1 Overview of the Thesis

1 Overview of the Thesis

How are beautiful singing voices developed? What features of the singing voice define

a voice of quality? Can these features be measured accurately and more importantly,

specifically taught? In 1967, Appleman, in The Science of Vocal Pedagogy, asserted

“vocal pedagogy cannot survive as an independent educational entity if the

physiological and physical facts, which comprise its core, remain subjects of sciolism

(superficial knowledge). Researchers must uncover and interpret scientific facts about

voice, voice quality and vocal pedagogy so that they might become realistic pedagogical

tools that may be employed by future teachers of voice” (Appelman, 1967, p. 5). This

thesis takes up Appleman’s call for research on the singing voice that both contributes

to our understanding of what constitutes a voice of quality and provides useful tools for

vocal pedagogues.

It includes published work in international peer reviewed journals (see Appendix W)

that extends previous research into the singing voice and its pedagogy. The hypotheses

are well motivated from the existing research and from established pedagogical

techniques. It forms the basis for future interdisciplinary work in assessing, describing

and training the classical singing voice.

1.1 Historical background to the thesis

Singing is the most natural of all musical arts. Wagner wrote ‘The human voice is the

practical basis of Music, and however far the latter may journey on her primal path, the

boldest combinations of the tone-setter, the most daring execution of the instrumental-

virtuoso, will always have to hark back to the purely singable, to find the law for their

1

achievements’ (Ellis, 1994). Throughout recorded history to the present day, singing has

played a significant role in cultural, social and emotional life.

1.1.1 Singing style in the western classical tradition

George Bernard Shaw said, “the notion that singing has deteriorated in the present

century is only a phase of the Good Old Times delusion … Every musical period suffers

from the illusion that it has lost the art of singing and looks back to an imaginary golden

age in which all singers had the secret of the bel canto taught by Italian magicians and

practiced in excelsis at the great Opera Houses of Europe” (Shaw, 1960).

Historical accounts of singing and singing technique do not adequately define vocal

quality. They provide information on famous singers and their teachers in each century

and give anecdotal accounts of singing training and subjective descriptions of vocal

quality. This tradition, often referred to as bel canto, broadly represents the classically

trained voice of opera and concert singers. Classical singing style has evolved over five

centuries in accordance with changing musical styles, compositional styles, cultural and

social factors and performing venues; all of which impacted on vocal production.

Scholars differ with respect to the advent of western classical singing: some attribute the

rise of the virtuoso singer in the late 16th to early 17th centuries as the beginning of

western classical singers (Duey, 1980; Stark, 1999); others date its advent later to the

development of specific operatic singing style which began in the 19th century (Potter,

1998).

2

1.1.2 Classical voice quality

1.1.2.1 Changing musical styles

The birth of opera (including favoli in musica, the earliest manifestations of opera) at

the beginning of the Baroque period marked the beginning of a new style of solo

singing. Composers like Monteverdi pioneered a new compositional style emphasising

the importance of the text, “making the words the mistress of the harmony and not the

servant”. The solo singer was the focus of the musical performance, in contrast with the

polyphonic Renaissance style, where the rules of counterpoint and voice-leading made

all voices and instruments equal (Rosand, 1989; Treitler & Strunk, 1998).

This new musical art form revived the humanist ideals of the ancient Greeks, with the

rise of the ‘musician poet’. Singing was regarded as an extension of oratorical speech.

As the singer was elevated to narrator, he had the freedom to interpret the poetic text in

his vocal performance. Singers of the day used extensive improvisation or

ornamentation to enhance their interpretation of the text and historical conventions in

these passagi (improvisations). The earliest singing treatises (Bacilly, 1668/1968;

Caccini, 1602/1930; Praetorius, 1619/1979) focussed on the art and interpretation of

ornamentation.

Musicologists attribute expressivity in early Baroque vocal music to the relationship

between text and music, but cannot ascertain the effect of these musical changes on

voice quality. However, the Italian singing style was imitated throughout 17th century

Europe. German and French writers praised the Italians’ ability to sing in both an

‘oratorical’ (Praetorius, 1619/1979) and expressive style, capturing the emotion of the

3

text (Mersenne, 1636/1979). Italians’ solo singing style was different to both the

Germanic (Praetorius, 1619/1979) and English (Playford, 1694/1972) choral singing

styles. The Italian language, with its sonorous vowel sounds, was ideal for enunciation

(Bacilly, 1668/1968) of poetic texts.

The impact of musical style in singing is demonstrated by the 17th century English

national singing style which followed an autonomous tradition, distinct from European

singing, as English opera developed a different compositional style as music drama with

spoken dialogue. By 1711, Handel’s opera Rinaldo opened in London, and England,

like the rest of Europe, was exposed to the Italian singing style.

1.1.2.2 Singers rise to fame

By the early 17th century, public performances of Italian opera were widespread.

Castrati dominated opera singing from the mid-17th century and were cast in the most

important roles in operas, while tenors and basses were reduced to minor roles. Castrato

singing was considered the highest art, superior to all other vocal qualities and these

singers were distinguished by extraordinary vocal power and flexibility unmatched by

other singers. Their powerful voices suited the new performance venues, as the first

public performances of opera were in rooms at least twice the size (Worsthorne, 1968)

of the first performance of Monteverdi’s Orfeo at the palace of Margherita Gonzaga-

Este in 1607 (Fenlon, 1984). Twentieth century critics speculate that the bel canto

tradition was synonymous with castrati and their highly florid vocal style (Duey, 1980;

Foreman, 1969).

4

1.1.2.3 Virtuosity

The earliest singing treatises (Bacilly, 1668/1968; Caccini, 1602/1930; Praetorius,

1619/1979) focussed on instructions for ornamentation, an intrinsic component of good

singing at that time. Later writers Tosi and Mancini described the art of vocal

embellishment: the “even, clear, flexible and moderately quick” trill as “indispensable

to singers” (Tosi, 1743/1967, ch 3). Tosi asserted that a singer who could “produce no

trill or only a faulty one will never become a great singer.” Tosi and Mancini also

identified distinct vocal registers with different sounds qualities, voce di petto (chest

voice) and voce di testa (head voice).

Eighteenth century composers Handel, Mozart and even Rossini incorporated the

virtuosic trend in singing into their scores, particularly for castrati and coloratura

sopranos. Singers extended this virtuosity in elaborate and improvised cadential figures.

Ornamentation as defined in the 18th century was different to the 20th century definition

of vibrato. For example, W.A. Mozart, in a letter to his father (Paris, 12 June 1778)

described the use of an ‘appropriate’ vibrato. He praised singer Joseph Meisner for his

cantabile singing (Anderson, 1991), but described his tendency to tremble, which

distorted the sound so much that Mozart could determine beats of a crotchet or quaver.

He said that “[t]he human voice trembles naturally – but in its own way – and only to

such a degree that the effect is beautiful. Such is the nature of the voice; and people

imitate it not only on wind instruments, but on stringed instruments too and even on the

clavier. But the moment the proper limit is overstepped, it is no longer beautiful –

because it is contrary to nature. It reminds me then of the organ when the bellows are

5

puffing” (Anderson, 1991, p. 552). Singing pedagogue Tosi also warned singers to sing

“without vocal trembling” (Tosi, 1743/1967).

1.1.2.4 Changing accompaniments

Changes to the ‘loudness’ of singing occurred in response to changes in sound quality

of the instruments used to accompany voice: for example, the harpsichord superseded

the lute as an accompanying instrument (Potter, 1998). “[I]nevitably, singers had to sing

louder” (Potter, 1998, p. 50). By the 19th century, voices had undoubtedly grown in size.

1.1.2.5 A new vocal style

In the 19th century, Rossini reportedly exclaimed “Alas for us, we have lost our bel

canto” (Jander & Harris, 2004). Changing vocal styles in opera, concert and oratorio

singing and a shift towards the modern idea of a classical voice gradually appeared in

19th century literature on voice (Potter, 1998).

Orchestras steadily grew in size alongside dedicated performance venues (rather than

domestic musical settings). By the mid-1800s, composers complained of voices that

could not be heard in larger halls (Cairns, 1969). Voices changed throughout the century

to suit these new performance requirements. Singers were no longer praised for their

virtuosic vocal ornamentation. Musical literature reports new classifications of voice or

‘fachs’ to define vocal timbre (tenore robusto, tenore di forza, Heldentenor, ‘Verdi

baritone’, ‘Falcon soprano’, ‘dramatic soprano’ and ‘lyrico spinto’). These fachs

describe vocal range and volume or carrying power.

6

The tenor voice, with an extended and powerful upper range, superseded the castrati

voices of previous centuries. Similarly, soprano voices extended upwards and specific

roles were written for baritones and mezzo soprani by composers such as Verdi.

Remarkable new vocal qualities emerged, which we would now recognise as ‘operatic’

For example, the first tenor high C by singer Duprez in Rossini’s William Tell in 1831

(Cairns, 1969) shocked unsuspecting audiences of the day.

Wagner and Verdi emphasised the relationship between text and music and cultivated a

speech influenced (Sprechgesang) or dramatic declamation style. Voices performing

epic Wagnerian works required unprecedented strength and endurance to be heard over

the largest orchestras. The term bel canto was used by Italians in reaction to the

Wagnerian singing style, and renewed patriotic support for the older traditions emerged

(Duey, 1980).

Surveys of the pedagogical literature (Burgin, 1973; Fields, 1947; Monahan, 1978)

identify the 1840s and Manuel Garcia II’s publications (Garcia, 1911) as the beginning

of a new teaching style based on scientific theory. Descriptions of vocal quality in

classical singing are supported by physiological instructions to achieve the quality. For

example, Garcia was first to advocate a low larynx position, and its new vocal quality.

Garcia’s observations of the larynx with the primitive laryngoscope marked the

beginning of a scientific approach to singing, which has influenced all subsequent vocal

study. By the late 19th century, the majority of pedagogical texts refer to anatomy,

physiology and the acoustics of voice production (Burgin, 1973; Fields, 1947;

Monahan, 1978).

7

1.2 Empirical research in singing voice

Vocal pedagogy in Western art music is now four centuries old, yet the definition of a

voice of quality remains as complex in the 21st century as it was in centuries before

acoustics and audio recordings. Empirical research into the singing voice uses acoustic

and perceptual methodologies to define vocal parameters associated with “good” vocal

quality and relate it to vocal functioning. Later studies report components of good

singing, identify differences between male and female singers, between singers and

speakers and between classical singers from pop or other styles of singing. These

studies also attempt to capture the qualities or features that distinguish novice from

professional singers and from local singers to singers of international renown.

While there is a proliferation of acoustic studies in singing voice throughout the

century, current writers (Callaghan, 2000; Miller, 1998) point out that it is difficult to

draw firm conclusions about the implications for singing pedagogy of many of these

scientific studies which typically use small numbers of subjects, rarely link the vocal

strategies used to perceptual results, often use student singers rather than professionals,

and rarely identify the teaching/learning approach of the subjects. Most importantly,

“The subtle individual properties that set one voice apart from another should not be

averaged out” (Miller, 1998, p. 299).

1.2.1 Vibrato

Vibrato is taken for granted as an intrinsic quality of the classical singing voice and in

pedagogical literature, as a component of tone quality in coordination, (Vennard, 1968)

richness (Seashore, 1938; Vennard, 1968) and vibrancy (Miller, 1996). Vibrato was

8

advocated as desirable in singing as a component of tone quality, justified by its

prevalent use by singers and instinctive acceptance by listeners. Vibrato is regarded,

especially in pedagogical circles, as a result of good singing, and, when successful,

almost imperceptible to the human ear. It is accepted in the singing literature that poor

vibrato is indicative of poor technique (Miller, 1996; Vennard, 1968) and inferior sound

quality. In classical singing, sound without vibrato, or straight-tone, has been described

as dull or spread (Vennard, 1968) and lacking freedom, power (Miller, 1996) and ring

(Vennard, 1968). A delayed vibrato onset indicated a faulty technique and an unnatural

voice quality (Miller, 1996).

Researchers have used acoustic, physiological and perceptual methods to study vibrato.

Vibrato is quantifiable by rate, the number of cycles per second, extent, the fluctuation

of pitch above and below the mean pitch, and onset, the delay from the onset of

phonation until the first vibrato cycle (Sundberg, 1988). Despite extensive research

(Prame, 1994, 1997; Shipp, Sundberg, & Hadlund, 1984; Sundberg, 1995a; Vennard,

1968) researchers have not reached consensus on what features constitute ideal vibrato.

Seashore (1932) defined good vibrato as “a pulsation of pitch, usually accompanied

with synchronous pulsations of loudness and timbre, of such extent and rate as to give a

pleasing flexibility, tenderness and richness to the tone” (Seashore, 1932, p. 7).

Seashore categorised vibrato as a musical ornament, invaluable for investigation

because it was used across all tones and impacted on tone quality. Using recorded

voices of famous singers, Seashore reported that vibrato was present in 95% of trained

singers’ voices and described it as a vocal ornament performed on every note. Later

9

studies investigate vibrato as an intrinsic component of classical vocal quality, distinct

from embellishment or ornamentation.

Ideal vibrato parameters vary by musical era and musical genre. Further, acoustic and

reports of appropriate vibrato do not accord with perceptual judgments of listeners

(Rothman, Diaz, & Vincent, 2000) and many studies define undesirable or defective

vibrato perceptually (Miller, 1996; Vennard, 1968) and acoustically (Keidar, Titze, &

Timberlake, 1984; Seidner, Nawka, & Cebulla, 1995) rather than attempt to identify

vibrato in voices of quality.

Physiological studies of vibrato produced varying results. Vibrato has been attributed to

crico-thyroid oscillations and 4 of 5 singers showed consistent lateral crico-arytenoid

muscle activity, oscillating at the same rate as the singers’ vibrato (Hirano, Hibi, &

Hagino, 1997). Comparing vibrato and straight tone, Hirano, Hibi & Hagino (1997)

found that some singers either widened or narrowed the supraglottic space and

hypopharynx when they produced vibrato. Vibrato has been linked to movement of the

larynx and many researchers drew comparisons to pre-existent neurological or existent

physiological tremors (Titze, Solomon, Luschei, & Hirano, 1994).

Studies seek to test singers’ conscious modification of vibrato: in the communication of

emotion (Howes, Callaghan, Davis, Kenny, & Thorpe, 2004; Sundberg, Iwarsson, &

Hagegard, 1995); loudness or sound pressure level (Michel & Myers, 1991; Titze et al.,

1999); and as consequence of vocal training (Brown, Rothman, Morris, & Sapienza,

2001; Brown, Rothman, & Sapienza, 2000; Ekholm, Papagiannis, & Chagnon, 1998;

10

Mendes, Rothman, Sapienza, & Brown, 2003; Wapnick & Ekholm, 1997). No studies

have assessed the effect of specific technical instruction on vibrato parameters.

1.2.2 Spectral energy

Classical and operatic singers produce an energy boost in their vocal quality, which is

perceived as a sound quality and also enables them to be heard without amplification.

Bartholomew (1934) first acknowledged a high energy spectral reinforcement as a

component of good voice quality. Sundberg (1974) described this energy as the

“singer’s formant”; a clustering of formants F3, F4 and F5 causing an energy peak in 2-

4 kHz. Acoustic research proposed that this energy is most likely to occur when the size

ratio of the pharynx to larynx tube/inlet is 6:1, which is achieved by lowering the larynx

in classical vocal quality (Sundberg, 1974; Titze, 1998; Titze, 2001).

The peak of high energy band between 2 and 4 kHz depends on individual singer and

also on vocal ranges (eg. soprano to bass) (Dmitriev & Kiselev, 1979). Soprano voices

do not consistently demonstrate spectral reinforcement in 2-4 kHz comparable to male

singers (Bartholomew, 1934; Sundberg, 1977, 1995b; Weiss, Brown, & Morris, 2001).

Studies show that high voices demonstrated a spectral reinforcement between 2.7 - 3.5

kHz (mezzo-soprano to high soprano) (Dmitriev & Kiselev, 1979) or 2.6 - 4.5 kHz but

concluded that this energy band was not as significant for female singers as for male

singers (Weiss et al., 2001). However, more recent studies of professional female

sopranos at a high performing level identified a stronger high level energy boost around

3 kHz (Barnes, Davis, Oates, & Chapman, 2004; Sundberg, 2001; Thorpe, Cala,

Chapman, & Davis, 2001) which may be associated with performing level and

subsequently singers’ audibility in large opera theatres.

11

Many studies consider the increase in vocal energy above 2 kHz a result of training in

vocal projection (Löfqvist & Mandersson, 1987). However, vocal projection or carrying

power is an important objective for young singers (Vurma & Ross, 2000), but which

can be developed to the detriment of vocal quality. Energy boosts above 2 kHz have

been associated with classical singers of the highest renown (Barnes et al., 2004; Thorpe

et al., 2001). Methods of measuring the energy distribution in voices give an indication

of acoustic features seen in good voices. Although neither proves the presence of a

singer’s formant, it provides a tool by which to measure and compare spectral energy in

the voice, perhaps throughout the development of a singer’s career (Lundy, Roy,

Casiano, Xue, & Evans, 2000).

More recent research has proposed methods to quantify this high range energy. Omori,

Kacker. Carroll, Riley & Blaugrund (1996) devised a method of objectively rating voice

quality by measuring the difference between the height of energy peaks associated with

fundamental frequency and singing power peak or energy boost (0-2 kHz and 2-4 kHz).

This result was identified as the singing power ratio (SPR). They compared singers with

non-singers and found that singers produced a significantly smaller SPR than non-

singers. These singers’ voices were described perceptually as demonstrating more

“ringing quality” and “richness” than non-singers. Singers with ≥ 4 years of training

were more likely to demonstrate a smaller SPR than singers with < 4 years training.

While SPR was not intended to define singer’s formant, it does provide a way in which

to compare inter and intra-subject high energy values.

12

1.2.2.1 Long term average spectra

In current singing research, the long term average spectrum (LTAS) is widely used as a

simple, non-invasive analysis for musical timbre and vocal features, both in speech and

singing. LTAS have been used to assess the effects of this characteristic energy boost >

2 kHz over time. An LTAS is the mean of all spectra over time (Baken & Orlikoff,

2000), and a measure of the dB level of the time-average of the power of the acoustic

signal at each frequency. As a research tool, LTAS facilitates the complex acoustic

evaluation of music, and captures the quantity of information produced in an entire

musical excerpt (Jansson & Sundberg, 1975).

In classical or operatic voice, an LTAS presents an energy boost around 3 kHz and this

has been linked to carrying power over an orchestra or audibility in an opera theatre

(Barnes et al., 2004; Sundberg, 1974; Thorpe et al., 2001; Vurma & Ross, 2000).

Energy in the 2-4 kHz regions which clusters the third, fourth and fifth formants, so

close in frequency that they appear as a unified peak, and amplifies a classically trained

singer to be heard over an orchestra (Sundberg, 1974). This amplification in male

voices, has been referred to the “2800” (Bartholomew, 1934), because of the frequency

level of the peak, or centre level of the singer’s formant, but is not as readily identified

in female voices (Weiss et al., 2001).

Methods used to generate LTAS curves have followed rigorous time and musical

stimuli criteria in order to reduce the effect of these variables on the final plot. LTAS

studies adhere to the recommended 20-30 seconds length of LTAS (Jansson &

Sundberg, 1975; Sundberg, 2001), although studies vary from short segments of 20

13

seconds (Howard, Szymanski, & Welch, 2002; Kenny & Mitchell, 2004, in press) to

samples lasting over a minute (Barnes et al., 2004; Mitchell & Kenny, 2004). After this

time, the plot identifies certain consistent features contained in the sound over time, and

averages out short-term variations in phonetic or articulatory strategies. This method

evens-out spectral changes associated with phonetic structures and musical stimuli and

gives a profile of the voice quality (Löfqvist & Mandersson, 1987). Sundberg proposed

that the energy boost above 2 kHz utilizes a bandwidth of spectra unfulfilled by the

orchestral energy levels (Sundberg, 1974).

LTAS have been presented as visual cues for vocal qualities and vocal genres (Borch &

Sundberg, 2002; Cleveland, Sundberg, & Stone, 2001; Sundberg, 1974). However,

quantitative interpretation of LTAS attempts to reduce the information in LTAS to a

single meaningful number to capture this energy boost in classical singing. Löfqvist and

Mandersson (1987) calculated spectral tilt, the difference between the between low (0-2

kHz) and high range (2-4 kHz) energy using the regression line between the two

prominent energy peaks. Novak and Vokral (1995) evaluated the spectral tilt of singing

students and found a steeper regression line in male voices (-3.65 – -4.93) than female

voices (2.71 – -3.33). These measurements were used, like SPR (Omori et al., 1996), to

provide an objective assessment of vocal quality. Similarly, Thorpe, Cala, Chapman &

Davis (Thorpe et al., 2001) compared the areas of energy 0-2 and 2-4 kHz rather than

the level of the energy peaks. These measurements have confirmed increased vocal

projection in high level professional singers (Barnes et al., 2004; Thorpe et al., 2001).

14

1.3 Pedagogical directions

Twentieth century vocal pedagogues asserted the need for singing teachers’

understanding of vocal functioning (Appelman, 1967; Miller, 1998). While a scientific

understanding of the voice is not a prerequisite for developing a good voice, it can

capture and document vocal pedagogy for future generations. Miller pointed out that

“for scientific research to be valid and have practical value in the studio, teachers of

singing must be involved, knowledgeable and interested. [Their] input, in areas of

expertise best understood by voice teachers, is essential” (Miller, 1998).

Singing pedagogy and singing literature in the last 25 years incorporates the new voice

science literature. Singing pedagogy conferences play an important role in the

dissemination of ideas and discussion on singing and singing technique. Bodies such as

the National Association of Teachers of Singing (NATS) in the United States, the

Australian National Association of Teachers of Singing (ANATS) and the British Voice

Association (BVA) are pivotal to the understanding and involvement of singing

pedagogues in singing research. Journals devoted to voice are directly linked to these

associations: Journal of Singing to NATS; Logopedics Phoniatrics Vocology to BVA

and Australian Voice to ANATS. Such publications facilitate dissemination of voice

research directly to singing pedagogues.

Pedagogical literature reports and disseminates scientific findings on voice. However,

some vocal historians are sceptical about the role of empirical research in singing and

they challenge researchers to undertake research relevant to the singing studio. Stark

was moved to comment that “voice scientists do not generally concern themselves with

15

elaborate historical or pedagogical constructs like bel canto, nor do they put much stock

in resonance images, ‘placing the voice’, or other traditional techniques of voice

training. Rather, they try to isolate specific physiological, acoustical, and aerodynamic

aspects of the singing voice in the controlled environment of the voice research

laboratory” (Stark, 1999).

Few empirical studies classify the components of good sound quality in singing,

although perceptual studies seek to describe good singing (Wapnick & Ekholm, 1997).

Recent singing studies have focussed on the singer’s intentions to achieve a good sound

and identified differences in singing with and without emotional connection or

enhanced projection (Barnes et al., 2004; Foulds-Elliott, Thorpe, Cala, & Davis, 2000;

Sundberg, 1997; Thorpe et al., 2001). Such studies imply that singers can consciously

manipulate their sound production.

Are great voices a natural gift or can they be developed by sound pedagogy? What

acoustic and perceptual features of the singing voice define a voice of quality? Can

these vocal features be accurately measured and taught? If they can be taught, what are

the best techniques for achieving the best vocal quality? Internationally renowned vocal

pedagogue Richard Miller stated “in any age, the main duties of a teacher of singing,

with regard to technique, have always been chiefly to (a) analyse vocal problems and

(b) design proper solutions for them.”

While singing pedagogy has absorbed research of the past 50 years, it has not directly

addressed specific vocal tools or singing techniques. This thesis seeks to provide

16

stronger links between empirical research and pedagogy in the singing studio, and

recognises the importance of singing pedagogues in this enterprise. Applied research in

singing must address the goals of singing pedagogy and systematically assess and

document the acquisition of vocal mastery. This is the first body of work to examine an

individual singing technique.

Thus, this thesis is situated at the intersection between voice research and vocal

pedagogy. It is devoted to the exploration of one commonly used vocal training

technique in classical singers – that of “open throat”.

It is the synthesising nature of a thesis that makes the last written the first presented.

The key points are presented in the first paper which summarises the body of work

presented in the thesis and ensures that it is accessible to the singers and pedagogues to

whom it is addressed.

17

REVIEW PAPER

Mitchell, H. F., & Kenny, D. T., (in review). Open throat: acoustic and perceptual support for pedagogical practice. Journal of Singing.

18

The open throat technique 1Journal of Singing, in review

Open throat: acoustic and perceptual support for pedagogical practice

Helen F. Mitchell and Dianna T. Kenny

From the Australian Centre for Applied Research in Music Performance (ACARMP), Sydney Conservatorium of Music, The University of Sydney, New South Wales 2006, Australia

Journal of Singing; in review

Music and song permeate every facet of human experience and play a significant role in cultural and social life. How are beautiful singing voices developed? What features of the singing voice define vocal beauty? Are listeners reliable in their assessments of these features? Can these features be accurately measured and taught? This paper reports on the first body of work of its kind to define and assess one feature of the technique of the classical singing voice – open throat. Such is its prominence in vocal pedagogy, Richard Miller commented ‘It would be hard to find a voice teacher who recommended singing with a closed throat’ (Miller 1996b) (p.58). The fact that a technique is universally prac-ticed does not in itself provide justification for its continued use; it must be validated empirically. Accordingly, we examined the acoustic and perceptual features of the classical singing voice while singers used and then reduced usage of open throat in a series of five studies. We confirmed that open throat technique is a useful and valid pedagogical technique that enhances vocal quality in the female classical voice. The methods and procedures developed to assess open throat can be used to assess other pedagogical training techniques, thereby contributing to the science of voice and vocal pedagogy.

Helen Mitchell, Australian Centre for Applied Research in Music Performance (ACARMP), The Conservatorium of Music, The Uni-versity of Sydney, New South Wales, Australia 2006 Phone: 61 2 9351 9644; Fax: 61 2 9351 9540 Email: [email protected]

INTRODUCTION

“Open throat” is a term regularly used in the singing studio. It is a pedagogical concept trans-mitted through the oral tradition of singing. Mitch-ell and Kenny (Mitchell et al. 2003; Mitchell and Kenny 2004a, 2004b, in press; Kenny and Mitch-ell in press, 2004) examined the technique in the modern singing studio using qualitative, acoustic, perceptual and statistical analyses to relate singing instructions, terminology and spectra to the sound qualities produced by instruction in this technique and validated its role in singing pedagogy.

Achieving vocal mastery

How are beautiful singing voices developed? What features of the singing voice define a voice of quality? Can these features be measured accu-rately and more importantly, specifically taught? Historically, teaching and evaluating singing has been guided by an oral tradition in which peda-gogical techniques are handed down from one generation of singing teachers to the next. Today, empirical research into the singing voice has the potential to benefit the singing community by doc-umenting and systematically assessing the acqui-sition of vocal mastery.

Since the 1980s technology to measure vocal

acoustics has become steadily more sophisticated and has been increasingly used in experimental research, with the findings of such research now being incorporated into texts on singing (Miller 1996b; Nair 1999; Sundberg 1977; Thurman and Welch 2000). International authority Richard Miller asserts “It is the responsibility of the sing-ing teacher in a scientific age to interpret and ex-pand vocal traditions through the means of current analysis so that the viable aspects of tradition can be communicated in a systematic way” (Miller 1998,p. 299).

However, it is often difficult to see how the general principles established in scientific stud-ies can be applied to the subtleties of developing individual vocal quality. To date, few studies link the vocal strategies used by singers to acoustic studies and perceptual judgments by pedagogues. This makes it difficult to draw firm conclusions about the implications of these scientific studies for singing pedagogy as they rarely identify the teaching/learning approach of the subjects (Miller 1998; Callaghan 2000).

In this paper, we report the first body of work to track a singing technique as practiced in the sing-ing studio using a comprehensive approach to all aspects of singing research (Mitchell and Kenny

19


2004a, 2004b, in press; Mitchell et al. 2003; Ken-ny and Mitchell in press, 2004). As it is pedagogi-cally informed and verified, this series of studies has the potential to substantially progress the fields of acoustic and perceptual assessment of voice in order to benefit the wider singing community and enhance pedagogical approaches.

DEFINING OPEN THROAT TECHNIQUE

Concepts relating to open throat can be traced throughout pedagogical and scientific singing liter-ature (Burgin 1973; Fields 1947; Monahan 1978). It is defined as a complex process that is both a pedagogical instruction and a perceived sensation or action that results in a specific sound quality. Use of the technique makes a difference to vocal quality. Indeed, Vennard defined open throat as the ‘condition agreed upon by most voice teachers as desirable for resonance’ (Vennard 1968, p.252). Current support for the use of open throat in sing-ing technique is widespread (Miller 1996a; Reid 1975, 1983). It elicits a sound quality which is per-ceived as resonant (Miller 1996b; Vennard 1968), round (Joiner 1998), free (Ware 1998), pure, (Ma-rafioti 1981) rich and warm (McKinney 1982) and is attributed to freedom from ‘constrictor tensions’ (Reid 1983)(p. 83). The sound quality is linked to vocal actions: the preparation to sing or inhala-tion (Hemsley 1998; Manèn 1987; Miller 1997b, 1997a); through the surprise breath or smelling the rose imagery (Hemsley 1998; Miller 1996b; Pu-ritz 1956); and visualizing space within the throat, through an ‘air-ball’ or ‘soap bubble’ (Herbert-Caesari 1951; Manèn 1987).

Do expert singing teachers actually agree on the definition of open throat technique, and if so, on what such a technique was meant to achieve? Subjective terminology used in singing pedagogy does not always indicate a specific vocal instruc-tion or action. Often terminology and meaning are not the same for each teacher. Communication of techniques in singing pedagogy can be improved by attempts to gain consensus on the use of ter-minology. This was the goal of our first study on open throat. We interviewed fifteen expert singing pedagogues to explore current thinking regard-ing terminology, pedagogy, sound quality and the perceived physiology associated with open throat technique (Mitchell et al. 2003).

Terminology of open throat

The majority of our fifteen vocal pedagogues interviewed agreed that open throat was essential

to good singing and more specifically to classical singing. Most included the technique as fundamen-tal in their singing training. Freedom, collar and depth were suggested as alternative terms to clarify meaning or to refer to an action. Pedagogues were aware of the need to tailor their terminology and instructions in the singing studio to each student’s vocal needs and learning styles. Despite the use of different terminology to describe the technique, there was consistency in the vocal instructions to achieve it and with respect to the sound qualities it produced. Pedagogues taught conscious control of open throat using laugh, sob, correct inhalation or maintaining the posture of inhalation.

Sound quality

Open throat produces a distinctive sound qual-ity recognised by most singing pedagogues. Table 1 [after (Mitchell et al. 2003)] identifies the most common terms used to describe the sound qual-ity. Pedagogues associated open throat with both a sound quality that was characterized by freedom, warmth and openness and an action that produced balance, coordination, evenness and consistency. The terms related to vocal ‘quality’ such as ‘warm’, ‘full’ or ‘round’, as well as the ‘functional’ terms such as ‘easy’ or ‘clean’ were used interchange-ably by participants in this study. Although the 15 participants offered 18 terms to describe sound quality, there were clear associations or similari-ties in their usage and application.

Physiological action

Singing pedagogues relied on their perceptions of sound qualities to determine the physiological processes at work in the production of the sound quality. Open throat was defined as a technique to maximize pharyngeal space and/or abduct the ventricular folds. Participants who consistently mentioned a presumed “abduction” or retraction of the false vocal folds or ventricular folds report-ed seeing the action of endoscopy and linked the technique to the term retraction. Those participants who used the term and technique of retraction did so to reduce constriction or tension to achieve a healthier sound quality as well as to achieve a spe-cific sound quality.

Recommendations

This study advanced previous work in clarifying terminology related to open throat (Burgin 1973; Fields 1947; Monahan 1978). In this study, peda-gogues’ use of open throat confirmed the perceived value of the technique. Pedagogues did not sepa-

20


rate quality and function descriptors and seemed comfortable with their use interchangeably in the studio. The research highlighted the usefulness of qualitative research as a tool for the generation of research questions and clarification of terminol-ogy used to describe vocal quality and to encour-age the use of anatomically correct terms in vocal pedagogy. For example, de-constriction is a more accurate term than retraction.

Having established the widespread usage of the term ‘open throat’ and its practice in the singing studio, the next challenge was to assess whether open throat had identifiable acoustic and percep-tual characteristics.

ACOUSTIC AND PERCEPTUAL VERIFICATION OF OPEN THROAT TECHNIQUE

The musical community has become increas-ingly fascinated with the link between acoustic measures of vocal quality and perceptual judg-ments of listeners. Prior to the work of Mitchell and Kenny (Mitchell and Kenny 2004a, 2004b; Kenny and Mitchell 2004, in press), two studies examined this link. Wapnick and Ekholm (Wap-

nick and Ekholm 1997) established 12 generally accepted perceptual criteria for the assessment of voice quality in classical singing (appropriate vi-brato, color/warmth, diction, dynamic range, ef-ficient breath management, evenness of registra-tion, flexibility, freedom throughout vocal range, intensity, intonation accuracy, legato line, and resonance/ring). Ekholm, Papagiannis and Cha-gnon (Ekholm, Papagiannis, and Chagnon 1998) used four of these criteria (“appropriate vibrato”, “resonance/ring”, “color/warmth”, and “clarity/focus”) and related them to objective measure-ments taken from acoustic analysis of the voice signal. Both studies required listeners to focus on specific vocal dimensions, as well as making an overall judgment of vocal quality. In both stud-ies, analysis of the listener rating scales of vocal quality revealed that the specific dimensions out-lined above were collinear and hence likely to be tapping into a single underlying construct, that of (overall) vocal quality, thereby rendering individ-ual assessments on each dimension at least partial-ly redundant. While focusing listeners by using a number of criteria may improve the consistency of judges’ responses, (Wapnick et al. 1993), studies have reported very high correlations between all dimension of voice quality studied (Ekholm, Pa-

Pedagogue 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 %Balanced/Coordinated □ □ □ □ □ □ □ □ □ □ □ □ □ □ 93Free □ □ □ □ □ □ □ □ □ □ 67Open □ □ □ □ □ □ □ □ 53Even/Consistent □ □ □ □ □ □ □ □ 53Warm □ □ □ □ □ □ □ □ 53With space □ □ □ □ □ □ □ □ 40Healthy/Safe □ □ □ □ □ 33Round □ □ □ □ □ 33Overtones/formants □ □ □ □ □ 33Easy/Flexible □ □ □ □ □ 33Clear □ □ □ □ 27Full □ □ □ 20Efficient □ □ □ 20With depth □ □ □ □ 20Clean □ □ □ 20Sexy/Juicy/Lusty □ □ 13Natural Voice □ □ 13Relaxed □ 7

Table 1: Sound qualities associated with open throat – pedagogues’ individual choices indicated by □.

21


pagiannis, and Chagnon 1998; Robison, Bounous, and Bailey 1994; Wapnick and Ekholm 1997). All of these dimensions were found to converge with the overall judgment of vocal quality. It may not be possible to separate individual features of good singing from the overall perception of a “good voice”.

Assessing acoustic and perceptual characteristics of open throat

Our next challenge was to determine whether this pedagogical technique could reliably produce desirable acoustic and perceptual changes in voic-es.

Six advanced female opera students, [3 sopranos, 3 mezzo-soprano] were asked to sing in three con-ditions: ‘optimal’ (O), using maximal open throat, ‘sub-optimal’ (SO), using reduced open throat and loud sub-optimal (LSO), which was the same as SO but with the additional instruction to sing as loudly as in O. Singers performed three musical tasks: messa di voce (a crescendo-diminuendo on a single note of long duration) on three pitches across their range, portions of an aria (Mozart: Ridente la Calma, K 152, bars 1-27) and a lied (Schubert: Du bist die Ruh D. 776 Op. 59, No. 3, bars 54 to 80) (Figure 1a and b). These were cho-sen as they require vocal skill and technical mas-tery (Miller 1996b) within the capacity of tertiary level students of opera (Ekholm, Papagiannis, and Chagnon 1998).

Singers’ voices were recorded to CD (Marantz CDR 630) using a high-quality microphone (AKG C-477) positioned on a head boom a constant 7 cm distance from the singer’s lips. This ensured we recorded only the voice energy, not the room reflections. We calibrated each recording in order to compare each singers’ recording with the others at the same SPL, heard as ‘loudness’. Recordings were anaysed with Soundswell (Hi-tech, Sweden) and Cool Edit software.

PERCEPTUAL DIFFERENCES OBSERVED IN OPEN THROAT

In this study, we explored whether particular sound qualities were associated with one vocal technique, open throat.

We asked singing pedagogues to judge the sing-ing of our sample of female classical singers as O (maximal open throat) and SO (reduced open throat) in 48 messa di voce and 30 song samples (including 6 repeated samples to test judges’ re-

sponse reliability). Fifteen expert singing peda-gogues made a forced choice decision (O or SO) on each sample they assessed. Correctly identified responses were counted by condition (O/SO), by judge and by singer.

The majority of listeners correctly recognised the use of open throat when it occurred. Collec-tively, judges were accurate in their identification of O (81.1% correct) and SO (91.0% correct) in messa di voce samples. In the individual song samples, listeners identified the use of open throat in 84% of O samples, and 69% of SO [Table 2 after (Mitchell and Kenny in press)]. They were more likely to make a correct identification in the Mozart task (85% correct) than in the Schubert task (68%). Listeners identified >83% of Mozart O and SO, and Schubert O. They were least reli-able in judging Schubert SO (53%).

The 15 judges identified the experimental condi-tion in 372 of 450 messa di voce samples (82.7%). Twelve of fifteen judges were moderately consist-ent in their judgments (k≥0.600); and three judges

&b8

3

Sop ‰ ‰

R

œ .œ

œ j

œ

J

œ .œœ

J

œ

Ù œœœ œœ œ œ œ œ

œ œ

œ œ

J

œ®

œœœ

&b

6

œ œ œ œ œœ

r

œ

r

œ

‰ ‰ ∑ ‰ ‰ ≈

R

œ .œœ

j

œ ≈

R

œ .œ

œ

j

œ≈R

œ

&b

12œ œœœ.

J

œ

R

œ

J

œ‰ ≈ R

œ .œ

œ

J

œ

j

œ

j

œ

j

œ≈ R

œ .œ

œ

J

œ

j

œ

j

œ

j

œ≈

R

œ

&b

18

J

œ

J

œ ≈

R

œ

J

œ

j

œ ≈

r

œœ œœœ.

j

œ

r

œ

j

œ

j

œ ≈

R

œ

J

œ

J

œ ≈

R

œ

J

œ

j

œ ≈

r

œ

&b

24

œ œœœ.

j

œ

rK

œ

œ

U

J

œ# œ œœ

œ œœ

œ œœœ

‰

≈

Ri - den - te la cal - ma nell' al - ma si_ des - ti, nell'

al - ma si_ desti; ne res - ti un seg - no di

sde-gno_e ti - mor. Ri - den - te la cal - ma nell' al - ma si des - ti; ne

res - ti più seg - no di sde-gno_e ti - mor, res - ti più seg - none

sde-gno_e ti - mor,

di

sde - gno_e ti - mor.di

&b

b

b 8

3

SopJ

œ .

J

œ

R

œœb

J

œ œb

J

œ œ

J

œ œ

J

œ

&b

b

b

6.œ

.œ

∑ œ œœ œ.œ œœ j

œ œ œœ œn œ œ

&b

b

b

12

œj

œ

∑ ∑

J

œ .

J

œ

R

œœb

J

œ

œb

J

œ

&b

b

b

18œ

J

œ œ

J

œ .œ.œn

∑ œ œœ œ.œ œ

&b

b

b

24œ j


.œ œ ‰

Dies Au - gen - zelt von dei - nem Glanz Al - lein er-

hellt, O füll es ganz, O füll es

ganz. Dies Au - gen - zelt von dei - nem

Glanz Al - lein er- hellt, O füll es

ganz, O füll es ganz!

Figure 1b: The Schubert Task, Du bist die Ruh D. 776 (Op. 59, No. 3), bars 54 to 80.

Figure 1a: The Mozart Task, Ridente la Calma, K 152, bars 1-27.

22


were inconsistent in their judgements. Listeners demonstrated reliability (fair to highly consistent) in their judgements through the duration of the song task, repeating their judgement of O or SO in an average of 86% of the six repeated samples in the perceptual test.

Conclusions

Listeners recognised a specific vocal quality in singers’ sound that they associated with the use of open throat technique. These findings suggest that there is a specific vocal quality in classical singing associated with the use of open throat technique, which is a perceptual reality to sing-ing pedagogues. Perceptual verification of a single vocal technique indicates that other pedagogical strategies can be assessed and evaluated making it feasible to track the methods by which good sing-ers are trained and which produce the most effec-tive outcomes in terms of vocal quality.

Singers rely on expert listeners’ judgments in auditions, competitions and examinations. Re-search indicates that listeners show some degree of reliability and consistency in their perceptual judgments of timbre, between vocal genres (from opera to music theatre) (Sundberg, Gramming, and Lovetri 1993), and between good and poor vocal and instrumental performance (Ekholm, Papagi-annis, and Chagnon 1998; Geringer and Madsen 1998; Saunders and Holahan 1997; Wapnick et al. 1993), in the assessment of excellence in overall voice quality (Wapnick and Ekholm 1997) and in rankings and ratings of performers in competi-

tive situations (Davidson and Da Costa Coimbra 2001).

ACOUSTIC CHARACTERISTICS OF OPEN THROAT TECHNIQUE

Having verified that open throat technique has reliable perceptual qualities, we then investigat-ed whether these were associated with consist-ent acoustic characteristics. A number of voice qualities [loudness, vibrato and long term average spectra (LTAS)] were assessed for their acoustic features. .

Loudness

Pedagogues suggested that the use of open throat resulted in ‘loudness’ (Mitchell et al. 2003). We used SO and LSO in experimental conditions to demonstrate that voice quality changes from O to SO were not simply the result of changes (i.e. re-ductions) to sound pressure levels (SPL) (Foulds-Elliott et al. 2000; Rossing, Sundberg, and Tern-strom 1986). A pilot test found that LSO did not produce a sufficiently discernible voice quality to be reliably distinguished from SO and was there-fore not included in the perceptual studies.

Vibrato

Vibrato is taken for granted as an intrinsic qual-ity of the classical singing voice and in pedagogi-cal literature, as a component of tone quality in coordination (Vennard 1968) richness (Seashore 1938; Vennard 1968) and vibrancy (Miller 1996b). Despite acoustical, physiological and perceptual studies (Vennard 1968; Prame 1994, 1997; Shipp, Sundberg, and Hadlund 1984; Sundberg 1995), it has been difficult to define its most desirable parameters. Consistent vibrato occurring within specified parameters, has been associated with a beautiful sound (Robison, Bounous, and Bailey 1994) and to listeners’ overall preference (Ekholm, Papagiannis, and Chagnon 1998). In the classical singing literature, a steady and even vibrato is universally promoted (Miller 1996b) while poor vibrato is considered indicative of poor technique (Vennard 1968; Miller 1996b) and inferior sound quality. In classical singing, sound without vibra-to, or straight-tone, has been described as dull or spread (Vennard 1968) and lacking freedom, pow-er (Miller 1996b) and ring (Vennard 1968). In fact, delay in vibrato onset is argued to be indicative of a faulty technique and an unnatural voice quality (Miller 1996b).

Few studies compare the same singers’ vibrato in

Task Hit % Miss %

All O 84% 16%

All SO 69% 31%

Mozart O 86% 14%

Mozart SO 84% 16%

Schubert O 83% 17%

Schubert SO 53% 47%

Table 2: Song responses. Percentages of correct (hit) and incorrect (miss) responses to the song task, by condition (optimal and sub-optimal) and by condition and task (Mozart and Schubert).

23


different tasks. When they do, they find it changes across musical styles (Easley 1932; Hakes, Shipp, and Doherty 1987), as a result of emotion (Sund-berg 1997; Rothman and Arroyo 1987; Gabriels-son and Juslin 1996), drama (Sundberg, Iwarsson, and Hagegard 1995) or indeed variations in ‘loud-ness’ (SPL) (Titze et al. 1999) or dynamics (Prame 1994). Our study is the first to assess the impact of a specific singing technique i.e. open throat on vibrato. As pedagogues associated evenness and consistency with the technique, we hypothesized that frequency modulations associated with vi-brato rate, extent and onset would vary outside acceptable or desirable parameters for SO and LSO compared to O, that is rate (VR) would be less consistent, extent (VE) would be reduced and onset (VO) would be delayed.

Vibrato results

We demonstrated reliable differences in vibrato parameters as a result of varying the degree to which singers applied the technique of open throat. Figure 2 [after (Mitchell and Kenny 2004b)] illus-trates the changes to vibrato parameters observed in spectrographs of the three experimental condi-tions. Hypotheses were confirmed for vibrato ex-tent and onset, that is, a reduction of open throat technique for these singers produced a significant decrease in VE in SO/LSO and a significant in-crease in VO in both SO/LSO. There was no statis-tically significant change for VR. However, visual inspection of the spectrographs (Figure 2) shows that reduction of open throat in the SO and LSO conditions was associated with greater irregularity of the vibrato pattern compared to O. There were no significant differences between SO and LSO on any of the vibrato parameters.

As vibrato is considered a key indicator of good singing, these findings suggest that open throat is important to the production of a good sound in classical singing. Since vibrato parameters largely define good singing technique in the literature, open throat would appear to be an essential ele-ment of sound vocal pedagogy. Inappropriate vibrato is indicative of poor singing in general; therefore further acoustic tests were required to test the differences in timbre.

Long-term average spectra (LTAS)

LTAS is widely used to represent singers’ sound (Sundberg 1974; Borch and Sundberg 2002) and its different vocal qualities based on energy changes that occur during different vocal tasks. Early re-searchers in this field have used LTAS to develop

exemplars of voice types (Sundberg 1974) or sing-ing genres (Cleveland, Sundberg, and Stone 2001; Borch and Sundberg 2002). LTAS curves have also been used to differentiate male and female voices (Mendoza et al. 1996), to note differences between singing and speaking voices (Barrichelo et al. 2001), solo and choral voice (Rossing, Sun-dberg, and Ternstrom 1987) and pop or country from opera singers (Borch and Sundberg 2002; Cleveland, Sundberg, and Stone 2001). Some LTAS have become exemplars of particular vocal qualities and of voices of quality. In classical or operatic voice, for example, an LTAS presents an energy boost around 3 kHz and this has been linked to carrying power over an orchestra or audibility in an opera theatre (Sundberg 1974; Barnes et al. 2004; Thorpe et al. 2001; Vurma and Ross 2000).

In research, we perform two simple measure-ments on LTAS. Figure 3 illustrates how the ra-tio measurements of singing power ratio (SPR) (Omori et al. 1996) and energy ratio (ER) (Thorpe et al. 2001) are derived from the LTAS curve. From the LTAS plots, the highest peaks in the 0-2 kHz and 2-4 kHz regions were labeled P1 and

0

2000

4000

6000

1 54320

0

2000

4000

6000

1 54320

1 543200

2000

4000

6000

Freq

uen

cy(H

z)

Time (seconds)

Time (seconds)

Time (seconds)

Freq

uen

cy(H

z)Fr

equ

ency

(Hz)

a

c

b

Figure 2 a-c: Examples of vibrato from subject 3 singing ‘Von deinem Glanz’ in each of the three conditions, optimal (O), sub-optimal (SO), and loud sub-optimal (LSO).

24


P2 respectively and the areas below the peaks, A1 and A2. The SPR (Omori et al. 1996) is the differ-ence between the energy peaks P1 and P2 while the ER (Thorpe et al. 2001) is the difference in energy area between 0-2 kHz and 2-4 kHz. A low ER or SPR represents a greater reinforcement in the 2-4 kHz region and better balance between the spectral energy 0-2 and 2-4 kHz. When there is less reinforcement in the 2-4 kHz region, ER re-sults follow SPR.

As inconsistent vibrato is considered indicative of poor singing, it was hypothesized that testing the energy distribution in our singers’ voices in each condition would identify the timbral changes associated with open throat. Hypotheses were gen-erated regarding the LTAS plots in O compared to SO: that peaks P1 (0-2kHz) and P2 (2-4kHz) would be reduced, that overall LTAS shape would demonstrate smaller or multiple energy peaks above 2 kHz in SO/LSO compared to O. Differ-ences between spectral peak height (SPR: singing power ratio,) and spectral area (ER: energy ratio) between the 0-2 and 2-4 kHz frequency ranges were performed to assess the effect of open throat on carrying power in the voices.

LTAS results

Visual inspection of long term average spec-tra (LTAS) confirmed differences between O and SO/LSO, and the O condition produced a rounder peak between 0-2 kHz indicating a warmer sound quality compared to SO (Figure 3). Despite these

findings, there was no significant difference in measurements of SPR (peak height) (Omori et al. 1996) or ER (peak area) (Thorpe et al. 2001) be-tween O and SO/LSO. There were however, sig-nificant differences between SO and LSO for P2, SPR and ER but hypotheses were not confirmed for O. These findings did not accord with differ-ences in vibrato extent and onset between O and SO/LSO.

As these results were not consistent with the vi-brato findings they suggest that while LTAS pro-vides information on energy distribution, meas-urements performed on the LTAS were unable to differentiate between experimental conditions, whereas the human ear produced the most reliable assessment of vocal quality. Plotting the differenc-es between O and SO/LSO pairs of LTAS clearly indicates the areas of spectral change (Figure 4). This appears to be the most sensitive measure of energy distribution differences between condi-tions, but we have no way of capturing and quan-tifying such data at this time.

DO ACOUSTIC MEASURES MATCH PERCEPTUAL JUDGMENTS?

The conflicting results from the perceptual and acoustic studies of open throat technique high-lighted potential problems with these conven-tional analyses as measures of vocal quality. In the final study, we compared perceptual ratings of vo-cal quality with acoustic measures performed on

Figure 3: Long term average spectra of singer 3 singing Mozart’s Ridente la calma in three experimental con-ditions, optimal, sub-optimal and loud sub-optimal (O, SO, LSO). Singing power ratio (SPR) is the difference between the spectral peaks (P1 and P2) and energy ratio (ER) is the difference between the area under the peaks (A1 and A2).

30

50

70

90

110

0 2000 4000 6000Frequency (Hz)

SPL(dBat7cm

) Optimal

Sub-optimal

Loud sub-optimal

= SPR

P1 P2

A2

A1

25


LTAS to assess whether SPR and ER were suffi-ciently sensitive to evaluate LTAS or vocal timbre and whether acoustic measures can predict a voice of quality?

Characteristic LTAS shapes, particularly of clas-sical or operatic voices of performers of the high-est regard have been accepted as an accurate visu-al representation of good sound. Operatic singing is positively associated with an increase in energy between 2-4 kHz (Omori et al. 1996; Thorpe et al. 2001; Vurma and Ross 2000; Barnes et al. 2004). While measures of SPR (Omori et al. 1996) and ER (Thorpe et al. 2001) are not intended to give information about singer’s formant, they should illustrate differences in energy distribution, or car-rying power. However, Vurma & Ross (Vurma and Ross 2000) found that increased energy above 2 kHz did not necessarily represent vocal quality judged perceptually.

To test the value of acoustic measures, we matched pedagogues’ perceptual ratings to acous-tic measures of each perceptual sample. Perceptual scores, SPR and ER were rank ordered for overall quality as defined in the singing literature (Omori et al. 1996; Thorpe et al. 2001; Barnes et al. 2004;

Vurma and Ross 2000). We then compared per-ceptual rankings with rankings of acoustic meas-ures (SPR and ER) to assess whether the acoustic characteristics matched the perceptual judgments of overall timbre.

Table 3 [after (Kenny and Mitchell 2004)] presents the rankings of each singer and the re-spective musical task and experimental condition, ordered by perceptual rank from highest to lowest. While we found the expected significant relation-ship between SPR and ER, there was no relation-ship between perceptual rankings of vocal quality of singers based on SPR or ER. LTAS measures were not consistent with perceptual ratings of vo-cal quality, and could not therefore be used to de-fine a voice of quality in our studies.

Measurements of comparative energy (SPR and ER) were inconclusive indicators of vocal qual-ity. A science of the singing voice cannot progress without addressing the problem inherent in accept-ing long-term average spectra as analogues for vo-cal quality without providing a link between per-ceptual cues and a visual representation of ‘quality in singing. This presents a major challenge to the current wisdom that acoustic parameters of voice

-20

0

20

40

60

80

100

0 1000 2000 3000 4000 5000

Optimal

Energydifference

Frequency (Hz)

SPL(dBat7cm

)

6000

Sub-optimal

Figure 4: Example of singer 1’s LTAS plots in the optimal and sub-optimal condition for the Mozart task. Below is an example of the calculation of energy difference between the optimal and sub-optimal conditions.

26


must emulate the established acoustic norms in or-der to achieve overall vocal excellence.

In Figure 5a-d, we compared exemplars of LTAS sampled from voices ranked high, middle or low on the perceptual rating scale. LTAS of the high-est-ranking singers [Figure 5a-b after (Kenny and Mitchell in press)] showed an increase in energy between 2-4 kHz, whereas LTAS of middle and low ranked singers (Figure 5c-d) lacked the uni-fied peak of energy increase above 2 kHz. Previ-ous similar findings (Borch and Sundberg 2002; Miller 1998; Vurma and Ross 2000) have inter-preted these visual cues in singers’ LTAS as a good sound. However, as third equal ranking singers 1 and 3, singing in the O condition produced this

unified energy peak > 2 kHz (the most ‘masculine’ LTAS shape) they should have had the ‘best’ voic-es perceptually. However, their LTAS were differ-ent to the highest-ranking singer (singer 5), who had a wider distribution of energy above 2 kHz. The SO plots of the lower ranked singers (Fig 3d) showed a spectral roll-off more consistent with plots for speech than singing.

CONCLUSIONS AND FUTURE DIRECTIONS

In this series of studies, we have demonstrated that open throat is a technical and perceptual re-ality to singers and singing pedagogues and pro-duces a specific vocal quality in classical singing that can be reliably identified by expert listeners. Through qualitative, acoustic and perceptual stud-ies, we have defined the term ‘open throat’ as a technique, an action and a sound quality.

While we know that singing improves over time and through training specific components of the training that produce improvements in vocal qual-ity have not hitherto been isolated (Ekholm, Papa-giannis, and Chagnon 1998; Mendes et al. 2003; Robison, Bounous, and Bailey 1994; Vurma and Ross 2000; Wapnick and Ekholm 1997). This series of studies presents a first attempt to study a single technique, to identify its perceptual and acoustic characteristics, and its effect on vibrato.

Vocal quality results from a complex combina-tion of acoustical parameters. To date, no single objective evaluation captures or characterizes vocal quality in a systematic way (Omori et al. 1996; Thorpe et al. 2001). We have demonstrated that acoustic analyses such as LTAS do not reli-ably match perceptual judgments by expert lis-teners and therefore cannot be used to define or predict vocal quality. We recommend that any fu-ture acoustic analyses or visual representations of voice must emulate the human ear. A new genera-tion of acoustic recording equipment and analysis software is emerging that has the potential to pro-vide much finer acoustic and psychoacoustic rep-resentations of vocal quality. Voice research will benefit from other areas of acoustic research (e.g. room acoustical quality (Nannariello, Osman, and Fricke 2002) and audio systems (Zwicker and Fastl 1999)) oriented to listener sensation and will use binaural measurements (using a dummy head microphone) to virtually place the listening sub-ject in the same acoustical environment as in the original recording(Møller et al. 1996). A singing quality model will be more appropriately derived

Table 3: Singer, task, condition and rankings for perceptual score, singing power ratio (SPR) and energy ratio (ER), sorted by perceptual score rank-ing from highest to lowest.

Perceptual Rank

SPR Rank

ER Rank Singer Task Condi-

tion

1 16 9 5 S O2 13 3 5 M O3 10 10 1 M O4 5 2 3 M O5 9 11 1 S O6 21 20 4 S O7 6 13 4 M O8 23 23 2 S O9 19 19 3 S O10 8 6 2 M O11 11 14 6 S O12 4 4 5 S SO13 3 12 1 M SO14 24 24 2 S SO15 15 8 6 M O16 7 7 5 M SO17 20 21 3 S SO18 18 18 6 S SO19 12 16 1 S SO20 1 5 3 M SO21 22 22 4 S SO22 17 15 2 M SO23 14 17 4 M SO24 2 1 6 M SO

27


by representing listener experience.

REFERENCES

Barnes, J.J., P.J. Davis, J. Oates, and J. Chapman. 2004. The relationship between professional operatic so-prano voice and high range spectral energy. J Acoust Soc Am 116 (1):530–538.

Barrichelo, V.M.O., R.J. Heuer, C.M. Dean, and R.T. Sataloff. 2001. Comparison of singer’s formant, speaker’s ring and LTA spectrum among classical singers and untrained normal speakers. Journal of Voice 15 (3):344-350.

Borch, D Z, and J Sundberg. 2002. Spectral distribu-tion of solo voice and accompaniment in pop music. Logopedics Phoniatrics Vocology 27:37-41.

Burgin, John Carroll. 1973. Teaching singing. Metuchen, N.J.: Scarecrow Press.

Callaghan, Jean. 2000. Singing and voice science. San Diego, California: Singular Pub. Group.

Cleveland, T. F., J Sundberg, and R.E. Stone. 2001.

Long-Term-Average-Spectrum characteristics of country singers during speaking and singing. Journal of Voice 15 (1):54-60.

Davidson, J.W., and D Da Costa Coimbra. 2001. In-vestigating performance evaluation by assessors of singers in a music college setting. Musicae Scientiae 5 (1):33-53.

Easley, E. 1932. A comparison of the vibrato in concert and opera singing. In The vibrato, edited by C. E. Seashore. Iowa: University of Iowa Press.

Ekholm, E., G. C. Papagiannis, and F. P. Chagnon. 1998. Relating objective measurements to expert evaluation of voice quality in Western classical sing-ing: critical perceptual parameters. Journal of Voice 12 (2):182-96.

Fields, V.A. 1947. Training the singing voice: an anal-ysis of the working concepts contained in recent contributions to vocal pedagogy. New York: King’s Crown Press.

Foulds-Elliott, SD., CW. Thorpe, SJ. Cala, and PJ. Davis. 2000. Respiratory function in operatic sing-ing: effects of emotional connection. Logopedics

20

40

60

80

100

120

0 1000 2000 3000 4000 5000 6000Hz

Singer 5 Schubert OSinger 5 Mozart O

SPL(dBat7cm

)

20

40

60

80

100

120

0 1000 2000 3000 4000 5000 6000Hz

SPL(dBat7cm

)

Singer 1 Mozart OSinger 3 Mozart O

Singer 1 Schubert O

20

40

60

80

100

120

0 1000 2000 3000 4000 5000 6000Hz

SPL(dBat7cm

)

Singer 6 Schubert O

Singer 5 Schubert SO

Singer 1 Mozart SO


20

40

60

80

100

120

0 1000 2000 3000 4000 5000 6000Hz

SPL(dBat7cm

)

Singer 2 Mozart SO

Singer 4 Mozart SO

Singer 6 Mozart SO

a b

c d

Figure 5: LTAS of highest (a), 3= (b), middle (c) and lowest (d) perceptually ranked singers. Legends correspond to task, Mozart (M) or Schubert (S), and singer number.

28


Phoniatrics Vocology 25 (4):151-68.Gabrielsson, A., and P.N. Juslin. 1996. Emotional Ex-

pression in Music Performance: Between the Per-former’s Intention and the Listener’s Experience. Psychology of Music 24:68-91.

Geringer, J.M., and C.K. Madsen. 1998. Musicians’ ratings of good versus bad vocal and string perform-ances. Journal of Research in Music Education 46 (4):522-534.

Hakes, J., T. Shipp, and E. T. Doherty. 1987. Acous-tic properties of straight tone, vibrato, trill and trillo. Journal of Voice 1 (2):148-156.

Hemsley, Thomas. 1998. Singing and imagination: a human approach to a great musical tradition. Oxford; New York: Oxford University Press.

Herbert-Caesari, Edgar F. 1951. The voice of the mind. London: Robert Hale.

Joiner, James Richard. 1998. Charles Amable Battaille: pioneer in vocal science and the teaching of singing. Lanham, Md.: Scarecrow Press.

Kenny, D.T., and H.F. Mitchell. 2004. Visual and audi-tory perception of vocal beauty: conflict or concur-rence? Paper read at 8th International Conference on Music Perception & Cognition (ICMPC8), August, at Evanston, IL, USA.

———. in press. Acoustic and perceptual appraisal of vocal gestures in the female classical voice. Journal of Voice.

Manèn, Lucie. 1987. Bel canto: the teaching of the clas-sical Italian song-schools its decline and restoration. Oxford: Oxford University Press.

Marafioti, P. Mario. 1981. Caruso’s Method of Voice Production. New York: Dover Publications. Original edition, 1922.

McKinney, James C. 1982. The Diagnosis and Correc-tion of Vocal Faults. Nashville, Tennessee: Broad-man Press.

Mendes, A.P., H. B. Rothman, C. M. Sapienza, and W. S. Brown, Jr. 2003. Effects of vocal training on the acoustic parameters of the singing voice. Journal of Voice 17 (4):529-543.

Mendoza, E, N Valencia, J Munoz, and H Trujillo. 1996. Difference in voice quality between men and women: use of the long-term average spectrum (LTAS). Jour-nal of Voice 10 (1):59-66.

Miller, R. 1998. The singing teacher in the age of voice science. In Vocal Health and Pedagogy, edited by R. T. Sataloff. San Diego, CA: Singular.

Miller, R. 1996a. Sotto Voce: Vocal Timbre in Piano Dynamic. Journal of Singing - The Official Journal of the National Association of Teachers of Singing 52 (5):23-24, 52.

Miller, Richard. 1996b. The structure of singing: sys-tem and art in vocal technique. New York: London: Schirmer Books.

———. 1997a. National schools of singing: English, French, German, and Italian techniques of singing re-visited. Rev. ed. Lanham, Md: Scarecrow Press.

———. 1997b. The Open Throat (La gola aperta). In On the Art of Singing, edited by R. Miller. New York: Oxford University Press.

Mitchell, H.F., and D.T. Kenny. 2004a. The effects of open throat technique on long term average spectra (LTAS) of female classical voices. Logopedics Pho-niatrics Vocology 29 (3):99-118.

———. 2004b. The impact of “open throat” technique on vibrato rate, extent and onset in classical singing. Logopedics Phoniatrics Vocology 29 (4):171-182.

———. in press. Can experts identify “open throat” technique as a perceptual phenomenon? Musicae Scientiae.

Mitchell, H.F., D.T. Kenny, M. Ryan, and P.J. Davis. 2003. Defining open throat through content analysis of experts’ pedagogical practices. Logopedics Phoni-atrics Vocology 28 (4):167-180.

Møller, H., M.F. M.F. Sørensen, C.B. Jensen, and D. Hammershøi. 1996. Binaural technique: do we need individual recordings? J. Audio Eng. Soc. 44:451-469.

Monahan, Brent Jeffrey. 1978. The art of singing: a compendium of thoughts on singing published be-tween 1777 and 1927. Metuchen, N.J.: Scarecrow Press.

Nair, Garyth. 1999. Voice tradition and technology: a state-of-the-art studio. San Diego: Singular Publish-ing Group.

Nannariello, J, M.R. Osman, and F.R. Fricke. 2002. Re-cent developments in the application of neural net-work analysis to architectural and building acoustics. Acoustics Australia 29:103-110.

Omori, K., A. Kacker, L. M. Carroll, W. D. Riley, and S. M. Blaugrund. 1996. Singing power ratio: quanti-tative evaluation of singing voice quality. Journal of Voice 10 (3):228-35.

Prame, E. 1994. Measurement of the vibrato rate of ten singers. J Acoust Soc Am 96 (4):1979-1984.

———. 1997. Vibrato extent and intonation in profes-sional Western lyric singing. J Acoust Soc Am 102 (1):616-621.

Puritz, Elizabeth. 1956. The teaching of Elisabeth Schumann. London: Methuen.

Reid, Cornelius L. 1975. Voice: psyche and soma. New York: Joseph Patelson Music House.

———. 1983. A dictionary of vocal terminology. New York: Joseph Patelson Music House.

Robison, C.W., B Bounous, and R Bailey. 1994. Vocal Beauty: A study proposing its acoustical definition and relevant causes in classical baritones and female belt singers. Journal of Singing 51:19-30.

Rossing, T.D., J Sundberg, and S. Ternstrom. 1986.

29


Acoustic comparison of voice use in solo and choir singing. J Acoust Soc Am 79 (6):1975-1981.

———. 1987. Acoustic comparison of soprano solo and choir singing. J Acoust Soc Am 82 (3):830-836.

Rothman, H. B., and A. Antonio Arroyo. 1987. Acoustic variability in vibrato and its perceptual significance. Journal of Voice 1 (2):123-141.

Saunders, T.C., and J.M. Holahan. 1997. Criteria-spe-cific rating scales in the evaluation of High School instrumental performance. Journal of Research in Music Education 45 (2):259-272.

Seashore, Carl E. 1938. Psychology of music. New York and London: McGraw-Hill Book Company.

Shipp, T., J. Sundberg, and S. Hadlund. 1984. A model of frequency vibrato. Paper read at Transcripts of the Thirteenth Symposium: Care of the Professional Voice, at New York.

Sundberg, J. 1977. The acoustics of the singing voice. Scientific American (March):82-91.

Sundberg, J. 1974. Articulatory interpretation of the “singing formant”. J Acoust Soc Am 55 (4):838-44.

———. 1995. Acoustic and psychoacoustic aspects of vocal vibrato. In Vibrato, edited by P. H. Dejonckere, M. Hirano and J. Sundberg. San Diego: Singular Publishing Group.

———. 1997. Expressivity in singing. TMH-QPSR 2 (3):13-19.

Sundberg, J., P. Gramming, and J. Lovetri. 1993. Com-parisons of pharynx, source, formant, and pressure characteristics in operatic and musical theatre sing-ing. Journal of Voice 7 (4):301-10.

Sundberg, J., J. Iwarsson, and H. Hagegard. 1995. A

singer’s expression of emotions in sung performance. In Vocal Fold Physiology: Voice Quality Control, ed-ited by O. Fujimura and M. Hirano. San Diego, Cali-fornia: Singular Publishing Group, Inc.

Thorpe, CW., SJ. Cala, J. Chapman, and PJ Davis. 2001. Patterns of breath support in projection of the singing voice. Journal of Voice 15 (1):86-104.

Thurman, L, and G. F. Welch, eds. 2000. Bodymind & voice: foundations of voice education. Rev. ed. ed. Collegevile, Minn: VoiceCare Network.

