~e~ns of the c~tting .angle of the tape.

There exist procedures whose application i

8
limited to certam musical demands, just as it is imaginable to develop a piece of m ·
1 . 1 . h" . 1 . US!c
exc usive Y Wit m. a smg e workmg method. Music and technique are so inter-related
that only by a umted effort can the artistic idea transcend the technical restrictions. STATISTIC AND PSYCHOLOGIC PROBLEMS OF SOUND

WERNER MEYER-EPPLER

1. Aleatoric Modulation
A process is said to be aleatoric (from Lat. alea=dice) if its course is determined in
general but depends on chance in detail. Calculation of these procedures can be effected
by statistical means. Musically, everything which is not 'written in the notes' is within
the aleatoric sphere. In traditional music, the significance of statistical questions has
been limited to the scientific investigation of the particular instrumental result attained
by an individual performer by the almost imperceptible variations of relevant musical
parameters, such as pitch, intensity, timbre and the duration of sound elements (16) 1 •
It has been correctly pointed out that up to the present all technical endeavours
to include these variations as a constructional element in the building of instruments,
have gone astray. As an irrefutable example we may cite the case of the cinema
organ.
At first appearance the critics seem to have all arguments on their side, but a closer
examination persuades us that a general human inadequacy cannot be blamed on to a
technical process. Fault is to be found not with the technical process but with the builder
of the organ, who could think of nothing more enterprising than mechanical regular
vibrato of amplitude or coupling of two pipes to produce beats. With the addition of
one more pipe, he would have been able to replace the sweet penetrating sound of the
cinema organ by a genuine 'choric effect'. 2
Certainly the resultant oscillatory modulation which is valid in all three dimensions
(pitch, intensity and timbre) is still far from corresponding to the aleatoric modulation
resulting from the playing of a string instrument. Some of its characteristics do, how-
ever, approach strict aleatoric modulation. This approximation may be considerably
extended by appropriate choice of frequency range and amplitude of oscillation in
the three pipes. It would nevertheless be misleading to attribute the ugly effect of the
regular vibrato to the technical 'perfection' of an instrument, thus attributing the bad
taste of the instrument builder to the nature of the technique. The contrary is true: from
a technical viewpoint instruments which correspond to an alleged public taste are
primitive and it will be seen that real perfection will eventually lead to instruments
which at the same time fulfil artistic requirements. The manufacturer is not to be excused
on the grounds that a greater degree of creative effort is demanded in order to find
better solutions.
Certainly it is in no way a simple task to effect a strict aleatoric modulation on a
mechanical basis with mechanical instruments. But it is impossible to discover why the
builders of electronic musical instruments insist on retaining the unbearable vibrato of
the cinema organ. It would require no additional technical expenditure to replace the
1 These numbers refer to the Bibliography at the end of the article.
2 The choric effect (swell on the great organ) may be described in terms of physics as a combined frequency and
amplitude modulation of internally independent spectral regions leading to an increased fusion of sounds.
54
55
 sinusoidal modulation by an aleatoric oscillation. Co~pletely adequate for this would
that the oscillation of amplitude and frequency may not be perceived as dynamic or
be a noise signal limited by filtering to its aesthetically most effective region around
pitch alterations. Gliding up and down of the tone whe~ the freque~cy ~s modulated
6 c.p.s. Apart from several patents registered in the Patent Office (e.g. 7) I have nowhere
stops for modulating frequencies of above 7 c.p.s. accordmg to exammatwns made by
found any evidence of this kind of modulation. The composer, however, who wishes to
Stevens and Davis (11). It does not, however, follow from this that a pitch movement
:vork with electronic means is not bound to look for an instrument manufacturer who
is to be observed in all cases of frequency modulation below t.1.is level. For this purpose,
IS prepared to build an aleatoric amplitude or frequency modulator into his gen-
as G. W. Stewart has shown with reference to the uncertainty relation of the conjugated
erators. In a studio he can produce these modulations for himself, and is thus enabled
variables duration and frequency, a frequency swing is required (i.e. a variation upwards
to exp~riment on the dependence of sound effect on the average frequency, the and downwards of the carrier frequency) of a size as great as that of the modulation
band Width and the r.m.s. amplitude of the modulating aleatoric oscillation.
