A Formant Range Profile for Singers

*,†Ingo R. Titze, *Lynn M. Maxfield, and *Megan C. Walker, *Salt Lake City, Utah, and †Iowa City, Iowa

Summary: Vowel selection is important in differentiating between singing styles. The timbre of the vocal instru-
ment, which is related to its frequency spectrum, is governed by both the glottal sound source and the vowel choices
made by singers. Consequently, the ability to modify the vowel space is a measure of how successfully a singer can
maintain a desired timbre across a range of pitches. Formant range profiles were produced as a means of quantifying
this ability. Seventy-seven subjects (including trained and untrained vocalists) participated, producing vowels with three
intended mouth shapes: (1) neutral or speech-like, (2) megaphone-shaped (wide open mouth), and (3) inverted-
megaphone-shaped (widened oropharynx with moderate mouth opening). The first and second formant frequencies (F1
and F2) were estimated with fry phonation for each shape and values were plotted in F1-F2 space. By taking four vowels
of a quadrangle /i, æ, a, u/, the resulting area was quantified in kHz2 (kHz squared) as a measure of the subject’s ability
to modify their vocal tract for spectral differences.
Key Words: Singing–Formants–Formant range–Singing vowels–Belting.

INTRODUCTION and females), the first and second resonances are often lowered
Singing styles are in part differentiated by the timbre of the voice. by widening the oropharynx7–9 to balance the harmonic ener-
In some instances, the use of a particular timbre may be a simple gies of fo, 2fo and 3fo. Mixed registration at the source can also
artistic choice by the vocalist, but it can also be an acoustic re- be used along with the vocal tract modifications. Sometimes the
quirement in many cases, determined by the size of the hall (if energy of 3fo is reinforced with the second resonance.9
singing is unamplified), the instrumental accompaniment, or the
[It is noted parenthetically that in the current study, as in many
dramatic content of the song. For example, if strings and flutes
other studies involving resonances of the vocal tract, a direct
are the primary accompaniment, a “brassy” voice timbre pro- measurement of the resonance frequencies fRn is not at-
duces a contrast in timbre. On the other hand, if brass and tempted. Rather, the assumption is made that the resonance
percussion dominate the accompaniment, a “flute-like” voice frequencies closely mirror the formant frequencies Fn, mea-
timbre produces a contrast. As instrumental accompaniment, per- sured as peaks in the output spectrum at the mouth. Hence,
formance spaces, and electronic enhancement vary, vocal timbre the approximation fRn ≈ Fn is assumed in subsequent discus-
may change in response. sions and the symbolic notation Fn is used10 with the full
According to the American Standards Institute, timbre is that understanding that Fn is not exactly equal to fRn.]
attribute of auditory sensation in terms of which a listener can
Continuing with classical operatic singing, lowering F1 and
judge that two sounds similarly presented and having the same
F2 also has the effect of creating a spectral distance between the
loudness and pitch are dissimilar.1 Acoustically, timbre is thus