Titze, I. R., R. Long, G. I. Shirley, E. Stathopoulos, L. O. Ramig, L. M. Carroll, and W. D. Riley. 1999. Messa di voce: an investigation of the symmetry of crescendo and decrescendo in a singing exercise. J Acoust Soc Am 105 (5):2933-40.

Vennard, William. 1968. Singing: the mechanism and the technic. 5th ed. New York: Fischer.

Vurma, A., and Y. Ross. 2000. Priorities in voice train-ing: carrying power or tone quality. Musicae Scien-tiae 4 (1):75-93.

Wapnick, J., and E. Ekholm. 1997. Expert consensus in solo voice performance evaluation. Journal of Voice 11 (4):429-36.

Wapnick, J., P Flowers, M Alegant, and L Jasinskas. 1993. Consistency in piano performance evaluation. Journal of Research in Music Education 41 (4):282-292.

Ware, Clifton. 1998. Basics of vocal pedagogy: the foundations and process of singing. New York: Mc-Graw-Hill.

Zwicker, E, and H Fastl. 1999. Psychoacoustics: facts and models. Berlin: Springer.

30

2 Defining open throat

Pedagogical literature advocates the use of open throat in singing. Historically, the term

open throat can be traced throughout vocal pedagogy literature (Burgin, 1973; Fields,

1947; Monahan, 1978) as an interpretation of freedom or lack of tension in the area of

the throat, resulting in a lack of constriction and a better tone quality. Current support

for the use of open throat in singing technique is widespread (Miller, 1996; Reid, 1975,

1983). Open throat is associated with techniques for breath and as factor of overall

resonance (Hemsley, 1998; Miller, 1996) and as an enhancement to vocal quality

(Miller, 1996; Reid, 1983; Vennard, 1968).

Terminology depicting space or openness in singing is familiar in both the singing

literature and in the singing studio. However, the definition and interpretation of the

instructions to elicit space or openness vary across singing pedagogues. In singing

pedagogy texts, there is diverse opinion as to the need and indeed success of the use of

open throat technique. The technique is used to correct constriction, necessary for the

production of certain vocal qualities and genres, as an enhancement to the sound or a

specific sound quality. Actions or gestures such as “yawn”, “sob”, “laugh” and “surprise

breath” are frequently found in the literature and may represent open throat technique.

Open throat has been associated with certain schools of singing, as a sound quality, in

certain national schools of singing throughout the literature. Different nationalistic

schools value vocal resonance for different sound qualities (Miller, 1997): the German

school favours conscious alteration of the pharyngeal size, different to speaking;

31

whereas the Italian school advocates less manipulation and a more natural sensation of

openness.

In the first study, we assessed the degree of consensus amongst singing pedagogues

regarding the definition of, and use in the singing studio of the technique called open

throat. Open throat was defined in other terms, such as “freedom”, “space”, “collar” or

“retraction”. Results indicated that all fifteen pedagogues described open throat

technique as fundamental to singing training and were positive about the sound quality

it achieved, especially in classical singing. Its use in singing has been characterised with

a sound quality which is free, warm and open. Open throat has been defined as a

technique to maximize pharyngeal space and/or abduct the ventricular folds.

This pedagogical approach has been well received in the scientific literature. Professor

David Howard, Editor of Logopedics Phoniatrics Vocology, noted in his editorial

review of our paper that our innovative study on open throat technique sat comfortably

with other international studies that examined physiological explanations for vocal

quality and acoustic synthesis of yawn and twang technique by Titze’s American

research group (Howard, 2003; Titze, Bergan, Hunter, & Story, 2003).

The results of the study were reported to the singing pedagogy community at the

Australian National Association of Teachers of Singing (ANATS) National Conference,

October 2002, Melbourne [Mitchell, H. F., Kenny, D. T., Ryan, M. Open throat:

pedagogical perceptions and practices].

32

PAPER 1

Mitchell, H. F., Kenny, D. T., Ryan, M., & Davis, P. J. (2003). Defining open throat through content analysis of experts' pedagogical practices.

Logopedics Phoniatrics Vocology, 28(4), 167-180

33

Defining ‘open throat’ through content analysis ofexperts’ pedagogical practices

Helen F. Mitchell, Dianna T. Kenny, Maree Ryan and Pamela J. Davis

From the Australian Centre for Applied Research in Music Performance (ACARMP), Sydney Conservatorium of Music,The University of Sydney, New South Wales 2006, Australia

Received 21 October 2002. Accepted 4 September 2003.

Logoped Phoniatr Vocol 2003; 28: 167�/180

‘Open throat’ is a term regularly used in the singing studio, but agreement across pedagogues as to its definition and functionhas not yet been assessed. Fifteen expert singing pedagogues participated in a qualitative study involving a semi-structuredinterview to explore current thinking regarding terminology, pedagogy, sound quality and the perceived physiology to achieveopen throat, as used in the singing studio. Most teachers included the use of the technique as a fundamental in singingtraining, and were positive about the sound quality it achieved, especially in classical singing. The purpose of the techniquewas described as a way of maximizing pharyngeal space and/or achieving abduction of the ventricular folds.

Key words: associated physiology, singing pedagogy, sound quality descriptors, terminology.

Helen F. Mitchell, Australian Centre for Applied Research in Music Performance (ACARMP), The Conservatorium of Music,University of Sydney, NSW 2006, Australia. Tel.: �/61 2 9351 1386. Fax: �/61 2 9351 9540. E-mail: [email protected]

INTRODUCTION

Historically, teaching and evaluating singing has been

guided by an oral tradition in which pedagogical

techniques are handed down from one generation of

singing teachers to the next. The oral tradition, with

its accompanying musical terminology, particularly

that used to describe performance quality, can be

confusing and indeed mystifying for those who use

and practise it (1).

Many of the terms used to describe ‘openness’,

‘space’ or ‘freedom’ as a characteristic of the singing

voice have been used inconsistently, and terminology

associated with sound quality among expert singing

pedagogues is diverse and idiosyncratic. Seashore

recommended that ‘musicians scrap a mass of the

current synonyms for tone quality, because these

words do not connote any demonstrable differences

in content. The diversity of words simply adds to the

confusion’ (2, p. 111). There is, and has always been a

recognition of an overall good sound and of its

components (3), but in this case of openness there is

need for clarification, especially in the area of profes-

sional classical singing.

Poor singing has been described as sound ‘getting

stuck’ in the throat (4). Awareness of the throat as a

technique in good singing, or the lack of awareness as

an explanation for bad singing, has appeared through-

out the singing pedagogy literature. It is believed that

sound should flow freely, with nothing impeding it,

like a ‘chimney’ for air (5), or an unobstructed passage

for a sound beam (6). Failure to achieve this free flow

has been perceived to have serious consequences for

the singer and the sound. ‘If the scholar should have

any defects, of the nose, the throat, or of the ear, let

him never sing but when the master is by, or somebody

that understands the profession in order to correct

him, otherwise he will get an ill habit, past all remedy’

(7, p. 187).

Concepts relating to open throat can be traced

throughout pedagogical and scientific singing litera-

ture. Vennard defines open throat as the ‘condition

agreed upon by most voice teachers as desirable for

resonance’ (8, p. 252). That is, the use of open throat

makes a fundamental difference to vocal quality. Titze

(9) and Titze and Story (10) described a ‘wide

pharynx’ as an acoustic enhancement to the first

formant and to the overall sound. The difference to

the sound has been described as rounder, larger (11),

free (12), pure (13), rich, warm (14), and concentrated

energy (15).

�ORIGINAL ARTICLE �

# 2003 Taylor & Francis. ISSN 1401-5439 Logoped Phoniatr Vocol 28

DOI: 10.1080/14015430310018856 34

At least two views of open throat have been

propagated in the pedagogical literature: the first

focuses on the action at laryngeal level (16), and the

other emphasizes pharyngeal involvement, at a level

extending to the soft palate or velum (6, 8). These

differing perspectives regarding the vocal mechanism

required to achieve open throat have inevitably

resulted in different teaching strategies and terminol-

ogy. Examples include actions involved in inhalation,

the surprise breath and smelling the rose (5, 17, 18).

Visualization of space within the throat, through an

‘air ball’ or ‘soap bubble’ (6, 19), and various other

configurations to alter the shape of the pharynx are

common throughout the literature. The lowered lar-

ynx, often along with widening of the oropharynx, is

also associated with open throat (6, 20, 21). Yawn is

both advocated (15, 21�/3), and also seen as poten-

tially distorting to the sound if extended too far by

creating tension and imbalance (24, p. 81). Forms of

laughing as a technique to achieve an open throat were

advocated in the late twentieth century (16, 24). Estill

(16) and Citardi et al . (25) advocate a method

involving retraction of the false vocal folds or abduc-

tion of the ventricular folds away from the midline.

Current pedagogical thought and teaching recom-

mend the use of more than one term to define the

relative openness of the pharynx and adjacent struc-

tures. Open throat and retraction are recognized in the

literature (16, 24) as well as in teaching, and a wealth

of other equivalents exist, such as throat widening,

space, or space at the back. From this point on, open

throat will be used as a generic term for all terminol-

ogy and technique descriptors.

Wapnick and Ekholm (26) advocated consistency in

the use of common terminology across singing peda-

gogues. They initiated links in the discovery and

application of terminology, perceptual judgment and

acoustic evaluation. Ekholm et al . (27) found striking

connections were made across judges when one sound

quality appeared dependent on another [resonance/

ring and clarity/focus (.94) and colour/warmth with

appropriate vibrato (.93)]. The methodology in this

study builds on these works by linking the study of a

technique with its consequent sound quality, to

ascertain the relationship between the desired sound

quality and the technique used to achieve it. This

study sought to determine agreement, links and

interdependency of descriptors on technique and

perceptual qualities in describing technique in singing.The focus of the study is on terminology and

techniques of teaching open throat as practised by

experienced singing pedagogues. In particular, we were

interested in the perceived role of open throat in

classical singing, and whether open throat was thought

to improve the overall sound quality.

METHODS

Participants

Participants were 15 experienced singing teachers with

established singing studios. All participants were

known to the researchers via affiliations with keymusic centres in Australia and were personally invited

to participate by one of the authors. Selection criteria

for inclusion in the study were based on reputation

within the singing profession, determined by renown

or prominence according to the calibre of the students

in the singing studio. Specifically, the participants:

1. taught singing at tertiary level in universities or

conservatoria;

2. had a successful private studio, defined as one

comprising of students who performed and audi-tioned at a big city or national level for positions

in opera companies, music theatre or professional

musical ensembles;

3. worked primarily in Australia or the United

Kingdom;

4. represented the range of current singing pedago-

gies in their teaching methods (investigators aimed

specifically to include both exponents and criticsof the Estill (16) method);

5. have had, or continue to have, a successful career

as a professional solo singer.

Participants were sent information about the project

and were invited to undertake a taped, semi-structured

interview regarding current vocal pedagogy, in parti-

cular, the terminology used by singing pedagogues to

explain singing techniques to their students. Ethical

approval to conduct the study was obtained from The

University of Sydney Ethics Committee. All the inter-views were conducted by the first author, and only the

first author was aware of the identity of participants.

Instruments

Questionnaire. Each participant completed a brief

questionnaire giving details of their teaching

experience and current singing students. The ques-

tionnaire was divided into three sections: demographicinformation, influences on singing pedagogy and

characteristics of their singing studio. Participants

were asked for information related to: age, profes-

sional education, number of years of teaching experi-

ence and proportion of time spent teaching or singing

professionally. They were also asked to name people

with whom they had studied, and workshops or

masterclasses they had attended that had influencedtheir teaching. Finally, pedagogues were asked to

classify themselves according to the level at which

168 H. F. Mitchell et al.

Logoped Phoniatr Vocol 2835

their students performed. Pedagogues were asked

to use the Bunch and Chapman taxonomy (28) to

estimate the percentage of their students in eachcategory (superstar, international, national/big city,

regional/touring, local community, singing teachers,

full-time students of singing, amateur, child) and in

the primary genres (opera, contemporary music thea-

tre, musical theatre, concert/oratorio/recital).

Semi-structured interview. A semi-structured inter-

view was developed to assess the beliefs and attitudes

of singing teachers related to open throat (29). Thismethod of data collection is described by Smith (30).

It was considered appropriate to the goals of the

current study because it ‘facilitates rapport/empathy,

allows a greater flexibility of coverage, enables the

interview to enter novel areas, and tends to produce

richer data’ (30, p. 12).

The interview schedule opened with general ques-

tions about the teaching of singing, for example, ‘Howdo you achieve a good sound in singing?’ and ‘How do

you correct a problematic sound?’ No direct questions

on open throat were asked but the topic was discussed

in depth once the pedagogue initiated discussion.

Reference to a key term or to an associated term,

such as open throat, throat widening, retraction or

space at the back, provided the impetus to begin a

discussion of open throat. If the pedagogue did notmention open throat during the interview, the inter-

viewer asked the question, ‘What do terms like open

throat, throat widening, retraction and space at the

back mean to you?’

The interview schedule focused on elucidating

pedagogues’ view of open throat, using categories

and terminology derived from the literature. These

were:

1. Relationships between the four most commonly

used terms: open throat, throat widening, retrac-

tion and space at the back2. Preferred term/terms

3. Sound qualities associated with these terms

4. Required technique to achieve the sound quality

5. Relationships between terms, actions and sound

qualities

6. Physical sensations and presumed physiology

associated with the terminology

Procedure

The interview was pilot tested on an experienced

singing pedagogue and modifications were made

to increase the clarity of the questions where necessary.

The interview schedule was used as a guide; specificopen-ended questions were followed by individualized

prompts and probes (30). A lexicon of interview

prompts was used to facilitate the flow of ideas

in the interview, although care was taken not

to interrupt the participant (31). Participants were

encouraged to elaborate on ideas through the use

of agreement prompts from the interviewer. The

interviewer used clarifying prompts to refine partici-

pants’ statements as necessary. Researcher agreement

was established regarding the appropriate use

of prompts during the interview prior to commence-

ment.

Interviews were conducted either in person (N�/4)

or by phone (N�/11) at a time and place convenient

to the participant, and permission to record the

interviews was obtained before proceeding. Tapes

were marked by participant number and transcribed

with transcriptions sent back to respondents for

verification and final approval before they were

qualitatively analysed. At this stage, participants

were given the opportunity to clarify anything that

they had said.

Analysis

Transcripts were analysed using in-depth qualitative

analysis, detailed in Smith (30) to identify relation-

ships between terminology, theory and teaching prac-

tice for each pedagogue (32). Transcripts were read in

depth to generate a profile of the topic responses. They

were then similarly coded and common themes were

identified. Significant or representative statements

were highlighted and collated across participants.

Recurrent themes were examined for inter-subject

similarities and differences. For example, mentions of

new terms were compared across pedagogues with

regard to term used, sound quality achieved and

technique applied. This type of qualitative data

collection and analysis has been used extensively in

research of singing evaluation and assessment (33, 34).

A coding scheme for the transcripts was developed.

Three of the four authors reviewed the codes and

independently selected exemplars of each code. This

type of interpretative analysis is described by Smith

(30) as a way of capturing the meaning of a partici-

pant’s responses.

These qualitative data form the basis of the results

presented below. Participants’ responses to the ques-

tion topics are grouped thematically in Results, in

accordance with the stated aims of the semi-structured

interview topics. The quotes were examined by an

expert pedagogue, within the context of the transcript,

to ensure they represented instances of the stated

theme, or illustrated a significant statement unique to

an individual or sub group.

Defining open throat 169


RESULTS

Participants

The 15 singing pedagogues interviewed for the study

were aged between 39 and 63 years, with an average

age of 49. Ten primarily taught in Australia and fivetaught in the UK. They dedicated an average of 95%

of their professional life to singing or the teaching of

singing. They had between 4 and 28 years of

experience in teaching singing, with an average of 20

years.

For nine participants (P1, 2, 3, 6, 7, 9, 10, 11, 12), an

average of 67% of their students performed in the top

four categories (superstar, international, national/bigcity, regional/touring) of the Bunch and Chapman

taxonomy (28). Participants’ studios contained an

average of 20% full-time singing students, and for six

participants (P1, 4, 5, 8, 13, 14), ]/50% of their studio

comprised full-time students at a tertiary institution.

Influences

Table 1a and Table 1b present data in response to the

two questions ‘Do you follow any particular school of

thought or pedagogy or scientific approach?’ and

‘How does that pedagogy or writer influence your

view on open throat, throat widening, retraction or

space at the back?’ respectively. From the table, it can

be seen that the major influences were divided equallybetween Estill (7 participants), Richard Miller (7

participants) and Janice Chapman (5 participants).

Seven participants also referred to their own experi-

ence or study as influential on their pedagogy.

Participants who cited experience as their key

influence made comments of the following type: ‘I

got to the point in my development where I just had to

know what those things meant for me personally, so Ithink it was a lot of my own exploration that brought

about a deeper level of understanding’ (P15).

Factors in the achievement of good sound

In response to the questions ‘How do you achieve a

good sound in singing?’ and ‘How might you correct a

problematic sound in singing?’ 50% of participants

offered an average of four key concepts that they

considered essential to good sound (range 1�/6 terms).

Good breathing was the most important factor for ten

participants (67%) followed by balance or coordina-

tion for eight (53%). Table 2a presents all the factors

cited by participants and the frequency of citation,

and Table 2b presents references to open throat.

Samples of typical comments on good sound are:

‘A good sound really relies so much on freedom and

good coordination and connection with air so that the

right muscles are engaged with the right amount of

energy and there’s no interference’ (P8). ‘I would aim

to establish a freedom of production, so that is looking

at breath, de-constriction, a free flow, a free pathway

for the sound’ (P9). ‘I’m looking for a looseness

around the neck and for the articulators to be loose so

that there’s no musculature resistance to the emission

of sound. I’m looking for something that old peda-

gogy would call space making’ (P15).

Five teachers referred to the subjective nature of its

definition. Specifically, a good sound ‘is a sound that

is non-abusive and appropriate to genre’ (P7); ‘is

subjective to people’s tastes, so I work towards making

sure that the sound is safe, and then I work on the

aesthetics of the sound’ (P12); ‘needs good aural

awareness and good aesthetic training’ (P4).

Of the eleven participants, three referred to specific

technical instructions to achieve a good sound. They

emphasized specific components of their pedagogical

approach and named the phenomenon of open throat

technique: ‘Well the first thing I work on is posture,

and the second thing I work on is breathing and the

third thing I work on is resonance . . . and the most

important thing about resonance is the open throat’

(P1); ‘I teach all my clients to retract, that’s one of the

Table 1a. Pedagogues most regularly cited as influential in the development of teaching techniques for singing

Influence Chapman Estill Miller

Pedagogues 1, 4, 7, 9, 15 1, 2, 3, 4, 7, 9, 10, 11, 13, 15 2, 4, 8, 9, 10, 13, 15% 33 67 47

Table 1b. Pedagogical influences of singing teachers with respect to open throat techniques

Influence Estill Miller Self/eclectic Chapman Caesari Other

Pedagogues 1, 2, 3, 7, 9, 10, 11, 12, 13 1, 4, 6, 8, 10, 11, 14 1, 2, 4, 5, 6, 8, 10, 13, 15 2, 9, 10, 11, 14, 15 11, 14 4, 11% 60 47 60 40 13 13



first things I teach them’ (P12); ‘The larynx is sitting in

a slightly lower position . . . basically the larynx is

being pulled down by one laryngeal depressor, namely

the sternothyroid’ which ‘maximizes pharyngeal

depth’ (P10).An additional two participants mentioned primal

sound in their account of achieving good sound. This

was later defined as an action containing a component

of open throat technique. ‘I’ll start with breathing,

support, posture and the primal sound’ (P2). ‘Sound is

related to primal expression, people say it’s over the

top, I say opera starts over the top’ (P14).

Associations between the four terms

After the open-ended questions exploring general

concepts of pedagogy, participants were offered four

common terms used in the singing literature and in

singing pedagogy: open throat, throat widening,

retraction and space at the back, and were asked

‘What do the terms mean to you?’ and ‘How do you

feel these terms are related?’

Nine participants (P1, 3, 5, 6, 7, 8, 11, 12, 14) agreed

that the terms were either interrelated or interchange-

able, through universal use in singing studios. How-

ever, each participant went on to define more specific

differences between terms and chose one or more that

they advocated in teaching: ‘Well, they do [mean the

same thing]. If you’re talking about open throat,

everybody knows what you mean*/it’s a generally

accepted term’ (P5).

Three preferred the term open throat (P5, 6, 14) and

consistently used the term throughout the interview.

Five (P1, 3, 7, 11, 12) of the nine interpreted all

terms as attempts to describe the action of retraction:

‘All of those [terms] to me mean that the false folds are

retracted, that there is no interference of the false folds

in the airflow’ (P3). ‘I would say that open throat and

retraction are the same thing’ (P7).

For six participants (P2, 4, 9, 10, 13, 15), each term

was defined differently, and they described subtle

differences between these terms. Statements gave either

positive or negative associations. For example:

Positive statements: ‘Retraction I use because I find

it’s actually a very useful technique when people are

constricted, I can hear false vocal fold interaction very

clearly . . . open throat is an end product of getting

something right physiologically . . . it doesn’t have any

meaning outside of the actual physiological basis’

(P2). ‘I think retraction for me would mean false folds

opening up or open wider, so that there’s a sense of

free emission of sound . . . Space for me, also carries a

physical sensation of stillness, so that it’s not forced in

any way, there’s no laryngeal depression via the base of

the tongue, or no artificial palate lifting that involve

tension of any of the other articulators’ (P15).

Negative statement: ‘Open throat is the beginning of

a yawn, relaxed, slightly soft back of the tongue,

‘‘high-ish’’ palate, wide pharynx. Throat widening

might be slightly less relaxed and it might make me

think a little bit about retraction’ (P4).

For seven participants (P2, 4, 6, 7, 12, 13, 15), space

at the back was perceived to be a different concept to

open throat because of the required action of the soft

palate in enlarging the space. ‘Space at the back . . . to

me means wide pharynx and a high palate . . . open

throat is for me achieved by having a correct inhala-

tion process, which in my terms is a relaxed abdominal

or splat breath’ (P2). ‘Throat widening to me can be

lower than ‘at the back’ . . . [which] I perceive to be in

the soft palate region’ (P13).

Two participants (P7, 8) defined throat widening as

an external action, and related less to the internal

action of open throat. In their pedagogy, it was not

perceived as a useful term: ‘Throat widening [is not the

same as retraction]; I prefer open throat, because then

it, for me, indicates that something is being opened

Table 2a. General factors necessary to achieve a good

sound, ranked by pedagogue frequency of citation

Factors %

Breathing 67Balance/coordination 53Posture 40Support 20Energy 13Primal sound 13Resonance 13Tongue 13Airflow 7Aural awareness 7Effort 7Placement 7Positioning 7Ring/twang 7

Table 2b. Specific open throat references as a necessary

factor to achieve a good sound, ranked by pedagogue

frequency of citation

Open throat references %

Ease/freedom 40De-constriction/not constricted/tense 33Open/open throat 20Depth 13Retraction 13



from the inside rather than the outside which is where

I teach people to make space’ (P7).

Two participants (P9, 10) expressed dissatisfactionwith all of the given terms because they were

potentially misleading: ‘Open throat, I think is a little

bit misleading for some people in that most singers

were raised on the old yawn sigh and the concepts of

yawning and making space’ (P9).

Preferred terms

Eight participants favoured one or more of the four

terms in their teaching. The other seven gave a

different term. Table 3 presents the use of preferredterminology.

New terms were given as (i) replacements for the

four given terms, or (ii) as better descriptors of similar

concepts; or (iii) as descriptors of new vocal pedagogy

believed to incorporate the same pedagogical goal.

Four participants (P8, 9, 11, 13) favoured the use of

the words free, freedom or freedom of tone.

‘I wouldn’t say I want an open throat . . . I wouldsay I want a free sound’ (P11). ‘A free sound, a

freedom of sound but the beginnings of what we call a

trained voice sound’ (P13).

‘Collar’ and ‘depth’ were discussed by four partici-

pants (P2, 9, 10, 15) and described as similar to, or

replacements for, open throat. For one, the concept

of collar is intrinsically related to the function of

retraction or open throat. ‘I teach [retraction] as afundamental of singing by getting the engagement of

the collar . . . the collar is both a postural and

a physiological function’ (P2).

For two participants (P10, 15), open throat was the

achievement of pharyngeal depth in classical singing.

It was also directly compared to the technique ‘collar’:

‘[The term collar], I’d call it low larynx or pharyngeal

depth’ (P10). ‘There is also a sense of laryngeallowering so that there feels like a great depth in the

vocal tract’ (P15).

Three of the participants (P2, 10, 15) would not use

a specific term in teaching but rather preferred to use

the students’ own language to describe their sensation

after teaching them the technique: ‘I would look for as

many synonyms as possible to describe the same thing,

and then I would probably ask a person to feed back

to me in their own language what we’ve just been

discussing so that we can actually have a point of

reference, so we’re both very clear as to what we’reboth talking about’ (P15).

Sound qualities associated with the terminology

Participants were asked ‘What sound quality do youassociate with these terms?’ They collectively produced

18 descriptions that they felt best characterized the

sound when open throat was used. Table 4 identifies

the most common sound terms used. ‘Balanced/

coordinated’ was the most important quality, named

by 14 of 15 teachers (93%). ‘Free’, ‘open’, ‘even/

consistent’ and ‘warm’ were each nominated by over

50% of pedagogues as the sound quality intrinsic tothe use of some form of open throat.

Fourteen participants (93%) said that using open

throat made the entire sound more balanced or

coordinated. They explained that using open throat

achieved: ‘a much better range of harmonics and

obviously from that, the formants’ (P3); ‘a resonating

spectrum [in which] frequencies balance quite beauti-

fully’ (P15).Eight pedagogues (53%) felt that the use of open

throat made the overall sound more even or consis-

tent: ‘If I told them to laugh, I was going to get a more

efficient result and a more effective sound, therefore a

rounder, bigger, more even, more secure, more stable,

more confident production of sound’ (P1).

Five pedagogues (P1, 2, 3, 7, 12) described healthy,

safe sound and then associated this with terms likeclear, clean and free: ‘The sound is clean . . . the sound

has no interference’ (P7). ‘The timbre changes when

retraction is achieved, and the voice sounds much

healthier, clearer, cleaner’ (P12).

Required technique to achieve the sound quality

In response to the questions ‘Suppose I were a new

student*/how would you tell me about some of these

terms?’ and ‘How would you teach me these techni-

ques?’ six pedagogues used some form of laugh and

five used cry or sob. A pre-yawn or start of a yawn was

mentioned by three, and focus by two. Three pedago-

Table 3. Terms favoured by each pedagogue as descriptors for open throat

Favoured term Open throat Retraction Open throat or retractioninterchangeably

New

Pedagogues 5, 6, 14 3, 7, 12 1, 2 4, 8, 9, 10, 11, 13, 15% 20 20 13 47



gues addressed the technique only through teaching

breathing. These are summarized in Table 5.

Nine pedagogues (P1, 2, 3, 5, 7, 8, 10, 11, 12)

believed that the action of open throat was

fundamental to the consistent production of good

sound. It was an important technique to maintain and

monitor throughout singing: ‘Otherwise you start to

use all the muscles that you shouldn’t deliberately be

using’ (P5). ‘It enables the mechanism to work

efficiently at source, so that you get a good sound

signal’ (P7). ‘It’s very much concerned with freedom

and the idea of just opening up and letting the sound

flow’ (P8). ‘Basically after [freedom], you’re adjusting

resonances’ (P11).Five pedagogues (P2, 7, 8, 9, 13) believed that open

throat needed to become automatic. They stated that

this was possible once a student had mastered

the action: ‘Yes, it’s there all the time when you’ve

got it*/it becomes part and parcel of your normal way

of singing. But it’s part of the breathing and support

function’ (P2). ‘There has to be a portion of the

concentration monitoring, until it becomes natural . . .

every time they open their mouths, they know it’s

going to happen’ (P13). ‘It is one of those things that a

student can become able to do automatically and

therefore it goes into the muscular memory and

it’s something that once you’ve experienced, you might

only monitor when you notice that it’s not happening’

(P7).

One pedagogue (P11) highlighted the problem of

actually learning to achieve the action of freedom

consistently: ‘It seems to me that you actually have to

have exercises that maintain that sensation, you can’t

just identify it and say that’s what’s happening’ (P11).

For two pedagogues (P6, 8), open throat was valued

as a tool that demanded more concentration in certain

ranges of the voice and particularly singing through

the passaggio: ‘I would say, particularly as one is going

into the passaggio . . . it helps people to have that

feeling of space at the back of the throat, and almost a

sensation of the voice going backwards’ (P6).

In the singing studio, three participants (P1, 8, 12)

said that it was a technique they regarded as important

to consider every time they taught: ‘[Retraction] is

something I teach every day. And with every student,

no matter how advanced’ (P1). ‘I use it all the time. I

think it’s like second nature . . . if I weren’t retracting,

I’d be constricting’ (P12).

Two participants (P4, 9) regarded open throat and

the teaching of it as a prescriptive tool or as a

technique to visit only when constriction was causing

a problem. They advocated addressing constriction

in other ways rather than discussing open throat:

‘[Open throat] isn’t an integral concept I consciously

and consistently consider . . . so on a needs basis.

Table 4. Sound qualities associated with open throat. Pedagogues’ individual choices indicated by I

Pedagogue 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 %

Balanced/coordinated I I I I I I I I I I I I I I 93Free I I I I I I I I I I 67Open I I I I I I I I 53Even/consistent I I I I I I I I 53Warm I I I I I I I I 53With space I I I I I I I I 40Healthy/safe I I I I I 33Round I I I I I 33Overtones/formants I I I I I 33Easy/flexible I I I I I 33Clear I I I I 27Full I I I 20Efficient I I I 20With depth I I I I 20Clean I I I 20Sexy/juicy/lusty I I 13Natural voice I I 13Relaxed I 7

Table 5. Pedagogical views on techniques to achieve open throat

Technique Laugh Cry or sob Yawn (start/pre) Focus Flow of breath Inhalation

Pedagogues 1, 3, 7, 9, 12, 15 1, 2, 9, 10, 15 2, 4, 9 5, 6 13, 14 8% 40 33 20 13 13 7



I don’t necessarily talk about open throat that much

. . . it’s not fundamental’ (P4). ‘[I use open throat] as

and when it matters. Some people have very little

constriction . . . often, if their breath is in place and if

that is buoyant enough, that takes care of a lot of

constriction anyway’ (P9).

Relationships between terms, actions and sound

qualities

Thirteen participants (80%) made a connection be-

tween breath or breathing-in and the concept of an

open throat. For eight participants, it was linked with

the inhalation process (P1, 2, 3, 8, 9, 10, 12, 15): ‘Open

throat . . . is for me achieved by having a correct

inhalation process . . . it’s a recoil diaphragmatic

breath’ (P2). ‘We want to maximize resonance and

freedom, maintaining that posture of inhalation’ (P8).

‘It’s a physical feeling that you get on inhalation and

that you maintain until the end of the phrase’ (P10).

‘I think space is initiated with inhalation’ (P15).

Similarly, seven participants (P1, 2, 7, 8, 9, 10, 15)

indicated that they would actively use the technique to

prepare for producing a note and linked it to the

inhalation process: ‘I would use it in association with

breath and taking breath because those two activities

are combined’ (P9). ‘You set it up with the intake of

breath.’ (P12).For four participants (P5, 7, 13, 14), it was an action

connected with the flow of breath and the continua-

tion of the breath throughout the phrase: ‘So there is

this lovely column of sound between the focus point

and the support in the voice’ (P5). ‘I talk more about

energizing the breath, energizing the voice, keeping the

tone consistent, working to the end of the phrase’

(P13). ‘It is to do with the flow of breath. If I can

perceive that the flow of breath is neither squeezed nor

dissipated . . . then generally the throat is open’ (P14).

Instructions to achieve an open throat included

actions to maximize space. Two participants (P2, 9)

used a pre-yawn or pre-yawn sigh to describe this to

students: ‘Sometimes I add instructions to maximize

the vocal tract by using, for example, the roar of a lion,

to feel how big those spaces can be and the yawn-sigh

effect, the pre-yawn sigh, I call it of allowing the sound

to come out without constricting the walls of the

pharynx’ (P2).

The technique of laugh was consistently linked to

the term retraction: ‘I use the device of laughter

because I know that that brings about the physical

action [of retraction]’ (P3). ‘[I would teach retraction]

via silent laughing, but I also use achieving a silent in

and out breath, so basically silent inspiration and

expiration . . . it’s a position and it’s a muscular

activity. What happens with the breath in terms of

airflow for instance happens as a result of it’ (P7).

For three participants (P1, 9, 15), laugh and cry orsob produced a subtle change to the sound quality:

‘a different colour’ (P1), ‘richer’ (P9) and a ‘different

harmonic readout within the tone’ (P15). ‘The other

way to open the throat is to cry in tune... on the

laughing basis, but I want to change the colour in the

sound. So I’m doing that for an aesthetic colour’ (P1).

Seven participants (P1, 2, 7, 8, 9, 10, 15) sponta-

neously commented on the value of open throat inclassical singing. It was considered (i) as a necessary

technique for making a classical sound and (ii) a

quality in the sound itself.

Classical technique: ‘Requires a degree of work

taking place in the larynx, and so therefore you need

to have the retracted position, the open throat, in

order to make it safe’ (P7). ‘The aim for classical is to

have the greatest sense of freedom involved, andfreedom with the laryngeal tract, or the vocal tract,

happens to occur with space-making.’ (P15)

Classical sound quality: ‘I think the Italian articu-

latory system sets up an open throat, and western

classical singing came from that . . . [open throat] is the

difference between a national sound and an interna-

tional sound’ (P2). ‘I think that [open throat] not only

improves but in terms of classical singing, it fines thequality that one requires if you want to be a classical

singer’ (P8). ‘It probably contributes acoustically to

the sound that is most satisfying in that particular

musical style’ (P9). ‘With depth, I will hear basically a

warmer, more classical sound’ (P10).

Physical sensations and presumed physiology associated

with the terminology

The physical action of open throat was conceptualizedby participants in two ways: in terms of anatomy or

physiology, and/or as a physical sensation.

Table 6 presents the anatomical and physiological

concepts mentioned in response to the question: ‘In

terms of physiology, which muscles would you expect

to be working, that is to say, what happens when you

use open throat?’

Eight participants consistently mentioned a pre-sumed ‘abduction’ or retraction of the false vocal folds

or ventricular folds which they linked to the term

retraction. Participants who referred to false vocal fold

movement reported seeing the action on endoscopy.

‘The false folds are retracted and there is no inter-

ference of the false folds in the airflow or indeed that

the false folds are not in any way pressing on the true

vocal folds’ (P3). ‘What I have noticed, when I look atpeople under endoscopy and stroboscopy, is that when

the ventricular bands are retracted the true vocal folds



have a much more symmetrical mucosal wave, and so

they behave much more efficiently and you hear that

efficiency in the purity of the tone’ (P12).

Seven participants stated that pharyngeal space was

involved in the technique they used: ‘[In retraction],

there is some lowering [of the larynx] and I think

there’s some widening of the larger spaces in the

laryngeal part of the pharynx’ (P2). ‘Open throat to

me is optimizing; perhaps all your pharyngeal space,

and that might involve a slight lowering of the larynx’

(P4). ‘I think it has more to do with pharyngeal depth

than it is to do with widening. Because as the larynx

descends, it comes slightly forward in the neck and if

you have a high, wide tongue position that’s slightly

further forward and you actually gain more pharyn-

geal depth as well as pharyngeal length’ (P10).

Six participants mentioned larynx lowering or the

use of sternothyroid as having an effect on open throat

or pharyngeal space: ‘I think that there has to be an

absence of the glossus muscles, that the muscles that

push down on the larynx from above are a hindrance

in getting an open throat. Sternothyroid, the true

depressor of the larynx, which actually functions

physiologically from below is much more valuable’

(P2). ‘I expect that there is a little bit of thyroid tilt, so

that the muscles involved in the thyroid tilting

mechanism are going to be involved [in retraction]’

(P3). ‘I would say the oropharynx or the middle

constrictor region, certainly tongue muscles, the

thyrohyoid, maybe sternothyroid as well would all

work as a unit to allow laryngeal lowering. Certainly

you are looking for a slackness in the muscles above

the larynx’ (P15).

Five participants (P2, 3, 7, 11, 12) commented that

there was not adequate research, and they could not

be completely sure what muscles were used or affected

by the action of open throat: ‘I can’t tell you that

really. I think that the research has just not been done’

(P2). ‘This is one of the things that needs more

research, we don’t absolutely know what happens

when you retract’ (P12).

Table 7 presents data on sensations associated with

the use of this technique. Participants described the

sensation they had or tried to achieve when they used

open throat. They named three locations of sensation

they associated with this technique: (i) internally

at laryngeal level; (ii) externally at laryngeal level;

and (iii) in the oropharynx and laryngopharynx. Table

8 presents the participants’ responses to the questions

of associated physical sensations when using open

throat.

Three participants described an internal sensation

at laryngeal level and specifically linked this to the

action of retraction: ‘[You feel] a widening . . . right at

the laryngeal area itself and also it feels much wider

behind the epiglottis and down in the pharynx itself’

(P1). ‘You feel it right inside the vocal mechanism. It

may just be that I’ve got a good imagination, but I feel

the giggle or the sob has got a very subtle, very internal

manoeuvre’ (P2).

For four participants, the action was felt externally

at laryngeal level. It was also visible in the movement

of the thyroid cartilage: ‘You can actually feel it

around the sides of the thyroid cartilage, because if

you’re widening on the inside there, then you do

actually get a widening [on the outside]. So you can

feel the effects of the retraction*/you can’t feel the

false vocal folds from outside’ (P7).

One participant (P10) was more explicit about this

external action, and linked it to vertical movement of

the larynx: ‘I feel it in the base of my neck . . . which is

actually sternothyroid having to kick in to pull the

larynx low enough so that you can get a thick enough

edge of vocal fold together for a slower vibratory rate’

(P10).

Six participants explained that the action of open

throat was felt throughout the oropharynx and

laryngopharynx: ‘One feels a kind of opening and a

Table 7. Pedagogues’ report of perceived sensation when they used open throat

Sensation Internally at laryngeal level Externally at laryngeal level In oropharynx and laryngopharynx

Pedagogues 1, 2, 3 3, 7, 10, 12 4, 6, 8, 9, 11, 15% 20 27 40

Table 6. Pedagogues’ perceptions on anatomical and structures involved when using open throat

Structure False vocal folds Pharynx Sterno-thyroid Larynx lowering Other structure

Pedagogues 1, 2, 3, 7, 11, 12, 13, 15 1, 2, 4, 8, 9, 10, 15 2, 10, 12, 15 2, 3, 4, 10, 15 7% 53 47 27 33 7



rising feeling at the back of the throat’ (P6). ‘[The

sensation] is all about pharyngeal muscles’ (P8).

‘Mainly, the sensation is within the throat region’

(P15).

Table 8 presents responses to the questions ‘Do you

think it is possible to have voluntary control of

muscles responsible for open throat, throat widening,

retraction and space at the back?’ and ‘Do you believe

it is possible to isolate and control the muscles used in

open throat, throat widening, retraction and space at

the back?’Thirteen participants (87%) considered that it was

possible to have voluntary control of the action of

open throat: ‘Conscious control of constriction or

retraction, nothing’s new really... my experience is that

of course it is’ (P11). ‘I certainly think you have

voluntary control of open throat, in that you can

optimize space’ (P4). ‘I think there is, up to a point,

voluntary control’ (P6). ‘It can take a long time, but

yes, you can.’ (P10). ‘I think to a certain extent... I

don’t know how efficiently [the muscles] work to

sustain that. I think you can generate a response,

particularly on the point of inhalation, but I’m sure

that there are postural considerations such as head

and neck alignment that help to maintain that

positioning’ (P9).

Two participants considered open throat as part of a

greater whole of singing technique, rather than a single

action: ‘The muscles controlling the larynx work of

their own accord’ (P5). ‘If a singer thought that they

had to have voluntary control every time they went out

on stage, they’d be in a mess. No they don’t have

voluntary control, but you do it, you get to the point

where... once you put the message into the subcon-

scious, it becomes habitual and then we just do it’ (P8).Seven participants believed that is was possible to

isolate and control the action. These comments were

specific to retraction: ‘I’m primarily looking for the

isolation simply of getting the false folds retracted so

that we have a de-constricted space at vocal fold level’

(P3). ‘Yes, I believe so, it has certainly been my

experience’ (P7). ‘I feel as though you can [isolate]. I

don’t know that I can actually control that, it may be

something else that’s moving’ (P9). ‘[Isolation] defi-

nitely, yes. However, I don’t know if anyone could

feedback accurately exactly what muscles they were

working with at any given point in time’ (P15).

Five participants believed that isolation of open

throat was not possible and that the action was a

component of a larger action or gesture in singing: ‘[Awhole motion] would certainly be my approach, rather

than trying to isolate elements’ (P6). ‘I don’t think you

can isolate one little thing. It’s part of a bigger picture,

and I think sometimes, when you try to isolate bits, it

can become too compartmentalized and a student

actually ends up thinking about bits and forgets about

singing’ (P13). ‘I don’t think that one can take one

muscle and say, exercise it, for example. They all workin a greater union’ (P15).

DISCUSSION

The majority of participants interviewed for this study

agreed that open throat was essential to good singing

and more specifically to classical singing. Specific

sound qualities in good singing were identified and

some were linked directly to open throat technique.

Participants acknowledged the complex nature ofinterdependent techniques needed in good singing.

Good breathing, balance or coordination, and free-

dom or ease of sound were considered essential to the

production of a good sound in singing. These concepts

have often been linked in the literature; for example,

Miller (24) believed that freedom in sustained singing

is produced by balanced breathing.

With respect to terminology regarding open throat,open throat and retraction were preferred, from the

given terms, to describe the phenomenon. Other terms

such as freedom, collar and depth were suggested as

better terms. Individual participants preferred one of

these terms to describe their concept of open throat,

although some used terms interchangeably. Reasons

for preferences given include: clarity of meaning;

preference for a specific pedagogy’s terminology; orreference to the type of action used to achieve the

technique.

Different pedagogues preferred different terminolo-

gies: the descriptors used reflected preferences to

current and historical singing pedagogy or pedago-

gues. For example, Miller (24) referred to open throat,

Estill referred to retraction (16) and Chapman to

collar, and the use of these terms were reflected in thecomments of this sample. However, the sound quali-

ties associated with all these terms were very similar.

Table 8. Pedagogues’ view on degree to which open throat was voluntary, involuntary, could be isolated or was part of

a coordinated action

Control Voluntary Involuntary Isolated action Coordinated action

Pedagogues 1, 2, 3, 4, 6, 7, 9, 10, 11, 12, 13, 14, 15 5, 8 2, 3, 7, 8, 9, 11, 12 4, 5, 6, 13, 15% 87 13 47 33



Singing teachers associated open throat with both a

sound quality that was characterized by freedom,

warmth and openness and an action that produced

balance, coordination, evenness and consistency. The

terms related to vocal quality such as ‘warm’, ‘full’ or

‘round’, as well as the functional terms such as ‘easy’

or ‘clean’, were used interchangeably by participants

in this study. The use of function and quality

descriptors was addressed in Wapnick and Ekholm

(26), who found that there were differing preferences

for use of quality or function terms. Although experts

tended towards one category of descriptor, the high

correlations of all descriptors suggest that sound

quality and function are a unitary concept in evaluat-

ing good sound. The participants in this study did not

separate these types of terms and seemed comfortable

with their use in the studio.

Participants in this sample demonstrated an under-

standing and interest in anatomical and physiological

relationships between sound quality and technique,

citing presumed muscle action such as ventricular fold

abduction or pharyngeal widening and associated

physical sensations. Most considered it feasible to

have voluntary control over the action of open throat,

but did not achieve consensus over the ability to

isolate the action from good singing technique. The

techniques used by participants in this sample in-

volved laughing, crying, conscious maintenance of

inhalation postures and (pre-) yawn. Scientific inves-

tigations related to each of these vocal actions provide

interesting hypotheses for what may happen physiolo-

gically when participants teach open throat technique

(25, 35�/9).

Laughing is advocated by pedagogues such as Estill

(16) and Miller (24) as a fundamental tool in each

pedagogical style for (i) the achievement of false vocal

fold retraction and (ii) for the balance of onset and

release respectively. The majority of those who advo-

cated false vocal fold retraction (ventricular fold

abduction) said that isolation of this action was

possible. This view is consistent with Estill (16, 25)

who recommends independent control of components

of the vocal mechanism. It is important to note that

no studies have been published providing evidence of

voluntary control of movement of the ventricular folds

laterally away from the midline. Current thinking is

divided as to whether retraction (25) occurs in open

throat and if there is a response to an instruction to

‘retract your ventricular folds’ or ‘retract’ from a rest

position.

Participants who used laugh as a technique did so to

achieve retraction. Although the use of this term has

been absorbed into singing pedagogy, it lacks anato-

mical validity. The anatomical term ‘abduction’ refers

to a lateral movement away from the body midline

with the opposite term, ‘adduction’ referring to a

lateral movement towards the body midline. The term

abduction appears to be the movement that some

pedagogues referred to as retraction which is not

usually used anatomically in that sense. However, the

use of the term retraction by the participants inter-

viewed here clearly meant more than a relaxation to a

rest position from a constricted vocal pattern; it also

implied a lateral movement away from the approxima-

tion to the midline. What is not known is whether

there is any voluntary control of the lateral position of

the ventricular folds as advocated by Estill. Those

participants who used the term and technique of

retraction did so to reduce constriction or tension to

achieve a healthier sound quality as well as to achieve

a specific sound quality.

Physiologically, abduction of the ventricular folds

presents a puzzle as they are ‘incapable of becoming

tense’ (40, p. 117). Histological reports have described

a rostral extension of the thyroarytenoid muscle up to

13 mm above the glottis, which at the level of the

ventricle, 2�/3 mm above the glottis, lies lateral to it,

and at this level it is referred to as the ventricular

muscle (41). The thyroarytenoid muscle is an adductor

of the vocal folds, and will have the same action on the

ventricular folds and so cannot be responsible for

abduction of the ventricular folds. There is voluntary

control over constriction of the ventricular and vocal

folds with adduction of the ventricular folds, and

changes in the vocal vibratory waveform are observed

in strained/tense and also in pressed phonation.

Reduced tension of the larynx and a degree of

expansion in the laryngeal ventricle was reported in

‘covered singing’, a technique used when singing

through the passaggio (42).

Changes in the lateral position of the ventricular

folds may alter the shape and position of the ventricle,

which could be responsible for other resonator char-

acteristics or even the vibratory pattern of the vocal

folds. In studies of specific vocal register by Welch

et al .(38), they found that male falsettists demon-

strated a systematic increase in pharyngeal and

ventricular area as well as shape and position of the

laryngeal ventricle across a series of octaves. They

suggested that darkening or covering was achieved

while still maintaining ‘singer’s formant’, by opening

the ventricular space and lowering the larynx. Refer-

ence to the ventricle and its perceived role in singing

were made by the Bel Canto school, and interpreted in

the twentieth century by practitioners such as Manen

(19) to include direct reference to the ventricular folds,

and she concluded that they are altered in length and

width, which directly affected the ventricle of Mor-

gagni and produced the ‘specific timbres of Bel Canto’

(19, p. 34).



However, it is not yet established whether the

opening of the ventricular space and lowering the

larynx is influenced by instruction and adjustments

made for the technique of open throat. Sundberg (39)

proposed that the observed distinctive high-energy

spectrum in singer’s formant is present when a specific

size relationship occurs between the diameter of the

pharynx and the opening of a narrowed laryngeal

inlet or tube (size ratio pharynx to larynx opening

areas �/6:1), which is more likely to occur when the

larynx is lowered. Participants in this study did not

specifically link the use of an open throat with ‘ring’,

terms which have previously been linked (43),

although they did suggest it facilitated a better balance

or more desirable distribution of harmonics.

Some participants in this sample advocated cry as

an instruction and discussed larynx lowering as an

action in open throat. Certain participants speculated

that the contraction of the sternothyroid muscle might

influence space created in the pharynx, larynx low-

ering or other associated movements in the vocal

mechanism. There is some support for this hypothesis.

Sternothyroid contraction is believed to influence the

thickening of the vocal folds in low pitches, but it is

uncertain as to its effect on higher pitches in trained

singers (44). The sternothyroid muscle attaches to

the thyroid cartilage, and it is possible that on short-

ening, it makes some impact on the vocal fold

configuration. Roubeau et al . (44) found that the

activity of the strap muscles, including the sternothyr-

oid, varied according to fundamental frequency, with

most activity occurring at the extremes of range. Hong

et al . (10, 45) concurred that the strap muscles did

affect fundamental frequency through the mechanism

of laryngotracheal pulling or bending. These studies

support the views of the participants’ interviewed that

contraction of the sternothyroid muscle has some

effect on the laryngeal configuration.

Vilkman et al . (46) agreed that changes to the

vertical position of the larynx had effects not only on

fundamental frequency but also may contribute to

abduction and adduction functions within the larynx.

This supports the benefits of lowered larynx as used in

classical singing. Shipp (47) determined that thyro-

hyoid and sternothyroid muscles were responsible for

subjects’ vertical laryngeal positioning while singing,

which matches the participants’ beliefs of laryngeal

control and manipulation.

Pedagogical methods linking breath, or the intake

of breath, was deemed by pedagogues in this sample to

be vital in the production of open throat, and achieved

using a variety of images or gestures in order to locate

a sound quality and sense of space or freedom (5).

Sonninen’s (48) pioneering study into external laryn-

geal frame musculature supported this pedagogical

theory, and identified anterior and inferior displace-

ment of the thyroid cartilage by sternothyroid and

tracheal exertion via inhalation, which may account

for vibratory changes in the vocal folds and acoustic

changes in fundamental frequency.

Many participants in this sample talked about

widening of the pharynx or a sensation of space

within the pharynx. They also associated this with

breath, and an action initiated through inhalation.

Anatomically, there is evidence that the pharynx can

widen or lengthen during singing. The pharynx is

comprised of three constrictors so, anatomically, only

lengthening of the vocal tract, or protrusion of the

tongue, should enlarge the pharyngeal resonator. The

yawn/yawn-sigh has been used to encourage open

throat, but has been questioned as a tool for achieving

it, as it may add unnecessary tensions (24). Pre-yawn

or yawn-sigh was named in this study as other

pedagogical devices to achieve open throat. Boone

and McFarlane (49) studied the yawn-sigh under naso-

endoscopy and identified characteristic lowered larynx

and widened pharynx across subjects, as well as

retracted elevation of the tongue which supports the

participants’ application of yawn or pre-yawn to

enlarge the pharyngeal resonator. Acoustically, this

yawn-sigh produced lower second and third formants.

The notion of pharyngeal size and its impact on the

increased intensity in sound was reported by the

participants interviewed in this study. Sonninen et al .

(50) investigated the size of the pharynx during open

and covered singing, in loud and quiet examples, and

proposed that the lower pharynx was in fact narrower

during the loud covered condition than in the open

condition. However, it is qualified by its restriction to

the low level of pharynx area studied. Although

pedagogues here referred to wide pharynx, they also

quantified depth via laryngeal lowering as important

to open throat. The narrowing of epiglottal area, and

its role in the production of singer’s formant is

investigated by Titze (9), who proposed the benefits

of adopting a wide pharynx, to achieve a lower first

formant to enhance the sound, and increase the

singer’s formant with a narrowed laryngeal inlet or

tube (39).

Recent studies (10, 36) investigated various voice

qualities including the ‘yawny’ voice quality and

concluded that this vocal tract manipulation resulted

in a distinctive increase in vocal tract volume, through

lengthening the vocal tract and widening the oral

cavity, which resulted acoustically in a closer grouping

of the first two formants. These findings confirm

pedagogical statements in this study related to the

perceived impact of the use of the sob/cry/yawn

manipulations to achieve greater space and lower the

larynx, but do not investigate the concept of tension.



Studies of control of different sound qualities,

focusing on different vocal genres have identified

different vocal ‘postures’ in singing (35, 37). Physio-logical movements were different across singing styles

and were believed to account for changes to the sound

quality. When the vocal tract was compared across a

single singer, operatic technique versus belting techni-

que elicited a lower larynx, wider larynx tube and

more separated ventricular folds. Interestingly, the

instruction to change the sound quality produced the

physiological variations.

CONCLUSIONS AND RECOMMENDATIONS

This study assessed the use of language about sound

quality associated with the technique of open throat.

Although the 15 participants offered 18 terms todescribe sound quality, clear associations or simila-

rities in their usage and application were apparent.

This study advances previous work in clarifying

terminology related to open throat (51, 52). Commu-

nication of techniques in singing pedagogy can be

improved by attempts to inform the use of terminol-

ogy. Consensus of pedagogues is displayed in the

evaluations of singing performance (33), although theway in which pedagogues interpret a sound is often

difficult to define verbally.

It also expands Wapnick and Ekholms’ (26)

attempts at matching pedagogical evaluation with

perceptual descriptors. Singing pedagogues rely on

their perceptions of sound qualities to determine the

physiological processes at work in the production of

the sound quality. The human ear must complementacoustical study in integrating the complex dimensions

of the human voice. Language in singing pedagogy,

despite Seashore’s (2) ideals, is destined to remain

subjective. However, in order to spread information

more effectively, we would make the following recom-

mendations:

1. Singing pedagogy and voice research need to

become more integrated. De-constriction is a

more anatomically correct term than retraction.

This would help to convey meaning more effec-

tively between science and art.2. Qualitative research is a useful tool for the

generation of research questions and elucidation

of study related to the voice, in particular, to the

sound quality.

3. Subsequent research needs to establish the asso-

ciated characteristics of open throat and the

perceptual quality accompanying this technique

(26, 27).4. In this study, open throat was considered impor-

tant in current pedagogy. Pedagogues were aware

of the technique’s value as well as the need to

tailor their terminology and instructions in the

singing studio to each student’s vocal needs andlearning styles.

REFERENCES

1. Jorgenson D. A History of Conflict. NATS Bulletin1980; 36: 31�/5.

2. Seashore CE. Psychology of Music. New York andLondon: McGraw-Hill Book Company; 1938.

3. Bartholomew W. A physical definition of ‘‘good voicequality’’ in the male voice. J Acoust Soc Am 1934; 6:25�/33.

4. Tosi PF. Opinioni de’cantori antichi, e moderni, o sienoOsservazioni sopra il canto figurato 2nd edn. London:JWilcox: 1743

5. Puritz E. The teaching of Elisabeth Schumann. London:Methuen; 1956.

6. Herbert-Caesari EF. The voice of the mind. London:Robert Hale; 1951.

7. Agricola JF. In: Julianne C Baird (ed.). Introduction tothe art of singing; translated. Cambridge: CambridgeUniversity Press; 1995.

8. Vennard W. Singing: the mechanism and the technic, 5thedn. New York: Fischer; 1968.

9. Titze I. Voice Research: The Wide Pharynx. Journal ofSinging 1998; 55: 27�/8.

10. Titze IR, Story BH. Acoustic interactions of the voicesource with the lower vocal tract. J Acoust Soc Am1997; 101: 2234�/43.

11. Joiner JR. Charles Amable Battaille: pioneer in vocalscience and the teaching of singing. Lanham (Md):Scarecrow Press; 1998.

12. Ware C. Basics of vocal pedagogy: the foundations andprocess of singing. New York: McGraw-Hill; 1998.

13. Marafioti PM. Caruso’s Method of Voice Production.New York: Dover Publications; 1981.

14. McKinney JC. The Diagnosis and Correction of VocalFaults. Nashville, Tennessee: Broadman Press; 1982.

15. Reid CL. A dictionary of vocal terminology: an analysis.New York: Joseph Patelson Music House; 1983.

16. Estill J. Primer of Basic Figures, 2nd edn. Santa Rosa:Estill Voice Training Systems; 1996.

17. Miller R. On the art of singing. New York: OxfordUniversity Press; 1996.

18. Hemsley T. Singing and imagination: a human approachto a great musical tradition. Oxford and New York:Oxford University Press; 1998.

19. Manen L. Bel canto: the teaching of the classical Italiansong-schools its decline and restoration. Oxford: OxfordUniversity Press; 1987.

20. Potter J. Vocal authority: singing style and ideology.New York: Cambridge University Press; 1998.

21. Garcia M. Hints on singing, New and revised edition.London: Ascherberg, Hopwood and Crew; 1894.

22. Shakespeare W. The art of singing: based on theprinciples of the old Italian singing masters, and dealingwith breath-control and production of the voice,together with exercises, Revised edition. Boston: OliverDitson; 1921.

23. Lehmann L. How to sing, New revised and supplemen-ted edition translated by Clara Willenbucher. New York:Dover Publications; 1993.



24. Miller R. The structure of singing: system and art invocal technique. New York and London: SchirmerBooks; 1996.

25. Citardi MJ, Yanagisawa E, Estill J. Videoendoscopicanalysis of laryngeal function during laughter. Ann OtolRhinol Laryngol 1996; 105: 545�/9.

26. Wapnick J, Ekholm E. Expert consensus in solo voiceperformance evaluation. J Voice 1997; 11: 429�/36.

27. Ekholm E, Papagiannis GC, Chagnon FP. Relatingobjective measurements to expert evaluation of voicequality in Western classical singing: critical perceptualparameters. J Voice 1998; 12: 182�/96.

28. Bunch M, Chapman J. Taxonomy of singers used assubjects in scientific research. J Voice 2000; 14: 363�/9.

29. Patton MQ. Qualitative evaluation and research meth-ods, 2nd edn. Newbury Park, California: Sage Publica-tions; 1990.

30. Smith JA. Semi-Structured Interviewing and QualitativeAnalysis. In: Langenhove LV (ed.). Rethinking Methodsin Psychology. London:SagePublications;1995. pp.9�/26.

31. Mathieson CM, Barrie CM. Probing the Prime Narra-tive: Illness, Interviewing, and Identity. Qual Health Res1998; 8: 581�/601.

32. Miles MB, Huberman MA. Qualitative Data Analysis:An Expanded Sourcebook. California: Sage Publica-tions Ltd; 1994.

33. Davidson JW, Coimbra D. Investigating performanceevaluation by assessors of singers in a music collegesetting. Musicae Scientiae 2001; 5: 33�/53.

34. Kokotsaki D, Davidson JW, Coimbra D. Investigatingthe Assessment of Singers in a Music College Setting:The Students’ Perspective. Research Studies in MusicEducation 2001; 16: 15�/32.

35. Lovetri J, Lesh S, Woo P. Preliminary Study on the Abilityof Trained Singers to Control the Intrinsic and ExtrinsicLaryngeal Musculature. J Voice 1999; 13: 219�/26.

36. Story BH, Titze IR, Hoffman EA. The relationship ofvocal tract shape to three voice qualities. J Acoust SocAm 2001; 109: 1651�/67.

37. Sundberg J, Gramming P, Lovetri J. Comparisons ofPharynx, source, formant, and Pressure Characteristicsin Operatic and Musical Theatre Singing. J Voice 1993;7: 301�/10.

38. Welch GF, Sergeant DC, MacCurtain F. Some PhysicalCharacteristics of the Male Falsetto Voice. J Voice 1989;2: 151�/63.

39. Sundberg J. Articulatory interpretation of the ‘‘singingformant’’. J Acoust Soc Am 1974; 55: 838�/44.

40. Zemlin WR. Speech and Hearing Science: Anatomy andPhysiology, 4th edn. Boston: Allyn and Bacon; 1998.

41. Hirano MS. K. Histological Color Atlas of the HumanLarynx. San Diego: Singular Publishing Group; 1993.

42. Hertegard S, Gauffin J, Sundberg J. Open and CoveredSinging as Studied by Means of Fiberoptics, InverseFiltering, and Spectral Analysis. J Voice 1990; 4: 220�/

30.43. Acker BF. Vocal Tract Adjustments for the Projected

Voice. J Voice 1987; 1: 77�/82.44. Roubeau B, Chevrie-Muller C, Lacau Saint Guily J.

Electromyographic activity of strap and cricothyroidmuscles in pitch change. Acta Otolaryngol 1997; 117:459�/64.

45. Hong KH, Ye M, Kim YM, Kevorkian KF, Berke GS.The role of strap muscles in phonation*/in vivo caninelaryngeal model. J Voice 1997; 11: 23�/32.

46. Vilkman E, Sonninen A, Hurme P, Korkko P. Externallaryngeal frame function in voice production revisited: areview. J Voice 1996; 10: 78�/92.

47. Shipp T. Vertical laryngeal position during continuousand discrete vocal frequency change. J Speech Hear Res1975; 18: 707�/18.

48. Sonninen A. The external frame function in the controlof pitch in the human voice. Ann NY Acad Sci 1968;155: 68�/90.

49. Boone DR, McFarlane SC. A critical view of the yawn-sigh as a voice therapy technique. J Voice 1993; 7: 75�/80.

50. Sonninen A, Hurme P, Laukkanen AM. The externalframe function in the control of pitch, register, andsinging mode: radiographic observations of a femalesinger. J Voice 1999; 13: 319�/40.

51. Burgin JC. Teaching singing. Metuchen (NJ): ScarecrowPress; 1973.

52. Monahan BJ. The art of singing: a compendium ofthoughts on singing published between 1777 and 1927.Metuchen (NJ): Scarecrow Press; 1978.



2.1 Conclusion

The similarities in pedagogues’ description of laryngeal gestures to achieve the

technique or sound quality and their consensus regarding its value, particularly to

classical singing, generated the hypotheses for subsequent acoustic studies. Pedagogues’

subjective descriptions of sound qualities indicated the difficulty in communication of

vocal pedagogy.

To date, research into the singing voice has described acoustic properties of voice and

its visual representation with few or no links to pedagogy. However, there are two

empirical studies which link acoustic measures with perceptual judgements and have

provided a basis for the projects in this thesis. One, by Wapnick and Ekholm (1997)

established 12 generally accepted perceptual criteria applied to the assessment of voice

quality in classical singing, and the other, by Ekholm, Papagiannis and Chagnon (1998)

applied four of these criteria (“appropriate vibrato”, “resonance/ring”, “color/warmth”,

and “clarity/focus”) and related them to objective measurements taken from acoustic

analysis of the voice signal. This thesis extends this work by investigating an individual

singing technique.

Hypotheses on the effects of open throat were generated from pedagogical beliefs and

practices derived in the first study (Mitchell, Kenny, Ryan, & Davis, 2003). These were

tested acoustically and perceptually in the second study and compared with factors of

optimal singing quality which have been identified in the literature and defined

acoustically. As open throat was considered an important component of good singing,

the following studies compare singing with and without application of open throat

48

technique to factors of good singing quality identified in previous voice science

literature. One such factor is vibrato.

49

3 Vibrato

There is now both a vast literature on vocal acoustic properties and the visual

representation of voice with many unanswered questions about what conclusions can be

drawn from such analyses. Current singing pedagogy texts absorb these findings into its

descriptions of voice (Callaghan, 2000; Callaghan & Wilson, 2004; Miller, 1996; Nair,

1999; Sundberg, 1988; Thurman & Welch, 2000), but these in turn lead to many

unanswered questions about what conclusions can be drawn from the scientific voice

literature (Callaghan, 2000; Miller, 1998).

Empirical singing studies define ideal vibrato parameters of rate, extent and onset by

singing genre, singing style and decade of study. Such studies explore the optimal

parameters for good vibrato, but most of these produce means based on multiple

singers’ data. Each parameter varied significantly across studies and results were related

to the decade of study, the style of singing and the musical stimuli performed. Studies in

vibrato rate (Horii, 1989) or extent (Seidner et al., 1995) in individual singers have been

unable to define consistent intra- or inter-singer vibrato parameters in professional

singers. Prame’s (1994) detailed model of vibrato identified the unique factors of

vibrato rate: intra-tone (features of vibrato cycles within an individual tone), intertone

(comparisons of vibrato rates in the 25 tones analysed for each artist) as well as inter-

artist (variations of mean rates across artists). Loudness or differences in SPL have been

associated with changes in vibrato parameters (Michel & Myers, 1991; Titze et al.,

1999), yet pedagogy perceives an evenness in vibrato across dynamic changes (Miller,

1996).

50

Few studies compare the same singers’ vibrato parameters in different tasks. While

vibrato is believed to be intrinsic to a singer’s sound, it does change across musical

styles (Easley, 1932; Hakes, Shipp, & Doherty, 1987). Differing emotional instructions

have affected vibrato parameters (Rothman & Arroyo, 1987; Sundberg, 1997), but to

date, no study has assessed the impact of a specific singing technique on vibrato, rather

than respective vibrato parameters of a group of singers. A singer’s technical intention

may play as great a role in the production of vibrato as any musical or stylistic changes

they make to their singing.

Six advanced singing students sang in two conditions: “optimal” (O), using maximal

open throat, “sub-optimal” (SO), using reduced open throat and “loud sub-optimal”

(LSO) to control for the effect of loudness. Studies that compare different vocal

instructions have identified changes in overall sound pressure (SPL) or power levels as

a direct result of an experimental condition (Foulds-Elliott et al., 2000; Rossing,

Sundberg, & Ternstrom, 1986, 1987; Thorpe et al., 2001). Therefore, the LSO condition

was included, in accordance with Foulds-Elliot et al. (2000) to take into account

possible effects of loudness, which may impact on vocal timbre.

The first acoustic study analyses changes to vibrato parameters of rate, extent and onset

using maximum open throat and reduced open throat technique. Analyses revealed

significant reductions in vibrato extent and increases in onset time, but no change to

vibrato rate in SO compared to O. This study was the first to link changes in vibrato

parameters with the use of a technical instruction rather than a particular singing genre

or musical era. Results of this study were presented to an international music

51

conference, Music and Gesture [Mitchell, H. F., & Kenny, D. T., Evaluation of open

throat technique on classical singing, International conference presentation at Music and

Gesture, August 2003, East Anglia, UK].

52

PAPER 2

Mitchell, H. F., & Kenny, D. T. (2004). The impact of open throat technique on vibrato rate, extent and onset in classical singing.

Logopedics Phoniatrics Vocology, 29(3), 99-118.

53

The impact of ‘open throat’ technique on vibrato rate,extent and onset in classical singing


From the Australian Centre for Applied Research in Music Performance (ACARMP), The Conservatorium of Music, TheUniversity of Sydney, New South Wales, Australia

Received 8 September 2003. Accepted 3 August 2004.


