ARTICLE IN PRESS

Curbing—The Metallic Mode In-between

An empirical study qualifying and categorizing restrained sounds known as
Curbing based on audio perception, laryngostroboscopic imaging, acoustics,
LTAS, and EGG

*Mathias Aaen Thuesen, †Julian McGlashan, and ‡Cathrine Sadolin, *Aarhus and ‡Copenhagen, Denmark, and
†Nottingham, UK

Summary: Objectives. This study aims to study the categorization Curbing from the pedagogical method Com-
plete Vocal Technique as a reduced metallic mode compared with the full metallic modes Overdrive and Edge by means
of audio perception, laryngostroboscopic imaging, acoustics, long-term average spectrum (LTAS), and electroglottography
(EGG).
Methods. Twenty singers were recorded singing sustained vowels in a restrained character known as Curbing. Two
studies were performed: (1) laryngostroboscopic examination using a videonasoendoscopic camera system and the
Laryngostrobe program; and (2) simultaneous recording of EGG and acoustic signals using Speech Studio. Images were
analyzed based on consensus agreement. Statistical analysis of acoustic, LTAS, and EGG parameters was undertaken
using Student paired t tests.
Results. The reduced metallic singing mode Curbing has an identifiable laryngeal gesture. Curbing has a more open
setting than Overdrive and Edge, with high visibility of the vocal folds, and the false folds giving a rectangular ap-
pearance. LTAS showed statistically significant differences between Curbing and the full metallic modes, with less energy
across all spectra, yielding a high second and a low third harmonic. Statistically significant differences were identified
on Max Qx, Average Qx, Shimmer+, Shimmer−, Shimmer dB, normalized noise energy, cepstral peak prominence,
harmonics-to-noise ratio, and mean sound pressure level (P ≤ 0.05).
Conclusion. Curbing as a voice production strategy is statistically significantly different from Overdrive and Edge,
and can be categorized based on audio perception. This study demonstrates consistently different laryngeal gestures
between Curbing and Overdrive and Edge, with high corresponding differences in LTAS, EGG and acoustic
measures.
Key Words: Complete Vocal Technique–Curbing–Audio perception–Laryngostroboscopic imaging–Long-term average
spectrum

INTRODUCTION Other studies on nonclassical styles argue that comparing non-

The two modes of singing, Overdrive and Edge, from the ter- classical singing—or rather contemporary commercial music
minology of Complete Vocal Technique (CVT), have recently (CCM)—with classical styles makes little sense, as classical mea-
been investigated using audio perception, laryngostroboscopic sures of vocal production are not viable and useful ways of
imaging, acoustics, long-term average spectrum (LTAS), and assessing nonclassical styles; hence, CCM should be consid-
electroglottography (EGG), resulting in a call for more re- ered on its own merits.7,8 However, relatively little research has
search into voice with similar methods of comparison.1 The been done on the style of CCM,7 which has been explained by
current paper on Curbing provides an example of such an anal- Zangger Borch et al as an outcome of the styles viewed histor-
ysis of other types of sounds known as Curbing, with a “reduced ically as “unhealthy” or detrimental to vocal health.9 Conversely,
metallic” characteristic, as compared with the full metallic char- recent studies into voice production of, for example, “belting”
acteristic of Overdrive and Edge.1–5 This line of comparative or “full metallic singing” argue for the health of such types of
inquiry follows recent research done into nonclassical styles of voice production, and point to the phonatory function being the
singing, which historically have been rarer than studies of clas- crux of the issue.1,6 Studying CCM styles of singing echoes the
sical singing styles in spite of the more common use of notion that research into nonclassical singing is at least as im-
nonclassical styles.6 portant as research into classical singing,10 and that a terminology
for classifying voice accordingly is called for. It has been argued
that the first step in developing a method for teaching specific
Accepted for publication January 12, 2017. sounds is to understand “the premises and conditions of the
Part of the findings from this study was presented at the 2016 Pacific Voice Confer-
ence, October 2016, in Warsaw.
desired sounds.”1
From the *Department of Business Communication, Aarhus University, Aarhus, Denmark; The present study contributes to exploring CCM styles by in-
†Department of Otorhinolaryngology, Queen’s Medical Centre Campus, Nottingham Uni-
versity Hospitals, Nottingham, UK; and the ‡Complete Vocal Institute, Copenhagen, Denmark.
vestigating sounds known as “Curbing” and compares findings
Address correspondence and reprint requests to Mathias Aaen Thuesen, Department of on Curbing with previous findings in the area done on Over-
Business Communication, Aarhus University, Jens Chr. Skous vej 4, 8000 Aarhus C,
Denmark. E-mail: [email protected]
drive and Edge to investigate such conditions and premises. The
Journal of Voice, Vol. ■■, No. ■■, pp. ■■-■■ term “Curbing” stems from the terminology of the CVT2 method,
0892-1997
© 2017 The Voice Foundation. Published by Elsevier Inc. All rights reserved.
which has been developed by author Cathrine Sadolin and
http://dx.doi.org/10.1016/j.jvoice.2017.01.010 investigated by Brixen et al3–5 and McGlashan et al1,11 as a genre
 ARTICLE IN PRESS
2 Journal of Voice, Vol. ■■, No. ■■, 2017