frequency. For a frequency modulation with pitch-level oscillation of 3 c.p.s. to become
One requires a sound generator operating within a frequency region of up to audible, a carrier frequency of e.g. 440 c.p.s. (a') must oscillate at a rate of at least
c.100 c.p.s. (heterodyne generator) (1), together with a band-pass filter whose three times per sec. between 437 and 443 c.p.s. If for aleatoric modulation of amplitude
pass range does not exceed 10 c.p.s. In this way an aleatoric oscillation is ob- and frequency a frequency range (defined by the middle of a narrow noise frequency
tained where the average instantaneous frequency lies at the centre of the trans- band) of below 7 c.p.s. and for frequency modulation the corresponding frequency
mission range of the filter, and :where the amplitude bscillates aleatorically with swing stipulated by Stewart (half the width of the noise frequency band) is selected, t~e
an average instantaneous frequency equal to half the band width of the filter. unsystematic course of the modulating function becomes perceptible, i.e. the dynamic
With a band-pass filter with a free transmission range of from 5 to 15 c.p.s., the or pitch structure of the music becomes to a greater or lesser extent aleatoric. ~
average instantaneous frequency of the aleatoric oscillation is at 10 c.p.s., the procedure of this kind can attract the composer, and cases are known where aleatonc
average oscillatory frequency at 5 c.p.s. characteristics have been introduced (by mathematical, not by technical means) into
The aleatoric oscillation enables us to modulate a periodic carrier oscillation the structure of the composition (J. Cage).
e.g. a sinusoidal oscillation. If A (t) is the filtered aleatoric oscillation and
sin 2 1t v t is the sinusoidal oscillation, (vis the instantaneous frequency of the osc- 2. Analysis of Compositions by Methods of Information Theory
illation in this formula), the following modulation forms may be distinguished: Apart from the application of statistical methods by composers as described above,
(i) Modulation of amplitude: A (t). sin 2 1t v t; other statistical methods are utilised in analysing musical compositions. Here it is a
(ii) Frequency or phase modulation: sin (2 7t v t- A (t).) matter of examining musical creations with the mathematical means provided by
These two types of modulation may also appear simultaneously. statistical information theory. One assumes as a premise that music, just as language, is
made up of defined or at least definable single elements which can be nominated and
For the realisation of the modulation we require in (i) a modulator, in (ii) a described, i.e. it does not form a structureless continuum. It is hardly necessary to recall
cont:ollable delay-line in so far as the oscillation generator's frequency cannot . that the sum of these elements does not result in the music, no more than component
be drrectly affected through a reactance tube (i.e. a tube operating electrically syllables form a poem. Nevertheless, it seems that by concentrating only on the 'semantic'
as a capacity or self-induction). For example, a stepwise controllable delay-line aspects of music and by the elimination of all the emotional-aesthetic qualities, in-
is to be found in the vibrato arrangement of the Hammond organ. Oscillator interesting aspects of musical structure may be discovered which cannot be found out
controls with reactance tubes are described by Flanagan (4) and Weber (14). by any other means (13). In this way special criteria of form are added to older ones,
They contain a pentode as a controllable frequency selecting element. which are designed for the examination of latent work methods of the compositional
A timbre modulation may be obtained by means of a controllable filter, e.g. process.