Mitchell, Kenny et al . (2003) identified ‘open throat’ as integral to the production of an even and consistent sound in classicalsinging. In this study, we compared vibrato rate, extent and onset of six advanced singing students under three conditions:‘optimal’ (O), representing maximal open throat; ‘sub-optimal’ (SO), using reduced open throat; and loud sub-optimal (LSO),using reduced open throat but controlling for the effect of loudness. Fifteen expert judges correctly identified the soundproduced when singers used open throat with 85% accuracy. Having verified the technique perceptually, we used a series ofunivariate repeated measures ANOVAs with planned orthogonal contrasts to test the hypotheses that frequency modulationsassociated with vibrato rate, extent and onset would vary outside acceptable or desirable parameters for SO and LSO.Hypotheses were confirmed for vibrato extent and onset but not for rate. There were no significant differences between SO andLSO on any of the vibrato parameters. As vibrato is considered a key indicator of good singing, these findings suggest thatopen throat is important to the production of a good sound in classical singing.

Key words: open throat, singing technique, vibrato, vocal pedagogy.

Helen Mitchell, Australian Centre for Applied Research in Music Performance (ACARMP), The Conservatorium of Music, TheUniversity of Sydney, New South Wales, Australia 2006. Tel.: �/61 2 9351 9644. Fax: �/61 2 9351 9540. E-mail:[email protected]

INTRODUCTION

The principal goal of classical and operatic singing

pedagogy is the production of a voice of quality that is

achieved through instruction in specific vocal techni-

ques and exercises to achieve the desired overall

sound. Instruction in the use of ‘open throat’ techni-

que to achieve a good sound is widespread (1�/4).

Historically, the term ‘open throat’ can be traced

throughout the vocal pedagogy literature (5�/7) as a

way of describing freedom or lack of tension in the

area of the throat, resulting in a lack of constriction

and a better tone quality. As early as 1935, Bartholo-

mew commented: ‘If the various tricks of the trade

that voice teachers use to improve quality are ana-

lyzed, most or all of them will be found to be a device

for directly or indirectly enlarging the throat.’ (8)

According to the singing literature, ‘open throat’ is a

complex maneuver that involves a pedagogical in-

struction and a perceived sensation or action that

results in a specific sound quality. Previous pedagogi-

cal strategies have linked ‘open throat’ to an indirect

action in the preparation to sing, on inhalation in the

imposto (9, 10), appoggio or as part of ‘breath

management’ (3, 11), or in the surprise breath or

smelling the rose (3, 9, 12). Pedagogues also used

imagery though visualization of space within the

throat, through an ‘air ball’ or ‘soap bubble’ to achieve

the posture of ‘open throat’. The action of yawn (13,

14) or yawn-sigh (2) was recommended, but pedago-

gues cautioned against yawning to the point of distor-

tion of sound through unnecessary tension (2, 3).

Perceptually, open throat produces a tone ‘free of

constrictor tensions’ ((2), p. 83). The sound quality

attributed to open throat is perceived in resonance

(3, 15), roundness (16), freedom (17), purity (18),

richness and warmth (19). Use of the technique

facilitates technical dexterity and better use of other

techniques, such as ‘ring’ (1, 2).

There are differing pedagogical perspectives of the

vocal mechanics involved in ‘open throat’ technique.

Pedagogues who explain their pedagogy in physiolo-

gical terms link their preferred strategy to achieving

open throat to lowered larynx, pharyngeal width and

raised soft palate (3, 11, 15, 20). Singing pedagogy also

hypothesized that ‘open throat’ counters constriction



DOI: 10.1080/14015430410001033 54

which may result in a reduction of pharyngeal space

(2). Pedagogues generally understand that open throat

is not necessarily an ‘enlarged pharynx’ or action that

may excessively depress the hyoid bone, tongue (2) or

larynx (3) as that may cause tension, and limit

resonance in the sound quality.

A recent study examined current understanding and

use of open throat in the classical singing studio.

Mitchell et al . (4) interviewed 15 singing pedagogues

and results indicated that all 15 pedagogues described

‘open throat’ technique as fundamental to singing

training. Although some pedagogues preferred differ-

ent terms, such as ‘retraction’, ‘freedom’ or ‘collar’,

they were in fact describing the same underlying

phenomenon of ‘open throat’. They recognized it in

the final sound as balanced and coordinated, free,

even or consistent, warm and open. Expert pedago-

gues in this study believed that open throat maximizes

pharyngeal space and/or abducts the ventricular folds

(4). While some vowels, for example [a], cannot be

produced without pharyngeal constriction (21) and

hence pharyngeal space cannot be maximized to the

same degree on every vowel, in the bel canto (Western

classical singing) tradition, the ideal is to achieve the

same warm, open voice quality on every note (10).

Story et al . (22) using MRI (magnetic resonance

imaging) techniques have confirmed that different

vocal postures are associated with different vocal

qualities such as ‘yawny’ and ‘twangy’. ‘Yawny’

involves widening of the oral cavity and lengthening

of the vocal tract while ‘twangy’ is produced by slight

constriction of the oral cavity and shortening the vocal

tract. Open throat is related to the ‘yawny’ quality

(2, 11) but pedagogically the yawn is considered too

exaggerated for classical singing. A pre-yawn or cry/

sob instruction is preferred by most pedagogues (2, 3,

23, 24). Story et al . (22) analyzed these physiologically

established vocal qualities acoustically but not percep-

tually, but recommended that perceptual assessment

was necessary to complement the acoustic and phy-

siological findings.

It may be possible to understand good classical

singing through acoustic features associated with

subjective descriptions of acceptability or beauty in

singing. Vibrato, for example, is widely accepted as a

component of singing, but despite acoustical, physio-

logical and perceptual studies (15, 25�/28), it has been

difficult to define its most desirable parameters. It is,

however, possible to trace different preferences over

the century (29) and also to assess its defects or

undesirable qualities (3, 15). However, acoustic studies

of undesirable qualities of vibrato have not achieved

consensus in their definitions (30, 31).

In this study, we investigated the technique of open

throat by identifying acoustic differences from the

same singers when they consciously manipulated the

technique in their singing. As pedagogues in a

previous study by these authors (4) reported thatsingers normally use open throat in good singing, and

were conscious of using the technique, we believed that

experienced singers could manipulate their use of this

technique. The aim of this study was 1) to test the

perceptual distinctiveness of the sound produced by

singers using open throat and 2) to determine the

measurable differences between a sound quality using

maximal open throat and a sound that used a lesserdegree of open throat, while holding other aspects of

singing technique or sound quality constant.

METHOD

Participants

Six female singers (three sopranos and three mezzo-

sopranos) volunteered to participate in this study.

They were advanced students with excellent techniqueof an experienced singing pedagogue, who is a

Lecturer in Vocal Studies and Opera at a state

Conservatorium of Music in Australia. It is the

premier institution for musical education in the

country and has produced singers of international

repute. The pedagogue had 30 years’ teaching experi-

ence, a private studio consisting of international and

national singers (32) and is considered to be one of thetop five singing teachers in Australia. Criteria for

participant selection included, through this pedago-

gue’s assessment, singers who: 1) had a good classical

singing technique for their level of training and

experience; and 2) understood and demonstrated

skillful control of ‘open throat’ or ‘retraction’ techni-

ques in their singing. The institutional human ethics

committee approved the study.Prior to the voice recording, participants completed

a questionnaire seeking information on age, years of

singing study, number of years of study with each

singing teacher and highest qualifications attained or

currently undertaken in music and/or singing, and

singer type (soprano or mezzo-soprano). The partici-

pants were also asked to classify the genres of singing

they performed in public (opera, classical, choral,music theatre and contemporary) in accordance with

the Bunch and Chapman taxonomy (32) and to

estimate the percentage each style played in their total

performing career.

The demographic information of the participants is

presented in Table 1. Participants were aged between

23 and 30 years, with a mean of 26 years. All had

studied singing for at least 7 years (average 9.8 years)and had spent an average of 5 years studying with

their present singing teacher. Each singer held a

172 H. F. Mitchell and D. T. Kenny


qualification in singing or music (four had Bachelor of

Music degrees and two had diplomas, in music and/or

singing) and five of six were currently undertaking asecond degree in singing (three postgraduate Diploma

of Opera and two Bachelor of Music degrees). All

defined the majority of their singing as operatic

(�/50%), with the second most common style classical

(�/20%), in accordance with the Bunch and Chapman

taxonomy (32) of singing voices. All reported that they

were in good health and able to perform the tasks.

Procedure

Singers were sent information about the project and

were invited to take part in an acoustic and perceptual

study of singing technique. They were required to

attend a single recording session lasting approximately

one hour. They were advised that the object of the

study was to investigate the acoustical and perceptual

features of the use of open throat in singing and toinvestigate the sound qualities associated with the use

of open throat.

Experimental protocol for assessing sound qualities

produced by open throat

A protocol was developed to assess the effect of ‘open

throat’ technique in singing. Two musical tasks werechosen to investigate the use of the technique in two

song excerpts. Prior to singing, each singer selected the

sequence of their tasks before commencing the experi-

ment by selecting a blank card, the reverse side of

which represented one of the tasks, to reduce the

possible effects of task-order.

Singers sang each of the two song excerpts under

three conditions: optimal, sub-optimal and loud sub-optimal. ‘Optimal’ (O) condition was necessary to

provide a sound quality with the best technique that

they could. This involved the maximal use of open

throat. ‘Sub-optimal’ (SO) condition involved the use

of a reduced (open throat) technique but still with an

acceptable singing technique and without consciously

altering any other aspect of their technique. It was

hypothesized, from interviews with pedagogues (4),that the SO condition would result in a reduction of

sound pressure level (SPL), so a third condition ‘loud

sub-optimal’ (LSO) involved the same instruction as

the SO condition, but with the added instruction that

the singer should try to achieve a louder dynamic than

in SO. This addressed SPL as an additional variable to

the SO condition and was similar to technical condi-

tions established by Foulds-Elliot et al . (33) when SPLwas a variable of degrees of emotional connection in

operatic singers.Tab

le1.

Su

mm

ary

of

sub

ject

cha

ract

eris

tics

.D

escr

ipti

on

of

sub

ject

nu

mb

er,

age,

the

nu

mb

ero

fyea

rsea

chsu

bje

cth

as

tak

ensi

ng

ing

less

on

s,th

en

um

ber

of

tea

cher

sea

chsu

bje

cth

as

ha

d,

the

nu

mb

ero

fyea

rsea

chsu

bje

ctst

ud

ied

wit

hth

eir

thre

em

ost

rece

nt

tea

cher

s(

T1� /T

3)

,th

eh

igh

est

leve

lo

fed

uca

tio

nco

mp

lete

dby

each

sub

ject

,th

ecu

rren

tleve

lo

fed

uca

tio

nu

nd

erta

ken

,vo

ice

typ

e,p

rop

ort

ion

of

each

sub

ject

’sca

reer

spen

tsi

ng

ing

op

era

,cl

ass

ica

l,ch

ora

l,m

usi

cth

eatr

e,

con

tem

po

rary

or

oth

erst

yle

so

fm

usi

c.

Su

bje

ctA

ge

(yea

rs)

Yea

rso

fle

sso

ns

No

.o

fte

ach

ers

T1

T2

T3

Hig

hes

ted

uca

tio

nC

urr

ent

edu

cati

on

Vo

ice

typ

eO

per

aC

lass

ica

lC

ho

ral

Mu

sic

thea

tre

Co

nte

mp

Oth

er

12

68

41

13

Ba

chel

or

Ba

chel

or

Mez

zo8

0%

20

%2

25

73

1.5

34

.5B

ach

elo

rP

ost

gra

du

ate

So

pra

no

80

%2

0%

33

01

22

10

2D

iplo

ma

N/A

So

pra

no

97

%3

%4

23

10

24

5.5

Ba

chel

or

Po

stg

rad

uat

eS

op

ran

o5

0%

45

%5

%5

25

10

27

.52

.5D

iplo

ma

Po

stg

rad

uat

eM

ezzo

73

%2

5%

2%

62

81

26

51

.51

.5B

ach

elo

rB

ach

elo

rM

ezzo

50

%3

5%

5%

10

%

Evaluating open throat 173


Each task was performed twice in the O and SO

conditions while the LSO condition was performed

only once. In total, each singer performed eachmusical task five times.

Instructions

A pedagogue was present during the recording ses-

sions to provide accompaniment for warm-ups and

practice of the tasks where necessary and to instruct

singers to achieve the required vocal postures for eachexperimental condition. For example, she instructed

them to pay attention to producing the most open

sound in their throat in the O condition, and a lesser

degree in the SO and LSO conditions. Some singers

asked how to produce the LSO condition and the

pedagogue instructed them to ‘use more twang’, as

taught in their lessons (23).

Reliability check

Prior to the commencement of the study, the partici-

pating students and pedagogue had several

practice sessions in which the singer was instructed

to use either O or SO at random. The pedagogue

indicated which technique she thought the singer

had used. Practice with each singer continued until

both the singer and the pedagogue reached 100%agreement on the occasions that the technique was

applied.

The musical tasks

Musical tasks were chosen in order to test different

demands of good singing, but were not musically

difficult. They were designed to test the use of openthroat, and contained musical features derived from a

previous qualitative study on the use of the technique,

where use or lack of the technique was deemed to be

particularly valuable or noticeable (4). These features

were: high tessitura, sustained or legato singing,

dynamic range control, and vocal agility.

The Mozart song Ridente la Calma , K 152, bars 1�/

27 (Fig. 1a) was selected as it is a nominally simplesong in the Italian language (34) with a mixture of

common musical statements involving repeated legato

lines as well as the initial stylized leaps of a major 4,

and short scale figures. All six singers sang this aria in

the same key (F-major).

The third verse of the Schubert lied, Du bist die Ruh

D. 776 (Op. 59, No. 3) (Fig. 1b), bars 54�/80 was

chosen for its demanding vocal control, sustainedmusical line and high climactic tessitura. The three

sopranos and three mezzo-sopranos sang this in an

appropriate key depending on soprano or mezzo-

soprano voices (E-flat, D-flat and C-major).

Recording

Participants were given time to warm up in the singing

studio and become familiar with the room before

recording. Recording levels for each singer were set

during this time. The acoustic signal was recorded

digitally (Behringer Ultragain preamplifier/Marantz

CDR 630) via a high-quality microphone (AKG C-

477) positioned on a head boom a constant 7 cm

distance from the subject’s lips. This ensured that

direct energy of the performers’ voices was recorded

rather than room reflections, enabling us to use a

studio environment with low ambient noise rather

than an anechoic studio (35).

Fig. 1a. The Mozart task, Ridente la Calma , K 152,

bars 1�/27 with the numbering of 37 notes selected for

analysis.

Fig. 1b. The Schubert task, Du bist die Ruh D. 776

(Op. 59, No. 3), bars 54�/80 with the numbering of 20

notes selected for analysis.



Calibration was carried out in each recording by

using a recording of two pink noise samples on

audiotape played immediately following each record-ing session at the same recording gain used for

recording the singer’s voice. Pink noise is a broadband

steady state signal which enables calibration of an

audio playback system across the frequency range for

tasks such as subjective testing and, if the analysis is

just done on computer, the noise is at least as good as

any other steady-state known signal. For calibration of

absolute sound pressure levels (SPL), the microphoneof a sound level meter (Rion NL-06 SPL) was held

adjacent to the AKG microphone from a speaker

(Bose Lifestyle) from which the pink noise was played.

The SPL (in dB at 7 cm) recorded by the sound level

meter was noted for the two noise signals and used

later for calibration.

The CDs of the recordings were marked by

subject number. The audio recordings were acquiredto computer at 16 kHz with a Loughborough Sound

Images PC/C32 board using Phog Version 2.0

and analyzed using Soundswell Version 4.00 (Hitech,

Sweden) software to produce nine channels of raw

and calculated data. A real-time digital SPL meter

was used to set a recording gain so that the pink

noise calibration level recorded in Phog was the

same (�/ 0.5 dB) as the SPL level noted duringrecording.

Perceptual test of open throat

Because open throat is a pedagogical concept that has

not yet been verified empirically, it was necessary to

determine whether expert pedagogues could reliably

distinguish the sound quality associated with this

instructional technique. Accordingly, a perceptual

test was designed to determine the degree to whichexpert pedagogues agreed that a sound was indicative

of the application of open throat. Judges who assessed

singers were not the singers’ pedagogues and did not

know any of their identities.

Listeners assessed 30 tracks (6 singers�/2 condi-

tions (O and SO)�/2 musical tasks) in random order,

including six repeats, also selected at random. (A pilot

test found that LSO did not produce a sufficientlydiscernible voice quality to be correctly distinguished

from SO. As LSO was created as acoustic confirma-

tion of SO, it was considered unreasonable to expect

the human ear to detect so small a difference in timbre.

The LSO condition was therefore not included in the

perceptual study.)

The perceptual test was conducted in a quiet

environment and samples were played from a SonyCD Walkman (DEJ885W) via circum-aural closed-

back stereo monitoring headphones (Sennheiser HD

270). Prior to presentation of stimuli, participants

were given information on the two singing conditions,

O and SO, and were presented with the musical scoreof each musical task. They were asked to judge

whether or not the singer was using open throat

technique in each sample.

Measurement of sound pressure level (SPL)

To calculate mean SPL, the SPL channel of the

Soundswell program was calibrated with the two

known SPL levels of the pink noise samples for eachsinger. The histogram function was applied to the SPL

channel in the Soundswell program for each complete

rendition of musical task and generated the mean,

standard deviations (SD), and percentile values of

SPL in dB.

Vibrato measurement

Vibrato cycles were measured manually from the

spectrogram feature in the Soundswell program using

a fine resolution Hamming analysis window, FFT

length 1024, with a bandwidth of 30 Hz. A cursor

point logged the time and frequency of each vibrato

peak and trough and the data exported to spread-

sheets. Frequency was measured on a high harmonic

across all subjects, as resolution increases with thepartial number (26). Fundamental frequency of each

logged point was calculated by dividing it by its partial

number. Data were manipulated within spreadsheets

(Excel, Microsoft 2000). Thirty-seven notes were

measured in the Mozart task and 20 in the Schubert.

The vibrato onset (VO) in seconds (s) was measured

from the initiation of phonation on the particular note

until the first conclusive peak of the vibrato cycle.Onset was only considered following the start of a

word or after a break in phonation; 18 VO occurrences

were studied in the Schubert task and 23 in the Mozart

task.

The vibrato rate (VR) in hertz (Hz) and vibrato

extent (VE) in semitones (ST) were measured from the

first identifiable peak until the last trough of the note.

The VR of each cycle was measured as the timedifference between adjacent vibrato peaks and calcu-

lated in Hz. The VR for each note was calculated as

the mean of all successive VR cycles of the note. If

vibrato peaks and troughs could not be discerned by

visual inspection of the spectrograph, a nil result was

recorded.

Vibrato extent in semitones (ST) was calculated by

dividing the peak-to-trough ST values by 2 (28). Afrequency point in Hz of a peak or trough was

calculated in ST using the formula:



F0 in ST�12(log10(F0) � log10 16:35)

log10 2[1]

The semitone difference between adjacent cycle

peaks and troughs was calculated as:

Peak-to-trough(ST)

�12(log10(peak F0) � log10(trough F0))

log10 2[2]

In this study, VE in ST represents the distance from

peak to the mean F0 of the vibrato cycle or from the

mean F0 of the cycle to the next trough (28).

Frequency values from successive vibrato cycles wereaveraged for each note studied.

Study 1: Perceptual verification of open throat. The

total number and percentages of correct responses foreach O and SO sample for each listener and for the

group as a whole were calculated. Intra-rater reliability

was also assessed by calculating the percentage of each

repeated pair that was correctly identified by each

judge.

Study 2: Vibrato parameters of open throat. The

study design was a repeated measures (three depen-

dent measures: vibrato rate (VR); vibrato extent (VE);

vibrato onset (VO)) randomized complete block with a

2 (task (Mozart versus Schubert))�/3 (condition

(optimal, sub-optimal and loud sub-optimal)) factor-

ial structure with planned orthogonal contrasts oneach of the three dependent variables.

Data were analyzed firstly through a series of fixed

and mixed effect univariate ANOVAs with three

factors (task, condition, subject). In the first series,

all three factors were entered as fixed effects; in the

second, subject was entered as a random effect.

Finally, the data were subjected to analysis using the

general linear model (GLM) with planned orthogonalcontrasts. Since all analyses yielded the same out-

comes, only the results of the contrasts are presented.

Main effects for task and condition and interaction

effects (contrasts) were calculated for each dependent

measure (VR, VE and VO). In the first contrast, O was

compared to SO and LSO and in the second contrast

SO was compared to LSO.

Hypotheses

The following hypotheses were generated:

1) Study 1: Expert judges would reliably identify the

use of open throat in samples of classical singing.

2) Study 2: SPL would decrease from O to SO.3) Study 2: Frequency modulations associated with

vibrato rate, extent and onset would vary outside

acceptable or desirable parameters established in

singers’ O singing for SO and LSO.

RESULTS

Hypothesis 1: Perceptual verification of open throat

Expert listeners correctly identified 85.3% O and

80.0% SO. Intra-judge reliability using the six repeated

samples was 83.33% (that is to say, judges, on average,

rated five out of six singing sample pairs correctly).

Hypothesis 2: SPL would decrease from the O to SO

condition

Hypothesis 2 stated that SPL would decrease from the

O to SO condition. Hypothesis 2 was not confirmed

from O to SO F(1,5)�/0.010, p�/0.924). The mean SPL

for O was 97.38 dB (SD 1.86) in the Mozart task and97.65 dB (SD 2.61) in the Schubert task. In the SO

condition, the mean SPL was lower (96.18 dB in

Mozart and Schubert). In LSO, each singer’s SPL was

associated with SPL levels that were either similar to

the O condition (6 examples B/2 dB different) and in 9

of the 12 LSO, samples increased SPL from the level

of the original O in the LSO condition (LSO means:

99.18 dB, 98.92 dB).

Vibrato

Data screening. The distributions for each depen-

dent measure (VR, VE, and VO) were assessed for

normality and outliers. Examination of skewness and

kurtosis statistics indicated that the distributions were

relatively normally distributed. Univariate tests for

homogeneity of variance (Levene’s Test of Equality of

Error Variances) for each of the dependent measures

(VR: F�/5.7, p B/0.001; VE: F�/2.65, p B/0.001; VO:F�/9.699, p B/0.001) indicated that this assumption

was violated for all three measures. Univariate F-tests

(with Bonferroni adjustment for three tests: p B/0.017)

indicated that task, condition and subject were all

significant, hence a very conservative p-value of p B/

0.005 was set for all subsequent analyses.

Descriptives. The complete vibrato data are pre-

sented in Table 2, showing the overall means from

averaging all measured notes sung by each singer, in

each condition, for vibrato rate (VR), vibrato extent

(VE) and vibrato onset (VO) in each musical task

(Mozart and Schubert).Differences in VR, VE and VO were considered

separately for effects of condition and task.



Hypothesis 3: Frequency modulations associated with

vibrato rate, extent and onset would vary outside

acceptable or desirable parameters established in

singers’ O singing for SO and LSO

Hypothesis 3a: For vibrato extent, singers’ F0 would

show greater variation from the mean in SO/LSO

compared to O

Hypothesis 3a stated that VE would show greaterincreases or decreases from mean fundamental fre-

quency in SO and LSO compared to O. Hypothesis 3a

was confirmed as overall means of VE by condition

indicated a significant reduction from O to SO/LSO in

both musical tasks. For VE, there was a main effect for

contrast 1 (F(1,5)�/20.821, p�/0.006) but not for

contrast 2 (F(1,5)�/1.448, p�/0.283). There was no

effect for musical task (F(1,5)�/0.032, p�/0.865). Fig. 2presents these data. There was also a significant

interaction between musical task and contrast 1

(F(1,5)�/12.136, p�/0.018), but not for musical task

and contrast 2 (see Fig. 2); that is, VE was wider for

both Mozart and Schubert in the O condition, but the

same for both musical tasks in the SO and LSO

conditions. Table 2 shows that all subjects’ mean VE

decreased in both the SO and LSO conditions by 0.09

ST to 0.92 ST, with an average reduction of 0.50 ST.

Inspection of the narrow standard deviations con-

firmed that singers’ reductions of VE in SO and LSO

were similar for all subjects. Although VE in semitones

changed overall in each condition, the singers’ varia-

bility remained slight (SD5/0.25) for notes within a

condition. The time plots in Fig. 3 plot the consecutive

VE of the notes measured for vibrato cycles in each

musical task (37 in Mozart, 20 in Schubert) in the O

and SO conditions for subjects 4 and 5. These present

data from subject 4, who showed a typical decrease

from O to SO, and subject 5, who showed a smaller

reduction in VE. In some SO vibrato cycles, there was

no measurable VE. Subjects 3 and 4 maintained the

Table 2. Mean vibrato rates in Hz, mean vibrato extent in semitones and mean vibrato onset in seconds, for the

Mozart and Schubert musical tasks in each condition, optimal (O), sub-optimal (SO) and loud sub-optimal (LSO);

standard deviations (SD) are given in parentheses.

Rate Extent Onset

Condition Condition Condition

Task Subject O SO LSO O SO LSO O SO LSO

Mozart 1 5.96 5.93 5.92 1.16 0.68 0.63 0.09 0.22 0.21(0.22) (0.28) (0.40) (0.15) (0.14) (0.16) (0.07) (0.17) (0.17)

2 5.63 5.32 5.36 1.26 0.77 0.73 0.15 0.16 0.21(0.28) (0.40) (0.29) (0.14) (0.16) (0.15) (0.17) (0.17) (0.09)

3 5.58 5.58 5.58 1.63 0.81 0.81 0.19 0.24 0.24(0.40) (0.29) (0.27) (0.16) (0.15) (0.21) (0.17) (0.09) (0.12)

4 5.65 5.60 5.51 1.48 0.60 0.60 0.12 0.28 0.27(0.29) (0.27) (0.31) (0.15) (0.21) (0.17) (0.09) (0.12) (0.10)

5 6.80 6.78 5.51 1.00 0.72 0.60 0.09 0.22 0.27(0.27) (0.31) (0.24) (0.21) (0.17) (0.20) (0.12) (0.10) (0.07)

6 7.27 6.28 6.63 0.85 0.52 0.67 0.10 0.27 0.25(0.31) (0.24) (0.36) (0.17) (0.20) (0.24) (0.10) (0.07) (0.11)

Mean 6.15 5.93 5.97 1.23 0.68 0.67 0.12 0.23 0.24(0.71) (0.53) (0.61) (0.29) (0.10) (0.08) (0.04) (0.04) (0.03)

Schubert 1 5.79 5.57 5.63 1.00 0.58 0.61 0.12 0.56 0.63(0.31) (0.33) (0.28) (0.20) (0.15) (0.15) (0.13) (0.43) (0.47)

2 5.23 5.06 5.18 1.32 1.19 0.88 0.22 0.26 0.42(0.32) (0.14) (0.14) (0.14) (0.21) (0.25) (0.10) (0.12) (0.16)

3 5.68 5.67 5.63 1.32 0.60 0.51 0.13 0.32 0.46(0.23) (0.26) (0.44) (0.25) (0.19) (0.15) (0.09) (0.27) (0.44)

4 5.66 5.70 5.65 1.44 0.61 0.58 0.13 0.22 0.10(0.23) (0.49) (0.37) (0.24) (0.19) (0.16) (0.10) (0.17) (0.07)

5 6.70 6.77 6.68 1.07 0.85 0.71 0.18 0.18 0.23(0.25) (0.26) (0.51) (0.19) (0.18) (0.12) (0.08) (0.13) (0.16)

6 5.96 5.50 5.88 0.90 0.77 0.81 0.18 0.20 0.15(0.73) (0.52) (0.75) (0.20) (0.14) (0.19) (0.10) (0.12) (0.11)

Mean 5.84 5.71 5.78 1.18 0.77 0.68 0.16 0.29 0.33(0.49) (0.57) (0.50) (0.22) (0.23) (0.14) (0.04) (0.14) (0.20)



two highest mean VEs overall, but also the greatest

variance (SD 0.25 and 0.24 ST respectively) but only inthe Schubert task (Table 2). The greatest VE seems to

be in the Mozart task, not the Schubert task (Table 2).

Hypothesis 3b: Vibrato rate would show greater

variations in SO/LSO compared to O

For VR, the hypothesis that vibrato would show

greater variations in SO and LSO compared to O

was not confirmed for task or condition. Overall

means of VR were not significantly different in each

task (F(1,5)�/2.415, p�/0.181). There was no effect of

condition for contrast 1 (F(1,5)�/2.828, p�/0.153) or

contrast 2 (F(1,5)�/0.629, p�/0.464), that is, VR in O,

SO and LSO conditions was not significantly different.

There were no interaction effects for task and condi-

tion. Fig. 4 presents the results of this analysis

graphically.

The mean VR across all subjects, conditions and

tasks was 5.92 Hz (SD 0.11). Four of the subjects

displayed a relatively similar mean VR across all

conditions and tasks (5.3�/5.8 Hz), whereas two

Condition

LSOSOO

Ext

ent i

n se

mito

nes

1.4

1.2

1.0

0.8

0.6

Mozart

Schubert

Fig. 2. Vibrato extent. Main effect of condition incontrast 1, that is to say, between optimal (O) and sub-

optimal (SO)/loud sub-optimal (LSO). Interaction

effect of musical task (Mozart and Schubert) for

contrast 1, that is, between optimal (O) and sub-

optimal (SO)/loud sub-optimal (LSO).

Condition

LSOSOO

Rat

e in

Hz

6.5

6.0

5.5

5.0

Mozart

Schubert

Fig. 4. Vibrato rate: no statistical effects of task

(Mozart and Schubert) or condition (optimal (O)

and sub-optimal (SO)/loud sub-optimal (LSO)).

0.

1.1.2.

0

050

0 5 10 15 20 25 30 35

0.0

1.01.52.0

0 5 10 15 20 25 30 35

0 5 10 15 200.00.5

0.5

1.1.2.

050

0 5 10 15 200.

1.1.2.

0

050

Mozart Note

Mozart Note

Schubert Note

Schubert Note

Exte

nt

in S

T

Exte

nt

in S

TE

xte

nt

in S

T

Exte

nt

in S

T

Optimal Sub-optimal

(a)

(b)

0.5

0.5

Fig. 3. Time plots of mean vibrato extent in semitones in the Mozart task, notes 1�/37 and the Schubert task in

the optimal and sub-optimal conditions: a�/subject 4; b�/subject 5.



subjects each showed a higher mean VR of 6.8 Hz as

shown in Table 2. Subject 6 was an outlier.

Hypothesis 3c: Vibrato onset would be longer in SO/

LSO compared to O

Hypothesis 3c stated that VO would be longer in SO/

LSO compared to O. Hypothesis 3c was confirmed as

overall means indicated a significant increase from O

to SO/LSO in both musical tasks. For VO, there was a

main effect for contrast 1 only (F(1,5)�/13.118, p�/

0.015), that is, VO was significantly less in the O

condition than in either SO or LSO. There were no

differences in VO between SO and LSO (F(1,5)�/0.905,

p�/0.385) and there was no effect for musical task

(F(1,5)�/1.559, p�/0.267) There were no interaction

effects between task and contrasts, and VO was

affected by condition in the same way in each task.

Fig. 5 presents the results of this analysis graphically.

A visual example of the changes to vibrato para-

meters by condition is given in Fig. 6. Each panel of

Fig. 6 represents the same singer and the same 3-

second portion of ‘Von deinem Glanz’ in the Schubert

musical task across each condition (O, SO and LSO).

The vibrato cycles are less regular or consistent in each

of the SO and LSO conditions, and in the LSO in

particular, the cycles are erratic and unstable.

There was no vibrato on 75 of a total of 684 notes

chosen for measurement. In the Mozart condition, 73

notes had no vibrato and 46 of these occurred in SO.

An example of this is shown in Fig. 6c.

DISCUSSION

We hypothesized that open throat technique would

produce audible differences in vocal quality and this

was confirmed by the perceptual study. Listeners were

reliable in their judgements and consistently identified

the technique in repeated samples. Singers’ vocal

quality produced discernable sound quality differences

in 85.3% of O and 80.0% of SO samples, which

listeners attributed to the pedagogical instruction of

‘open throat’. This study confirmed pedagogues’

recognition of a fundamental technique in classical

singing previously defined by expert pedagogues by a

specific sound quality (4). Further analysis of the

perceptual qualities of this technique is the subject of

ongoing research by these authors.

This study had a number of methodological

strengths. For example, we considered it important

Condition

LSOSOO

Ons

et in

sec

onds

0.4

0.3

0.2

0.1

Mozart

Schubert

Fig. 5. Vibrato onset: main effect of condition in

contrast 1, that is to say, between optimal (O) and sub-

optimal (SO)/loud sub-optimal (LSO).

Fig. 6a�/c. Examples of vibrato from subject 3 singing

‘Von deinem Glanz’ in each of the three conditions,

optimal, sub-optimal and loud sub-optimal (O, SO,

LSO) respectively.



to demonstrate that voice quality produced during SO

was not simply the result of changes (that is to say,

reductions) in SPL. The inclusion of a second

comparison group, LSO, indicated that SPL did not

account for differences between O and SO. Foulds-

Elliot (33) and Rossing (36) have used similar study

designs. The hypothesized reduction in SPL did not in

fact occur; however, there were differences in the

energy distributions of each condition that will be

explored further in future studies on this technique. A

second methodological strength was the use of a

homogenous group of singers of similar skill and

training who could consistently apply the technique.

While other studies have used the same sex subjects

(37, 38), singers in previous studies have ranged from

novice to advanced students. Studies assessing voice

qualities in which subjects vary in amount of singing

training or experience associate certain qualities with

levels of training, and find differences in vibrato or

energy which could be attributed to training. In some

cases, singing studies are unable to correlate length of

training with improved sound quality or any measur-

able acoustic feature of voice (39), suggesting it is not

length of training per se but quality or type of training

that is the critical factor.

Having established that the overall sound quality of

open throat technique is perceptually identifiable, we

subsequently assessed vibrato, a major defining fea-

ture of vocal quality. Although vibrato is taken for

granted as a feature of the classical singing voice, as a

component of tone quality, ‘richness’ (15), ‘vibrancy’

(3) or ‘resonance’ (40), it is widely accepted in the

singing literature that vibrato outside acceptable rate

and extent parameters is indicative of poor technique

(3, 15) and inferior sound quality. Vibrato rate and

extent are important ways to communicate emotion in

singing. For example, deep, slow vibrato is perceived

as sad, while fast, irregular vibrato communicates fear

(41). Vibrato may vary for other reasons; for example,

for dramatic purposes (42); in response to stylistic

requirements of particular genres (43, 44); or as a

result of changes to SPL in messa di voce (45); or for

dynamic variation (25) towards the ends of notes.

However, consistent vibrato within specified para-

meters has been linked to beautiful sound (38) and

to listeners’ overall preference (37). In the classical

singing literature, a steady and even vibrato is

universally promoted (3).

In this study, we demonstrated reliable differences in

vibrato parameters as a result of varying the degree to

which singers applied the technique of open throat.

Reduction of open throat technique for these singers

produced a significant decrease in VE in SO/LSO and

increase in VO in both SO/LSO. Singers spanned the

conventional VR parameters, from 5 Hz to �/6 Hz,

and although visual inspection showed that VR was

less even in SO/LSO, this was not statistically sig-

nificant. While singers vary vibrato for emotional

reasons, or suitability in different musical genres, these

six singers’ vibrato changed as a result of using

optimal or insufficient open throat technique. Since

vibrato parameters largely define good singing tech-

nique in the literature, open throat would appear to be

an essential element of sound vocal pedagogy.

Different genres require differing vibrato extent. For

example, singers use a wider vibrato extent in operatic

singing compared to concert singing (43) and in

classical singing compared to baroque or early music

style (44). Informal studies have found that the same

singers increased vibrato extent in dramatic lied or

Verdi operatic arias in comparison to Schubert’s ‘Ave

Maria’ (26). Changes to vibrato extent in an individual

singer are normally associated with emotional intent

(for example, agitated and peaceful in (42)). In this

study, singers reduced VE by, on average, more than

40% from O to SO/LSO. Decreased vibrato extent

associated with reduced open throat suggests that

instruction and technique are also relevant to the

production of classical voice quality, which was

perceptually equated with poor singing technique.

These findings suggest that open throat is indeed a

specific technique used to achieve a classical or

operatic quality. Our understanding of what produces

changes in vibrato extent may need to be extended to

include not only the demands of genre and emotional

intention, and to a lesser extent, the production of

particular vowels, but also its deliberate cultivation

through instruction in particular pedagogical techni-

ques such as open throat.

A singer’s vibrato rate has been identified as an

intrinsic component of his or her sound quality (28)

and is the most consistent parameter (46, 47). The

vocal system has a natural tendency to oscillate in the

5�/6 Hz frequency range. Results from this study

support this finding. Singers’ rate did not vary outside

of their O range, regardless of condition. Changes to

VR have been identified in different musical genres

(43), but even in these, VR was the most consistent

factor when the same singers sang opera or concert

songs. Our findings confirm previous data on vibrato

rate. Visual inspection of the spectrographs shows that

reduction of open throat in the SO and LSO condi-

tions was associated with greater irregularity of the

vibrato pattern compared to O, although these

differences did not reach statistical significance when

mean rates were compared across conditions. This is

probably due to the fact that mean VR could not take

into account occasions where no cycle was measured

because singers did not produce vibrato. Such cycles

occurred only in SO and LSO. Additionally, there were



large inter-subject differences of up to 1 Hz across

different tasks and conditions.

In pedagogical terms, if vibrato onset was delayedor absent in sustained pitches or during rapid syllables,

then the voice was not ‘natural’ and this would

indicate a faulty technique. This view concurs with

Ekholm et al . (37) who associated late onset of vibrato

(�/0.5 seconds) with the overall perception of poor

vibrato and this also was highly correlated with lower

ratings of the overall voice quality. In terms of

measurement, Prame (25) did not directly addressVO, although he rejected the first vibrato cycle in the

mean VR or VE, because the pattern had not

stabilized. In our study, both SO and LSO conditions

revealed a marked increase in VO (Table 2 and Fig.

6b�/c), which is associated with poorer overall vibrato

and inferior application of technique.

Pedagogical implications

The challenge for vocal pedagogues is to assess each

individual voice and to devise a program of technical

work to improve the basic sound. In Western classical

singing, good technique is integral to vocal quality,

and poor technique is identified by faults perceived

within the overall sound (48, 49). Research into

singing, by contrast, has hitherto focused on identify-ing acoustic parameters that characterize a ‘good’

voice by reference to findings that have been generated

by averaging data from a number of voices. The

dilemma and paradox in this approach is that the

averaged findings may not adequately represent any of

the individual voices used to generate the desirable

acoustic parameters that become the benchmark of

vocal quality. Pedagogy focuses on individual techni-ques to achieve the complete sound and evaluating

teaching techniques individually is necessary to verify

their effectiveness and reliability as teaching tools.

This study extends previous work in its attempts to

assess and verify the efficacy of specific teaching tools

by reference to their acoustic and perceptual impact

on vocal quality. While replication of these findings is

required prior to confident recommendations thatopen throat technique be applied in singing studios,

this study has demonstrated that individual teaching

techniques can be empirically assessed and their effects

verified both perceptually and acoustically.

CONCLUSION

Pedagogical and perceptual studies agree that sub-

standard vibrato is immediately identifiable and

indicates inadequate technique or poor vocal quality.Whether reduced vibrato extent and delayed vibrato

onset are consequences of reduced ‘open throat’

technique specifically, or the primary factors perceived

as ‘open throat’ or indeed any technique used in good

singing needs further investigation. Vibrato alone doesnot necessarily carry the acoustic cues of ‘vocal

quality’ to listeners (50), yet perceptually, in this study,

the best quality was associated with open throat

technique. Further acoustic analysis of O and SO

sound qualities is necessary to identify which para-

meters most closely concur with the holistic percep-

tions of expert pedagogues.

Auditory-perceptual aspects of a voice represent apsychological reality for both the singer and the

listener. Linking an acoustic parameter such as vibrato

to a commonly used instruction, such as open throat,

has the potential to provide a common terminology

for ease of communication between singers, their

teachers, and clinicians. Future research may also

define ‘open throat’ as a set of physiological para-

meters.

ACKNOWLEDGEMENTS

We thank Ms Maree Ryan for her pedagogical advice

and support, Dr Densil Cabrera and Peter Thomas for

their advice on acoustical matters and the six singers

for their willing participation in the project.

REFERENCES

1. Reid CL. Voice: Psyche and Soma. New York: JosephPatelson Music House; 1975.

2. Reid CL. A dictionary of vocal terminology. New York:Joseph Patelson Music House; 1983.

3. Miller R. The structure of singing: system and art invocal technique. New York; London: Schirmer Books;1996.

4. Mitchell HF, Kenny DT, Ryan M, Davis PJ. Definingopen throat through content analysis of experts’ peda-gogical practices. Logoped Phoniatr Vocol 2003; 28:167�/80.

5. Fields VA. Training the Singing Voice: an analysis of theworking concepts contained in recent contributions tovocal pedagogy. New York: King’s Crown Press; 1947.

6. Burgin JC. Teaching singing. Metuchen, N.J.: ScarecrowPress; 1973.

7. Monahan BJ. The art of singing: a compendium ofthoughts on singing published between 1777 and 1927.Metuchen, N.J.: Scarecrow Press; 1927.

8. Bartholomew WT. The Role of Imagery in VoiceTeaching. In: Music Teachers National AssociationProceedings for 1935; Oberlin, Ohio; 1935. p. 78.

9. Hemsley T. Singing and imagination: a human approachto a great musical tradition. Oxford; New York: OxfordUniversity Press; 1998.

10. Manen L. Bel canto: the teaching of the classical Italiansong-schools its decline and restoration. Oxford: OxfordUniversity Press; 1987.



11. Miller R. The Open Throat (La gola aperta). In: MillerR, editor. On the Art of Singing. New York: OxfordUniversity Press; 1997. p. 17�/9.

12. Puritz E. The teaching of Elisabeth Schumann. London:Methuen; 1956.

13. Shakespeare W. The Art of Singing. Ditson: New York;1921.

14. Garcia M. Hints on Singing. Translated from the Frenchby Beata Garcia. London: E. Ascherberg; 1894.

15. Vennard W. Singing: the mechanism and the technic. 5thed. New York: Fischer; 1968.

16. Joiner JR. Charles Amable Battaille: pioneer in vocalscience and the teaching of singing. Lanham, Md.:Scarecrow Press; 1998.

17. Ware C. Basics of vocal pedagogy: the foundations andprocess of singing. New York: McGraw-Hill; 1998.

18. Marafioti PM. Caruso’s Method of Voice Production.New York: Dover Publications; 1981.

19. McKinney JC. The Diagnosis and Correction of VocalFaults. Nashville, Tennessee: Broadman Press; 1982.

20. Herbert-Caesari EF. The voice of the mind. London:Robert Hale; 1951.

21. Baer T, Gore JC, Gracco LC, Nye PW. Analysis of vocaltract shape and dimensions using magnetic resonanceimaging: vowels. J Acoust Soc Am 1991; 90(2 Pt 1):799�/828.

22. Story BH, Titze IR, Hoffman EA. The relationship ofvocal tract shape to three voice qualities. J Acoust SocAm 2001; 109: 1651�/67.

23. Estill J. Primer of Basic Figures. 2nd ed. Santa Rosa:Estill Voice Training Systems; 1996.

24. Chapman J, Davis PJ. Primal Singing Australian Voice1998; 4: 9�/11.

25. Prame E. Measurement of the vibrato rate of ten singers.J Acoust Soc Am 1994; 96: 1979�/84.

26. Prame E. Vibrato extent and intonation in professionalWestern lyric singing. J Acoust Soc Am 1997; 102: 616�/

21.27. Shipp T, Sundberg J, Hadlund S. A Model of Frequency

Vibrato. In: Van Lawrence L, editor. Transcripts of theThirteenth Symposium: Care of the Professional Voice;1984; New York: The Voice Foundation; 1984. p. 116�/

17.28. Sundberg J. Acoustic and Psychoacoustic Aspects of

Vocal Vibrato. In: Dejonckere PH, Hirano M, SundbergJ, editors. Vibrato. San Diego: Singular PublishingGroup, Inc.; 1995. p. 35�/62.

29. Rothman HB, Diaz JA, Vincent KE. Comparinghistorical and contemporary opera singers with histor-ical and contemporary Jewish cantors. J Voice 2000; 14:205�/14.

30. Keidar A, Titze IR, Timberlake C. Vibrato Character-istics of Tenors Singing High Cs. In: Van Lawrence L,editor. Proceedings of the Thirteenth Symposium: Careof the Professional Voice; 1984; New York City: TheVoice Foundation; 1984.

31. Seidner W, Nawka T, Cebulla M. Dependence of theVibrato on Pitch, Musical Intensity, and Vowel inDifferent Voice Classes. In: Dejonckere PH, HiranoM, Sundberg J, editors. Vibrato. San Diego: SingularPublishing Group, Inc.; 1995. p. 63�/82.


33. Foulds-Elliott S, Thorpe C, Cala S, Davis P. Respiratoryfunction in operatic singing: effects of emotional con-nection. Logoped Phoniatr Vocol 2000; 25: 151�/68.


35. Cabrera D, Davis P, Barnes J, Jacobs M, Bell D.Recording the Operatic Voice for Acoustic Analysis.Acoustics Australia 2002; 30: 103�/8.

36. Rossing TD, Sundberg J, Ternstrom S. Acoustic com-parison of soprano solo and choir singing. J Acoust SocAm 1987; 82: 830�/6.


38. Robison CW, Bounous B, Bailey R. Vocal Beauty: Astudy proposing its acoustical definition and relevantcauses in classical baritones and female belt singers.Journal of Singing 1994; 51: 19�/30.

39. Vurma A, Ross Y. Priorities in voice training: carryingpower or tone quality. Musicae Scientiae 2000; 4: 75�/93.

40. Callaghan J. Singing and voice science. San Diego,California: Singular Publishing. Group; 2000.

41. Gabrielsson A, Juslin PN. Emotional Expression inMusic Performance: Between the Performer’s Intentionand the Listener’s Experience. Psychology of Music1996; 24: 68�/91.

42. Sundberg J, Iwarsson J, Hagegard H. A Singer’sExpression of Emotions in Sung Performance. In:Fujimura O, Hirano M, editors. Vocal Fold Physiology:Voice Quality Control. San Diego, California: SingularPublishing Group, Inc.; 1995. p. 217�/31.

43. Easley E. A Comparison of the Vibrato in Concert andOpera Singing. In: Seashore CE, editor. The Vibrato.Iowa: University of Iowa Press; 1932. p. 269�/75.

44. Hakes J, Shipp T, Doherty ET. Acoustic Properties ofStraight Tone, Vibrato, Trill and Trillo. J Voice 1987; 1:148�/56.

45. Titze IR, Long R, Shirley GI, Stathopoulos E, RamigLO, Carroll LM, et al. Messa di voce: an investigation ofthe symmetry of crescendo and decrescendo in a singingexercise. J Acoust Soc Am 1999; 105: 2933�/40.

46. Titze IR, Solomon NP, Luschei ES, Hirano M. Inter-ference between normal vibrato and artificial stimula-tion of laryngeal muscles at near-vibrato rates. J Voice1994; 8: 215�/23.

47. Titze IR, Story B, Smith M, Long R. A reflex resonancemodel of vocal vibrato. J Acoust Soc Am 2002; 111(5 Pt1): 2272�/82.

48. Davidson JW, Da Costa Coimbra D. Investigatingperformance evaluation by assessors of singers in amusic college setting. Musicae Scientiae 2001; 5: 33�/53.

49. Stanley M, Brooker R, Gilbert R. Examiner Perceptionsof Using Criteria in Music Performance Assessment.Research Studies in Music Education 2002; 18: 43�/52.

50. Rothman HB, Rullman FJ, Arroyo AA. Inter- andIntrasubject Changes in Vibrato: Perceptual and Acous-tic Aspects. J Voice 1990; 4: 309�/16.



3.1 Conclusion

Vibrato is an accepted tone quality of the classical singing voice (Miller, 1996;

Vennard, 1968). It is widely accepted in the singing literature that poor vibrato is

indicative of poor technique (Miller, 1996; Vennard, 1968) and inferior sound quality or

timbre. Previous studies have not considered the application of vocal techniques to

refine vocal features such as vibrato in singing. These results indicate that for these

singers, open throat was an integral technical component of their production of a good

sound.

From the results of the vibrato study, it was hypothesized that the energy distribution in

these singers’ voices in each condition would also reflect the timbral changes associated

with open throat. Long-term average spectra have been used to represent vocal genres

and also vocal qualities, in particular, the carrying power identified in operatic and

classical voices. This methodology was used to further assess the differences in quality

achieved by open throat.

66

4 Long term average spectra

In singing literature, particular LTAS models or exemplars have become synonymous

with vocal qualities and voices of quality. Singing analyses present exemplars of voice

types (Sundberg, 1974) or singing genres (Borch & Sundberg, 2002; Cleveland et al.,

2001). LTAS curves have been used to make obvious differentiations between males

and females (Mendoza, Valencia, Munoz, & Trujillo, 1996), singers and speakers

(Barrichelo, Heuer, Dean, & Sataloff, 2001), solo voice and choral voice (Rossing et al.,

1987) and pop or country from opera singers (Borch & Sundberg, 2002; Cleveland et

al., 2001). Research designs to compare multiple spectra indicate that it is important that

only one factor be varied in any given comparison (Barnes et al., 2004; Rossing et al.,

1987; Thorpe et al., 2001). Visual inspection of LTAS in each case confirmed

differences between pairs of samples.

Early studies using LTAS present visual data for analyses (Jansson & Sundberg, 1975;

Rossing et al., 1987) and later studies seek to quantify the data contained in the LTAS

to a single meaningful number (Barnes et al., 2004; Novak & Vokral, 1995; Omori et

al., 1996; Thorpe et al., 2001). These have been used to predict quality in voice

professionals (Novak & Vokral, 1995), to compare voices and also as a measure of

goodness or a voice of quality by level of professional singing (Barnes et al., 2004;

Thorpe et al., 2001). While exemplars illustrate the most common or expected features

by singer or singing genre, they carry little statistical power in the identification or

report of vocal quality.

67

More recent research linked features of these exemplars with performers of the highest

regard in singing performing circles (Barnes et al., 2004; Thorpe et al., 2001). LTAS

evaluation methods have followed these trends identified through visual inspection of

singers’ spectra and assess vocal quality through measures of central tendency or

variability between singers (Novak & Vokral, 1995; Sundberg, 2001).

Current practice in singing research accepts LTAS as representative of timbre. The

analysis has been used to track gross differences between singers and speakers

(Barrichelo et al., 2001) and inherent similarities between stylised country singing and

speech (Cleveland et al., 2001). It follows that LTAS should identify timbral differences

in the same group of subjects performing under different conditions.

LTAS are used to assess the acoustic output and timbre in extended singing. In this next

paper, we assessed the impact of open throat on visual displays of LTAS as well as on

conventional analysis methods of singing power ratio (SPR) (Omori et al., 1996) and

energy ratio (ER) (Thorpe et al., 2001). We measured the spectral distribution of

acoustic energy in the same group of singers in their singing of a classical song and a

romantic lied. Our aim was to confirm timbral differences identified from changes to

vibrato from O to SO.

68

PAPER 3

Mitchell, H. F., & Kenny, D. T. (2004). The effects of open throat technique on long term average spectra (LTAS) of female classical

voices. Logopedics Phoniatrics Vocology, 29(4), 171-182.

69

The effects of open throat technique on long termaverage spectra (LTAS) of female classical voices


From the Australian Centre for Applied Research in Music Performance (ACARMP), Sydney Conservatorium of Music,The University of Sydney, New South Wales 2006, Australia

Received 8 October 2003. Accepted 13 May 2004.


In the third of a series of studies on open throat technique, we compared long term average spectra (LTAS) of six advancedsinging students under three conditions: ‘optimal’ (O), representing maximal open throat, ‘sub-optimal’ (SO), using reducedopen throat, and loud sub-optimal (LSO) to control for the effect of loudness. Using a series of univariate repeated measuresANOVAs with planned orthogonal contrasts, we tested the hypotheses that sound pressure level (SPL) and the ratio of spectralenergy in peaks and areas between 0�/2 kHz and 2�/4 kHz would be reduced in SO and LSO compared to O. There weresignificant differences between SO and LSO but hypotheses were not confirmed for O. These findings do not accord withdifferences in vibrato extent and onset between O and SO/LSO (Mitchell and Kenny, in press). These results suggest that whileLTAS provides information on energy distribution, measuring spectral energy areas appears to be the most sensitive measureof energy distribution between conditions. Plotting the differences between O and SO/LSO pairs of LTAS clearly indicates theareas of spectral change. The findings from this study also indicate that LTAS are not sufficiently sensitive to measure vocaltimbre as they were not consistent with perceptual or other acoustic studies of the same samples.

Key words: long-term average spectra, open throat, singing technique, vocal pedagogy.

Helen Mitchell, Australian Centre for Applied Research in Music Performance (ACARMP), The Conservatorium of Music, TheUniversity of Sydney, New South Wales, Australia 2006. Tel.: 61 2 9351 9644. Fax: 61 2 9351 9540. E-mail: [email protected]

INTRODUCTION

In the singing literature and current pedagogy, open

throat technique is considered important to the

achievement of a good sound, in particular, as a

necessary component of resonance (1�/3). In a recent

study, pedagogues agreed that the technique of open

throat produces an audible quality in the sound,

equating its use with an even and consistent, balanced

and coordinated, free, warm and open quality in the

overall sound (4). They described the technique as a

way of maximizing pharyngeal space and/or achieving

abduction of the ventricular folds that could be

achieved by laugh, sob, correct inhalation or main-

taining the posture of inhalation (4).

Despite the routine use of ‘open throat’ technique in

the singing studio, there has to date been only one

scientific study that has addressed the vocal quality

believed to be produced by this technique (5). Hy-

potheses generated from pedagogical beliefs about

open throat were tested acoustically by Mitchell and

Kenny (5). In this study, we compared vibrato rate,

extent and onset of six advanced singing students

under three conditions: ‘optimal’ (O), representing

maximal open throat, ‘sub-optimal’ (SO), using re-

duced open throat, and loud sub-optimal (LSO), using

reduced open throat but controlling for the effect of

loudness. Fifteen expert judges correctly identified the

sound produced when singers used open throat with

87% accuracy. Having verified the technique percep-

tually, we used a series of univariate repeated measures

ANOVAs with planned orthogonal contrasts to test

the hypotheses that frequency modulations associated

with vibrato rate, extent and onset would vary outside

acceptable or desirable parameters for SO and LSO.

Hypotheses were confirmed for vibrato extent and

onset but not for rate. There were no significant

differences between SO and LSO on any of the vibrato

parameters. As vibrato is considered a key indicator of

good singing, these findings suggested that open

throat is important to the production of a good sound

in classical singing.

In this study, we examined the spectral distribution

of acoustic energy in the same six singers’ voices in



DOI: 10.1080/14015430410015722 70

each condition to identify the timbral changes in SO

and LSO. Positive descriptions of spectral energy such

as ‘resonance/ring’ (6), the singer’s formant (7) have

been linked to production of a beautiful sound and

with an efficient and aesthetically pleasing distribution

of energy or harmonics and loudness in classical

singing (4).

Long-term average spectra (LTAS) have been used

to analyse musical timbre (8) and also vocal features,

in both speech (8, 9) and singing (10, 11). An LTAS

gives an overall impression of an entire excerpt (12),

identifies certain consistent features contained in the

sound over time, and averages out short-term varia-

tions in phonetic structure (13). After 20�/30 seconds,

the LTAS represents the individual’s sound spectra

(12, 14). The LTAS, when taken over a sufficiently

long production, suppresses the short-term variations

that are due to notes and lyrics, while retaining such

long-term invariance in the spectrum as can be

expected to reflect a particular phonatory and/or

articulatory strategy. The LTAS plot is a measure of

the dB level of the time-average of the power of the

acoustic signal at each frequency.Rossing et al. (15) used LTAS to track differences in

the same singers and found that soprano singers’

LTAS plots were different depending on whether the

singing was solo or choral, and soft or loud. Jansson

and Sundberg (12) noted that the nature of the musical

task, the key, the instrument on which it was

performed, and timbral differences on the same

instrument all influenced the LTAS plot. These data

indicate that in research designs comparing LTAS, it is

important that only one factor be varied in any given

comparison. This might involve the same singer sing-

ing and speaking the same task (11), singing the same

task in different styles (15), different singers singing

the same task (16) or instruments playing in different

musical keys or producing the sound in different ways

or with different musical expression (12).

Long-term average spectra, particularly in singing,

have been accepted as analogues for vocal quality.

Vocal projection has also been conceptualized in this

way. Vocal projection and the ability to be audible

above an orchestra are vital to the classical singing

voice (17, 18). Classical and operatic singers demon-

strate a characteristic spectral energy above 2 kHz, to

enable amplification over an orchestra, which does not

show amplification in this frequency band (12, 16, 19).

This carrying power was identified by Bartholomew

(20) as a spectral reinforcement around 2800 Hz,

produced by male and low female singers. Sundberg

classified this as the ‘singer’s formant’ (21). The

existence of this energy band has been widely debated

as an element in the soprano voice (20, 22�/24). Some

studies have identified greater energy in sopranos of

international standard than other classically trained

sopranos (15, 23) and concluded that energy above 2

kHz may be as important for sopranos as for other

voice types. However, in some cases, an increased

energy above 2 kHz was perceived as detrimental to

vocal quality when assessed perceptually (18). It is

therefore important to consider carrying power as

energy distribution and not simply as an indicator of

quality.

The typical method of assessing vocal quality

through LTAS has been through an investigation of

differences in measures of central tendency and

variability between singers (10, 14). There is a need

to explore intra-singer differences under different

performance requirements or conditions because in-

dividual spectra may be unique. Averaged data may

produce curves that do not adequately represent the

unique qualities of individual singing voices and may

indeed produce a curve that does not represent any

voice (11, 14, 16, 25). The individuality of spectra is

evident when the same singer singing portions of the

same aria produces different spectral plots and both

plots are considered in making a mean trend of

spectral changes (25). Studies using the same singers

but different musical tasks (e.g. country singers singing

an anthem and a song (11)) show marked differences

in spectra between the two conditions. This demon-

strates the importance of using the same musical

sample or stimuli when examining LTAS derived

under different experimental conditions.

A conventional way of reducing the information in

the LTAS to a single meaningful number is to compute

the ratio of energies in a low and a high frequency

band. Often 2 kHz is chosen as the delimiting

frequency (25, 26). In LTAS, particularly in clinical

practice, the slope between these peaks is calculated

(13). In singing, measures of spectra compare energy

peak height (singing power ratio) and peak area

(energy ratio) between 0�/2 kHz and 2�/4 kHz. The

difference between the height of the major peaks

between 0�/2 kHz and 2�/4 kHz (26) quantifies the

relative energy between 2�/4 kHz although it does not

account for the shape of these energy peaks. Thorpe et

al. assessed the area under the LTAS curve at 0�/2 kHz

and 2�/4 kHz, rather than the highest peaks of energy,

to identify the areas in which the energy is reinforced

or reduced when singers sang with varying levels of

emotional connection (25). It also enables effective

comparison between individual singers singing in

different conditions and between groups of singers.

Comparing pairs of LTAS, where the data have been

normalized to remove the sound pressure level (SPL)

variables (11), produced little difference between

LTAS of country singers’ speech and singing.



In this study, we assessed whether open throat

technique is a necessary component of good classical

vocal quality and measured the spectral distribution ofacoustic energy with the same group of singers who

consciously manipulated the technique in their singing

of a classical song and a romantic lied. Our aim was to

use LTAS to discern measurable differences in spectral

energy between an optimal sound quality, using

maximal open throat, and a sound which used a lesser

degree of open throat, produced without consciously

altering other components of their singing techniqueor sound quality. This approach represents a major

extension on current work that describes vocal fea-

tures at only one point in time.

METHOD

Participants

Six female singers (three sopranos, three mezzo-

sopranos) volunteered to participate in this project.They were advanced students with excellent technique

of an experienced singing pedagogue, who was a

Lecturer in Vocal Studies and Opera at a state

Conservatorium of Music. Criteria for participant

selection included, through this pedagogue’s assess-

ment, singers who: 1) had a good classical singing

technique for their level of training and experience;

and 2) understood and demonstrated skilful control of‘open throat’ or ‘retraction’ techniques in their sing-

ing. The institutional human ethics committee ap-

proved the study.

Prior to the voice recording, participants completed

a questionnaire seeking information on age, years of

singing study, number of years of study with each

singing teacher, highest qualifications attained or

currently undertaken in music and/or singing, andsinger type (soprano or mezzo soprano). The partici-

pants were also asked to classify the genres of singing

they performed in public (opera, classical, choral,

music theatre and contemporary) and to estimate the

percentage each style played in their total performing

career.

The demographic information of the participants is

presented in Table 1. Participants were aged between23 and 30 years, with a mean of 26 years. All had

studied singing for at least 7 years (average 9.8 years)

and had spent an average of 5 years studying with

their present singing teacher. Each singer held a

qualification in singing or music (four had Bachelor

of Music degrees and two had diplomas, in music and/

or singing) and five of six were currently undertaking a

second degree in singing (three postgraduate Diplomaof Opera and two Bachelor of Music degrees).

All defined the majority of their singing as operatic

(�/50%), with the second most common style classical

(�/20%), in accordance with the Bunch and Chapman

(27) taxonomy of singing voices. All reported that they

were in good health and able to perform the tasks.

Procedure

Singers were sent information about the project and

were invited to take part in an acoustic and perceptual

study of singing technique. They were required to

attend a single recording session lasting no more than

an hour and were told that the object of the study was

to investigate acoustical and perceptual features of the

use of open throat in singing and to discover the sound

qualities associated when a singer uses some form of

open throat technique compared with when a singer

does not use open throat.

Instruments

Protocol. A protocol was developed to assess the

effect of ‘open throat’ technique in singing. Two

musical tasks were developed to investigate the use

of the technique in two song excerpts. Musical tasks

were chosen in order to test different demands of good

singing, but were not musically difficult. They were

specifically designed to test the use of open throat, and

contained musical features derived from a previous

qualitative study on the use of the technique (4), where

use or lack of the technique was deemed to be

particularly valuable or noticeable. These features

were: high tessitura, sustained or legato singing,

dynamic range control, and vocal agility. Each musical

task took an average of 74 seconds to perform.

The Mozart song Ridente la Calma , K 152, bars

1�/27 (Figure 1a) was selected as it is a nominally

simple song in the Italian language (28) with a mixture

of common musical statements involving repeated

legato lines as well as the initial stylized leaps of a

major 4, and short scale figures. All six singers sang

this aria in the same key (F major).

The third verse of the Schubert lied, Du bist die Ruh

D. 776 (Op. 59, No. 3), bars 54�/80 (Figure 1b) was

chosen for its demanding vocal control, sustained

musical line and high climactic tessitura. The three

sopranos and three mezzo-sopranos sang this in an

appropriate key depending on soprano or mezzo-

soprano voices (E-flat, D-flat and C major).

Prior to singing, each singer selected the sequence of

their tasks before commencing the experiment by

selecting a blank card, the reverse side of which

represented one of the tasks, to reduce the possibility

that task order may influence the results.

Effects of open throat technique on LTAS 101


Recording

Participants were given time to warm up in the singing

studio and become familiar with the room before

recording. Recording levels for each singer were set

during this time. The acoustic signal was recorded

digitally (Behringer Ultragain preamplifier/Marantz

CDR 630) to CD via a high-quality microphone

(AKG C-477) positioned on a head boom a constant

7-cm distance from the singer’s lips. This ensured that

direct energy of the performers’ voices was recorded

rather than room reflections, enabling us to use a

studio environment with low ambient noise rather

than an anechoic studio (29).Calibration was carried out in each recording by

using a recording of two pink noise samples on

audiotape played immediately following each record-

ing session at the same recording gain used for

recording the singer’s voice. Pink noise is a broadband

steady state signal of known sound pressure level. Pink

noise enables calibration of an audio playback system

across the frequency range for tasks such as subjective

testing and if the analysis is just done on computer, the

Tab

le1.

Su

mm

ary

of

pa

rtic

ipa

nt

cha

ract

eris

tics

.D

escr

ipti

on

of

sin

ger

nu

mb

er,

age,

the

nu

mb

ero

fyea

rsea

chsi

nger

ha

sta

ken

sin

gin

gle

sso

ns,

the

nu

mb

ero

f

tea

cher

sea

chsi

nger

ha

sh

ad

,th

en

um

ber

of

yea

rsea

chsi

nger

stu

die

dw

ith

thei

rth

ree

mo

stre

cen

tte

ach

ers

(T

1� /T

3)

,th

eh

igh

est

leve

lo

fed

uca

tio

nco

mp

lete

dby

each

sin

ger

,th

ecu

rren

tleve

lo

fed

uca

tio

nu

nd

erta

ken

,vo

ice

typ

e,p

rop

ort

ion

of

each

sin

ger

’sca

reer

spen

tsi

ng

ing

op

era

,cl

ass

ica

l,ch

ora

l,m

usi

cth

eatr

e,

con

tem

po

rary

or

oth

erst

yle

so

fm

usi

c

Sin

ger

Ag

e(y

ears

)Y

ears

of

less

on

sN

o.

of

tea

cher

sT

1T

2T

3H

igh

est

edu

cati

on

Cu

rren

ted

uca

tio

nV

oic

ety

pe

Op

era

Cla

ssic

al

Ch

ora

lM

usi

cth

eatr

eC

on

tem

pO

ther

12

68

41

13

Ba

chel

or

Ba

chel

or

Mez

zo8

0%

20

%2

25

73

1.5

34

.5B

ach

elo

rP

ost

gra

du

ate

So

pra

no

80

%2

0%

33

01

22

10

2D

iplo

ma

N/A

So

pra

no

97

%3

%4

23

10

24

5.5

Ba

chel

or

Po

stg

rad

uat

eS

op

ran

o5

0%

45

%5

%5

25

10

27

.52

.5D

iplo

ma

Po

stg

rad

uat

eM

ezzo

73

%2

5%

2%

62

81

26

51

.51

.5B

ach

elo

rB

ach

elo

rM

ezzo

50

%3

5%

5%

10

%

Fig. 1a. The Mozart task musical score.

Fig. 1b. The Schubert task musical score in the

soprano key.



noise is at least as good as any other steady-state

known signal. In addition, a tone has a disadvantage

over noise by being more susceptible to interference in

the environment, and hence would give less repeatable

results. For calibration of absolute sound pressure

levels (SPL), the microphone of a sound level meter

(Rion NL-06 SPL) was held adjacent to the AKG

microphone from a speaker (Bose Lifestyle) from

which the pink noise was played. The SPL (in dB at

7 cm) recorded by the sound level meter was noted for

the two noise signals and used later for calibration.

Experimental conditions

Singers sang each of the two song excerpts under three

conditions: optimal, sub-optimal and loud sub-opti-

mal. ‘Optimal’ (O) condition was necessary to provide

a sound quality with the best technique that they

could. This involved the maximal use of open throat.

