Emotional Expression Code in Opera and Lied Singing
Dr.Eliezer Rapoport
Emotional expression in singing is encoded at the microlevel of the single individual tone, and is
deciphered by FFT (Fast Fourier Transform) analysis. FFT
spectrograms of vocal tones, from recordings of opera arias and lieder sung by
famous artists, reveal a large variety of (tonal) temporal structures, composed
of several initial stages, that are correlated in a
very systematic way with the emotions expressed in the text and the music. Over
50 distinct types of vocal tone structures (to be called modes) were
identified, and classified into eight main categories: Neutral-Soft, Calm,
Expressive, Transitional-Multistage, Intermediate, Short, Excited, and
Virtuoso, in a very well-defined hierarchical scheme. These various types of
temporal structures, and in particular the tone beginnings, are controlled by
the interplay of seven mechanisms: 1) onset of phonation (voicing); 2) vibrato;
3) excitation of higher harmonic partials, (particularly the singing formant);
4) transition - a gradual pitch increase from the onset to the sustained stage;
5) sforzando - an abrupt pitch increase at the very onset of the tone; 6) pitch
change within the tone; and 7) unit pulse.
The
large number and variety of these temporal structures (modes) are so shaped
deliberately (but unawarely), by the singers for expressing the large richness
and variety of emotions from sadness to joy, and from happiness to anger.
Furthermore, these modes indeed constitute the language of emotional expression
in singing deciphered in the present study. A notation, based on the operating
mechanisms was developed, and is in effect the alphabet. The various modes are
the vocabulary, and the categories and their use in the melodic phrase
constitute the grammar and syntax rules of this language.
INTRODUCTION
AND OVERVIEW
This
work, originally intended as a study of timbre in voices of great artists,
involved a very detailed examination of a large number of vocal FFT (Fast
Fourier Transform) spectrograms of great opera and lied
singers, from compact disc recordings. The observation that an individual vocal
tone has a characteristic temporal structure (music is a sequence of temporal events)
containing, quite often, several initial stages, the large number and variety
of such structures, and the
Further examination of the context in which the
various modes appeared, revealed a clear separation of the modes into groups according
to the emotional context of the music and the text - in such a way that a
certain group (or category) of modes that appear in a certain emotional
context, such as rage, definitely will not appear in a contrary emotional
context such as happiness or joy. To each type of emotion there corresponds typical groups of modes. The conclusion was
that, being in some way characteristic of the singer, these modes are indeed
related to the singer's way of expression of the various emotions in the
literary text and the musical score. Then, a systematic study was carried out
by examining selected excerpts and phrases from opera arias and lieder of
well-defined emotional character, from sadness to rage or triumphant joy, based
on listening to the music, and on the text. To cite representative examples:
(1) Schubert Ave Maria D.839, sung by Felicity Lott, expressing a devotional
prayer and peace of mind; (2) Amelia's aria "Morro, ma prima in
grazia...", from Verdi Un Ballo in Maschera, (Act III, scene I) sung by
Margaret Price, expressing deep sadness and resignation; (3) Riccardo's Aria
"La rivedra nell'estasi..." from Verdi Un Ballo in Maschera, (Act I,
scene I), sung by Luciano Pavarotti, expressing love; (4) Alfredo's aria
"De miei bollenti spiriti" from Verdi La Traviata, sung by Pavarotti
(and by Alfredo Kraus), expressing happiness; (5) Riccardo's aria "E
scherzo od a follia..: ', from Verdi Un Ballo in Maschera, (Act I, scene II),
sung by Pavarotti, expressing ridicule; (6) Riccardo's aria "Si, rivederti,
Amelia a nella tua belta..." from Act III, scene II, of Verdi Un Ballo in
Maschera, (Pavarotti) expressing an outburst of emotions; (7) Calaf's phrase
"figlio del Cielo...", expressing challenge, determination, and
utmost tension, and the riddle solving scene, expressing excitement and victory
- both from Puccini Turandot (Pavarotti); (8) the second half of Schubert lied
"Der Doppleganger", sung by Dietrich Fischer-Dieskau, expressing
utmost rage; and (9) the bell song from Delibes Lakmas, sung by Edita Gruberova
- for virtuoso and coloratura singing. These, and many others, were
meticulously
Furthermore, the grouping of modes of similar
structure and similar emotional character into a category was found very
effective: eight such categories, belonging to a very distinct hierarchical
scheme, cover the whole range of emotions, from the mildest and most gentle
(lowest category in the hierarchy) to the most vigorous and excited (highest in
the hierarchy). These categories are, in hierarchical order: 1. Neutral-Soft
(N), 2. Calm (C), 3. Expressive (R),
4. Transitional-Multistage (T), 5. Short (S), 6. Intermediate (I), 7. Excited
(X), and 8. Virtuoso (Vr) - to be further described in the section
labelled "Results and Systematics".
A clear differentiation exists: singing modes
corresponding to salient (long, high pitch, or metrically accented) tones in
the score are very different from modes corresponding to short, unaccented,
ornamental tones, and always belong to different categories (the salients
usually belong to the higher category). Secondly, the salients are the modes
(and tones) that determine the emotional character of the aria (or lied). Thus,
among the salient modes those of examples (1) and (2) belong to the Calm (C)
category, whereas the ornamental tones belong to the Neutral-Soft (N) category.
Salients of examples (3) and (4), belong to the T
category. In example (6) some salients belong to the T category,
and others to the X category, as are those of examples (7), and (8). A large
variety of short, ornamental modes occur in example (5), and are classified in
the S category. Many Vr category modes are found in example (9). The various
modes, categories, the classification scheme, and the various mechanisms are
described in detail in the sections on "Results and Systematics",
"Mechanisms" and "Theoretical Interpretation".
The present work belongs in the discipline of physics
and acoustics, and is concerned with the information encoded in the acoustic
signals that constitute the vocal tones. It deals with the singer role in the
singer-listener communication process. Perception and cognition - signal
processing on the listener's part belong to psychoacoustics. However,
perception of changes and transients, similar to edge detection in vision is
unquestionable, as phrased by Sundberg (1982): "It seems to be a
perceptual universal: changing stimuli are easier to process than
quasi-stationary stimuli", and: "the beginning and ending of a sound
are probably very revealing". So is the importance of the attack part in
timbre, as demonstrated by Risset (1991) for wind instruments, and the brain's
perceptive and cognitive ability to process, integrate and assign definite
pitch to short-duration acoustic signals, as demonstrated in psychoacoustic
experiments by d'Alessandro and Castellengo (1994). All these lead to the
assumption that the structural features in the acoustical signals, so
artistically shaped by the singer, are (subconsciously) perceived and
deciphered by the listener.