and style-free approach. In the CVT method, “Curbing” relates pants’ CVT-related training ranged from 1 to 15 years, with a
to the audio perception of sounds that have a reduced metallic median of 4.6 years of CVT-specific training. Singers had various
“slightly plaintive or restrained sound” compared with Over- nationalities: 13 Danish singers, 4 Norwegian singers, 1 Ice-
drive, which is a “direct sound with a more shout-like character,” landic singer, and 2 Dutch singers. The singers self-identified
and Edge, which has a “light and somewhat aggressive sound their various styles of singing, including rhythmical (1 singer),
with a more screamy character.”4 pop (9 singers), rock (3 singers), classical (2 singers), jazz (1
A recent study of two other CVT categorizations, namely singer), soul (2 singers), experimental (1 singer), and gospel (1
“Overdrive” and “Edge,” showed statistically significant differ- singer). Data for the study were obtained from all 20 partici-
ences between the two categories of sounds,1 and argued for a pants; however, data from four participants were excluded from
multidimensional approach to studying voice based on audio per- the study of EGG and laryngostroboscopic imaging as the par-
ception, laryngostroboscopic imaging, acoustics, LTAS, and EGG. ticipants either could not tolerate the examination (1 singer) or
The findings on Overdrive and Edge indicated a dominant second the data quality was not adequate for interpretation or could not
harmonic, as well as a progressive constriction of the supra- be obtained (3 singers).
glottic structures, leading to the definition of the two modes of
singing as “metallic.” Particularly, the following intentional con- Instrumentation
strictions in the larynx were identified: ventricular folds covering The study was conducted in two stages. The first stage was con-
visual parts of the vocal folds, anterior-posterior narrowing of cerned with obtaining the laryngostroboscopic images, which
the arytenoid-cuneiform complexes and petiole, an anteromedial were recorded with a videonasoendoscopic camera system
rolling in of the cuneiforms, an acute angle between the ary- (ENFV2 videoscope [KeyMed, Southend-on-Sea, UK] and
epiglottic folds and the epiglottis, narrowing of the piriform OTVS7 camera [Olympus, Southend-on-Sea, UK]). A
sinuses, and a heightened larynx position.1 These findings are Laryngostrobe system (Laryngograph Ltd, Wallington, UK) cap-
similar to findings from previous studies of “belting”12,13 as well tured the images while simultaneously acquiring the acoustic and
as the findings of Hanayama et al14 on the metallic voice, where EGG signals. A head-mounted omnidirectional microphone re-
pharyngeal wall narrowing, laryngeal rise, aryepiglottic, and lateral corded the acoustic signal, and two electrodes placed on both
laryngeal constrictions were found. sides of the singers’ larynges recorded the EGG signal. The La-
Using an identical approach to the McGlashan et al paper, the ryngograph microprocessor synchronized and processed the two
present study analyses and defines the mode of “Curbing” and signals.
compares the findings on “Curbing” with the previous findings The second stage of the study was concerned with obtaining
on Overdrive and Edge to further test the proposed method and an acoustic signal and an EGG signal with Speech Studio (vers.
its validity for developing a voice classification system, as well 3.56; Laryngograph Ltd). A Sony (Surrey, UK) ECM 989 mi-
as exploring the voice classification “Curbing” itself. The purpose crophone was used to record the audio signal, and it was kept
of the study is, therefore, to investigate the categorizations of at a constant distance of 25 cm from the singer. Audio and EGG
voice offered by CVT by means of laryngostroboscopic imaging, were captured simultaneously. The Laryngograph microproces-
acoustics, LTAS, EGG, and audio perception. Consequently, this sor synchronized and processed the signals for analysis in the
study tests and evaluates the following hypotheses: Speech Studio software.
H1: There is no visual difference between the laryngeal ges-
Vocal task
tures of “Curbing” and the previous findings on Overdrive and
All participants were instructed to sustain15 the vowel [I] on C4
Edge
for male participants (262 Hz) and on B4 for female partici-
H2: There is no significant difference in the LTAS profile of pants (494 Hz). The vowel was sustained for approximately 2–3
“Curbing” compared with the previous findings on Overdrive seconds.
and Edge
H3: There is no significant difference in the multidimensional Laryngeal gestures
acoustic properties of “Curbing” compared with the previous find- The study used a separation of the vocal tract according to
ings on Overdrive and Edge McGlashan et al’s division into six levels,1 as visible in Fig. 1,
and focused on levels 1–4. The six levels relate to (1) the vocal
H4: There is no significant difference in the multidimensional fold level; (2) the ventricular folds; (3) the arytenoid cartilages
EGG-derived measurements of “Curbing” compared with the pre- or cuneiform, epiglottis, and aryepiglottic folds; (4) piriform fossa
vious findings on Overdrive and Edge. and posterior pharyngeal wall of the hypopharynx; (5) soft palate,
uvula, oropharynx, and the back of the tongue; and (6) the oral
METHODS and nasal cavities. Still images from the laryngostroboscopic foot-
Participants ages was chosen based on blind checks by one of the authors
The study analyzed data from 20 professional singers (10 men (CS) by listening for “Curbing” sounds without viewing the
and 10 women), all with training in the categorization Curbing. imaging material or the electroglottograph signals. Having chosen
The median age of the singers was 34, ranging from 20 to 52 a representative sound example for the mode of “Curbing,” we
years. The singers’ singing experience ranged from 2 to 26 years, used the corresponding still image from the footage for further
with a median of 13.5 years of singing experience. The partici- analysis based on visual clarity. The further analysis of laryn-
 ARTICLE IN PRESS
Mathias Aaen Thuesen, et al Qualifying and Categorizing Curbing 3

FIGURE 1. The six levels of the vocal tract. Reprinted from McGlashan et al.1

geal gestures was based on a detailed evaluation of the similarities EGG provided a resonance-free measure,1 which precluded data
and differences of Curbing as compared with the findings on impact from vocal tract resonance (area function17), yielding a
Overdrive and Edge in a previous study.1 The analyses and com- more accurate description of the laryngeal activity.
parisons were done by two of the authors (MAT and CS) until To further quantify acoustic differences between Curbing and
a consensus agreement was obtained on each gesture for each Overdrive and Edge, the study included measures of Shimmer+,
still image. The still images were examined for men and women, Shimmer−, Shimmer dB, harmonics-to-noise ratio (HNR), nor-
respectively, and each identified laryngeal gesture was rated on malized noise energy (NNE), cepstral peak prominence (CPP),
a scale of degree ranging from 0 to 4; 0 (not at all), 1 (a little), and mean sound pressure level (SPL). These measures com-
2 (more), 3 (a lot), or 4 (almost fully). In total, 18 stills for Curbing bined with the EGG made up a multidimensional voice profile.1,18
were examined and compared with the findings on 18 stills of Statistical significance was tested using the paired Student t test
Overdrive and 18 stills of Edge. for men, women, and men and women combined.