of the RC- or RL-type. E. E. Schneider (10) describes this type of filter and the Formal criteria of various orders may be determined. The first order contains all
way in which it works. If a transposing filter is utilised the pass band is con- observations concerning the statistical distribution of the sound elements themselves
trolled in the simplest way by a phase-shift generator (6). without reference to any mutual relations. 1 The second order ('Markoff chains') and
The composer who disposes of the possibilities of aleatoric modulation will. be sur- all higher orders take into account, in examination, the frequency of transfer from
prised to discover that this kind of modulation leads him directly into a world of one element to another or between further distant elements, and their contextual
phenom~na, pr~vio~sly described as 'noises'. By broadening the frequency range of the relationships; considerable mathematical problems arise from these examinations. As a
modulatmg osc1llatwn and increasing the r.m.s. amplitude the resultant sound becomes characteristic which serves as a form criterion of the first or higher orders, information
more similar to a noise, i.e. a sound whose exact pitch cannot be determined. entropy may be derived from an observation of the rate of appearance of elements or
These. few observations clarify the fact that aleatoric sound procedures need not
1 We consider for example in statistical terminology the 'mean' or 'average' pitch level (respectively value, intensity,
necessanly produce aleatoric (i.e. non-predictable) sensations. For this it is essential spectral density, speed, et~.) of the compositi~n and the absol}lte _or relativ.e distrib.ution ~f the approJ?riate char~c­
teristics around this average value. The adopt1on of an averag~ng mterval,, '·~· the t1m.e reg1on ?Ver wJ:tlch avera~lll:g
that the frequency range of the applied aleatoric modulation be selected in such a way is extended is essential for the comparison of values; for a p1ece ofmus1c 1s not stnctly stationary m the stat1st1c
sense and s'o not liable to come under the application of the ergodic hypothesis.
56
57
groups of elements. 1 Up to the present, mathematical results exist only for literary works seconds the sequence of notes may be interchanged without affecting the perception of
(5) but there is no reason why the same procedures should not be applied to musical the sound. The acoustical stimuli in this case associated to the not~s, are called 'con-
compositions. ' ditionally equal'; the indifferentiable sensations are called 'metamenc :
Metamerism is a phenomenon which may be generally observed m the world of
3. Structural Characteristics of Valency Regions auditory perception. . . .
Filtered noise offers us a simple example of metamensm. Although this no1~e
Though a specific statistical structure may be ascertained in a musical composition presents a quite different oscillographic picture at different times (as a~l ale~tonc
fronr an examination of the score, it is by no means certain that this structure, con- procedures should) the sound perception is unaffected by these alteratiOns 111 the
ceived at an intellectual level, is automatically transferred into sound. There are no
oscillation image. . . . .
relationships between acoustical stimuli which may be derived from the score and the Finally, it is to be emphasised that sound elements which ~re J~xtaposed 111 tl~e can
corresponding sensations of sound which would permit us to consider these sensations ' have the effect that identical physical vibration procedures give nse to totally differer:t
as some kind of 'mapping' of stimuli within the physiologico-psychological range. To sensations. The phenomenon has been particularly observ~d in ~he case of sy~thetlc
clarify the details of these fairly complicated relationships we are compelled to adopt · soun ds sueh as 'p' , 't' , 'k' which may be perceiVed 111 a totally. different
exp1osive .
a special notation for that property of an acoustical stimulus which determines the manner, depending on the vowels which are juxtaposed to them (3). To exp~am ;his one
similarity or disparity of a sensation. With reference to the terminology custom- cannot attribute it to masking which has already been ~owr: f?r .a l~ng time oecause
arily used in physiological optics, we will call this quality the valency of the stimulus the influencing is effected by the following vowel (regressive ~Issmulatwn) as well as by
(9). the proceeding. At this point it may be mentioned that practlcal~y all statements about
These valencies may be represented in a multi-dimensional space. The components timbre have been effected by means of phonetic methods applied to spoken sounds.
of the valencies, e.g. frequencies, time- and place-co-ordinates act as 'co-ordinates' of Contrary to instrumental timbres, vowel timbres ha~e t~e parti~ular advant~ge that ~ey
this space. In an abstract valency space of this type every sensation indicates a place of may be recognised by untrained hearers thus mak111g It possible to descnbe prec1s~ly
sensation. In opposition to what one might conjecture these places of sensation are not
even small alterations.
distributed arbitrarily close together but are separated from each other by different
!imina. 2 The resulting cellular structure is the metric field of the valencies.