Seashore (1938) introduced the concept of
"Musical performance score", wherein pitch frequencies in a
particular performance (mostly from Harold
Emotional
expression in speech, music and singing
Emotional
expression in speech and in theater acting was a subject of several
investigations by Seashore (1938), Fonagy (1981), Kotlyar and Morozov (1976),
Williams and Stevens (1972), Clynes (1983), and others, and the state of
knowledge on this subject was reviewed by Sundberg on several occasions
(Sundberg 1977, 1982, 1987, 1994). In essence the means for emotional
expression in speech are changing pitch, loudness, and tempo-changing pitch
being the dominant factor in intonation. In singing, emotional expression takes
place in three levels: 1. text, 2. musical score, 3.
performance. On the second level (the composer's level - or the written music)
rules of emotional expression, and in particular rules of excitement and
relaxation, were studied by musicologists for many years (Cooke 1983; Cohen
1971, 1983). Universal factors creating tension or excitement are associated
with abruptness, increased loudness, pitch, tempo etc., whereas gradual change
is typical of a calm or relaxed mood. On the performance level, as the pitch
had been determined beforehand by the composer, the degrees of freedom for
emotional expression and personal interpretation left for the performing artist
in keyboard instruments are changing loudness, tempi, tone durations and
connections - staccato and legato.
Singers, in addition to the flexibility of changing
resonators and registers, and thus the spectral timbre of the voice from light
to deep, resonating etc., have an additional ability to shape the temporal
structure of the individual tones and to create a large variety of nuances (the
modes characterized in the present work), including some pitch transitions and
changes within the individual tone, all serving the purpose of emotional
expression (along with timbral embellishment). This is microintonation -
intonation on the microscale within the individual tone. The present work shows
that the universal principles of excitement and relaxation are valid down to
the microlevel of an individual tone. Intonation in speech serves as a means of
emotional expression - the counterpart in music is the melodic line, written by
the composer along with the other elements discussed above. The singer
Emotional expression in singing can be considered as
encoded into the acoustic signals as frequency modulation. The analogy to radio
transmission, where a carrier wave in the MHz range is being
frequency-modulated (FM) by information carrying signals (sound, speech, and
music) in the 100 Hz-15 KHz range, is that in singing, pitch (range 100-1,400
Hz) is the carrier wave, frequency modulated by emotion carrying signals in the
6-12 Hz range. In speech, a distinction is made between the lexical level
carrying the verbal information, and the prosodic level (intonation) carrying
emotions. The analogy in singing is: the melodic line (pitch) is the lexical
level on which the 6-12 Hz emotional messages are superposed as a prosodic
level on the microscale (microintonation). The author is accumulating evidence
pointing to the direction that the frequency range of 6-12 Hz is of special
neurophysiological importance in relation to neurologic oscillators (Titze
1995) in the 6-12 Hz range. Such oscillators function as brain pacemaker
mechanisms (biological clock) that are operational in rhythm and time counting.
A unit pulse, or elementary beat, in the 6-12 Hz range, was encountered in the
present work, being very dominant in singing melodies of rapid tempi (section
on the "Unit Pulse").
METHODOLOGY
AND PROCEDURES
Opera
arias and lieder recorded on compact discs were used as the source. Recorded
material from concert halls, opera houses or recording studios represents real
singing, in real (theater or public) atmosphere, with emotions. Singing in a
laboratory under experimental conditions in psychoacoustic experiments might
lead to very different results, with the emotional aspect frequently
overlooked.
Selected musical phrases from performances by great
artists were recorded into the computer (Silicon Graphics) and subjected to FFT
analysis (see the following section). Conversion to lower sampling frequency,
32, 24, 16, or 12 kHz for saving memory space and computing time, was done,
depending on the harmonic spectral range of the particular artist, with very
slight or no loss of quality - 12 kHz was found quite adequate in most cases,
even for soprano singers. Eight kilohertz sampling rate is quite sufficient for
baritone and bass singers. Examination of an FFT spectrogram, or parts of it (a
single tone or a part of a single tone), was done simultaneously with listening
in order to find the correspondence between the spectral appearance and the
musical contents.
A large number of excerpts from opera arias and lieder
- two singing media of very different character, were studied in order to
ensure that the observed modes and features are general, frequently occurring,
and are not some isolated or
Fast Fourier
Transform (FFT) analysis
FFT
analysis is a very efficient and powerful technique for analysis of acoustic
and musical signals. Detailed description of FFT or DFT (Discrete Fourier
Transform) methods can be found in books on digital signal processing,
(Oppenheim and Schafer 1975, 1989). The FFT program decomposes a signal into
its frequency components, and the FFT spectrogram is a frequency-time diagram
that gives at each instant in time all the spectral information: component
frequencies and their intensities contained in the signal. The FFT spectrogram
is a picture obtained by "freezing" a temporal event (or events),
visually displaying the full information contained in it. In the present work
The WAVES+ program by Entropic Research Lab. Inc. was used, with default
values: step size 2 ms, Hanning window 512 points frame length (fft order 9) -
about 50 ms for sampling frequency of 12 kHz - leading to 20 Hz frequency
resolution, and 50 ms nominal temporal resolution. Actual temporal resolution
is much better due to repetition every 2 ms (step size). In the FFT
spectrograms displayed in the following sections, the x-axis denotes time
(fullscale 1.9 s, smallest division 10 ms), y-axis; frequency in Hz (fullscale
6,000 Hz, smallest division 50 Hz), and the instantaneous intensity of each
frequency component originally presented as colorscale (or colormap) on the
computer display, is translated in the b/w print with grey background in the
following way: the intensity of each frequency component is proportional to the
width of the line in the spectrogram. White and black on the grey background
correspond to high and low intensity, respectively.
It is known already from the days of von Helmholtz and
Ohm that the human ear is also a frequency analyzer, some sort of a filter bank
(Plomp 1964, 1968). The ear and the FFT program are two parallel frequency
analysis systems of different characteristics. The ear has psychoacoustic
constraints: logarithmic scale, nonlinearity, difference tones, critical bands
and masking - the computer: finite resolution as determined by the particular algorithm, and the minimum number of points required for the
FFT.
