetto register than in modal. The same observations have
been made in several other investigations, using EGG or inverse
filtering.4,5,14 In general, our measurements of glottal area
agreed with these observations.
In his videokymographic investigation3 of register transition,
Svec observed a persistent chink in falsetto. Our analysis of the
glottal area agreed with this observation only in about half of
our untrained subjects. Videokymography, however, observes
vocal fold oscillations on a singular line, thus reducing the
comparability with data on glottal area. Our software detected
Matthias Echternach, et al High-Speed Imaging and EGG in Register Transitions 649

TABLE 1.
Pearson Correlation (R2 values) for the Different Measurements of the Open Quotient
Anterior Middle Posterior Area EGG35 DECOM
Anterior 1.00 0.78 0.35 0.44 0.01 0.08
Middle 0.78 1.00 0.60 0.55 0.00 0.09
Posterior 0.35 0.60 1.00 0.52 0.02 0.16
Area 0.44 0.55 0.52 1.00 0.03 0.09
EGG35 0.01 0.00 0.02 0.03 1.00 0.44
DECOM 0.08 0.09 0.16 0.09 0.44 1.00

a persistent chink as soon as a chink occurred anywhere along CONCLUSIONS

the longitudinal axis. Nevertheless, this software reported a per-
sistent chink only in a small number of sequences concerning
the measurement at the three trajectories. Summarizing, as
pointed out already by Rubin and Hirt in their study of trained
subjects,8 falsetto register would be frequently but not necessar-
ily associated with a persistent chink.
The EGG is a standard method for evaluation of vocal
registers.5 Regardless of the criterion used for determining
the open quotient, DECOM or 35% criteria, our electroglotto-
graphic data showed an increased open quotient during falsetto
register. However, open quotients derived from HSDI and from
EGG show a considerable scatter. It should be recalled, that the
measurement of the open quotient in both methods is very dif-
ferent. HSDI detects the open quotient based on changes of the
superficial optical glottal area, thus neglecting the vertical
closure. EGG, on the other side, measures in-depth electrical
conductance and therefore reflects both vertical and horizontal
contact. However, this method is not able to differentiate be-
tween total or partial glottal closure.
It should be kept in mind that high-speed imaging has
limitations as well. First, in contrast to the EGG, most laryngo-
scopic examinations are limited to vowels produced with
a fronted tongue and an elevated epiglottis tip. In the present
study, the vowel /i/ was used, having a first formant in the
vicinity of 300 Hz. At the same time, fundamental frequency
was increased up to 440 Hz. As a consequence, the first formant
may have affected the vocal fold vibrations.15–18
Secondly, the examinations were performed with rigid laryn-
goscopy, which is likely to influence the conditions of the vocal
fold vibrations. Unfortunately, commercially available light sour-
ces do not allow a detailed analysis of high-speed images col-
lected by flexible laryngoscopy at a frame rate of 4000 images
per second.
Apart from the differences in the open quotient between the
modal and falsetto registers, our data also reveal that strong irreg-
ularities occurred during register transition in nearly half of the
subjects (10 out of 18). Similar observations on trained singers
were made in an earlier high-speed imaging investigation.8 Ac-
cording to Titze, different modes of vibration might occur during
register transitions.19 Furthermore, changes of the vibratory
mechanism might cause different contributions to the pitch
jumps that are often observed in register transitions.20 Our data
suggest that irregularities typically observed in audio and electro-
glottographic signals during register transitions are mainly
