Vous êtes sur la page 1sur 28

Bartk, Lendvai and the Principles of Proportional Analysis

Author(s): Roy Howat

Source: Music Analysis, Vol. 2, No. 1 (Mar., 1983), pp. 69-95
Published by: Wiley
Stable URL: http://www.jstor.org/stable/853953
Accessed: 03-06-2016 18:21 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.

Wiley is collaborating with JSTOR to digitize, preserve and extend access to Music Analysis

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms



It is now over ten years since Erno Lendvai's analytical work on Bart6k's music first
appeared in English in book form, and over 25 years since it first appeared in
Hungarian.' Nearly all Bart6k analysis since then has referred to Lendvai, and some
writers discuss or develop aspects of his theory. This theory involves, and attempts to
link together, intervallic structures in harmony, tonality and melody, the use of
rhythm and metre, and the organization of forms in terms of large- and small-scale
proportion. This article concentrates on proportion and on Lendvai's claim that
Bart6k organized many pieces around the ratio known as 'golden section'. Lendvai's
proportional claims have been summarized, and occasionally extended, by various
writers2; of these, Fennelly, Kramer and Solomon challenge Lendvai's methods or
conclusions, and Solomon extends his challenge with an alternative method of
analysis. This critical disquiet suggests a need, in the first place, for an evaluation of
the musical relevance of proportional analysis, and in the second, for a wider-ranging
examination of the technical issues involved.

1: Mathematical preliminaries

The golden section, first documented in Euclid's Elements as 'extreme and mean
ratio', is the division of a fixed length in two so that the ratio of the shorter portion to
the longer portion equals the ratio of the longer portion to the whole length. The
section's exact value is irrational, its decimal places continuing indefinitely; it
approximates to 0.618034 . . . (a little under two-thirds) of the length measured: C is
the golden section of the line AB, with the longer portion lying to the left; D is the
opposite golden section of AB, with the shorter portion to the left. The special
property of golden section is that D also divides the portion AC in golden section. The
system can be further filled in or extended, forming a network of golden sections and
symmetrical divisions - something no other ratio will do - and this is accepted as
the main reason for the ratio's importance in the structure of organic forms and for its
repeated appearance through many epochs in the history of art.4
Golden section has another unique property: any summation number series with a
'memory' of two terms - that is, in which each term is the sum of the preceding two
terms - gradually approaches, as its values increase, a geometric series with a golden
section ratio. The basic series of this type begins with 0 and 1, giving 0, 1, 1, 2, 3, 5, 8,

? MUSIC ANALYSIS 2:1, 1983 69

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

Fig. 1: Golden Section


0.618034 . . .

0.618034 . ..

13, 21, 34, 55, 89, 144 . . . ; this is known as the 'Fibonacci' series and represents -
from 1, 2 upwards - nearest whole-number accuracy to golden section. For example,
34 x 0.618034 = 21.013 . . . ; 34 + 0.618034 = 55.013 . . . The next possible series,
known as the Lucas sequence, begins 1, 3, giving 1, 3, 4, 7, 11, 18, 29, 47, 76.. .
and yields nearest whole numbers to golden section from 4, 7 upwards.5 This
whole-number way of representing golden section is obviously important if the ratio is
to work in musical forms where divisions normally have to conform also to whole
numbers of bars, beats, semitones and so on.