RESULTS AND
SYSTEMATICS
The
results of the present study are acoustic signals analyzed by FFT, and
classified according to their characteristic temporal structure and shape, in
correlation with their musical and artistic origin and emotional context.
Whereas the acoustic signals in themselves belong to the realm of acoustics and
physics, the information extracted from them sheds light on apparently
nonscientific areas of art, esthetic values, human emotions, expression, and
communication. However, communication between human beings (and also by animals
and birds - see Cohen 1983) is performed through acoustic signals in speech,
and those are meant to convey emotional messages. This is also the case in
artistic singing. FFT analysis was found to be an efficient tool for
deciphering the code of emotional expression in artistic singing, and a
discipline analogous to linguistics is unveiled. Indeed, the analogy between
music and linguistics was demonstrated in numerous studies (Jackendoff and
Lerdahl 1982).
In this section we describe the singing modes and
categories, along with representative examples. The various operating
mechanisms are described in the section on "Mechanisms". The
classification and hierarchical scheme were deduced on the phenomenological
level from the correlation of the structural characteristics of the modes and
their emotional and musical origin. A theoretical interpretation based on
postulates and principles is presented in the section labelled
"Theoretical Interpretation".
The eight categories of singing modes
Representative
modes from amongst the many identified and characterized in the present study, are schematically presented in Fig. 1, in the form of
time variation of the fundamental frequency and higher partials of the singing
formant. The horizontal axis denotes time; vertical axis, frequency; dotted
line; singing formant. Figure 2 (section on "Notation for Short
Tones") shows modes of the S category (to be discussed in the sections
labelled "The Unit Pulse", "Notation for Short Tones",
"The Short (S) category" and "Performance Scores No. 6 and
7"). Actual examples of FFT spectrograms in the frequency-time
representation from opera arias are presented in Figs. 3-14.
First
category: Neutral-Soft (N)
In
the Neutral-Soft category we include modes associated with the mildest feelings
- sometimes at the border of neutrality that appear in the unaccented
ornamental tones in many lieder and arias, such as in the opening of Schubert
Ave maria (sung by Felicity Lott). Many N modes appear in the recitative-like
opening of the
Schematic presentation of the singing mode categories.
Horizontal: time; Vertical: frequency
[ in 500 Hz] units; Dotted line: singing formant.
To
this category belong the most gradual of the singing modes, and as such they
appear in singing of music of calm, relaxed, subdued, or sad character. They
appear when rather long-duration tones are sung,
typically lasting 0.6-2.0 s. Figure 4 shows a typical C mode tone (C1 in Fig.
1, P-E-V in the notation of the section on "Theoretical
Interpretation"). This is the syllable "cia" from the single
word phrase "Fran-cia", expressing longing
to her country of birth; from Elisabetta's aria from Don Carlos by Verdi, sung
by Monserrat Caballe. Other examples can be found as salients in Schubert lieder:
Ave Maria, Nacht and Traume, and Ganymed, sung by Felicity Lott (Rapoport
1996), and in many others. As can be seen in Fig. 4, a C mode has a
characteristic structure evolving gradually in three stages: starting in
phonation of only the lower formants (first stage), then the singing formant is
excited (second stage), and finally vibrato sets in the third stage (see
description of mechanisms in the section on "Mechanisms").
Third
category: Expressive (R)
In
the R category other ways of organization in time of the three mechanisms
(phonation, excitation, and vibrato) are possible. The initial phase of vibrato
is also significant. A full description of the R category is presented in the
section labelled "The Expressive (R) category".
Fourth
category: Transitional-Multistage (T)
A
higher degree in excitement than in the C or R modes is achieved by introducing
a third element: pitch transition; a gradual increase in pitch in one or two
stages from the onset to the sustained stage, mostly practiced by tenors. This
is not
Fig. 4. C mode. ("cia" from
"Fran-cia", Verdi Don Carlos, Elisabetta's aria "Tu the le
Vanita", sung by Monserrat Caballe).
Score
No. 3). The aria "De miei bollenti spiriti" from La Traviata by
Verdi, and the preceding recitative, express great happiness. Pavarotti and
Alfredo Kraus use the T1 and T2 modes extensively in this aria. Further on in
the aria at the climax of happiness, the phrase "io vivo quasi in
ciel" is repeated five times, each time leading to a climax - sung by
Kraus in the T1 mode.
Another type of multistage transition is included as a
subcategory here, though it could belong to the virtuoso class. This is
portamento (the term used by singers), or glide (glissando) - the term used by
Seashore (1938). Portamento is performed mostly by soprano singers either
ascending or on a leap to a high note - to a climax. (The inverse is a
descending portamento). The transition is performed with vibrato, effectively
climbing on it. An ornamental, rather long and gradualencompassing seven
semitones, glide transition G1 taken from Oscar's ballade in Verdi Un Ballo in
Maschera, sung by Kathleen Battle, is shown in Fig. 6. A dramatic portamento is
shown in Fig. 7. The ascending glissando achieves extra qualities adding a
dramatic effect and embellishment and is perceived at moderate glissando as
timbre effect, and at longer glissandi as a chromatic ascending scale (pitch
aspect). It was studied previously from the point of view of pitch perception
(d'Alessandro and Castellngo 1994). In addition to the display of virtuosity,
technical mastery, and timbre embellishment on the part of the singer, a deeper
Fig. 6. G mode. (from Oscar's Ballade,
Verdi Un Ballo in Maschera, sung by Kathleen Battle).
Fifth
category: Intermediate (I)
This
category contains a variety of tonal evolutions that are somewhat intermediate
between the Expressive (R) and the Excited (X) modes. They contain a short previbrato
stage, or decreasing pitch at the tone onset. Such modes appear in the singing
of Maria Callas (Rapoport 1995).
Fig. 7.
A dramatic G mode (portamento) (from Verdi Don Carlos, Elisabetta's aria
"Tu the le vanita..", sung by Monserrat
Caballe).
Sixth
category: Short (S)
The
first group of modes to be classed in the S category are w modes, shown in Fig.
8, and marked in Performance Score No. 5. Such modes have some wiggly structure
of excessive jitter, a precursor of vibrato, and thus occupy an intermediate
Fig. 8. w modes; a sequence: w2 w3 w3 w2 w2. Five opening eighths
from "Ah voi, signori.." phrase sung by
Ulrica, Verdi Un Ballo in Maschera, sung by Christa Ludwig.