EGG and acoustic recordings Long-term average spectrum

EGG was used as a noninvasive method for obtaining data of LTAS was used as a means for obtaining an “objective voice
the vocal fold activity,16 usable for distinguishing between the classification”19,20 of the measured modes based on acoustic fre-
modes of singing.1 To obtain the EGG and acoustic signals, the quency responses, and included both narrowband and broadband
Laryngograph and electroglottograph were used. A head- analyses of the obtained sustained vowel. The acoustic signal
mounted, pressure-sensitive, omnidirectional electret microphone was used in Speech Studio to generate the spectra, ranging from
with a ±2 dB, 100 Hz–10 kHz, noise level of 26 dB (Sound Pres- 0 Hz to 8000 Hz. Narrowband was used to analyze harmonic
sure Level, A Weighting [SPLA]), and a dynamic range of 88 dB structures, whereas broadband was used to analyze the for-
was used. The EGG consisted of two gold-plated electrodes placed mants. Both broadband and narrowband data were normalized
on the surface of the skin over the thyroid cartilage of the par- for loudness for each sample21 to allow for better comparisons
ticipant. Both signals were acquired simultaneously by the according to a correction factor, similarly to previous studies per-
Laryngograph microprocessor, recorded by Speech Studio onto formed with LTAS significance testing.1 A correction factor of
a laptop. the difference in dB between Overdrive and Edge and Over-
drive and Curbing at the fundamental frequency of the sample
EGG and acoustic measures note (494 Hz for women, and 262 for men) was used to normalise
The EGG was used to obtain and enable analysis of the vocal for loudness. The normalization was done at the closest inte-
fold activity during phonation, enabling the comparison of dif- gral value (500 Hz for women and 250 Hz for men). The Curbing
ferences and similarities between the vocal modes tested. Based data were compared with the Overdrive and Edge data from a
on the EGG-derived measures, a multidimensional voice profile previous study,1 and test of significance was done by Student t
was generated, which included Qx, QxSD, Fx, FxSD, jitter first, tests, for men, women, and men and women combined respec-
jitter second, jitter factor, and relative average perturbation. The tively between Curbing and the previous study findings.
 ARTICLE IN PRESS
4 Journal of Voice, Vol. ■■, No. ■■, 2017

TABLE 1.
Summary of Laryngeal Gestures and a Quantified Comparison of Curbing, Overdrive, and Edge
Maximum Potential Total Score Total Score Total Score
Level Laryngeal Gesture Score (4*18) for Overdrive for Edge for Curbing
2 How much of the vocal folds are covered by the 72 29 52 13
false folds?
Vocal folds seen along their length 72 33 15 54
False folds are active, giving a rectangular 72 31 9 60
appearance
3 Anterior-posterior narrowing between 72 37 61 25
arytenoid or cuneiform complex and petiole
Cuneiforms rolled in anteromedially 72 37 65 19
Laryngeal inlet is triangular shaped 72 34 62 15
Acute angle between the aryepiglottic folds and 72 43 64 30
the epiglottis
4 Narrowing of the piriform sinuses 72 40 61 23
Higher larynx 72 38 66 51
Bold indicates the new findings on Curbing as compared to the Overdrive and Edge findings from the previous study of Overdrive and Edge as refiners of
belting.

RESULTS seemed to be very little narrowing of the piriform sinuses (23

Laryngeal gestures out of 72 as highest possible score); however, the larynx seemed
The following observations of the laryngeal movements in levels to be elevated (51 out of 72 as highest possible score for higher
1–4 were made in the study (Table 1). larynx). The combination of seeing the vocal folds at their entire
For Curbing (see Fig. 2), there seemed to be identifiable ac- length, a rectangular-shaped false fold, and a higher larynx seems
tivity at levels 2–4. In level 2, the false folds were not to be characteristic of the laryngeal setting of Curbing.
predominantly covering the vocal folds, yielding free visual access
to the vocal folds—visible along their length (54 out of 72 as Long-term average spectrum
highest possible score) and the activity. Interestingly, the false The LTAS analysis yielded some interesting findings for the mode
folds were active in terms of giving a rectangular appearance Curbing, with significant differences visible from Overdrive and
(60 of out 72 as highest possible score). On level 3, there was Edge on both narrowband and broadband measures. On narrow-
some anterior and posterior narrowing taking place between the band measure, it seems that there is a significant amount of energy
arytenoid and cuneiform complexes and the petiole (25 out of in the area between 438 and 563 Hz for women and between 219
72 as highest possible score), but there seemed to be little rolling and 375 Hz for men. There seems to be less harmonic structure
in of the cuneiforms anteromedially (19 out of 72 as highest pos- in the spectra from around 5100 Hz for women and from around
sible score) and little triangular-shaped appearance of the laryngeal 3 kHz for men. Generally, Curbing seems to have less energy in
inlet (15 out of 72 as highest possible score). A somewhat acute the spectrum than in both Overdrive and Edge, which seems to
angle between the aryepiglottic folds and the epiglottis was ob- be very prominent for men, particularly before loudness correc-
served (30 out of 72 as highest possible score). On level 4, there tion, but also significantly after normalization. This is particularly

