Evan Jones

Assistant Professor, Music Theory

School of Music, Florida State University
Tallahassee, Florida 32306 USA
jones_e@music.fsu.edu

Iannis Xenakiss Nomos Alpha for solo cello (1966)

exploits the three principal methods of sound production on a cello stringbowing, plucking, and
col legno (hitting the string with the stick of the
bow)to a remarkably extensive degree. Each of
these methods of sound production has been the focus of acoustic research. Conceptually, the simplest of these three sound production methods is
the plucked string, the physics of which is summarized in Backus (1977) and elsewhere. The spectrum of harmonic partials activated by the action
of plucking depends upon the location along the
string where plucking occurs. The sinusoidal components of the resulting waveform will reect, in
their respective amplitudes, the degree to which
the point of plucking approached or coincided with
their own peaks and nodes. The displacement of
the string caused by plucking forms a hump that
propagates along the string; upon reaching one end
of the string, the hump is reected to form a dip
that travels back in the reverse direction. Through
Fourier analysis, this rapid circulation of humps
and dips can be reduced to the harmonic partials
that comprise, in varying amplitudes, the spectrum
of the generated sound.
The idealized image of the waveform produced
by bowing a string is that of a sawtooth shape (produced by the bow sticking to and slipping against
the string in quick alternation) in which the amplitude of the nth harmonic is that of the fundamental divided by n. In order that no slipping occurs
during the long period of sticking during a bowed
stroke, a minimum bow force is required, which
rises in proportion to an increasing bow speed but
which tends towards a nite non-zero minimum at
very slow bow speeds (Schelleng 1973). The
minimum-required bow force can as much as double when the fundamental frequency coincides
with the air resonance or the principal wood resoComputer Music Journal, 26:1, pp. 7386, Spring 2002
2002 Massachusetts Institute of Technology.

An Acoustic Analysis
of Col Legno Articulation
in Iannis Xenakiss
Nomos Alpha

nancecommonly known as wolf tones (Raman

1918). Using a machine described by Saunders
(1937), Schelleng (1973) identied a dropoff from
perfect inverse frequency harmonic strength between the fth and tenth partials. Response was
found to fall off sooner with less exible strings or
bow pressure closer to the minimum.
The physics of a struck string, however, has been
studied more from the perspective of the mechanism of a piano than that of col legno playing on a
string instrument. Studies such as Bacon and
Bowsher (1978) and Suzuki (1987), for example, focus on issues not entirely relevant to the situation
of a col legno attack on a cello. For example, they
discuss the hardness of the hammering agent,
which in the case of a piano hammer is assumed always to be less hard than a cello bow stick. Certain
aspects of sound production are common to both
instrumental contexts. When a string is plucked or
struck, for example, a peak is created at a point in
the wave-action of the string where some of the
sounds harmonic components would otherwise
have a node. This has the result of nullifying those
components, a process known as harmonic damping. In the case of a bowed string, however, the effects of harmonic damping are circumvented.

Parametric Extremes in Xenakiss Nomos Alpha

An acoustic description of Nomos Alpha is of interest because of the wide variety of extended techniques that the composer juxtaposes in quick
succession. In this piece, Xenakis achieves extremes of affect (setting new standards of potential
for what might be characterized as eerieness, aggression, and hyperactivity), extremes of register
(exploiting a range of over seven octaves, retuning
the lowest string down an octave to C1 32.7 Hz
and vaulting as high as E8 5274 Hz), and extremes of timbre (from the composers dictum forJones

73

Some of the most extreme effects in Nomos Alpha

seem to be linked together in the music in a perhaps purposeful web of allusions between acoustic
event-classes. Although this article only explores
the acoustic properties of one end of this continuum, I would propose a critical reading of the
acoustic space inhabited by the pieceone that ties
together diverse textural and timbral elements
based on a phenomenological criterion of adja-