vowel-determining formants (F1 and F2) and those formants often
largely related to the frequency spectrum in the voice. Regula-
referred to as the singer’s formant cluster (F3, F4, and F5). This
tion of the spectrum involves adjusting the relative amplitudes
distance allows the listener to more clearly distinguish chiar-
of individual harmonics, which can be accomplished at the source
oscuro (bright-dark) timbre in a singer’s voice. One form of
or by the choice of vowels and allophonic variations of these
brightness, also referred to as vocal ring (Bartholomew, 1932),11
vowels. Specifically, recent investigations have shown that the
is related to the increased energy in the singer’s formant cluster
strength of the second harmonic relative to the first harmonic
(Sundberg, 1977),12 a hallmark of unamplified singing in the op-
is a major difference between vocal styles.2–6 For example, to
eratic style. However, lowering F1 and F2 by simply using a small
maintain a female belt or a call timbre at medium-high pitches
mouth opening (as in the case of a speech-like /u/ vowel) would
(G5–C5 and occasionally up to F5), as is frequently required
reduce the radiated output power too much for unamplified
in music theater styles, the first resonance frequency of the vocal
singing. Hence, stretching of the oropharyngeal space beyond
tract fR1 is often raised to its upper limit (much higher than in
what is typical for speech allows a moderate (/ɔ/-like) mouth
conversational speech) to reinforce a second harmonic (2fo) that
opening to be maintained while F1 and F2 are lowered.
is already strong at the source with modal (chest) registration.
The purpose of this investigation was to determine the degree
However, because there is a limit in how high this resonance
to which singers with limited access to sophisticated laborato-
frequency can be raised, music theater singers are limited in their
ry instrumentation can develop a personalized F1–F2 formant plot
fo range if they want to retain the dominance of 2fo in the spec-
to establish the fo range over which contrasting singing styles
trum. Conversely, in operatic styles in the C4 to C5 range (males
can be performed. It is known that vocal tract resonances are
affected by vocal tract length and cross-sectional area varia-
Accepted for publication August 22, 2016.
From the *National Center for Voice and Speech, Salt Lake City, Utah; and the †De-
tion (Fant, 1960,13 Section 2.3), and that F1 and F2 largely
partment of Communication Sciences and Disorders, University of Iowa, Iowa City, Iowa. determine vowel perception.14 Therefore, the degree to which
Address correspondence and reprint requests to Ingo R. Titze, National Center for Voice
and Speech, Salt Lake City, UT 84101. E-mail: [email protected]
an individual can stretch the dimensions of the vocal tract de-
Journal of Voice, Vol. 31, No. 3, pp. 382.e9–382.e13 termines the range of vowel modification or adjustment available
0892-1997
© 2017 The Voice Foundation. Published by Elsevier Inc. All rights reserved.
to that individual.15 We propose that providing an understand-
http://dx.doi.org/10.1016/j.jvoice.2016.08.014 ing of an individual’s formant range profile (FRP), in a manner
Ingo R. Titze, et al A Formant Range Profile for Singers 382.e10