FIGURE 2. Examples of videonasoendoscopic images of Curbing showing the identified laryngeal gestures.
 ARTICLE IN PRESS
Mathias Aaen Thuesen, et al Qualifying and Categorizing Curbing 5

FIGURE 3. Female narrowband LTAS—Curbing compared with Overdrive and Edge.

observable after the second harmonic and onward. The male normalized Curbing samples differed on 251 of those com-
Curbing samples seem to indicate relatively less energy through- pared with Overdrive and on 242 compared with Edge, with P
out the entire spectra after 406 Hz, whereas the female Curbing values ranging from 0.05 to 0.00. Of these P values, 217 dif-
samples seem to indicate relatively less energy from around 781 Hz fered with P = 0.00 comparing Curbing with Overdrive, and 231
and throughout the rest of the spectra. Whereas Overdrive and Edge comparing Curbing with Edge. This is also the case for gender-
both seem to show a relatively clear harmonic structure in the area specific comparisons, where female Curbing differs on 213
from 5 kHz and upward for men and 8 kHz for women, Curbing measures compared with Overdrive, and 237 compared with Edge,
does not seem to have any clear harmonic structure after 5 kHz and male Curbing differs on 231 measures compared with Over-
for women and from around 3 kHz for men. Both Overdrive and drive, and 244 compared with Edge (summary of comparisons
Edge seem to have a peak around the second harmonic and a low is presented in Table 2). The very few measures that were not
fundamental, whereas Curbing seems to show a peak at around statistically significantly different are commented on below for
the fundamental and the lower second harmonic (see Figs 3 and mean measures of Curbing, and for men and women respectively.
4 for men and women respectively). Comparing mean and normalized measures of Curbing to mean
Comparing the Curbing data with the findings on Overdrive and normalized measures of Overdrive and Edge, we found that
and Edge, we found that there seems to be a high statistically there seems to be a little statistical difference in the area between
significant difference between Curbing and both Overdrive and 250 and 375 Hz when compared with Overdrive (P values ranging
Edge throughout the LTAS narrowband spectra. In total, 256 P from 0.12 to 1.00) and in the area between 250 and 469 Hz when
values were calculated based on the narrowband data, and the compared with Edge (P values ranging from 0.06 to 0.95). In the

FIGURE 4. Male narrowband LTAS—Curbing compared with Overdrive and Edge.

 ARTICLE IN PRESS
6 Journal of Voice, Vol. ■■, No. ■■, 2017

TABLE 2. TABLE 3.
Summary of P Values From Two-tailed t Test on LTAS Nar- Summary of P Values From Two-tailed t Test on LTAS
rowband for Curbing Broadband for Curbing
N P = 0.00 P = 0.05 or below N P = 0.00 P = 0.05 or below
Female C/O 256 96 213 Female C/O 128 115 115
Female C/E 256 124 237 Female C/E 128 125 125
Male C/O 256 113 231 Male C/O 128 47 47
Male C/E 256 137 244 Male C/E 128 126 126
Male + Female C/O 256 217 251 C/O (mean) 128 118 118
Male + Female C/E 256 231 242 C/E (mean) 128 126 126