2: Analytical issues

The analytical issues emerge from a comparison of the methods of recent proportional
studies of a large chronological range of music. In analyses of proportions in
mediaeval music, Powell (1980), van Crevel (1959, 1964) and Trowell (1979) always
measure proportions by the music's notated pulse, or tactus, and consider the possi-
bility of various systems of proportion. By contrast, Pascoe (1973), dealing with a
range of composers from Bach to Bart6k, defines the proportions of many pieces
according to the duration of recorded performances, and measures them against a grid
of exclusively golden section divisions. Szentkirilyi (1976) examines Bart6k's Second
Sonata for Violin and Piano, nominally following Lendvai, but his proportional
analysis is inexact since only three fairly wide arithmetical ranges are defined, taking
in almost all possibilities. By contrast, Camp's study of Mozart sonata movements
(1968) measures proportions precisely but deals with only a single aspect of each
movement - the double bar - and explicitly avoids consideration of any other
musical processes. Lendvai's own methods vary, sometimes concentrating on a single
event, elsewhere going into intricate detail; sometimes using the bar as the unit of
measurement despite varying bar lengths, elsewhere counting by beats specifically to
take account of changing bar lengths.
Evidently there is a need for firmer criteria, even if provisional, in order to
determine whether proportional analysis reveals anything significant about the music,
or about the composer's intentions or intuitions: otherwise, it is all too easy to make
up attractive patterns by selecting from the music's multiplicity of events. This essay
therefore has parallel aims: to propose some practical criteria for setting about
proportional analysis; to check Lendvai's Bart6k analyses against these criteria; and as
a result to reach some exact conclusions about Bart6k's music.
It should be said right away that the results of analysis are independent of the
question whether the proportional patterns found are the result of 'conscious' or
'subconscious' design, a question relevant to this study since Bart6k is not known ever

70 MUSIC ANALYSIS 2:1, 1983

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

to have spoken explicitly of proportional structure in his music. Controversy, in fact,

has long raged in various disciplines over the question of how accurately the human
mind can intuitively evaluate proportions either spatially or temporally, or whether
there are any instinctive or intuitive preferences for certain ratios. Gustav Fechner's
experiments last century with golden section rectangles (Fechner 1876), fortified by
Adolf Zeising's treatises (1854, 1884), still fuel debate, with significant contributions
recently from J. Benjafield (1976) and I. C. McManus (1980). If, then, a logical and
accurate system of proportions in musical form has occurred without the composer's
conscious knowledge, this intuitive aspect is confirmed, and the discovery has signifi-
cance for fields of study outside music, as well as for the aesthetics of our intuitive
response to musical form. Conversely, if a composer is known to have designed such a
system consciously, its importance to that composer at least is established, and this
knowledge implies that the composer may have believed it to be a way of aiding the
intuitive understanding of the music. The evidence on this question is discussed
following the music analyses, as it is dependent on the analytical findings and, more
intriguingly, reveals connections with Debussy and Ravel, two composers Bart6k held
in particularly high esteem.

(a) Measuring time and proportion

Kramer (1973) gives some examples of composers describing their application of

proportions, including Kienek, Nono and Stockhausen. The scores themselves may
tell us this: in Stockhausen's Fresco and Mikrophonie II, for example, durations of
sections are determined exactly by clock timings. Bart6k's music is different, with a
fixed metre and notated length which are open to interpretative nuance, despite the
durations supplied in some of his scores. Moreover, to play a movement or section at
Bart6k's suggested metronome speed in some cases produces timings quite at variance
with those he supplies. The solution comes from Bart6k's explanations that timings
and metronomic indications are 'suggested only as a guide to the executants' (preface
to the violin-and-piano score of the 1938 Violin Concerto) and, above all, that the
timings were 'notated from an actual performance'. That is, the music was written
well before these timings were determined. So if Bart6k intended a proportionally
exact structure, it could be located only in the notated metre or rhythm. Lendvai's
decision to measure Bart6k's music by the score and not by the clock is therefore
reasonable: ' . . . music breathes in metric pulsations and not in absolute measure-
ment of time . .. In music, passing time is made realisable by beats or bars whose role
is more emphatic than the duration of the performance' (1971:26). Still, options
should not be excluded in advance: there are cases where it is useful to try various
ways of measuring, and the music's rhythmic character should always be a factor in
deciding whether, in a piece with variable metre, the bar-line or the beat should be
regarded as decisive.6