Another class contains a large variety of modes of
short-duration tones: quarters, eighths, and sixteenths in rapid passages and
staccatto tones, appearing mostly in music of joyful character but also in
other contexts. The S modes are shown schematically in Fig. 2, and further
discussed in the section on The Short (S) category. They are totally different
from modes of the other categories, and are derived from different mechanisms
(sections on "The Unit Pulse", "Notation for Short Tones",
"The Short (S) category" and "Performance Scores No. 6 and
7"). Some of these modes are condensed or shortened versions of the T and
I modes, and some consist only of the transition part (T and D modes in Fig.
2). The large majority of the S modes shown in Fig. 2 are derived from three
fundamental units: the unit pulse A (or a), the two-pulse unit M (or m), and
the three-pulse M3, or m3 (fully described in sections on "The Unit
Pulse" and "The Short (S) category", notation section
"Notion for Short Tones"). M and M3 modes are shown in Fig. 9 (also
marked in Score No. 5). A, m, and M modes are shown in Fig. 10 and marked in
Score No. 6.
Fig. 9.
M modes; the second sequence of the same five eighths from "Ah voi,
signori.." (Christa Ludwig). Sequence: M !M3! M3 M M3.
Seventh
category: Excited (X)
To
this category belong the strongest of all singing modes. X modes appear in
singing musical phrases of the highest excitement (rage, tension, triumphant
cry) when singers sing in an accentuated way at the top of their voice. The
very famous phrase: "Figho del cielo! io chiedo
d'affrontar la prova" from Puccini's Turandot is sung in a sequence of X
modes by Luciano Pavarotti. The four notes: "io chiedo d'a-" are shown in Fig. 11. X modes start with accentuated vibrato
(sforzando - section on "Mechanisms"), with the first vibrato period
strong and steep, as shown schematically in mode Exl in Fig. 1.
The X modes are not only a result of loudness. Even in
singing at low loudness (piano), transitibn from C mode singing to X mode
singing takes place at a very small increase in tension in the vocal expression
or in the text. This happens in the single word "Si!" - shown in Fig.
12 - from the phrase: "Carlo qui verra!...Si! the parta a scordi omai" (Verdi Don Carlos, the aria of
Elisabetta: Tu the le Vanita), reflecting a momentary tension and anxiety after
a passage of a prayer of utmost sadness ("S'ancor si piange in
cielo") sung in C modes (see also section labelled "Performance Score
No. 5" in this context).
Fig. 10.
A and M modes, from Cherubino's aria "Non so piu
cosa son cosa faccio" (from Mozart Le Nozze di Figaro, sung by Frederica
von Stade).
Eighth
category: Virtuoso (Vr)
Figure 13 shows an example of virtuoso staccato
singing (Virtuoso mode V1) by the soprano Edita Gruberova of an excerpt from
the Bell Song of Delibes Lakme.
Fig. 11 Four X modes. The Phrase "io chie - do d' a.. " (from "Figlio del Cielo..", Puccini Turandot, sung by Pavarotti).
There
is a concentration of all the energy in only three or four partials, in sharply
perfect pitch, no deviations, no vibrato, and no temporal structural
complexities. It represents sheer technical perfection and artistic mastery. No
extramusical emotion is conveyed, as the explicit wish on the composer's part
is to give the artist the opportunity to exhibit her virtuosity and artistry -
"her", as this category exists only in soprano-coloratura roles.
Figure 14 shows a coloratura excerpt in which an ingenious, artistic, and
masterful use of a legato technique is made in a rapid succession of short
tones of a duration comparable to half vibrato period (with an outwardly
appearance similar to vibrato), and achieving a continuous, smooth and very
polished transition from one tone to another, within the rapid tempo of the
music. Such coloratura singing was discussed previously by Sundberg (1982,
1987, 1994), and by d'Allessandro and Castellengo (1994) from the point of view
of pitch perception, whereas here the emphasis is on its functional role in
Fig. 12. X mode. The one-syllable word "Si!" (Verdi Don Carlos, Elisabetta's aria "Tu the le vanita", sung by Monserrat Caballe, see text on "Seventh category").
In this
case there is no "attack" stage as there is a continuous legato transition from
one tone to another, either in a small interval or in a leap. This is
MECHANISMS
A straightforward deduction from the FFT spectrograms examined in the present study is that the various modes are the result of operation and interplay of seven mechanisms, and the timing at which each of them enters into operation. These are the results of voluntarily activated physiological mechanisms (controlled by the brain) for the purpose of emotional expression. These mechanisms are:
1. Phonation: The term phonation, or voicing, usually means production or utterance of vocal sound.
Here it is used particularly to denote sound production in the lower formants.
2. Excitation of higher harmonic partials - especially the singing formant (2-3 kHz for male singers and 3-4 kHz for female singers - see Sundberg 1987).
3. Vibrato.
4. Transition: A gradual pitch increase at the tone onset up to the pitch of the sustained state.
5. Sforzando: An abrupt pitch increase at the tone onset, lasting usually 50-80 ms.
6. Pitch change within the individual tone: a gradual increasing, decreasing, or ondulating pitch change (not vibrato) within the tone.
7. Unit pulse or elementary beat that tightens and releases tension in the vocal folds (see section on "The Unit Pulse").
An additional mechanism - tremolo, an irregular vibrato, imparts pathos or even a feeling of horror.
(Schubert's lied "Der Doppelganger" sung by Fischer-Dieskau). All these mechanisms are perfectly
controlled by the singer (vibrato and higher harmonics excitation are controlled by loudness, but vibrato
is independent to some extent). Change in pitch within the individual tone is an additional flexibility the
human voice has and the artists use to their advantage. This is a facility keyboard instruments do not
have, wind instruments have to a limited extent, and string instrument players can use more at freedom.
The various categories are characterized by the number, combination, and sequence order of the operating mechanisms (see section on "Theoretical Interpretation"). We present here a brief discussion on vibrato (mechanism no. 3) and transition (no. 4), and a simple neurophysiological model of the unit pulse (no. 7).
Vibrato
Seashore (1938) pointed out the important-role of vibrato in rendering beauty, richness, tenderness, pleasing flexibility and expressiveness to artistic singing, and remarked that though vibrato contributes to expressiveness, it does not in itself differentiate among the various emotions. Seashore also remarked that the vibrato extent (in the songs that he and Harold Seashore examined) was one semitone, and that vibrato occurs on all tones.