area from 688 to 8000 Hz, all measures vary significantly between pared with Edge, and male Curbing differs on 47 measures
Curbing and Edge and between Curbing and Overdrive, with P (P = 0.00) compared with Overdrive, and on 126 measures
values of 0.00 throughout the comparisons. All differences between (P = 0.00) compared with Edge. The few measures that were not
Curbing and Overdrive and Edge are to be found in the first and statistically significantly different are commented on below for
second harmonic structures on mean measures. mean measures of Curbing, and for men and women respectively.
Comparing the female Curbing and female Overdrive samples Comparing broadband findings on female Curbing with female
specifically, we found that there seems to be no statistically sig- Overdrive specifically, we found that there seems to be no sta-
nificant difference between the two in the area from 250 to tistically significant differences in 13 of the 128 measures. There
344 Hz, from 406 to 719 Hz, from 2469 to 2594 Hz, and from seems to be no or weak statistically significant differences between
2969 Hz to 3031 Hz (P values ranging from 0.06 to 0.63). Sim- Curbing and Overdrive for women at 63–188 Hz (with P values
ilarly, comparing the female Curbing and female Edge samples 0.55, 0.79, and 0,7) and from 2313 to 2813 Hz (P values ranging
specifically, we found that there seems to be no statistically sig- from 0.08 to 0.29).
nificant difference between the two in the area from 250 to 375 Hz Similarly, for female Curbing compared with female Edge,
(P values ranging from 0.30 to 0.95) or in the area between 469 there are only three measures that differ, which are found at 63 Hz
and 563 Hz (P values ranging from 0.10 to 0.15). Further- (P = 0.11), 563 Hz (P = 0.35), and 625 Hz (P = 0.15). Inspect-
more, the measures at 688 Hz, 7719 Hz, and 6719 Hz show P ing the visual representation of the female broadband (see Fig. 5),
values approximating 0.10, indicating a weak statistically sig- we found that Curbing seems to follow a similar formant struc-
nificant difference at these measures. The statistically significant ture as Overdrive from around 3 kHz, with a clearly reduced
findings on 213 out of 256 measures seem to be accounted for amount of energy across all spectra.
as well after the second harmonic for women. Comparing the male samples of Curbing with male samples
Comparison of the male Curbing and male Overdrive samples of Overdrive yields the least amount of statistically signifi-
specifically shows similar results with no statistically signifi- cantly different measures, with only 47 measures differing
cant difference from 31 Hz to 375 Hz (P values ranging 0.07– between male Curbing and male Overdrive. The male samples
0.77), and weak statistically significant differences at 4594 Hz seem to show high statistically significantly different values at
(P = 0.08), 4625 Hz (P = 0.06), 4781–4906 Hz (P values ranging 63–2188 Hz (P = 0.00), contrary to the findings in female samples.
from 0.06 to 0.15), 5063 Hz (P = 0.06), and 6625–6656 Hz However, there seems to be little or weak statistically signifi-
(P = 0.10 and 0.13). Similarly, comparing the male Curbing and cant differences from 2250 to 4125 Hz (P values ranging from
male Edge samples specifically, we found that there seems to 0.06 to 0.40), from 4250 to 5250 Hz (P values ranging from 0.06
be little statistical difference in the area from 31 Hz to 375 Hz to 0.35), and from 5500 to 7250 Hz (P values ranging from 0.06
(P values ranging from 0.06 to 0.67), with all spectra above to 0.24). Contrastingly, this does not apply to male Curbing com-
varying with statistically significant differences (P = 0.00). pared with male Edge, where only two measures are not
Comparing Curbing on broadband measures seems to indi- statistically significantly different. These are at 563 Hz and
cate an even further statistically significant difference between 625 Hz, with P values of 0.50 and 0.15, respectively. Inspect-
Curbing and Overdrive, and between Curbing and Edge. On the ing the visual representation of the male broadband (see Fig. 6),
broadband LTAS, 128 P values were calculated, of which Curbing we found that male Curbing, similarly to the female sample, also
differed on 118 from Overdrive and 126 from Edge, with P values seems to follow a formant structure similar to Overdrive, with
ranging from 0.05 to 0.00, of which Curbing differed from Over- reduced energy in all the spectra. However, both the female and
drive on 118 measures with P = 0.00 and from Edge on 126 the male samples seem to indicate a large drop in energy in the
measures, with P = 0.00. Curbing seems to have a significantly higher frequencies from around 7500 Hz, compared with both
different first formant as compared with the coinciding first Overdrive and Edge (Table 4).
formant of Overdrive and Edge.
The statistically significant differences pointed out in the mean EGG and acoustic measures
summary in Table 3 are also visible in gender-specific compari- An overview of the means and standard deviations used for sta-
sons, where female Curbing differs on 115 measures (P = 0.00) tistical evaluation of EGG and acoustics can be seen in Table 5,
compared with Overdrive, and on 125 measures (P = 0.00) com- with the reported significances visible in Table 5. Because of
 ARTICLE IN PRESS
Mathias Aaen Thuesen, et al Qualifying and Categorizing Curbing 7

FIGURE 5. Female broadband LTAS—Curbing compared with Overdrive and Edge.

the relatively small sample size (n = 16), P values less than 0.10 (P = 0.00), NNE (P = 0.00), CPP (P = 0.00), HNR (P = 0.00),
are partially highlighted, indicating findings that may be inter- and mean SPL (P = 0.00). This does vary slightly when com-
esting regardless of the slightly weaker statistically significant paring the female Curbing with the female Overdrive, where the
differences. Max Qx and NNE measures do not seem to have as high a sta-
Comparing Curbing with Overdrive, we found that there seems tistically significant difference (P = 0.10 for both measures).
to be statistically significant differences in terms of numbers of Comparing the male Curbing with the male Overdrive, we found
periods (P = 0.02), Max Qx (P = 0.06), Average Qx (P = 0.05), that there seems to be a lower statistically significant differ-
Shimmer+ (P = 0.00), Shimmer− (P = 0.00), Shimmer dB ence on the Max Qx measure (P = 0.05). In the mean comparison

FIGURE 6. Male broadband LTAS—Curbing compared with Overdrive and Edge.