(b) Accuracy and musical significance

The compromise about any golden section construction in music is that measurement
by beats or bars, involving whole numbers (or even fractions), cannot attain the exact
irrational value of the ratio. But this compromise is present in nature where the golden
section often appears most clearly as the Fibonacci series (for example, the spiral
arrangements of seeds in many plants, documented in Cook 1903 and Thompson

MUSIC ANALYSIS 2:1, 1983 71

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

1917). Music, too, is dependent on other similar compromises, the most obvious one
in Western music being equal temperament. Proportions in music are not an isolated
function in themselves: they can be neither articulated nor detected except in terms of
the music's dramatic and structural events. Therefore any proportional system has to
accommodate itself to the music's audible structure, not only in terms of whole beats
or bars but also occasionally in terms of larger units - for example, sequences of
four-bar groups if the rhythmic impetus prohibits their disruption. In short, the
large-scale proportions must be co-ordinated with the smaller-scale proportions of
metre and rhythm. The main analytic criterion, then, is: if an inaccuracy occurs in a
proportional system, can it be accounted for in terms of musical necessity?7

(c) Lendvai's analyses8

Two qualities make Lendvai's analyses especially fascinating. The first is the link
between his theory and the organic world: Lendvai stresses Bart6k's expertise in
natural history and his frequently expressed view of music as an organic phenomenon.
The other quality is the cohesion of Lendvai's ideas which embrace all aspects of
musical structure. For example, golden section is investigated not only relative to
formal duration but also as a feature of Bart6k's use of modes and intervals. Lendvai
notes that pentatony shows golden section properties in its distinctive intervals (tone,
minor third, fourth and minor sixth = 2, 3, 5 and 8 semitones); balancing this, and
the chromatic systems derived from it, is the symmetrically-based harmonic series
(consisting of whole-number multiples of the fundamental) from the first twelve
harmonics of which an 'acoustic' scale is derived (C, D, E, F?, G, A, Bb) and is
demonstrated to be prominent at moments of resolution in Bart6k's music.9 Lendvai
also shows how a model of an 'axis' grouping of minor-third-related tonalities,
dividing the octave symmetrically into four parts, can account for tonal relationships
normal in Bart6k - especially for his predilection for the tonal span of a tritone. In
the Sonata for Two Pianos and Percussion, for example, the tonic C defines an 'axis'
completed by the three related poles Eb, F?/Gb and A (C and F? are mutual
'counterpoles', like A and Eb); G, Bb, C?/Db and E form the 'dominant axis', and F,
Ab, B and D the 'subdominant axis'.10
Doubts arise about Lendvai's claims concerning proportion because he launches his
analyses without first discussing the questions of criteria and method raised above;
although he takes up some of these questions later, it is often only after liberties,
sometimes convenient ones, have been taken, as seen below. His analyses and his
proportional calculations are often imprecise. Numbers of bars begun are sometimes
confused with bars completed, and more generally, two essentially different golden
section phenomena within a movement are thrown together: in some cases an impor-
tant musical event is identified as a movement's golden section without the support of
further analysis, while in other cases thorough analyses of proportions inside a
movement or a section relate the proportions more integrally to the music's formal
Lendvai does not attempt, either, to pursue the possibility of proportional connec-
tions with earlier music. In this he may have been misled by an oversimple view of the
Classical style as built almost entirely on symmetrical principles: 'As a rule, the
themes in the music of Haydn, Mozart and Beethoven are divided into 8 + 8, 4 + 4
and 2 + 2 bars' (1971:77). But classical phrase structure shows extensive subversion
of such symmetries. Indeed elision of the last bar or two bars of a regular group into