‘Sub-optimal’ (SO) condition involved the use of a

reduced (open throat) technique but still with an

acceptable singing technique and without consciously

altering any other aspect of their technique. It was

hypothesized that the SO condition would result in a

reduction of SPL, so a third condition ‘loud sub-

optimal’ (LSO) involved the same instruction as the

SO condition, but with the added instruction that the

singer should try to achieve a louder dynamic than in

SO. This addressed the possibility of SPL as the only

variable responsible for the differences between SO

and O. This was similar to technical conditions

established by Foulds-Elliot et al. (30) when SPL

was a variable of degrees of emotional connection in

operatic singers.

Each task was performed twice in the O and SO

conditions while the LSO condition was performed

only once because of the slight possibility of adverse

effects on vocal health. In total, each singer performed

each musical task five times.

Instructions

A pedagogue was present during the recording ses-

sions to provide accompaniment for warm-ups and

practice of the tasks where necessary and also selected

the terminology and instructions necessary for each

singer to achieve the experimental conditions. For

example, she instructed them to pay attention to

producing the most open sound in their throat in the

O condition, and a lesser degree in the SO and LSO

conditions. Some singers asked how to produce the

LSO condition and the pedagogue instructed them to

‘use more twang’, as taught in their lessons (31).

Analysis

The audio recordings were digitized at 16 kHz using

Phog Version 2.0 (Hitech Development, Sweden)

software and reopened in Soundswell Version 4.0

(Hitech Development, Sweden) to produce nine chan-

nels of raw and calculated data (including SPL and

F0). A real-time digital SPL meter was used to set a

recording gain so that the pink noise calibration level

recorded in Phog was the same (�/0.5 dB) as the SPLlevel noted during recording.

Long-term average spectra

Long-term average spectra (LTAS) analyses (band-

width 300 Hz) (32, 33) of the 36 song samples

demonstrated the differences in spectral energy for

the singers, conditions and musical tasks. LTAS

contained only voiced data (13, 32). A 300 Hz LTAS

bandwidth is less sensitive than a narrow-band analy-sis to movement in the partials when the singer is at

higher fundamental frequencies. A complication when

using LTAS is to account for loudness, which is

problematic as frequencies above 2 kHz increase faster

than those below 2 kHz as sound pressure level (SPL)

increases (34).

In order to relate these LTAS plots to the known dB

of the calibration tone, an LTAS was performed onsteady five-second portions of the pink noise sample

for each singer, using the same 300 Hz filter. The Sect

tool in Soundswell does not compute the overall

equivalent level (Leq) for an LTAS. Therefore, to

calculate the absolute SPL of each calibration tone, it

was necessary to find the mean SPL of the pink noise

LTAS. Each point of the LTAS curve was linearized

(y�/10 (x/10)), then all data points were summed, andthis total converted back to dB (x�/10 �/log(y)). The

mean SPL of pink noise (in decibels down from full

scale) was subtracted from the known SPL of the

calibration noise measured during recording to pro-

duce the calibration offset. This calibration offset (in

dB) was applied to each LTAS. In earlier research

using LTAS methodology, White (32) arrived at a

‘calibration offset’ by calculating the mean SPL of asung /i/ vowel by adding the peak level (dB) of the

LTAS with the known level. Using LTAS area to

calculate the mean SPL of pink noise was used in

preference to calculating the LTAS peak height and

was considered a valid method for evaluating the

overall level (dB) of the pink noise at 7 cm.

LTAS analysis

Acoustic analysis focused on two frequency areas: 0�/2kHz and 2�/4 kHz. From the LTAS plots, the highest

peaks in the 0�/2 kHz and 2�/4 kHz regions were



labelled P1 and P2 respectively. Peak levels (in dB) and

peak centre frequencies (in Hz) were calculated for

each singer in each task and condition.Three measures were used to quantify these LTAS

data. Linear regression was used to quantify the

relationship between the shapes of each singer’s

LTAS by condition (O, SO, LSO), in each musical

task. In the regression coefficients, a slope of 1 and an

intercept of 0 would imply a perfect correlation, thus

deviations demonstrate differences observed between

condition results. This methodology draws compar-isons between the changes in the same task with

different overall qualities (11, 16).

The singing power ratio (SPR), described by Omori

et al. (26) compares the peak levels of P1 and P2. The

SPR is the difference between the level of P1 and P2 in

dB (LP1�/LP2) and is a measure of spectral slope

between 0�/2 kHz and 2�/4 kHz. When the energy is

focused in the 0�/2 kHz region, it results in higher SPRresults and there is typically less energy reinforcement

in the 2�/4 kHz region. A low SPR indicates a stronger

energy peak �/2 kHz. Although SPR does not prove

the presence of a singer’s formant, it enables inter- and

intra-singer comparison of spectral energy in the voice

(35, 36).

The energy ratio (ER), described by Thorpe et al.

(25) compares overall energy between the area between0�/2 kHz (A1) and the area between 2�/4 kHz (A2). It is

calculated by taking the difference between average

energy values for the two frequency areas (A1�/A2). A

low ER represents a greater reinforcement in the 2�/4

kHz region whereas a high ER represents a smaller

energy boost �/2 kHz. Areas below the LTAS curve

were calculated in the same way as the calibration

tone. When there is less reinforcement in the 2�/4 kHzregion, ER results follow SPR. As ER assesses the

area of the LTAS curve, the energy differences between

singers’ O/SO, O/LSO and SO/LSO LTAS data were

plotted graphically to compare the energy distribution

at each frequency point.

Other factors that may influence the LTAS plot,

such as fundamental frequency (F0) distribution, and

the pitch range of each musical task was considered(12). Histograms of the F0 were computed in Sounds-

well for each musical task.

Study design

The study design was a repeated measures (three

dependent measures: SPL; P1; P2) randomized com-

plete block with 2 (composer (Mozart versus

Schubert))�/3 (condition (optimal, sub-optimal and

loud sub-optimal)) factorial structure with plannedorthogonal contrasts on the three dependent variables.

This design was chosen because the study hypotheses

were mainly concerned with the effects of experimental

condition on SPL, P1 and P2 and not between-singer

variability. The chosen design absorbs singer varia-

bility within the error term to reveal the main effects

and interactions of interest.

Data were analysed firstly through a series of fixed

and mixed effect univariate ANOVAs with three

factors (task, condition, singer). In the first series, all

three factors were entered as fixed effects; in the

second, singer was entered as a random effect. Finally,

the data were subjected to analysis using the general

linear model (GLM) with planned orthogonal con-

trasts. Since all analyses yielded the same outcomes,

only the results of the GLM are presented here. Main

effects for task and condition and interaction effects

(contrasts) were calculated for each dependent mea-

sure (SPL, P1, P2). In the first contrast, O was

compared to SO and LSO and in the second contrast

SO was compared to LSO.

Hypotheses

Hypotheses were generated regarding the spectral

distribution of acoustic energy and sound pressure

levels produced by singers using open throat compared

to a reduced open throat condition. The loud sub-

optimal (LSO) condition was used as a control for

loudness between O and SO. Therefore we compared

O with SO/LSO. Our specific hypotheses were as

follows:

1. Overall sound pressure level (SPL) will decrease

from O to SO and from O to LSO. Expert

pedagogues stated that the use of open throat

would increase their perception of ‘loudness’ (4).2. Following from hypothesis 1, long-term average

spectral distribution, that is, peak levels P1 and P2

(in dB), will decrease in SO/LSO. Voice acoustics

predict that the magnitude of the level difference

between P1 and P2 will become smaller with

increasing SPL, except at extremes of effort and

vowel but we wanted to test whether this relation-

ship also pertains during the use of a perceptuallydistinct singing technique (4, 5).

3. There will be a difference in energy distribution

shown in the shape of the LTAS curve between O

and SO/LSO, based on expert pedagogical de-

scriptions and perceptual judgements (4, 5).

4. SPR measurements will increase in SO/LSO

compared to O. Lower SPR is associated with

higher vocal quality and greater length of training.The pedagogical literature (1�/3, 37) and our

previous study (4) have indicated that length of

training and vocal quality are also associated with



open throat. We therefore test the hypothesis that

open throat will be associated with lower SPR.

5. ER measurements will increase in SO/LSO com-pared to O. In singing, ER results tend to follow

SPR. That is, lower SPR and ER are associated

with higher vocal quality (36).

6. Changes observed in each condition will be

consistent with reductions in vibrato extent and

increases in vibrato onset in SO/LSO compared to

O (5). Given the statistical confirmation that SO/

LSO were different to O (5), we investigate howthese findings affect the LTAS energy distribution.

RESULTS

Descriptives

Figure 2a and b present the LTAS of each singer in

each condition for the Mozart and Schubert tasks

respectively. The complete SPL and spectral peak data

are presented in Table 2, showing the overall means of

SPL and levels of P1 (0�/2 kHz) and P2 (2�/4 kHz)

with corresponding centre frequency (Hz) by eachsinger, in each condition and in each musical task

(Mozart and Schubert).

Differences in SPL, P1 and P2 were considered

separately for effects of condition and task.

Hypothesis 1: overall sound pressure level (SPL) will

decrease from O to SO and from O to LSO

Hypothesis 1 stated that SPL will be lower in the SO

condition than in O and from O to LSO. Hypothesis 1

was confirmed between SO and LSO. For SPL, there

was a main effect for contrast 2 only (F(1,5)�/18.948,

p�/0.007), that is, SPL was significantly higher in theLSO condition than SO. Overall means indicated no

significant difference between SPL of O and SO/LSO

(F(1,5)�/0.010, p�/0.924). There was no effect of task

(F(1,5)�/0.001, p�/0.974), and SPL was affected by

condition in the same way in each task. Figure 3

presents these data graphically.

The mean SPL for the O condition over the six

singers was 97.4 dB (SD 1.9) in the Mozart task and97.7 dB (SD 2.6) in the Schubert task. In the SO

condition, the mean SPL was significantly lower,

particularly in the Mozart task. In both the SO and

LSO condition, each singer’s SPL was significantly

different to O (Table 2). The LSO condition was

associated with SPL levels that were either similar to

the O condition (six examplesB/2 dB different) and in

nine examples, increased SPL from the level of theoriginal O in the LSO condition (LSO means: 99.2 dB,

98.9 dB).

Hypothesis 2: peak levels P1 and P2 (in dB) will

decrease in SO/LSO

Hypothesis 2 was confirmed for P1. For P1, there was

a main effect for contrast 1 only (F(1,5)�/17.359,

p�/0.009), that is, the level of P1 was significantly

greater in the O condition than SO/LSO. There was no

significant difference between SO and LSO (F(1,5)�/

4.911, p�/0.078). There was no effect of task (F(1,5)�/

1.392, p�/0.291). There were no interactions between

musical task and condition. These data are presented

in Fig. 4. In the SO condition P1 was associated with a

considerable decrease (�/3 dB) in 9 of the 12 (6

singers�/2 musical tasks) SO cases compared to the

O condition. In the LSO condition P1 was reduced

�/3 dB in half of 12 LSO cases, but was greater in one

take compared to the O condition.

Hypothesis 2 was not confirmed for P2. For P2,

there was a main effect for contrast 2 only (F(1,5)�/

15.330, p�/0.011) that is, levels of P2 were significantly

higher in LSO than in SO but SO/LSO were not

significantly different to O (F(1,5)�/1.112, p�/0.340).

Fig. 5 presents these data. There was no effect of

musical task (F(1,5)�/1.643, p�/0.256) and there were

no interactions between task and condition for P2.

Changes in the P2 by condition were more difficult to

quantify. In the SO condition, P2 was associated with

a considerable decrease (�/3 dB) in 7 of the 12 SO

cases compared to the O condition. Some were

markedly reduced, such as 8�/10 or 20 dB lower (Table

2). One SO sample was associated with a greater SPL

(�/3 dB) compared to the O condition.

Hypothesis 3: there will be a difference in energy

distribution shown in the shape of the LTAS curve

between O and SO/LSO

Hypothesis 3 stated that there will be a difference in

energy distribution shown in the shape of the LTAS

curve between O and SO/LSO, based on expert

pedagogical descriptions and perceptual judgements.

This hypothesis was tested using linear regression

where all singers’ LTAS data were normalized, with

P1 at 0 dB and the LTAS distribution compared

between O and SO/LSO and between SO and LSO in

each musical task, for each singer (after Cleveland

et al . (11)). Hypothesis 3 was not confirmed. Correla-

tions between individual singers’ LTAS across the

three conditions are presented in Table 3. All correla-

tions were greater than 0.90. Singer 5 was an outlier

between O/LSO. Slopes were all close to 1 and

approaching unity, and intercepts vary from close to

zero in the Schubert task, to �/3.92�/�/3.05 in the

Mozart.



Fig. 2a. The Mozart task: LTAS for the six singers in the Mozart musical task, showing mean sound pressure

levels (dB at 7 cm) on the ordinate at frequencies from 0�/6000Hz on the abscissa. The bold, thin and dashed lines

show optimal, sub-optimal and loud sub-optimal conditions respectively. Each LTAS is marked by subject

number 1�/6.



Hypothesis 4: SPR measurements will increase in SO/

LSO compared to O

Hypothesis 4 stated that SPR will increase in SO/LSO

compared to O, that is, the difference between peak

levels between 0�/2 kHz and 2�/4 kHz would indicate

differences in carrying power (energy in 2�/4 kHz

region). Hypothesis 4 was not confirmed for O and

SO/LSO. In testing the differences (in dB) of SPR, the

effect of task approached significance (F(1,5)�/8.928,

p�/0.031), that is, the difference between P1 and P2 in

the Schubert task was slightly greater than in Mozart,

but both musical tasks showed the same overall

pattern across conditions. There was an effect for

contrast 2 only, (F(1,5)�/45.152, p�/0.001), that is, the

Fig. 2b. The Schubert task.



Table 2. Measurements of sound pressure level (SPL) at 7 cm, peak levels in dB (P1, P2) from the 0�/2 and 2�/4 kHz regions of a long term average spectra and

centre frequencies in Hz for P1 and P2 for the Mozart and Schubert musical tasks in each condition, optimal (O), sub-optimal (SO) and loud sub-optimal

(LSO)

SPL P1 P2

Condition Condition Condition

O SO LSO O SO LSO O SO LSO

Task Singer dB (Hz) Database (Hz) dB (Hz) dB (Hz) dB (Hz) dB (Hz)

Mozart1 99.3 97.0 98.3 97.4 (750) 90.9 (750) 91.6 (750) 77.2 (3125) 76.0 (3000) 79.0 (3000)2 95.3 92.8 97.1 98.2 (750) 90.6 (750) 97.6 (750) 73.7 (3125) 65.9 (3000) 77.3 (3000)3 98.6 97.6 98.6 98.5 (750) 92.2 (750) 95.3 (750) 84.4 (3000) 74.8 (3125) 80.2 (3125)4 97.9 98.0 98.4 96.8 (750) 93.8 (750) 92.5 (750) 74.8 (3125) 71.8 (3125) 73.5 (3125)5 98.3 95.6 102.8 99.4 (750) 91.5 (750) 94.8 (750) 79.2 (3000) 73.3 (3000) 80.4 (3000)6 94.8 96.0 99.9 94.9 (750) 94.2 (750) 98.6 (750) 72.7 (3500) 78.3 (3125) 83.7 (3250)Mean (SD) 97.4 (1.9) 96.2 (1.9) 99.2 (2.0) 97.5 (1.6) 92.2 (1.5) 95.1 (2.8) 77.0 (4.3) 73.4 (4.3) 79.0 (3.4)

Schubert1 98.0 96.0 98.7 93.6 (625) 89.3 (625) 90.9 (625) 75.9 (3125) 69.9 (2875) 76.6 (2875)2 98.4 90.9 96.3 103.1 (625) 91.0 (750) 97.0 (750) 78.3 (3125) 58.3 (3125) 72.9 (3250)3 100.6 99.1 100.9 100.8 (750) 96.9 (750) 98.7 (750) 79.7 (3000) 73.7 (3000) 80.6 (3125)4 99.7 99.4 101.4 99.6 (750) 99.3 (750) 99.6 (750) 74.9 (3250) 72.6 (3250) 75.1 (3375)5 93.7 94.7 96.9 97.2 (500) 92.3 (625) 91.6 (625) 76.0 (3375) 76.4 (3000) 77.9 (3375)6 95.5 97.0 98.8 99.1 (625) 96.8 (625) 96.1 (625) 78.9 (3500) 77.7 (3375) 82.8 (3375)Mean (SD) 97.6 (2.6) 96.2 (3.1) 98.8 (2.1) 98.9 (3.3) 94.3 (3.9) 95.7 (3.7) 77.3 (1.9) 71.4 (7.0) 77.6 (3.6)

10

8H

.F

.M

itchell

an

dD

.T

.K

enny

Lo

gop

edP

ho

nia

trV

oco

l2

9

79

difference between P1 and P2 was significantly lower

in LSO than in SO. There was no significant difference

between O and SO/LSO (F(1,5)�/5.101, p�/0.133).

There was no interaction effect between task and

condition. These data are presented in Fig. 6.

The SPR of each task and condition was calculated

for each singer to compare the relationship of energypeaks between 0�/2 kHz and 2�/4 kHz for inter-subject

differences in task and intra-subject effects of condi-

tion. Individual SPR results are presented in Table 4.

Lower SPR results, indicating more energy between

2�/4 kHz, from O to SO were seen in three Mozart SO

and two Schubert SO cases. The SPR was lower for all

singers in LSO compared with SO.

Hypothesis 5: ER measurements will increase in SO/

LSO compared to O

Hypothesis 5 stated that ER will increase in SO/LSO

compared to O. The ER was calculated for each

musical task in each condition for each subject.

Individual subjects’ ER by task and condition are

presented in Table 5. The changes to ER results by

singer corresponded to SPR results in the Mozart and

Schubert tasks from SO/LSO compared to O, that is,ER decreased with condition across singers. In the

Schubert task, subject 5 was an outlier.

Hypothesis 5 was not confirmed for O and SO/LSO.

There was a weak trend showing the effect of task on

ER (F(1,5)�/5.278, p�/0.070), that is, ER in the

Schubert task was slightly greater than in Mozart,

but both musical tasks showed the same overall

pattern across conditions. There was an effect forcontrast 2 only, (F(1,5)�/36.913, p�/0.002), that is, ER

was significantly less in LSO than in SO. There was no

Fig. 3. Main effect for SPL (in dB at 7 cm) between

sub-optimal and loud sub-optimal. Means for two

musical tasks (Mozart and Schubert) in each of the

three conditions, optimal, sub-optimal and loud sub-optimal.

Fig. 4. Main effect for P1 between optimal and sub-

optimal/loud sub-optimal. Means for the two musicaltasks (Mozart and Schubert) in each of the three

conditions, optimal, sub-optimal and loud sub-opti-

mal.

Fig. 5. Main effect for P2 between sub-optimal and

loud sub-optimal. Mean differences between the two

musical tasks (Mozart and Schubert) in each of the

three conditions, optimal, sub-optimal and loud sub-

optimal.



significant difference between O and SO/LSO (F(1,5)�/

3.531, p�/0.119). There was no interaction effect

between task and condition. These data are presented

in Fig. 7.

The energy distribution differences in the LTAS of

each singer were compared by subtracting SO from O

(illustrated in Fig. 8), LSO from O and LSO from SO.

The energy differences for all six singers between O/

SO, O/LSO and SO/LSO in the Mozart and Schubert

tasks are presented in Fig. 9a�/c. Data points above

zero represent an increase in energy at the particular

frequency whereas data points below zero indicate a

decrease in energy between conditions.

At peak level P1 (between 500 Hz and 1 kHz), Fig.

9a shows that all singers reduced acoustic energy

between O and SO. Similarly, in Fig. 9b, singers

showed a reduction in the energy of P1 in LSO

compared to O. Conversely, there was an increase for

all singers between 0�/2 kHz between LSO and SO.

Changes around peak level P2 were more complex.

The differences between O and SO, and O and LSO

were characterized by a series of peaks and troughs at

points between 2�/4 kHz rather than a consistent

reduction or increase in energy at all data points.

Between SO and O, these differences are associated

with a reduction of SPL (Fig. 9a), and between LSO

and O, the changes represent a combination of

increases and decreases in SPL. Fig. 9c shows the

subtracted differences between LSO and SO where the

majority of these data points show an increase in

acoustic energy. This indicates that for these singers

the shape of the SO and LSO LTAS plots was similar,

but with greater overall energy in LSO.

Fig. 9a and b show a prominent increase of energy

at 2 kHz. In Fig. 2a and b, the SO and LSO conditions

showed a narrower P2, predominantly in the Mozart

musical task. The variation in energy around 2 kHz in

Fig. 9a and b shows that, rather than a consistent

increase of energy, the shape of the peaks had

changed. The shape of the P2 peak differs for each

singer depending on condition, illustrated in the thin

and dashed lines in Fig. 2a and b.

Table 3. Correlations (r), intercepts and slope of linear regressions of LTAS values calculated in the optimal and

sub-optimal, and optimal and loud sub-optimal conditions of the Mozart and Schubert Tasks

O-SO O-LSO SO-LSO

Task Singer r Intercept Slope r Intercept Slope r Intercept Slope

Mozart1 0.97 �/0.48 0.94 0.95 0.36 0.80 0.99 0.79 0.852 0.93 0.68 1.10 0.95 �/1.72 0.83 0.94 �/3.92 0.693 0.96 �/0.78 0.87 0.98 �/1.26 0.83 0.97 �/0.46 0.914 0.96 �/2.48 0.91 0.93 �/3.07 0.78 0.98 �/0.65 0.875 0.91 1.61 0.88 0.86 3.05 0.80 0.98 2.29 0.956 0.98 �/0.44 0.91 0.95 0.88 0.80 0.96 1.08 0.79

Schubert1 0.97 �/0.62 0.84 0.95 �/0.44 0.77 0.99 0.28 0.942 0.96 �/0.94 0.94 0.97 �/1.38 0.80 0.97 �/2.90 0.793 0.96 �/1.42 1.00 0.99 �/0.43 1.00 0.98 �/0.05 0.954 0.92 �/1.51 0.83 0.92 �/0.87 0.71 0.99 0.32 0.855 0.96 �/0.87 0.85 0.96 �/0.05 0.72 0.98 0.31 0.836 0.95 0.81 0.81 0.93 �/1.95 0.72 0.96 �/0.46 0.88

Fig. 6. Main effect for singing power ratio (SPR)

differences between P1 and P2 between sub-optimal

and loud sub-optimal. Means of difference between P1and P2 in the two musical tasks (Mozart and

Schubert) in each of the three conditions, optimal,

sub-optimal and loud sub-optimal.



Hypothesis 6: Changes observed in each condition will

be consistent with reductions in vibrato extent and

increases in vibrato onset in SO/LSO compared to O

Hypothesis 6 stated that changes in vibrato observed

in each condition will be reflected in the correspond-

ing LTAS plots for each singer. Specifically, the

reduction in vibrato extent and onset would affect

each singer’s LTAS curve in SO/LSO compared to O.

In each musical task, fundamental frequency distribu-

tion (F0) was affected by an absence or decrease of

vibrato. Both a reduction of extent and an increase inthe delay of onset meant that for the time period of

each musical task, fewer pitches were sung. Mean F0

in each condition and in each task varied little across

conditions (B/20 Hz) in 23 of 24 SO and LSO takes

compared to O. The changes to the distribution of F0

across tasks are presented in Fig. 10 for each task in

each condition. These panels, of O, SO and LSO F0

distribution respectively, show a decrease in the spreadin F0 covered during each task. In Fig. 10a, there is

one mean distribution, for the same key used by all

singers, but in Fig. 10b, there are three separate plots

representing the respective keys used by sopranos and

mezzos.

Fig. 11 presents normalized LTAS of singer 3’s

Mozart task with inset panels of F0 distribution. It

presents an exemplar of the effects seen because ofvibrato extent changes to the F0 and the correspond-

ing shape changes to the LTAS across conditions.

DISCUSSION

Long-term average spectra of singers in this study

showed that there were significant differences between

SO and LSO but not between O and SO/LSO in

spectral distribution of energy peaks and area between

0�/2 kHz and 2�/4 kHz and in overall SPL. For the

individual peaks P1 and P2, there was a significant

difference in the level of P1 in O compared with SO/

Table 4. Measurements of singing power ratios long

term average spectra analysed for the Mozart and

Schubert musical tasks in each condition, optimal

(O), sub-optimal (SO) and loud sub-optimal (LSO)

SPR

Condition

Task Subject O SO LSO

Mozart1 20.2 14.9 12.62 24.4 24.7 20.33 14.0 17.4 15.04 22.0 22.0 19.05 20.2 18.2 14.56 22.2 15.8 14.9

Schubert1 17.8 19.5 14.32 24.8 32.7 24.13 21.1 23.2 18.14 24.7 26.6 24.55 21.1 15.9 13.76 20.1 19.0 13.4

Table 5. Measurements of energy ratios over the long

term average spectra analysed for the Mozart and

Schubert musical tasks in each condition, optimal

(O), sub-optimal (SO) and loud sub-optimal (LSO)

ER

Condition

Task Subject O SO LSO

Mozart1 17.0 14.9 13.12 19.6 19.2 15.13 12.1 16.8 14.24 19.1 17.3 14.15 17.4 15.2 12.06 17.4 13.1 11.1

Schubert1 16.1 17.9 13.62 22.2 29.4 20.33 18.5 20.2 15.84 20.1 20.9 18.15 16.2 14.1 12.66 17.2 16.1 10.2

Fig. 7. Main effect for energy ratios (ER) between the

sub-optimal and loud sub-optimal conditions. Meansof the difference between ER in the two musical tasks

(Mozart and Schubert) in each of the three conditions,

optimal, sub-optimal and loud sub-optimal.



LSO, but for P2, there was a difference between SO

and LSO. Based on the regression plots, there were

highly statistically significant correlations between the

shapes of singers’ LTAS across all conditions, indicat-

ing the lack of sensitivity of such measures in detecting

timbral differences.

SPL has an effect on LTAS slope, such that an

increase in SPL typically increases high frequency

energy at a faster rate than lower frequency energies

(34, 36). Therefore, it was important to generate more

than one experimental condition to test the impact of

open throat technique in singing. In addition, we

chose to use average SPL as a separate measure from

power levels in these analyses to detect differences

between conditions (36). The inclusion of a second

comparison group, LSO, indicated that SPL did not

account for differences between O and SO. Foulds-

Elliot (30) and Rossing (15) have used similar study

designs. The hypothesized reduction in SO SPL did

not in fact occur; however, there were differences in the

energy distributions of each condition. While SPL

would be expected to be higher than both P1 and P2,

the SPL was the average of the sound pressure (or

power) level, rather than the levels of the average

powers that were accumulated into the LTAS in the

respective frequency regions for P1 and P2. Because

the distribution of power in typical sounds is heavily

skewed toward low powers, with brief excursions to

very high powers, the average of the power level will

tend to be lower than the level of the average power.

The average SPL more accurately reflects the typical

sound level than the level of the average power.There were statistically significant correlations be-

tween conditions in the shape of singers’ normalized

LTAS. The variations in positive and negative inter-

cepts are indicative of spectral differences greater than

4 kHz. When Cleveland et al. (11) compared country

singers’ speech and singing they found a slope close to

1 and an intercept close to zero indicating similarity

between speech and singing plots in country singers.

While strong similarities might be expected in country

singing and speaking, a far greater difference would be

expected between optimal and sub-optimal attempts

in classical singing, but this was not the case. LTAS is

used in clinical contexts as an indicator of the natural

voice sound (8), and therefore should have been ideal

for assessing the changes in the timbre associated with

each condition.

As the LTAS curves were so similar, detailed

analyses of spectral distribution on the calibrated

data were essential. The increased P1 in O confirms

that ‘open throat’ is often characterized by a more

dominant fundamental partial. Subjectively, this

would correspond to a rounder and less pressed

sound. Calibrating these LTAS enabled precise infor-

mation on the relative or absolute level of the peaks

(15, 32), which would have been lost in normalized

LTAS curves (11).

Measures of SPR and ER reduce LTAS information

to a single number, which can be used to compare

intra- and inter-singer energy. The difference in energy

between 0�/2 kHz and 2�/4 kHz is indicative of the

carrying power of the voice. A smaller difference

between P1 and P2, or SPR, identified LSO as the

most effective for vocal projection. In LTAS, a smaller

difference between peak levels P1 and P2 implies

improved carrying power and a better classical quality

(10, 18). This is the case for male singers in production

of the singer’s formant (21). For SO, there was a

reduction of P2 compared to LSO, and therefore

reduction in carrying power. In the case of LSO,

Vurma and Ross (18) have previously reported that a

small difference between peaks was not necessarily

associated perceptually with an improved vocal qual-

ity.

Measuring the difference between peaks provides a

representation of energy rather than of quality. As a

form of measurement, Lundy et al. (35) also found

SPR (26) problematic in assessing energy differences

between students’ sung and spoken vowels. They

found no differences between the SPR of each. In

addition to the level of the peak, the centre frequency

of P2 has been used to define an energy boost in the

Fig. 8. Example of singer 1’s LTAS plots in the

optimal and sub-optimal condition for the Mozart

task. Below is an example of the calculation of energy

difference between the optimal and sub-optimal con-ditions.



singing voice. Centre frequencies in O, SO and LSO

ranged from 3000�/3500 Hz in all conditions which

was in keeping with Dmitriev and Kiselevs’ (38)

findings for female singers: 2700 and 3500Hz

(mezzo-soprano to high soprano). Although Weiss et

al. (22) have questioned the legitimacy of the singer’s

Fig. 9a�/c . (a) Subtracted energy between the optimal and sub-optimal conditions, for each singer in the Mozarttask, shown in the left column, and Schubert, in the right column. (b) Subtracted energy between the optimal and

loud sub-optimal conditions. (c) Subtracted energy between the loud sub-optimal and sub-optimal conditions.



Fig. 10a�/b. (a) The Mozart task: mean fundamental frequency distribution for all six singers showing the

histogram distribution on the ordinate from 0�/0.2 occurrences at frequencies from 200�/900Hz on the abscissa in

optimal, sub-optimal and loud sub-optimal conditions. (b) The Schubert Task, in the three musical keys used by

the six singers.



formant in sopranos, they also found stronger spectral

peaks around 3 kHz. These female singers’ spectral

reinforcement was significantly less than that of male

singers.

Using conventional measures to analyse the differ-

ences in LTAS by condition did not correspond to the

previous study, where SO/LSO were found not to

indicate acceptable vibrato. However, visual inspection

of Fig. 2a and b shows that, while the difference

between peaks was much closer in the LSO condition,

the shape of P2 and energy above 2 kHz was in

complete contrast to O.

Measuring ER obtained specific information point

by point in LTAS to describe the differences in LTAS

by condition. Overall ER results followed SPR.

Exploring these ER results, through the subtracted

data (after Thorpe et al . (25)) effectively demonstrated

the changes to singers’ LTAS plots at every point.

Thorpe et al. found increases at all data points

between spectra in normal singing and using vocal

projection, but with considerable energy increases

above 2 kHz. Fig. 9a�/c clearly indicates that the

overall energy distribution fluctuates between condi-

tions; specifically, the shape of a singer’s LTAS

changed by condition. Subtracted ER shows that

while levels of P2 (in dB) are regularly matched in

LSO compared to O where the centre frequency is

close to or above zero, the level of points to either side

decrease below zero. This clearly indicates the change

in LTAS shape between O/LSO. Between O/SO, the

results are similar, but P2 in SO is lower than O, so the

dB level fluctuations occur below zero, illustrating a

decrease in overall energy distribution.

Subtracted data between SO/LSO produced results

more like Thorpe et al. (25) as all subjects demon-

strated increases in energy or spectral density over 2

kHz. This implies that there was not a difference in

LTAS shape between SO/LSO, but rather an increase

in energy level at every frequency point. As LSO was

devised as a control for loudness, specifically to

remove SPL as sole variable for the SO condition

(30), it is interesting to note that despite the differences

in overall SPL, LSO is a louder replica of SO.

Explanations for unusual peaks or troughs between

O/SO and O/LSO can be traced in Fig. 2. For example,

greater energy around 2 kHz does not represent a new

peak, but rather, highlights a change in energy

distribution above and below this frequency. The

variability of increase or decrease is far greater in the

O/SO and O/LSO comparisons, and if peaks cancel

out troughs, and so on, this may explain an equalizing

in overall ER results or no statistical difference

between O and SO/LSO. Computing ER as a single

meaningful number across the entire 0�/2 kHz and 2�/4

kHz regions may not be helpful in identifying changes

in timbre or energy. However, it appears to be a more

sensitive measure compared to regression plots and

peak height as a way of visualizing changes across

LTAS.

Fig. 11. Normalized examples of singer 3’s LTAS

plots in optimal, sub-optimal and loud sub-optimal

conditions for the Mozart task. Insets represent the

histogram of fundamental frequency for the same

rendition of the musical task.



Calculating a mean of all singers’ energy increases

may not represent the energy increase of any singer

(25). The individuality of change in the subtracted

data for these singers suggests unique changes to their

vocal quality, and to their energy spectra (35). A mean

of all subjects’ performance of both musical tasks

would have obscured rather than illuminated the

unique qualities of each voice. The mean line through

ER differences illustrates the trends of change by

condition, but does not represent the changes to any

particular voice.

Changes to F0 distribution (illustrated in Fig. 10)

were accounted for by a reduction in vibrato (5), as the

musical tasks were identical in O, SO and LSO. These

singers’ O vibrato had a wider pitch range compared

to SO and LSO. This in turn may be reflected in the

spectral distribution of the LTAS plot showing peaks

in the 2�/4 kHz region of SO and LSO, where they are

distinctly narrower or include multiple peaks. Inter-

estingly, Sundberg (39) linked vibrato to the distribu-

tion of partials in the sound spectrum. Increased

vibrato extent has also been linked specifically to

spectral changes in sopranos’ solo singing compared

to choral singing, where it increased the level (dB) of

P2, and changed its overall distribution and the shape

of the peak (15). Although it was hypothesized that

sopranos produced this energy above 2 kHz by

increasing vibrato extent rather than clustering higher

formants as in male singer’s formant, results in this

study show increases in P2 by around 3 dB and not the

15 dB identified by Rossing et al. (15). Further

investigation of the effect of vibrato extent on spectra

of individual notes is needed. The changes of vibrato

parameters were not linked to such a significant

reduction in P2, suggesting that singers still increased

LSO energy above 2 kHz in other ways.

The broader P2 in O is possibly but not necessarily

due to vibrato. If open throat involves a larger mouth

opening, for example, this would reduce the Q factor

of the formant resonances, which would broaden but

also lower the peaks. P2 could still be higher in O than

in SO if the total output is higher in O. In any case, the

fact that partials are moving around vigorously due to

F0 modulation in vibrato is in itself not sufficient to

explain a broader peak in the spectrum envelope.

Rather, concomitant motions of the pharynx walls can

be expected to perturb the formant frequencies, which

would smear the peaks. Further study of these possible

explanations for the shapes of the peaks is needed.The instruction to achieve the unusual condition of

LSO by using more ‘twang’ may explain the difference

in singers’ P2 shape from O to LSO. Titze (40)

compared twang to ring quality and although he

found them similar spectrally, defined twang differ-

ently as a narrowing of the pharynx. This pedagogical

instruction has been associated with production of

resonance in the classical voice (21, 25, 40). However,

for these singers, LSO produced a similar peak levelbut different spectral distributions over the whole peak

compared to O.

Pedagogical implications

Singing pedagogues rely on a series of singing

techniques to produce the best possible sound from

each student. Pedagogues evaluate each voice on an

individual basis and internally assess what combina-

tion of instructions is necessary for the particular

voice to achieve its full potential. As researchers strive

to define a ‘good sound’ in singing through acoustic

measurement, it is important to assess a voice as apedagogue would, using the overall sound to devise

appropriate vocal exercises to achieve optimal quality

for that voice.

Teachers have a mental image of the particular

beautiful sound individual students are capable of

producing and use verbal explanation, vocal modelling

and feedback to teach this. The acoustic information

currently available has limited application in thisprocess. Mitchell and Kenny confirmed that the use

of open throat technique is useful in producing a good

classical sound and demonstrated its impact on the

regularity and evenness of singers’ vibrato rate, extent

and onset when the technique is reduced (5). Pedago-

gues did not explicitly attribute the singer’s formant or

increased carrying power to the use of open throat

technique but they did suggest its use facilitated abetter balance or more desirable distribution of

harmonics.

Poor vocal quality could not be defined in terms of

its spectral distribution in a way that we would have

expected. As our previous investigations confirmed

that SO and LSO did not reach the standards of good

classical singing, the purported link between vibrato

and overall quality illustrated through LTAS appearsto occur in the narrow or multiple peaks above 2 kHz.

Acoustic implications

LTAS is an ideal representation of a singer’s sound ina natural setting, that is, an entire performance rather

than single notes. However, despite�/60 seconds of

musical stimuli, the effect of task approached signifi-

cance for measures of SPR and ER. Jansson and

Sundberg (12) investigated LTAS by task and condi-

tion and in musical task noted different LTAS shapes

when the same instrument played the same scale in a

different musical key or played a triad figure comparedto a chromatic scale, which produced narrower and

more rounded peaks above 2 kHz. In light of these



findings, LTAS should be examined in the context of

specific tasks and conditions.

If LTAS does change depending on the group ofpitches analysed (12, 41), it is important in future to

address one variable in any experiment: the same

singer singing in two ways or two different singers

singing the same musical task. In this study, we

considered each singer’s spectra individually. The

mean line of subtracted ER does not represent any

one singer, and virtually cancels out the effect of the

individual singers. Previous templates of LTAS curveshave been produced from averages generated from

groups of singers (14). These averages may, in fact, not

represent any one voice and may be an artefact of

clustered measurement.

Assessment of LTAS is complex. Visual inspection

identifies obvious differences between the LTAS of

each condition; yet performing a variety of conven-

tional analyses could not identify statistically signifi-cant intra- or inter-singer differences to support the

vibrato findings, where SO/LSO represented poor

singing. Future research needs to evaluate in greater

detail factors that contribute to the shape of singers’

LTAS, both acoustically and perceptually.

CONCLUSION

Lessening the amount of open throat used in singing

produced few statistically significant differences to

these singers’ LTAS in each condition. Our perceptual

and acoustic work using open throat technique as the

dependent variable in assessment has shown that

conventional acoustic measures of voice quality donot correspond to perceptual assessments of vocal

beauty or mastery. A science of the singing voice

cannot progress without addressing the problem that

standard acoustic measures may not be sufficiently

sensitive to track perceived changes. We propose to

classify acoustic descriptors of these voices through

perceptual ratings of expert listeners to better evaluate

LTAS in singing and explore the role of LTAS as arepresentation of timbre in singing as well as energy.

This will complete the initial assessment of a single

technique and its impact on overall classical quality as

taught in the singing studio.

ACKNOWLEDGEMENTS

We thank the six singers for their willing participation

in the project and Ms Maree Ryan for her expert

pedagogical advice. The authors also thank ProfessorPamela Davis for early ideas on the research question

and study design. We are grateful for the helpful

comments of the reviewers in the revision of this

paper.

REFERENCES

1. Hemsley T. Singing and imagination: a human approachto a great musical tradition. Oxford, New York: OxfordUniversity Press; 1998.

2. Miller R. The structure of singing: system and art invocal technique. New York, London: Schirmer Books;1996.

3. Vennard W. Singing: the mechanism and the technic, 5thed. New York: Fischer; 1968.

4. Mitchell HF, Kenny DT, Ryan M, Davis PJ. Definingopen throat through content analysis of experts’ peda-gogical practices. Logoped Phoniatr Vocol 2003; 28:167�/80.

5. Mitchell HF, Kenny DT. The impact of ‘open throat’technique on vibrato rate, extent and onset in classicalsinging. Logoped Phoniatr Vocol. In press.


7. Robison CW, Bounous B, Bailey R. Vocal Beauty: Astudy proposing its acoustical definition and relevantcauses in classical baritones and female belt singers.Journal of Singing 1994; 51: 19�/30.

8. Lofqvist A. The long-time-average spectrum as a tool invoice research. Journal of Phonetics 1986; 14: 471�/5.

9. Mendoza E, Munoz J, Naranjo N. The Long-TermAverage Spectrum as a Measure of Voice Stability. FoliaPhoniatr Logop 1996; 48: 57�/64.

10. Novak A, Vokral J. Acoustic Parameters for theEvaluation of Voice of Future Voice Professionals. FoliaPhoniatr Logop 1995; 47: 279�/85.

11. Cleveland TF, Sundberg J, Stone RE. Long-Term-Average-Spectrum Characteristics of Country SingersDuring Speaking and Singing. J Voice 2001; 15: 54�/60.

12. Jansson EV, Sundberg J. Long-Time-Average-SpectraApplied to Analysis of Music. Part I: Method andGeneral Applications. Acustica 1975; 34: 15�/9.

13. Lofqvist A, Mandersson B. Long-Time Average Spec-trum of Speech and Voice Analysis. Folia PhoniatrLogop 1987; 39: 221�/9.

14. Sundberg J. Level and center frequency of the singer’sformant. J Voice 2001; 15: 176�/86.

15. Rossing TD, Sundberg J, Ternstrom S. Acoustic com-parison of soprano solo and choir singing. J Acoust SocAm 1987; 82: 830�/6.

16. Borch DZ, Sundberg J. Spectral distribution of solovoice and accompaniment in pop music. LogopedPhoniatr Vocol 2002; 27: 37�/41.

17. Murbe D, Sundberg J, Iwarsson J, Friedmann P, Hof-man G. Longitudinal study of solo singer educationeffects on maximum SPL and level in the singers’formant range. Logoped Phoniatr Vocol 1999; 24:178�/86.

18. Vurma A, Ross Y. Priorities in voice training: carryingpower or tone quality. Musicae Scientiae 2000; 4: 75�/93.

19. Sundberg J. The Science of the Singing Voice. DesKalb,Illinois: Northern Illinois University Press; 1988.



20. Bartholomew W. A physical definition of ‘good voicequality’ in the male voice. J Acoust Soc Am 1934; 6: 25�/

33.21. Sundberg J. Articulatory interpretation of the ‘singing

formant’. J Acoust Soc Am 1974; 55: 838�/44.22. Weiss R, Brown WS Jr, Morris J. Singer’s formant in

sopranos: fact or fiction? J Voice 2001; 15: 457�/68.23. Sundberg J. The singer’s formant revisited. Voice 1995;

4: 106�/19.24. Sundberg J. The Acoustics of the Singing Voice.

Scientific American 1977; March: 82�/91.25. Thorpe C, Cala S, Chapman J, Davis P. Patterns of

breath support in projection of the singing voice. J Voice2001; 15: 86�/104.

26. Omori K, Kacker A, Carroll LM, Riley WD, BlaugrundSM. Singing power ratio: quantitative evaluation ofsinging voice quality. J Voice 1996; 10: 228�/35.



29. Cabrera D, Davis P, Barnes J, Jacobs M, Bell D.Recording the Operatic Voice for Acoustic Analysis.Acoustics Australia 2002; 30: 103�/8.

30. Foulds-Elliott S, Thorpe C, Cala S, Davis P. Respiratoryfunction in operatic singing: effects of emotional con-nection. Logoped Phoniatr Vocol 2000; 25: 151�/68.

31. Estill J. Primer of Basic Figures, 2nd ed. Santa Rosa:Estill Voice Training Systems; 1996.

32. White P. A study of the effects of vocal intensityvariation on children’s voices using long-term averagespectrum (LTAS) analysis. Logoped Phoniatr Vocol1998; 23: 111�/20.

33. White P. Formant Frequency Analysis of Children’sSpoken and Sung Vowels Using Sweeping FundamentalFrequency Production. J Voice 1999; 13: 570�/82.

34. Nordenberg M, Sundberg J. Effect on LTAS of vocalloudness variation. TMH-QPSR, KTH 2003; 45: 93�/100.

35. Lundy DS, Roy S, Casiano RR, Xue JW, Evans J.Acoustic Analysis of the Singing and Speaking Voice inSinging Students. J Voice 2000; 14: 490�/3.

36. Barnes JJ, Davis PJ, Oates J, Chapman J. The relation-ship between professional operatic soprano voice andhigh range spectral energy, J Acoust Soc Am 2004. Inpress.

37. Miller R. The Open Throat (La gola aperta). In: MillerR, editor. On the Art of Singing. New York: OxfordUniversity Press; 1997. p. 17�/9.

38. Dmitriev L, Kiselev A. Relationship between theformant structure of different types of singing voicesand the dimensions of supraglottic cavities. FoliaPhoniatrica 1979; 31: 238�/41.

39. Sundberg J. Perceptual Aspects of Singing. J Voice 1994;8: 106�/22.

40. Titze I. Acoustic Interpretation of Resonant Voice. JVoice 2001; 15: 519�/28.

41. Howard DM, Szymanski J, Welch GF. Listeners’Perception of English Girl and Boy Choristers. MusicPerception 2002; 20: 35�/49.



4.1 Conclusion

Visual inspection of long-term average spectra (LTAS) confirmed differences between

O and SO in this study, but results showed that conventional acoustic measures on

LTAS did not produce statistically significant differences between the experimental

conditions. These results were not consistent with the vibrato findings and suggest that

measures of SPR [the difference between the peak height (Omori et al., 1996)] and ER

[the difference between peak areas (Thorpe et al., 2001)] are not sufficiently sensitive to

evaluate LTAS.

These findings represent a major challenge to current wisdom that acoustic parameters

of voice can identify changes in vocal timbre. Subsequent studies compare acoustic

output to perceptual assessments. The fourth study addressed the perceptual significance

of open throat technique and its impact on overall singing quality. The fifth study

considered the effectiveness of acoustic analysis in assessing vocal quality.

90

5 Perceptual verification

Evaluation of singing in tertiary institutions revealed that the assessment of singing was

based initially on the whole performance, before appraisal of specific technical and

performance components (Davidson & Da Costa Coimbra, 2001; Stanley, Brooker, &

Gilbert, 2002). Music performance studies, particularly in singing, find that judges do

not explicitly discuss an overall ‘good sound’, but rather, focus on aspects of technical

control (interaction between the voice and body, the role of diction on control, register

control, and adequate support for the sound) (Davidson & Da Costa Coimbra, 2001).

Lower marks were associated with a greater number of suggestions for technical

improvements.

Formal criteria are designed to complement rather than replace subjective responses to

performance assessment (Johnson, 1997). Research studies that use specific rating

scales (Boyle, 1992) have reported high correlations of all objective criteria with the

overall judgement (Ekholm et al., 1998; Geringer & Madsen, 1998; Saunders &

Holahan, 1997). As the individual dimensions are attributed to the final mark awarded,

it may not be possible to isolate the separate features that constitute overall quality.

Although some examiners believe criteria-specific evaluation interfered with the holistic

assessment of performance, they reported that use of criteria focussed their evaluation

on elements of performance, and was useful for student feedback. Perceptual studies in

music must address the global approach of listeners before focusing listeners to a single

91

technical component (Stanley et al., 2002). This enables judges to justify their response

and provide specific feedback for students.

Our aim was to determine the measurable differences between an optimal sound quality

a singer can produce involving the use of open throat, and a sound which uses a lesser

degree of the open throat technique, without altering other components of their singing

technique or sound quality. In Western classical singing, the requirements for good

vocal training and final assessment focus on the overall perception of a good sound.

Technical considerations focus on any indicators of poor quality (Davidson & Da Costa

Coimbra, 2001) rather than a series of specific technical attributes. However,

pedagogical literature focuses on a wide variety of techniques which contribute to this

final sound.

Fifteen experienced singing pedagogues and adjudicators assessed 30 samples generated

from the data collected for studies 2 and 3, and rated each sample as O or SO. Listeners

were presented with the definitions of each experimental condition, and of the technique

of open throat. The results of this perceptual study confirmed that listeners could

reliably identify the presence of open throat technique and were consistent in their

judgements of repeated samples.

92

PAPER 4

Mitchell, H. F., & Kenny, D. T. (in press). Can experts identify open throat technique as a perceptual phenomenon? Musicae Scientiae.

93

Perceptual identification of open throat 1Musicae Scientiae in press

Can experts identify “open throat” technique as a perceptual phenomenon?


From the Australian Centre for Applied Research in Music Performance (ACARMP), Sydney Conservatorium of Music, The University of Sydney, New South Wales 2006, Australia

Received 25 July 2004. Accepted 7 February 2005.

Musicae Scientiae; in press

Singing pedagogues have identified ‘open throat’ as a widely used pedagogical tool in the singing studio and a critical technique to achieve good classical vocal quality. This study is the first to assess the perceptual validity of open throat. Fifteen expert singing pedagogues assessed 48 messa di voce and 24 song samples with six repeats of six advanced singing students under two conditions: ‘optimal’ (O), representing use of maximal open throat technique and ‘subop-timal’ (SO), representing reduced open throat. Correctly identified responses were counted by condition (O/SO), by judge and by singer. Data were analyzed using Cohen’s Kappa. Hypotheses that correct identification would be greater than chance were confirmed for both messa di voce and the song samples, with thirteen of the fifteen judges correctly identifying 82.7% of song samples as O or SO. Singers’ self-ratings attributed their best singing to maximal use of open throat technique. These results indicate that listeners are consistent in making a dichotomous choice in identify-ing use of a singing technique through auditory-perceptual evaluation. The findings from this study suggest that there is a sound quality associated with the use of open throat technique, which is a perceptual reality to singing pedagogues and singers and that the specific vocal quality in classical singing that it produces can be reliably identified by expert listeners. If singers and expert listeners’ of singing accord, focussing future research on specific singing techniques could enhance singing pedagogy.

Helen Mitchell, Australian Centre for Applied Research in Music Performance (ACARMP), The Conservatorium of Music, The Uni-versity of Sydney, New South Wales, Australia 2006 Phone: 61 2 9351 9644; Fax: 61 2 9351 9540 Email: [email protected]

INTRODUCTION

In this paper, we report the first study to assess the perceptual validity of a specific singing technique – open throat (OT) – as taught in today’s singing studio. Previous perceptual studies (Ekholm, Pa-pagiannis, & Chagnon, 1998; Mendes, Rothman, Sapienza, & Brown, 2003; Robison, Bounous, & Bailey, 1994; Vurma & Ross, 2000; Wapnick & Ekholm, 1997) have demonstrated that singing improves over time and through training but have not identified specific components of the training that produce improvements in vocal quality and thus add little to the singing pedagogical literature. In order to evaluate the contribution of one sing-ing technique (OT), we have assessed the acoustic aspects of voice quality produced with and with-out this technique and the perceptual impressions of listeners, firstly their capacity to identify voices that used the technique and then to describe the vocal quality produced.

Open Throat Technique (OT)

The term “open throat” occurs frequently in the

vocal pedagogy literature (Burgin, 1973; Fields, 1947; Monahan, 1978). It is defined as a complex process that is both a pedagogical instruction and a perceived sensation or action that results in a specific sound quality. Current support for the use of OT in singing technique is widespread (Miller, 1996; Reid, 1975, 1983). When OT is used, peda-gogues perceive the sound as resonant (Miller, 1996; Vennard, 1968), round, (Joiner, 1998), free (Ware, 1998), pure, (Marafioti, 1981) rich and warm (McKinney, 1982). The sound quality is attributed to freedom from ‘constrictor tensions’ (Reid, 1983, p. 83). In the pedagogical literature, the sound quality attributed to open throat is linked to the preparation to sing, on inhalation (Hemsley, 1998; Manèn, 1987; Miller, 1997a, 1997b) through the surprise breath or smelling the rose imagery (Hemsley, 1998; Miller, 1996; Puritz, 1956) and visualizing space within the throat, through an ‘air-ball’ or ‘soap bubble’ (Herbert-Caesari, 1951; Manèn, 1987) to achieve the posture of OT.

The technique of OT is a pedagogical concept transmitted through the oral tradition of singing. Mitchell, Kenny, Ryan & Davis (2003) assessed

94


current pedagogical practices and perceptions of OT. They noted that expert singing pedagogues identified OT as a critical technique in vocal pro-duction of classical singing. Despite the use of dif-ferent terminology to describe the technique (such as freedom, collar and retraction), OT was defined by these pedagogues as a technique to maximize pharyngeal space and/or abduct the ventricular folds. There was consistency in the vocal instruc-tions to achieve it. Pedagogues teach conscious control of OT using laugh, sob, correct inhalation or maintaining the posture of inhalation. They con-sistently reported a perceptually distinctive sound quality as a result of mastery of this technique and described their perception of the final sound qual-ity as balanced, coordinated, free, even, consist-ent, warm and open.

Hypotheses generated from pedagogical beliefs and practices derived in the first study on OT were subsequently tested acoustically by Mitchell and Kenny (2004b) who compared the use of open throat technique (optimal, O) to a reduction of the technique (sub-optimal, SO) using female classi-cal singers. Reduced OT resulted in significant re-ductions in vibrato extent and increases in vibrato onset time but no change to vibrato rate between the two conditions. These singers relied on OT to produce acceptable vibrato parameters. However, according to the literature, inappropriate vibrato is indicative of poor singing in general; therefore further acoustic tests were required to test the dif-ferences in timbre. Visual inspection of long term average spectra (LTAS) confirmed differences be-tween O and SO for each singer, and the O con-dition produced a rounder peak between 0-2 kHz indicating a warmer sound quality compared to SO. Despite these findings, LTAS did not demon-strate significant differences between SO and O in conventional measures performed on long term average spectra energy peak height [singing pow-er ratio (Omori, Kacker, Carroll, Riley, & Blau-grund, 1996)] and energy peak area [energy ratio (Thorpe, Cala, Chapman, & Davis, 2001)].

Technology to measure voice acoustically has become more sophisticated and is used increas-ingly in experimental voice research, the findings of which are now being incorporated into texts on singing (Miller, 1996; Nair, 1999; Sundberg, 1977; Thurman & Welch, 2000). To date, research into the singing voice has described acoustic prop-erties of voice and its visual representation with few or no links to pedagogical descriptors or per-ceptual judgments. The implications of acoustic studies of the singing voice within singing lit-

erature and pedagogy are difficult to ascertain, since most studies do not assess acoustically or perceptually the vocal strategies used in the sing-ing studio (Foulds-Elliott, Thorpe, Cala, & Davis, 2000; Thorpe et al., 2001). Indeed, few studies link acoustic findings with perceptual judgments (Ekholm et al., 1998) or discuss their pedagogical implications (Callaghan, 2000; Miller, 1998).

This study seeks to provide stronger links be-tween empirical research and singing pedagogy, and recognises the importance of singing peda-gogues in this enterprise.

Perceptual Studies

Perceptual evaluation of musical performance is important for a number of reasons. Opera singers, their teachers and audiences are primarily inter-ested in how the voice sounds and use the sensory information derived from listening as the prima-ry basis for determining the quality of the sound (Kitch & Oates, 1994). Despite recent advances in the physiological and acoustic measurement of voice, “…voice quality is fundamentally per-ceptual in nature” (Kreiman, Gerratt, Kempster, Erman, & Berke, 1993, p. 21). The auditory-per-ceptual aspects of a voice represent a psychologi-cal reality for both the singer and the listener and have the potential to provide a common terminol-ogy for ease of communication between singers, their teachers, and clinicians. Development of a reliable and valid method of communicating one’s perception has significant implications. For ex-ample, audience appreciation ensures continuing engagement of performers; expert listeners adju-dicate examinations, auditions and other competi-tions, acts which may seal the fate of performing artists. Because such evaluations can have signifi-cant impacts on musical careers, it is important to establish whether listeners are reliable in their judgments, both in comparison to other listeners and across time.

Research indicates that listeners show some de-gree of reliability and consistency in their percep-tual judgments of timbre, between vocal genres (from opera to music theatre) (Sundberg, Gram-ming, & Lovetri, 1993), and between good and poor vocal and instrumental performance (Ekholm et al., 1998; Geringer & Madsen, 1998; Saunders & Holahan, 1997; Wapnick, Flowers, Alegant, & Jasinskas, 1993), in the assessment of excellence in overall voice quality (Wapnick & Ekholm, 1997) and in rankings and ratings of performers in competitive situations (Davidson & Da Costa Coimbra, 2001).

95


Recent evidence indicates that judges initially assess the performance as a whole before ‘justi-fying’ their ranking by assessing specific criteria (Stanley, Brooker, & Gilbert, 2002). Judges do acknowledge, however, that providing them with specific criteria focuses their assessment on certain elements of performance and that this provides the basis for useful, specific feedback to students.

Perceptual studies of vocal quality generally re-quire listeners to focus on specific vocal dimen-sions, rather than, or as well as, making an overall judgement. Focusing listeners by using a number of rating criteria improves the consistency of judges’ responses (Wapnick et al., 1993). However, analy-sis of available rating scales that request listeners to make several judgments of vocal quality reveal that these specific dimensions may all be tapping into a single underlying construct, that of vocal quality, thereby rendering individual assessments on each dimension at least partially redundant. For example, a number of researchers (Ekholm et al., 1998; Robison et al., 1994; Wapnick & Ekholm, 1997) have found very high correlations between all dimensions of voice quality studied, (for ex-ample, “appropriate vibrato”, “resonance/ring”, “color/warmth”, and “clarity/focus”). All of these dimensions were found to converge with the over-all judgement of vocal quality.

There are, of course, other limits to the accu-racy of perceptual judgements. Perceptual stud-ies produce an unnatural assessment setting for teachers and assessors of singing. Listeners have commented on their difficulty in assessing sing-ing when the duration of the sample is insufficient (Ekholm et al., 1998). In forced choice situations in which listeners are asked to make dichotomous choices, listeners can reliably make gross judg-ments such as identifying girl from boy singers (Howard, Szymanski, & Welch, 2002). However, listeners could not identify singers from speakers during speaking rather than singing tasks (Brown, Rothman, Morris, & Sapienza, 2001; Brown, Rothman, & Sapienza, 2000), or the same sing-ers singing at different pitches throughout their singing range (Erickson, 2003; Erickson & Perry, 2003). This lack of identification may have been due to the variability in the stimuli rather than a lack of sensitivity in the listeners, and these issues are addressed in the design of this study.

Perceptual identification of open throat

To date, there has been no research that assesses the sound qualities associated with particular vo-cal techniques. In Western classical singing, the

focus remains on the overall good sound quality (Davidson & Da Costa Coimbra, 2001) whereas pedagogical literature focuses on mastery of a wide variety of techniques to achieve this final sound (Miller, 1998; Stark, 1999).

In this study, we investigated the perceptual sig-nificance of ‘open throat’ technique by compar-ing the same female classical singers when they consciously used maximum open throat technique and reduced open throat technique in their singing of messa di voce (a crescendo-diminuendo on a single note of long duration), a classical song and a romantic lied. The aim of the study was to assess expert pedagogues’ perception of OT technique in singing, to explore its acoustic correlates, and to assess the degree of agreement between percep-tual and acoustic factors.

METHOD

Participants

ListenersListeners were 15 experienced singing peda-

gogues, 12 females and 3 males aged between 37 and 76 with a mean of 54 years. Six partici-pants had a postgraduate qualification in sing-ing, five had a diploma of music or singing, and three a bachelor degree in music. One cited exten-sive international performing experience as their qualification in singing. All had taught singing for 4 to 40 years, with an average for 20 years. Thirteen of fifteen participants taught singers in a Conservatorium of Music. Overall, participants’ singing studios comprised an average of 39.5% of operatic students and 36.7% classical students. For 11 pedagogues, the majority of their studio comprised these two genres. Six pedagogues also taught ≥20% of musical theatre students in their studio. Eleven taught a proportion of international and national level singers, nine at big city or re-gional/touring and eight at local community level. Participants were either known to the researchers via affiliations with key music centres in Australia or volunteered in response to an advertisement in a national singing organization newsletter. Partici-pants were sent information about the project and were invited to take part in a perceptual study of singing technique. They were required to partici-pate in a single listening session at a time and lo-cation convenient to them.

Prior to commencement, participants completed a questionnaire regarding their musical and teach-

96


ing experience and current singing studio. Each participant was asked for demographic informa-tion and to classify their singing studio according to their students’ performing levels and singing styles (Bunch & Chapman, 2000).

SingersSix female singers, three sopranos and three

mezzo-sopranos volunteered to participate in this project. They were advanced students with excel-lent technique studying with an experienced singing pedagogue, a Lecturer in Vocal Studies and Opera at a state Conservatorium of Music in Australia. This institution is considered the premier institu-tion for musical education in the country, having produced several singers of international repute. Criteria for participant selection included, through this pedagogue’s assessment, singers who: (1) had a good classical singing technique for their level of training and experience; and (2) understood and demonstrated skilful control of ‘open throat’ or ‘retraction’ techniques in their singing. Singers were sent information about the project and were required to attend a single recording session. They were told that the object of the study was to inves-tigate acoustical and perceptual features of the use of open throat in singing.

Prior to the voice recording, participants com-pleted a questionnaire. Singer participants were aged between 23 and 30, with a mean of 26 years. All had studied singing for at least 7 years (aver-age 9.8 years) and had spent an average of 5 years studying with their present singing teacher. Each singer held a qualification in singing or music (four had Bachelor of Music degrees and two had diplomas, in music and/or singing) and 5 of 6 were currently undertaking a second degree in singing (3 postgraduate Diploma of Opera and 2 Bachelor of Music degrees). All defined the majority of their singing as operatic (>50%), with the second most common style classical (>20%) (Bunch & Chap-man, 2000). All reported that they were in good health and able to perform the tasks.

Singer Protocol

The Musical TasksThree musical tasks were chosen in order to test

different demands of good singing, but were not musically difficult. They were designed to test the use of open throat, and contained musical features derived from a previous qualitative study on the use of the technique (Mitchell et al., 2003), where use or lack of the technique was deemed to be particularly valuable or noticeable. These features

were: high tessitura, sustained or legato singing, dynamic range control, and vocal agility.

The messa di voce, a crescendo-diminuendo on a single note is a tool for accomplishing and maintaining uniform timbre throughout the sing-ing range (Miller, 1996). The singers sang three messe di voce across their passagio beginning with pitches determined by their voice type, soprano or mezzo soprano (A4, G4, G-flat4).

The Mozart song Ridente la Calma, K 152, bars 1-27 (Figure 1a) was selected as it is a nominally simple song in the Italian language (Wapnick & Ekholm, 1997) with a mixture of common musi-cal statements involving repeated legato lines as well as the initial stylised leaps of a major 4, and short scale figures. All 6 singers sang this aria in the same key (F major).

The third verse of the Schubert lied, Du bist die Ruh D. 776 (Op. 59, No. 3) (Figure 1b), bars 54 to 80 was chosen for its demanding vocal control, sustained musical line and high climactic tessitu-ra. The three sopranos and 3 mezzo-sopranos sang this in an appropriate key depending on voice type (E-flat, D-flat and C major).

Experimental conditionsSingers sang each of the two song excerpts un-

der three conditions: optimal, sub-optimal and loud sub-optimal. ‘Optimal’ (O) provided the best possible sound quality the singer could produce using their best open throat technique. ‘Sub-opti-mal’ (SO) condition involved the use of a reduced (open throat) technique but still with an acceptable singing technique and without consciously alter-ing any other aspect of their technique. The loud

Figure 1a: First 7 bars used as perceptual stimuli of Mozart song Ridente la Calma, K 152.

&b8

3‰ ‰

R

œ .œ

œ j

œ

J

œ .œœ

J

œ

Ù œœœ œœ œ œ œ œ

œ œ

œ œ

J

œ®

œœœ

&b

6

œ œ œ œ œœ

r

œ

r

œ

‰ ‰ ∑ ‰ ‰ ≈

R

œ .œœ

j

œ ≈

R

œ .œ

œ

j

œ≈R

œ

≈

Ri - den - te la cal - ma, nell' al - ma si - desti, nell'

al - ma si desti.

&b

b

b 8

3

J

œ .

J

œ

R

œœb

J

œ œb

J

œ œ

J

œ œ

J

œ

.œ œœ j


&b

b

b

6.œ

.œ

Dies Au - gen zelt, Von dei - nem Glanz Al - lein er -

hellt.

Figure 1b: Bars 54 to 60 used as perceptual stimuli of the Schubert lied, Du bist die Ruh D. 776 (Op. 59, No. 3).

97


sub-optimal condition provided additional acous-tic information discussed in earlier papers (Kenny & Mitchell, in press; Mitchell & Kenny, 2004a, 2004b).

Each task was performed twice in the O and SO conditions. On completion of each O and SO con-dition in each task, participants were asked to self-rate their singing on a 10 cm visual analogue scale (VAS) for ‘openness’ and ‘my best singing’. Each singer completed six of these descriptive self-rat-ing scales.

Reliability check

A pedagogue was present during the recording sessions to provide accompaniment for warm-ups and practice of the tasks where necessary. Prior to the commencement of the study, the students and the pedagogue had several practice sessions in which the singer was instructed to use either O or SO at random. The pedagogue indicated which version she thought the singer had used. Practice with each singer continued until both the singer and the pedagogue reached 100% agreement on the occasions that the technique was applied or reduced.

Recording

The voice was recorded using a high-quality mi-crophone (AKG C-477) positioned on a head boom a constant 7 cm distance from the singer’s lips. This ensured that direct energy of the performers’ voices was recorded rather than room reflections, enabling us to use a studio environment with low ambient noise rather than an anechoic studio (Ca-brera, Davis, Barnes, Jacobs, & Bell, 2002). The signal was then amplified using a Behringer Ul-tragain preamplier and digitally recorded to a CD recorder (Marantz CDR 630).

Calibration was carried out in each recording by playing pink noise samples immediately follow-ing each recording session at the same recording gain used for recording the singer’s voice. For cal-ibration of absolute sound pressure levels (SPL), a sound level meter (Rion NL-06 SPL) was placed adjacent to the AKG microphone 7 cm from a speaker (Bose Lifestyle) from which the pink noise was played. The SPL shown on the sound level meter was noted for the pink noise signal and used later for calibration.

Pink noise enables calibration of an audio play-back system across the frequency range for tasks such as perceptual testing and comparative analy-sis on a computer. Pink noise is at least as good as

any other steady-state known signal. A pure sine wave tone is more susceptible to interference in the environment, and hence would give less re-peatable results.

This data set was used as a basis for perceptual and acoustic evaluations.

Perceptual Test

StimuliThe audio recordings of the singers were digit-

ally extracted from CD using Audiograbber soft-ware (www.audiograbber.com) on a standard PC computer to wav audio format in stereo at 41000 Hz and 16 bit sample rate. This methodology en-sures as close as possible that the original recorded sound was played back to the judges (no filtering or normalization was applied to minimize the ef-fects of digital artifacts).

The recorded pink noise for each singer was used to equalize the peak levels of each sample to ensure that relative SPL for each singer was the same. The amplification tool in Cool Edit calcu-lated the SPL necessary to equalize each sample’s peak SPL, thus making each recording relative to the level (in dB) of the other samples produced on the day.