 ARTICLE IN PRESS
8 Journal of Voice, Vol. ■■, No. ■■, 2017

of Curbing and Overdrive, Curbing seems to differ with statis-

TABLE 4.
Paired Samples Measures Across EGG and Acoustics
tical significance on a total of 11 out of 21 measures.
Comparing Curbing with Edge, we found that there seems to
Paired Measure Mean SD be statistically significant differences in terms of number of
Group 1 C no. of periods 1078,80 852,67 periods explained by duration of the sung note, duration
OD no. of periods 859,84 595,51 (P = 0.02), Max Qx (P = 0.01), CPP (P = 0.00), and mean SPL
E no. periods 738,74 519,79 (P = 0.00). This varies slightly when comparing gender-
Group 2 C mean SPL (dB) 88,47 6,64 specific findings. Comparing female Curbing with female Edge,
OD mean SPL (dB) 101,94 5,18
E mean SPL (dB) 99,02 3,74
we found that there seems to be no differences from the mean
Group 3 C average Fx (Hz) 378,81 113,28 comparison. However, when the male Curbing was compared
OD average Fx (Hz) 370,73 114,46 with the male Edge, the number of periods, as well as the CPP,
E average Fx (Hz) 391,89 124,76 does not vary with statistical significance (P = 0.32 and 0.13 re-
Group 4 C SD Fx (Hz) 3,97 10,72 spectively). Out of the 21 measures, the mean Curbing differs
OD SD Fx (Hz) 1,72 2,26 from the mean Edge on a total of 5 out of 21 measures.
E SD Fx (Hz) 4,71 8,94 When comparing findings across the P values, it seems that
Group 5 C jitter first (%) 1,37 2,18
OD jitter first (%) 0,89 1,06
the main differences of Curbing compared with Overdrive and
E jitter first (%) 4,95 16,30 Edge can be found on measures of shimmer, NNE, CPP, HNR,
Group 6 C jitter second (%) 1,12 2,55 and mean SPL, as well as some Qx measures, where P values
OD jitter second (%) 0,53 0,63 approximate 0.00.
E jitter second (%) 3,27 10,90
Group 7 C jitter factor (%) 1,77 4,02 DISCUSSION
OD jitter factor (%) 0,84 0,99 In a recent paper, the two modes, Overdrive and Edge, from the
E jitter factor (%) 5,01 16,45
Group 8 C RAP (%) 0,88 1,19
terminology of Complete Vocal Institute were investigated by
OD RAP (%) 0,70 0,97 McGlashan et al using and correlating between audio percep-
E RAP (%) 3,32 10,75 tion, laryngostroboscopic imaging, acoustics, LTAS, and EGG,
Group 9 C Shimmer dB (dB) 0,68 0,30 calling for more research into voice with similar methods of
OD Shimmer dB (dB) 0,39 0,21 comparison.1 The current paper on Curbing provides an example
E Shimmer dB (dB) 0,91 0,72 of a similar analysis of other types of sounds known as Curbing,
Group 10 C Shimmer+ (%) 7,53 3,30 with a “reduced metallic” characteristic, as compared with the
OD Shimmer+ (%) 4,24 2,22
E Shimmer+ (%) 10,17 7,99
full metallic characteristics of Overdrive and Edge.1–5 The find-
Group 11 C Shimmer− (%) −7,64 3,76 ings support the establishment of Curbing as its own mode of
OD Shimmer− (%) −4,28 2,37 singing—a particular larynx and vocal tract position and the ac-
E Shimmer− (%) −9,35 4,32 companying source and filter interaction,22 resulting in consistent
Group 12 C average Qx (%) 50,95 8,53 auditory perceptual properties. The findings support a rejection
OD average Qx (%) 53,10 9,44 of all the proposed hypotheses. H1 can be rejected because Curbing
E average Qx (%) 53,09 12,50 does indeed feature a specific laryngeal gesture, different from
Group 13 C SD Qx (%) 2,80 2,12
OD SD Qx (%) 2,52 3,23
Overdrive and Edge. H2 can be rejected as the LTAS reveals a
E SD Qx (%) 5,22 7,10 large number of statistically different frequency responses on
Group 14 C max Qx (%) 56,88 8,01 both LTAS narrowband and broadband. H3 and H4 are rejected
OD max Qx (%) 59,31 10,91 as Curbing differs from Overdrive and Edge, with statistical sig-
E max Qx (%) 62,53 11,51 nificance on both EGG-derived and acoustic measurements.
Group 15 C min Qx (%) 42,82 12,46 When Curbing is compared with Overdrive and Edge with nor-
OD min Qx (%) 46,43 14,32 malized LTAS, the broadband reflects that Curbing differs from
E min Qx (%) 42,24 22,96
Group 16 C NNE (dB) −21,77 6,95
Overdrive on 118 of 128 measures, and from Edge on 126 of
OD NNE (dB) −29,24 7,14 128 measures, with P values of 0.00. Similarly, the narrow-
E NNE (dB) −19,70 7,40 band analysis indicates that Curbing differs from Overdrive on
Group 17 C HNR (dB) 19,18 3,74 251 of 256 measures, and from Edge on 242 of 256 measures,
OD HNR (dB) 24,44 3,16 of which 217 for Overdrive and 231 for Edge are observed with
E HNR (dB) 17,88 4,59 P = 0.00. Although the LTAS measurements proposed here do
Group 18 C CPP 6,21 1,27 not necessarily translate directly into metallic measures, it is
OD CPP 7,75 1,25
E CPP 7,27 1,68
indeed discernible that Curbing seems to have similar struc-
tures to Overdrive and Edge in some frequency ranges, but with
Note: Average Fx, average fundamental frequency derived from EGG; Average
Qx, average contact quotient; C, Curbing; CPP, cepstral peak prominence; less energy across all the frequencies observed. This could be
E, Edge; HNR, harmonics-to-noise ratio; N, 16 for each pair; NNE, normal- related to the “metallic character” also recognized by Hanayama
ized noise energy; OD, overdrive; RAP, relative average perturbation; SD,
standard deviation; SPL, sound pressure level; SD Fx, standard deviation of et al,14 who concluded that metallic voice seems to have an in-
fundamental frequency derived from EGG; SD Qx, standard deviation of creased second formant. Similarly, Curbing was observed with
contact quotient.
a high first harmonic and a low second harmonic for both men
 ARTICLE IN PRESS
Mathias Aaen Thuesen, et al Qualifying and Categorizing Curbing 9

TABLE 5.
Comparative P Values From EGG and Acoustic Two-tailed t Test (P = 0.00, P < 0.05–0.10) With a 95% Confidence Interval
Pair 1: Female Pair 2: Female Pair 3: Male Pair 4: Male Pair 5: Mean Pair 6: Mean
Curbing Curbing Curbing Curbing Curbing Curbing
Compared Compared Compared Compared Compared Compared
With Female With Female With Male With Male With Mean With Mean
Measures Overdrive Edge Overdrive Edge Overdrive Edge
Number of periods 0,02 0,03 0,50 0,32 0,02 0,02
Mean SPL 0,00 0,00 0,00 0,00 0,00 0,00
Average Fx 0,17 0,87 0,31 0,35 0,12 0,33
SD Fx 0,28 0,19 0,28 0,66 0,36 0,65
Jitter first 0,53 0,84 0,38 0,33 0,28 0,31
Jitter second 0,42 0,81 0,35 0,33 0,28 0,31
Jitter factor 0,44 0,82 0,35 0,33 0,28 0,31
RAP 0,86 0,53 0,39 0,34 0,51 0,31
Shimmer DB 0,04 0,35 0,00 0,37 0,00 0,23
Shimmer+ 0,02 0,35 0,00 0,37 0,00 0,23
Shimmer− 0,03 0,62 0,00 0,41 0,00 0,31
Average Qx 0,14 0,13 0,24 0,79 0,05 0,25
SD Qx 0,82 0,37 0,24 0,26 0,69 0,17
Max Qx 0,10 0,05 0,38 0,08 0,06 0,01
Min Qx 0,50 0,62 0,58 0,41 0,37 0,73
NNE 0,10 0,50 0,01 0,71 0,00 0,42
HNR 0,00 0,72 0,00 0,21 0,00 0,49
CPP 0,00 0,00 0,00 0,13 0,00 0,00
Note: Average Fx, average fundamental frequency derived from EGG; Average Qx, average contact quotient; C, Curbing; CPP, cepstral peak prominence; E,
Edge; HNR, harmonics-to-noise ratio; Max Qx, maximum contact quotient; Min Qx, minimum contact quotient; N, 16 for each pair; NNE, normalized noise
energy; OD, overdrive; RAP, relative average perturbation; SD Fx, standard deviation of fundamental frequency derived from EGG; SD Qx, standard devia-
tion of contact quotient; SPL, sound pressure level.
Bold indicates P ≤ 0.05.