cency. That is, a certain listening strategy is assumed whereby a series of resemblances are
noticed between adjacent timbral or textural entities, and these are judged to be valid connections in
part by virtue of their very adjacency. Though this
reading will nd a fuller exposition in a subsequent
study, I will make brief allusion to it here to contextualize the present study.
At the center of the web of interrelations between acoustic phenomena, I locate the beatcount successions described above, in which
Xenakis indicates the exact number of beats per
second that should be heard at each of several
time-points through the course of a sustained
double-stop consisting of two notes very close in
pitch. Perhaps because of their striking nature, Xenakiss beat-count successions comprise the hub of
my analytic reading of Nomos Alpha, around
which at least three other acoustic phenomena revolve. Xenakis frequently pairs such beat-count
successions with double-stop glissandi, which are
ubiquitous throughout Nomos Alpha and occur
with both tremolo and legato articulations. The
two phenomena share important identifying characteristics and make similar technical demands of
the performer. We may draw a further connection
between the beats themselves and the unnaturally
low, slow tunes played on the cellos lowest string
(which is re-tuned down an octave).
Finally, Xenakis makes use of a third acoustic
phenomenon, a recurring col legno texture, denoted
consistently in the score by the abbreviation fcl
(frapper col legno). In his legend, Xenakis notes
fcl frapper avec le bois de larchet (to hit the
string with the stick of the bow). Figure 1 illustrates this technique in temporal proximity to several beat-count successions and to a double-stop
glissando immediately preceding the rst fcl indication. In Figure 1, the beat-count successions are
notated with double staves, and that the numbers
between the staves indicate the desired beats per
second created by the given de-tuned double-stop.
Although any of these acoustic phenomena would
make worthy subjects of a timbral analysis project
such as this, I have decided to focus exclusively on
what is perhaps the least traditional timbral eventclass in the piece: the recurring col legno texture.

74

Computer Music Journal

bidding vibrato completely to the frequent

employment of col legno, sul ponticello, and tremolo effects). Given the inclinations to disorder in
the variety of effects in Nomos Alpha, it is not difcult to agree with Grifths (1975), who questions
the signicance of the logical constructions governing both in-time and out-of-time pitch procedures
in the piece. A detailed explication of these procedures is provided in Formalized Music (Xenakis
1992).
Xenakis even attains extremes of pitch-space delimitation: witness the frequent glissandi of various
speeds and scopes, frequent quarter-tone indications, and frequent notation of a precise number of
beats per second to be heard in slightly mistuned
unison double-stops. In conventional styles, such
beating would usually be disguised by vibrato, but
since Xenakis forbids vibrato, the beats are more
readily perceptible here. However, Xenakiss notated requirements of varying numbers of beats per
second between the two notes of a near unison
(which I call beat-count successions) would seem
be at odds with Benades view of the difculty of
producing beats between string instruments
pitches. Owing to jittera variation by as much
as a fth of a semitone in the pitch of a bowed
notestring ensembles blend comparatively effortlessly and attain a characteristically warm sound
without offending the ear with clashing beats (Cremer 1973). Notwithstanding this jitter, however,
such closely tuned double stops as indicated in Nomos Alpha are easily heard to ebb and ow closer
and farther apart, and the beats themselves are often clearly perceptible.

A Proposed Acoustic Event-Class Space

Figure 1. Rhythmic col

legno attacks, beat-count
succession: Nomos Alpha,
page 8, lines 23.

My agenda in this examination of Xenakiss use of

fcl sounds is to reveal exactly what acoustic
component-phenomena this manifestation of extended technique, the col legno, entails.

Realizations of Col Legno: Taps and Slaps

An elaboration of the possible acoustic products of
the fcl technique is necessary because there are two
distinct, perhaps compositionally unanticipated, aspects of the resulting sound. The most typical
sound of col legno involves the gently percussive
articulation of a brief high note, to which I will refer as a tap. The pitch of such a tap corresponds
to the distance from the bridge at which the bow
comes in contact with the string, and is the same
pitch that would be heard if a cellists nger had
been placed at the same point and the remaining
few inches of string were either plucked or bowed.
Taps are especially easily audible over several successive col legno attacks on the same ngered
pitch. They change over a very wide pitch range
over only a few inches of string, of course, since
the bottom two inches of the A string on the cello
can produce any note from about C6 to beyond the
threshold of human pitch perception. In addition to

taps, however, a rather aggressive striking with the

wood of the bow on either an unstopped string or a
string ngered no higher than approximately a perfect fth above the open string results in the noisy
bouncing of the string against the ngerboard. This
is a much more aggressively percussive slap
sound, and it characterizes many instances of the
fcl sounds in recordings of Nomos Alpha.
The pinpointing of either tap or slap sounds on
printouts of the evolving harmonic spectra of Xenakiss fcl segments on commercial recordings of
Nomos Alpha is very difcult and usually requires
the isolation of a highly localized instant of sound
to observe. To identify both tap and slap sounds
graphically, I will be examining brief excerpts from
Rohan de Sarams recording of Nomos Alpha, issued in 1991 as part of the Arditti Quartets comprehensive 35-year retrospective of Xenakiss
chamber music (Xenakis 1991). Many listeners,
even string players familiar with col legno technique, may nd it difcult to pick out the various
sonic components of col legno sounds. (Similarly,
while most people are capable of identifying, classifying, and producing sustained vowel sounds, most
will fail to perceive even the lower partials as distinct from the fundamental frequency of a sung
Jones

75

Figure 2. Nomos Alpha,

page 5, line 6.