similar to a voice range profile for pitch and loudness, can help formant peaks. It was anticipated that recording fidelity would
the individual better determine appropriate use of vowels for not be an issue for obtaining formant frequencies; hence, a variety
various stylistic vocal needs. of microphones, including Shure SM48s (Niles, IL) and inter-
nal microphones on students’ laptops, were used to obtain the
METHODS acoustic signal. The project was conducted in a space with
Vowels were divided into three categories, based on stylized vocal minimal background noise (below 65 dB) to reduce any inter-
tract shapes: neutral (speech-like), megaphone (belt or call), and ference of the acoustic signal, but a sound isolation booth was
inverted megaphone (opera-like), following shape descriptions not deemed necessary for frequency analysis.
of Titze and Worley3 for males and Titze et al4 for females. Table 1 Students were instructed to produce five speech vowels
lists the vowels we placed in each category. The first (neutral) (/i,e,ɑ,o,u) in a manner that felt like their “normal” speech pro-
category consists of vowels that are consistent with an individ- duction, but in vocal fry instead of high-pitched phonation. Next,
ual’s general speech gestures. The second category, the they were instructed to produce the “megaphone” vowels (æ, a,
megaphone, or divergent, mouth shape, is based on the above ɔ) by widening the mouth opening as much as possible while
studies and other pedagogical literature.16,17 It is known to produce maintaining the intelligibility of the vowel. Finally, they were
a timbre that is often described as open, yell-like, or brassy. This instructed to produce the “inverted megaphone” vowels (I, ɛ, ʌ,
shape facilitates the 2fo proximity to F1. It is created with a wide ʊ) with a moderate mouth opening, but widening the orophar-
opening at the lips, a narrow pharynx, and sometimes an el- ynx as much as possible while still maintaining the intelligibility
evated larynx (p. 22).17 It is characteristic of musical theater of the vowel. As the students carried out these tasks, one or more
singing, some jazz and pop singing, and high-energy gospel of the authors monitored the outputs perceptually and provid-
singing. The third category, the inverted megaphone or conver- ed further instruction when necessary to encourage individual
gent mouth shape, produces a timbre described as full and deep subjects to widen their pharynx or mouth more. Some subjects
if higher harmonics are sufficiently present, but also a “hooty” expressed difficulty in sustaining vocal fry on all vowel con-
or “fluty” quality when first harmonic (fo) energy dominates (p. figurations, but with practice, all were able to complete the task.
23).17 To create a dark “hooty” timbre, one must have a wide Once the tokens had been recorded, students used either Praat
oropharyngeal space and a narrower space toward the front of (Amsterdam, The Netherlands) or Voce Vista (Princeton, NJ) to
the mouth. Lessac18 has called this timbre the “Y-buzz” in speech, measure and record approximate frequencies of F1 and F2. In a
in contrast to the “call” timbre from which belting is derived. formal research setting, vocal tract resonance frequencies fRn would
The corresponding vocal tract shapes are not unique to humans. have been obtained19 instead of formant frequencies, but an im-
Chimpanzees are known to produce a “pant-hoot” and a “scream,” portant objective was for students to be able to obtain these
two sounds that have a similar dichotomy in vocal tract shape measurements by themselves, quickly and economically. For stu-
and related sound quality. dents using Voce Vista, the default settings (10 Hz bandwidth,
and power spectrum frequency scale of 0–5000 Hz) were pre-
Subjects served. A time point was selected in the middle of a steady portion
Data were gathered as a regular classroom activity during the of the vocal fry audio recording. The power spectrum was then
Summer Vocology Institute conducted at the University of Utah. visually inspected and frequencies for the first two peaks, F1 and
The class was entitled Voice for Performers, but only some reg- F2, were recorded. For students using Praat, the spectrogram
istrants were performers; others were speech-language pathologists was viewed with a range of 0–5000 Hz, window length of
interested in studying the singing voice. Students ranged in age 0.025 second, and a dynamic range of 90 dB. A steady segment
from their early 20s to over 50 years old, and experience in vocal of the audio signal, 250 milliseconds in length, was selected,
training, singing, and/or acting voice similarly varied widely across and the first two formants were retrieved using the following set-
the subject pool. In total, 77 subjects (54 females and 23 males) tings: maximum formant of 5000 Hz, window length of
participated in this study. Data were collected over a period of 0.025 second, and dynamic range of 90 dB. As the vocal tract
3 years and subsequently compiled by the authors for final configurations changed and formant frequencies shifted accord-
analysis. ingly, the “number of formants” setting was adjusted manually
for each token, usually between five and seven, until the formant
Equipment and procedures contours reasonably matched the intensity maximums visible in
Students sustained each vowel shape in vocal fry (as opposed the spectrogram.
to high-pitched phonation) to energize the vocal tract with a dense Using one of these two analysis tools (Voce Vista or Praat),
frequency spectrum, thereby obtaining the best resolution of F1 and F2 were extracted by the students for the following se-
quences: speaking vowels—/i, e, a, o, u/; inverted megaphone
mouth shape—/ɪ, ɛ, ʌ, ʊ/; and megaphone mouth shape—/æ, a,
TABLE 1. ɔ/. Students created their own personalized F1–F2 chart as a ref-
Vowels Used for Three Mouth Shape Categories erence tool for future use in helping determine vowel choices
Mouth Inverted on certain pitches in various singing styles. Recordings of 24
Shape Neutral Megaphone Megaphone of the subjects were reviewed by the authors, using the Praat
software, to check for validity of the students’ measurements.
Vowels /i, e, ɑ, o, u/ /æ, a, ɔ/ /I, ɛ, ʌ, ʊ/
In addition to the personalized F1–F2 charts produced by the
382.e11 Journal of Voice, Vol. 31, No. 3, 2017