Strictly speaking it is not constant but depends on the velocity of alteration of the 4. Pitch Loudness Level
stimuli. Sounds, for example, carry their metric field over a large area with them. They The newcomer to physic-psychological acoustics will be stupefied when the results of
are only differentiable when they follow each other in close sequence. This phenomenon experiments show him that a note does not remain a note, ~nd a so~d not a sou_nd of
is known as conversion. Where the stimuli are presented in an isolated form the valency ' unalterable perceptible qualities, if alterations are effected 111 one of Its valency dimen-
locations can only be roughly di..t'ferentiated. The network is narrowed, as the number sions, which at first glance would appear insignificant. Thus, ~or example, _he ~ann~t
of stimuli per time unit is increased but at the same time it shrinks together. fully realise that a pure tone (physically - a sinusoidal sound s~gnal) alters m pit~h, 1f
All alterations of stimuli are rendered ineffective by the cellular structure of the the amplitude is enlarged or diminished, where the frequency IS constant. He w1ll be
valency-region if they are smaller than the difference Iimina at the appropriate velocity equally amazed to learn that this phenomenon has nothing to do with th~ ~ound
of alteration of the stimuli. It is not possible to effect an arbitrarily narrow scale either generating instrument and thus cannot result from a technical inadequacy, but. Is 111 fact
for pitch level or for intensity. Thus, for example, at a volume of 80 phons (forte), at a a property of the human ear (the cochlea). Only after detailed study and expenments of
moderately fast rate of succession, we are unable to distinguish more than about 200 one's own can one realise the fact known to acousticians, that the lou_dness _level of a
pitch levels in the whole sound area. We obtain similar limitations in examining the sinus tone of constant amplitude will alter by many orders of magmtude if the fre-
temporal sequence of compositional elements. Within a time interval of some milli- quency is altered, and that the amount of alteration depends largely on the sound
1 If pi is the frequency of occurrence of the element 'i' in the work to be analyzed, the first-order information- pressure amplitude chosen. . . .
entropy is defined as:- In returning to sounds, complex tones and noises, one ';'Ill be discouraged by the
multitude of occurring events (11). For example, it will be discovered that the loudness
which may be applied to the total inventory of different elements. In the second order, the frequencies pi are
level of a complex tone cannot be calculated from the levels or at least the sound-
replaced by the joint frequencies p(i,j) which indicate the occurrence of the pair of elements 'ij'. Thus the second- press~re amplitudes of its components by ~Y known_ mathematical means but that
order information-entropy is strange physico-psychological relationships 111tervene 111 the_ form of t_he S?ne-curve.
H 2 = - :?:?P (i,j) log2 p (i,j)
1 J If the resultant loudness level is correctly determined, a new difficulty anses: It does not
2 The idea of a metric field may be illustrated by an example from phonetics. Vowels are, as is well known, charac- follow from the explanation of the loudness level whether certain componen~s of the
terised by two formant regions each of a width of 300 c.p.s. Tests enable one to establish how many c.p.s. are complex tone do at. all contribute to the compound sound resultant; they sometrmes are
equal to shift the middles of the formant regions to effect a noticeable alteration of vowel colour. M. Joos found
that frequency alterations of approximately a semi-tone are necessary (8). The metrical field of vowel timbres rendered imperceptible by masking.
then resembles honeycombs with diameters of approximately one whole tone.
59
58
--------- -- -- - - - ------- -- --

5. Triple Pitch Quality 6. The Psychological Effect of Infra Sound

Research into electric methods of sound and noise generation has revealed a great H. Burris-Meyer of the Stevens Institute of Technology in Hoboken has demon-
number of perceptive phenomena which can only be discovered in traditional instru- strated that listeners may perceive rhythmically sound oscillations which lie below the
mental sound, where the ear has been prepared for it by electro-acoustic experiments. lowest frequency threshold (i.e. below 16 c.p.s.). In a play 'The Emperor Jones' infra-
This is similar to the impression received from exotic languages; when one hears them acoustical sound beats were generated behind the stage by means of a super-din1ensional
for the first time one can hardly observe their phonetic characteristics. Only by con- 'electronic drum' which carried a significant intensity when the rhythm was tense.
centrated study can one teach one's ear to distinguish the unaccustomed sounds and Despite the fact that the actor could hear nothing of this drum rhythm he soon began
sound combinations. . to synchronise his spoken part with the acoustically imperceptible drum beats. Thus
One of the most significant discoveries for the composer in the sphere of sound per- : sound, i.e. vibrations, may generate a physical effect even if they lie outside the per-
ception is the triple pitch level quality. Every musician is aware of the fact that a note ception area. It should not be necessary to demonstrate the consequences of this
has a double quality, the one- its absolute pitch, running parallel to the frequency, the psychological discovery for electronic music.
other- the chroma, a quality which recurs cyclically within each octave (only for
frequencies of up to c. 4500 c.p.s.). It sounds obvious if we say that we hear a note a'
Bibliography:
if the fundamental of the frequency is 440 c.p.s. But what happens if we remove this
fundamental by electrical means, leaving only the harmonics with frequencies of 880 (1) R. R. Bennett and A. S. Fulton, J. Appl. Phys. 22 (1951), p. 1187.