The files were then edited in Cool Edit Pro 1.2 (www.adobe.com) and final samples saved to CD.

ProcedureThe perceptual test was conducted in a quiet en-

vironment and samples were played on CD from a Sony CD Walkman (DEJ885W) via circum-aural closed-back stereo monitoring headphones (Sen-nheiser HD 270). This enabled the study to be con-ducted in participants’ singing studios. This meth-odology was favoured as optimal sound quality, rather than using a computer sound card (Erick-son & Perry, 2003) or sending tapes to participants (Ekholm et al., 1998; Wapnick & Ekholm, 1997). Prior to presentation of stimuli, participants were given information on the two singing conditions, O and SO, based on qualitative results in a previ-ous study (Mitchell et al., 2003). Pedagogues were instructed to assess each singing sample for use of open throat technique. In the first study, peda-gogues were presented with single notes in messa di voce and in the second study, extended singing samples in both O and SO conditions.

Single notes as one of the test stimuli were used to assess the timbral qualities of open throat technique without potential distractors such as

98


intonation (Geringer & Madsen, 1998) or diction (Wapnick & Ekholm, 1997). The term ‘timbre’ has attracted different definitions over time. Tradition-ally, ‘timbre’ was defined as the perceptual differ-ence between two sounds which are equal in pitch and loudness (ANSI, 1973). Following this defini-tion, musical ‘timbre’ has been investigated as the difference between pairs of sounds rather than the ‘timbre’ of a single sound. More recently, ‘timbre’ has been defined as overall sound quality (Handel & Erickson, 2004). Since the aim of this study was to assess the perceived overall sound quality when a pedagogical technique is applied, we assessed the messa di voce responses individually.

Study 1: Messa di voce

For the messa di voce pairs, participants were asked to assess timbre in each sung note. Each pair contained a combination of O and/or SO by the same singer. There were four possible paired com-binations: O/O, O/SO, SO/O and SO/SO. Each listener had a trial session with four sample pairs, O/O, O/SO, SO/O, SO/SO before commencing the study. There were 720 samples of messa di voce in 360 pairs.

Study 2: Song samples

In the second study, participants were presented with individual samples of Mozart or Schubert. They were presented with the musical score of each musical task and asked to assess each sample for use of open throat technique. Each listener had a trial session with four samples, 2 O, 2 SO for each of the musical tasks (Mozart and Schubert) before starting the second perceptual test. Each of the 15 judges responded to 24 pairs of singing samples plus six repeats.

Study design

Assessment of inter-rater agreement between judges was calculated using Cohen’s Kappa (Siegel & Castellan, 1988). Kappa varies between -1 and 1, where 1 is perfect agreement and 0 is agreement not better than expected by chance. Although its use in medical settings has been criticised because it does not adjust for bias (the difference in rated prevalence of a disorder between judges), this is not a problem in a controlled experimental design where the prevalence of the two conditions being rated is identical.

As a test statistic, Kappa can verify that agree-ment exceeds chance levels. It is an omnibus in-dex of agreement and does not make distinctions among various types and sources of disagreement.

Further, Kappa is influenced by distributions and base-rates. As a result, Kappas are seldom com-parable across studies, procedures, or populations. Thus, attempts to categorize ranges of kappa as “good”, “fair”, or “poor” may be inappropriate. Krippendorf (1980) advocates agreement of at least .70.

Altman (1991) argued that any value of Ka-ppa below .50 would indicate “poor” agreement. Fleiss (1981) recommends the following Kappa benchmarks: < .40 = poor agreement; 0.40-0.75 = intermediate to good agreement; > .75 = excel-lent agreement. Gardner (1995) argues that Kappa should exceed .70 before one proceeds with ad-ditional data analyses. McGinn, Wyer, Newman, Leipzig & Guyatt, (2004) state that a safe rule of thumb is that .5 indicates moderate agreement, and anything lower is considered fair to poor agree-ment.

RESULTS

MESSA DI VOCE

In the first analysis, judges’ individual ratings of samples as either O or SO were assessed. Table 1 shows the results of these ratings.

A number of tests were conducted to determine the reliability of the judges in their ratings of the samples (Cicchetti & Feinstein, 1990). All confi-dence intervals were set at 95%. All indices dem-onstrated high levels of accuracy in the ratings of these judges. The measure of efficiency (ie correct overall classification rate) was 0.8597 (95% CI: 0.8322 - 0.8843); sensitivity (ie the proportion of true positives rated as positive) = 0.8111 (95% CI: 0.7668 - 0.8502) and specificity (ie the number of true negatives rated as negative) = 0.9083 (95% CI: 0.8737 - 0.9361). The chance levels of effi-ciency were 50%; of sensitivity, 45.14% and of specificity, 54.86%. The overall misclassification rate was 0.1403 (95% CI: 0.1157 - 0.2094); the false positive rate = 0.0917 (95% CI: 0.0639 - 0.1263); and the false negative rate = 0.1889 (95% CI: 0.1498 - 0.2332).

Cohen’s Kappa was calculated to test the hy-pothesis that Kappa = 0. The hypothesis was not confirmed (z=19.40, p =0.0000. Cohen’s Kappa = 0.7194 (95% CI: 0.6690 - 0.7699), indicating agreement substantially above chance among this sample of judges on this perceptual test.

The reliability of the best and worst judges were

99


tested individually. For the best judge, the measure of efficiency = 0.9792 (95% CI: 0.8893 - 0.9995); sensitivity = 0.9583 (95% CI: 0.7888 - 0.9989) and specificity = 1.00 (85%); overall misclassifi-cation rate = 0.0208 (95% CI: 0.0005 - 0.1398); and false negative rate = 0.0417 (95% CI: 0.0011 - 0.2112). Cohen’s Kappa = 0.9583 (95% CI: 0.8776 - 1.0391) indicated almost perfect agreement in the best judge’s responses. (z=6.65, p =0.0000)

For the worst judge, the measure of efficiency = 0.6458 (95% CI: 0.4946 - 0.7784); sensitivity = 0.5833 (95% CI: 0.3664 - 0.7789) and specificity = 0.7083 (95% CI: 0.4891 - 0.8738). The overall misclassification rate = 0.3542 (95% CI: 0.2216 - 0.5389); the false positive rate = 0.2917 (95% CI: 0.1262 - 0.5109); and the false negative rate = 0.4167 (95% CI: 0.2211 - 0.6336). Cohen’s Kappa = 0.2917 (95% CI: 0.0232 - 0.5601) indicated poor agreement in the worst judge’s responses (z=2.04, p =0.0417). These data are presented in Table 2.

Song samples

In the individual song samples, listeners identi-fied the use of open throat in 84% of O samples, and 69% of SO (Table 3). They were more likely to make a correct identification in the Mozart task (85% correct) than in the Schubert task (68%). Listeners identified > 83% of Mozart O and SO, and Schubert O. They were least reliable in judg-ing Schubert SO (53%).

In 180 song samples, for O, judges rated 80 of 90 correctly in Schubert and 70 of 90 correctly in Mozart, that is, judges were less consistent in rat-ing the Mozart task (k=.3) than the Schubert task (k=.47) in both O and SO.

Reliability of judges

The 15 judges identified the experimental con-dition in 372 of 450 samples (82.7%). There is a 95% chance that this sample comes from a popu-lation where the true proportion comes from either side of 0.827. Kappa analysis was performed to assess listener consistency. Twelve of fifteen judg-es were moderately consistent in their judgments (k≥0.600); judges 1, 5 and 7 were inconsistent in their judgements. These data are presented in Ta-ble 4.

Individual singers

Individual singers’ O samples, in both Mozart and Schubert were more readily identified by lis-teners than SO. Figure 2a-d presents these data graphically, showing the correct identifications and misclassifications of O and SO for each sing-er in each musical task. Four of six singers (1-4) were consistently identified correctly as singing with open throat in O and less open throat in SO. Singers 5 and 6 were outliers. Judges incorrectly judged Singer 6’s intentions in Mozart O and SO

All judgments

Sample O SO

O 81.11% 18.89%

(89.85%) (17.22%) 50.00%

SO 9.17% 90.83%

(10.15%) (82.78%) 50.00%

(45.14%) (54.86%) N = 720

Table 1: Messa di voce ratings. Percentages of all correctly and incorrectly identified samples by Overall percentages of correctly and incorrectly judged samples. Top row represents row percent-ages and bottom row in parentheses represents col-umn ratings.

Best Judge (3) Worst Judge (1)

O SO O SO

O 23 1 14 10

(95.8%) (4.2%) (58.3%) (41.7%)

SO 0 24 7 17

(0.0%) (100.0%) (29.2%) (70.8%)

Table 2: Messa di voce responses of the best and worst judges. The top row represents the total number of judgments from 24 responses to O and 24 to SO and the bottom row represents the corre-sponding percentages of correctly and incorrectly identified samples.. Task Hit % Miss %

All O 84% 16%

All SO 69% 31%

Mozart O 86% 14%

Mozart SO 84% 16%

Schubert O 83% 17%

Schubert SO 53% 47%

Table 3: Song responses. Percentages of correct (hit) and incorrect (miss) responses to the song task, by condition (optimal and sub-optimal) and by condition and task (Mozart and Schubert).

100


and identified her O as SO and SO as O in over half of the 30 judgements. Singer 5 was incorrect-ly identified in Schubert SO only, that is, judges viewed her singing in SO as O.

Singers’ self-perceptions of their sound

Overall means of singers’ VAS results indicated that singers associated their performance of the O condition with a high degree of openness and strongly representative of their best singing (Fig-ure 3). Subject 6 was dissatisfied with all of her singing on the test day and cast “informal votes” when asked to rate her singing. Her responses were omitted from the final result.

These responses are in accordance with the sing-ers’ intentions in their performance in each con-dition. Pearson r varied between the responses to best singing and openness according to condition and task. The O condition responses were moder-ately correlated in the Mozart task (r=0.45), but highly correlated in the Schubert (r=0.84) but nei-ther was statistically significant (p>0.05). In the

SO condition, responses were both very highly correlated (r=0.98, r=0.94), and significantly dif-ferent from zero (p<0.01, p<0.05). At least 90% of the singers’ perception of good singing can be attributed to their perception of openness.

DISCUSSION

This study confirmed that the use of open throat technique is a perceptual reality to singers and that it produces a specific vocal quality in classi-cal singing that can be reliably identified by expert listeners. While previous acoustic studies proved inconsistent in differentiating between experimen-tal conditions O and SO, the human ear produced the most reliable assessment of vocal quality.

Messa di voce

Collectively, judges were accurate in their iden-tification of O and SO and in each messa di voce sample, the majority of listeners correctly recog-nised the use of open throat when it occurred. Re-liability of listeners’ judgements ranged from fair to highly consistent. The high proportion of cor-rect judgements of O and SO suggests that these singers produced a specific sound quality or tim-bre that expert pedagogues could reliably differen-tiate, thus strengthening the claim that open throat technique is associated with a particular desirable vocal quality.

In a previous study, expert pedagogues confirmed their collective understanding of the technique of open throat and its use as an essential pedagogical instruction in the singing studio (Mitchell et al., 2003). This study confirmed their assertions in the previous study that open throat technique produc-es a specific audible quality, readily identifiable by vocal pedagogues.

Perceptually, we considered it an important methodological advance to assess open throat tech-nique in single notes (Erickson, 2003; Erickson & Perry, 2003) before progressing to extended sing-ing samples (Ekholm et al., 1998; Howard et al., 2002; Robison et al., 1994; Wapnick & Ekholm, 1997). While extended singing samples are more representative of listeners’ and pedagogues’ expe-riences in the singing studio and in singing assess-ment, testing timbre of single notes removes any influence of intonation, diction and musical stim-uli on judgements and focuses listeners directly to core sound quality. Proficiency in the messa di voce is a key indicator of vocal mastery (Miller, 1996) and these samples provided listeners with

Listener Kappa Asymp. Std. Error(a)

Approx. T(b)

Approx. Sig.

1 0.400 0.167 2.196 0.028

2 0.867 0.091 4.747 0.000

3 0.733 0.123 4.053 0.000

4 0.800 0.109 4.392 0.000

5 0.467 0.161 2.556 0.011

6 0.867 0.090 4.790 0.000

7 0.200 0.179 1.095 0.273

8 0.600 0.145 3.316 0.001

9 0.733 0.124 4.017 0.000

10 0.667 0.133 3.727 0.000

11 0.733 0.120 4.168 0.000

12 0.667 0.133 3.727 0.000

13 0.667 0.136 3.660 0.000

14 0.600 0.134 3.586 0.000

15 0.800 0.107 4.472 0.000

Table 4: Reliability of judges’ responses. Cohen’s Kappa, Asymptomatic standard error (not assum-ing the null hypothesis), asymptotic standard error (assuming the null hypothesis) and approximate significance for listeners 1-15.

101


sufficient vocal information to respond to timbre. As the definition of ‘timbre’ is generally accept-

ed as overall or intrinsic sound quality (Handel & Erickson, 2004), it was less important to the aims of this part of our study to report the difference between one sound and another (ANSI, 1973).

Song samples

Twelve of fifteen expert listeners consistently identified the use of open throat technique in O and a reduction of the technique in SO in the song samples. Listeners demonstrated reliability in their judgements through the duration of the song task, repeating their judgement of O or SO in the

0

5

10

15

1 2 3 4 5 6

0

5

10

15

1 2 3 4 5 6

0

5

10

15

1 2 3 4 5 6

0

5

10

15

1 2 3 4 5 6

Mozart OptimalO hitO miss

Schubert Sub-OptimalSO hitSO miss

Schubert OptimalO hitO miss

Mozart Sub-OptimalSO hitSO miss

Tota

lres

po

nse

s

Singer

Singer

Singer

Singer

Tota

lres

po

nse

sTo

talr

esp

on

ses

Tota

lres

po

nse

s

b

c

d

a

Figure 2 Correct identifications and misclassifications for each singer (1-6) in (a) Mozart optimal (O); (b) Mozart sub-optimal (SO); (c) Schubert optimal and (d) Schubert sub-optimal.

102


six repeated samples in the perceptual test. These results support previous research that indicates high intra-judge agreement (Davidson & Da Costa Coimbra, 2001; Ekholm et al., 1998) in assessing overall vocal quality.

However, when Mitchell and Kenny assessed the use of open throat in these singers through acoustic characteristics of vibrato (Mitchell & Kenny, 2004b) and energy distribution in LTAS (Mitchell & Kenny, 2004b), their findings did not accord with experts’ perceptual evaluation (Ken-ny, Davis, & Oates, in press). While the vibrato analysis revealed significant reductions in vibrato extent and increases in onset time in SO compared to O, conventional measures of singing power ra-tio (SPR) (the difference between the peak height) and energy ratio (ER) (the difference between peak area) on LTAS of O and SO were not statistically significantly different (Mitchell & Kenny, 2004b). Changes in vibrato rate and extent, inappropriate to genre are considered indicative of poor sing-ing so it was hypothesized that testing the energy distribution in these singers’ voices in each condi-tion would identify the timbral changes associated with open throat. LTAS results were not consistent with the vibrato findings for O/SO, and SPR and ER were not sensitive to the timbral differences between O/SO identified perceptually by expert

listeners in this study.We examined our hypotheses using both extend-

ed singing samples as well as single notes (as in the messa di voce) because eachmusical task tests a different demand of good singing and the former more closely resembles singing as heard and as-sessed in the studio and in formal assessment situations (Davidson & Da Costa Coimbra, 2001; Ekholm et al., 1998; Howard et al., 2002; Robison et al., 1994; Wapnick & Ekholm, 1997).

An interesting and somewhat unexpected finding in our study was that listeners were more likely to identify the correct experimental condition in the Mozart song sample rather than Schubert. Previ-ous perceptual studies have recommended using the same musical stimuli to reduce inter-subject variables that may influence responses (Howard et al., 2002; Saunders & Holahan, 1997); yet, these findings indicated an effect of musical task. In-deed, as open throat is of critical importance in high sustained singing (Mitchell et al., 2003), the scale motif and high tessitura of the Schubert task may have proved too difficult for singers to truly reduce open throat in SO. Conversely, the Mozart task was less vocally demanding, opening up the possibility that singers require a less robust tech-nique to make the desired sound.

When Mitchell and Kenny (2004a) assessed these singers’ vocal qualities in O and SO through the energy distribution in LTAS, the effect of mu-sical task approached significance, that is, both SPR and ER were slightly greater in Schubert than in Mozart. According to current singing literature, lower ER and SPR should indicate improved sing-ing quality, by boosted energy above 2 kHz (Omori et al., 1996; Thorpe et al., 2001). A future study by these authors will examine the significance of SPR and ER rankings with perceptual judgment of overall quality (Kenny & Mitchell, in press).

Examination of correct judgements by singer in-dicated singers 1-4 were correctly identified more often than singers 5 and 6. In testing the sound quality of open throat by singer, singers 1-4 were consistently identified correctly in O and SO. Singer 5 and 6 were outliers in certain conditions and tasks. Singer 6, in the Mozart task, produced the most inconsistent responses amongst the judg-es. Her O and SO were incorrectly identified in over 50% of occurrences, suggesting there was little difference between her O and SO, and that judges collectively were unclear whether she was attempting to use maximum open throat. On the other hand, listeners had little trouble identifying

0

10

20

30

40

50

60

70

80

90

100

Mozart Schubert

OOpenness

OBestS

inging

SOOpenness

SOBestS

inging

OOpenness

OBestS

inging

SOOpenness

SOBestS

inging

VASof100m

m

Figure 3: Overall means of singers’ VAS scores from 1-100, for singers 1-5 (singer 6 omitted) to report the degree of openness and representation of her best singing for each musical task (Mozart and Schubert) in each condition (optimal and sub-optimal).

103


singer 5’s Schubert O, but were misled by her SO, judging it as O in 8 of 15 cases. The overall high quality of singer 5’s voice may have misled judges to rate her SO as O.

For an accomplished singer, it may be possible to sing using reduced open throat technique but still produce an aesthetically pleasing sound, or equally, use an adequate open throat technique but produce an aesthetically displeasing sound as in the case of singer 6. Clearly, there are many fac-tors that converge in producing an aesthetically pleasing vocal quality, of which open throat is but one. Listeners have clearly attended to multiple dimensions of the vocal production in arriving at their perceptual judgments of the singers. In fact, singers 5 and 6 were the highest and lowest scor-ing singers respectively in a study of overall vo-cal quality (Kenny & Mitchell, in press) but these rankings showed no relationship with correspond-ing rankings of SPR and ER, which are intended as objective measures of voice or voice quality.

Singers in this study were acutely aware of the impact of open throat technique on their overall singing sound. Singers 1-5 equated good singing with use of maximal open throat technique and felt their overall singing sound decreased when they reduced the technique. Matching a perform-er’s intention to a pedagogical concept is valuable to singing pedagogy. Singer 6 was dissatisfied with her singing on the day of recording, which accorded with listeners’ judgements of her vocal quality and the similarities between her O and SO. It may be, as one listener pointed out, that it is ‘hard to tell if it is a good singer singing badly, or a bad singer singing well’.

Sophisticated acoustic measurement technolo-gies are now being routinely applied to voice sci-ence and vocal pedagogy the goal of which is to provide objective measurement of voice. To date, studies that attempt to define a good voice per-ceptually and acoustically (Ekholm et al., 1998; Wapnick & Ekholm, 1997) or assess vocal mas-tery throughout musical education (Mendes et al., 2003; Vurma & Ross, 2000) by linking perceptual ratings with acoustic measurements have not as-sessed the relationship between vocal mastery and the development of specific technical proficiencies. If singers’ and expert listeners’ reports of sing-ing accord, focussing research on specific sing-ing techniques could enhance singing pedagogy. Matching optimal singing quality with performer technique or intention enhances communication between singers and teachers. These findings will

be strengthened by replication in male voices and professional singers.

Current work in singing has not sufficiently in-corporated perceptual ratings and descriptions of sound quality. Vocal quality results from a com-plex combination of acoustical parameters that in-teract. To date, no single objective evaluation cap-tures or characterizes vocal quality in a systematic way (Omori et al., 1996; Thorpe et al., 2001). The future of experimental studies in the teaching of singing depends on their incorporation into a the-ory of the singing voice and on their relation to current practice.

Most singers and teachers of singing understand vocal functioning and control through audiation, sensory feedback and mental imagery. Teachers have a mental image of the particular sound indi-vidual students are capable of producing and use verbal explanation, vocal modeling and feedback to teach this. Despite the fact that teachers and as-sessors of singing process sound using sensory and auditory methods as their primary feedback (Kitch & Oates, 1994), few perceptual studies use these professional skills in assessing vocal qual-ity. If expert listeners can reliably identify a tech-nique, such as open throat, in the overall singing sound, future perceptual and acoustic studies of voice may provide more useful information for the dissemination of current vocal practices, their as-sessment and documentation in singing pedagogy. It is now feasible to track the methods by which good singers are trained. In linking the use of spe-cific pedagogical techniques with perceptual judg-ment and acoustic measurement, future studies in voice can provide a basis for a more systematic pedagogy of the singing voice.

ACKNOWLEDGEMENTS

We thank Ms Maree Ryan for her pedagogical advice and support, Mr Peter Thomas and Dr Den-sil Cabrera for their advice on acoustical matters, the six singers for their willing participation in the project, and the reviewers for helpful comments on an earlier draft of the manuscript.

REFERENCES

Altman DG (1991). Practical Statistics for Medical Re-search. Chapman and Hall.

ANSI. (1973). Psychoacoustical terminology: S3.20. New York: American National Standards Institute.

Brown, W. S., Jr, Rothman, H. B., Morris, R. J., & Sa-

104


pienza, C. M. (2001). Acoustic analyses of trained singers perceptually identified from speaking sam-ples. Journal of Voice, 15(1), p. 25-35.

Brown, W. S., Jr., Rothman, H. B., & Sapienza, C. M. (2000). Perceptual and acoustic study of profession-ally trained versus untrained voices. Journal of Voice, 14(3), 301-309.

Bunch, M., & Chapman, J. (2000). Taxonomy of sing-ers used as subjects in scientific research. Journal of Voice, 14(3), 363-369.

Burgin, J. C. (1973). Teaching singing. Metuchen, N.J.: Scarecrow Press.

Cabrera, D., Davis, P., Barnes, J., Jacobs, M., & Bell, D. (2002). Recording the operatic voice for acoustic analysis. Acoustics Australia, 30, 103-108.

Callaghan, J. (2000). Singing and voice science. San Diego, California: Singular Pub. Group.

Cicchetti, D. V., & Feinstein, A. R. (1990). High agree-ment but low kappa: II. Resolving the paradoxes. J. Clin. Epidemiol, 43, 551-558.

Davidson, J. W., & Da Costa Coimbra, D. (2001). In-vestigating performance evaluation by assessors of singers in a music college setting. Musicae Scientiae, 5(1), 33-53.

Ekholm, E., Papagiannis, G. C., & Chagnon, F. P. (1998). Relating objective measurements to expert evaluation of voice quality in Western classical sing-ing: critical perceptual parameters. Journal of Voice, 12(2), 182-196.

Erickson, M. L. (2003). Dissimilarity and the Clas-sification of Female Singing Voices: A Preliminary Study. Journal of Voice, 17(2), 195-206.

Erickson, M. L., & Perry, S. R. (2003). Can listeners hear who is singing? A comparison of three-note and six-note discrimination tasks. Journal of Voice, 17(3), 353-369.

Fields, V. A. (1947). Training the singing voice: an analysis of the working concepts contained in recent contributions to vocal pedagogy. New York: King’s Crown Press.

Fleiss, JL. (1995) Statistical Methods for Rates and Proportions (2nd edition). New York: Wiley 1981.

Foulds-Elliott, S., Thorpe, C., Cala, S., & Davis, P. (2000). Respiratory function in operatic singing: ef-fects of emotional connection. Logopedics Phoniat-rics Vocology, 25(4), 151-168.

Gardner, W. (1995). On the reliability of sequential data: measurement, meaning, and correction. In John M. Gottman (Ed.), The analysis of change. Mahwah, N.J.: Erlbaum.

Geringer, J. M., & Madsen, C. K. (1998). Musicians’ ratings of good versus bad vocal and string perform-ances. Journal of Research in Music Education, 46(4), 522-534.

Handel, S., & Erickson, M. L. (2004). Sound Source Identification: The Possible Role of Timbre Transfor-

mations. Music Perception, 21(4), 587-610.Hemsley, T. (1998). Singing and imagination: a human

approach to a great musical tradition. Oxford; New York: Oxford University Press.

Herbert-Caesari, E. F. (1951). The voice of the mind. London: Robert Hale.

Howard, D. M., Szymanski, J., & Welch, G. F. (2002). Listeners’ perception of English girl and boy choris-ters. Music Perception, 20(1), 35-49.

Joiner, J. R. (1998). Charles Amable Battaille: pioneer in vocal science and the teaching of singing. Lanham, Md.: Scarecrow Press.

Kenny, D.T., Davis, P. & Oates, J. (2004). Music per-formance anxiety and occupational stress amongst opera chorus artists and their relationship with state and trait anxiety and perfectionism. Journal of Anxi-ety Disorders, 18, 6, 757-777.

.Kenny, D. T., & Mitchell, H. F. (in press). Acoustic and perceptual appraisal of vocal gestures in the female classical voice. Journal of Voice.

Kitch, J. A., & Oates, J. (1994). The perceptual features of vocal fatigue as self-reported by a group of actors and singers. Journal of Voice, 8(3), 207-214.

Kreiman, J., Gerratt, B. R., Kempster, G. B., Erman, A., & Berke, G. S. (1993). Perceptual evaluation of voice quality: review, tutorial, and a framework for future research. Journal of Speech & Hearing Research., 36(1), 21-40.

Krippendorf, K. (1980). Content Analysis: An Intro-duction to its Methodology. Sage Publications, Thou-sand Oaks, CA.

Landis, J., & Koch, G. (1977). The measurement of observer agreement for categorical data. Biometrics, 33(1), 159-174.

Manèn, L. (1987). Bel canto: the teaching of the clas-sical Italian song-schools its decline and restoration. Oxford: Oxford University Press.

Marafioti, P. M. (1981). Caruso’s Method of Voice Pro-duction. New York: Dover Publications.

McKinney, J. C. (1982). The Diagnosis and Correction of Vocal Faults. Nashville, Tennessee: Broadman Press.

McGinn, P. T., Wyer, C., Newman, T. B., Leipzig, R., & Guyatt, G. (2004). Tips for learners of evidence-based medicine: Measures of observer variability (kappa statistic). Canadian Medical Association Jour-nal, 171(11), 1369-1373.

Mendes, A. P., Rothman, H. B., Sapienza, C. M., & Brown, W. S., Jr. (2003). Effects of vocal training on the acoustic parameters of the singing voice. Journal of Voice, 17(4), 529-543.

Miller, R. (1996). The structure of singing: system and art in vocal technique. New York: London: Schirmer Books.

Miller, R. (1997a). The Open Throat (La gola aperta).

105


In Miller, R. (Ed.), On the Art of Singing (pp. 17-19). New York: Oxford University Press.

Miller, R. (1998). The singing teacher in the age of voice science. In Sataloff, R. T. (Ed.), Vocal Health and Pedagogy (pp. 297-300). San Diego, CA: Sin-gular.

Miller, R., 1926-. (1997b). National schools of singing : English, French, German, and Italian techniques of singing revisited / Richard Miller. ([Rev. ed.]. ed.). Lanham, Md.: Scarecrow Press.

Mitchell, H. F., & Kenny, D. T. (2004a). The effects of open throat technique on long term average spectra (LTAS) of female classical voices. Logopedics Pho-niatrics Vocology, 29(3), 99-118.

Mitchell, H. F., & Kenny, D. T. (2004b). The impact of “open throat” technique on vibrato rate, extent and onset in classical singing. Logopedics Phoniatrics Vocology, in press.

Mitchell, H. F., Kenny, D. T., Ryan, M., & Davis, P. J. (2003). Defining open throat through content analy-sis of experts’ pedagogical practices. Logopedics Phoniatrics Vocology, 28(4), 167-180.

Monahan, B. J. (1978). The art of singing: a compen-dium of thoughts on singing published between 1777 and 1927. Metuchen, N.J.: Scarecrow Press.

Nair, G. (1999). Voice tradition and technology: a state-of-the-art studio. San Diego: Singular Publishing Group.

Omori, K., Kacker, A., Carroll, L. M., Riley, W. D., & Blaugrund, S. M. (1996). Singing power ratio: quan-titative evaluation of singing voice quality. Journal of Voice, 10(3), 228-235.

Puritz, E. (1956). The teaching of Elisabeth Schumann. London: Methuen.

Reid, C. L. (1975). Voice: psyche and soma. New York: Joseph Patelson Music House.

Reid, C. L. (1983). A dictionary of vocal terminology. New York: Joseph Patelson Music House.

Robison, C. W., Bounous, B., & Bailey, R. (1994). Vo-cal Beauty: A study proposing its acoustical defini-tion and relevant causes in classical baritones and fe-

male belt singers. Journal of Singing, 51, 19-30.Saunders, T. C., & Holahan, J. M. (1997). Criteria-spe-

cific rating scales in the evaluation of High School instrumental performance. Journal of Research in Music Education, 45(2), 259-272.

Siegel, S., & Castellan, J. N. J. (1988). Nonparametric Statistics for the Behavioural Sciences (Second ed.): McGraw Hill International.

Stanley, M., Brooker, R., & Gilbert, R. (2002). Exam-iner perceptions of using criteria in music perform-ance assessment. Research Studies in Music Educa-tion, 18, 43-52.

Stark, J. A. (1999). Bel canto: a history of vocal peda-gogy. Toronto, Buffalo: University of Toronto Press.

Sundberg, J. (1977). The acoustics of the singing voice. Scientific American(March), 82-91.

Sundberg, J., Gramming, P., & Lovetri, J. (1993). Com-parisons of pharynx, source, formant, and pressure characteristics in operatic and musical theatre sing-ing. Journal of Voice, 7(4), 301-310.

Thorpe, C., Cala, S., Chapman, J., & Davis, P. (2001). Patterns of breath support in projection of the singing voice. Journal of Voice, 15(1), 86-104.

Thurman, L., & Welch, G. F. (Eds.). (2000). Bodymind & voice: foundations of voice education (Rev. ed. ed.). Collegevile, Minn: VoiceCare Network.

Vennard, W. (1968). Singing: the mechanism and the technic (5th ed.). New York: Fischer.

Vurma, A., & Ross, Y. (2000). Priorities in voice train-ing: carrying power or tone quality. Musicae Scien-tiae, 4(1), 75-93.

Wapnick, J., & Ekholm, E. (1997). Expert consensus in solo voice performance evaluation. Journal of Voice, 11(4), 429-436.

Wapnick, J., Flowers, P., Alegant, M., & Jasinskas, L. (1993). Consistency in piano performance evalua-tion. Journal of Research in Music Education, 41(4), 282-292.

Ware, C. (1998). Basics of vocal pedagogy: the founda-tions and process of singing. New York: McGraw-Hill.

106

5.1 Conclusion

Expert pedagogues recognised the voice qualities associated with use of open throat

technique and reduction of the technique. Measurements performed on the LTAS in

paper 3 were unable to differentiate between experimental conditions, whereas the

human ear produced the most perfect and integrated assessment of vocal quality.

The fifth paper addressed these problematic acoustic measurements applied to LTAS

and compared them to listeners’ ratings of the voices by ranking perceptual and acoustic

measurements for overall goodness as defined in the singing literature (Barnes et al.,

2004; Omori et al., 1996; Thorpe et al., 2001; Vurma & Ross, 2000).

107

6 Vocal quality

To date, few scientific studies link acoustic measures with perceptual judgments. There

are two such studies that provide some basis for our proposed project. One, by Wapnick

and Ekholm (Wapnick & Ekholm, 1997) established 12 generally accepted perceptual

criteria applied to the assessment of voice quality in classical singing, and the other, by

Ekholm, Papagiannis and Chagnon (Ekholm et al., 1998) applied four of these criteria

(appropriate vibrato, resonance/ring, color/warmth, and clarity/focus) and related them

to objective measurements taken from acoustic analysis of the voice signal. However,

most current research relies on acoustic findings generated by averaging data from

groups of voices in order to characterize a good voice. The dilemma and paradox in this

approach is that the averaged findings may not adequately represent any of the

individual voices used to generate the desirable acoustic parameters that become the

benchmark of vocal quality.

Studies linking perceptual judgment and acoustic evaluation of performance quality in

singing encounter numerous problems. These include descriptors of voice quality,

selection of samples to evaluate and the absence of a gold standard against which to

assess even expert judgments, since inter-rater reliability is generally very poor, and

there are not well established relationships between acoustic profiles and perceptual

judgements.

Establishing criteria specific rating schemes does guide a listener’s thought process and

focuses them on the task. However, Ekholm, Papagiannis et al. (1998) found high

108

correlations between perceptual criteria an overall judgment of quality. Using a number

of criteria, rather than a single overall judgement may provide an illusion of objectivity.

Despite the widespread use of LTAS in music and voice research to date, there has been

little consideration given to the methodologies used to analyse and compare LTAS. As

early as 1975, Jansson and Sundberg (1975) recognised the benefits of using LTAS

analysis to show the individuality of each measured timbre. Their preliminary studies

identified the impact of musical task, length of time, performer or instrument and

performance instruction on the shape and reproducibility of LTAS (Jansson &

Sundberg, 1975). To date, there has been no similar study specifically for the singing

voice, and in particular, to test the value of current analysis methods performed on

LTAS in singing.

More recently, LTAS have been used to quantify perceptual cues guiding or misleading

listeners to judge singing samples as boys or girls (Howard et al., 2002). Exemplars of

LTAS were presented to explain the acoustic cues which identified girls as girls, boys as

boys or girls as boys. This indirectly assessed the vocal quality through perceptual cues

and acoustic output. While there has been major progress in the use of LTAS in singing

research, no study cross referencing perceptual response to LTAS exemplar exists for

operatic or classical voices, or as an explicit investigation of vocal quality and acoustic

cues. In light of the extensive vocal literature, this lack of perceptual evaluation

represents a considerable defect of current acoustic evaluations.

109

Further, conventional measures of SPR and ER on LTAS have not been sufficiently

sensitive to evaluate vocal quality to a given instruction within the same singer or across

singers (Mitchell & Kenny, 2004). Having established that ER and SPR may not be

representative of vocal quality, this study aims to assess the sensitivity of SPR and ER

to evaluate and report LTAS similarities and differences in singing. Central tendencies

in LTAS have been used to represent voice types (soprano to baritone) (Sundberg,

2001) or mean trends of spectral changes between pairs of LTAS (Barnes et al., 2004;

Mitchell & Kenny, 2004; Thorpe et al., 2001). These averages of LTAS may, in fact,

not represent any one voice and may be manufactured by clustered measurement

(Mitchell & Kenny, 2004).

In our fifth paper LTAS measurements were examined with respect to the perceptual

ratings of singers. Acoustic measurements were ranked for vocal quality, specifically,

carrying power after the work of (Barnes et al., 2004; Omori et al., 1996; Thorpe et al.,

2001). As expected, there was a strong relationship between the two acoustic measures,

but there was no relationship between perceptual rankings of vocal beauty and acoustic

rankings of vocal quality. It was argued that the human ear makes the most effective

acoustic evaluation and current analyses used to evaluate singers’ LTAS are not

sufficiently sensitive to discriminate vocal quality and the conclusions of this paper

suggest that LTAS should not be viewed as templates for overall vocal quality.

110

PAPER 5

Kenny, D. T., & Mitchell, H. F. (in press). Acoustic and perceptual appraisal of vocal gestures in the female classical voice Journal of

Voice.

111

ARTICLE IN PRESS

Acoustic and Perceptual Appraisal of VocalGestures in the Female Classical Voice

Dianna T. Kenny and Helen F. Mitchell

New South Wales, Australia

Summary: Long-term average spectra (LTAS) have identified features in thesounds of singers and have compared different vocal qualities based on energychanges that occur during different vocal tasks. In this study, we comparedthe perceptual ratings of vocal quality of expert pedagogues with acousticmeasures performed on LTAS. Fifteen expert judges rated 24 samples withsix repeats of six advanced singing students under two conditions: “optimal”(O), which represented the application of the maximal open throat technique;and “suboptimal” (SO), which represented the application of the reduced openthroat technique. LTAS were performed on each singing sample, and twoconventional assessments of peak energy height [singing power ratio (SPR)]and peak area [energy ratio (ER)] were calculated on each LTAS. Perceptualscores, SPR, and ER were rank ordered. We then compared perceptual rankingswith rankings of acoustic measures (SPR and ER) to assess whether theseacoustic measurements matched the perceptual judgments of vocal quality.Although we found the expected significant relationship between SPR andER, there was no relationship between perceptual ratings of vocal samples orsingers based on SPR or ER. These findings suggest that because LTASmeasures are not consistent with perceptual ratings of vocal quality, suchmeasurements cannot define a voice of quality. Future research with LTAS toassess vocal quality should consider alternative measures that are more sensi-tive to subtle differences in vocal parameters.

Key Words: Long-term average spectra—Perceptual rankings—Voicequality.

Accepted for publication December 9, 2004.From the Australian Centre for Applied Research in Music

Performance (ACARMP), Sydney Conservatorium of Music,The University of Sydney, New South Wales, Australia.

Address correspondence and reprint requests to DiannaKenny, Australian Centre for Applied Research in Music Perfor-mance (ACARMP), The Conservatorium of Music, The Uni-versity of Sydney, New South Wales, Australia 2006. E-mail:[email protected]

Journal of Voice, Vol. ■, No. ■, pp. ■

0892-1997/$30.00� 2005 The Voice Foundationdoi:10.1016/j.jvoice.2004.12.002

1

INTRODUCTION

What qualities of the singing voice define goodvocal quality and what constitute its specific vocalfeatures? This question has come into sharp focussince the advent of acoustic analysis. Few empiricalstudies classify good sound quality in singing,although perceptual studies have attempted to de-scribe the features of good singing1–3 and link these toacoustic features of voice.

Expert listeners are generally reliable in theirjudgments of overall quality. Ratings of good andpoor performance are correlated most strongly with

112

mailto:[email protected]

ARTICLE IN PRESS

DIANNA T. KENNY AND HELEN F. MITCHELL2

tone quality and intonation.1,2,4 Strong correlationshave also been observed among different descrip-tors that assess quality, such as color/warmth, reso-nance/ring, clarity/focus, and appropriate vibrato,which indicate that these factors converge on thesame underlying construct of overall vocal quality.Similar findings were reported in Merritt et al,5

where listeners matched their perception of perfor-mance anxiety of speakers with a variety of vocaland physical features. The ratings on the separatevocal and performance features (eg, physical ease,presence, gesture, eye contact, pace, clarity, andeye contact) all intercorrelated highly. Rating theseparate features added nothing to the overall ratingof the total performance. It seems that judges applypersonal constructs to assist with their judgmentstrategies as they often cannot articulate the compo-nents of sound quality on which their judgmentswere based.6

Vocal pedagogues in the singing studio mustassess each voice and devise a technical and aes-thetic program to improve its basic sound. “Openthroat” is one technical component taught in manymodern singing studios, and pedagogues agree thatit produces an identifiable quality in the sound,which is described as “even and consistent,” “bal-anced and coordinated,” “round,” and “warm.”Mitchell et al,4 in subsequent studies, comparedthe open throat technique (O) with a reduction ofthe technique described as suboptimal (SO) in thesame advanced singing students and found that ex-perts could identify the technique in 83% ofsamples.7,8

Applied research in voice must address the goalsof singing pedagogy in assessing and refining avoice of quality and generate perceptually viablemeasures to systematically define voice. Ideally,we need exemplars of vocal quality and to find mea-surements that enable us to rank voices in accor-dance with acoustic measures. One key questionaddressed in this article is whether acoustic featuresare associated with perceptual preferences of overallquality. To date, research into the singing voice hasdescribed acoustic features of voice and its visualrepresentation with few or no links to either percep-tual preferences or pedagogy.

Recent research has struggled to identify acousticcues that attract the highest perceptual rankings from

Journal of Voice, Vol. ■, No. ■, 2005

expert listeners.2,3,9 The voice can be representedacoustically in many ways. In current singing re-search, long-term average spectra (LTAS) are widelyapplied to represent timbre and vocal features, bothin speech and in singing. An LTAS gives an overallimpression of an entire excerpt by identifying certainconsistent features contained in the sound overtime.10 In singing samples, it averages out short-termvariations in phonetic structure11 and is a measureof the decibel level of the time-average of the powerof the acoustic signal at each frequency. A conven-tional way of assessing LTAS is to reduce the infor-mation it contains to a single meaningful number, bycomputing the ratio of energies in a low- and a high-frequency band.12,13 In singing, measures of spectracompare energy peak height [singing power ratio(SPR)] and peak area [energy ratio (ER)] between0–2 and 2-4 kHz. The difference between the heightof the major peaks between 0–2 and 2–4 kHz12

quantifies the relative energy between 2 and 4 kHz,although it does not account for the shape of theseenergy peaks. Assessing the areas under the LTAScurve at 0–2 and 2–4 kHz, rather than at the highestpeaks of energy, also identifies the areas in whichthe energy is reinforced or reduced.13 Both SPR andER enable effective comparison between individualsingers and between groups of singers.

LTAS exemplars of voice types14 and singinggenres15,16 are available. LTAS curves have also dif-ferentiated male and female voices,17 singers andspeakers,18 solo voice and choral voice,19 and popor country from opera singers15,16 based on thechanges in energy distribution necessary for thesedifferent vocal tasks.

In singing literature, particular LTAS models orexemplars are associated with particular vocal quali-ties; high-range energy within the LTAS identifiesvoices that produce a “ringing quality”2,3 or carryingpower or amplification over an orchestra,9,13,14

which is considered essential to operatic voices. InBarnes et al,20 the LTAS provides information aboutthe formant structure. However, it is premature toassert a connection between projection as demon-strated in LTAS and vocal quality, as Thorpe et al13

and Barnes et al20 have been the only studies onvocal projection in which the precise standard ofthe singing subjects has been specified and a homo-geneous group of singers of national and interna-tional reputation have been used.21 However, with

113

ARTICLE IN PRESS

VOCAL GESTURES IN FEMALE CLASSICAL VOICE 3

conservatory students, Vurma and Ross9 reportedno correlation between measurable carrying powerand the perception of tone quality by experts. Theyalso found no association between these measuresand length of training.

To date, such LTAS patterns have not been consis-tently associated with “good” or “beautiful” singing.Bartholomew22 associated “good voice quality” withhigh energy but did not resolve whether qualitywas associated with energy. Since Bartholomew22

suggested that male singers produced a specificenergy around 2800 Hz, and Sundberg14 identifiedthe “singer’s formant” as a critical component ofoperatic singing, the spectral distribution of the clas-sical and operatic voice has been measured againstthe standards established by this pioneering research.From this research, the energy produced in an oper-atic voice to project over an orchestra is consistentlyassociated with performers of the highest standardsin operatic singing.13,20 Ekholm et al2 matchedacoustic features to specific descriptors, but like ear-lier studies,22 they reported relationships betweenacoustic and descriptive features and presented ex-emplars of high-energy only in the most highlyrated voices.

LTAS have also been used to “validate” perceptualcues. For example, Howard et al23 and White24 con-ducted studies in which raters were asked to identifychild voices as male or female. Both studies foundtypically different LTAS for each gender and similar-ities between the spectra of girls identified as boysand boys identified as girls. LTAS is therefore reli-able in distinguishing male from female child andadult voice.23–25 However, this is a long way fromdetermining vocal quality among like voices, partic-ularly singing voices (eg, experienced adult femaleclassical singers).

A variation of the study type that applies percep-tual ratings to validate acoustic findings is withacoustic parameters to predict qualities, such as emo-tion in voice. These studies have found judges to beinconsistent and unreliable26 and/or to have awide range of responses to the proposed emotionalcontent of the music.27–30 Where listeners are consis-tent in identifying emotional content, stimuli havea tendency to be exaggerated, for example, in vibratoparameters,30 dynamics,31 or tone onset.32

Moreover, in a previous study, Mitchell andKenny33 tested the energy distribution of singingwith and without the use of open throat technique.Although visual inspection of LTAS illustrated smalldifferences, conventional measures performed onthese LTAS (SPR and ER) did not conclusively dif-ferentiate the high energy produced with and withoutopen throat across all singers. There is no doubtthat an operatic singer must produce energy in sucha way as to amplify his/her voice over an orchestra,but the visual presentation of operatic vocal qualitydoes not necessarily reflect its perception by expertlisteners. Similarly, Titze et al34 found the acousticoutput in “twangy” quality boosted energy around3 kHz. If vocal qualities other than operatic alsoproduce marked energy between 2 and 4 kHz, tonequality may not easily be demonstrated in the acous-tic measure of LTAS. The timbral effects of theopen throat technique have not been tested withrespect to the LTAS parameters and rankings oroverall perceptual rankings. Identifying trends in ERand SPR across all singers may result in an averagethat does not represent an actual voice and may notadequately represent the timbre of any individualvoices.

In this study, we tested whether the spectral mea-sures made on LTAS in female classical singerscould assess the difference in vocal quality betweenO and SO techniques and whether the ratings ofpedagogues of overall quality matched LTAS.

METHOD

ParticipantsListeners

Listeners were 15 experienced singing peda-gogues, 12 women and 3 men, aged between 37 and76 years, with a mean of 54 years. Six participantshad a postgraduate qualification in singing, five hada diploma of music or singing, and three a bachelorsdegree in music. One cited extensive internationalperforming experience as their qualification in sing-ing. All had taught singing for 4 to 40 years, with anaverage for 20 years. Thirteen of fifteen participantstaught singers in a Conservatorium of music. Over-all, the singing studios comprised an average of39.5% operatic students and 36.7% classical stu-dents. For 11 pedagogues, most of their studio com-prised these two genres. Six pedagogues also taught


114

ARTICLE IN PRESS


�20% of musical theater students in their studio.Eleven taught a proportion of international and na-tional level singers, nine at a big city or regional/touring, and eight at the local community level.Participants were either known to the researchersvia affiliations with key music centers in Australiaor volunteered in response to an advertisement ina national singing organization newsletter. Partici-pants were sent information about the project andwere invited to take part in a perceptual study ofsinging technique. They were required to participatein a single listening session at a time and locationconvenient to them.

Before commencement, participants completed aquestionnaire detailing their musical and teachingexperience and current singing studio. Each partici-pant was asked for information related to age,number of years teaching, and highest qualificationsattained in music and/or singing. Participants wereasked to classify their singing studio according totheir students. Specifically, they gave the proportionof their private studio and studio at a musical conser-vatorium. Finally, participants were asked to classifytheir studio according to their primary singing genres(opera, contemporary music theater, musical the-ater, concert/oratorio/recital, choral) and the level atwhich their students performed (superstar, interna-tional, national/big city, regional/touring, local com-munity, full-time students of singing, amateur) andestimate the percentage each played in the per-forming career of their students.21

SingersSix female singers, three sopranos and three

mezzo-sopranos, volunteered to participate in thisproject. They were advanced students with excellenttechnique of an experienced singing pedagogue, whois a lecturer in vocal studies and opera at a stateConservatorium of Music in Australia. It is thepremier institution for musical education in thecountry and has produced singers of internationalrepute. Criteria for participant selection includedsingers who (1) had a good classical singing tech-nique for their level of training and experience and(2) understood and demonstrated skillful control of“open throat” or “retraction” techniques in theirsinging.


Before the voice recording, participants com-pleted a questionnaire seeking information on age,years of singing study, number of years of studywith each singing teacher and highest qualificationsattained or currently undertaken in music and/orsinging, and singer type (soprano or mezzo soprano).The participants were also asked to classify thegenres of singing they performed in public (opera,classical, choral, music theater, and contemporary)and to estimate the percentage each style played intheir total performing career.

Singer participants were aged between 23 and 30years, with a mean of 26 years. All had studiedsinging for at least 7 years (average 9.8 years)and had spent an average of 5 years studying withtheir present singing teacher. Each singer held aqualification in singing or music (four had Bachelorof Music degrees and two had diplomas, in musicand/or singing) and five of six were currently under-taking a second degree in singing (three postgraduateDiploma of Opera and two Bachelor of Musicdegrees). All defined most of their singing as oper-atic (�50%), with the second most common style asclassical (�20%), in accordance with the Bunchand Chapman21 taxonomy of singing voices. Allreported that they were in good health and couldperform the tasks.

Singers were sent information about the projectand were invited to take part in an acoustic andperceptual study of singing technique. They wererequired to attend a single recording session lastingno more than an hour and were told that the objectof the study was to investigate acoustical and percep-tual features of the open throat in singing and todiscover the sound qualities associated when a singeruses some form of open throat technique comparedwith when a singer does not use the open throattechnique.

Singer protocolA protocol was developed to assess the effect

of “open throat” technique on singing. Two musicaltasks were chosen to investigate the applicationof the technique in two song excerpts. Before singing,each singer selected the sequence of their tasksbefore commencing the experiment by selecting ablank card, the reverse side of which representedone task, to reduce the possible effects of task order.

115

ARTICLE IN PRESS


The musical tasksMusical tasks were chosen to test different de-

mands of good singing, but they were not musicallydifficult. They were designed to test the applicationof open throat and contained musical features de-rived from a previous qualitative study on the appli-cation of the technique,4 where application or lackof application of the technique was deemed to beparticularly valuable or noticeable. These featureswere high tessitura, sustained or legato singing, dy-namic range control, and vocal agility.

The Mozart song Ridente la Calma, K 152, bars1–27 (Figure 1A), was selected as it is a nominallysimple song in the Italian language1 with a mixtureof common musical statements involving repeatedlegato lines as well as the initial stylized leaps of amajor 4 and short scale figures. All six singers sangthis aria in the same key (F major).

The third verse of the Schubert lied, Du bist dieRuh D. 776 (Op. 59, No. 3) (Figure 1B), bars 54 to80, was chosen for its demanding vocal control,sustained musical line, and high climactic tessitura.The three sopranos and the three mezzo-sopranossang this in an appropriate key depending on sopranoor mezzo-soprano voices (E-flat, D-flat, and Cmajor).

Experimental conditionsSingers sang each of the two song excerpts under

three conditions: O, SO, and loud SO (LSO). Oinvolved the maximal application of open throat.SO involved a reduced (open throat) techniquebut still with an acceptable singing technique andwithout consciously altering any other aspect of theirtechnique. It was hypothesized, from interviews withpedagogues,4 that the SO condition would result ina reduction of sound pressure level (SPL), so a thirdcondition, LSO, involved the same instruction asthe SO condition, but with the added instruction thatthe singer should try to achieve a louder dynamic thanin SO. This condition addressed SPL as an additionalvariable to the SO condition and was similar totechnical conditions established by Foulds-Elliottet al26 in her study of emotional connection.

Each task was performed twice in the O and SOconditions, whereas the LSO condition was per-formed only once because of the slight possibility

of adverse effects on vocal health. In total, eachsinger performed each musical task five times.

InstructionsA pedagogue was present during the recording

sessions to provide accompaniment for warm-upsand practice of the tasks where necessary and toinstruct singers to achieve the required vocal pos-tures for each experimental condition. For example,she instructed them to pay attention to producingthe most open sound in their throat in the O condi-tion, and to a lesser degree in the SO and LSOconditions. Some singers asked how to produce theLSO condition, and the pedagogue instructed themto “use more twang,” as taught in their lessons.35

RecordingParticipants were given time to warm up in the

singing studio and become familiar with the roombefore recording. Recording levels for each singerwere set during this time. The voice was recordedwith a high-quality microphone (AKG C-477, AKGAcoustics, Vienna, Austria) positioned on a headboom, a constant 7-cm distance from the lips of thesinger. This placement ensured that the direct energyof the voices was recorded rather than room reflec-tions, which enabled us to use a studio environmentwith low ambient noise rather than an anechoicstudio.36 The signal was then amplified with a Beh-ringer Ultragain preamplier and digitally recordedto a CD recorder (Marantz CDR 630, Marantz JapanInc., Kanagawa, Japan).

Calibration was carried out in each recording byplaying pink noise samples immediately after eachrecording session at the same recording gain appliedfor recording the voice. For calibration of absoluteSPLs, the a sound level meter (Rion NL-06 SPL,Rion Co. Ltd, Tokyo, Japan) was placed adjacentto the AKG microphone 7 cm from a speaker (BoseLifestyle, Bose Corporation, Framingham, MA),from which the pink noise was played. The SPLshown on the sound level meter was noted for thepink noise signal and applied later for calibration.

Pink noise enables calibration of an audio play-back system across the frequency range for tasks suchas perceptual testing and comparative analysis ona computer. Pink noise is at least as good as anyother steady-state known signal. A pure sine wave


116

ARTICLE IN PRESS


FIGURE 1. A. We used bars 1 to 7 as perceptual stimuli of the Mozart song Ridente la Calma, K 152. B. We used bars 54 to 60as perceptual stimuli of the Schubert lied, Du bist die Ruh D. 776 (Op. 59, No. 3).

tone is more susceptible to interference in the envi-ronment and hence would give less repeatableresults.

This data set was applied as a basis for the percep-tual and acoustic evaluations.

Preparation of perceptual CDsThe audio recordings of the singers were digitally

extracted from the CDs on a standard PC to waveaudio format in stereo at a 16-bit, 41,000-Hzsample rate [with Audiograbber software (www.audiograbber.com)]. This methodology ensures asclose as possible that the original recorded soundwas played back to the judges (no filtering or normal-ization was applied to minimize the effects ofdigital artifacts).

The recorded pink noise for each singer wasapplied to equalize the peak levels of each sampleto ensure that relative SPL for each singer wasthe same. The amplification tool in Cool Edit Pro1.2 (www.adobe.com) calculated the SPL necessaryto equalize the peak SPL of each sample, thus


making each recording relative to the level (in deci-bels) of the other samples produced on the day.

The files were then edited in Cool Edit. We usedthe first 7 bars of Mozart and bars 54 to 60 of theSchubert musical tasks for the perceptual study. Eachconsisted of an entire musical phrase lasting around20 seconds. Only samples with correct intonationwere selected in the study. We used 24 CD tracksin the study: six singers singing Mozart O and SOand Schubert O and SO. CD tracks of the sampleswere presented in random order (generated atwww.randomizer.org), and six additional repeatstracks were generated, three O and three SO, ofwhich three were the Mozart task and three wereSchubert.

Pilot studiesTwo pilot studies were conducted to consider the

methodological issues discovered in this perceptualstudy. The first pilot study was conducted, witha single expert pedagogue, to determine whetherthe three experimental conditions performed by thesingers were detectable in a perceptual study. The

117

ARTICLE IN PRESS


samples of three subjects were presented in eachsinging condition (O, SO, and LSO) in random order.LSO compared with SO did not produce a suffi-ciently discernible voice quality to be correctly dis-tinguished by an expert pedagogue. The pilot testlistener marked all LSO samples as SO and remarkedthat it was not possible to differentiate between thetwo conditions, SO and LSO. As LSO was createdas acoustic confirmation of SO, it is unreasonable toexpect the human ear to detect so small a differencein timbre. The LSO condition was generated toensure acoustic differences could be attributed to theopen throat technique and not solely to loudness.The LSO condition was therefore deleted from theperceptual study.

The second pilot study addressed normalizationof SPL of perceptual samples. Normalizationreduces the possible effects of SPL as a variable. Totest the perceptual difference in normalized samples,samples were normalized with the amplification toolin Cool Edit, saved, and then reduced back to theiroriginal SPL. Pairs of samples, each containing onenormalized and returned and one original, werepresented to two expert pedagogues. Pedagoguesdescribed a discernable difference in quality, saying,for example, “the overall recording quality is re-duced” or the overall timbre had changed, with dif-ferences in the “woofy” and “tweety” balance inthe recording.

Although other perceptual studies have normal-ized SPL across stimuli,23 singing pedagogue sub-jects in Wapnick and Ekholm1 commented on thedifficulty of evaluating voices not presented withrealistic loudness. On the basis of the pilot study,we decided not to normalize samples but insteadmake them relative to known SPL, based on SPLat the recording.

Perceptual testProcedure

The perceptual test was conducted in a quiet envi-ronment, and samples were played from a SonyCD Walkman (DEJ885W; Sony Corporation, Tokyo,Japan) via closed-back stereo monitoring head-phones (Sennheiser HD 270, Sennheiser, Tullamore,Ireland). This setup enabled the study to be con-ducted in the singing studios of the participants.With headphones, it was possible to eliminate room

effects from the listening environment. This method-ology was favored as optimal sound quality, ratherthan with a computer sound card37 or sending tapes toparticipants.1,2 Only one listener took the test on anyoccasion, and each listener took the test once. Levelswere checked before each test was conducted andset at a consistent and comfortable SPL volume foreach subject.

Before presentation of stimuli, participants weregiven information on the two singing conditions, Oand SO, and were presented with the musical score ofeach musical task. They were asked to rate the overallvocal quality of each sample on a 10-point scale ata standard relevant to advanced singing students at anational Conservatorium of Music. They were alsoasked to give qualitative comments and recommen-dations about the sound quality. Each listener hada trial session with four samples, two O and twoSO, in each musical task before starting the percep-tual test.

Acoustic analysisThe audio recordings were digitized at a 16-kHz

sample rate with Phog Version 2.0 (Hitech, Sweden)software and analyzed with Soundswell Version 4.0(Hitech, Sweden), with A-weighting for SPL pinknoise calibration and subject SPL measurement.

LTAS analyses (bandwidth 300 Hz) wereperformed on the 24 musical task files.25,38 LTAScontained only voiced data.11,25 A 300-Hz LTASbandwidth is less sensitive than a narrow-band anal-ysis to movement in the partials when the singer isat higher fundamental frequencies. A complicationwith LTAS is to account for loudness, which is prob-lematic as frequencies above 2 kHz increase fasterthan those below 2 kHz as SPL increases.39

To relate these LTAS plots to the known decibelof the calibration tone, an LTAS was performed onsteady 5-second portions of the pink noise samplefor each singer, with the same 300-Hz bandwidth.The Sect tool in Soundswell does not compute theoverall equivalent level (Leq) for an LTAS.Therefore, to calculate the absolute SPL of eachcalibration tone, it was necessary to find the meanSPL of the pink noise LTAS. Each point of the LTAScurve was linearized [y � 10^(χ/10)], then all datapoints were summed, and this total was convertedback to decibels [χ � 10·log(y)]. The mean SPL of


118

ARTICLE IN PRESS


pink noise (in decibels down from full scale) wassubtracted from the known SPL of the calibrationnoise measured during recording to produce the cali-bration offset. This calibration offset (in decibels)was applied to each LTAS. In earlier research withLTAS methodology, White25 arrived at a “calibrationoffset” by calculating the mean SPL of a sung /i/vowel by adding the peak level (in decibels) of theLTAS with the known level. Applying the LTASarea to calculate the mean SPL of pink noise waspreferential to calculating the LTAS peak height andwas considered a valid method for evaluating theoverall level (in decibels) of the pink noise at 7 cm.

Acoustic analysis focused on two frequency areas:0–2 kHz and 2–4 kHz. From the LTAS plots, thehighest peaks in the 0–2-kHz and 2–4-kHz regionswere labeled P1 and P2, respectively. Peak levels (indecibels) and peak center frequencies (in Hertz)were calculated for each singer in each task andcondition. We used two measures to quantify theseLTAS data. The SPR, described by Omori et al12

compares the peak levels of P1 and P2. The SPRis the difference between the level of P1 and P2 indecibels (LP1 – LP2) and is a measure of spectralattenuation between 0–2 kHz and 2–4 kHz. Whenthe energy is focused in the 0–2-kHz region, it resultsin higherSPR results and there is typically less energyreinforcement in the 2–4-kHz region. A low SPRindicates a stronger energy peak �2 kHz. AlthoughSPR does not prove the presence of the formant ofa singer, it enables inter- and intrasinger comparisonof spectral energy in the voice.12,40

The ER, described by Thorpe et al,13 comparesoverall energy between the area 0–2 kHz (A1) andthe area 2–4 kHz (A2). It is calculated by taking thedifference between average energy values for the twofrequency areas (A1 – A2). A low ER represents agreater reinforcement in the 2–4-kHz region,whereas a high ER represents a smaller energy boost�2 kHz. Areas below the LTAS curve were calcu-lated in the same way as the calibration tone. Whenthere is less reinforcement in the 2–4-kHz region, ERresults follow SPR.

Study designThe reliability of the ratings of the judges was

tested with the scores of the six repeats against theoriginal scores. Intraclass correlation coefficients


TABLE 1. Ranked Mean Perceptual Scores for EachSinger With Standard Deviations in Each Musical Task(Mozart and Schubert) in Each Condition (0 and SO)

Rank Singer Task Condition Mean SD

1 5 Schubert O 7.60 1.882 5 Mozart O 7.07 1.493 1 Mozart O 6.93 1.444 3 Mozart O 6.93 2.195 1 Schubert O 6.93 1.496 4 Schubert O 6.80 1.937 4 Mozart O 6.73 1.838 2 Schubert O 6.67 1.729 3 Schubert O 6.27 1.8710 2 Mozart O 5.80 1.3711 6 Schubert O 5.40 1.6412 5 Schubert SO 5.40 2.0313 1 Mozart SO 5.20 1.2114 2 Schubert SO 5.20 1.5215 6 Mozart O 4.67 1.5416 5 Mozart SO 4.40 1.5917 3 Schubert SO 4.13 1.6418 6 Schubert SO 3.87 1.2519 1 Schubert SO 3.73 1.7520 3 Mozart SO 3.67 1.2921 4 Schubert SO 3.53 1.6422 2 Mozart SO 3.47 1.4123 4 Mozart SO 3.47 1.3024 6 Mozart SO 2.67 1.23

(ICCs) were calculated with a two-way randomeffects model (absolute agreement).

The study design was a repeated-measures [24ratings per listener] randomized complete block witha 2 [Task (Mozart vs. Schubert)] x 2 [Condition(O and SO)] factorial structure. The main effects fortask and condition and interaction effects (contrasts)were calculated for each dependent measure.

Because the design of this study included foursamples from each of the six singers, the 24 samplesare potentially nonindependent, which would consti-tute a violation of the assumptions of correlationand other parametric tests. Before analyses of thesedata, each dependent measure was tested for serialdependency with autocorrelation. The criterionvalue for serial dependency was P � 0.404. Resultsindicated that each of the three measures wereindependent: perceptual ratings, P � 0.111; ER,P � 0.238; SPR, P � 0.151.

119

ARTICLE IN PRESS


FIGURE 2. Highest ranked perceptual samples. LTAS, exemplars of the qualitative positive and negative descriptions of the soundquality of pedagogues, with details of rank, singer, task, condition, and perceptual score.

RESULTS

DescriptivesMean rating scores and standard deviations calcu-

lated from averaging all measured notes sung byeach singer, in each condition, for O and SO condi-tions in each musical task (Mozart and Schubert) arepresented in Table 1. Data are ranked by mean score.

Intrajudge reliability and test–retest reliabilityIntraclass correlations with two-way random

effects model of absolute agreement (ICC(2,1))compared the exact scores of the repeated sampleswith the original ratings. For the O condition,ICC � 0.7843 [CI 95% 0.4822 � 0.9613] F � 9.52,P � 0.0001. For SO, ICC � 0.7579 [CI 95% 0.4338� 0.9558] F � 8.60, P � 0.0001; that is, listenersmatched their exact rating in the first rendition andthe repeat in nearly five of six cases. Therefore,judges were reliable in their absolute ratings.

Overall ratingsOverall means of score by condition indicated a

significant reduction in SO compared with O in both

musical tasks. For mean score, there was a maineffect for condition (F(1, 5) � 111.872, P � 0.000),but not for musical task (F(1, 5) � 1.497, P �0.276); that is, ratings were significantly higher forO than for SO in both the Mozart and the Schuberttasks. There was no interaction between musicaltask and condition (F(1, 5) � 0.355, P � 0.577).

Singer 5 was rated highest overall, in both Mozartand Schubert tasks (7.60, 7.07), and singers 1 and3 were ranked third with a score of 6.93. For the Ocondition, the mean scores were 6.34 for Mozartand 6.61 for Schubert. The O condition scoresranged from 5.40 to 7.60. For SO, singer 5 was ratedhighest for Schubert with a score of 5.40. For SO,the mean scores were 3.81 for Mozart and 4.31 forSchubert. SO scores ranged from 2.67 to 5.40. Oscores consistently scored a higher rating than didSO. Subject 6 in Mozart O was an outlier.

LTAS analysisFigures 2–5 present LTAS performed on selected

samples in the perceptual study. LTAS may provideacoustical interpretation of the perceptual ratings, orconversely, acoustic cues that influenced ratings may


120

ARTICLE IN PRESS


FIGURE 3. Third-ranked perceptual samples. LTAS, exemplars of the qualitative positive and negative descriptions of the soundquality of pedagogues, with details of rank, singer, task, condition, and perceptual score.

be evident in the LTAS. In qualitative interpretationof each sample, listeners categorized their responsesinto positive and negative comments about the vocaland technical quality and suggestions for improve-ment. In addition to LTAS, each figure providesthe overall perceptual ranking, task, condition, andscore of each sample. Finally, each figure presentsexemplars of the qualitative responses given bylisteners for these samples. These analyses wereconducted to explore the perceptual responses tothe individual samples. They illustrate the personalresponse of the listeners to the most important quali-ties contained in the sample: of the voice, the tech-nique, and the overall musical performance.

The LTAS plots of singer 5 (Figure 2) are mark-edly different than those of the other singers; thatis, above 2 kHz, she produces a different distributionof energy from the other singers.

Singers 1 and 3 produce similar LTAS plots inboth tasks and are still rated highly (Figure 3). Lowerscoring O samples showed lower amplitudes below


3 kHz and the center frequency of energy � 2 kHz,P2, at a higher frequency.

Midscoring samples included one O and three SOsamples. Figure 4 presents these LTAS data.

For lowerscoring SO samples, allLTAS plots showless-defined or prominent peaks, particularly � 2kHz. The lowest scoring SO samples (Figure 5)show a series of spectral peaks and a spectral rolloff with rising frequency with no amplified peaksabove 2 kHz.

As spectral differences were unique to each singer,the SPR of each task and condition was performed tocompare the relationship of energy peaks between0–2 and 2–4 kHz. The SPR results of the perceptualsamples are presented in Table 2, ranked from thebest to the worst SPR values. Lower SPR resultsindicate more energy between 2 and 4 kHz. The SPRranks did not correspond to the perceptual rankings.