Vibrato is the phenomenon of periodic frequency modulation (pitch modulation), accompanied by amplitude modulation of the same frequency, in phase, in opposite phase, or of double frequency, depending on whether the strongest harmonic partial is lower, higher, or exactly on the strongest formant, respectively (Sundberg 1987). It should therefore be realized that frequency modulation originates in the vocal folds, whereas amplitude modulation originates in the resonating cavities.
Scotto Di Carlo (1991), describing the physiological aspects of artistic singing, pointed out that vibrato is a result of a perfect equilibrium among the different muscular coordinating mechanisms involved in singing. In addition to its esthetic aspect, vibrato is functional in enabling the voice to float above the orchestral sound and avoid masking. Vibrato is also a means of increasing vocal efficiency, and a very good indicator of vocal health.
Detailed observations and measurements of the vibrato rate and extent by the present author on vocal tones belonging to the various categories of singing modes reveal at least four types of vibrato that appear in different contexts. The vibrato extent is about 100 cent (one halftone) in the X, T, and G modes, and 60-70 cent in the C and R modes. This was found for several singers. In addition, some
"embryonic" vibrato of vibrato extent of 15-30 cent
appear (Cs subcategory). The fourth case is a nonsinusoidal vibrato. The
vibrato frequency and extent are not constant even within one tone (Frame
1994), and change along a melodic phrase in a rather complicated way, sometimes
in concurrence and sometimes nonconcurrently with increasing pitch. This is
obviously related to the expressiveness of vibrato and its relation to tension
increase along the melodic phrase, and deserves further study.
Transition
Two
types of transition (gradual pitch increase) were encountered in the present
study: (1) concave pitch increase, with slope increasing with time (T modes)
and (2) convex or linear pitch increase with decreasing or constant slope,
respectively (K modes, section on "Notation"). They can be further
characterized by the extent in a unit of 100 cents or number of semitones, and
by the duration and can be denoted as T(ql, q2), or K(ql, q2), respectively,
where q2 denotes the extent in semitones, and ql the duration in seconds. The
author prefers to introduce a new unit B, the neurophysiological beat, that
amounts to the unit pulse or one vibrato pulse - about 120-150 ms. By vibrato
pulse we define a vibrato period between two successive minima in the vibrato
line in the FFr spectrogram. (As vibrato is considered an approximate sine
wave, this corresponds to a period between -112 and +37c/2). This is the unit
of time counting in artistic singing - based on results of the present study
(see next section). This unit is not absolutely constant. Like a heart beat, it
varies according to the emotional state, and from one singer to another. Its
choice represents some sort of normalization. Some measurements on T modes
yield values of 2-6 semitones for extent, 2-5 B for duration.
The Unit
Pulse - a simple neurophysiological model
The
vocal folds are two elastic membranes operated by muscular systems (Scotto Di
Carlo 1991, Sundberg 1987) commanded by the brain via the nervous system. In
singing or in speech, excess air pressure from the lungs forces the vocal folds
apart, releasing a pulse of air stream which in turn reduces pressure between
the vocal folds, and their elasticity closes them again. The excess pulmonary
pressure again opens the vocal folds, and the process is repeated. Thus a
periodically repetitive sequence of emitted air pulses is produced. This is
sound, and its fundamental frequency is determined by the mass (length and
thickness) of the vocal folds, and of most importance, by their tension.
Increased tension results in increasing frequency (increasing pitch), and
tension release results in decreasing pitch (Scotto Di Carlo 1991, Sundberg
1987). The shape of a singing signal in the frequency-time FFT spectrogram is
thus an exact replica of the vocal folds tension
132 E.
RAPOPORT
in a tension-time diagram. In
particular, examination of the signals shown in Fig. 10 teaches us that these A
and M signals reflect directly a rapid tightening of the vocal folds followed
by immediate tension release (A mode), this same process is repeated twice in
the M mode. This process is obviously commanded by the brain, and we have here
direct evidence that a pulse is sent from the brain commanding the tightening
of the vocal folds in order to sing at the desired frequency (corresponding to
that tension in the vocal folds). The action of this pulse is rapid tightening,
followed by immediate tension release at a time span corresponding to 6-12
pulses/s (6-12 Hz). In principle, a double-duration tone could be obtained by a
pulse of double duration. In reality a two-pulse mode is obtained. This points
out very clearly that the command for tightening the vocal folds for a
double-duration tone is two single-pulse commands - not one double-duration
pulse. Thus there exists a unit pulse command, a quantum for tightening the
vocal folds consisting of tension increase followed by tension release. For
double and triple-duration tones, two unit pulses and three units
pulses respectively, are required. The author examined a large number of lieder
and arias of short tones and rapid tempi in which there are rhythmical
repetitive units such as: (eighth, eighth, quarter), or: (eighth, quarter) -
the corresponding vocal modes are: unit pulse corresponding to the eighth, and
two unit pulses corresponding to the quarter. (Three examples are presented in
the sections on "The Short (S) category" and "Performance Scores
No. 6 and 7").
The unit pulse is an element of time counting in
artistic singing and provides a new insight and a new point of view in the
interpretation of vibrato, sforzando, and coloratura singing. Vibrato appears
above a certain excitement threshold and is actually a sequence of unit pulses;
periodic tightening and release cycles produced by a repetitive sequence of
unit pulse commands. (Subthreshold excessive jitter was encountered in the w
modes). This explains why vibrato always starts in the direction of increasing
frequency (tightening of the vocal folds) - the first unit pulse being either a
small or a large one. Likewise, the last vibrato pulse always terminates
towards frequency decrease (tension release). A long-duration tone is always
composed of an integral number of whole vibrato pulses. In this way the brain
counts tonal time duration in the number of vibrato pulses (or unit pulses).
In the sequence of coloratura tones of the Virtuoso
Vr2 modes (Fig. 14), each tone is essentially a unit pulse A
mode. The sforzando part of the first vibrato pulse of an X mode is actually a
more vigorous unit pulse. X modes very often terminate in a stronger and
shorter unit pulse.