72 MUSIC ANALYSIS 2:1, 1983

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

the first bar or two bars of a new group is exploited in Haydn, Mozart and Beethoven
to produce movements in which the Fibonacci series and simpler whole-number ratios
are perceptible: see Pousseur (n.d.) and Rutter (1977).
One might argue that symmetry in Classical music exists as an instinctive model for
exploitation and variation. But then Lendvai argues that in Bart6k's chromatic,
golden-section-based movements (specifically, the first movement of the Music for
Strings) 'the law of balance and symmetrical periodisation is replaced by regularities
of tense asymmetry' (1971:77), when in fact the Music for Strings itself opens with
phrase groups of 4 + 4 + 4 + 4 + 4 + 6 + 4 bars. This suggests that if golden section
is indeed influential in that movement's layout (Lendvai's analysis is examined
below), it works in balance with symmetrical elements, parallel to the balance
Lendvai postulates between Bart6k's diatonic and chromatic (pentatonically-derived)
Early in his study (1971:18) Lendvai supplies a list of the first category of pro-
portional occurrences - those where a single dominating structural event exactly
marks a point of golden section. Some of those cases are calculated accurately, for
example in 'Free Variations' and 'Broken Chords' ('Divided Arpeggios') from Mikro-
kosmos - though in the former piece no account is taken of bar lengths which vary
between 3 and 9. Other cases are less accurate, for example 'Diary of a Fly' (Mikrokos-
mos) in which the golden section point is in b. 61 (after 112.5 x 0.618 = 69.5 units of
4) and not at the sff in b.59 as Lendvai claims. In two other cases - the first
movement recapitulation points in Contrasts and the Sonata for Two Pianos and
Percussion - Lendvai's calculations are inaccurate by a bar as a result of confusion in
his numbering system between bars begun and bars completed. In neither of these
examples does Lendvai allow for bars of different length, though to do so, in fact, fails
to annul the inaccuracy. There are therefore two drawbacks to these examples. First,
in isolation they tell us very little else about the pieces concerned. Second, in view of
the inaccuracies, it could be argued that those particular occurrences are fortuitous, or
are merely the more precise manifestations of a very approximate intuition of golden
section, revealing little about Bart6k's working methods.
Nevertheless, other accurate examples are not hard to find. The first movement of
the Third Piano Concerto consists of 554 crotchet beats whose main point of golden
section, after 342, falls exactly at the beginning of b. 117 - the movement's recapitu-
lation. In the second movement of the Fifth Quartet, the central turning point of the
ABCC'B'A' arch form in b. 35 (letter C, Piu lento) is placed with equal golden section
exactitude, dividing the movement's total of 223 crotchet beats 138:85 - a particu-
larly impressive example, marked clearly in the middle of the bar. The opening
movement of the Divertimento is the most arithmetically precise example so far. Its
total of 1681 quaver beats yields golden section to nearest whole numbers of 1039:642.
1039 quaver beats lead exactly to the beginning of b. 129, the movement's fff climax
on a unison tonic F. To complement this, 642 quavers lead exactly to the beginning of
b. 80, where the music, having completed the exposition, turns (at Piu tranquillo) to
the tonic's 'counterpole' of B major. This positioning, shown in Fig. 2, is identical
with Fig. 1 above, which gave the basic format of two complementary golden sections.
Such precision and consistency challenge the possibility of random occurrence. If we
try, as a control, the alternative method of counting by bars rather than beats, no such
accurate proportions result in the Third Piano Concerto and Divertimento examples,
but they do in the Fifth Quartet example in Perle's analysis (1967:7, discussed below).
Lendvai also mentions an overall golden section occurrence in the Divertimento's first

MUSIC ANALYSIS 2:1, 1983 73

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

movement (1971:18), but confuses the issue by describing the movement's metrical

produceunits' (the movement
the required contains
total of 1681 bars
quavers of 4,
when 8, 7 and by
multiplied 8, and his 563
3), and
by identifying the golden section as the 'recapitulation', which is a dubious label for bs
129-30 (if that is indeed the place he means).

Fig. 2: first movement of Divertimento

climax on
fff unison F
(b. 129) 1681

1039 642

642 4 1039-

modulation to B
(b. 80)

Further analysis of these scores (except for the Fifth Quartet) does not reveal a
similarly proportional structure in their smaller detail. But the possibility of pro-
portional logic dominating the details of a whole movement or passage is subsequently
taken up by Lendvai in analysis of four movements.