In this study, ER was calculated for each musicaltask in each condition for each subject. The ER dataof the samples of each singer are presented in Table

121

ARTICLE IN PRESS


20

40

60

80

100

120

0 1000 2000 3000 4000 5000 6000

Hz

SPL

(dB

at 7

cm

)Positive Comments

"good foundation tone, voice is still building" "good round sound" "still solid core" "a tone with roundness, depth"

Negative Comments

"sound is constricted and unstable" "lacks freedom" "lacks vibrancy and warmth" "little colour"

Rank Singer Task Condition Mean Score SD min max

11 6 Schubert O 5.40 1.64 3 8

12 5 Schubert SO 5.40 2.03 2 8

13 1 Mozart SO 5.20 1.21 3 7

14 2 Schubert SO 5.2 1.52 3 7

Singer 6 Schubert O


Singer 1 Mozart SO


FIGURE 4. Mid-ranked perceptual samples. LTAS, exemplars of the qualitative positive and negative descriptions of the soundquality of pedagogues, with details of rank, singer, task, condition, and perceptual score.

3, ranked from the best to the worst ER. The changesto ER results by singer corresponded to SPR resultsin the Mozart and Schubert tasks from SO comparedwith O; that is, ER decreased with condition acrosssingers. The rankings of ER did not correspond toperceptual rankings.

Relationships between the rankingsTable 4 presents the rankings of each singer and

the respective musical task and experimental condi-tion, ordered by perceptual rank from highest tolowest.

The Pearson correlation coefficients were calcu-lated for each of the three dependent measures (per-ceptual rating, SPR, and ER). Table 5 shows thatalthough there is a very highly significant relation-ship between ER and SPR (as expected), there isno relationship between either of these acoustic mea-sures and perceptual judgments.

The Spearman rho correlation coefficients werecalculated for the rankings of samples and singers

on each dependent measure by the judges. Table 6displays these results.

These data indicate that there was no relationshipbetween perceptual ratings of samples or singersbased on ranked ER and SPR.

A series of ICCs (ICC2,1) were calculated to deter-mine the degree to which judges were consistent intheir rankings of samples and singers based on theirrankings with respect to ER and SPR. In the firstICC, perceptual ratings were compared with ER forsamples. The ICC was 0.012 (F � 0.987,P � 0.511), which indicated no consistency betweenthe two rankings. When this test was completed forsingers, the ICC 0.029 (F � 0.972, P � 0.527) indi-cated no consistency between the two rankings. Inthe second ICC, perceptual ratings were comparedwith SPR for both samples and singers. The ICCfor samples was 0.47 (F � 1.89, P � 0.07), whichindicated moderate consistency between the tworankings. When this test was conducted for singers,the ICC was 0.029 (F � 0.972, P � 0.527), which


122

ARTICLE IN PRESS


FIGURE 5. Lowest ranked perceptual samples. LTAS, exemplars of the qualitative positive and negative descriptions of the soundquality of pedagogues, with details of rank, singer, task, condition, and perceptual score.

indicated no consistency between the two rankings.The ICC between rankings of singers on ER andSPR was 0.6 (F � 2.5, P � 0.016), which indicated asignificant relationship between the rankings ofsingers.

DISCUSSION

In this study, we recorded six female singers whoperformed two songs with two different singing tech-niques: “open throat” (O) and “reduced open throat”(SO). As expected, the audio samples related to theO production were judged by expert listeners tobe sung with a better technique than the audio sam-ples related to the SO production. These perceptualrankings, however, did not match the spectral mea-surements of two spectral parameters (SPR and ER,as measured on LTAS). We conclude that these twospectral parameters (SPR and ER) are not useful orrelevant in assessing the vocal quality differencesbetween “open throat” and “reduced open throat”singing techniques. Because we found that the maxi-mum open throat produced a vocal quality in the


singing voice that was judged perceptually by expertlisteners to produce the highest overall vocal quality,we therefore tentatively argue that optimal openthroat is key to vocal quality and that it is this aspectof vocal quality that has not been detected in LTAS.

Reliability of judgesListeners were consistent in their ratings of overall

quality, producing exactly the same score, on aver-age, for five of the six repeated samples. There wereno statistically significant differences between rat-ings of the two musical tasks. Overall preferencesfor particular singers were also remarkably similar,and judges ranked their favorite singers (5, 1, and 3)consistently higher than the others. Of these, singer 5scored the highest overall ratings.

Role of perceptual studies in assessing voiceThis study is a methodological improvement over

previous twinned acoustic and perceptual stud-ies.2,3,23 We reduced the information in LTAS toa single number to make it possible to compare

123

ARTICLE IN PRESS


TABLE 2. Ranked SPR for Each Singer, in EachMusical Task (Mozart and Schubert) in Each Condition

(O and SO)

Rank Singer Task Condition SPR

1 3 Mozart SO 11.062 6 Mozart SO 12.013 3 Mozart O 13.004 1 Mozart SO 16.105 5 Schubert SO 16.596 4 Mozart O 17.657 5 Mozart SO 17.838 2 Mozart O 18.139 1 Schubert O 18.5110 1 Mozart O 18.7011 1 Schubert SO 18.8312 6 Schubert O 18.9013 5 Mozart O 19.5114 4 Mozart SO 19.6515 6 Mozart O 19.8116 5 Schubert O 20.2817 2 Mozart SO 21.1818 6 Schubert SO 21.5619 3 Schubert O 22.3420 3 Schubert SO 23.7621 4 Schubert O 24.2722 4 Schubert SO 25.3323 2 Schubert O 25.6524 2 Schubert SO 33.20

intersinger differences across pairs of LTAS aswell as intrasinger differences across pairs ofLTAS.12,13,20,33 Neither ER nor SPR measurementswere corroborated by the perceptual ratings of over-all quality. LTAS may successfully differentiate be-tween different qualities (opera versus country, pop,or speech), but this does not identify the timbralsubtleties that may be contained in LTAS. Ideally,a meaningful acoustic measurement would be com-parable with a perceptual rating by expert listeners.Although there may be acoustic cues in these singingsamples that elicit listener preferences, providing alink between these and a visual representation ofsuperior quality in LTAS presents a complex chal-lenge for the field. Although more consistency seemsto exist between LTAS and energy, we still do notknow how this affects quality as available studies didnot include a perceptual component.10,13,14,20 Also,SPR and ER may be uncorrelated to the perceived“goodness” of vocal quality because there are at

TABLE 3. Ranked ER for Each Singer, in EachMusical Task (Mozart and Schubert) in Each

Condition (O and SO)

Rank Singer Task Condition ER

1 6 Mozart SO 9.412 3 Mozart O 11.103 3 Mozart SO 11.534 5 Mozart O 14.375 5 Schubert SO 14.606 2 Mozart O 14.937 5 Mozart SO 15.108 6 Mozart O 15.269 5 Schubert O 15.4510 1 Mozart O 15.7211 1 Mozart SO 15.8212 1 Schubert O 15.9513 4 Mozart O 16.1314 6 Schubert O 16.1515 2 Mozart SO 16.1816 1 Schubert SO 16.9017 4 Mozart SO 17.1018 6 Schubert SO 18.9619 3 Schubert O 19.5120 4 Schubert O 20.0121 3 Schubert SO 20.2722 4 Schubert SO 20.6823 2 Schubert O 22.3324 2 Schubert SO 29.32

least three ways in which a singer can increase theenergy in the higher harmonics, which thus affectboth SPR and ER in a similar direction: (1) produc-tion of the formant (good quality), (2) employingpressed phonation (bad vocal quality, at least froma classical perspective), and (3) deliberate additionof twang to boost carrying power.

Acoustic measures such as SPR and ER per-formed on LTAS were never intended to providevocal quality descriptors,22 but when the lowestranked singer on the perceptual ratings achieved thehighest rankings in SPR and ER as in this study,these findings indicate that such measures on LTAScannot define vocal quality as perceived by ear. TheLTAS may contain subtle information on classicalsinging quality, such as differences between openthroat technique and a SO open throat condition. TheLTAS of O and SO conditions produced in this studywere both different than LTAS produced in otherstudies examining speech quality,11,18 pop singing,16


124

ARTICLE IN PRESS


TABLE 4. Singer, Task, Condition, and Rankings forPerceptual Score, SPR, and ER, Sorted by Perceptual

Score Ranking from Highest to Lowest

Perceptual SPR ERRank Rank Rank Singer Task Condition

1 16 9 5 Schubert Optimal2 13 3 5 Mozart Optimal3 10 10 1 Mozart Optimal4 5 2 3 Mozart Optimal5 9 11 1 Schubert Optimal6 21 20 4 Schubert Optimal7 6 13 4 Mozart Optimal8 23 23 2 Schubert Optimal9 19 19 3 Schubert Optimal10 8 6 2 Mozart Optimal11 11 14 6 Schubert Optimal12 4 4 5 Schubert Suboptimal13 3 12 1 Mozart Suboptimal14 24 24 2 Schubert Suboptimal15 15 8 6 Mozart Optimal16 7 7 5 Mozart Suboptimal17 20 21 3 Schubert Suboptimal18 18 18 6 Schubert Suboptimal19 12 16 1 Schubert Suboptimal20 1 5 3 Mozart Suboptimal21 22 22 4 Schubert Suboptimal22 17 15 2 Mozart Suboptimal23 14 17 4 Mozart Suboptimal24 2 1 6 Mozart Suboptimal

or country singing.15 We therefore argue that energydistribution above 2 kHz may indicate a fundamentalquality of classical and operatic singing but is notnecessarily an indicator of overall vocal quality.Measures of SPR and ER may concur more closelywith perceptual studies of vocal projection, carryingpower, or potential amplification over an orchestra.

Exemplars of vocal quality rated by listenersWe compared exemplars of LTAS sampled from

high-, mid-, and low-perceptual ranks and presentedthese with representative qualitative descriptors.LTAS of the highest-ranking singers showed anincrease in energy between 2 and 4 kHz, whereasLTAS of mid- and low-scoring singers lacked theunified peak of energy increase above 2 kHz. Previ-ous similar findings9,13,14,18,41 have interpreted thesevisual cues in LTAS as representative of classicalvocal quality. However, as the third ranking singers1 and 3 in both the Mozart and the Schubert O


TABLE 5. Pearson Correlations Among PerceptualRating, SPR, and ER of the 24 Samples

Rating SPR

0.101SPR P � 0.638

0.064 0.960ER P � 0.767 P � 0.000

produced this unified energy peak �2 kHz (the most“masculine” LTAS shape), they should have hadthe “best” voices perceptually. However, their LTASwere different than those of the highest-rankingsinger (singer 5), who had a wider distribution ofenergy above 2 kHz. The low SO plots (Figure 5)showed a spectral rolloff more consistent with plotsof speech than with singing,11,42 but with reducedenergy from 0 to 2 kHz. Therefore, measurementsof comparative energy between 0–2 and 2–4 kHz,such as SPR or ER, were inconclusive indicators ofvocal quality. High-scoring singers were praised fortheir overall vocal quality, or combinations of posi-tive qualities that resulted in a “balanced” sound.4

Judges recognized an overall good quality in themid-scoring voices, but they identified technicalflaws that impacted on vocal production. In gen-eral, judges were more likely to comment on faultyvocal production or technique, particularly in thoseranked lowest. Negative comments across all singerswere more descriptive and focused on technical sug-gestions for improvement in the sound. Conversely,those singers ranked highly attracted little comment.It seems that a beautiful voice is difficult to describe.

LTAS, and measures applied to LTAS, are notsensitive to quality differences that were identifiedby other analyses, such as vibrato. In previousstudies, these voices were examined for vibrato andspectral differences in O and SO,7,33 but although thedifferences were significant for vibrato parameters,these differences were not confirmed by spectralanalysis, which suggests that current measures thatassess voice, such as LTAS, are not sufficientlysophisticated to represent voice. Spectrographicwork in other fields43–47 has not conclusively identi-fied the same voice in different recordings despitesimultaneous perceptual and visual cues. If this isthe case, it may not be possible to convincingly

125

ARTICLE IN PRESS


TABLE 6. The Spearman Rho for Rankings ofSamples and Singers for Perceptual Rating,

SPR, and ER

Samples (n � 24) Rating SPR

0.308SPR P � 0.143

�0.006 0.169ER P � 0.977 P � 0.431

Singers (n � 6)0.214

SPR P � 0.315�0.314 0.014

ER P � 0.135 P � 0.947

illustrate good singing by visual representation alone.Indeed, with any acoustic analysis, isolation maylead to incorrect interpretation of acoustic cues inthe singing voice. The danger of such a methodis the creation of an ideal, but nonexistent, voiceof quality.

In measuring vocal quality in singing, the inter-pretation of LTAS is as subjective as is the perceptualassessment of expert listeners. Our listeners wereconsistent in their perceptual judgments and reli-able in their assessment of vocal quality. Althoughspectral assessments, such as LTAS, allow for acomparable holistic visual assessment of the voice,future voice research must assign a prominentplace to perceptual rating and avoid confusing spec-tral energy distributions with vocal quality.

Ideally, LTAS should provide an acoustic resourceto assess the spectra of a performance,10 rather thanof a single note.12,48,49 An LTAS plot, like an ex-tended listening sample, provides an overall impres-sion of the most regular features of the sound ofa singer. Given the advantages of LTAS, in that itstabilizes to a regular pattern over time10,33 whileretaining the individuality generated by its musicalstimuli and the individual singer, one would expectthe final LTAS curves to show a positive relationshipwith perceptual ratings. This result was not con-firmed in this study.

Current acoustic information available to singingstudents and singing pedagogues has limited appli-cation in the singing studio. Vocal quality is basedon multiple factors, which are unlikely to be inde-pendent. The next generation of acoustic analyses

must complement the human ear in integrating thecomplex dimensions of the human voice. Futureefforts could well be directed toward establishingperceptual benchmarks of vocal quality and theirassociated acoustical parameters. To remain viable,acoustic research in voice must attempt to emu-late the human ear.

Acknowledgments: We thank Ms. Maree Ryan for herpedagogical advice and support. We also thank Mr.Peter Thomas and Dr. Densil Cabrera for their advice onacoustical matters and the six singers for their willingparticipation in the project. We are grateful for the helpfulcomments of the reviewers in the revision of this paper.

REFERENCES

1. Wapnick J, Ekholm E. Expert consensus in solo voice per-formance evaluation. J Voice. 1997;11:429–436.

2. Ekholm E, Papagiannis GC, Chagnon FP. Relating objectivemeasurements to expert evaluation of voice quality in West-ern classical singing: critical perceptual parameters. J Voice.1998;12:182–196.

3. Robison CW, Bounous B, Bailey R. Vocal beauty: a studyproposing its acoustical definition and relevant causes inclassical baritones and female belt singers. J Singing.1994;51:19–30.

4. Mitchell HF, Kenny DT, Ryan M, Davis PJ. Defining openthroat through content analysis of experts’ pedagogicalpractices. Logoped Phoniatr Vocol. 2003;28:167–180.

5. Merritt L, Richards A, Davis PJ. Performance anxiety: lossof the spoken edge. J Voice. 2001;15:257–269.

6. Thompson WF, Diamond PCT, Balkwill L-L. The adjudica-tion of six performances of a Chopin etude: a study ofexpert knowledge. Psychol Music. 1998;26:154–174.

7. Mitchell HF, Kenny DT. The impact of “open throat” tech-nique on vibrato rate, extent and onset in classical singing.Logoped Phoniatr Vocol. 2004;29:171–182.

8. Mitchell HF, Kenny DT. Can experts identify “open throat”-technique as a perceptual phenomenon? Musicae Sci-entiae. In press.

9. Vurma A, Ross Y. Priorities in voice training: carryingpower or tone quality. Musicae Scientiae. 2000;4:75–93.

10. Jansson EV, Sundberg J. Long-time average spectra appliedto analysis of music. Part I: method and general applica-tions. Acustica. 1975;34:15–19.

11. Lofqvist A, Mandersson B. Long-time average spectrumof speech and voice analysis. Folia Phoniatr Logoped.1987;39:221–229.

12. Omori K, Kacker A, Carroll LM, Riley WD, BlaugrundSM. Singing power ratio: quantitative evaluation of singingvoice quality. J Voice. 1996;10:228–235.


126

ARTICLE IN PRESS


13. Thorpe C, Cala S, Chapman J, Davis P. Patterns of breathsupport in projection of the singing voice. J Voice.2001;15:86–104.

14. Sundberg J. Articulatory interpretation of the “singing for-mant.” J Acoust Soc Am. 1974;55:838–844.

15. Cleveland TF, Sundberg J, Stone RE. Long-term averagespectrum characteristics of country singers during speakingand singing. J Voice. 2001;15:54–60.

16. Borch DZ, Sundberg J. Spectral distribution of solo voiceand accompaniment in pop music. Logoped Phoniatr Vocol.2002;27:37–41.

17. Mendoza E, Valencia N, Munoz J, Trujillo H. Differencein voice quality between men and women: use of the long-term average spectrum (LTAS). J Voice. 1996;10:59–66.

18. Barrichelo VMO, Heuer RJ, Dean CM, Sataloff RT. Com-parison of singer’s formant, speaker’s ring and LTA spec-trum among classical singers and untrained normalspeakers. J Voice. 2001;15:344–350.

19. Rossing TD, Sundberg J, Ternstrom S. Acoustic comparisonof soprano solo and choir singing. J Acoust Soc Am. 1987;82:830–836.

20. Barnes JJ, Davis PJ, Oates J, Chapman J. The relationshipbetween professional operatic soprano voice and high rangespectral energy. J Acoust Soc Am. 2004;116:530–538.

21. Bunch M, Chapman J. Taxonomy of singers used as subjectsin scientific research. J Voice. 2000;14:363–369.

22. Bartholomew W. A physical definition of “good voice qual-ity” in the male voice. J Acoust Soc Am. 1934;6:25–33.

23. Howard DM, Szymanski J, Welch GF. Listeners’ perceptionof English girl and boy choristers. Music Percept. 2002;20:35–49.

24. White P. Long-term average spectrum (LTAS) analysisof sex and gender-related differences in children’s voices.Logoped Phoniatr Vocol. 2001;26:97–101.

25. White P. A study of the effects of vocal intensity variation onchildren’s voices using long-term average spectrum (LTAS)analysis. Logoped Phoniatr Vocol. 1998;23:111–120.

26. Foulds-Elliott S, Thorpe C, Cala S, Davis P. Respiratoryfunction in operatic singing: effects of emotional connec-tion. Logoped Phoniatr Vocol. 2000;25:151–168.

27. Behrens G, Green S. The ability to identify emotional con-tent of solo improvisations performed vocally and on threedifferent instruments. Psychol Music. 1993;21:20–30.

28. Howes P, Callaghan J, Davis PJ, Kenny DT, Thorpe C. Therelationship between measured vibrato characteristics andperception in Western operatic singing. J Voice. 2004;18:216–230.

29. Siegwart H, Scherer K. Acoustic concomitants of emotionalexpression in operatic singing: the case of Lucia in Ardigli incensi. J Voice. 1995;9:249–260.

30. Sundberg J, Iwarsson J, Hagegard H. A singer’s expressionof emotions in sung performance. In: Fujimura O, Hirano


M, editors. Vocal Fold Physiology: Voice Quality Con-trol. San Diego, CA: Singular Publishing Group; 1995:217–231.

31. Kamenetsky S, Hill D, Trehub S. Effect of tempo anddynamics on the perception of emotion in music. PsycholMusic. 1997;25:149–160.

32. Gabrielsson A, Juslin PN. Emotional expression in musicperformance: between the performer’s intention and thelistener’s experience. Psychol Music. 1996;24:68–91.

33. Mitchell HF, Kenny DT. The effects of open throat tech-nique on long term average spectra (LTAS) of female classi-cal voices. Logoped Phoniatr Vocol. 2004;29:99–118.

34. Titze I, Bergan CC, Hunter EJ, Story B. Source and filteradjustments affecting the perception of the vocal qualitiestwang and yawn. Logoped Phoniatr Vocol. 2003;28:147–155.

35. Estill J. Primer of Basic Figures. 2nd ed. Santa Rosa, CA:Estill Voice Training Systems; 1996.

36. Cabrera D, Davis P, Barnes J, Jacobs M, Bell D. Recordingthe operatic voice for acoustic analysis. Acoust Australia.2002;30:103–108.

37. Erickson ML, Perry SR. Can listeners hear who is singing?A comparison of three-note and six-note discriminationtasks. J Voice. 2003;17:353–369.

38. White P. Formant frequency analysis of children’s spokenand sung vowels using sweeping fundamental frequencyproduction. J Voice. 1999;13:570–582.

39. Nordenberg M, Sundberg J. Effect on LTAS of vocal loud-ness variation. TMH-QPSR. 2003;45:93–100.

40. Lundy DS, Roy S, Casiano RR, Xue JW, Evans J. Acousticanalysis of the singing and speaking voice in singing stu-dents. J Voice. 2000;14:490–493.

41. Sundberg J. Level and center frequency of the singer’sformant. J Voice. 2001;15:176–186.

42. Novak A, Vokral J. Acoustic parameters for the evaluationof voice of future voice professionals. Folia PhoniatrLogoped. 1995;47:279–285.

43. Koenig BE. Spectrographic voice identification: a forensicsurvey. J Acoust Soc Am. 1986;79:2088–2090.

44. Hakes J, Shipp T, Doherty ET. Acoustic properties ofstraight tone, vibrato, trill and trillo. J Voice. 1987;1:148–156.

45. Hoit JD, Jenks CL, Watson PJ, Cleveland TF. Respiratoryfunction during speaking and singing in professional coun-try singers. J Voice. 1996;10:39–49.

46. Hollien HF. The Acoustics of Crime: The New Science ofForensic Phonetics. New York: Plenum Press; 1990.

47. Rose P, editor. Forensic Speaker Identification. New York:Taylor & Francis; 2002.

48. Cleveland TF. Acoustic properties of voice timbre typesand their influence on voice classification. J Acoust SocAm. 1977;61:1622–1629.

49. Sundberg J. The Science of the Singing Voice. DeKalb, IL:Northern Illinois University Press; 1988.

127

ICMPC8 CONFERENCE PROCEEDINGS

Kenny, D.T., & Mitchell, H.F., Visual and auditory perception of vocal beauty: conflict or concurrence? In: Lipscomb S, Ashley R,

Gjerdingen R, Webster P, editors. 8th International Conference on Music Perception & Cognition (ICMPC8); August 2004; Evanston, IL

USA: Causal Productions; 2004. p. 171-174.

128

1

VISUAL AND AUDITORY PERCEPTION OF VOCAL BEAUTY: CONFLICT OR CONCURRENCE?

Dianna T. Kenny & Helen F. Mitchell

Australian Centre for Applied Research in Music Performance (ACARMP) Conservatorium of Music, University of Sydney

ABSTRACT

In this study, we compared perceptual ratings of vocal beauty with acoustic measures performed on LTAS. Fifteen expert judges rated 24 samples with six repeats of six advanced singing students under two conditions: ‘optimal’ (O), representing maximal open throat and ‘suboptimal’ (SO), representing reduced open throat technique in singing. LTAS were performed on each singing sample and conventional measures of energy in singing, using peak height and peak area were calculated on each LTAS. Perceptual scores and acoustic measures of ratios of peak height and peak area of LTAS were rank ordered. While there was a significant relationship between LTAS measures, there was no relationship between perceptual ratings of vocal samples or singers based on LTAS measures. These findings suggest that since LTAS measures are not consistent with perceptual ratings of vocal beauty, they cannot be used to define a beautiful voice.

1. INTRODUCTION What features of the singing voice define vocal beauty? Can these features be measured accurately and more importantly, specifically taught? Since the 1980s technology to measure vocal acoustics has become more sophisticated and has been increasingly used in experimental research, with the findings of such research now being incorporated into texts on singing [e.g. 1-3]. While there is a proliferation of acoustic studies of the singing voice it is difficult to draw firm conclusions about the implications for singing pedagogy of many of these scientific studies, which typically use small numbers of subjects, rarely link the vocal strategies used to perceptual results, often use student singers rather than professionals, and fail to identify the teaching/learning approach of the subjects.

There is now both a vast literature on vocal acoustic properties and many unanswered questions about what conclusions can be drawn from such analyses. There are many ways that the voice can be represented acoustically, but the most common is the measurement of energy production. Long-term average spectra (LTAS) are used to represent singers’ sound [4, 5] and its different vocal qualities based on energy changes that occur during different vocal tasks. A conventional way of reducing the information in the LTAS to a single meaningful number is to compute the ratio of energies in a low and a high frequency band. Often 2 kHz is chosen as the delimiting frequency [6, 7]. Do these ratios determine a beautiful voice? This question has come into sharp focus since the advent of acoustic analysis. The utility of experimental findings for the teaching of singing depends on their incorporation into a theory of the singing voice and on their relation to current practice.

However, most current research relies on acoustic findings generated by averaging data from groups of voices in order to characterize a good voice [6, 8]. The dilemma and paradox in this approach is that the averaged findings may not adequately represent any of the individual voices used to generate the desirable acoustic parameters that become the benchmark of vocal quality. Further, vocal quality derives from a combination of acoustical parameters that interact. “The subtle individual properties that set one voice apart from another should not be averaged out” [9] (p. 299).

Studies of voice quality have taken two main forms. In the first type of study, LTAS is used to “validate” perceptual cues [10]. In the second type of study, acoustic findings are “validated” by perceptual ratings [11, 12]. There are two such studies that provide some basis for this project. One, by Wapnick and Ekholm [13] established 12 generally accepted perceptual criteria applied to the assessment of voice quality in classical singing, and the other, by Ekholm, Papagiannis and Chagnon [11] applied four of these criteria (“appropriate vibrato”, “resonance/ring”, “color/warmth”, and “clarity/focus”) and related them to objective measurements taken from acoustic analysis of the voice signal.

In this study, we investigated pedagogues’ rating of female classical singers using maximum open throat [14] and reduced open throat in their singing of a classical song and a romantic lied [15, 16]. We matched perceptual ratings to acoustic measures of each perceptual sample to determine whether acoustic analysis matched perceptual judgments of overall timbre. The aim of the study was to assess the degree of agreement between acoustic analysis and perceptual ratings of vocal beauty.

2. METHOD

Participants Listeners were 15 experienced singing pedagogues, 12 females and 3 males aged between 37 and 76 with a mean of 54 years. Six participants had a postgraduate qualification in singing, five had a diploma of music or singing, and three a bachelor degree in music. One cited extensive international performing experience as their qualification in singing. All had taught singing for between 4 and 40 years, with an average of 20 years. Thirteen of fifteen participants taught singers in a Conservatorium of music. Participating pedagogues taught, on average, 39.5% operatic students and 36.7% classical students. For 11 pedagogues, the majority of their studio comprised these two genres. Six pedagogues also taught ≥20% of musical theatre students. Eleven taught a proportion of international and national level singers, nine at big city or regional/touring and eight at local community level.

129

2

Subjects and Stimuli Six advanced singing students, studying opera at a Conservatorium of Music sang two song excerpts, Mozart song Ridente la Calma and Schubert lied Du bit die Ruh under two conditions: optimal, and sub-optimal. ‘Optimal’ (O) condition was necessary to provide a sound quality with the best technique that they could. This involved the maximal use of open throat. ‘Sub-optimal’ (SO) condition involved the use of a reduced (open throat) technique but still with an acceptable singing technique and without consciously altering any other aspect of their technique.

The acoustic signal was recorded digitally (Behringer Ultragain preamplifier/Marantz CDR 630) via a high-quality microphone (AKG C-477) positioned on a head boom a constant 7 cm distance from the singer’s lips. Pink noise was played from a speaker (Bose Lifestyle) and the microphone of a sound level meter (Rion NL-06 SPL) was coupled to the AKG microphone for calibration of absolute sound pressure levels (SPL).

The audio recordings of the singers were digitally extracted using Audiograbber software (www.audiograbber.com) on a standard PC computer to wav audio format in stereo at 41000 Hz and 16 bit sample rate. The files were then reopened in Cool Edit Pro 1.2. The first 7 bars of Mozart and bars 54 to 60 of the Schubert musical tasks were used for the perceptual study. Each consisted of an entire musical phrase lasting around 20 seconds.

Procedure The perceptual test was conducted in a quiet environment and samples were played from a Sony CD Walkman (DEJ885W) via circum-aural closed-back stereo monitoring headphones (Sennheiser HD 270). Prior to presentation of stimuli, participants were given information on the two singing conditions, O and SO, and were presented with the musical score of each musical task. They were asked to judge whether or not the singer was using open throat technique in each sample. They were asked to rate each sample on a 10-point scale. Judges who assessed singers were not the singers’ pedagogues and did not know any of their identities.

Listeners assessed 30 tracks [6 singers x 2 conditions, (O and SO) x 2 musical tasks] in random order, including six repeats, also selected at random.

Acoustic Analysis Long term average spectra (LTAS) analyses (filter 300 Hz) of the 24 song samples demonstrated the differences in spectral energy for the singers, conditions and musical tasks. LTAS contained only voiced data [17].

From the LTAS plots, the highest peaks in the 0-2 kHz and 2-4 kHz regions were labeled P1 and P2 respectively. Two measures were used to quantify these LTAS data. The singing power ratio (SPR), described by Omori et al. [13] compares the maximum energy by peak levels of P1 and P2. The SPR is the difference between the level of P1 and P2 in dB and is a measure of spectral slope between 0-2 kHz and 2-4 kHz. When the energy is focused in the 0-2kHz region,

it results in higher SPR results and there is typically less energy reinforcement in the 2-4 kHz region. Although SPR does not prove the presence of a singer’s formant, it enables inter- and intra-singer comparison of spectral energy in the voice [18].

The energy ratio (ER), described by Thorpe et al. [12] compares overall energy between 0-2 kHz and 2-4 kHz. It is calculated by taking the difference between average energy values for the two frequency ranges. A low ER represents a greater reinforcement in the 2-4 kHz region and better balance between the spectral energy 0-2 and 2-4 kHz. Areas below the LTAS curve were calculated in the same way as the calibration tone. When there is less reinforcement in the 2-4 kHz region, ER results follow SPR.

Study Design Perceptual scores, SPR and ER were rank ordered. We then compared perceptual rankings with rankings of acoustic measures (SPR and ER) to assess whether the acoustic characteristics matched the perceptual judgments of overall timbre.

Reliability of the judges’ ratings was tested with the scores of the six repeats against the original scores. Intraclass Correlation Coefficients were calculated with a two-way random effects model (absolute agreement).

The study design was a repeated measures [24 ratings per listener] randomized complete block with a 2 [Task (Mozart vs Schubert)] x 2 [condition (Optimal and Sub-optimal)] factorial structure. Main effects for task and condition and interaction effects (contrasts) were calculated for each dependent measure.

3. RESULTS Intraclass correlations with two-way random effects model of absolute agreement [ICC(2,1)] compared the exact scores of the repeated samples with the original ratings. For the O condition, ICC = .7843 [CI 95% .4822 - .9613] F = 9.52, p < .0001. For SO, ICC = .7579 [CI 95% .4338 - .9558] F = 8.60, p < .0001, that is, listeners matched their exact rating in the first rendition and the repeat in nearly 5 of 6 cases. Therefore, judges were reliable in their absolute ratings.

Overall means of score by condition indicated a significant reduction in SO compared to O in both musical tasks. For mean score, there was a main effect for condition (F(1, 5) = 111.872, p = .000), but not for musical task (F(1, 5) = 1.497, p=.276); that is, ratings were significantly higher for O than for SO in both the Mozart and Schubert tasks. There was no interaction between musical task and condition (F(1, 5) = .355, p = .577).

Table 1 presents the rankings of each singer and the respective musical task and experimental condition, ordered by perceptual rank from highest to lowest.

130

3

Singer 5 was rated highest overall, in both Mozart and Schubert tasks (7.60, 7.07), and singers 1 and 3 were ranked third with a score of 6.93. The singers’ O condition scores ranged from 5.40 to 7.60. For SO, singer 5 was rated highest for Schubert. Singers’ SO scores ranged from 2.67 to 5.40. Singers’ O scores consistently scored higher rating than SO.

Pearson’s correlation coefficients were calculated for each of the three dependent measures (Perceptual rating, SPR and ER). While there was a highly significant relationship between ER and SPR, as expected (r=.960* p=.000) there was no relationship between either of these acoustic measures and perceptual judgments (ER, r= 0.064, p=.767; SPR, r=0.101p=.638).

Spearman’s rho correlation coefficients were calculated for judges’ rankings of samples and singers on each of the dependent measures. Results indicated that there was no relationship between perceptual ratings of samples or singers based on ranked ER (r=-0.006, p=.977) and SPR (r=0.308, p=.143).

Figure 1a and b present LTAS exemplars sampled from the highest ranked singer 5, and three singers ranked 3= singing in O. Figure 1c presents the middle ranked singers, singing in both O and SO and finally 1d illustrates the lowest perceptually ranked singers, all in SO.

20

40

60

80

100

120

0 1000 2000 3000 4000 5000 6000

Hz

dB

S5M5

(a)

20

40

60

80

100

120

0 1000 2000 3000 4000 5000 6000

Hz

dB

M1M3S1

(b)

20

40

60

80

100

120

0 1000 2000 3000 4000 5000 6000

Hz

dB

S6S5M1S2

(c)

20

40

60

80

100

120

0 1000 2000 3000 4000 5000 6000

Hz

dB

M2M4M6

(d)

Figure 1: LTAS of highest (a), 3= (b), middle (c) and lowest (d) perceptually ranked singers. Legends correspond to task, Mozart (M) or Schubert (S), and singer number.

Perceptual Rank

SPR Rank

ER Rank Singer Task Condition

1 16 9 5 S O 2 13 3 5 M O 3 10 10 1 M O 4 5 2 3 M O 5 9 11 1 S O 6 21 20 4 S O 7 6 13 4 M O 8 23 23 2 S O 9 19 19 3 S O

10 8 6 2 M O 11 11 14 6 S O 12 4 4 5 S SO 13 3 12 1 M SO 14 24 24 2 S SO 15 15 8 6 M O 16 7 7 5 M SO 17 20 21 3 S SO 18 18 18 6 S SO 19 12 16 1 S SO 20 1 5 3 M SO 21 22 22 4 S SO 22 17 15 2 M SO 23 14 17 4 M SO 24 2 1 6 M SO

Table 1: Singer, task, condition and rankings for perceptual score, singing power ratio (SPR) and energy ratio (ER), sorted by perceptual score ranking from highest to lowest.

131

4

4. DISCUSSION These findings suggest that since LTAS rankings were not consistent with perceptual ratings of vocal beauty, they cannot be used to define a beautiful voice. Perceptual ratings were significantly higher for O than for SO in both the Mozart and Schubert tasks. There was no interaction between musical task and condition. While there was a significant relationship between SPR and ER, there was no relationship between perceptual ratings of vocal samples or singers based on SPR or ER. Intraclass correlations with two-way random effects model of absolute agreement (ICC(2,1)) indicated that judges were reliable in their ratings.

Characteristic LTAS shapes, particularly of classical or operatic voices have been accepted as an accurate visual representation of good sound in the genre and operatic singing is positively associated with an increase in energy between 2-4 kHz [6, 7, 12]. While measures of SPR [7] and ER [6] are not intended to give information about singer’s formant, they should illustrate differences in energy distribution, or carrying power. However, Vurma & Ross [12] found that increased energy above 2 kHz did not necessarily represent vocal quality judged perceptually. In this study, measurements of comparative energy were inconclusive indicators of vocal quality.

We compared exemplars of LTAS sampled from high, mid and low perceptual ranks. LTAS of the highest-ranking singers showed an increase in energy between 2-4 kHz, whereas LTAS of mid and low scoring singers lacked the unified peak of energy increase above 2 kHz. Previous similar findings [5, 9, 12] have interpreted these visual cues in singers’ LTAS as a good sound. However, as third ranking singers 1 and 3 in O produced this unified energy peak > 2 kHz (the most ‘masculine’ LTAS shape) they should have had the ‘best’ voices perceptually. However, their LTAS were different to the highest-ranking singer (singer 5), who had a wider distribution of energy above 2 kHz. The low SO plots (Fig 1d) showed a spectral roll-off more consistent with plots of speech rather than singing.

Linking perceptual rankings with corresponding LTAS plots expands upon previous twinned perceptual and acoustic studies [7, 8]. We argue that previous templates such as shapes of LTAS curves are no longer acceptable as indicators of vocal quality. They have been produced from averages generated from groups of singers. These averages may, in fact, not represent any voice and may be an artefact of clustered measurement. A science of the singing voice cannot progress without addressing the problem inherent in accepting long-term average spectra as analogues for vocal quality without providing a link between perceptual cues and a visual representation of ‘goodness’ in singing. We have demonstrated that conventional acoustic measures of voice quality do not correspond to perceptual measures of vocal beauty. This represents a major challenge to current wisdom that acoustic parameters of voice must emulate the established acoustic norms in order to achieve overall vocal excellence.

5. REFERENCES 1. Miller, R., The structure of singing: system and art in vocal

technique. 1996, New York: London: Schirmer Books.

2. Sundberg, J., The Science of the Singing Voice. 1988, DesKalb, Illinois: Northern Illinois University Press.

3. Thurman, L. and Welch, G.F., eds. Bodymind & voice: foundations of voice education. 2000, VoiceCare Network: Collegevile, Minn.

4. Sundberg, J. (1974). Articulatory interpretation of the "singing formant". J Acoust Soc Am, 55(4), 838-844.

5. Borch, D. Z., & Sundberg, J. (2002). Spectral distribution of solo voice and accompaniment in pop music. Log Phon Vocol, 27, 37-41

6. Thorpe, C., Cala, S., Chapman, J., & Davis, P. Patterns of breath support in projection of the singing voice. J Voice, 2001. 15(1), p. 86-104.

7. Omori, K., Kacker, A., Carroll, L.M., Riley, W.D., and Blaugrund, S.M., Singing power ratio: quantitative evaluation of singing voice quality. J Voice, 1996. 10(3): p. 228-35.

8. Sundberg, J. Level and center frequency of the singer's formant. J Voice, 2001. 15(2), p. 176-186.

9. Miller, R., The Singing Teacher in the Age of Voice Science, in Vocal Health and Pedagogy, R.T. Sataloff, Editor. 1998, Singular: San Diego, CA. p. 297-300.

10. Howard, D.M., Szymanski, J., and Welch, G.F., Listeners' Perception of English Cathedral Girl and Boy Choristers. Music Perception, 2002. 20(1): p. 35-49.

11. Ekholm, E., Papagiannis, G.C., and Chagnon, F.P., Relating objective measurements to expert evaluation of voice quality in Western classical singing: critical perceptual parameters. J Voice, 1998. 12(2): p. 182-96.

12. Vurma, A., & Ross, Y. (2000). Priorities in voice training: carrying power or tone quality. Musicae Scientiae, 4(1), 75-93.

13. Wapnick, J. and Ekholm, E., Expert consensus in solo voice performance evaluation. J Voice, 1997. 11(4): p. 429-36.

14. Mitchell, H.F., Kenny, D.T., Ryan, M., and Davis, P.J., Defining open throat through content analysis of experts' pedagogical practices. Log Phon Vocol, 2003. 28: p. 167-180.

15. Mitchell, H.F. and Kenny, D.T., The impact of "open throat" technique on vibrato rate, extent and onset in classical singing. Log Phon Vocol, in press.

16. Mitchell, H.F. and Kenny, D.T., The effects of open throat technique on long term average spectra (LTAS) of female classical voice. Log Phon Vocol, in press.

17. White P. A study of the effects of vocal intensity variation on children's voices using long-term average spectrum (LTAS) analysis. Logoped Phoniatr Vocol 1998; 23:111-120.

18. Barnes JJ, Davis PJ, Oates J, Chapman J. The relationship between professional operatic soprano voice and high range spectral energy. J Acoust Soc Am 2004; in press.

132

6.1 Conclusion

We are a society of vocal critics. We recognise a bad voice, a good voice and a truly

great voice when we hear them. Yet scientific characterisations and descriptions of

beautiful voices have to date been inadequate. There is now a vast literature on vocal

acoustic properties and the visual representation of voice as well as many unanswered

questions about what conclusions can be drawn from such analyses. This systematic

body of research has assessed the degree to which acoustic measures can inform our

perception of vocal quality and the development of quality through vocal pedagogy.

In addition to the above-mentioned published papers, and on the basis of the issues

outlined above, it follows naturally that what is needed are new methods of assessing

vocal quality. Few studies employ singers of the highest performing standards or

include a perceptual component by expert listeners in combination with spectral

analysis. The papers above highlight the problems in current singing research and the

scientific literature and find that current methods used to analyse singing are insufficient

to capture and define voices of quality.

133

7 Overall conclusions and future directions

In this section, the main findings of each study are presented along with implications for

future research. Limitations and applications of this research to such future projects are

also discussed.

7.1 Review of the findings

This body of work has defined and examined a major pedagogical technique called

‘open throat’ and assessed its acoustic and aesthetic correlates. This research

demonstrated that open throat does have unique acoustic indicators and vibrato

differences were apparent between those singers using open throat compared to those

who were not. As inconsistent vibrato is considered indicative of poor singing, it was

hypothesized that testing the energy distribution in singers’ voices in each condition

would identify the timbral changes associated with open throat. Visual inspection of

LTAS confirmed differences between open throat and sub optimal open throat (that is,

reduced open throat), but conventional measurements of singing power ratio (SPR) (the

difference between the peak height) and energy ratio (ER) (the difference between peak

area), were not statistically significant. These results were not consistent with the

vibrato findings and suggested that conventional measures of SPR and ER are not

sufficiently sensitive to evaluate LTAS. In addition, there was no relationship of any

kind between perceptual rankings of vocal beauty and acoustic rankings of vocal

quality.

Current literature has used the shape of the long term average spectra (LTAS) as the

‘objective’ measure of quality but this research has found that, although acoustic

134

analysis LTAS and other indicators of vocal quality such as singing power ratio (SPR)

and energy ratio (ER) have contributed to our understanding of vocal production, it is

clear that these methods are somewhat limited in their capacity to define vocal quality,

since we have found no relationship between these standard acoustic measures and

perceptual ratings of vocal quality by expert listeners.

These studies have assessed the degree to which acoustics can inform our perception of

vocal quality and its development through vocal pedagogy. In singing the human ear is

the more perfect than any acoustic measure in evaluating vocal quality, it remains the

most effective method of assessing voice. Pedagogues process the whole sound,

interpret and assess technical features from the overall quality. Our understanding of

what produces changes in vocal quality may need to be extended to include not only the

demands of genre and emotional intention, but also its deliberate cultivation through

instruction in particular pedagogical techniques such as open throat. These studies pose

a number of unanswered questions which will be addressed in future research.

7.2 Originality of the research

This is the first body of work to document a technique used throughout historical and

current singing pedagogy and assess its value in classical singing training. In addition, it

tested conventional acoustic analyses in defining overall vocal quality, and represents a

major challenge to current wisdom that acoustic parameters of voice must emulate the

established acoustic norms in order to achieve overall vocal excellence.

Each project design challenged current research methodologies used in singing research

and reported in the singing literature. Rigorous and novel acoustic and perceptual

135

methodologies developed for these studies will form the basis for future research.

Methodologies used in the recording process have ensured consistency across data

collection, and close-miking removed room reflections, based on the most recent

research in recording operatic voices (Cabrera, Davis, Barnes, Jacobs, & Bell, 2002).

Individual papers discuss the innovations in calibration techniques (Mitchell & Kenny,

2004) and their future use of particular software used to analyse voice.

Previous scientific studies have used single subjects or averaged groups of subjects to

generate an overview of a perfect voice. Through this series of studies we have rejected

this methodological style, because it creates a mythical voice that is representative of no

actual voice.

7.3 Limitations of the research

It is believed that the homogenous group of female singers used in this study was a

methodological strength. However, we recommend replication of these results using

other singer participants to confirm the data presented above. Future studies must test

the use of the technique in male voices and also in professional singers to ascertain its

value in the singing studio. Previous studies on singing students have used cross-

sections of singers at various stages in their training (Ekholm et al., 1998; Novak &

Vokral, 1995; Vurma & Ross, 2000), but singers in this study were of similar age and

level of vocal training. The numbers in the singer subject pool were comparable to

many other LTAS and singing studies (Barnes et al., 2004; Cleveland et al., 2001;

Thorpe et al., 2001). Using two experimental musical tasks increased the information,

but statistical trends observed between tasks made it important to consider each task

individually.

136

Physiological verification of open throat was beyond the scope of this current project.

Pedagogical interpretations of the instructions to achieve the technique were considered

within the current literature (Mitchell et al., 2003) and we speculated on the possible

anatomical configurations that could account for the action and the sound quality.

Whether or not it is possible to isolate the action of open throat from other components

of good singing technique is still unclear and there may be more than one physiological

action involved to achieve the technique. Confirmation of the use of open throat was

made solely on expert perceptual identification.

Technology used in these studies has been superseded by music industry standard

software. Recording to CD recorder has been replaced by recording to computer hard

disk and studies should demand testing standard microphones that were not available or

affordable when these studies commenced. While Soundswell is innovative software

with multiple channel analysis tracks through its integration with a real-time

phonetogram program (Phog), the voice channel files were not compatible with other

software. Perceptual wave files were initially created from Swell at a bit rate of 16000

Hz, which was a third of CD quality. Therefore perceptual samples were created for

these projects using Cool Edit Pro (now Adobe Audition). There are numerous

commercial and freeware recording and analysis software packages which are more

widely used across different fields of study (for example Protools, PRAAT and Adobe

Audition) and these deserve consideration for use in future studies.

137

The recording techniques used in this singing research provided ideal conditions to

consistently test vocal energy. Using microphones mounted on head booms 7 cm from

the subjects’ lips removed any effect of room reflections. However, this produces

unnatural stimuli for a perceptual recording, unlike singing in a concert hall or theatre

space and akin to singing while standing next to a teacher in the singing studio. While

listeners in this study are familiar with such proximity of their singing students, it may

be that future recording process must absorb this effect, and make multiple recordings

conducive to perceptual evaluations, using for example, reverberation in Protools to

place the recorded sound in a church or a concert hall acoustic setting.

7.4 Future directions

Future research will be conducted with a team of music researchers, audio engineers,

singers and singing pedagogues representing the disciplines of singing pedagogy,

musicology, psychology, acoustics and audio engineering who have the capacity to

revolutionise the conceptual, theoretical and methodological bases for future studies

related to the pedagogy, development and description of singing voices of quality.

7.4.1 Establishing a gold standard

Research studies to date have not used voices judged perceptually as the most beautiful.

Future research must establish exemplars of vocal quality in professional singers by

obtaining vocal samples of professional classical singers, collected under identical

acoustic and musical conditions.

Vocal quality derives from a combination of acoustical parameters that interact. The

simple correlation studies performed in this thesis cannot deal with the non-linear

138

interactions of multiple parameters. Voice quality is likely to be based on multiple

factors, which are not necessarily independent. Models of vocal quality should be

generated from multiple acoustical parameters using non-linear statistics, such as neural

network and genetic algorithm techniques (Nannariello, Osman, & Fricke, 2002;

Zwicker & Fastl, 1999); thereby modelling the more subtle aspects of quality that are

lost in the residual of traditional correlation studies.

Expert pedagogues, speech pathologists and audio engineers will evaluate their singing

acoustically and perceptually.

7.4.2 Recording techniques

The singing research field must develop new precision acoustical measurement

techniques. New methods of measuring vocal sound must be attuned to subjectively

assessed aspects of vocal quality. Such measurement methods will yield a model of

vocal quality that can be used for objective vocal assessment, as well as improving

vocal quality through teaching.

Precision acoustical measurement techniques must include near-field measurements of

the singers (Cabrera, 2003) as well as binaural measurements suitable for subjective

evaluation. Binaural techniques (using a dummy head microphone) virtually place the

listening subject in the same acoustical environment as the original dummy head, with

only minor artefacts. Unlike conventional headphone techniques, binaural reproduction

yields an externalised auditory image with intact distance cues (Nannariello et al.,

2002). With the precise control of audio signal gain, the sound level of the reproduced

139

song can be maintained from the original level, which is of primary importance in

auditory preference studies.

In future studies, we will make use of music industry hardware and software for the

recording process. The acoustic signal will be recorded digitally to hard disc (24-bit

audio, to accommodate the very large dynamic range of singers) using a measurement-

grade microphone positioned on a head boom a constant 7 cm distance from the singer’s

lips (Cabrera et al., 2002). This close microphone tracks singer movements, as well as

rendering room acoustical effects negligible (Cabrera et al., 2002). A second

measurement microphone will be fixed at a distance of 1 m from the singer, and a

dummy head (Neumann KU100) at 2 m.

Prior to recording, the room will be acoustically optimised. The first-order reflections in

the recording room will be suppressed by the addition of sound absorptive material. In

addition, we will measure the transfer function from an artificial mouth at the singer

position to each microphone, to account for room acoustical effects and these will be

factored out in the analysis.

To assist the singer with timing and pitch accuracy and singing in an unnatural

environment, each singer will hear an identical backing track of piano accompaniment.

Closed headphones will be used for vocal and piano foldback so singers can hear both

the piano and themselves in the headphones. We will apply reverberation while

recording so singers have the sense of recording in a real acoustic space eg church or

140

big hall. This reverberation is recorded on discrete channels so that analysis of voice

and if necessary reverberation can be made separately.

7.4.3 Perceptual assessors

A significant methodological innovation in the perceptual studies would be to compare

singing pedagogues with other groups of expert judges, such as speech pathologists and

audio engineers. The latter two groups are highly attuned to minute variations in vocal

timbre, but have hitherto not been used in studies of vocal quality in singers. The use of

commercial audio engineers has significant implications for singers and singing

pedagogues as well as the music industry. The established scientific language used by

this group to describe vocal characteristics has not generalized to other vocal specialists

who have yet to develop a precise descriptive vocabulary.

Expert audio/mastering engineers are able to acoustically define their perceptions of

vocal quality by making voices as beautiful as possible, using a defined set of sound

enhancement computer software in industry standard, Protools. Comparing these

enhanced recordings, acoustically and perceptually, with the original voice sample will

demonstrate what sound qualities these voices are required to achieve for an optimal

sound quality and expertly capture the acoustic measures needed to equate with

perceptual judgments of a voice of quality.

Panels of singing pedagogues will be used to assess and discuss vocal samples in an

adjudication setting. This would enable discussion in the use of subjective language

among judges to reach consensus which would allow researchers to build a descriptive

taxonomy of voice.

141

7.4.4 Integrating science into a systematic pedagogy

The acoustic information currently available to singing students and singing pedagogues

has limited application in the singing studio. Yet, the traditional approach to pedagogy

can be a lengthy and often frustrating process for both teacher and student. Linking

perceptual judgment with acoustic measurement has the potential to provide a basis for

a more systematic pedagogical approach leading to more consistent and predictable

results in less time than current approaches.

Previous studies on vocal quality have used cross-sections of singers at various stages in

their training. Future research should document longitudinally the acoustic and

perceptual changes of vocal quality and skill acquisition in response to particular

pedagogical techniques. This approach has the potential to progress the art and science

of vocal pedagogy beyond sciolism.

It was beyond the scope of the current project to test the development and outcome of

the technique across time. Such an approach represents a major extension on current

work that describes vocal features at one point in time.

142

8 Conclusion

Musical pedagogy has evolved through an oral, non-scientific tradition. Empirical

research does not invent new ways to sing, but can objectively compare singing

traditions and pedagogies by systematically assessing and documenting the acquisition

of vocal mastery in order to identify which elements of technical and aesthetic training

produce the most beautiful voices, thereby contributing to the science of voice and a

more efficient singing pedagogy.

In future singing research, sophisticated technologies and research methods will

enhance acoustic and perceptual measurement and pedagogical practices in the musical

arts, thereby enhancing the musical development and career prospects of the next

generation of young performers and teachers. The methods that we have begun to

develop to assess vocal quality in this series of studies will be applicable to all vocal

styles including pop and music theatre and to instrumental performance. These studies

will hopefully benefit professional musicians, music students and the music industry.

143

9 References

Anderson, E. (Ed.). (1991). The letters of Mozart and his family. London: Macmillan

Press.

Appelman, D. R. (1967). The science of vocal pedagogy. Bloomington, Ind.: Indiana

University Press.

Bacilly, B. d. (1668/1968). Remarques curieuses sur l'art de bien chanter. English A

commentary upon the art of proper singing [by] Benigne de Bacilly. Translated

and edited by Austin B. Caswell. Brooklyn, N.Y.: Institute of Medival Music.

Baken, R. J., & Orlikoff, R. F. (2000). Clinical measurement of speech and voice. San

Diego: Singular Publishing Group.

Barnes, J. J., Davis, P. J., Oates, J., & Chapman, J. (2004). The relationship between

professional operatic soprano voice and high range spectral energy. J Acoust Soc

Am, 116(1), 530–538.

Barrichelo, V. M. O., Heuer, R. J., Dean, C. M., & Sataloff, R. T. (2001). Comparison

of singer's formant, speaker's ring and LTA spectrum among classical singers

and untrained normal speakers. Journal of Voice, 15(3), 344-350.

Bartholomew, W. (1934). A physical definition of "good voice quality" in the male

voice. J Acoust Soc Am, 6, 25-33.

Borch, D. Z., & Sundberg, J. (2002). Spectral distribution of solo voice and

accompaniment in pop music. Logopedics Phoniatrics Vocology, 27, 37-41.

Boyle, J. D. (1992). Evaluation of music ability. In Colwell, R. (Ed.), Handbook of

Research on Music Teaching and Learning (pp. 247-265). New York: Schirmer

Books.

144

Brown, W. S., Jr, Rothman, H. B., Morris, R. J., & Sapienza, C. M. (2001). Acoustic

analyses of trained singers perceptually identified from speaking samples.

Journal of Voice, 15(1), p. 25-35.

Brown, W. S., Jr., Rothman, H. B., & Sapienza, C. M. (2000). Perceptual and acoustic

study of professionally trained versus untrained voices. Journal of Voice, 14(3),

301-309.

Burgin, J. C. (1973). Teaching singing. Metuchen, N.J.: Scarecrow Press.

Cabrera, D. (2003). Pink noise and calibrations. In Mitchell, H. F. (Ed.). Sydney.

Cabrera, D., Davis, P., Barnes, J., Jacobs, M., & Bell, D. (2002). Recording the operatic

voice for acoustic analysis. Acoustics Australia, 30, 103-108.

Caccini, G. (1602/1930). Le nuove musiche (Riproduzione dell'edizione del 1601 / a

cura di Francesco Mantica. ed.). Roma: Monteverdi, Claudio.

Cairns, C. (1969). The memoirs of Hector Berlioz. London.

Callaghan, J. (2000). Singing and voice science. San Diego, California: Singular Pub.

Group.

Callaghan, J., & Wilson, P. (2004). How to Sing and See. Sydney: Cantare.

Cleveland, T. F., Sundberg, J., & Stone, R. E. (2001). Long-Term-Average-Spectrum

characteristics of country singers during speaking and singing. Journal of Voice,

15(1), 54-60.

Davidson, J. W., & Da Costa Coimbra, D. (2001). Investigating performance evaluation

by assessors of singers in a music college setting. Musicae Scientiae, 5(1), 33-

53.

145

Dmitriev, L., & Kiselev, A. (1979). Relationship between the formant structure of

different types of singing voices and the dimensions of supraglottic cavities.

Folia Phoniatrica, 31(4), 238-241.

Duey, P. A. (1980). Bel canto in its golden age: a study of its teaching concepts. New

York: Da Capo Press.

Easley, E. (1932). A comparison of the vibrato in concert and opera singing. In

Seashore, C. E. (Ed.), The vibrato (pp. 269-275). Iowa: University of Iowa

Press.

Ekholm, E., Papagiannis, G. C., & Chagnon, F. P. (1998). Relating objective

measurements to expert evaluation of voice quality in Western classical singing:

critical perceptual parameters. Journal of Voice, 12(2), 182-196.

Ellis, W. A. (Ed.). (1994). Richard Wagner's Prose Works (Vol. iv A music-school for

Munich). Lincoln: University of Nebraska.

Fenlon, I. (1984). Monteverdi's Mantuan Orfeo. Early Music, 12(2), 163-172.

Fields, V. A. (1947). Training the singing voice: an analysis of the working concepts

contained in recent contributions to vocal pedagogy. New York: King's Crown

Press.

Foreman, E. V. (1969). A comparison of selected Italian vocal tutors of the period circa

1550 to 1800. Unpublished diss., U. of Illinois.

Foulds-Elliott, S., Thorpe, C., Cala, S., & Davis, P. (2000). Respiratory function in

operatic singing: effects of emotional connection. Logopedics Phoniatrics

Vocology, 25(4), 151-168.

Garcia, M. (1911). Garcia's Treatise on the Art of Singing. London: Leonard & Co.

146

Geringer, J. M., & Madsen, C. K. (1998). Musicians' ratings of good versus bad vocal

and string performances. Journal of Research in Music Education, 46(4), 522-

534.

Hakes, J., Shipp, T., & Doherty, E. T. (1987). Acoustic properties of straight tone,

vibrato, trill and trillo. Journal of Voice, 1(2), 148-156.

Hemsley, T. (1998). Singing and imagination: a human approach to a great musical

tradition. Oxford; New York: Oxford University Press.

Hirano, M., Hibi, S., & Hagino, S. (1997). Physiological aspects of vibrato. In

Dejonckere, P. H., Hirano, M. & Sundberg, J. (Eds.), Vibrato (pp. 9-33). San

Diego: Singular Publishing Group, Inc.

Horii, Y. (1989). Frequency modulation characteristics of sustained /a/ sung in vocal

vibrato. Journal of Speech & Hearing Research, 32(4), 829-836.

Howard, D. M. (2003). Editorial. Logopedics Phoniatrics Vocology, 28(4).

Howard, D. M., Szymanski, J., & Welch, G. F. (2002). Listeners' perception of English

girl and boy choristers. Music Perception, 20(1), 35-49.

Howes, P., Callaghan, J., Davis, P. J., Kenny, D. T., & Thorpe, C. (2004). The

relationship between measured vibrato characteristics and perception in Western

operatic singing. Journal of Voice, 18(2), 216-230.

Jander, O., & Harris, E. T. (2004). Singing, Grovemusic online. Retrieved 12 January

2004, 2004

Jansson, E. V., & Sundberg, J. (1975). Long-Time-Average-Spectra applied to analysis

of music. Part I: method and general applications. Acustica, 34, 15-19.

Johnson, P. (1997). Performance as experience: the problem of assessment criteria.

British Journal of Music Education, 14(3), 271-282.

147

Keidar, A., Titze, I. R., & Timberlake, C. (1984). Vibrato characteristics of tenors

singing high Cs. Paper presented at the Proceedings of the Thirteenth

Symposium: Care of the Professional Voice, New York City.

Kenny, D. T., & Mitchell, H. F. (2004, August). Visual and auditory perception of vocal

beauty: conflict or concurrence? Paper presented at the 8th International

Conference on Music Perception & Cognition (ICMPC8), Evanston, IL, USA.

Kenny, D. T., & Mitchell, H. F. (in press). Acoustic and perceptual appraisal of vocal

gestures in the female classical voice. Journal of Voice.

Löfqvist, A., & Mandersson, B. (1987). Long-time average spectrum of speech and

voice analysis. Folia Phoniatrica et Logopedica, 39, 221-229.

Lundy, D. S., Roy, S., Casiano, R. R., Xue, J. W., & Evans, J. (2000). Acoustic analysis

of the singing and speaking voice in singing students. Journal of Voice, 14(4),

490-493.

Mendes, A. P., Rothman, H. B., Sapienza, C. M., & Brown, W. S., Jr. (2003). Effects of

vocal training on the acoustic parameters of the singing voice. Journal of Voice,

17(4), 529-543.

Mendoza, E., Valencia, N., Munoz, J., & Trujillo, H. (1996). Difference in voice quality

between men and women: use of the long-term average spectrum (LTAS).

Journal of Voice, 10(1), 59-66.

Mersenne, M. (1636/1979). Harmonie universelle. In MacClintock, C. (Ed.), Readings

in the history of music in performance. Bloomington: Indiana University Press.

Michel, J. F., & Myers, R. D. (1991). The effects of crescendo on vocal vibrato. Journal

of Voice, 5(4), 292-298.

148

Miller, R. (1996). The structure of singing: system and art in vocal technique. New

York: London: Schirmer Books.

Miller, R. (1997). National schools of singing: English, French, German, and Italian

techniques of singing revisited (Rev. ed.). Lanham, Md: Scarecrow Press.

Miller, R. (1998). The singing teacher in the age of voice science. In Sataloff, R. T.

(Ed.), Vocal Health and Pedagogy (pp. 297-300). San Diego, CA: Singular.

Mitchell, H. F., & Kenny, D. T. (2004a). The effects of open throat technique on long

term average spectra (LTAS) of female classical voices. Logopedics Phoniatrics

Vocology, 29(3), 99-118.

Mitchell, H. F., & Kenny, D. T. (2004b). The impact of "open throat" technique on

vibrato rate, extent and onset in classical singing. Logopedics Phoniatrics

Vocology, 29(4), 171-182.

Mitchell, H. F., Kenny, D. T., Ryan, M., & Davis, P. J. (2003). Defining open throat

through content analysis of experts' pedagogical practices. Logopedics

Phoniatrics Vocology, 28(4), 167-180.

Mitchell, H. F., & Kenny, D. T. (in press). Can experts identify “open throat” technique

as a perceptual phenomenon? Musicae Scientiae.

Monahan, B. J. (1978). The art of singing: a compendium of thoughts on singing

published between 1777 and 1927. Metuchen, N.J.: Scarecrow Press.

Nair, G. (1999). Voice tradition and technology: a state-of-the-art studio. San Diego:

Singular Publishing Group.

Nannariello, J., Osman, M. R., & Fricke, F. R. (2002). Recent developments in the

application of neural network analysis to architectural and building acoustics.

Acoustics Australia, 29, 103-110.

149

Novak, A., & Vokral, J. (1995). Acoustic parameters for the evaluation of voice of

future voice professionals. Folia Phoniatrica et Logopedica, 47, 279-285.

Omori, K., Kacker, A., Carroll, L. M., Riley, W. D., & Blaugrund, S. M. (1996).

Singing power ratio: quantitative evaluation of singing voice quality. Journal of

Voice, 10(3), 228-235.

Playford, J. (1694/1972). An introduction to the skill of musick. New York: Da Capo

Press.

Potter, J. (1998). Vocal authority: singing style and ideology. New York: Cambridge

University Press.

Praetorius, M. (1619/1979). Syntagma musicum. In MacClintock, C. (Ed.), Readings in

the history of music in performance. Bloomington: Indiana University Press.

Prame, E. (1994). Measurement of the vibrato rate of ten singers. J Acoust Soc Am,

96(4), 1979-1984.

Prame, E. (1997). Vibrato extent and intonation in professional Western lyric singing. J

Acoust Soc Am, 102(1), 616-621.

Reid, C. L. (1975). Voice: psyche and soma. New York: Joseph Patelson Music House.

Reid, C. L. (1983). A dictionary of vocal terminology. New York: Joseph Patelson

Music House.

Rosand, E. (1989). Monteverdi's mimetic art: L'incoronazione di Poppea. Cambridge

Opera Journal, 1, 113-137.

Rossing, T. D., Sundberg, J., & Ternstrom, S. (1986). Acoustic comparison of voice use

in solo and choir singing. J Acoust Soc Am, 79(6), 1975-1981.

Rossing, T. D., Sundberg, J., & Ternstrom, S. (1987). Acoustic comparison of soprano

solo and choir singing. J Acoust Soc Am, 82(3), 830-836.

150

Rothman, H. B., & Arroyo, A. A. (1987). Acoustic variability in vibrato and its

perceptual significance. Journal of Voice, 1(2), 123-141.

Rothman, H. B., Diaz, J. A., & Vincent, K. E. (2000). Comparing historical and

contemporary opera singers with historical and contemporary Jewish cantors.

Journal of Voice, 14(2), 205-214.

Saunders, T. C., & Holahan, J. M. (1997). Criteria-specific rating scales in the

evaluation of High School instrumental performance. Journal of Research in

Music Education, 45(2), 259-272.

Seashore, C. E. (1938). Psychology of music. New York and London: McGraw-Hill

Book Company.

Seashore, C. E. (Ed.). (1932). The vibrato. Iowa: University of Iowa Press.

Seidner, W., Nawka, T., & Cebulla, M. (1995). Dependence of the vibrato on pitch,

musical intensity, and vowel in different voice classes. In Dejonckere, P. H.,

Hirano, M. & Sundberg, J. (Eds.), Vibrato (pp. 63-82). San Diego: Singular

Publishing Group.

Shaw, G. B. (Ed.). (1960). How to become a music critic. London: Rupert Hart-Davis.

Shipp, T., Sundberg, J., & Hadlund, S. (1984). A model of frequency vibrato. Paper

presented at the Transcripts of the Thirteenth Symposium: Care of the

Professional Voice, New York.

Stanley, M., Brooker, R., & Gilbert, R. (2002). Examiner perceptions of using criteria in

music performance assessment. Research Studies in Music Education, 18, 43-52.

Stark, J. A. (1999). Bel canto: a history of vocal pedagogy. Toronto, Buffalo: University

of Toronto Press.

151

Sundberg, J. (1974). Articulatory interpretation of the "singing formant". J Acoust Soc

Am, 55(4), 838-844.

Sundberg, J. (1977). The acoustics of the singing voice. Scientific American(March),

82-91.

Sundberg, J. (1988). The science of the singing voice. DesKalb, Illinois: Northern

Illinois University Press.

Sundberg, J. (1995a). Acoustic and psychoacoustic aspects of vocal vibrato. In

Dejonckere, P. H., Hirano, M. & Sundberg, J. (Eds.), Vibrato (pp. 35-62). San

Diego: Singular Publishing Group.

Sundberg, J. (1995b). The singer's formant revisited. Voice, 4, 106-119.

Sundberg, J. (1997). Expressivity in singing. TMH-QPSR, 2(3), 13-19.

Sundberg, J. (2001). Level and center frequency of the singer's formant. Journal of

Voice, 15(2), 176-186.

Sundberg, J., Iwarsson, J., & Hagegard, H. (1995). A singer's expression of emotions in

sung performance. In Fujimura, O. & Hirano, M. (Eds.), Vocal Fold Physiology:

Voice Quality Control (pp. 217-231). San Diego, California: Singular Publishing

Group, Inc.

Thorpe, C., Cala, S., Chapman, J., & Davis, P. (2001). Patterns of breath support in

projection of the singing voice. Journal of Voice, 15(1), 86-104.

Thurman, L., & Welch, G. F. (Eds.). (2000). Bodymind & voice: foundations of voice

education (Rev. ed. ed.). Collegevile, Minn: VoiceCare Network.

Titze, I. (1998). Voice research: the wide pharynx. Journal of Singing, 55(1), 27-28.

Titze, I. (2001). Acoustic interpretation of resonant voice. Journal of Voice, 15(4), 519-

528.

152

Titze, I., Bergan, C. C., Hunter, E. J., & Story, B. (2003). Source and filter adjustments

affecting the perception of the vocal qualities twang and yawn. Logopedics

Phoniatrics Vocology, 28, 147-155.