It is known that various neural and pacemaker
oscillators exist in the brain. It is possible that the unit pulse is produced
by such neural oscillators. The pleasant and expressive effect of vibrato might
be explained by the existence in the brain of oscillators functioning also in
hearing, in such a way that they resonate in response to vibrato - also to
vibrato of string instruments. The unpleasant effect
In a recent publication Titze (1995) wrote: "The
act of singing involves the coupling of an active array of oscillators:
mechanical, neural, and acoustic" and "overriding the mechanical and
acoustic oscillators are the neurologic oscillators". He further wrote:
"Evidence is growing that vocal vibrato, for example, may be a cultivated
vocal tremor: a collection of tremor frequencies in the 4-7 Hz region is
produced in the central nervous system. This can be thought of as a
low-frequency narrow-band noise produced by muscle activation. A peripheral
oscillator (partly mechanical and partly sensory-neural) is tuned to a specific
frequency (around 5 Hz) by reflexes. When this oscillator is
excited by the tremor spectrum through laryngeal muscles, a relatively stable
vibrato frequency results, typically between 5-6 Hz". The jitter in
singing is actually the tremors to which Titze refers, becoming more pronounced
in the w modes. The latter actually have certain temporal structure, indicating
the action of two or three unit pulses (Fig. 8, and section on
"Performance Score No. 5").
THEORETICAL
INTERPRETATION
Hitherto,
on the empirical-phenomenological level, interpretation and classification of the various singing modes was achieved by
meticulous observation, characterization, deduction, and correlation. We
proceed now to a theoretical interpretation and deduction based on postulates
and principles. The first step is the introduction of notation that provides a
means of description of a vocal tone (or mode) from its very onset through its
many stages of development to its termination. This notation also provides a
general framework for a unified treatment of the subject, and even enabled the
author to make further predictions that were afterwards confirmed by observing
new modes in relation to their emotional message, to which no attention was paid,
and no thought was given before (R category). It was also realized that in
effect, we are dealing here with a language, the language of emotional
expression in artistic singing, having this notation as its actual alphabet,
the singing modes as its vocabulary, and the categories, hierarchical
classification scheme, and the organization of the various modes in a melodic
phrase as its grammar and syntax rules, respectively.
Notation
The
following notation is introduced:
p - denotes phonation (lower
formants, usually the fundamental and two or three harmonic partials) at low
intensity (with respect to the intensity at the sustained stage).
P - phonation (lower formants) at full intensity.
p* -
phonation of only the fundamental, with possibly the second harmonic suppressed
to very low intensity-mezza-voce.
e, E -
excitation of the singing formant, low intensity and full intensity,
respectively.
v, V -
vibrato - low vibrato extent (amplitude), and full sustained stage's value,
respectively.
i - a gradual increase.
Example: iv - a gradual increase in vibrato extent
from v to V.
t, T -
transition (small and large extent, respectively): a gradual pitch increase
from the onset to the sustained state, with concave shape (slope increasing
with time).
k, K - a
gradual pitch increase with convex shape - slope decreasing with time, small
and large extent, respectively.
j, J - jitter (small and large
extent, respectively).
s - a symbol preceding
another symbol denoting that the latter is of small extent. Thus sv is still of
smaller extent than v, se is of smaller extent than e.
w - a wiggly tone of excessive jitter having a
structure more pronounced than jitter, and less than vibrato - intermediate
between jitter and vibrato (Fig. 8).
- a symbol denoting stopping
or termination of the mechanism whose symbol is to the left of the ";"
symbol. Thus T; denotes the termination of the transition stage. Two mechanisms
that stop simultaneously are to be enclosed in brackets followed by the ";"
symbol. Example: (ET);.
_ - denotes the sustained state. All the symbols to
the left of the = sign continue to operate at the sustained state (unless
explicitly denoted by All symbols to the right of the = sign belong to the
tonal decay stage.
d, D - decreasing pitch within the
tone, small and large extent, respectively. o - vibrato
starting at phase between +ð/6 and
ð/4 (R category).
F - tone ending in a fuzzy
way due to a fricative consonant or (phonetic) stop.
L - legato connection of two
tones. All the symbols to the left of the L sign continue to operate.
[, ] - beginning of the tone, end of tone,
respectively. (To be used if needed, as sometimes in the case of a legato
connection).
Processes
that take place simultaneously are denoted without spacing. Thus pe denote low
intensity phonation and excitation occurring simultaneously, whereas two
mechanisms whose onset takes place one after the other, with at least a 150 m
(1 B) separation in time between onsets, are denoted by a hyphen separating
For more quantitative description of the (temporal)
structure and multistage development of the tone - the time duration in seconds
of each stage can be inserted. Example: a tone; PZE-T;V=D]
can be written PZE0.14-T0.32;V1.2D0.51]. The equivalent, expressed in B units,
is: PZE1-T2;V8-D3] - (see section on
"Transition").
Notation for
Short Tones (S category)
Modes
of the Short (S) category are shown schematically in Fig. 2, along with the
corresponding notation, explained in the following:
A, a - a short tone (unit pulse or elementary beat)
consisting of an abrupt increase in pitch followed by a steep decrease, or
release - visually resembling an inverted U, or a vibrato period from -7E/2 to
37U2 large and small extent, respectively (Fig. 10). It is considered as the
fundamental unit.
m, M - a
short tone resembling the small letter m, consisting actually of a double beat
(Fig. 10).
m3, M3 - an m type tone comprising three beats. Small
and capital letter denotes (as before) small, and large extent, respectively.
! - a cut. Thus, !a denotes a beat "a" whose initial (pitch
increase) part is cut - unphonated - the phonated tone starts at phase 7c/5. Similarly,
a! denotes the beat having its terminal part cut by
n/5 (unphonated), similarly for !m and m!. These modes appear in singing
staccato tones in rapid tempi.
+ - stretching (along the time axis). +m will denote a
(double beat) m mode of which the left beat is stretched or extended, as
compared to its right beat, and the reverse situation is in m+.
Postulates
and Principles
Vocal, and other musical tones are
acoustic signals (periodically-varying air pressure pulses) and are fully
characterized by the physical entities: frequency, amplitude, duration - or
more precisely: the ensemble of the fundamental frequency and harmonic
partials, intensity, and duration.
Emotions
and expressiveness are encoded in the acoustic signals (the vocal tones) by
introducing changes and complexities, particularly as frequency changes and
modulations, periodic or nonperiodic, thus shaping the tone's temporal and
spectral structure. Every feature has meaning and nothing is haphazard.
All
the acoustic information is contained in the FFT spectrogram down to the most
subtle detail.
The
mechanisms 1-7: phonation, vibrato, excitation of higher harmonic (singing
formant), transition, sforzando, pitch change, and unit pulse - whose role in
shaping the vocal tones' temporal structures was elucidated in the present
study, corresponding to the various types of frequency modulation encoded
emotional messages.