and women, compared with the subsequent high second har- Although Curbing displays distinctive laryngeal gestures that
monic and low fundamental of Overdrive and Edge. These seem to be prominent for the mode in particular (the rectangular
findings correlate with what Brixen et al4 found on a different appearance of the false folds and the visibility of the vocal folds
cohort of singers, namely that on an Fast Fourier Transforma- along their length), it seems that some of the prominent and dis-
tion (FFT)-based narrowband analysis of Curbing, Overdrive, tinctive features of Overdrive and Edge are not observable as
and Edge, Curbing seems to have a high first formant and a low equally prominent in Curbing. Both Overdrive and Edge are ob-
second formant, compared with Overdrive and Edge, where a served with the cuneiforms rolling in anteromedially and the
low first formant and a high second formant were noted. The laryngeal inlet appearing triangular shaped (with Overdrive re-
findings in this study correlated with Brixen et al’s findings, chal- corded as 37 and 34 on the two measures, and Edge scoring 65
lenge the current conception of metallic voice as relating solely and 62, respectively). Curbing, in comparison, scores only 19
to a high second harmonic, and pose that metallic voice might on the cuneiforms rolling in anteromedially, and scoring as low
be a combination of harmonic structures, which is the case with as 15 on the laryngeal inlet, providing a triangular-shaped inlet.
Curbing, Overdrive, and Edge. This is also the case for the activity of the false folds in covering
Conversely, metallic voice has been related to aryepiglottic the visibility of the vocal folds, where Overdrive scores 29, Edge
constriction,23 which seems to correlate well with the findings scores 52, and Curbing scores only 13. This would suggest that
on Curbing as well as with previous findings on Overdrive and although there are similar features between Curbing and Over-
Edge. Curbing scored 30 out of a possible 72 on this laryngeal drive and Edge, Curbing is indeed visually recognizable and
gesture, translating to a medium amount of aryepiglottic con- displays its own set of distinctive laryngeal gestures.
striction, compared with the higher Overdrive score of 43 and This argument is in line with the proposition by McGlashan
Edge score of 64. Moreover, Curbing is observed with a dis- et al1 that Overdrive and Edge can be understood as differing
tinctive laryngeal gesture visible in a combination of two particular on the merits of a progressive constriction of the supraglottic
activities in the larynx, namely the vocal folds being seen along laryngeal structures, with a continuous narrowing of the laryn-
their length (54 out of 72) and the false folds being active in so geal inlet moving from Overdrive to Edge. Situating Curbing
far as giving a rectangular appearance (60 out of 72). These two on this continuum of progressive narrowing of the laryngeal inlet,
features seem characteristic of Curbing compared with Over- Curbing is observed with only a medium progressively narrow-
drive, scoring 33 and 31, and Edge, scoring 15 and 9, on the ing compared with the even increased narrowing in the mode
same measures. of Overdrive, with Edge being the most narrowed of the three.
 ARTICLE IN PRESS
10 Journal of Voice, Vol. ■■, No. ■■, 2017