Figure 3. Waveform representation of rst attack in

Figure 2 excerpt.

Figure 4. Waveform representation of all of Figure 2

excerpt; melody of tap
tones notated in pitch
names.

Figure 2

Figure 3

Figure 4

vowel.) I am motivated, then, by the unawareness

of tap/slap components among musicians, including (presumably) most composers who employ col
legno techniques, and even string players, to whom
these complicated col legno sounds are intimately
familiar.

Taps

76

Computer Music Journal

My examination of tap sounds on this recording begins with the excerpt reproduced in Figure 2, from
the bottom of page 5 of the score. The instantaneous shifts in dynamic level, combined with the

Figure 5. Fast Fourier

Transform of Figure 2
excerpt.

steady sixteenth-note tempo to which the composer expects the fcl attacks to adhere, require a
dexterity of bow control in manipulating the contact point, particularly to produce the rapid
changes of volume as directed. Once the initial fff
dynamic is discarded for the ensuing p, Mr. de
Sarams unusually prominent tap-components indicate where his bow is striking the string in an eminently audible melody of tap sounds over the
course of the A-sharp-and-a-half (three quartertones above A) and the G-sharp in the second measure of Figure 2. Each note of this melody seems
almost to precede the articulation of the ngered,
notated pitch; this particularly clear in Figure 3,
which reproduces the waveform representation of
the rst attack of the A-sharp-and-a-half. This
melody is notated with letter names in Figure 4,
above the waveform representation of the recorded
performance, but below other letter names describing the ngered pitches given in the score.

One can deduce from the taps pitches that Mr.

de Sarams bow begins on the string around where
most cellos ngerboards end (F6), recoils a little (as
far from the bridge as D6), and then pursues the
higher dynamic marking on the G-sharp by approaching the bridge (as close to the bridge as B
6),
nally recoiling again in anticipation of the quieter
dynamic after the barline. However, it is very difcult to conrm ones own aural experience convincingly by comparison with a Fourier analysis of
the recorded excerpt. Figure 5 reproduces part of
the Fast Fourier Transform (FFT) of this long
stretch of music. A few peaks that coincide with
the melody of taps taken in dictation
are discernible: several of the taps around F6
1400 Hz show up as the highest peaks (as marked)
at time points 1.05 sec, 1.8 sec, and 2.5 sec, corresponding very closely to the points at which Mr.
de Sarams tap-melody is heard to intersect that
pitch level. But the veritable forest of fcl attacks
Jones

77

Figure 6. Nomos Alpha,

page 6, lines 12.

Figure 7. Waveform representation of second-last

attack in Figure 6 excerpt.

Figure 6

Figure 7

conspire to obscure any continuous, clear-cut

curve one might have hoped to see from 1400 Hz
down to 1175 Hz, up to 1865 Hz, and back down
to 1175 Hz.
It is necessary to restrict ones purview drastically to be able to conrm ones aural sensations
through computer sound analysis. Figure 6 reproduces an fcl excerpt from the top of the sixth page
of the score. Figure 7 provides the waveform representation of the second-last attack in this excerpt
on Rohan de Sarams recording, in which the rapid
higher-frequency oscillation is apparent at the outset of the sound. Over the course of this particular
fcl section, the pitches of the tap sounds migrate
gradually higher (in tandem with the tempo of the
rhythmic articulations), beginning around E7
2637 Hz, and nishing (by the time of Figure 7)
around G  7 3322 Hz. While this nal frequency
is not immediately apparent in the waveform repre-

sentation, it is easily observable on the graphic representation of its FFT (given in Figure 8) at about
32003600 Hz. Moreover, this frequency has somewhat of a timbral personality, evidenced by its own
apparent harmonic spectrum, consisting of incontestable overtones to the tap sound itself at frequencies of 65006800 Hz and 1000010500 Hz.
This harmonicity conrms the validity of the aural
sensation of the tap-component in the fcl articulation.
I have reported the same excerpt to have begun
with tap sounds at approximately E7 in Rohan de
Sarams hands; one hopes that similarly encouraging evidence might appear in an FFT of the other
end of this excerpt. Indeed, as shown in Figure 9,
another spectrum of at least two harmonics (approximating 25002700 Hz and 51005300 Hz) correspond to the perceived tap pitch of E7 at the
outset of this fcl gure. These two FFT graphs

78

Computer Music Journal

Figure 8. Fast Fourier

Transform of second-last
attack in Figure 6 excerpt.

