In 1958 John Cage composed "Aria", for a voice of any range. 

This piece is dedicated to Cathy Berberian, one of the most talented voices in 20th century. However this composition has become an obligatory stage for a singer who is interested in contemporary music. 

The first thing that one can notice is the particular but at the same time simple musical notation traceable in the score: a sequence of curved lines each one roughly describing ⁽¹⁾ the pitch path requested for the emission of sound (see fig. 1). It's a relative pitch: the position of the curve on the sheet does not involve a frequency relation with previous or followings curves but simply points out a "change of state" and a time (time runs ideally from left to right). 

Moreover, through different colors placed on the curves, the score prescribes to the interpreter a timbre change. The different timbres are not predetermined by Cage on the score but chosen by the performer during the rehearsals. 

Cathy Berberian, for example, chose in this way: 

• dark blue = jazz; 

• red = alto; 

• black with a parallel dotted line = sprechstimme; 

• black = dramatic; 

• violet = Marlene Dietrich;

•  yellow = coloritura; 

• green = folk; 

• orange = oriental; 

• blue = baby; 

• brown = nasal.

In the score there are also black squares that indicate noisy events (with an undetermined pitch). The text, pronounced according to the intonation suggested by the curves, contains vowels, consonants, words and phrases in five different languages: Armenian, Russian, English, French and Italian. This is the complete text:

In the recorded version that we wanted to analyze Paul Hillier is helped by six singers, for a total of seven performers: everyone manages one or more timbres.

With the WaveSurfer software (version 1.8.2) written by Jonas Beskow and Kare Sjolander has been possible to analyze a fragment from the "Aria" recording performed by Paul Hillier with the Theatre of Voices ⁽²⁾.

This fragment lasts approximatively 1' 13'' and is taken by the "John Cage: Litany for the Whale" compact disc (Harmonia Mundi 907187). Wavesurfer described the pitch contour following these specifications:
algorithm ESPS;

• max pitch value: 4000 Hz;

• min pitch value: 50 Hz;

• analysis window lenght: 0.0075 s;

• frame interval: 0.01 s.

The obtained results, if compared with the score page regarding the analyzed fragment, show the extreme performance exactness by the Theatre of Voices.

Fig.1: the score page regarding the analyzed fragment.

Fig.2: WaveSurfer. Above there is the time-amplitude representation. Below the pitch contour.

It's easy to notice how the heights profile reconstructed by WaveSurfer is similar to the curved lines traced by Cage's hand. Observe the likeness between the second curve from left on the score and the pitch contour delimited by marker number 2 on WaveSurfer. See also the likeness between the fourth curve from left on the score and the pitch contour delimited by marker number 4 on WaveSurfer. Finally is also noticeable the likeness between the first curve from right on the score and the pitch contour delimited by marker number 12.

After the pitch contour we also analyzed the formantic profiles of the various voices.

First of all it's necessary to point out that in the analyzed fragment all 7 performers are involved.

According to the WaveSurfer markers and to the curves on the score, we can say that:

the first curve is relative to the first female voice: 1(F);

the second and third curves are relative to the first male voice: 1(M);

the fourth curve is relative to the second male voice: 2(M);

the fifth curve is relative to the third male voice: 3(M);

the sixth curve is relative to the second and third female voices: 2-3(F-F);

the seventh curve is relative to the second female voice: 2(F);

the eighth curve is relative in part to the fourth female voice 4(F) and in part to the second male voice 2(M);

the ninth curve is relative in part to the fourth male voice 4(M) and in part to the second and third female voices 2-3(F-F);

the tenth curve is once again relative to the second female voice: 2(F);

the eleventh curve is once again relative to the first male voice: 1(M);

the twelfth curve is once again relative to the first female voice: 1(F).

This is, resuming, the order of the entries:

1(F) 1(M) 1(M) 2(M) 3(M) 2-3(F-F) 2(F) 4(M) 2(M) 4(M) 2-3(F-F) 2(F) 1(M) 1(F).

Reconstruction of this order has been possible also through the analysis conducted on the first four formants of the voice of each performer.

Fig.3: first formant of the different voices.

Fig.4: second formant of the different voices.

Fig.5: third formant of the different voices.

Fig.6: fourth formant of the different voices.

In figure 3 the red profile indicates the variation in time of the first formant frequency.

The blue profile on figure 4 indicates, instead, the variation in time of the second formant frequency.

And so on: the brown on figure 5 refers to third formant and the black on figure 6 refers to the fourth.

All the graphics here transcribed indicate on the x axis the time in frames 10 ms long; on the y axis the frequency in Hz. They have been produced through the TclTk script "Pitch And Formant Viewer", written by Lorenzo Seno. This script translates in bidimensional plots a series of formants and pitch contour data included respectively in .frm and .f0 files generated through the TclTk "SNACK" package.

Moreover the "Pitch And Formant Viewer" interface allows, for every formant, to limit the frequency bandwidth (BW%) relevant for the graphic realization: a good way to eliminate from the graphic not so relevant data.

Compare the timing of figures 3,4,5 and 6 with the timing of figure 2 to detect the belongings of a formantic profile to one of the 7 voices involved in the recording.

(1) But a better approximation than that of a five-line staff! - [back to text]

(2) A group of singers, including the composer Terry Riley. - [back to text]

Walter Cianciusi (2005)