Similarly, these findings resemble previous findings on on audio perception, laryngostroboscopic imaging, acoustics, LTAS, and
“belting,”12,14,23,24 such as the study by Lawrence who concludes EGG. J Voice. 2016;doi:10.1016/j.jvoice.2016.09.006.
2. Sadolin C. Complete Vocal Technique. Copenhagen: Shout Publishing;
that “belting” has a constricted pharynx accompanied with closed 2010.
ventricular spaces,25 arguing for a more constricted setting needed 3. Brixen EB, Sadolin C, Kjelin H. On acoustic detection of vocal modes. 132nd
to obtain sounds such as belting. It can be argued that Curbing Convention of the Audio Engineering Society; 2012; Budapest, Hungary.
shares some characteristics with Overdrive and Edge, and sub- 4. Brixen EB, Sadolin C, Kjelin H. Acoustical characteristics of vocal modes
sequently belting, in terms of laryngeal gestures; however, it differs in singing. The 134th Convention of the Audio Engineering Society; 2013;
Rome, Italy.
significantly on EGG and acoustic measures such as Shimmer+, 5. Brixen EB, Sadolin C, Kjelin H. The importance of onset features in listeners’
Shimmer−, and Shimmer dB compared with Overdrive and Edge, perception of vocal modes in singing. Paper presented at the 137th
particularly. Moreover, investigating the EGG and acoustics com- Convention of the Audio Engineering Society 2014; Los Angeles, CA.
parison further, we found that Curbing has a significantly different 6. Zangger Borch D, Sundberg J. Some phonatory and resonatory characteristics
mean SPL from that of Overdrive and Edge. Whereas Overdrive of the rock, pop, soul, and Swedish dance band styles of singing. J Voice.
2010;25:532–537.
and Edge, along with varying forms of belting, have been re- 7. Lovetri J. Contemporary commercial music. J Voice. 2008;22:259–262.
ported with high dB SPL,13,26,27 Curbing is observed with a 8. Lovetri J. The confusion about belting: a personal observation. J N Y Singing
somewhat relatively lower dB SPL and overall spectral energy, Teach Assoc. 2012;10:4–7.
which is also visible in the long-term average spectra, even after 9. Zangger Borch D, Sundberg J, Lindestad P-Å, et al. Vocal fold vibration
normalization. Correspondingly, Curbing is observed with drops and voice source aperiodicity in “dist” tones: a study of timbral ornament
in rock singing. Logoped Phoniatr Vocol. 2004;29:147–153.
in spectral energy around 3 kHz on both narrowband and broad- 10. Kovačić G, Boersma P, Domitrovic H. Long-term average spectra in
band measures. This area is also known as the singer’s formant professional folk singing voices: a comparison of the Klapa and Dozivacki
cluster,28 where increased amounts of energy in this spectra is a styles. Inst Phon Sci. 2003;25:53–64.
known strategy for loudness and subsequent heightened dB SPL, 11. McGlashan J, Sayles M, Sadolin C, et al. Vocal effects in singing: a study
which corresponds well with the observed loudness of Curbing. of intentional distortion using laryngostroboscopy and electrolaryngography.
Paper presented at the 10th International Conference on Advance in
Conclusively, the presented findings corroborate previous sug- Quantitative Laryngology, Voice and Speech Research 2013; Cincinnati, OH.
gestions that a multidimensional assessment of voice using audio 12. Estill J. Belting and classic voice quality—some physiological differences.
perception, laryngostroboscopic imaging, EGG, acoustics, and Med Probl Perform Art. 1988;3:37–43.
LTAS is indeed relevant and yields interesting measures for com- 13. Titze I. Belting and a high larynx position. J Singing. 2007;63:557–558.
parisons of different types of voice productions, which enabled 14. Hanayama EM, Camargo ZA, Tsuji DH, et al. Metallic voice: physiological
and acoustic features. J Voice. 2009;23:62–70.
the categorization of voice based on metallic characteristics. 15. Clapham RP, van As-Brooks CJ, van Son RJ, et al. The relationship between
acoustic signal typing and perceptual evaluation of tracheoesophageal voice
CONCLUSION quality for sustained vowels. J Voice. 2015;29:517, e523-519.
Based on the current study, restrained sounds with a reduced me- 16. Bier SD, Watson CI, McCann CM. Using the perturbation of the contact
quotient of the EGG waveform to analyze age differences in adult speech.
tallic character can be categorized in terms of Curbing, differing
J Voice. 2014;28:267–273.
conceptually from Overdrive and Edge with statistically signif- 17. Sundberg J. The Science of the Singing Voice. DeKalb, IL: Northern Illinois
icant differences across all proposed measures of audio perception, University Press; 1987.
laryngostroboscopic imaging, acoustics, LTAS, and EGG. Further 18. Laryngograph. Speech Studio Users Guide. Wallington, UK: Laryngograph;
research into the metallic character is suggested to investigate 2011.
19. Johnson A, Kempster G. Classification of the classical male singing voice
the conditions under which “metallic” sounds can be pro-
using long-term average spectrum. J Voice. 2010;25:538–543.
duced, which muscles are involved in producing the Curbing 20. da Silva PT, Master S, Andreoni S, et al. Acoustic and long-term average
mode, and how to best conceptualize and measure this strategy spectrum measures to detect vocal aging in women. J Voice. 2010;25:411–
of voice production. Moreover, further studies comparing the find- 419.
ings from this study with other conceptualizations of voice should 21. Sundberg J, Thalen M. Respiratory and acoustical differences between belt
and neutral style of singing. J Voice. 2015;29:418–425.
be undertaken to test how other restrained as well as non-
22. Fant G. Acoustic Theory of Speech Production: With Calculations Based
restrained sounds with or without a reduced metallic character on X-ray Studies of Russian Articulations, Vol. 2. Berlin, Germany: Walter
can be conceptualized. Finally, it is proposed to study the bound- de Gruyter; 1971.
ary conditions and variation in muscle activity between the various 23. Yanagisawa E, Estill J, Kmucha S, et al. The contribution of aryepiglottic
conceptualizations (Curbing, Overdrive, and Edge) to further the constriction to “ringing” voice quality—a videolaryngoscopic study with
acoustic analysis. J Voice. 1989;3:342–350.
understanding of these singing strategies.
24. Yanagida S, Nishizawa N, Mizoguchi K, et al. Voice onset time for the
word-initial voiceless consonant /t/ in Japanese spasmodic dysphonia—a
Acknowledgment comparison with normal controls. J Voice. 2015;29:450–454.
The authors thank the singers for willingly undergoing exami- 25. Lawrence V. Laryngeological observations on belting. J Res Singing.
nation, as well as the Complete Vocal Institute for funding the 1979;2:26–28.
26. Miles B, Hollien H. Whither belting? J Voice. 1990;4:64–70.
research.
27. Schutte H, Miller D. Belting and pop, nonclassical approaches to the female
middle voice: some preliminary considerations. J Voice. 1993;7:142–
REFERENCES 150.
1. McGlashan J, Thuesen MA, Sadolin C. Overdrive and edge as refiners of 28. Weiss R, Brown WS Jr, Morris J. Singer’s formant in sopranos: fact or
“belting”?: an empirical study qualifying and categorizing “belting” based fiction. J Voice. 2001;15:457–468.