1320, 1760 c.p.s., etc.? Or if we take away the fundamental 440 c.p.s. and second (2) H. Burris-Meyer, J. Acoust. Soc. Amer. 13 (1941/42), p. 16.
harmonic 880 c.p.s.? We learn from experiments that the perceived pitch level remains (3) F. S. Cooper, P. C. Delattre, A. M. Liberman, J. M. Borst and L. J. Gerstman,
the same: a'. One may take away many of the lower harmonics without altering this. J. Acoust. Soc. Amer. 24 (1952), p. 597.
If this 'mutilated' note is interrupted for only an interval of a second, the sensation is (4) J. L. Flanagan, Electronics, Dec. 1949.
completely altered. Instead of the 'residual tone' on a' we now hear another pitch (5) W. Flicks, Studium Generale 6 (1953), p. 506.
which lies approximately in the region of the strongest remaining harmonics and is (6) W. A. Gunther, Frequenzanalyse akustischer Einschwingvorgange; Juris-Verlag
called 'formant pitch'. With some practice one can hear the residual pitch and the Zfuich 1951.
formant pitch simultaneously, recognising their unmistakable varied qualities. Both (7) J. M. Hanert and D. Hancock jr., US-Pat. 2, 432, 152 of 1947.
pitch sensations are independent of one another and so, although it may sound para- (8) M. Joos, Acoustic Phonetics; Language Monograph No.23, 1948.
doxical, a one-voice sequence of notes takes on the character of a two-voice line through (9) W. Meyer-Eppler, special number 'Elektronische Musik' of Techn. Hausrnitt. d.
the counter movements of residual and formant pitch. Sequences of notes may be NWDR 6 (1954), p. 29.
generated in which even the experienced listener will not be able to tell whether the (10) E. E. Schneider, Philos. Magazine (7) 36 (1945), p. 371.
sequence rises or falls. The singular effect of certain carillons depends on this conflict (11) S. S. Stevens and H. Davis, Hearing, its Psychology and Physiology; Wiley/
between residual and formant tones. Chapman & Hall, New York/London, 3rd Edition, 1948.
Without going into the far-reaching consequences of a triple pitch quality of the (12) G. W. Stewart, J. Acoust. Soc. Amer. 2 (1931), p. 325.
theory of the mechanics of hearing, I merely wish at this point to emphasise that at this (13) E. Sudau, Echtheitsuntersuchungen an ausgewahlten Kammermusikwerken von
stage of our research the significance of the form of oscillation apart from the spectral J. S. Bach; Diss. Kiel, 1948. -
structure of the musical sound signal becomes clear. If a stationary sound signal is (14) D. Weber, Elektron 4 (1951), p. 97.
supplied with a periodic envelope, under certain conditions the frequency of this (15) J. C. Webster and E. D. Schubert, J. Acoust. Soc. Amer. 26 (1954), p. 754.
envelope will become audible as a residual tone: This casts a new light on the problems (16) F. Winckel, Klangwelt unter der Lupe; Max Hesses Verlag, Berlin und Wunsiedel
of rhythmic formation, in that this may without effort be considered as the continuation 1952.
of the pitch phenomenon in the lowest frequencies.
The timbre of notes with so called 'gliding formants' belongs to this order of
problems. That a sound, whose spectrum lacks the even harmonics is known as 'a
hollow sound' (e.g. the clarinet), quite independently of the actual pitch of the funda-
mental, ought to have to do with the form of oscillation rather than with the structure
of the spectrum. Similarly with sounds which, for example, lack all harmonics divisible
by 3 or 5. They also have properties which remain invariable although the frequency of
the fundamental note is altered; we still lack a suitable notation for their sound
characteristics. Examination of spectra of this type has been limited to the period in
which one has been able to work with electric sound generators.
60 61