Titze, I. R., Long, R., Shirley, G. I., Stathopoulos, E., Ramig, L. O., Carroll, L. M., et al.

(1999). Messa di voce: an investigation of the symmetry of crescendo and

decrescendo in a singing exercise. J Acoust Soc Am, 105(5), 2933-2940.

Titze, I. R., Solomon, N. P., Luschei, E. S., & Hirano, M. (1994). Interference between

normal vibrato and artificial stimulation of laryngeal muscles at near-vibrato

rates. Journal of Voice, 8(3), 215-223.

Tosi, P. F. (1743/1967). Opinioni de' cantori antichi e moderni. English Observations

on the florid song translated into English by Mr. Galliard. London: William

Reeves.

Treitler, L., & Strunk, O. (Eds.). (1998). Foreword to Il quinto libro de' madrigali with

the "Declaration" of Giulio Cesare Monteverdi (Vol. 4). New York and London:

W. W. Norton & Co.

Vennard, W. (1968). Singing: the mechanism and the technic (5th ed.). New York:

Fischer.

Vurma, A., & Ross, Y. (2000). Priorities in voice training: carrying power or tone

quality. Musicae Scientiae, 4(1), 75-93.

Wapnick, J., & Ekholm, E. (1997). Expert consensus in solo voice performance

evaluation. Journal of Voice, 11(4), 429-436.

Weiss, R., Brown, W. S., Jr., & Morris, J. (2001). Singer's formant in sopranos: fact or

fiction? Journal of Voice, 15(4), 457-468.

Worsthorne, S. (1968). Venetian opera in the seventeenth century. Oxford.

153

Zwicker, E., & Fastl, H. (1999). Psychoacoustics: facts and models. Berlin: Springer.

154

APPENDICES FOR PAPER 1

155

Appendix A: Project 1 Ethics Proposal

Ethics Application Project Outline

Reference 1/06/20 Supralaryngeal and Pharyngeal Widening During Singing

Many pedagogical texts advocate ‘opening the throat’ in singing in order to achieve a

certain vocal quality and timbre. The broad objectives of this research are to describe

current views and practices on pedagogical approaches that advocate voluntary control

of widening or ‘retraction’ of the ventricular folds (Estill, 1996) as well as other

approaches for pharyngeal or throat widening, for example those taught by Miller

(1996) and others (Reid, 1983). All of these approaches are collectively termed as

‘supralaryngeal and pharyngeal widening’ in this study. This is the first of a series of

novel research projects which will assess the attitudes of pedagogues and performers

towards the different pedagogical approaches as well as discovering the physiological

changes that occur, if any, when different pedagogical approaches related to

supralaryngeal and pharyngeal widening are employed. These studies will inform

pedagogy and contribute basic knowledge to singing science.

This study seeks to describe the knowledge of professional and semi-professional

performers and experienced teachers with regard to supralaryngeal and pharyngeal

widening through qualitative interviewing, in order to illustrate different terminology

and approaches used in singing pedagogy. Both classical operatic and contemporary

music theatre singing will be studied as these styles reflect the current demand on many

classical teachers’ studio time. It may also be that there are intrinsic differences across

these different styles. The objective is to study a set of different pedagogical

156

instructions used by singing teachers focused on the control of the laryngeal and

pharyngeal region.

The interview format is designed to gain information about teacher and performer

approaches to singing pedagogy. The interview will commence with a series of open-

ended and general questions with regard to all aspects of how they achieve a good

sound while singing, and factors that they believe enhance or facilitate good singing

practice. When and if they mention one of the terms associated with ‘supralaryngeal and

pharyngeal widening’, a more direct line of questioning will follow seeking more

specific information related to terminology, sound and awareness, pedagogy,

physiology and perception. If interviewees do not spontaneously mention the concepts

of interest, a question to the effect of 'Have you ever heard of throat widening?' will be

asked and then the interviewees will be invited to disclose their knowledge of the

technique.

The purpose of this project is to address the knowledge and awareness of trained singers

teachers about supralaryngeal and pharyngeal widening. It will also determine how and

when it is used in the singing studio. The interviewing will give a clear response of the

knowledge and preference of the singers for such pedagogical techniques from an

empirical point of view. The research will convey the knowledge, preference and

reasons for expressed preferences of singers and teachers for supralaryngeal and

pharyngeal widening in pedagogy and establish whether other terms are used, for

example via imagery as an access to these experts’ views on the likely effects of

supralaryngeal and pharyngeal widening on voice quality and physiology. The results of

157

this study will provide empirical information used in singing training, its underlying

similarities and differences as well as draw attention to areas needing further scientific

study.

Participants will be from a variety of backgrounds. In the interview, teachers will be

asked to indicate the main pedagogical influences on their style of teaching.

Approximately the same number of teachers who indicate Estill (1996) as a major

influence will be sought as those who indicate more classically mainstream pedagogues

such as Miller (1996). Some leading teachers will be approached directly and invited to

participate in the study. Others will be recruited through a general advertisement.

Respondents will be sent a Subject Information Sheet which will provide information

about the project. All teacher subjects will be encouraged to recommend to their

students, who perform regularly, to take part in the same study. These subjects will

contact the research team directly in order to participate.

With regard to demographic details, subjects will be asked a few questions about their

own singing background in order to ascertain the level of teaching or performing. The

interviews will be conducted in private, with a single subject, either teacher or singer in

a venue convenient for them. For leading teachers not resident in Sydney, the interview

will be conducted by telephone. Prior to the interview, subjects will be asked about their

teaching and performing background. They will be asked to classify themselves

according to a taxonomy (Bunch and Chapman, 2000)

158

The study will provide a basis in order to develop a framework for future, more specific

studies which will assess the association between any changes in the configuration or

movement of the supralaryngeal and pharyngeal structures and any reproducible voice

quality change within subjects and across subjects.

Questions on which the interview will be based are shown below. Please note that in the

open-ended interview, these questions will be developed in the conversation rather than

in the order shown.

The conversation will begin with very general questions on pedagogy: 'How do you

achieve a good sound?' and 'What techniques would you use to correct a problematic

sound?'. At any mention of a term associated with supralaryngeal and pharyngeal

widening, the questions below will be worked into the converstion.

1. Based on your experience, what do the terms Open Throat, Throat Widening,

Retraction and Space at the Back (OT/TW/R/SB) mean to you?

2. How do you feel these terms are related?

How would you describe the practice, if not one of the terms above?

3. What sound quality do you associate with these words?

What differences can you hear when someone is using OT/TW/R/SB compared

to someone who is not?

4. How does the sound quality let you know that someone is using it?

5. In your own singing, when do you/would you use it?

6. Why do you/would you use OT/TW/R/SB?

7. Suppose I was a new student – how would you tell me about some of these terms?

159

Is it something you teach?

How do you teach it?

8. Do you follow any particular ‘school of thought’ or pedagogy or scientific approach?

How does that pedagogy/writer influence your view on OT/TW/R/SB?

9. To what extent do you think that the use of OT/TW/R/SB improves or enhances the

final singing sound?

10. For you, what part of the singing process is OT/TW/R/SB involved in?

All … beginning … preparation?

11. In terms of physiology, which muscles would you expect to be working, that is to

say, what actually happens when you use OT/TW/R/SB?

12. Where do you feel it?

13. Do you think it is possible to have voluntary control of muscles responsible for

OT/TW/R/SB? Either through sensation or imagery?

Do you believe it is possible to isolate and control the muscles used in

OT/TW/R/SB?

160

Appendix B: Project 1 Subject Information Sheet

National Voice Centre

The University of Sydney

SUBJECT INFORMATION SHEET A RESEARCH STUDIES

Project Title: Current Vocal Pedagogy Investigators: Associate Professor Pamela Davis Associate Professor Dianna Kenny Ms Maree Ryan Ms Helen Mitchell Project Contact person: Associate Professor Pamela Davis, phone 9351 5352 or 0411 449 274. This project has been approved by the University of Sydney Institutional Ethics Committee. Any persons with concerns or complaints about the conduct of the research study can contact the Manager of the Ethics and Biosafety Administration on 02-9351-4811. 1. You are invited to participate in the project described below. You will, at all times, be free to decline to continue with a particular task, or to cease participating in the project. Approximately two weeks after sending this Information Sheet, Ms Mitchell will contact you by telephone to discuss whether you wish to participate and if so, to make an appointment. 2. Background to project This project will describe aspects of current vocal pedagogy, its use and its application in the singing studio. 3. Description of project - methods and demands If you choose to participate in this study you will be required to: • Fill in a short questionnaire with generalized details of your background and current teaching in

singing. You will also be asked to self-classify yourself and/or your students as to what genre of music you perform and at what level, in accordance with a set taxonomy.

• Participate in a thirty-minute semi-structured interview conducted by Helen Mitchell, discussing current vocal pedagogy and its application in your singing studio. The interview will be taped, with your permission, for later transcription. Taping of the interview will not begin until we are ready to start the interview itself. When the interview has been transcribed, you will be sent a copy and have the opportunity to confirm your answers.

• Recommend to your students who perform regularly to participate as Performer Subjects and to provide with Information Sheet B which will be supplied to you. The students would be interviewed with similar questions but with a focus on their learning. The data analyses and reports from this study will not enable any person to link specific student and teacher answers.

161

Confidentiality of all your records will be maintained by coding your results with a number and not your name. The results of the analyses may be published but your name or any groups will not be identified (i.e. the data will not enable readers of any of the output, including drafts of data analysis provided to the various investigators which may include papers for publication) to link specific data with voice type, age, or other features which may enable identification of individuals. There will be a debriefing session after all subjects have been interviewed. 4. Possible risks There are no risks involved in the participation of this project. We will be delighted to answer any further questions that you may have. If you are satisfied to continue with the study, please contact one of us to arrange an appointment. Yours sincerely, Associate Professor Pamela Davis Associate Professor Dianna Kenny Ms Maree Ryan Ms Helen Mitchell

162

Appendix C: Project 1 Subject Consent Sheet



CONSENT SHEET RESEARCH STUDIES

Project Title: Current Vocal Pedagogy Investigators: Associate Professor Pamela Davis Associate Professor Dianna Kenny Ms Maree Ryan Ms Helen Mitchell 1. I acknowledge that I have read the above statement which explains the nature, object and the possible risks of the investigation, and the statement has been explained to me to my satisfaction. Before signing this document I have been given the opportunity to ask questions relating to any possible physical harm I may suffer as a result of my participation and I have received satisfactory answers. I have also been informed that I may not receive any benefits from participating in this study. I have been offered a debriefing session after the research has been completed. 2. My decision whether or not to participate will not prejudice my future relations with The National Voice Centre or any of the investigators listed above. If I decide to participate, I am free to withdraw my consent and to discontinue participation at any time without prejudice. 3. I agree that research data gathered from the results of the study may be published provided my name is not used. __________ _________________________________ DATE Signature of Subject 4. I have fully explained to the subject _______________________________ the nature and purpose of the programme and the procedures to be employed as described above and such risks as are involved in their performance. __________ _________________________________ DATE Signature of Responsible Investigator 5. The foregoing is an accurate summary of the explanation which was made to the volunteer, and the undersigned witnessed the signatures. __________ _________________________________ DATE Signature of Auditor-Witness

163

Appendix D: Project 1 Subject Questionnaire



Questions for Participants

1. Age

2. Please outline your professional education

3. How many years have you been teaching?

4. With whom have you trained?

5. Please list masterclasses and workshops you have attended that have influenced your teaching.

6. What proportion of your professional life and income is generated through singing and the

teaching of singing?

7. Where do your students perform and at what level? Please use the Bunch/Chapman

Taxonomy given on the next pages*.

*Bunch, Meribeth and Chapman, Janice, ‘Taxonomy of Singers Used as Subjects in Scientific Research’, Journal of Voice, 14.3 (September 2000), 363-369

164

Categories of Singers * Singers have been divided into nine categories based on proven performance achievement.

1. Superstar • A singer of world-wide fame and recognition who commands the highest fees • Is supported by powerful marketing and a personal entourage • Is never out of the public eye • Commands media attention • Professional support may include teachers, coaches, repetiteurs and choreographers

2. International • A singer who performs in the international arena • Is known to the profession and amongst the public according to specific genre or style category • Commands good fees • Agency representation • Has high levels of training and support

3. National/Big City

• A singer recognisable at the national and big city level, who usually sings in the vernacular • Has good levels of pay • Agency representation • Possible marketing • Training and support according to what can be afforded

4. Regional/Touring (often seasonal)

• Normally a young singer gaining experience • May still be in training • May have an agent • Rarely has marketing

5. Local Community (often semiprofessional)

• Usually paid soloists and club singers who have sporadic engagements • Paid local prevailing rates for performance • Have variable training

6. Singing Teachers

• Who do not fit any of the other categories

7. Full-time Students of Singing (ages 18-25 in:) • Tertiary specialist training courses • Universities • Music and drama colleges

8. Amateur

• Sings for pleasure (unpaid but not unrewarded)

9. Child • Prepubertal voice

Types of Singing Opera … Contemporary Music Theatre … Musical Theatre … Concert/Oratorio/Recital … Recording Artist … Pop … Rock … Rap … Cabaret and Club … Jazz … Folk … Gospel and Soul … Country and Western … Pub and Karaoke … Church ad Cathedral … World Music … Vocal Groups (inc. Barbershop) … Session Singer … Busker/Street Singer.

* This categorisation is taken directly from the ‘Taxonomy of Singers’ of Meribeth Bunch and Janice Chapman. Bunch, Meribeth and Chapman, Janice, ‘Taxonomy of Singers Used as Subjects in Scientific Research’, Journal of Voice, 14.3 (September 2000), 363-369

165

1** SUPERSTAR % 3 NATIONAL/BIG CITY (con’d)

%

1.1 Opera 3.2 Contemporary Music Theatre 1.2 Contemporary Music Theatre 3.2a Major Principal 1.3 Musical Theatre 3.2b Minor Principal 1.4 Concert/Oratorio/Recital 3.2c Ensemble 1.5 Recording Artist Primarily 3.3 Musical Theatre 1.6 Pop 3.3a Major Principal 1.7 Rock 3.3b Minor Principal 1.8 Rap 3.3c Ensemble 1.9 Cabaret and Club 3.4 Concert/Oratorio/Recital 1.10 Jazz 3.5 Recording Artist Primarily 1.11 Folk 3.6 Pop 1.12 Gospel and Soul 3.7 Rock 1.13 Country and Western 3.8 Rap 3.9 Cabaret and Club 2 INTERNATIONAL % 3.10 Jazz 2.1 Opera Principal 3.11 Folk 2.2 Contemporary Music Theatre

Principal 3.12 Gospel and Soul

2.3 Musical Theatre 3.13 Country and Western 2.4 Concert/Oratorio/Recital 3.14 Pub and Karaoke 2.5 Recording Artist Primarily 3.15 Church/Cathedral 2.6 Pop 3.15a Soloist 2.7 Rock 3.15b Professional Chorister 2.8 Rap 3.15b1 Adult 2.9 Cabaret and Club 3.15b2 Child 2.10 Jazz 3.16 World Music 2.11 Folk 3.17 Vocal Groups (inc. Barbershop) 2.12 Gospel and Soul 3.18 Session Singer 2.13 Country and Western 3.19 Busker (specify type, Opera,

pop, etc)

2.14 World Music (specify type) 2.15 Vocal Groups (inc. Barbershop) 4 REGIONAL/TOURING % 4.1 Opera 3 NATIONAL/BIG CITY % 4.1a Major Principal 3.1 Opera 4.1b Minor Principal 3.1a Major Principal 4.1c Chorus/Cover 3.1b Minor Principal 4.2 Contemporary Music Theatre 3.1c Chorus 4.2a Major Principal 4.2b Minor Principal 4.3c Chorus/Cover

166

4** REGIONAL/TOURING (con’d)

% 6 SINGING TEACHER %

4.3 Musical Theatre 6.1 University/Schools/Colleges 4.3a Major Principal 6.2 Private 4.3b Minor Principal 4.3c Ensemble/Cover 7 FULL-TIME VOICE % 4.4 Seasonal Music Theatre/Touring

Co STUDENT

Summer Rep/Pantomime 7.1a Post graduate specialist 4.5 Concert/Oratorio/Recital training courses 4.6 Pop 7.1b University and College 4.7 Rock 7.1c Music and Drama School 4.8 Rap 4.9 Cabaret and Club 8 AMATEUR % 4.10 Jazz 8 (sings for pleasure) 4.11 Folk 4.12 Gospel and Soul 9 CHILD % 4.13 Country and Western 9.1 Opera Principal 4.14 World Music (specify type) 9.2 Contemporary Music Theatre

Principal

4.15 Vocal Groups (inc Barbershop) 9.3 Musical Theatre Principal 9.4 Oratorio Soloist 5 LOCAL COMMUNITY 9.5 Recording artist (Often semi-professional) 9.6 Pop

5.1 Opera (principal parts only) 9.7 Rock

5.2 Contemporary Music Theatre 9.8 Rap

5.3 Musical Theatre 9.9 Children’s Choir

5.4 Concert/Oratorio/Recital

5.5 Church Soloist

5.6 Pop

5.7 Rock

5.8 Rap

5.9 Cabaret and Club

5.10 Jazz

5.11 Folk

5.12 Gospel and Soul

5.13 Country and Western

5.14 Pub and Karaoke

5.15 World Music

5.16 Vocal Groups (inc. Barbershop)

5.17 Busker

**The table and wording of the table used in this questionnaire is taken from the Bunch and Chapman Taxonomy. The % column is my addition for ease of use in this study. Bunch, Meribeth and Chapman, Janice, ‘Taxonomy of Singers Used as Subjects in Scientific Research’, Journal of Voice, 14.3 (September 2000), 363-369

167

APPENDICES FOR PAPERS 2 AND 3

168

Appendix E: Project 2 Ethics Proposal

Ethics Application Project Outline

Reference 02/05/08 Musical terminology, particularly in the field of performance has been a source of

uncertainty and mystification for those who use and practice it. Many of the terms used

to describe ‘openness’ as an intrinsic element of the singing voice have not come under

sufficient scrutiny to determine a consistent terminology or associated sound quality

among expert singing pedagogues. Seashore (1938) recommended that ‘musicians scrap

a mass of the current synonyms for tone quality, because these words do not connote

any demonstrable differences in content. The diversity of words simply adds to the

confusion.’(p 111) There is, and has always been a recognition of an overall ‘good’

sound, but in this case of ‘openness’, there is need of clarification, especially in the area

of singing at a professional and/or expert level.

The use of specific words to describe a sound quality may or may not be effective in the

singing studio. A term used by a teacher to elicit a particular ‘sound quality’ is arguably

effective only if it achieves the desired effect, and not a subsequent incidental change.

The ability to divide and distinguish different sound elements with different singing

instructions from the greater whole of an ‘excellent sound’ attracts significant support in

current vocal pedagogy (Estill 1996, 1997, Miller 1996).

This study seeks to discover if a set of specific instructions to a group of singers can

change their sound quality and if this is demonstrable when a recording of their singing

is played to a panel of expert judges. The ability to distinguish precisely which

169

instruction evokes which sound is important because it maximises teaching

effectiveness and efficiency. However, it remains difficult to determine precisely the

quality of sound achieved by individual instructions. We rely not on words to describe

the sound, but the ears and experience of the pedagogues working with the most

successful singers at the current time to judge the sound.

These instructions relate to the area of ‘opening or widening’ any of the vocal tract

structures (eg. ventricular folds, and/or pharynx) from an already unconstricted starting

point in the sound. The aim of the project is to find out if a precise verbal instruction or

actual gesture can trigger a sound quality difference. That is to say, can a singer

demonstrate a difference in sound quality in a specific way, when they are given a

precise instruction believed to change the sound in terms of openness?

From the earlier study (Ethics Approval 1/06/20), we have learnt from interviews with

leading pedagogues and from some of their students a number of terms used to describe

sound qualities associated with the concept of open throat in singing. This project will

use these terms in an attempt to derive actual associations with sound quality in singing.

Methodology

Singers: Ten female singer subjects (sopranos and mezzo-soprano) will be recruited. Ms

Maree Ryan, a co-investigator from the Conservatorium of Music, is one of Australia’s

leading singing pedagogues and has many students achieve national and international

level singing. She will select singers on the basis of her expert judgement on their

expected ability to perform the proposed protocol. They will be asked to attend a

170

recording session at the Conservatorium of Music or the National Voice Centre at a

time that is convenient for them and the researchers.

Protocol for singer subjects

Singers will be asked to sing a portion of ‘Ridente la calma’ (a Mozart aria), and a

portion of ‘Du bist die Ruh’ (a Schubert lied). They will perform each under two

conditions: the first with a healthy but with a ‘not truly open’ sound; and secondly with

a ‘fully open’. Ms Ryan will select the terminology and instructions necessary for each

subject to achieve both these sounds at an optimal level. She will also be able to

accurately discern, from her wide experience, whether or not each singer is achieving

her best possible rendition for the two conditions. Singers may be asked to repeat a take

if she or the singer judge it to be less than the usual standard for that singer. Singers will

also be asked to sing a messa di voce, from the most quiet singing to the loudest and

then back to the most quiet, on a number of notes throughout their range.

Sound recording: The acoustic signal will be recorded digitally (Behringer Ultragain

preamplifier/ Marantz CDR 630) via a high-quality microphone (AKG C-477)

positioned on a head boom a constant 7 cm distance from the subject's lips so that direct

energy of the performer's voice will be recorded rather than room reflections, enabling

us to use a studio environment at the Conservatorium of Music with low ambient noise

rather than an anechoic chamber for all the recordings. A sound level meter (Rion) will

be used to calibrate the sound pressure level by reference to the level of a calibration

tone played during the recording session at the same recording gain. The singer subjects

will be asked to describe their own assessment of the performance in both conditions,

171

both of sensation and sound quality. We will adopt a qualitative approach for the

description of these experiences, as this feedback is likely to be subtle and variable. The

performers will also self-rate their performance immediately after recording each of the

conditions on a 100 mm visual analogue scale (VAS) for: warmth, freedom throughout

range, consistent sound quality, warmth, appropriate vibrato, openness, my best singing.

Perceptual analysis: The 60 excerpts (~45 secs in duration) from the original recordings

will be randomly ordered onto a CD, with 20% extra duplication to assess intra-judge

reliability, and be played to expert pedagogues for evaluation. The same excerpts used

for the acoustic analysis will be analysed perceptually. The recordings will be played by

Ms Mitchell to the pedagogue subjects, with short breaks between excerpts and asked to

rate the performances in a questionnaire on a 100 mm Visual Analogue Scale. They will

be asked about a number of criteria (warmth, freedom throughout range, consistency of

sound quality, openness, brilliance/ring, appropriate vibrato and overall performance

excellence), but the particular focus will be on associations with the concept of

‘openness’. Subjects will be able to ask to hear any particular example again, at the end

of the session. They will work in a quiet environment and the sound will be delivered by

quality speakers from a CD player. They will listen to each perceptual set within the

same day. Ms Mitchell will be present during the perceptual ratings. Intra-rater

reliability which will be set at >70% point by point agreement for judgments of

duplicate performances.

Acoustic: The overall characteristics of the recorded voice signals from the experiments

will be described using Soundswell signal analysis software (HiTech, Sweden). Long

172

term average spectra (LTAS) calculated from the voiced segments will be used to

describe of the singer's spectra in addition to an analysis of messa di voce spectra. In

order to quantify the acoustic characteristics of any changes arising from different tasks,

we will select the same climatic section of 45 seconds in duration from the song. We

will characterise the LTAS by means of its peak frequency, level of the fundamental

relative to level of F1, average spectral slope, and the presence, frequency, and

magnitude of any spectral peak in the 2-4 kHz frequency range. We will also compare

performances in the conditions using as an index the relative change in the range 0-2 vs

2-4 kHz range and will correlate the spectral energy in the 0-2 and 2-4 kHz regions with

sound pressure level. Spectrographic analysis will be performed and the presence and

depth of vibrato or any other feature relevant to the performers' voices recorded will be

measured.

173

Appendix F: Project 2 Singer Information Sheet



RESEARCH STUDY INTO THE USE OF SUPRALARYNGEAL AND PHARYNGEAL WIDENING (OPEN THROAT) IN SINGING

SINGER SUBJECT INFORMATION SHEET

You are invited to take part in a research study into the use of supralaryngeal and pharyngeal widening (open throat) in singing. The study is being conducted by Helen Mitchell, a postgraduate student at The University of Sydney, Associate Professor Pamela Davis (Director of the National Voice Centre), Associate Professor Dianna Kenny (Director of Postgraduate Studies, National Voice Centre) and Ms Maree Ryan (Lecturer in Singing, The Conservatorium of Music), and will form the basis of Helen Mitchell’s PhD at The University of Sydney.

The object is to investigate acoustical and perceptual analysis of the use of open throat in singing and to discover the sound qualities associated when a singer uses some form of open throat technique compared with when a singer does not use open throat.

If you agree to take part in this study, you will be asked to complete a short questionnaire that tells us a bit about you (your age, education, occupation and singing experience) and then to attend a recording session at the Conservatorium of Music or the National Voice Centre for a single recording at a time arranged with the researchers, lasting no more than an hour.

At this recording session, you would be asked to sing for a recording a portion of Ridente la calma (Mozart), a portion of Du bist die Ruh (Schubert), some scales and messe di voce across your vocal range and then give feedback on your view of each performance. We will ask you to sing all of these things in two ways: the first with the maximum degree of openness you would use in your technique, and the second with a smaller degree of openness, but still with an acceptable singing technique. Both these conditions will be around the same volume, so that the test is of the difference in sound quality. There are no risks involved in the participation of this project.

All aspects of the study, including results, will be strictly confidential and only investigator Helen Mitchell will have access to information on participants. Results will be coded with a number and not your name and the data will not enable readers of any of the output to link specific data with voice type, age, or other features which may enable identification of individuals. A report of the study may be submitted for publication, but individual participants will not be identifiable in such a report. There will be a debriefing session on completion of all data collection.

Participation in this study is entirely voluntary: you are not obliged to participate and - if you do participate - you can withdraw at any time. Whatever your decision, it will not affect your relationship with any of the researchers, or with the National Voice Centre.

The ethics committee of The University of Sydney has approved this study. If you have a concern or complaint about the conduct of a research study, you can contact the Manager for Ethics and Biosafety Administration, University of Sydney on (02) 9351 4811.

When you have read this information, Helen Mitchell will discuss it with you further and answer any questions you may have. If you would like to know more at any stage, please feel free to contact Helen Mitchell on 9517 5435 or [email protected].

174

Appendix G: Project 2 Singer Consent Sheet



CONSENT SHEET

RESEARCH STUDIES

Project Title: The Use of Supralaryngeal and Pharyngeal Widening (Open Throat) in Singing Investigators: Associate Professor Pamela Davis Associate Professor Dianna Kenny Ms Maree Ryan Ms Helen Mitchell 1. I acknowledge that I have read the above statement which explains the nature, object and the possible risks of the investigation, and the statement has been explained to me to my satisfaction. Before signing this document I have been given the opportunity to ask questions relating to any possible physical harm I may suffer as a result of my participation and I have received satisfactory answers. I have also been informed that I may not receive any benefits from participating in this study. I have been offered a debriefing session after the research has been completed. 2. My decision whether or not to participate will not prejudice my future relations with The National Voice Centre or any of the investigators listed above. If I decide to participate, I am free to withdraw my consent and to discontinue participation at any time without prejudice. 3. I agree that research data gathered from the results of the study may be published provided my name is not used. __________ _________________________________ DATE Signature of Subject 4. I have fully explained to the subject _______________________________ the nature and purpose of the programme and the procedures to be employed as described above and such risks as are involved in their performance. __________ _________________________________ DATE Signature of Responsible Investigator

175

Appendix H: Project 2 Singer Questionnaire The Use of Supralaryngeal and Pharyngeal Widening (Open Throat) in Singing

Questionnaire for Singer Subjects Page 1 of 3



The Use of Supralaryngeal and Pharyngeal Widening (Open Throat) in Singing

SINGER SUBJECT DEMOGRAPHIC QUESTIONNAIRE

We are interested in finding out about you. Please complete the following questionnaire by ticking the box that best describes you or writing in the space provided. When you have finished, please hand the questionnaire back to one of the researchers.

The questions begin on the next page.

176



1. How old are you?

_____________years 2. At what age did you begin individual singing lessons?

_____________years 3. For how many years (or months) in total have you had individual singing lessons?

_____________years (or _____________months)

4. How many different singing teachers have you had?

_____________teachers 5. Please specify the duration of your training with each singing teacher (beginning with

your first teacher):

Teacher 1: ____years or ____months Teacher 2: ____years or ____months

Teacher 3: ____years or ____months Teacher 4: ____years or ____months Teacher 5: ____years or ____months Teacher 6: ____years or ____months

177



6. What is the highest level of singing education you have completed?

Postgraduate degree at University or Conservatorium

Bachelor degree at University or Conservatorium

Diploma of Music

Higher School Certificate or equivalent

7. What is your current level of singing education (if this applies)?

Postgraduate degree/diploma at University of Conservatorium

Bachelor degree at University or Conservatorium

8. What voice category are you? (please tick)

Soprano.....................................

Mezzo soprano .........................

Contralto ..................................

9. What types of singing do you perform in public? Please circle the relevant singing type

and write in the space the percentage of your total performance time spent performing each type of singing.

Type of Singing % of performance time

Opera

Other classical (eg. Song and concert repertoire, church music)

Choral singing

Musical Theatre

Contemporary styles

Other (please specify)

Thank you for completing this questionnaire

178


179

Appendix I: Vibrato onset raw data

Schubert Vibrato Onset Subjects 1-2

word SO 1 SSO 1 SLSO 1 SO 2 SSO 2 SLSO 2 1 Dies 0.63 0.26 0.04 0.41 0.43 0.48 2 Au 0.14 0.91 0.05 0.23 0.04 0.46 3 zelt 0.15 0.81 0.59 0.26 0.22 0.44 4 Dei 0.08 1.00 0.44 0.28 0.07 0.25 5 glanz 0.08 0.06 0.64 0.31 0.23 0.24 6 lein 0.09 0.09 0.31 0.10 0.19 0.25 7 hellt1 0.08 0.06 0.07 0.10 0.34 0.28 8 hellt 2 nil nil nil nil nil nil 9 ganz 0.03 1.02 1.62 0.26 0.27 0.79

10 ganz 0.09 0.90 0.50 0.30 0.31 0.55 11 Dies 0.16 0.13 0.54 0.25 0.41 0.29 12 Au 0.13 0.52 0.69 0.26 0.09 0.42 13 zelt 0.02 0.81 0.74 0.19 0.26 0.62 14 Dei 0.15 0.08 0.94 0.25 0.24 0.56 15 glanz 0.07 0.40 0.58 0.24 0.23 0.47 16 lein 0.08 0.49 0.79 0.12 0.48 0.24 17 hellt1 0.06 0.06 0.07 0.08 0.33 0.27 18 hellt 2 nil nil nil nil nil nil 19 ganz 0.03 1.25 1.53 0.23 0.29 0.56 20 ganz 0.11 1.24 1.24 0.03 0.28 0.39

180


word SO 3 SSO 3 SLSO 3 SO 4 SSO 4 SLSO 4 1 Dies 0.03 Absent 0.67 0.20 0.15 0.06 2 Au 0.18 0.28 0.61 0.05 0.24 0.15 3 zelt 0.02 0.41 0.63 0.02 0.05 0.03 4 Dei 0.26 0.05 0.60 0.25 0.06 0.04 5 glanz 0.06 0.40 0.46 0.05 0.17 0.23 6 lein 0.22 0.24 0.08 0.07 0.40 0.06 7 hellt1 0.17 0.19 0.09 0.06 0.24 0.06 8 hellt 2 nil nil nil nil nil nil 9 ganz 0.05 0.05 0.21 0.06 0.27 0.04

10 ganz 0.05 0.36 0.39 0.04 0.23 0.22 11 Dies 0.17 Absent 0.22 0.25 0.08 0.07 12 Au 0.03 0.58 0.58 0.03 0.05 0.04 13 zelt 0.03 0.19 0.22 0.03 0.65 0.04 14 Dei 0.24 0.19 0.22 0.22 0.06 0.03 15 glanz 0.08 0.25 0.19 0.20 0.32 0.17 16 lein 0.21 0.06 0.76 0.25 0.55 0.03 17 hellt1 0.05 0.07 0.07 0.18 0.18 0.07 18 hellt 2 nil nil nil nil nil nil 19 ganz 0.24 1.05 0.38 0.04 0.03 0.15 20 ganz 0.22 0.69 1.96 0.28 0.20 0.22

181


word SO 5 SSO 5 SLSO 5 SO 6 SSO 6 SLSO 6 1 Dies 0.28 0.20 0.19 0.23 0.25 0.23 2 Au 0.23 0.18 0.44 0.04 0.22 0.04 3 zelt 0.16 0.18 0.04 0.20 0.26 0.38 4 Dei 0.21 0.05 0.44 0.20 0.22 0.23 5 glanz 0.23 0.22 0.22 0.30 0.32 0.27 6 lein 0.06 0.06 0.19 0.05 0.04 0.04 7 hellt1 0.06 0.07 0.21 0.25 0.10 0.21 8 hellt 2 nil nil nil nil nil nil 9 ganz 0.24 0.19 0.04 0.21 0.08 0.04

10 ganz 0.27 0.33 0.20 0.20 0.21 0.22 11 Dies 0.26 0.19 0.26 0.22 0.22 0.08 12 Au 0.04 0.20 0.18 0.19 0.20 0.05 13 zelt 0.19 0.17 0.21 0.18 0.20 0.25 14 Dei 0.23 0.05 0.06 0.39 0.04 0.21 15 glanz 0.23 0.22 0.18 0.04 0.41 0.05 16 lein 0.06 0.05 0.38 0.05 0.06 0.04 17 hellt1 0.08 0.06 0.07 0.08 0.07 0.11 18 hellt 2 nil nil nil nil nil nil 19 ganz 0.23 0.19 0.24 0.22 0.44 0.05 20 ganz 0.23 0.57 0.66 0.19 0.20 0.19

182

Mozart Vibrato Onset Subjects 1-2

word MO 1 MSO 1 MLSO 1 MO 2 MSO 2

MLSO 2

1 den 0.05 0.23 0.24 0.07 Absent Absent 2 te 0.06 0.07 0.07 0.08 0.07 0.34 3 la 0.07 0.06 0.06 0.21 0.27 Absent 4 cal 0.03 0.22 0.22 0.22 0.23 Absent 5 ma nil nil nil nil nil nil 6 re 0.04 0.41 0.26 0.17 0.04 0.24 7 sti 0.05 0.20 0.22 0.05 0.05 0.06 8 se 0.18 0.43 0.21 0.26 0.41 0.22 9 gno nil nil nil nil nil nil

10 gno e nil nil nil nil nil nil 11 mor nil nil nil nil nil nil 12 den 0.05 0.19 0.22 0.24 0.05 Absent 13 te 0.04 0.35 0.05 0.07 0.06 0.25 14 la 0.08 0.09 0.08 0.08 0.38 0.26 15 cal 0.04 0.04 0.05 0.25 0.07 Absent 16 ma nil nil nil nil nil nil 17 al 0.07 0.21 0.45 0.29 Absent 0.23 18 ma nil nil nil nil nil nil 19 si 0.05 0.19 0.23 0.22 0.24 0.04 20 des 0.07 0.07 Absent 0.24 0.26 0.22 21 sti 0.20 0.05 0.07 0.06 0.05 0.06 22 re 0.07 0.39 0.06 0.03 0.06 0.24 23 sti 0.05 0.03 0.59 0.05 0.05 0.07 24 se 0.04 0.55 0.54 0.04 0.25 0.40 25 gno nil nil nil nil nil nil 26 gno e nil nil nil nil nil nil 27 mor nil nil nil nil nil nil 28 mor 2 nil nil nil nil nil nil 29 res 0.06 0.06 0.23 0.06 Absent Absent 30 sti 0.05 0.05 0.06 0.05 0.24 0.29 31 se 0.25 0.56 0.21 0.20 0.22 0.23 32 gno nil nil nil nil nil nil 33 gno e nil nil nil nil nil nil 34 ti 0.21 Absent 0.05 0.25 0.05 0.22 35 mor nil nil nil nil nil nil 36 di 0.22 0.43 0.46 0.20 Absent 0.20 37 mor nil nil nil nil nil nil

183



MLSO 4

1 den 0.25 0.42 Absent 0.22 Absent 0.22 2 te 0.20 0.19 Absent 0.05 0.36 0.37 3 la 0.06 0.22 0.05 0.21 0.22 0.05 4 cal 0.37 0.19 Absent 0.19 0.19 0.22 5 ma nil nil nil nil nil nil 6 re 0.19 Absent Absent 0.22 0.37 0.23 7 sti 0.20 0.20 0.21 0.06 0.38 0.41 8 se 0.21 0.20 0.25 0.04 Absent 0.23 9 gno nil nil nil nil nil nil

10 gno e nil nil nil nil nil nil 11 mor nil nil nil nil nil nil 12 den 0.06 0.04 0.25 0.06 Absent 0.21 13 te 0.03 0.20 0.23 0.06 0.39 0.21 14 la 0.19 0.18 0.11 0.05 0.21 0.22 15 cal 0.21 0.16 0.41 0.04 0.34 0.40 16 ma nil nil nil nil nil nil 17 al 0.10 0.22 0.22 0.07 0.21 0.21 18 ma nil nil nil nil nil nil 19 si 0.20 Absent 0.40 0.05 Absent 0.19 20 des 0.22 0.22 0.41 0.04 Absent Absent 21 sti 0.21 0.20 Absent 0.05 0.20 0.41 22 re 0.19 0.58 0.24 0.05 0.22 Absent 23 sti 0.21 0.22 0.38 0.21 0.05 0.38 24 se 0.20 0.38 0.20 0.20 Absent 0.44 25 gno nil nil nil nil nil nil 26 gno e nil nil nil nil nil nil 27 mor nil nil nil nil nil nil 28 mor 2 nil nil nil nil nil nil 29 res 0.23 0.25 0.26 0.18 Absent 0.25 30 sti 0.21 0.20 0.24 0.22 0.36 0.23 31 se 0.22 Absent 0.04 0.23 0.39 Absent 32 gno nil nil nil nil nil nil 33 gno e nil nil nil nil nil nil 34 ti 0.17 0.19 0.18 0.05 Absent 0.23 35 mor nil nil nil nil nil nil 36 di 0.23 Absent Absent 0.23 Absent Absent 37 mor nil nil nil nil nil nil

184



MLSO 6

1 den 0.07 0.21 0.21 0.25 Absent 0.21 2 te 0.06 Absent 0.16 0.18 Absent Absent 3 la 0.05 0.50 0.19 0.20 Absent 0.23 4 cal 0.05 Absent 0.04 0.05 Absent 0.25 5 ma nil nil nil nil nil nil 6 re 0.06 0.23 Absent 0.04 Absent 0.23 7 sti 0.05 0.36 0.21 0.05 Absent 0.41 8 se 0.22 0.35 0.18 0.05 Absent 0.21 9 gno nil nil nil nil nil nil

10 gno e nil nil nil nil nil nil 11 mor nil nil nil nil nil nil 12 den 0.05 0.06 0.54 0.06 Absent 0.19 13 te 0.05 0.19 0.18 0.03 0.42 0.19 14 la 0.04 0.06 0.20 0.05 0.23 Absent 15 cal 0.03 0.17 0.19 0.15 Absent 0.19 16 ma nil nil nil nil nil nil 17 al 0.09 0.06 0.37 0.08 0.23 0.12 18 ma nil nil nil nil nil nil 19 si 0.06 0.19 0.35 0.04 Absent 0.24 20 des 0.23 0.20 0.21 0.04 Absent 0.25 21 sti 0.19 0.18 0.18 0.04 0.23 0.41 22 re 0.04 0.06 0.06 0.06 Absent 0.25 23 sti 0.20 0.39 0.39 0.04 0.32 0.42 24 se 0.07 0.36 0.19 0.21 Absent Absent 25 gno nil nil nil nil nil nil 26 gno e nil nil nil nil nil nil 27 mor nil nil nil nil nil nil 28 mor 2 nil nil nil nil nil nil 29 res 0.06 0.05 0.20 0.07 Absent Absent 30 sti 0.07 0.35 0.38 0.04 Absent Absent 31 se 0.05 0.43 0.34 0.21 Absent Absent 32 gno nil nil nil nil nil nil 33 gno e nil nil nil nil nil nil 34 ti 0.17 0.18 0.33 0.16 0.18 Absent 35 mor nil nil nil nil nil nil 36 di 0.15 0.13 0.16 0.24 Absent Absent 37 mor nil nil nil nil nil nil

185

Appendix J: Vibrato extent raw data

Schubert Vibrato Extent Subjects 1-2

word SO 1 SSO 1 SLSO 1 SO 2 SSO 2 SLSO 2 1 Dies 0.87 0.82 0.80 1.22 1.22 0.87 2 Au 0.97 0.46 0.77 1.33 1.00 0.92 3 zelt 1.20 0.70 0.88 1.20 1.05 1.12 4 Dei 1.20 0.48 0.63 1.26 1.19 1.17 5 glanz 1.29 0.85 0.76 1.41 1.29 1.52 6 lein 1.14 0.67 0.63 1.32 1.16 1.17 7 hellt1 0.94 0.44 0.39 1.36 0.92 0.78 8 hellt 2 0.59 0.44 0.49 1.07 0.63 0.71 9 ganz 1.11 0.47 0.45 1.57 1.33 0.92

10 ganz 1.10 0.53 0.65 1.26 1.36 0.78 11 Dies 0.76 0.65 0.42 1.27 1.40 0.71 12 Au 1.12 0.65 0.70 1.20 1.22 0.84 13 zelt 1.13 0.81 0.80 1.33 1.34 0.93 14 Dei 0.99 0.52 0.50 1.56 1.24 0.77 15 glanz 1.27 0.69 0.66 1.54 1.36 1.19 16 lein 1.07 0.57 0.53 1.29 1.30 0.62 17 hellt1 0.82 0.46 0.42 1.36 1.07 0.56 18 hellt 2 0.61 0.42 0.49 1.12 0.98 0.65 19 ganz 0.94 0.33 0.43 1.51 1.57 0.71 20 ganz 0.79 0.61 0.72 1.28 1.20 0.61

186


word SO 3 SSO 3 SLSO 3 SO 4 SSO 4 SLSO 4 1 Dies 0.93 absent 0.33 1.62 0.38 0.78 2 Au 1.40 0.48 0.46 1.82 0.75 0.54 3 zelt 1.45 0.70 0.58 1.52 0.58 0.78 4 Dei 1.60 0.86 0.39 1.68 0.60 0.62 5 glanz 1.31 0.77 0.71 1.63 0.56 0.69 6 lein 1.25 0.83 0.68 1.33 0.66 0.69 7 hellt1 1.21 0.67 0.61 1.21 0.46 0.46 8 hellt 2 1.07 0.51 0.60 1.04 0.44 0.34 9 ganz 1.30 0.60 0.78 1.66 0.69 0.64

10 ganz 1.56 0.36 0.40 1.62 0.81 0.68 11 Dies 1.28 absent 0.30 1.14 0.41 0.43 12 Au 1.47 0.30 0.35 1.62 0.90 0.73 13 zelt 1.66 0.38 0.71 1.62 1.00 0.88 14 Dei 1.96 0.74 0.69 1.63 0.83 0.65 15 glanz 1.46 0.74 0.54 1.52 0.68 0.62 16 lein 1.17 0.75 0.61 1.35 0.56 0.53 17 hellt1 1.18 0.60 0.40 1.16 0.31 0.37 18 hellt 2 0.96 0.45 0.34 1.00 0.29 0.33 19 ganz 1.09 0.30 0.43 1.37 0.68 0.41 20 ganz 1.18 0.81 0.34 1.23 0.55 0.48

187



10 ganz 0.77 0.84 0.78 0.98 0.77 0.94 11 Dies 1.27 1.03 0.96 0.74 0.71 0.42 12 Au 1.21 1.03 0.84 0.66 0.62 0.68 13 zelt 1.22 1.13 0.82 0.84 0.68 0.80 14 Dei 1.28 0.82 0.73 0.87 0.75 0.85 15 glanz 1.19 0.74 0.77 0.77 0.74 0.77 16 lein 1.12 0.78 0.72 1.07 1.02 0.78 17 hellt1 0.85 0.60 0.68 0.94 1.02 1.03 18 hellt 2 0.82 0.73 0.72 0.79 0.77 1.27 19 ganz 0.87 0.72 0.78 0.71 0.61 0.74 20 ganz 0.96 0.76 0.73 1.00 absent 1.07

188

Mozart Vibrato Extent Subjects 1-2


MLSO 2

1 den 1.22 0.72 0.80 0.93 absent absent 2 te 1.21 0.69 0.62 1.21 0.70 0.52 3 la 1.21 0.91 0.65 1.09 0.60 absent 4 cal 1.02 0.96 0.80 1.09 0.29 absent 5 ma 1.19 0.83 0.71 1.20 0.75 0.50 6 re 0.85 0.61 0.59 1.00 0.34 0.50 7 sti 1.23 0.84 0.70 1.01 0.68 0.76 8 se 1.25 0.72 0.50 1.42 0.68 0.70 9 gno 1.23 0.90 0.89 1.18 1.06 0.77

10 gno e 1.27 0.55 0.49 1.22 1.03 0.88 11 mor 1.29 0.60 0.65 1.18 0.94 0.87 12 den 1.02 0.39 0.57 1.13 0.23 absent 13 te 1.26 0.74 0.46 1.32 1.05 0.93 14 la 1.29 0.69 0.74 1.29 0.70 0.57 15 cal 1.23 0.65 0.71 1.28 0.46 absent 16 ma 1.48 0.74 0.80 1.28 0.78 0.89 17 al 0.90 0.40 0.36 1.16 absent 0.35 18 ma 1.20 0.69 0.88 1.58 0.86 0.86 19 si 1.23 0.67 0.77 1.37 0.81 0.80 20 des 1.29 0.57 absent 1.28 0.91 0.75 21 sti 1.38 0.67 0.69 1.45 0.89 0.93 22 re 1.19 0.63 0.45 1.15 0.91 0.53 23 sti 1.38 0.82 0.59 1.34 0.79 0.73 24 se 1.02 0.70 0.40 1.25 0.74 0.79 25 gno 1.25 0.87 0.78 1.34 0.90 0.86 26 gno e 0.99 0.66 0.69 1.35 0.93 0.86 27 mor 1.02 0.60 0.71 1.38 0.97 0.84 28 mor 2 1.20 0.71 0.68 1.31 0.81 0.95 29 res 1.05 0.59 0.39 1.10 absent absent 30 sti 1.17 0.69 0.60 1.37 0.78 0.56 31 se 1.10 0.47 0.35 1.20 0.52 0.77 32 gno 1.18 0.77 0.89 1.31 0.79 0.78 33 gno e 1.14 0.61 0.84 1.69 0.91 0.74 34 ti 0.83 absent 0.49 1.46 1.00 0.80 35 mor 0.95 0.72 0.54 1.13 0.59 0.77 36 di 1.21 0.48 0.33 1.28 absent 0.30 37 mor 1.02 0.51 0.63 1.26 0.90 0.69

189



MLSO 4

1 den 1.90 1.15 absent 1.46 absent 0.59 2 te 1.74 0.34 absent 1.63 0.53 0.68 3 la 1.61 0.36 0.63 1.52 0.50 0.58 4 cal 1.53 0.69 absent 1.43 0.50 0.38 5 ma 1.36 0.92 0.76 1.48 0.58 0.58 6 re 1.93 absent absent 1.64 0.37 0.53 7 sti 1.55 0.80 0.98 1.47 0.74 0.59 8 se 1.68 0.80 0.83 1.53 absent 0.62 9 gno 1.81 1.23 1.00 1.76 0.78 1.02

10 gno e 1.59 0.78 1.11 1.38 0.72 0.62 11 mor 1.67 0.52 0.67 1.62 0.87 0.44 12 den 1.27 0.60 0.24 1.30 absent 0.51 13 te 1.65 1.09 0.76 1.79 0.68 0.67 14 la 2.04 0.55 0.71 1.29 0.52 0.41 15 cal 1.65 1.09 0.85 1.33 0.74 0.39 16 ma 1.68 0.63 0.61 1.50 0.61 0.83 17 al 1.35 0.35 0.31 1.15 0.34 0.44 18 ma 1.33 0.90 1.00 1.68 0.64 0.99 19 si 1.65 absent 0.75 1.32 absent 0.54 20 des 1.77 0.72 1.07 1.41 absent absent 21 sti 1.23 0.90 absent 1.58 0.60 0.71 22 re 1.62 0.84 1.02 1.24 0.53 absent 23 sti 1.45 1.08 1.13 1.46 0.69 0.54 24 se 1.81 0.76 0.43 1.47 absent 0.54 25 gno 1.60 0.69 absent 1.56 absent 0.67 26 gno e 1.85 1.01 1.36 1.51 0.72 0.70 27 mor 1.68 0.51 1.24 1.51 0.44 absent 28 mor 2 2.07 0.95 0.94 1.64 0.72 0.49 29 res 1.56 1.12 0.36 1.36 absent 0.55 30 sti 1.61 1.16 0.80 1.57 0.72 0.62 31 se 1.64 absent 0.81 1.39 0.49 absent 32 gno 1.22 0.89 0.93 1.56 absent 0.76 33 gno e 1.70 1.06 1.23 1.54 0.46 0.49 34 ti 1.48 0.86 0.87 0.79 0.56 35 mor 1.78 0.74 0.30 1.63 0.43 0.58 36 di 1.64 absent absent 1.38 absent absent 37 mor 1.74 0.90 0.58 1.80 0.73 0.54

190



MLSO 6

1 den 1.04 0.72 0.60 0.94 absent 0.33 2 te 1.24 absent 0.62 0.96 absent absent 3 la 0.99 0.63 0.64 0.78 absent 0.31 4 cal 1.05 absent 0.44 0.93 absent 0.47 5 ma 1.10 0.67 0.76 1.10 absent 0.77 6 re 1.09 0.68 absent 0.78 absent 0.41 7 sti 1.03 0.67 0.64 0.69 absent 0.58 8 se 0.85 0.69 0.63 0.72 absent 0.58 9 gno 1.12 0.71 0.74 0.95 0.51 0.84

10 gno e 0.88 0.55 0.50 0.86 absent 0.57 11 mor 1.06 0.66 0.62 0.92 0.59 0.84 12 den 0.99 0.71 0.62 1.21 absent 1.00 13 te 1.04 0.79 0.67 0.67 0.42 0.57 14 la 0.93 0.70 0.58 0.61 0.50 absent 15 cal 0.85 0.72 0.74 0.72 absent 0.50 16 ma 1.13 0.85 0.87 0.94 0.77 0.72 17 al 0.96 0.65 0.69 1.36 0.32 1.22 18 ma 1.12 0.81 0.54 0.80 absent 0.54 19 si 1.08 0.74 0.64 0.79 absent 0.56 20 des 0.93 0.57 0.52 0.78 absent 0.65 21 sti 1.01 0.66 0.69 0.78 0.62 0.93 22 re 1.04 0.59 0.56 1.06 absent 0.58 23 sti 1.15 0.93 0.84 0.85 0.40 0.64 24 se 1.01 0.89 0.44 0.75 absent absent 25 gno 1.06 0.81 0.63 0.76 absent absent 26 gno e 0.80 0.84 0.64 0.86 0.47 0.63 27 mor 0.83 0.56 0.39 0.88 0.36 0.83 28 mor 2 1.12 0.66 0.65 0.83 0.56 0.96 29 res 0.99 0.70 0.52 0.97 absent absent 30 sti 0.96 0.85 0.66 0.78 absent absent 31 se 1.03 0.85 0.43 0.64 absent absent 32 gno 0.97 0.66 0.57 0.76 absent 0.54 33 gno e 0.76 0.91 0.40 0.69 0.57 0.72 34 ti 0.97 0.68 0.52 0.63 0.45 absent 35 mor 0.94 0.64 0.62 0.90 0.59 0.60 36 di 0.86 0.75 0.35 0.76 absent absent 37 mor 1.01 0.61 0.61 1.00 0.66 0.77

191

Appendix K: Vibrato rate raw data

Schubert Vibrato Rate Subjects 1-2



192


word SO 3 SSO 3 SLSO 3 SO 4 SSO 4 SLSO 4 1 Dies 5.77 absent 6.25 6.01 7.49 6.41 2 Au 5.46 5.78 5.77 5.78 5.61 5.89 3 zelt 5.81 5.74 5.57 5.92 6.05 5.80 4 Dei 5.44 5.46 5.17 5.67 5.88 5.61 5 glanz 5.71 5.76 6.40 5.52 5.30 5.52 6 lein 5.62 5.76 5.75 5.68 5.56 5.79 7 hellt1 5.80 5.77 5.85 5.59 5.27 5.57 8 hellt 2 6.01 6.24 5.95 5.57 5.39 5.66 9 ganz 5.41 5.68 5.49 5.76 5.58 5.59

10 ganz 5.33 5.58 5.44 5.43 5.25 5.20 11 Dies 6.14 absent 5.49 6.17 6.18 6.71 12 Au 5.69 4.94 5.34 5.66 5.41 5.40 13 zelt 5.75 5.37 5.28 5.91 5.83 5.58 14 Dei 5.46 5.41 5.33 5.64 5.45 5.51 15 glanz 5.78 5.77 5.45 5.51 5.59 5.63 16 lein 5.77 5.77 5.75 5.69 5.82 5.69 17 hellt1 5.77 5.84 5.84 5.42 5.69 5.34 18 hellt 2 5.98 5.82 6.24 5.51 5.37 5.26 19 ganz 5.48 5.60 5.78 5.60 5.53 5.72 20 ganz 5.32 5.77 4.46 5.16 5.79 5.14

193




194

Mozart Vibrato Rate Subjects 1-2


MLSO 2

1 den 5.93 5.91 5.57 5.40 absent absent 2 te 5.88 5.86 5.89 5.50 5.23 5.56 3 la 5.92 5.85 5.77 5.75 5.43 absent 4 cal 5.70 5.88 5.99 5.94 5.88 absent 5 ma 5.86 5.77 5.79 5.51 5.64 5.13 6 re 5.96 6.49 5.77 5.72 5.57 5.56 7 sti 6.06 6.09 6.12 5.36 5.19 5.18 8 se 6.28 6.13 5.63 5.90 5.56 5.26 9 gno 5.95 5.82 5.81 5.27 5.19 5.11

10 gno e 5.95 5.68 5.77 5.50 4.96 5.18 11 mor 5.84 5.42 5.59 5.36 5.23 5.11 12 den 6.59 6.97 6.74 6.19 5.83 absent 13 te 5.63 5.73 5.59 5.49 5.34 5.46 14 la 5.91 5.85 5.75 5.89 5.41 5.66 15 cal 5.93 5.66 5.96 5.73 5.52 absent 16 ma 6.16 5.93 5.96 5.67 5.13 5.41 17 al 6.19 6.00 6.71 6.50 absent 5.56 18 ma 5.85 6.02 5.65 5.47 5.13 5.36 19 si 6.06 5.97 5.92 5.83 5.26 5.56 20 des 6.12 6.11 absent 5.46 5.13 5.46 21 sti 5.89 6.05 6.17 5.44 5.03 5.18 22 re 6.31 5.78 6.11 5.72 5.38 5.41 23 sti 5.95 6.07 6.45 5.27 5.19 5.69 24 se 5.91 6.13 6.17 5.72 5.28 5.41 25 gno 5.58 5.80 5.74 5.27 4.84 5.13 26 gno e 5.89 5.56 5.67 5.78 5.17 5.45 27 mor 5.70 5.76 5.70 5.20 4.92 5.10 28 mor 2 5.75 5.58 5.72 5.64 5.27 5.27 29 res 6.29 5.97 5.18 6.01 absent absent 30 sti 5.90 6.00 5.86 5.51 5.56 5.26 31 se 5.98 6.06 6.11 6.02 5.64 5.46 32 gno 5.74 5.69 5.89 5.38 5.60 5.00 33 gno e 5.68 5.92 5.65 5.66 5.69 5.22 34 ti 5.95 absent 6.32 5.64 5.46 5.09 35 mor 5.88 5.68 5.43 5.45 4.89 5.04 36 di 6.37 6.41 7.30 6.02 absent 6.67 37 mor 5.83 5.86 5.68 5.20 5.01 5.16

195



MLSO 4

1 den 5.74 6.17 absent 6.11 absent 6.25 2 te 5.58 5.78 absent 5.96 5.88 6.27 3 la 5.92 6.94 5.57 5.63 5.57 5.57 4 cal 5.68 5.76 absent 6.05 5.83 5.00 5 ma 5.70 5.32 5.26 5.57 5.69 5.48 6 re 5.65 absent absent 6.02 5.26 5.28 7 sti 5.79 5.82 5.79 5.46 5.26 5.41 8 se 5.93 5.39 5.72 5.86 absent 5.56 9 gno 5.45 5.59 5.88 5.58 5.44 5.46

10 gno e 5.57 5.69 5.80 5.81 5.10 5.26 11 mor 5.30 5.55 5.26 5.33 5.16 5.09 12 den 5.97 5.91 6.70 6.28 absent 6.07 13 te 5.37 5.23 5.56 5.85 5.57 5.69 14 la 5.48 5.89 5.57 6.03 5.72 6.27 15 cal 5.63 5.53 5.28 5.60 5.41 6.07 16 ma 5.71 5.56 6.25 5.85 5.57 5.57 17 al 5.82 6.13 6.07 6.21 6.46 6.07 18 ma 5.16 5.26 5.57 5.74 5.41 5.26 19 si 5.33 absent 5.26 5.58 absent 5.36 20 des 5.42 5.24 5.26 5.51 absent absent 21 sti 5.46 5.64 absent 5.13 5.77 5.57 22 re 5.67 5.68 5.79 5.64 5.72 absent 23 sti 5.62 5.94 5.56 5.31 5.62 5.26 24 se 5.58 5.81 5.26 5.63 absent 5.26 25 gno 5.26 6.17 absent 5.41 absent 5.29 26 gno e 5.38 5.28 5.27 5.58 5.90 5.48 27 mor 5.62 5.51 5.27 5.31 5.90 28 mor 2 5.50 5.66 5.26 5.20 5.41 5.26 29 res 5.67 5.64 5.90 5.56 absent 5.00 30 sti 5.75 5.60 5.66 5.25 5.26 5.28 31 se 5.52 absent 5.26 5.63 5.88 absent 32 gno 5.16 5.42 5.26 5.27 absent 5.00 33 gno e 5.37 5.33 5.57 5.69 5.41 6.07 34 ti 5.64 5.56 5.37 absent 5.05 35 mor 5.45 5.43 5.90 5.24 5.23 5.05 36 di 6.31 absent absent 6.17 absent absent 37 mor 5.39 5.17 5.37 5.37 6.25 5.88

196



MLSO 6

1 den 7.23 6.83 6.90 6.45 absent 5.88 2 te 7.04 absent 6.83 6.90 absent absent 3 la 7.04 7.42 6.83 7.08 absent 5.26 4 cal 6.52 absent 6.85 7.24 absent 6.07 5 ma 6.61 6.22 6.97 7.33 absent 6.28 6 re 6.83 6.70 absent 6.75 absent 7.11 7 sti 6.68 6.90 7.06 7.22 absent 7.69 8 se 6.77 6.90 6.46 7.39 absent 7.50 9 gno 6.85 6.83 6.83 8.22 6.70 7.35

10 gno e 6.63 6.63 6.86 7.12 absent 6.62 11 mor 6.55 6.58 6.68 7.38 6.87 7.01 12 den 6.79 6.46 7.69 7.13 absent 6.25 13 te 6.96 6.97 6.70 7.81 7.52 6.48 14 la 6.63 6.95 6.83 6.99 5.96 absent 15 cal 6.76 6.68 6.79 8.14 absent 5.61 16 ma 6.78 6.71 6.69 7.63 5.88 6.97 17 al 6.69 6.27 7.74 6.96 6.67 6.46 18 ma 6.69 7.04 6.67 7.81 absent 6.97 19 si 6.95 6.98 6.97 7.35 absent 6.19 20 des 7.06 6.69 6.67 7.18 absent 6.90 21 sti 6.85 6.79 6.53 7.17 6.67 6.79 22 re 6.46 6.68 6.59 7.12 6.62 23 sti 6.68 6.25 6.70 6.82 6.48 8.28 24 se 6.80 6.46 6.55 7.32 absent absent 25 gno 6.90 7.14 6.67 7.53 absent absent 26 gno e 6.89 6.77 6.80 7.38 6.70 6.43 27 mor 6.87 6.83 6.67 6.72 4.55 6.19 28 mor 2 6.88 6.87 6.69 8.05 6.62 6.81 29 res 6.68 6.27 6.27 7.22 absent absent 30 sti 6.89 6.90 6.67 6.87 absent absent 31 se 7.10 7.42 6.90 7.31 absent absent 32 gno 6.75 6.81 6.46 7.47 absent 7.49 33 gno e 6.67 7.14 7.54 7.59 6.79 7.35 34 ti 6.80 7.01 6.83 7.84 6.25 absent 35 mor 6.55 6.68 6.87 6.65 4.98 4.95 36 di 6.92 6.97 6.97 7.51 absent absent 37 mor 6.69 6.63 6.58 6.22 5.53 6.03

197


198

Appendix L: Long term average spectra raw data

All Schubert Optimal

Schubert Optimal Raw Data Hz SO1 SO2 SO3 SO4 SO5 SO6

0 125 -57.87 -61.8 -60.29 -61.5 -57.66 -60.62 250 -47.65 -54.59 -49.29 -52.61 -48.14 -50.75 375 -41.88 -44.04 -41.14 -42.89 -40.37 -42.46 500 -37.69 -33.25 -35.86 -36.4 -34.06 -34.63 625 -35.39 -27.18 -30.76 -31.86 -33.44 -31.13 750 -37.22 -26.71 -28.49 -29.68 -39.8 -33.55 875 -42.07 -31.44 -31.26 -31.23 -51.57 -43.41

1000 -40.76 -42.22 -39.88 -38.01 -51.79 -52.13 1125 -38.9 -52.82 -45.33 -43.17 -54.08 -51.08 1250 -38.62 -50.22 -44.48 -44.12 -58.46 -52.57 1375 -39.69 -50.38 -44.76 -48.57 -56.82 -55.1 1500 -44.19 -53.71 -44.08 -53.39 -49.52 -51.49 1625 -50.22 -55.45 -44.35 -53.81 -46.29 -47.16 1750 -52.33 -54.4 -47.32 -53.33 -48.31 -45.17 1875 -56.36 -56.13 -52.81 -55.32 -55.02 -46.5 2000 -61.01 -59.3 -59.58 -56.71 -61.97 -51.66 2125 -63.6 -61.15 -59.93 -57.61 -64.26 -59.1 2250 -63.49 -63.86 -57.75 -60.85 -61.74 -64.3 2375 -62.03 -64.17 -56.87 -62.62 -59.77 -64.96 2500 -59.93 -62.67 -56.85 -61.76 -59.86 -63.34 2625 -56.31 -62.21 -57.66 -60.36 -57.66 -60.93 2750 -53.94 -61.46 -54.16 -57.1 -55.22 -61.57 2875 -53.18 -59.03 -50.69 -54.97 -55.4 -62.66 3000 -53.19 -54.14 -49.61 -54.86 -56.25 -59.47 3125 -54.37 -51.5 -50.62 -54.54 -57.19 -55.05 3250 -54.5 -52.18 -53.21 -54.38 -55.8 -51.44 3375 -54.92 -54.62 -55.33 -54.66 -54.56 -51.25 3500 -57.46 -56.6 -56.89 -55 -55.41 -53.98 3625 -61.27 -57.46 -60.02 -57.15 -57.46 -55.99 3750 -65.62 -58.11 -65.62 -60.78 -60.96 -58.13 3875 -71.04 -60.05 -73 -64.87 -64.16 -58.91 4000 -76.56 -62.09 -79.03 -68.58 -65.64 -59.87 4125 -80.66 -65.12 -81.54 -70.29 -65.81 -62.64 4250 -84.75 -69.59 -83.06 -73.1 -67.92 -66.52 4375 -89.38 -73.35 -84.56 -78.21 -74.03 -71.05

199

4500 -92.27 -77.43 -83.12 -82.58 -83.85 -72.78 4625 -90.69 -81.09 -79.24 -83.72 -89.52 -72.73 4750 -89.29 -82.65 -78.49 -81.75 -90.26 -74.44 4875 -90.16 -82.94 -80.32 -78.93 -91.62 -77.97 5000 -93.06 -84.34 -82.36 -78.99 -92.83 -81.62 5125 -94.76 -92.17 -85.46 -82.4 -92.1 -83.84 5250 -93.79 -92.45 -87.25 -86.29 -89.04 -85.11 5375 -92.35 -92.19 -85.87 -88.34 -85.15 -83.22 5500 -92.32 -91.21 -85.15 -86.64 -82.72 -81.77 5625 -93.8 -90.73 -83.22 -84.14 -82.72 -82.32 5750 -94.66 -90.38 -79.97 -81.17 -82.62 -80.9 5875 -94.01 -89.99 -78.12 -78.68 -82.78 -79 6000 -92.16 -89.01 -78.71 -78.34 -85.02 -79.02 6125 -89.68 -87.07 -80.43 -80.65 -87.59 -80.74 6250 -87.72 -85.07 -81.66 -85.52 -88.42 -81.29 6375 -86.54 -83.76 -83.1 -88.64 -88.3 -80.28 6500 -86.14 -83.18 -84.15 -88.4 -87.34 -79.5 6625 -87.09 -82.92 -84.72 -88.72 -86.78 -79.73 6750 -88 -82.29 -85.53 -89.75 -87.52 -82.01 6875 -88.27 -81.68 -86.53 -88.53 -88.88 -84.26 7000 -88.7 -81.71 -86.97 -87.93 -88.93 -83.44

200

All Schubert Sub-optimal

Schubert Sub-optimal Hz SSO1 SSO2 SSO3 SSO4 SSO5 SSO6

0 125 -60.59 -62.52 -61.33 -62.08 -58.3 -61.43 250 -52.06 -60.88 -52.24 -54.37 -48.71 -52.1 375 -46.02 -50.76 -44.17 -45.55 -42.92 -44.26 500 -42.11 -44.3 -38.64 -39.84 -38.91 -37.21 625 -39.71 -39.4 -34.09 -33.74 -38.31 -33.4 750 -40.67 -38.84 -32.39 -30.04 -43.3 -34.89 875 -43.08 -43.82 -35.2 -31.34 -47.97 -42.66