students, the authors used the F1 and F2 values measured by the As two single-subject examples, consider a female singer and
students and calculated the area of the quadrangle created by a male singer characterized by the average of the formant fre-
plotting F1 against F2 for the vowels /i/, /æ/, /a/, and /u/. These quencies shown in Figure 1. Further, consider that an attempt
vowels were chosen as they generally presented the extreme high is made to sing these vowels on a G4 musical note, the middle
and low values for the two formant frequencies. Area for the quad- of the C4–C5 octave where most of the timbre adjustments occur
rangle was calculated from these points as in singing. Figure 2 shows the average female and male formant

( x1 y2 − x2 y1 ) + ( x2 y3 − x3 y2 ) + ( x3 y4 − x4 y3 ) + ( x4 y1 − x1 y4 )
A=
2

where xn and yn are the frequency values of F1 and F2, respec- frequencies for the five speaking vowels. Superimposed are grid
tively, for each vowel. lines that represent the first three harmonics for the note G4
(392 Hz). Note that F1 of the vowels /u/ and /i/ is close to fo.
RESULTS The timbre for /i/ and /u/ would be flute-like because of the dom-
Figure 1 shows the F1–F2 vowel space for the five selected speech- inance of first harmonic energy. Similarly, F1 for the /ɑ/ vowel
like vowels across all subjects, females on the left and males is close to 2fo, which means that the timbre for /ɑ/ would be brass-
on the right. The vowels /æ/ and /a/ are excluded because they like. (Brass instruments generally have a strong second harmonic.)
are in a special class for singing in a belt style. The /i/ and /u/ The timbre with these corner vowels can change dramatically
vowels best define the range of F2, ranging from 600 Hz to with small changes in fo. In some singing styles a sudden timbral
2900 Hz in females and 500 Hz to 2600 Hz in males. change is desirable, whereas in other styles a sudden timbral

FIGURE 1. F1–F2 chart for five selected speaking vowels, females left and males right.

FIGURE 2. F1–F2 formant space for five averaged speech vowels with harmonic grid lines superimposed for the note G4 (392 Hz).
Ingo R. Titze, et al A Formant Range Profile for Singers 382.e12

FIGURE 3. F1–F2 formant space for the averaged specialized singing vowels with harmonic grid lines superimposed for the note G4 (392 Hz).

change is avoided by choosing vowels that do not have for- Females experience the F1—2fo proximity (or near-tuning) at
mants close to a specific harmonic. The vowels /o/ and /e/, for slightly higher notes, around A4–B4, or 440–490 Hz. For /æ/
example, would appear to produce a stable timbre for a range and /a/, F1 is about 100 Hz higher than reported by Peterson and
of notes below and above G4. Barney14 for males and females, indicating that a formant stretch
Figure 3 shows the F1–F2 space for style-specific singing took place. However, on average the students did not come close
vowels, the megaphone-shaped vowels /æ/ and /a/ and the in- to the 1000–1200 Hz F1 values achieved by top-rank profes-
verted megaphone-shaped vowels /I/, /ɛ/, /Ʌ/, /ɔ/ and /ʊ/. Note sional theater singers.4
that the inverted megaphone-shaped vowels have F1 frequen- Figure 4 shows FRPs for four individuals, two males and two
cies that lie between fo and 2fo, indicating a timbre balance that females who showed the greatest difference in the area of the
could likely be maintained over several notes in either direc- vowel quadrangle /i/-/æ/-/a/-/u/. If vowels are produced accord-
tion from G4. Classically trained singers use these vowels in the ing to the procedure outlined earlier, this quadrangle measures
passaggio region of their voice, between D4 and A4. On the con- the degree of articulatory stretch that the individual is capable
trary, the megaphone-shaped vowels have F1 above 2fo, favoring of producing. The vowels /æ/ and /a/ stretch F1 and the vowels
this second harmonic in the spectrum. For the males, there is /i/ and /u/ stretch F2. Since formant frequencies can be mea-
almost a tuning of F1 to 2fo on this one note, but exact tuning sured in kHz, the area of the quadrangle here is expressed in
to any single harmonic is not required for stylistic distinctions. kHz2. The two females produced FRP areas of 0.195 kHz2

FIGURE 4. F1–F2 quadrangle for individuals with the smallest and the greatest FRP.
382.e13 Journal of Voice, Vol. 31, No. 3, 2017