The
zero of the hierarchy is a tone devoid of any expression or human emotion such
as, a sound produced by a mechanical or electrical oscillator, a violin played
by a motor-driven bow or a clarinet blown by a mechanical pump. The human voice
always has some jitter endowing it with human warmth (Rasch 1984, McAdams
1982).
The
universal rules of excitement and calmness (Cohen 1971, 1983) in th4 simple
form, relating abruptness to excitement, and gradual change to calmness or
relaxation, are valid down to the microlevel of an individual tone and forn the
basis for the classification of the singing modes in categories according t( a
hierarchical scheme. A tone having a gradually developing structure i
associated with an emotional state of a calm and subdue nature, whereas a tone
of a more abrupt structure is associated with an emotion of more excited nature
Mutual Exclusivity - signals of contrary type should not occur in the same
musical phrase expressing a certain definite emotion, unless, intentionally,
the two signals are explicitly meant to express contrary emotions (such as
momentary tension or anxiety in a calm musical phrase).
The
smaller the number of the operating mechanisms, and the more graduall: they
enter into operation - the more calm the mode will be. Inversely, the mor mechanisms operating together and more abruptly - the
more excited the mode (and the emotion expressed) will be.
Classification
of the modes and the hierarchical scheme
The
principles, and in particular No. 6, and No. 8, with the help of the notatiol
enable us to classify the singing modes into categories and to construct a
hierarch, from first principles, simply on the basis of the number of operating
mechanism and their way of operation (gradual - sequential, or abrupt -
simultaneous) - ii the following way:
1. One operating mechanism: phonation (p or P). This
is the uninteresting zer, mode. Addition of jitter produces modes: pj, pJ and
PJ, and even those ar rather rare.
2. The lowest category in the hierarchy will comprise
two mechanisms: phonatio + excitation; this leads to the modes: pj-e, pje,
pje-E, PjE, and PJE. To thes we add modes of another two-mechanism combination:
phonation + vibrato
3. Three operating mechanisms: phonation + excitation
+ vibrato, operating gradually and in succession: this leads to the C category,
with modes such as p-e-v-EV, pe-v-EV, peiv-VE, PE-v-V, P-E-V, and pV-e-E. Modes
pev and pevEV, might be included in the C category, but are preferably
included in the R category.
4. Three operating mechanisms: phonation + excitation
+ vibrato, operating together. This leads to the Expressive (R) category (see
below), with modes pev-EV, pev, opev, oPEV, and PEV.
5. Four mechanisms: phonation + excitation +
transition + vibrato. This leads to the modes of the T category: T1 modes;
pet;V=], PET;V=], PET1;T2;V=], T2 mode: PE-T;V=] (with a steep T part), and the
G subcategory modes: pevt;V=], pevT;VE=], pieivT;VE=], PEivT;V=], and modes
terminating in portamento: PET;V=D], PEivT;V=D], for example.
6. Five mechanisms: phonation + sfozado + excitation +
transition + vibrato. This leads to more vigorous modes of the T and G
category: PEZ;-T;V=], PEZ;-TV-T;=], and G modes
PEZVT;=], PEZI;TVZ2;-T;=], and others.
7. Four mechanisms: phonation + sfozando + excitation
+ vibrato operating simultaneously. This leads to the modes X1 (PZEV), and X2
(PzEV) of the X category; the strongest in the hierarchy. These modes can
terminate in descending portamento: PZEV=D], and PzEV=D].
All
these modes, and many others, occur in actual singing.
OUTSTANDING
CATEGORIES: R AND S
These
two outstanding categories deserve special discussion as they exhibit some outstanding
features still more subtle than those in the other categories.
The
Expressive (R) category
This
category was arrived at from theoretical considerations. Let us consider
"pev" modes of the type pev, pev-EV, up to PEV, with the three
mechanisms operating simultaneously. While pev and pev-EV could be considered
as being the least gradual of the C category, mode PEV might be considered as
an X mode less the Z element, and therefore expressing more excitement than a C
mode, but some release of tension compared to X modes. As Seashore (1938)
already remarked on the role of vibrato in endowing pleasantness and
expressiveness to the tone - these pev modes are definitely more expressive
than C modes. and of less excitement
Another consideration is the initial phase of a
"pev" mode. It might be that an initial phase of
ð/2 changes the
character of the mode to a brighter, lighter, or more joyful - or at least less
somber or emphasized than vibrato starting with phase 3/2ð (or -ð/2). Up to
now, the subject of phase was not treated except in sforzando in X modes, where
the first vibrato cycle is sharper than the others, being actually an A unit
pulse (see S modes), and the tone starts actually on the low pitch side. In
arias in Mozart Le Nozze di Figaro, pev modes with initial phase
ð/2 (to be
denoted oR, or opev modes) were encountered in rather joyful emotional context,
and also in the drinking song (Brindisi) in Verdi La Traviata - leading to the
conclusion that the initial phase of the first vibrato cycle has expressive significance
- this is in the region of the "attack" of the tone. All these led to
the conclusion that the pev modes belong to a distinct category - structurally
and emotionally.
The Short
(S) category
1. w modes
w modes are modes with features in
between mere jitter and vibrato, and are essentially N modes that more
excitement was built into them in the form of excessive jitter. An example of w
modes is presented in Fig. 8. The effect of time counting in unit pulses
(section on "The Unit Pulse") is already apparent in w modes, and we
can distinguish w2, w3, and w4 modes consisting of two, three, and four beats,
respectively. These are precursors to fully developed unit pulse modes M, M3
(and a 4-fold unit).
2. Unit pulse S modes:
A careful examination of a number of arias and lieder of fast tempi: Cherubino's aria "non so piu cosa son cosa faccio" from Mozart Le Nozze di Figaro, Riccardo aria "E scherzo od a follia..." from Verdi Un Ballo in Maschera, Schubert lied Der Musensohn, and others reveal that the dominant factor in such arias or lieder of rapid tempi is rhythm, and repetitive rhythmical units. From the temporal structural point of view we already encountered the existence of the two units: the A and a modes as a single unit, and the m and M modes as a double unit. The m3 (or M3) modes are also important in this respect, considered here as a triple unit comprising three single A or a units. These units were found to play a dominant role in this type of singing. Furthermore, many of the other short modes are simply derived from these two basic units.