Figure 9. Fast Fourier

Transform of rst attack
in Figure 6 excerpt.

Figure 8

Figure 9

Jones

79

Figure 10. Nomos Alpha,

page 6, line 5.

Figure 11. Waveform representation of two adjacent

fcl attacks from middle of
Figure 10 excerpt.

Figure 10

Figure 11

serve as mutual conrmations, propelling the phenomenon of tap sounds beyond the status of aural
illusion.

The slap sound is much less easily denable in

acoustical terms than is the tap. Very similar to the

characteristic slap of a single snare-drum articulation, the slap sound is most familiar to string players as the dening sonic attribute of what is known
as Bartok pizzicato, a style of plucking the string
so as to displace the string along an axis roughly
perpendicular to the direction in which the string
runs. When released, the string snaps against the
ngerboard with a sharp, explosive report. Compo-

80

Computer Music Journal

Slaps

Figure 12. Fast Fourier

Transform of the two
fcl attacks graphed in
Figure 11.

nents that are likely part of the resulting sound

whether produced with Bartok pizzicato or with
an aggressive col legno attackinclude the resonating of the wood of the ngerboard in response to
the strings impacts (possibly more than one), the inharmonic vibration of the string after one or more
interactions with the ngerboard, and the inhibition
of certain harmonics of the pitch of the string.
This discussion of the various traces left by slap
sounds is, however, necessarily restricted to a narrow focus, stemming exclusively from simple visual comparisons of the graphic representations of
the sounds themselves. Just as the tap sounds discussed earlier are not necessarily heard on each of a
series of col legno attacks (through which, never-

theless, an audible overall rising or falling of the

taps pitches gives a picture of how the bow sticks
contact point progresses along the string), so too
may a slap sound occur on one of a series of such
attacks and not the next.
Figure 10 reproduces a brief excerpt from page 6
of the score. Figures 11a and 11b contrast two temporally adjacent fcl attacks from the middle of this
excerpt on Rohan de Sarams recording, the rst of
which (around time 1.13 sec) has no slap component, and the second of which (around time 1.40
sec) has a high degree of slap. As seen earlier in this
study, the high-frequency attack noise seems to
precede the establishment of the deeper frequencies
associated with the pitch of the ngered note. FigJones

81

Figure 13. Fast Fourier

Transform of two adjacent
fcl attacks from middle of
Figure 14 excerpt.

ure 11b clearly features a much noisier attack,

which is itself seemingly preceded by some aspect
of the attack at a high frequency but low amplitude. The low-amplitude, high-frequency components observable at the outset of each example
might correspond to the initial impact of the bow
stick against the string, and the explosive report
that follows on its heels in the case of Figure 11b
could conceivably correspond to the impact of the
string against the ngerboard. It might be proposed
that the amplitude of the resulting noise reects
the greater capacity for resonance of the ngerboard than either the string itself or the stick of the
bow. On the other hand, the motion of the string

will be disrupted to a much greater degree by its

impact against the ngerboard (occurring along a
long segment of the string) than by the bows impact against the string (occurring along only a few
millimeters of the string. This may account for the
greater impulsiveness and amplitude of the slap
against the ngerboard compared to the impact of
the bow against the string.
These same two attacks are subjected to FFTs in
Figures 12a and 12b (which are, respectively, the
FFTs of Figures 11a and 11b). The peaks of Figure
12a are duplicated very faithfully in Figure 12b. Although those in Figure 12b appear to be of lesser
amplitude (and less jagged), the reality of the situa-

82

Computer Music Journal

Figure 14. Nomos Alpha,

page 1, line 4.

Figure 15a. Nomos Alpha,

page 7, line 4; page 8,
line 1.

Figure 15b. Fast Fourier

Transform of part of Figure
15a excerpt.

Figure 14

Figure 15a

Figure 15b

tion is that 12b simply contains more noise than

12a. Notice the near total absence of at valleys
in 12b reaching as far down as the zero-plane,
whereas such valleys are quite prominent in the
upper third of 12a. It would seem that the slap of
Figures 11b and 12b is less harmonic an event than
those illustrated in Figures 11a and 12a.
A similar result is gleaned from a comparison of

Figures 13a and 13b, which show the FFT analysis

of two temporally adjacent attacks from a different
excerpt (Figure 14). Rohan de Sarams rendering of
Figure 14 contains even more starkly differing attacks than seen previously in Figures 11 and 12; as
Mr. de Saram subjects a greater and greater length
of the C string to his fcl attacks, his realization of
Xenakiss crescendo makes an increasing proporJones