and 0.911 kHz2, whereas the two males produced FRP areas of Acknowledgment
0.337 kHz2 and 0.625 kHz2. This work was supported by National Institute on Deafness and
Other Communication Disorders Grant R01 DC012045, awarded
to Ingo R. Titze, Principal Investigator.
DISCUSSION AND CONCLUSIONS
A primary objective of this study was to determine if singers REFERENCES
with limited access to laboratory instrumentation could personalize 1. American National Standards Institute (ANSI). 1994, American National
an F1–F2 vowel chart for the purpose of understanding voice Standard Acoustical Terminology. ANSI S1.1-1994 (R1999).
2. Bjorkner E. Musical theater and opera singing—why so different? A study
timbre in singing styles. The original intent was to produce a
of subglottal pressure, voice source, and formant frequency characteristics.
maximum FRP for each individual. This was not accomplished J Voice. 2008;22:533–540.
with our set of instructions. To produce such a maximum range, 3. Titze IR, Worley AS. Modeling source-filter interaction in belting and
it is likely that an articulatory model will first be needed to de- high-pitched operatic male singing. J Acoust Soc Am. 2009;126:1530–1540.
termine which configurational stretches will produce the ultimate 4. Titze IR, Worley AS, Story BH. Source-vocal tract interaction in female
operatic singing and theater belting. J Sing. 2011;67:561–575.
formant stretches. With that determination, a better set of stretch
5. Bourne T, Garnier M. Physiological and acoustic characteristics of the female
instructions may result, ultimately leading to an FRP that can music theater voice. J Acoust Soc Am. 2012;131:1586–1594.
be considered maximum for vowel availability toward a given 6. Sundberg J, Thalen M. Respiratory and acoustical differences between belt
style and a given pitch range. This study was only a first step and neutral style of singing. J Voice. 2015;19:418–425.
in this direction. 7. Sundberg J, Gramming P, Lovetri J. Comparisons of pharynx, source,
formant, and pressure characteristics in operatic and musical theatre singing.
Vocal timbre is in part predictable on the basis of the prox-
J Voice. 1993;7:301–310.
imity of specific harmonics to specific formants. Unlike in 8. Stone RE Jr, Cleveland TF, Sundberg PJ, et al. Aerodynamic and acoustical
most man-made musical instruments, in which an attempt is measures of speech, operatic, and Broadway vocal styles in a professional
made to preserve a distinct timbre of the instrument by simul- female singer. J Voice. 2003;17:283–297.
taneously tuning multiple harmonics to resonances, the biological 9. Miller D. Resonance in Singing: Voice Building Through Acoustic Feedback.
Princeton, NJ: Inside View Press; 2008.
vocal instrument relies on highly variable vocal tract shapes to
10. Titze IR, Baken RJ, Bozeman K, et al. Toward a consensus on symbolic
adjust the output spectrum. Results here showed that the area notation of harmonics, resonances, and formants in vocalization. J Acoust
of the vowel quadrangle (in kHz2) could be more than a factor Soc Am. 2015;137:3005–3007.
of 2 across individuals. Future studies could address the effect 11. Bartholomew WT. A physical definition of “good voice-quality” in the male
that voice training can have on the ability to stretch the voice. J Acoust Soc Am. 1934;6:25–33.
12. Sundberg J. The Science of the Singing Voice. DeKalb, IL: Northern Illinois
FRP for a given individual. Also of interest would be the effect
University Press; 1987.
that native language, race, or body size could have on the 13. Fant G. Acoustic Theory of Speech Production. The Hague: Mouton; 1960.
FRP. 14. Peterson GE, Barney HL. Control methods used in a study of the vowels.
Quantification of the degree of reduction of formant space was J Acoust Soc Am. 1952;24:175–184.
not addressed well in this study. Future studies could address a 15. de Boer B. Investigating the acoustic effect of the descended larynx with
articulatory models. J Phon. 2010;38:679–686.
quantification of the oropharyngeal stretch for the centralized
16. Estill J. Basic Figures and Exercises Manual. New York, NY: Imagetech;
vowels /I/, /ɛ/, /ᴧ/, /ɔ/, and /ʊ/ with mouth openings that allow 1992.
maximum sound radiation from the lips. Here the question would 17. Bozeman KW. Practical Vocal Acoustics: Pedagogic Applications for
be: How small can the area in the F1–F2 space be to preserve Teachers and Singers. Hillsdale, NY: Pendragon Press; 2013.
both the timbre and vowel intelligibility when the objective is 18. Lessac A. The Use and Training of the Human Voice: A Bio-Dynamic
Approach to Vocal Life. Mountain View, CA: Mayfield Publishing; 1996
to get the most sound to the listener? Measurement of lip opening
1967, reprint.
and magnetic resonance imaging or computed tomography 19. Epps J, Smith JR, Wolfe J. A novel instrument to measure acoustic
scanning would be the obvious methodology to quantify the resonances of the vocal tract during phonation. Meas Sci Technol.
stretch. 1997;8:1112.