caused by irregularities in vocal fold vibration.
The now commercially available HSDI equipment is a useful
tool for gaining information about vocal fold vibration for fun-
damental frequencies up to about 440 Hz.
Our findings basically confirm observations earlier made by
Rubin and Hirt8 and others that the open quotient is greater in
falsetto than in modal and that vocal fold vibration irregularities
occur during the register transition. Despite good in-class corre-
lations, we also found systematic but only weak correlations of
the open quotient measured from HSDI with that measured
from EGG, regardless if the DECOM or the EGG 35% criterion
was applied. These divergencies imply that the open quotient
may differ depending on whether it is based on EGG or on glot-
tal area or transglottal flow data.
REFERENCES
1. Garcia M. Memoire sur la voix humaine presente a lA`cademie des Sciences
en 1840. 2nd ed.). Paris: Imprimerie d’E Duverger; 1847.
2. Röhrs M, Pascher W, Ocker C. Untersuchungen über des Schwingungsver-
halten der Stimmlippen in verschiedenen Registerbereichen mit unter-
schiedlichen stroboskopischen Techniken. Folia Phoniatr (Basel).
1985;37:113–118.
3. Svec JG. On Vibration Properties of Human Vocal Folds. University of
Groningen; 2000.
4. Sundberg J, Högset C. Voice source differences between falsetto and modal
registers in counter tenors tenors and baritons. Logoped Phoniatr Vocol.
2001;26:26–36.
5. Henrich N. Mirroring the voice from Garcia to the present day: some in-
sights into singing voice registers. Logoped Phoniatr Vocol. 2006;31:3–14.
6. Pressman J. Physiology of the vocal cords in phonation and respiration.
Arch Otolaryngol. 1942;35:355–398.
7. Timcke R, von Leden H, Moore P. Laryngeal vibrations: measurements of
the glottic wave. I. The normal vibratory cycle. AMA Arch Otolaryngol.
1958;68:1–19.
8. Rubin HJ, Hirt CC. The falsetto. A high speed cinematographic study.
Laryngoscope. 1960;70:1305–1324.
9. Henrich N, D’Alessandro C, Doval B, Castellengo M. Glottal open quotient
in singing: measurements and correlation with laryngeal mechanisms,
vocal intensity, and fundamental frequency. J Acoust Soc Am.
2005;117(3 Pt 1):1417–1430.
10. Hoppe U, Rosanowski F, Dollinger M, Lohscheller J, Schuster M,
Eysholdt U. Glissando: laryngeal motorics and acoustics. J Voice.
2003;17:370–376.
11. Preissler W. Stimmumfänge und Gattungen. Arch Sprach- und Stimmphy-
siologie usw. 1939;3:65–85.
12. Roubeau B, Henrich N, Castellengo M. Laryngeal vibratory mechanisms: the
notion of vocal register revisited. J Voice 2008;. Epub ahead of print Jun 5.
13. Bloothooft G, Van Wijck M, Pabon P. Relations between vocal registers in
voice breaks. Eurospeech. 2001;2001:39–42.
14. Salomão GL, Sundberg J. Relation between perceived voice register and
flow glottogram parameters in males. J Acoust Soc Am. 2008;124:546–551.
650
15. Titze IR. Nonlinear source-filter coupling in phonation: theory. J Acoust
Soc Am. 2008;123:2733–2749.
16. Titze IR, Riede T, Popolo P. Nonlinear source-filter coupling in phonation:
vocal exercises. J Acoust Soc Am. 2008;123:1902–1915.
17. Rothenberg M. Cosi’ Fan Tutte and what it means. In: Baer T, Sasaki C,
Harris KS, eds. Vocal Fold Physiology: Laryngeal Function in Phonation
and Respiration. San Diego: College Hill Press; 1986:254–263.
Journal of Voice, Vol. 24, No. 6, 2010
18. Fant G. Glottal flow: models and interaction. In: Proceedings of sympo-
sium on voice acoustics and dysphonia, Gotland. J Phon. 1986;14:
393–399.
19. Titze IR. Principles of Voice Production. NJ: Prentice Hall; 1994.
20. Svec JG, Schutte HK, Miller DG. On pitch jumps between chest and
falsetto registers in voice: data from living and excised human larynges.
J Acoust Soc Am. 1999;106:1523–1531.