83

Figure 16. Magnication of

two frequency regions
from Fast Fourier Transform in Figure 15b.

tion of slap inevitable. There is no way to achieve a

col legno attack at a dynamic level of fff on an
open C string (the lowest of the four strings), or on
a low ngered note on the C string, without the
string impacting the ngerboard to some degree. In
the midst of his crescendo, Mr. de Saram proceeds
from the spectrum of Figure 13a to that of Figure
13b in a fraction of a second. Much of the lowerfrequency harmonic material is similar between
the two FFTs: the peak around 550 Hz, for exam-

ple, is prominent in each graph. The two FFTs,

however, rapidly assert their differences. While Figure 13a contains practically no harmonic material
above 1500 Hz (the hump at 2150 Hz is perhaps a
tap component), Figure 13b hardly reaches the zeroplane at all.
Any attempt to narrow down the nature of all
this noise might begin by isolating some of the frequencies involved. Despite the continuous, random, indistinguishably noisy aspect conveyed by

84

Computer Music Journal

Figure 16. Continued

both Figures 12b and 13b, individual frequencies

are discernible with more careful applications of
computer capabilities. Figure 15a contains an excerpt spanning pages 78 of the score, which Rohan
de Saram realizes in quite a slappy manner, and
Figure 15b shows an exceptionally black FFT
representation of part of the excerpt. Not much
useful information is easily gleaned from Figure
15b, but the most distant horizon of Figure 15b (the
topmost contour) reveals a distinctly rolling prole.
These hills and valleys are discriminable into remarkably pronounced areas of resonance, as seen in
Figures 16a and 16b, which reveal formant-like
humps in the frequency prole at 21002400 Hz,
33003600 Hz, and 920010000 Hz. I hypothesize

that these areas of resonance are not the formants

of the instrument but are instead associated with
the modes of vibration of the ngerboard itself, in
reference to which I know of no published literature. The ngerboard to which I have immediate
access, however, seems to resonate at a fundamental pitch approximating 190 Hz. I would anticipate
that a similar set of inharmonic partials apply to its
modes of vibration as to a xylophone bar, which
like a cellos ngerboardis wooden and is characterized by a hollowed-out underside. As in the case
of the xylophone, the hollowing-out of the underside of a cellos ngerboard might have evolved
in order to attain a more harmonic set of natural
vibrations.
Jones

85

Conclusion

References

It is my hope that the present study will serve not

only to widen a compositional understanding of the
complex character of the col legno articulation, but
also to point towards the importance of the details
of a performative realization of a piece to the analysis thereof. A useful step for future research would
be to analyze other performances of Nomos Alpha,
as well as instances of the col legno technique in
other pieces, in order to reach a broader acoustic
understanding of this effect and to make such an
understanding available both to performers and to
composers. More generally, much research has yet
to be done into questions of the musical ramications of various instruments construction and
acoustic properties, in exploration of the musical
possibilities inherent in such acoustic ndings. It is
the task of a study such as the present one, I believe, to accommodate and synthesize not only the
bare facts of an instruments makeup and the range
of its output, but also the degree to which the composers understanding of the medium informed the
construction of the music, as well as issues surrounding the realization of structural aspects of the
music in performance.

Backus, J. 1977. The Acoustical Foundations of Music,

2nd ed. New York and London: W. W. Norton.
Bacon, R. A., and J. M. Bowsher. 1978. A Discrete
Model of a Struck String. Acustica 41:2127.
Cremer, L. 1973. The Inuence of Bow Pressure on the
Movement of a Bowed String. Catgut Acoustical Society Newsletter 19:2125.
Grifths, P. A. 1975. Xenakis: Logic and Disorder. Musical Times 116:32931.
Raman, C. V. 1918. On the Mechanical Theory of Vibrations of Bowed Strings. Calcutta: Baptist Mission
Press, Indian Association for the Cultivation of Science.
Saunders, F. A. 1937. Mechanical Action of Violins.
Journal of the Acoustical Society of America 9:8198.
Schelleng, J. C. 1973. The Bowed String and the
Player. Journal of the Acoustical Society of America
53(1):2641.
Suzuki, K. 1987. Model Analysis of a Hammer-String
Interaction. Journal of the Acoustical Society of
America 82(2):11451151.
Xenakis, I. 1991. Nomos Alpha. Paris: Montaigne Auvidis MO 782005.
Xenakis, I. 1992. Formalized Music: Thought and Mathematics in Music, rev. ed. Stuyvesant, New York: Pendragon Press.

86

Computer Music Journal