Referring to Fig. 2 and the schematic presentation of the modes of the Short (S) category with the corresponding notation, it is easy to comprehend how the various S modes are derived from the a, A, m, and M modes (and sometimes also from the m3 and M3 modes) by the operation of the !, T, t, D, d, and + operations. Referring now to Cherubino's aria, the basic repetitive rhythmic unit is two eighths and a quarter, (1/8 1/8 1/4) sung by Frededrica von Stade as an (a a m) rhythmic unit (in many variations, like "chopped" a!, !a, !m, m! etc. which are due to staccato and phonetic consonants). In "E scherzo od a follia..." the repetitive rhythmical unit is a sixteenth and an eighth (1/16 1/8) sung by Pavarotti as a repetitive (a m) unit. In Schubert lied Der Musensohn a repetitive small rhythmic unit (1/8 1/4) is repeated twice and combines with a third unit (1/8 3/8) to form a (1/8 1/4 1/8 1/4 1/8 3/8) larger rhythmic unit that is repeated nine times in the first melodic stretch of the lied. These are sung by Barbara Hendricks as (a m a m a m3) repetitive rhythmic units. The m3 beat has always the rather peculiar shape of the structure !aAAa! (or oaAAa!). In these examples (Performance Score No. 6 - Cherubino, and No. 7 - Hendricks) von Stade and Hendricks count in eighths, and Pavarotti in sixteenths. It seems that in such arias, due to the rapid tempi, control is switched to the direct command of the brain, the latter activating a special pacemaker counting mechanism (oscillator) producing a special pulsetrain - operating as an "automatic pilot". The singer, unawarely, is still in some control (phonation) and is capable of some variety and expressiveness in this staccato singing by "chopping" and manipulating the unit pulses by use of the !, d, t, and + operations.
PERFORMANCE
SCORES
With the results of the present study; singing modes, categories, hierarchy, and notation, we have new tools to return to Seashore's idea of the performance score - a very useful tool indeed for the analysis and comparison of performances and interpretation by various artists. A performance score is constructed as follows: above each note in the vocal score the mode category (from the FFT spectrogram) of the singing mode is written. By marking different categories in different colors, one can have a very clear visual differentiation among the various modes. The correspondence between the singing modes and the notes in the vocal score, the text, and the orchestral or piano roles with their harmonic and chordal structure can be analyzed and compared to Schenker schemes. Performances of different artists can be compared note to note on the same vocal score. The use of the detailed notation can reveal details and subtleties. We present here seven performance scores to illustrate the role of the various mode categories in the melodic phrase. Modes shown in Figs. 3, 5, 8, 9, and 10 are marked in the corresponding scores.
Performance
Score No. 1
Score
No. 1 is the phrase "I fior
Performance
score No. 1
Performance score No. 2
Score
No. 2 is the opening of the aria of Amelia: "Morro, ma prima in grazia..."
from the third act of Un Ballo in Maschera by Verdi,
where the singing of Margaret Price and of Maria Callas are compared. Margaret
Price singing expresses deep sorrow and sadness, accepting with resignation her
tragic fate. This five-phrase excerpt starts in the C-N salient-ornamental
combination, except for the emphasis on the second syllable at the very
beginning - in oR mode. In the second phrase as tension increases the N modes
are replaced by w4 and w3 modes. The third phrase, with "1' unico figlio
mio", has R, I, I and X, and zG2 as salients, and m modes as ornamentals.
The fifth phrase returns to resignation in the C-N combination. Maria Callas
expresses more pathos and even protest, as she does not
Performance
score No. 2.
Performance
scores No. 3 and 4
Score
No. 3 shows the opening of Riccardo's aria "La rivedra
nell'estasi..." from the first act of Un Ballo in Maschera by Verdi, sung
by Luciano Pavarotti, opening in a T1 mode, with three more T1 modes at the
three climaxes of this excerpt. In this aria Riccardo expresses his love and
admiration for Amelia and her beauty, and the T modes with their brilliance do
exactly this - indeed imparting to the aria its exhilaration and ecstasy. Score
No. 4 - "Si, rivederti, Amelia" is essentially the same aria repeated
in the fourth act in F major - but in a totally different emotional context.
This time it is an outburst of emotions with a strong determination to go for
the last time to see Amela and admire her beauty at the risk of the threat on
his life. There is much more tension and excitement in this aria, and this is
expressed by Pavarotti singing the salients and climaxes in X and strong T2
modes and the ornamentals in w and m modes.
Performance score No. 5
Score
No. 5 is another one-line phrase "Ah voi, signori..." of the fortune
teller Ulrica (Christa Ludwig) from the first act of Un Ballo in Maschera by
Verdi. This phrase follows Riccardo's aria "E scherzo od a follia",
sung by Pavarotti in a large variety of a and m modes,
mocking and making ridicule of Ulrica's prophesy. Ulrica addresses the
conspirators among the crowd singing in a very somber mood, asking them if her
words dismay them and that they, knowing what in their heart, they would not
dare ridicule her. The phrase opens in five ascending eighths sung in w3 and w2
modes (the unit pulse is also operational in w modes as can be seen in Fig. 8).
The rhythmic pattern is that w2 and w3 modes alternate. This sequence repeats
for the second time exactly the same notes, no change in loudness except for a
very subtle accentuation, but in remarkably strong M3 and M modes. This time
the text is: "voi non osate ridere" (you dare not laugh). The
emotional context of this very short phrase is responsible for this remarkable
change in the modes. The tension in the text and the dramatic situation are
translated directly to abruptness in the tightening of the vocal folds, as can
be seen in Fig. 9, showing the second five eighths sequence.
Performance score No. 5.
Performance
scores No. 6 and 7
Score No. 6 shows the beginning of Cherubino aria "Non so piu cosa son cosa faccio..." from Le Nozze di Figaro by Mozart, sung by Frederica von Stade, and Score No. 7 shows the beginning of Schubert lied: Der Musensohn, sung by Barbara Hendricks. These scores are sung in a variety of a, m and m3 modes, and were described in the section labelled "The Short (S) category".
Performance score No. 7.
APPENDIX:
A LIST OF ARIAS AND LIEDER STUDIED IN THE PRESENT WORK
Schubert
Lieder
Singers: (1) Felicity Lott, Innovative Music Productions Ltd. IMP-PCD 898. (2) Barbara Hendricks, (a) EMI CDS 754909 2, (b) EMI 7475492. (3) Gundula Janovitz, DG 437943-2. (4) Dietrich Fischer-Dieskau, Philips 411051-2. (5) Brigitte Fassbaender, Sony Classical SK 53104. (6) Kathleen Battle, DG 419237-2. (7) Mitsuko Shirai, C