1000 -41.62 -55.87 -43.11 -38.94 -48.25 -49.87 1125 -41.97 -64.74 -47.48 -45.75 -51.96 -49.33 1250 -44.48 -63.71 -47.41 -45.63 -55.74 -51.44 1375 -46.15 -63.3 -47.56 -48.29 -54.12 -53.01 1500 -49.5 -64.3 -45.29 -51.65 -50.34 -52.45 1625 -52.1 -64.42 -45.29 -51.35 -45.66 -50.25 1750 -53.26 -65.58 -49.52 -51.46 -44.03 -49.1 1875 -55.99 -69.27 -52.88 -53.25 -48.17 -50.32 2000 -57.6 -72.43 -56.93 -55.73 -58.15 -54.42 2125 -60.36 -74.78 -60.48 -55.21 -64.7 -59.67 2250 -64.42 -76.65 -61.81 -56.82 -65.83 -62.73 2375 -65.45 -77.54 -62.74 -57.34 -64.37 -63.01 2500 -63.59 -77.69 -65.43 -59.29 -64.56 -64.33 2625 -60.58 -78.44 -62.3 -63.23 -65.31 -63.15 2750 -59.68 -77.97 -60.71 -59.88 -60.63 -62.15 2875 -59.18 -77.33 -58.41 -58.71 -55.79 -62.81 3000 -59.78 -74.8 -55.57 -59.5 -54.2 -61.49 3125 -60.44 -71.56 -56.23 -56.9 -55.29 -57.73 3250 -61.26 -73.12 -58.04 -56.68 -58.08 -54.38 3375 -63.32 -77.12 -58.9 -59.06 -59.17 -52.44 3500 -65.06 -78.15 -60.78 -60.86 -58.18 -53.18 3625 -67.01 -78.7 -61.26 -62.11 -58.15 -55.35 3750 -71.47 -76.74 -64.04 -63.58 -59.7 -57.17 3875 -76.04 -77.34 -68.93 -62.79 -60.56 -58.27 4000 -79.17 -80.32 -71.31 -63.51 -60.28 -58.4 4125 -82.69 -82.62 -74.83 -67.27 -61.59 -59.53 4250 -85.48 -85.96 -79.34 -70.13 -65.67 -63.3 4375 -88.42 -87.33 -83.08 -74.97 -70.7 -68.46 4500 -86.53 -87.83 -85.69 -80.49 -73.31 -71.13 4625 -86.79 -87.62 -85.24 -81.48 -75.08 -71.36 4750 -88.43 -87.89 -84.92 -79.11 -78.24 -72.75

201

4875 -88.2 -91.78 -85.89 -78.47 -82.31 -75.83 5000 -89.84 -98.32 -88.29 -80.24 -83.14 -78.16 5125 -89.94 -101.46 -89.76 -84.39 -82.09 -81.53 5250 -89.14 -102.69 -90.05 -90.1 -82.55 -85.41 5375 -89.75 -102.64 -91.11 -89.81 -83.88 -87.65 5500 -88.78 -102.5 -92.52 -85.69 -83.43 -89.4 5625 -88.43 -103.29 -92.89 -84.9 -82.42 -91.08 5750 -89.63 -103.28 -92.49 -86.6 -83.52 -90.79 5875 -91.21 -102.64 -91.97 -89.18 -85.39 -89.16 6000 -94.39 -102.19 -90.89 -89.99 -86.61 -87.53 6125 -97.13 -101.2 -89.4 -88.56 -87.8 -86.63 6250 -96.25 -100.29 -88.01 -87.05 -89.32 -86.51 6375 -94.22 -98.92 -86.74 -87.78 -92.41 -85.37 6500 -94.21 -98.37 -86.49 -90.32 -94.52 -83.18 6625 -95.7 -97.91 -87.03 -90.9 -94.1 -81.6 6750 -97.04 -97.14 -88.26 -90.71 -93.54 -80.57 6875 -97.47 -96.06 -90.47 -90.94 -93.3 -80.52 7000 -97.83 -96 -92.25 -90.24 -93.12 -82.25

202

All Schubert Loud sub-optimal

Schubert Loud Sub-optimal Hz SLSO1 SLSO2 SLSO3 SLSO4 SLSO5 SLSO6

0 125 -59.42 -62.3 -61.1 -62.02 -58.14 -59.96 250 -49.93 -58.17 -51.83 -53.51 -48.69 -50.08 375 -44.11 -48.42 -43.11 -43.72 -43.53 -42.94 500 -40.5 -40.12 -37.49 -38.23 -40.08 -36.69 625 -38.13 -33.96 -32.92 -33.36 -39.04 -34.05 750 -38.82 -32.8 -30.6 -29.68 -41.36 -36.84 875 -40.7 -36.99 -32.92 -29.72 -42.9 -43.38

1000 -38.57 -48.33 -39.99 -35.53 -44.12 -43.37 1125 -38.58 -60.23 -43.15 -45.72 -47.96 -44.31 1250 -42.06 -56.55 -44.71 -46.18 -51.72 -47.51 1375 -44.63 -53.28 -46.84 -47.12 -51.01 -46.34 1500 -46.43 -51.2 -45.31 -49.55 -48.5 -44.62 1625 -46.53 -51.85 -44.69 -50.17 -43.92 -45.57 1750 -47.57 -53.76 -48.28 -49.61 -41.93 -42.82 1875 -49.95 -55.24 -52.34 -49.9 -45.69 -42.52 2000 -50.73 -59.38 -53.66 -50.32 -55.5 -47.59 2125 -53.09 -62.94 -55 -50.15 -61.95 -53.9 2250 -57.61 -64.31 -56.33 -52.77 -63.23 -56.66 2375 -59.37 -65.47 -56.57 -53.77 -63.78 -57.7 2500 -58.36 -66.09 -59.75 -54.19 -64.4 -58.69 2625 -54.55 -64.76 -59.04 -56.73 -63.73 -58.36 2750 -52.95 -64.53 -55.73 -59.17 -58.54 -58.06 2875 -52.45 -63.54 -52.53 -59.09 -54.24 -56.03 3000 -53.26 -61.6 -49.37 -58.81 -52.71 -53.17 3125 -53.86 -59.95 -48.71 -55.88 -53.15 -50.01 3250 -54 -59.5 -50.44 -54.49 -54.27 -47.67 3375 -55.64 -62.11 -53.13 -54.28 -54.75 -47.4 3500 -57.82 -62.82 -55.07 -54.96 -55.39 -47.76 3625 -60.74 -61.1 -56.39 -57.05 -56.7 -48.29 3750 -65.13 -58.39 -59.09 -57.93 -58.4 -49.82 3875 -69.33 -56.94 -65.26 -56.87 -58.2 -51.75 4000 -72.92 -58.87 -73.28 -56.95 -57.64 -53.58 4125 -76.92 -62 -76.5 -58.83 -58.99 -55.35 4250 -80.63 -65.26 -79.15 -60.63 -62.88 -58.3 4375 -86.44 -69.77 -82.72 -63.25 -67.55 -63.18 4500 -90.37 -71.51 -85.43 -67.88 -69.47 -67.45 4625 -87.79 -73.08 -83.78 -69.61 -70.97 -68.42 4750 -85.79 -76.53 -80.97 -68.72 -74.75 -70.63

203

4875 -85.1 -79.39 -80.24 -67.86 -80.34 -75 5000 -85.76 -83.53 -83.19 -68.27 -80.74 -78.8 5125 -84.39 -87.61 -88.88 -71.45 -78.52 -82.07 5250 -82.24 -87.38 -91.35 -77.77 -78.12 -84.64 5375 -83.12 -88.53 -90.1 -82.66 -78.86 -85.71 5500 -83.87 -89.45 -89.13 -80.48 -78.28 -87.08 5625 -84.26 -90.71 -89.12 -78.57 -77.26 -86.98 5750 -86.08 -91.79 -89.61 -80.11 -78.93 -83.88 5875 -86.82 -92.12 -88.28 -83.16 -83.74 -81.25 6000 -89.07 -92.12 -86.05 -84.51 -88.27 -80.33 6125 -92.79 -90.63 -84.14 -83.56 -89.9 -79.96 6250 -93.21 -89.64 -82.68 -82.2 -91.11 -79.26 6375 -91.13 -88.55 -81.78 -82.32 -92.94 -78.18 6500 -90.33 -87.52 -81.98 -84.45 -93.44 -77.9 6625 -90.9 -85.05 -81.96 -88.07 -92.8 -78.07 6750 -91.38 -82.55 -82.84 -89.25 -92.03 -78.36 6875 -90.67 -82.33 -85.72 -88 -91.99 -78.71 7000 -89.95 -83.88 -88.79 -86.49 -92.3 -78.76

204

All Mozart Optimal

Mozart Optimal Raw Data Hz MO1 MO2 MO3 MO4 MO5 MO6

0 125 -58.74 -59.97 -58.89 -58.95 -58.66 -60.4 250 -46.99 -48.98 -46.47 -47.97 -46.2 -49.58 375 -38.61 -40.74 -38.47 -39.6 -37.71 -41.41 500 -34.59 -36.86 -34.82 -35.39 -33.8 -37.58 625 -33.25 -33.2 -32.73 -33.73 -31.93 -35.89 750 -31.62 -31.67 -30.81 -32.54 -31.15 -35.26 875 -32.49 -35.56 -32.94 -34.18 -35.17 -38.98

1000 -37.86 -44.28 -39.06 -39.9 -43.23 -47.66 1125 -42.05 -49.43 -42.54 -45.76 -46.22 -52.85 1250 -44.83 -56.08 -46.71 -51.97 -52.2 -53.32 1375 -44.7 -54.63 -45.31 -55 -47.96 -49.34 1500 -41.65 -52.28 -42.05 -52.53 -43.68 -47.28 1625 -40.86 -51.56 -43.03 -51.42 -43.19 -48.4 1750 -44.64 -51.19 -47.38 -53.03 -45.34 -50.08 1875 -51.26 -51.75 -50.59 -56.83 -50.03 -51.92 2000 -55.15 -53.3 -53.51 -61.34 -54.98 -54.25 2125 -57.66 -55.49 -54.09 -63.76 -56.04 -56.14 2250 -59.15 -57.35 -54.32 -62.67 -56.77 -58.15 2375 -56.44 -60.03 -54.5 -61.02 -59.42 -62.94 2500 -54.17 -61.48 -54.75 -61.93 -60.11 -64.48 2625 -53.03 -61.89 -51.96 -62.29 -57.09 -64.5 2750 -51.96 -61.6 -48.1 -59.89 -54.59 -65.66 2875 -52.1 -59.68 -45.7 -56.97 -52.35 -62.38 3000 -52.14 -57.6 -44.89 -55.06 -51.39 -58.67 3125 -51.81 -56.11 -45.41 -54.53 -52.16 -57.56 3250 -51.94 -57.01 -47.09 -55.12 -54.9 -58.03 3375 -53.26 -58.09 -49.56 -55.84 -55.87 -57.64 3500 -56.25 -58.54 -51.76 -56.57 -54.8 -57.5 3625 -58.99 -60.09 -54.06 -58 -55.34 -58.1 3750 -62.18 -62.32 -58.44 -60.84 -57.29 -58.72 3875 -67.61 -63.78 -64.46 -64.09 -60.01 -59.56 4000 -72.79 -65.07 -69.47 -65.5 -63.46 -61.25 4125 -75.4 -67.35 -74.83 -66.18 -67.76 -63.26 4250 -78.74 -70.61 -79.15 -68.08 -71 -67.03 4375 -84.42 -73.83 -78.55 -72.55 -73.67 -70.67 4500 -88.46 -76.5 -77.29 -77.07 -74.27 -70.52 4625 -82.98 -80.42 -76.59 -79.02 -73.67 -71.33 4750 -79.29 -86.09 -75.68 -79.72 -75.68 -74.17

205

4875 -81.61 -90.87 -75.58 -79.23 -81.87 -76.91 5000 -87.97 -92.56 -77.84 -80.14 -86.99 -78.03 5125 -91.19 -92.25 -81.63 -83.43 -86.11 -79.53 5250 -91.58 -92.54 -84.18 -88.14 -84.72 -83.33 5375 -86.52 -92.2 -83.39 -88.87 -83.43 -88.17 5500 -81.52 -91.24 -81.72 -87.31 -83.01 -90.03 5625 -81.67 -90.72 -81.27 -85.14 -82.3 -89.7 5750 -86.66 -90.25 -81.26 -83.08 -81.22 -89.06 5875 -90.85 -89.91 -78.08 -81.72 -80.94 -87.9 6000 -90.9 -88.99 -73.96 -82.12 -80.5 -86.68 6125 -88.68 -87.08 -72.95 -83.69 -78.86 -85.71 6250 -84.24 -85.1 -74.29 -85.78 -78.21 -85.04 6375 -82.5 -83.75 -76.32 -86.94 -80.82 -84.9 6500 -84.38 -83.21 -78.89 -87.53 -83.62 -85.55 6625 -85.82 -82.95 -79.21 -88.79 -83.39 -86.04 6750 -85.59 -82.3 -77.33 -90.15 -82.99 -85.53 6875 -85.53 -81.72 -78.6 -90.17 -82.62 -85.16 7000 -85.26 -81.76 -82.11 -89.05 -83.05 -85.31

206

All Mozart Sub-optimal

Mozart Sub-optimal Hz MSO1 MSO 2 MSO3 MSO4 MSO5 MSO6

0 125 -60.12 -60.72 -60.08 -60.64 -60.47 -61.82 250 -49.82 -51.49 -49.47 -49.89 -49.86 -52.68 375 -42.63 -43.71 -41.72 -42.54 -42.43 -43.52 500 -40.01 -40.42 -38.54 -39.74 -39.83 -39.19 625 -39.25 -39.68 -37.56 -37.68 -39.38 -37.61 750 -38.13 -39.25 -37.12 -35.54 -39.09 -36.03 875 -39.09 -42.78 -39.79 -37.29 -42.07 -37.97

1000 -40.93 -51.19 -44.97 -40.79 -46.57 -44.84 1125 -44.83 -56.82 -49.08 -44.19 -50.79 -50.22 1250 -48.32 -60.96 -54.59 -50.87 -56.32 -54.84 1375 -47.13 -54.52 -53.5 -54.84 -52.8 -53.56 1500 -45.48 -51.81 -48.31 -53.22 -49.11 -48.83 1625 -47.35 -54.32 -47.89 -53.36 -48.83 -47.94 1750 -51.07 -56.43 -51.47 -54.08 -49.53 -51.36 1875 -54.63 -56.71 -55.22 -53.74 -52.29 -54.28 2000 -54.29 -57.13 -58.68 -52.53 -54.67 -55.79 2125 -54.99 -58.81 -59.56 -53.59 -56.57 -53.21 2250 -56.42 -61.25 -61.04 -55.81 -57.72 -52.69 2375 -59.91 -63.88 -62.75 -57.86 -61.15 -55.36 2500 -61.4 -63.92 -64.49 -61.43 -65.25 -60.16 2625 -60.14 -66.34 -62.34 -62.78 -65.51 -59.89 2750 -58.08 -67.9 -60.55 -64.08 -63.73 -60.49 2875 -54.63 -66.58 -57.92 -62.13 -60 -62.18 3000 -53.03 -64.72 -55.01 -58.37 -57.33 -56.57 3125 -55.55 -64.75 -54.47 -57.55 -58.01 -51.84 3250 -59.64 -66.95 -57.75 -60.68 -61.38 -52.2 3375 -60.24 -67.16 -61.54 -62.75 -60.75 -56.44 3500 -60.41 -66.44 -61.79 -61.76 -59.66 -57.27 3625 -61.63 -67.53 -62.98 -62.85 -59.92 -56.45 3750 -63.68 -68.5 -65.86 -64.11 -59.98 -58.12 3875 -67.42 -67.42 -69.23 -62.86 -59.82 -59.63 4000 -72.86 -67.44 -72.44 -62.46 -60.37 -59.59 4125 -78.96 -69.3 -77.42 -63.48 -62.65 -60.8 4250 -86.17 -72.23 -82.38 -66.45 -66.14 -63.32 4375 -87 -74.83 -84.89 -68.39 -67.78 -69.04 4500 -86.02 -76.56 -85.29 -70.47 -68.4 -73.22 4625 -88.69 -80.14 -83.63 -71.4 -70.31 -70.98 4750 -94.18 -87.34 -82.97 -73.21 -74.83 -70.3

207

4875 -94.6 -94.38 -83.98 -74.97 -80.1 -73.98 5000 -88.07 -96.83 -86.79 -74.14 -82.17 -79.03 5125 -83.11 -96.92 -87.93 -76.07 -81.99 -81.43 5250 -82.44 -96.44 -87.5 -79.26 -82.42 -82.61 5375 -85.43 -97.44 -89.05 -81.1 -83.46 -85.01 5500 -90.54 -98.08 -90.78 -82.53 -84.29 -87.3 5625 -93.1 -97.85 -91.06 -86.67 -84.73 -90.03 5750 -92.56 -96.77 -90.5 -89.9 -85.25 -89.99 5875 -90.19 -94.92 -90.18 -90.08 -85.44 -88.48 6000 -89.99 -93.54 -88.85 -86.97 -84.02 -88.02 6125 -92.37 -92.36 -86.73 -84.82 -83.58 -86.85 6250 -94 -91.26 -85.49 -85 -85.74 -84.49 6375 -93.84 -90.68 -84.6 -85.91 -89.01 -82.31 6500 -93.63 -89.31 -84.78 -87.68 -90.77 -80.99 6625 -92.83 -87.42 -86.54 -89.42 -91.21 -81.83 6750 -92.42 -86.8 -86.41 -89.9 -91.4 -83.68 6875 -92.63 -87.47 -83.74 -89.36 -91.12 -83.39 7000 -93.77 -88.69 -84.53 -88.14 -90.98 -81.49

208

All Mozart Loud sub-optimal

Mozart Loud Sub-optimal Hz MLSO1 MLSO2 MLSO3 MLSO4 MLSO5 MLSO6

0 125 -59.45 -60.04 -59.88 -60.41 -59.86 -61.12 250 -48.77 -48.99 -48.42 -49.19 -48.8 -50.06 375 -41.57 -40.89 -40.54 -41.87 -41.26 -40.64 500 -38.99 -37.67 -37.38 -39.63 -38.49 -35.82 625 -38.53 -34.65 -35.74 -38.54 -37.16 -33.42 750 -37.48 -32.26 -33.99 -36.86 -35.75 -31.56 875 -37.71 -35.76 -36.07 -38.04 -37.64 -33.37

1000 -38.73 -44.13 -40.88 -39.92 -40.62 -39.91 1125 -42.71 -48.9 -43.89 -43.19 -44.79 -45.63 1250 -47.48 -56.06 -48.92 -50.26 -51.27 -47.64 1375 -45.56 -50.86 -49.99 -52.95 -47.22 -44.17 1500 -43.39 -47.08 -48.11 -50.86 -43.46 -41.53 1625 -45.53 -49.04 -47.99 -51.22 -43.74 -42.16 1750 -49.29 -51.8 -51.49 -52.82 -45.23 -44.31 1875 -50.7 -49.34 -54.38 -52.78 -47.99 -45.43 2000 -50.16 -48.42 -56.36 -51.12 -50.09 -47.49 2125 -51.3 -50.52 -56.68 -52.18 -51.67 -47.02 2250 -52.08 -52.09 -55.82 -53.04 -52.56 -46.21 2375 -55.47 -55.06 -56.21 -53.56 -55.55 -48.88 2500 -59.11 -55.54 -58.48 -57.3 -57.5 -53.04 2625 -57.92 -56.57 -56.91 -58.67 -57.02 -53.95 2750 -55.85 -59.08 -55.23 -60.59 -57.47 -54.29 2875 -52.15 -55.37 -53.04 -60.73 -53.68 -53.38 3000 -50.07 -52.53 -50.55 -57.27 -50.21 -49.7 3125 -52.3 -54.42 -49.07 -55.81 -50.85 -46.72 3250 -57.12 -57.28 -51.78 -57.61 -54.09 -46.44 3375 -57.86 -56.64 -56.56 -58.41 -53.84 -48.33 3500 -58.08 -57.01 -56.61 -57.2 -53.01 -48.7 3625 -59.34 -58.44 -57.78 -58.18 -53.63 -49.61 3750 -61.67 -59.01 -62.26 -60 -53.92 -51.73 3875 -64.9 -56.88 -69.03 -59.47 -54.01 -52.58 4000 -69.3 -55.98 -73.06 -59.59 -54.22 -52.86 4125 -75.06 -58.48 -76.14 -60.6 -55.48 -54.44 4250 -82.65 -61.87 -78.5 -63.52 -58.86 -57.73 4375 -87.36 -61.38 -81.17 -66.81 -61.25 -61.22 4500 -86.83 -61.99 -82.38 -66.1 -60.92 -62.83 4625 -88.34 -66.17 -80.41 -63.99 -62.57 -62.41 4750 -91.85 -74.4 -79.13 -65.62 -68.13 -63.37

209

4875 -89.47 -82.87 -80.16 -71.03 -74.98 -66.74 5000 -85.5 -86.88 -82.41 -74.14 -77.59 -72.18 5125 -82.25 -84.2 -85.47 -74 -77.52 -76.48 5250 -79.51 -80.56 -87.05 -72.4 -77.98 -78.33 5375 -79.71 -83.34 -88.67 -70.56 -76.9 -80.05 5500 -83.76 -90.45 -88.3 -71.65 -74.15 -79.91 5625 -87.58 -91.65 -87.53 -76.61 -73.62 -79.9 5750 -86.51 -91.29 -86.66 -83.95 -75.69 -80.21 5875 -85.42 -85.28 -85.96 -85.28 -78.31 -80.84 6000 -85.2 -81.52 -84.61 -81.09 -78.68 -80.02 6125 -85.78 -83.04 -81.62 -79.11 -79.23 -79.29 6250 -88.81 -85.65 -79.61 -80.01 -81.54 -79.38 6375 -90.58 -84.48 -79.67 -82.55 -84.89 -78.75 6500 -89.73 -81.79 -81.06 -85.43 -87.65 -78.84 6625 -88.37 -78.94 -82.67 -87.59 -89 -79.67 6750 -87.26 -77.57 -83.92 -87.48 -88.88 -79.57 6875 -86.84 -78.55 -84.8 -86.03 -88.78 -78.07 7000 -87.71 -80.31 -84.82 -85.98 -88.8 -77.09

210

Appendix M: Music and Gesture Abstract

Evaluation of ‘open throat’ technique on sound production in classical singing

Helen F. Mitchell, Dianna T. Kenny, Pamela J. Davis and Maree Ryan

‘Open throat’ is thought to involve the maximizing of the pharyngeal space or abduction of the false

vocal folds (Miller 1996; Estill 1996). Mitchell, Kenny, Ryan & Davis (in press) recently reported

consensus among expert pedagogues that the technique of ‘open throat’ was fundamental to good

vocal pedagogy and the production of a beautiful sound, especially in classical singing. The sound

quality produced by this technique was variously described by these pedagogues as free, open,

balanced, coordinated and warm. In addition, open throat was thought to allow the singer to control

dynamic range more effectively than when the technique was not applied. This paper assesses the

acoustic qualities of open throat on the classical singing voice.

Six experienced female singers, deemed by an expert pedagogue to have mastered the technique of

‘open throat’, participated in the study. Singers were asked to perform the same arias under three

different conditions: (1) optimal technique, including maximum ‘open throat;’ (2) good technique

but with less open throat; (3) same as (2) but with the additional instruction to sing as loudly as in

condition 1. Measurements taken from spectrograms indicated that the singers’ vibrato parameters,

notably extent and onset, were affected by the experimental conditions and the decreased use of this

technique. Long-term average spectra plots confirmed timbral differences in centre frequency and

energy level of spectral peaks above 2 kHz.

The absence of open throat produced marked variability in sound quality on the acoustic measures.

This study therefore concluded that ‘open throat’ assisted in the production of a reliable and

consistent sound in classical singing. Further research will confirm its importance for male

classical singers and for singers in other genres.

211

References

Estill, J. (1996). Primer of Basic Figures. Santa Rosa, Estill Voice Training Systems.

Miller, R. (1996). The structure of singing: system and art in vocal technique. New York: London,

Schirmer Books.

212

APPENDICES FOR PAPERS 4 AND 5

213

Appendix N: Project 2 Pedagogue Information Sheet

Expert Perceptual Analysis of Singing Information Sheet for Pedagogue Subjects Page 1 of 1


SYDNEY CONSERVATORIUM OF MUSIC

AUSTRALIAN CENTRE FOR APPLIED RESEARCH IN MUSIC

PERFORMANCE

RESEARCH STUDY INTO EXPERT PERCEPTUAL ANALYSIS OF SINGING

PEDAGOGUE SUBJECT INFORMATION SHEET

You are invited to take part in a research study into the expert perceptual analysis of singing. The study is being conducted by Helen Mitchell, a postgraduate student at The University of Sydney, Associate Professor Pamela Davis, Associate Professor Dianna Kenny and Ms Maree Ryan (Lecturer in Singing, The Conservatorium of Music) and will form the basis of Helen Mitchell’s PhD at The University of Sydney.

The object is to discover associations between different terminology and sound qualities that you can hear in singing. We would use the information you give to present more useful information on how pedagogues interpret sound qualities into verbal descriptions. If you agree to take part in this study, you will be asked to complete a short questionnaire that tells us a bit about you (your age, sex, education, occupation and singing experience) and listen and give feedback on a series of singing examples, rating them for sound quality and excellence. To do this, you would be asked to attend a single session lasting less than an hour, at a time and location convenient to you to do this. At this perceptual analysis session, you will hear a number of examples of singing, and will be asked to rate them for a range of different types of sound qualities.

All aspects of the study, including results, will be strictly confidential and only investigator Helen Mitchell will have access to information on participants. Results will be coded with a number and not your name and the data will not enable readers of any of the output to link specific data with voice type, age, or other features which may enable identification of individuals. A report of the study may be submitted for publication, but individual participants will not be identifiable in such a report. There are no risks involved in the participation of this project. There will be a debriefing session on completion of all data collection. Participation in this study is entirely voluntary: you are not obliged to participate and - if you do participate - you can withdraw at any time. Whatever your decision, it will not affect your relationship with any of the researchers, or with the Australian Centre for Applied Research in Music Performance. The ethics committee of The University of Sydney has approved this study. If you have a concern or complaint about the conduct of a research study, you can contact the Manager for Ethics and Biosafety Administration, University of Sydney on (02) 9351 4811. When you have read this information, Helen Mitchell will discuss it with you further and answer any questions you may have. If you would like to know more at any stage, please feel free to contact Helen Mitchell on 0421 632 713 or [email protected].

214

Appendix O: Project 2 Pedagogue Consent Sheet Expert Perceptual Analysis of Singing Consent Sheet for Pedagogue Subjects Page 1 of 1




PERFORMANCE

CONSENT SHEET

RESEARCH STUDIES

Project Title: Expert Perceptual Analysis of Singing Investigators: Associate Professor Dianna Kenny Ms Helen Mitchell 1. I acknowledge that I have read the above statement which explains the nature, object and the possible risks of the investigation, and the statement has been explained to me to my satisfaction. Before signing this document I have been given the opportunity to ask questions relating to any possible physical harm I may suffer as a result of my participation and I have received satisfactory answers. I have also been informed that I may not receive any benefits from participating in this study. I have been offered a debriefing session after the research has been completed. 2. My decision whether or not to participate will not prejudice my future relations with the Australian Centre for Applied Research in Music Performance or any of the investigators listed above. If I decide to participate, I am free to withdraw my consent and to discontinue participation at any time without prejudice. 3. I agree that research data gathered from the results of the study may be published provided my name is not used. __________ _________________________________ DATE Signature of Subject 4. I have fully explained to the subject _______________________________ the nature and purpose of the programme and the procedures to be employed as described above and such risks as are involved in their performance. __________ _________________________________ DATE Signature of Responsible Investigator

215

Appendix P: Project 2 Pedagogue Questionnaire

Expert Perceptual Analysis of Singing Questionnaire for Pedagogue Subjects Page 1 of 3

Subject ID_____________




PERFORMANCE

Expert Perceptual Analysis of Singing

PEDAGOGUE SUBJECT DEMOGRAPHIC QUESTIONNAIRE

We are interested in finding out about you. Please complete the following questionnaire by ticking the box that best describes you or writing in the space provided. When you have finished, please hand the questionnaire back to one of the researchers.

The questions begin on the next page.

216


Subject ID_____________ 10. How old are you?

_____________years 11. What sex are you?

□ Female

□ Male 12. How many years have you been teaching singing?

_____________years 13. What is the highest level of singing education you have completed?

□ Postgraduate degree at University or Conservatorium

□ Bachelor degree at University or Conservatorium

□ Diploma in Music or Singing

□ Higher School Certificate or equivalent

14. What proportion of your singing studio is private and what proportion is at a University

or Conservatorium?

(a)_____________% private studio (b)_____________% university or conservatorium

Total of (a) and (b) = 100%

217


Subject ID_____________ 15. What types of singing do your students perform in public? Please circle the relevant

singing type and write in the space the percentage of your total studio.

Type of Singing % of performance time Opera

Other classical (eg. Song and concert repertoire, church music)

Choral singing

Musical Theatre

Contemporary styles

Other (please specify)

16. Of the major group of your students, what proportion of them perform at the following

levels? What types of singing do your students perform in public? Please circle the relevant singing type and write in the space the percentage of your total performance time spent performing each type of singing.

Level of Singing % of studio Superstar

International

National

Big City

Regional/Touring

Local Community

Full-Time Voice student

Amateur

Total 100%

Thank you for completing this questionnaire

218

Appendix Q: Project 2 Pedagogue Listener Response Sheet



Australian Centre for Applied Research in Music Performance

Expert Perceptual Analysis of Singing

Response Sheet

219

Subject ID_____________

Background to the Project

Expert pedagogues have defined “open throat” as a necessary component of classical singing technique. It was achieved by a widening of the pharynx or ‘retraction’ of the false vocal folds and elicited a sound quality described as, balanced and coordinated, free, open, even and consistent and warm (Mitchell, Kenny et al., in press).

In the study in which you have been invited to be a rater, sopranos and mezzo-sopranos, studying singing at a tertiary institution sang messe di voce on three notes in their range. They also sang excerpts from Ridente la calma by Mozart and Du bist die Ruh by Schubert. They were asked to sing in two different ways, which we have called Optimal and Technical.

These were defined as:

Optimal: the singer’s best singing sound, with maximum use of openness possible.

Technical: Use of an acceptable singing technique in all respects except that the singer has been instructed to use the minimum degree of openness possible.

We ask you to rate each track you hear to indicate which condition you think singers are using. Not all singers will perform each condition, and some conditions may be repeated. Excerpts are presented in a randomised order.

On the first CD, there are 24 messe di voce pairs

On the second CD, there are 30 song excerpts.

Instructions on rating are on the next page.

220

Subject ID_____________

Instructions PART 1: MESSA DI VOCE PAIRS

For each track, the same singer will sing twice. Please rate the sound quality for each of the two

samples as either optimal (O) or technical (T). There are four possible combinations for the messa

di voce pairs. If you think that singer sang with open throat on both occasions, you would complete

the boxes as in Example 1. If the singer sang technical on the first occasion and used maximum

open throat on the second, complete the boxes as given in Example 3 and so on.

Example 1 Example 3

Sample 1 Sample 2 Sample 1 Sample 2

O O T O

Example 2 Example 4


O T T T

We will first do a practice session using four singer pairs to familiarise you with the perceptual

discriminations required in this study. The investigator will play the four samples

PRACTICE MESSA DI VOCE PAIRS:

SINGER 1 SINGER 3


SINGER 2 SINGER 4


221

Messe di voce Pairs

Subject ID_____________ SINGER 1 SINGER 9 SINGER 17 Sample 1 Sample 2 Sample 1 Sample 2 Sample 1 Sample 2

SINGER 2 SINGER 10 SINGER 18 Sample 1 Sample 2 Sample 1 Sample 2 Sample 1 Sample 2







222

Singer Example Subject ID_____________

PART 2: SONG SAMPLES You will be required to make the same discriminations in this section as for the Messa di Voce task

except that you will be rating single excerpts from classical songs. You will hear 30 song tracks by

different singers, with a portion of a song, lasting about 20 seconds.

1. Please indicate which condition you think the singer in the excerpt is singing: Optimal or

Technical.

2. Make brief comments or recommendations in point form about the sound quality

3. Rate the singing in this sample out of 10.

Example Excerpt In the song track I heard, the singer: demonstrated ‘open throat’ demonstrated minimal ‘open throat’ OPTIMAL TECHNICAL Comments or Recommendations about the sound quality

•

•

•

Overall Rating: Circle a mark out of 10. 1 2 3 4 5 6 7 8 9 10

223

Singer X

Subject ID_____________ In the song track I heard, the singer:

demonstrated ‘open throat’ demonstrated minimal ‘open throat’ OPTIMAL TECHNICAL Comments or Recommendations

•

•

•

Overall Rating: Circle a mark out of 10. 1 2 3 4 5 6 7 8 9 10

224

Appendix R: Pedagogue Listener Debrief Document

Expert Perceptual Analysis of Singing Page 1 of 1




PERFORMANCE

EXPERT PEDAGOGUE DEBRIEF DOCUMENT RESEARCH STUDY INTO THE USE OF SUPRALARYNGEAL AND PHARYNGEAL WIDENING

(OPEN THROAT) IN SINGING Thank you for participating in this study at the Australian Centre for Applied Research in Music Performance – your input as a pedagogue is very important to the expansion and continuation of knowledge in this field.

In this study, the singers you have listened to were asked to sing in two different ways: the first with the maximum degree of openness they would use in their technique, and the second with a smaller degree of openness, but still with an acceptable singing technique. Both these conditions will be around the same volume, so that the test is of the difference in sound quality.

The aim of the study was to look at the use of open throat technique in singing, and this project was to discover if these instructions achieve a perceivable and intrinsic difference to the sound quality produced.

The words in the evaluation you have just completed were the result of an earlier study, where our sample of pedagogues gave a number of words that they associated with the use of open throat in singing. Words like freedom, consistency, warmth and openness were directly linked to the technique.

The more general terms were to ascertain if the overall sound was heard was better with or without use of the technique.

This study is not a test for pedagogues, and there are no correct or incorrect responses. We value your opinions and insights you have given into the examples played.

Thank you once again for taking part in this study.

On behalf of the research team,

Associate Professor Dianna Kenny

Ms Helen Mitchell

225


226

Appendix S: Perceptual judgments: messa di voce

Messa di voce perceptual responses Listeners 1-2

[Response code: 1=Optimal, 2=Sub-optimal]

Messa di voce LISTENER 1 LISTENER 2 sample sample

SAMPLE Subject singing 1st 2nd Answer1 Answer2 Answer1 Answer2

1 6 2 2 2 2 2 22 1 1 1 1 1 1 13 1 2 1 2 1 2 14 1 2 2 2 2 2 25 3 1 1 2 2 1 16 5 1 2 1 1 1 27 2 1 2 2 2 1 28 6 2 1 2 2 2 19 4 2 2 2 2 2 2

10 3 1 2 1 1 1 211 6 1 1 2 2 1 112 4 1 2 2 2 1 213 5 2 1 1 1 1 114 5 2 2 1 1 1 115 2 2 1 2 2 2 116 2 1 1 1 1 1 117 3 2 2 2 1 2 218 3 2 1 1 2 2 119 4 1 1 1 1 1 120 6 1 2 1 2 1 221 1 1 2 1 2 1 222 5 1 1 1 1 1 123 2 2 2 2 2 2 224 4 2 1 2 2 2 1

227


LISTENER 3 LISTENER 4 LISTENER 5 LISTENER 6 Answer1 Answer2 Answer1 Answer2 Answer1 Answer2 Answer1 Answer2

2 2 2 2 2 2 1 11 1 1 1 1 1 2 12 1 2 1 2 1 2 12 2 2 2 2 2 2 21 1 1 1 1 1 1 12 2 1 2 1 2 1 21 2 1 2 1 2 1 22 1 2 2 2 1 2 12 2 2 2 2 2 2 21 2 1 2 1 2 1 21 1 1 1 1 1 2 11 2 1 2 1 2 1 22 1 2 1 1 1 1 12 2 2 2 1 1 2 12 1 2 1 2 1 2 11 1 1 1 2 2 1 22 2 2 2 2 1 2 12 1 2 1 2 1 2 11 1 1 1 1 1 1 11 2 2 2 2 1 1 21 2 1 2 1 2 1 21 1 1 1 1 2 1 12 2 2 2 2 2 2 22 1 2 1 2 1 2 1

228



2 2 1 1 2 2 2 21 1 1 1 1 2 1 12 1 2 1 2 1 2 12 2 2 1 2 2 2 22 2 1 1 2 2 2 21 2 1 2 1 2 1 21 2 1 2 1 2 1 22 2 2 1 2 2 2 22 2 1 1 2 2 2 21 2 2 1 1 2 1 21 1 1 1 2 2 1 21 1 1 2 1 2 1 22 1 2 1 2 1 2 12 2 2 2 2 2 2 22 1 2 1 2 1 2 11 1 1 1 1 1 2 12 2 2 1 2 2 2 22 1 1 2 2 2 2 11 1 2 1 1 1 2 12 2 1 2 2 2 1 21 2 1 2 1 2 1 21 1 2 1 2 2 2 12 2 2 1 2 2 2 22 1 1 2 2 1 2 1

229



2 2 2 2 2 2 2 21 1 1 1 1 1 1 12 1 2 1 2 1 2 12 2 2 2 2 2 2 21 1 1 1 1 1 2 21 2 2 2 1 2 1 21 2 1 2 1 2 1 22 1 2 1 2 1 2 12 2 2 2 2 2 2 21 2 1 2 1 2 2 11 1 1 1 1 1 2 21 2 1 2 1 2 1 22 1 2 2 2 2 2 12 2 2 2 2 1 2 22 1 2 1 2 1 2 12 1 1 1 1 1 2 22 2 2 2 2 2 2 22 1 2 1 2 1 2 21 1 1 1 1 1 1 21 2 1 2 1 2 2 21 2 1 2 1 2 1 21 1 2 2 1 1 2 22 2 2 2 2 2 2 22 1 2 1 2 1 2 1

230

Messa di voce perceptual responses Listener 15

LISTENER 15 Answer1 Answer2

2 21 22 12 21 11 21 22 12 21 22 21 22 12 22 11 12 22 11 11 21 21 22 22 1

231


232

Appendix T: Perceptual judgments: song samples

Song sample perceptual responses Listeners 1-2

[Response code: O=Optimal, SO=Sub-optimal]

Song samples Listener 1 Listener 2 SAMPLE SINGER TASK CONDITION Response Rating Response Rating

1 6 Mozart SO SO 2 SO 1 2 1 Schubert O O 6 O 9 3 1 Mozart O O 9 O 8 4 5 Mozart O SO 6 O 9 5 6 Schubert SO O 5 SO 2 6 1 Mozart SO SO 4 SO 3 7 5 Schubert SO O 5 O 7 8 2 Mozart SO SO 4 SO 2 9 3 Schubert O SO 4 O 9

10 4 Mozart SO SO 3 SO 3 11 3 Mozart O O 5 O 9 12 4 Schubert SO O 5 SO 1 13 5 Schubert O O 9 O 10 14 5 Mozart SO SO 5 SO 1 15 2 Mozart O O 7 O 7 16 3 Schubert SO SO 5 SO 1 17 6 Schubert O O 7 O 5 18 2 Schubert O O 7 O 9 19 3 Mozart SO SO 4 SO 2 20 1 Schubert SO SO 3 SO 2 21 4 Schubert O SO 5 O 8 22 6 Mozart O SO 5 SO 3 23 2 Schubert SO O 6 SO 3 24 4 Mozart O O 7 O 8 25 4 Schubert O O 7 O 8 26 4 Schubert SO SO 4 SO 2 27 5 Mozart O O 9 O 9 28 6 Schubert SO O 5 SO 1 29 2 Mozart O O 5 O 7 30 2 Mozart SO SO 4 SO 1

233


Listener 3 Listener 4 Listener 5 Listener 6 Response Rating Response Rating Response Rating Response Rating

SO 2 SO 3 SO 4 SO 4 O 5 O 8 O 7 O 7 O 7 O 8 O 6 O 7 O 7 O 6 SO 5 O 8

SO 2 SO 4 SO 5 SO 3 O 6 SO 5 SO 5 O 7 O 5 SO 6 SO 5 O 6

SO 3 SO 3 SO 5 SO 5 O 6 O 7 SO 6 O 7 O 3 SO 4 O 6 SO 5 O 8 O 9 O 7 O 8

SO 2 SO 2 O 5 SO 4 O 8 O 9 O 6 O 8

SO 4 SO 5 O 5 SO 6 O 6 O 5 O 7 O 7

SO 3 SO 3 SO 6 SO 4 O 3 SO 4 SO 6 O 7 O 7 O 7 O 8 O 7

SO 3 SO 3 SO 5 SO 5 SO 2 SO 2 SO 6 SO 6 O 7 O 9 O 7 O 8

SO 3 SO 4 SO 6 O 7 SO 4 O 6 O 7 SO 4 O 6 O 9 O 6 O 8 O 9 O 9 O 7 O 8

SO 2 SO 2 SO 5 SO 4 O 5 O 9 O 8 O 8

SO 3 SO 4 SO 6 SO 4 O 3 O 7 O 7 O 7

SO 3 SO 3 SO 5 SO 4

234



SO 2 SO 5 SO 3 SO 3 O 4 O 8 O 6 O 7 O 5 O 7 O 5 O 7 O 5 O 8 O 5 O 7

SO 3 SO 5 SO 5 SO 6 O 5 O 6 O 7 SO 4

SO 2 O 8 SO 5 O 8 SO 3 SO 6 SO 4 SO 5 SO 3 O 6 O 5 O 7 SO 3 SO 6 SO 3 SO 2 SO 2 O 7 O 5 O 8 O 5 O 7 SO 3 SO 2

SO 2 O 8 O 7 O 8 O 5 SO 6 SO 6 SO 2

SO 3 O 7 SO 4 O 6 O 4 O 6 SO 3 SO 2

SO 2 O 7 SO 5 SO 6 O 2 SO 6 O 6 O 8 O 5 SO 6 SO 4 SO 2

SO 1 SO 6 SO 5 SO 4 SO 2 O 7 O 5 O 8 O 5 O 7 O 6 SO 2

SO 3 SO 6 O 7 SO 6 O 3 O 8 O 7 SO 5 O 5 O 7 O 6 O 7

SO 1 SO 6 SO 4 SO 2 O 5 O 8 O 7 O 9 O 3 SO 6 SO 4 SO 5 O 4 SO 6 O 7 SO 6

SO 2 SO 6 SO 5 SO 2

235



SO 1 SO 2 SO 3 SO 1 O 7 O 9 O 8 SO 5

SO 5 O 10 O 7 O 7 O 7 O 10 SO 8 O 8

SO 3 SO 3 SO 5 SO 3 SO 6 SO 6 SO 6 SO 4 SO 2 O 8 O 7 SO 4 SO 2 SO 2 SO 4 SO 1 O 7 O 7 O 8 SO 3

SO 2 SO 3 SO 4 SO 2 O 7 O 10 O 9 O 4

SO 3 SO 2 SO 5 SO 4 O 8 O 9 O 8 O 8

SO 4 SO 6 SO 5 SO 2 O 7 SO 5 O 7 SO 4

SO 5 SO 3 O 7 SO 5 SO 5 SO 6 O 8 SO 6 SO 6 O 7 O 9 SO 5 SO 2 SO 3 SO 5 SO 3 SO 4 SO 3 SO 6 SO 2 O 6 O 9 O 9 O 5 O 5 SO SO 4 O 4

SO 4 SO 6 O 7 SO 3 SO 6 O 10 O 7 O 7 O 7 O 9 O 9 O 6

SO 4 SO 2 SO 4 SO 2 O 7 O 10 O 8 O 8

SO 6 SO 3 SO 3 SO 4 O 5 SO 5 O 7 SO 4

SO 3 SO 3 SO 3 SO 2

236

Song sample perceptual responses Listeners 15

Listener 15 Response Rating

SO 4 O 8 O 6 O 7

SO 4 SO 4 SO 3 SO 3 O 9

SO 3 O 6

SO 3 O 6

SO 4 O 5

SO 5 SO 4 O 6

SO 3 SO 4 O 7

SO 3 SO 6 O 4 O 5

SO 3 O 8

SO 4 SO 5 SO 3

237

Appendix U: Perceptual LTAS

Perceptual Sample LTAS Subjects 1-2

Hz SO1 MO1 SSO1 MSO1 SO 2 MO2 SSO2 MSO2 0

125 -61.17 -59.84 -62.37 -60.4 -62.27 -60.8 -62.62 -61.48 250 -49.71 -47.57 -54.42 -49.66 -58.75 -49.06 -64.02 -51.69 375 -42 -39.1 -46.2 -42.05 -46.16 -40.3 -51.43 -43.43 500 -35.92 -36.35 -40.17 -40.06 -36.38 -37.84 -44.29 -41.36 625 -32.47 -39.1 -36.84 -42.93 -30.18 -39.45 -39.59 -44.02 750 -34.28 -42.4 -37.99 -42.35 -28.98 -41.79 -39.04 -46.3 875 -40.29 -39.57 -40.4 -38.88 -32.91 -43.34 -43.88 -47.11

1000 -40.89 -39.43 -40.62 -39.56 -43 -43.82 -55.18 -49.15 1125 -38.18 -42.79 -41.73 -44.83 -52.96 -47.72 -62.39 -54.27 1250 -36.32 -47.99 -43.28 -49.88 -53.96 -55.09 -62.91 -61.24 1375 -36.98 -48.13 -45.28 -50.4 -54.99 -52.31 -62.35 -55.17 1500 -42.08 -47.62 -49.12 -51.16 -55.86 -49.88 -61.85 -52.37 1625 -49.74 -49.15 -51.79 -51.54 -56.82 -51.28 -60.84 -55.3 1750 -50.79 -53.81 -50.59 -54.23 -59.85 -54.12 -62.54 -61.01 1875 -54.15 -56.95 -52.63 -57.17 -62.45 -55.87 -68.29 -62.12 2000 -58.42 -55.59 -54.29 -54.17 -61.29 -57.74 -71.99 -60.28 2125 -61.94 -56.35 -56.81 -55.32 -60.67 -57.12 -73.76 -60.53 2250 -62.12 -61.09 -61.2 -62.48 -62.35 -58.08 -76.5 -64.35 2375 -60.13 -64.62 -62.11 -65.53 -63.26 -61.2 -77.94 -65.78 2500 -57.56 -60.08 -61.11 -62.43 -63.12 -61.34 -77.83 -63.56 2625 -53.93 -56.16 -57.33 -62.31 -64.23 -60.63 -78.03 -65.9 2750 -52.01 -55.05 -55.92 -61.24 -63.85 -63.14 -76.95 -71.15 2875 -50.98 -55.49 -55.67 -56.85 -62.17 -62.04 -76.96 -67.16 3000 -51.13 -55.24 -57.29 -54.98 -58.31 -58.1 -75.14 -62.54 3125 -52.35 -55.3 -59.81 -56.78 -54.83 -55.97 -72.24 -62.72 3250 -51.51 -57.29 -59.74 -61.39 -54.63 -56.67 -73.82 -65.05 3375 -51.86 -59.86 -60.95 -63.49 -57.21 -57.64 -76.2 -65.72 3500 -55.39 -60.14 -62.26 -62.18 -59.59 -57.75 -77.1 -64.06 3625 -59.88 -60.25 -63.65 -62.92 -60.03 -58.7 -78.47 -63.89 3750 -65.1 -63.35 -68.62 -66.58 -60.18 -62.22 -77.54 -67.38 3875 -70.55 -69.59 -76.42 -70.36 -62.06 -66.47 -79.03 -66.46 4000 -74.65 -76 -78.82 -73.92 -64.72 -67.14 -79.77 -63.96 4125 -78.67 -78.5 -78.8 -79.18 -68.02 -68.64 -80.94 -65.33 4250 -83.62 -81.52 -80.96 -85.97 -72.51 -72.87 -84.79 -70.1 4375 -88.71 -88.37 -87.1 -92.64 -75.41 -75.82 -85.95 -74.65

238

4500 -89.92 -94.55 -88.6 -96.65 -78.69 -79.5 -87.13 -77.27 4625 -88.08 -97.18 -87.27 -98.21 -81.63 -84.34 -86.28 -78.46 4750 -87.64 -98.95 -87.13 -97.6 -82.61 -89.32 -85.77 -84.49 4875 -90.19 -99.39 -85.16 -95.68 -82.64 -94.1 -89.66 -94.3 5000 -93.54 -100.04 -86.34 -95.3 -83.82 -95.37 -96.55 -98.04 5125 -94.27 -99.37 -86.82 -95.89 -85.7 -95.18 -100.35 -100.34 5250 -92.06 -96.62 -85.34 -97.22 -87.23 -96.16 -103.28 -101.52 5375 -90.41 -94.78 -85.04 -98.14 -88.72 -95.87 -104.36 -101.53 5500 -91.57 -94.99 -83.61 -97.4 -89.58 -94.69 -104.48 -99.2 5625 -93.46 -96.59 -83.38 -97.52 -88.45 -94.37 -104.54 -97.76 5750 -93.48 -97.2 -85.25 -97.8 -88.41 -93.68 -104.07 -97.93 5875 -92.53 -96.76 -86.57 -96.71 -88.81 -93.62 -103.61 -97.35 6000 -90.82 -96.91 -89.64 -96.86 -88.77 -92.48 -103.18 -96.52 6125 -88.98 -96.59 -94.15 -98.06 -87.77 -90.19 -101.94 -94.4 6250 -86.61 -94.47 -95.21 -97.7 -85.64 -88.54 -101.09 -92.24 6375 -85.67 -92.14 -94.82 -96.45 -85.1 -86.98 -98.8 -91.93 6500 -85.46 -90.09 -94.08 -96.46 -84.65 -86.56 -96.88 -91.59 6625 -85.18 -88.81 -94.2 -96.32 -84.56 -86.44 -97.19 -89.87 6750 -85.83 -89.08 -95.1 -95.41 -85.66 -85.44 -96.96 -90.36 6875 -87.06 -90.5 -95.69 -94.93 -86.61 -84.53 -96.54 -90.83 7000 -87.52 -91.56 -96.68 -94.78 -88.19 -84.66 -88.17 -90.64

239


Hz SO3 MO3 SSO 3 MSO3 SO4 MO4 SSO4 MSO4 0

125 -62.1 -59.16 -62.43 -60.93 -62.02 -59.19 -62.72 -61.24 250 -55.37 -45.17 -57.12 -48.73 -54.1 -47.36 -58.56 -50.5 375 -43.55 -36.96 -45.69 -40.51 -42.94 -38.83 -46.99 -42.81 500 -35.68 -34.48 -38.51 -38.36 -35.6 -35.95 -40 -41.02 625 -28.24 -36.54 -32.04 -40.38 -30.12 -36.98 -31.88 -44.05 750 -24.9 -39.49 -29.15 -42.62 -27.35 -38.38 -27.43 -44.7 875 -27.17 -38.25 -31.72 -40.09 -28.67 -39.28 -28.41 -40.67

1000 -35.91 -36.14 -40.79 -38.04 -35.54 -40.5 -35.93 -39.9 1125 -43.43 -38.44 -45.96 -40.91 -41.91 -44.63 -44.05 -43.77 1250 -43.41 -45.55 -45.64 -48.15 -42.56 -51.4 -44.88 -52.44 1375 -42.52 -49.49 -45.45 -47.06 -46.29 -54.68 -47.07 -53.02 1500 -40.35 -51.62 -44.27 -49.61 -50.7 -53.73 -49.8 -52.47 1625 -41.09 -51.39 -44.35 -53.23 -52.4 -53.58 -50.04 -53.18 1750 -46.68 -51.58 -50.24 -51.83 -53.93 -56.78 -52.98 -54.41 1875 -55.6 -53.99 -59.13 -51.85 -54.28 -59.02 -59.56 -56.72 2000 -59.02 -54.12 -58.49 -52.07 -53.03 -61.66 -56.6 -56.22 2125 -57.95 -53.17 -59.62 -53.16 -54.43 -63.58 -53.85 -56.81 2250 -55.97 -55.8 -59.95 -57.36 -58.2 -66.8 -55.27 -62.15 2375 -54.2 -59.46 -61.11 -60.49 -59.37 -68.54 -54.52 -66.25 2500 -53.35 -56.73 -63.99 -56.95 -58.94 -66.03 -56.12 -62.82 2625 -55.1 -52.56 -60.87 -53.76 -58.07 -62.81 -60.28 -62.69 2750 -53.15 -49.29 -58.54 -55.06 -56.33 -59.39 -56.46 -67.78 2875 -49.35 -47.48 -56 -57.58 -54.5 -56.74 -55.71 -66.86 3000 -47.24 -47.61 -52.91 -52.23 -53.15 -54.8 -57.49 -60.52 3125 -47.69 -47.52 -53.2 -49.1 -51.87 -53.6 -53.51 -59.55 3250 -51 -48.16 -54.91 -50.94 -51.62 -54.54 -52.76 -63.16 3375 -54.22 -50.1 -55.66 -55.2 -52.15 -56.82 -55.22 -65.26 3500 -55.69 -51.76 -58.12 -55.36 -53.02 -56.06 -57.52 -62.95 3625 -57.7 -53.74 -59.45 -54.53 -55.61 -56.27 -59.66 -63.08 3750 -62.57 -58.48 -62.02 -58.01 -58.77 -59.23 -62.32 -66.36 3875 -69.75 -66.79 -65.95 -65.84 -61.9 -63.38 -63.74 -65.44 4000 -76.07 -76.15 -68.42 -71.9 -64.8 -67.7 -65.82 -62.32 4125 -77.93 -81.19 -73.4 -76.15 -66.67 -70.76 -68.27 -62.7 4250 -78.98 -82.14 -79.31 -77.67 -69.84 -73.21 -70.17 -67.27 4375 -80.31 -83.11 -81.8 -78.47 -75.15 -77.16 -75.06 -72.56 4500 -80.64 -82.68 -82.4 -78.71 -79.8 -84.34 -81.04 -74.49 4625 -77.65 -81.66 -81.92 -77.86 -80.76 -88.36 -79.85 -75.32 4750 -76.6 -81.57 -82.71 -77.94 -79.2 -88.06 -76.26 -79.46

240

4875 -78.61 -80.12 -84.17 -80.51 -76.73 -86.75 -75.66 -83.28 5000 -80.83 -79.37 -85.89 -81.82 -76.26 -85.82 -77.19 -83.31 5125 -82.8 -81.78 -87.03 -81.6 -79.05 -86.89 -80.42 -83.85 5250 -83.85 -85.42 -86.95 -83.63 -83.08 -89.34 -86.39 -87.64 5375 -82.1 -86.62 -87.6 -86.59 -85.35 -90.17 -85.9 -93.74 5500 -81.36 -85.15 -89.22 -86.4 -83.82 -89.04 -81.05 -95.72 5625 -79.59 -82.67 -89.61 -85.42 -82.32 -87.51 -80.36 -95.35 5750 -76.37 -81.26 -89.5 -84.38 -79.61 -86.41 -82.56 -95.41 5875 -74.17 -80.64 -89.43 -84.72 -76.27 -87.07 -86.13 -94.18 6000 -75.28 -80.23 -88.13 -83.28 -75.05 -88.66 -88.01 -91.97 6125 -80.01 -79.97 -86.11 -79.17 -76.64 -89.93 -85.45 -90.58 6250 -82.46 -80.78 -84.63 -77.18 -81.59 -90.08 -82.76 -90.93 6375 -82.47 -82.01 -83.66 -77.68 -86.29 -87.93 -83.51 -91.65 6500 -82.92 -82.52 -84.23 -77.67 -85.39 -87.05 -86.98 -90.58 6625 -82.84 -82.81 -85.7 -76.34 -85.04 -88.96 -88.08 -89.76 6750 -83.19 -82.99 -86.81 -76.98 -86.34 -90.98 -88.47 -90.64 6875 -84.22 -83.56 -88.6 -79.65 -84.91 -90.13 -90.77 -91.1 7000 -85.11 -85.35 -91.87 -82.4 -84.56 -88.5 -90.27 -90.56

241


Hz SO5 MO5 SSO5 MSO5 SO6 MO6 SSO6 MSO6 0

125 -58.47 -59.07 -59.58 -60.91 -61.65 -60.98 -62.53 -62.22 250 -48.01 -45.47 -48.44 -49.57 -51.97 -49.17 -54.25 -52.73 375 -38.79 -37.05 -41.82 -41.82 -41.12 -40.93 -44.33 -43.11 500 -31.68 -34.55 -36.36 -40.07 -31.33 -38.71 -35.1 -39.95 625 -30.85 -36.85 -35.27 -43.13 -27.29 -41.73 -31.01 -41.92 750 -37.1 -41.56 -40.62 -45.24 -29.35 -45.77 -32.78 -45.13 875 -49.05 -45.96 -47.57 -43.24 -38.87 -45.02 -41.32 -44.41

1000 -49.81 -46.46 -48.9 -43.8 -50.45 -46.47 -49.52 -44.65 1125 -51.72 -49.14 -51.34 -48.88 -48.15 -51.87 -48.37 -49 1250 -55.56 -54.35 -53.71 -57.01 -48.61 -52.91 -49.51 -55.37 1375 -54.12 -51.4 -52.93 -53.85 -51.67 -49.85 -49.31 -52.75 1500 -46.47 -46.64 -48.34 -50.53 -48.56 -49.82 -49.76 -50.95 1625 -43.12 -46.28 -43.56 -51.02 -42.75 -51.28 -48.8 -51.12 1750 -45.05 -50.93 -41.84 -53 -39.96 -54.29 -48.17 -54.96 1875 -51.87 -54.29 -45.89 -56.16 -40.87 -59.72 -51.03 -57.82 2000 -58.58 -54.38 -55.89 -56.47 -45.87 -59.23 -57.06 -54.25 2125 -60.71 -56.58 -63.04 -57.56 -53.67 -58.41 -62.27 -51.17 2250 -58.81 -61.16 -64.68 -61.43 -60.08 -62.62 -63.23 -53.65 2375 -56.95 -63.1 -61.86 -63.82 -60.51 -67.94 -63.14 -59.79 2500 -56.51 -58.37 -62.12 -64.04 -58.28 -65.01 -65.13 -59.7 2625 -55.02 -55.41 -64.34 -64.48 -55.64 -65.48 -63.6 -56.84 2750 -52.81 -55.85 -58.53 -67.6 -56.63 -68.82 -63.06 -58.69 2875 -52.36 -55.99 -53.4 -62.36 -58.66 -65.28 -63.67 -63.39 3000 -52.94 -54.53 -51.86 -57.9 -55.66 -60.52 -62.77 -57.35 3125 -53.99 -54.5 -53.48 -58.5 -50.5 -58.52 -59.2 -51.96 3250 -52.66 -55.78 -58.05 -62.67 -46.49 -58.89 -54.68 -52.04 3375 -51.13 -55.39 -58.22 -62.6 -46.19 -58.96 -52.57 -56.39 3500 -51.8 -54.06 -56.31 -60.19 -48.81 -58.58 -53.78 -57.77 3625 -53.87 -54.64 -56.64 -61.54 -50.99 -59.07 -56.47 -56.22 3750 -57.52 -56.96 -58.41 -64.99 -53.48 -60.73 -59.22 -57.4 3875 -60.41 -61.04 -59.24 -62.76 -54.3 -62.22 -60.39 -61.26 4000 -61.76 -64.62 -58.44 -59.95 -55.25 -63.47 -60.16 -62.02 4125 -61.64 -67.57 -59.06 -60.46 -57.7 -65.58 -61.85 -62.44 4250 -63.22 -70.69 -62.9 -63.44 -61.35 -69.06 -65.01 -63.79 4375 -68.98 -75.34 -68.23 -65.92 -66.14 -73.24 -68.61 -67.87 4500 -79.22 -80.65 -70.55 -67.84 -67.87 -74.28 -73.34 -73.79 4625 -86.68 -82.76 -71.91 -69.62 -68.02 -74.9 -72.19 -74.58 4750 -87.32 -85 -74.56 -73.62 -70.21 -76.91 -71.22 -73.32

242

4875 -89.44 -86.85 -78.82 -78.02 -74.53 -78.75 -74.19 -74.84 5000 -92.05 -87.52 -81.13 -80.6 -78.73 -80.4 -78.43 -77.64 5125 -91.88 -86.33 -80.51 -83.51 -80.45 -83.54 -82.44 -78.49 5250 -88.91 -85.51 -81.12 -87.53 -81.1 -88.09 -85.1 -79.68 5375 -82.68 -85.49 -82.54 -88.9 -77.92 -92 -86.59 -84.31 5500 -78.63 -85.59 -82.53 -85.46 -75.96 -93.22 -88.91 -92.08 5625 -78.31 -85.1 -80.96 -84.64 -76.48 -92.39 -91.53 -92.19 5750 -78.57 -85.53 -81.33 -85.02 -74.99 -90.49 -93.28 -90.57 5875 -79.53 -86.56 -82.3 -84.42 -73.05 -89.43 -93.36 -90.68 6000 -82.65 -88.04 -83.02 -84.03 -73.11 -88.72 -91.01 -91.29 6125 -85.46 -87.56 -84.54 -85.38 -75.03 -87.8 -89.1 -89.69 6250 -85.95 -85.72 -86.42 -88.39 -75.63 -86.8 -88.6 -86.24 6375 -85.78 -85 -90.1 -90.04 -74.56 -86.85 -87.03 -83.55 6500 -85.05 -83.84 -93.17 -90.36 -73.84 -88.15 -84.56 -82.96 6625 -84.55 -82.45 -92.41 -91.32 -74.24 -88.82 -82.73 -84.37 6750 -85.73 -82.35 -91.72 -92.14 -76.95 -87.98 -82.14 -85.2 6875 -87.51 -82.16 -91.44 -91.21 -79.99 -87.23 -82.32 -84.5 7000 -88.1 -82.06 -91.12 -90.86 -79.28 -87.49 -81.89 -85.36

243

Appendix V: ICMPC8 Abstract

APPRAISING VOCAL BEAUTY BY EAR AND EYE: CAN WE BELIEVE OUR SENSES? Dianna T Kenny and Helen F Mitchell

Background

What qualities of the singing voice define vocal beauty? This question has come into

sharp focus since the advent of acoustic analysis. There is now both a vast literature on

vocal acoustic properties and many unanswered questions about what conclusions can

be drawn from such analyses. These questions are both methodological and conceptual.

There are many ways that the voice can be represented acoustically, but the most

common is the measurement of energy production. Long-term average spectra (LTAS)

are used to represent singers’ sound and its different vocal qualities based on energy

changes that occur during different vocal tasks. Studies of voice quality have taken two

main forms. In the first type of study, LTAS is used to “validate” perceptual cues. In the

second type of study, acoustic findings are “validated” by perceptual ratings.

Aims

In this study, we investigated pedagogues’ rating of female classical singers using

maximum open throat technique and reduced open throat technique in their singing of a

classical song and a romantic lied. We matched perceptual ratings to acoustic measures

of each perceptual sample to determine whether acoustic analysis matched perceptual

judgements of overall timbre. The aim of the study was to assess the degree of

agreement between acoustic analysis and perceptual ratings of vocal beauty.

244

Method

Fifteen expert judges rated 24 samples with six repeats of six advanced singing students

under two conditions: ‘optimal’ (O), representing maximal open throat and ‘suboptimal’

(SO), representing reduced open throat. LTAS were performed on each singing sample

and sound pressure ratio (SPR) and energy ratio (ER) were calculated on each LTAS.

Perceptual scores, SPR and ER were rank ordered. We then compared perceptual

rankings with rankings of acoustic measures (SPR and ER) to assess whether the

acoustic characteristics matched the perceptual judgments of overall timbre.

The study design was a repeated measures [24 ratings per listener] randomized complete

block with a 2 [Task (Mozart vs Schubert)] x 2 [condition (Optimal and Sub-optimal]

factorial structure. Main effects for task and condition and interaction effects (contrasts)

were calculated for each dependent measure.

Results

Intraclass correlations with two-way random effects model of absolute agreement

(ICC(2,1)) indicated that judges were reliable in their ratings, matching their exact rating

in the first rendition and the repeat in 5 of 6 cases.

Perceptual ratings were significantly higher for O than for SO in both the Mozart and

Schubert tasks. There was no interaction between musical task and condition.

245

While there was a significant relationship between SPR and ER, there was no

relationship between perceptual ratings of vocal samples or singers based on SPR or

ER.

Conclusions

These findings suggest that since LTAS measures are not consistent with perceptual

ratings of vocal beauty, they cannot be used to define a beautiful voice.

Topic Areas

Aesthetic perception and response

Key words: LTAS, perceptual ratings, vocal beauty

246

Appendix W: Summary of publications and presentations arising from the thesis

Peer-reviewed journal articles Mitchell, H. F., & Kenny, D. T. (in review). Open throat: acoustic and perceptual

support for pedagogical practice. Journal of Singing.

Mitchell, H. F., Kenny, D. T., Ryan, M., & Davis, P. J. (2003). Defining open throat

through content analysis of experts' pedagogical practices. Logopedics

Phoniatrics Vocology, 28(4), 167-180.

Mitchell, H. F., & Kenny, D. T. (2004). The effects of “open throat” technique on long

term average spectra (LTAS) of female classical voice. Logopedics Phoniatrics

Vocology, 29(3), 99-118.

Mitchell, H. F., & Kenny, D. T. (2004). The impact of “open throat” technique on

vibrato rate, extent and onset in classical singing. Logopedics Phoniatrics

Vocology 29(4), 171-182.

Kenny, D. T. & Mitchell, H. F., (in press). Acoustic and perceptual appraisal of vocal

gestures in the female classical voice. Journal of Voice.

Mitchell, H. F., & Kenny, D. T., (in press). Can experts identify “open throat” technique

as a perceptual phenomenon? Musicae Scientiae.

Conference Papers Mitchell, H. F., Kenny, D. T., Ryan, M. Open throat: pedagogical perceptions and

practices. ANATS National Conference, October 2002, Melbourne.

Mitchell, H. F., & Kenny, D. T. Evaluation of open throat technique on classical

singing. Music and Gesture [University of East Anglia in association with the

European Society for the Cognitive Sciences of Music (ESCOM), the Society

247

for Music Analysis (SMA), and the Society for Education, Music, and

Psychology Research (SEMPRE)], August 2003, East Anglia, UK.

Kenny, D.T., & Mitchell, H.F., Visual and auditory perception of vocal beauty: conflict

or concurrence? In: Lipscomb S, Ashley R, Gjerdingen R, Webster P, editors.

8th International Conference on Music Perception & Cognition (ICMPC8);

August 2004; Evanston, IL USA: Causal Productions; 2004. p. 171-174.

248

Documents

1 Overview of the Thesis