(i) Sonata for Two Pianos and Percussion: First Movement

Lendvai selects bs 2-18 of the Sonata's first movement (Ex. 1) as 'a model of golden
section construction' (1971:18). He omits b. 1 (the pp F? timpani roll) from the
calculation, which is logically acceptable since the movement's thematic and rhythmic
activity begin only in b. 2. From b. 2 the theme follows a canonic series of 'tonic axis'
entries (on F? and C, bs 2-8), of 'dominant axis' entries (on G and Db, bs 8-11), and
finally of 'subdominant axis' entries, this time inverted (on D and Ab, bs 12-17),
leading to the climax of the introduction, the fff at the beginning of b. 18. According
to Lendvai the first tonic entry, the first 'dominant axis' entry, the first inverted
('subdominant axis') entry and the final fff form an exact golden section sequence,
following a series of 171/3, 28 and 46 units of 3. In addition the two cymbal clashes,
after 11 and 22 3 units, divide the tonic and dominant segments in golden section
(GS), the first segment by long-short division ('positive GS') and the second one
short-long ('negative GS'). The only discrepancy in Lendvai's figures is small: the
first 'dominant axis' entry comes after 18/3 units, not 171/3. To the nearest whole
numbers, this gives a series of events after 11, 18, 28 and 46 units, with the second
cymbal clash after 22 units forming a subdivision of 4:6 between 18 and 28 units. The
only breach of theoretical accuracy is that at 28 and 46 the sequence drops one unit
short of the Lucas summation numbers (4, 7, 11, 18, 29, 47 . . .) with which it
Obviously this breach is too small to invalidate the logic of the shape and its
structural reality: Lendvai has found an excellent example to establish his case. One

74 MUSIC ANALYSIS 2:1, 1983

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

Example l. Barf6k, Sonata for Two Pianos and Percussion I,bs 2-1B,(as in Lendvai's
short score reduction, 1971, 18-19)

Circled numbers refer to completed units of 3

b2P I

8--- 8-

b 1WDpI R -
Side drumbal c.. Side dm P

ybPP L .
I?7. 7 p

77FJ _ t

l~prdce y in erts~n fBosy& ake~adofKhnan7vei .7

MUSIC ANALYSIS 2:1, 1983 75

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

possible explanation for the small inaccuracy is that the exact Lucas numbers 29 and
47 would have necessitated further changes of bar length; while such changes were
musically acceptable (and proportionally necessary) in bs 4 and 6, from b. 9 onwards
they would impede the music's momentum. But there is another explanation, one that
reinforces Lendvai's case. Had he taken the rest of the movement's introduction into
account, noticing that the first dynamic peak (the cymbal clash in b. 6 after 11 units)
is exactly halfway to the second peak after 22 units, he might also have noticed that
this second clash is just one unit away from the halfway point to the fffclimax after 46
units; this climax, too, is within one unit of the exact mid-point to the first entry of the
movement's first subject (ff) which completes the preceding long crescendo (after 91
units) one bar after the Allegro arrives at b. 32. This symmetrical sequence of dynamic
peaks, providing a powerful large-scale rhythm to prepare the first entry of the
Allegro's first subject, runs in counterpoint with Lendvai's tonal golden section
sequence (Fig. 3). In addition, the accelerando leading to the main Allegro begins after
55 units (b. 21) and the Allegro arrives after 88 (b. 32) - within one unit of golden
section formed by the Fibonacci numbers 55 and 89. Evidently, therefore, it is useful
to scrutinize any cracks that appear in the initial constructive logic; in this case an
early exception has helped to prove the rule.

Fig. 3: first movement introduction in Sonata for Two Pianos and Percussion
(b. 2)
28 46

'tonic 18 INVERSION
) axis 'dominant' 'subdominant'
proportions axis axis
noted by 7 11
11 7 4 6
1. iTam-
tam 91

Allegro first subjcct

1111 2enters,
22 //
Cymbal Cymbal (b 33)
f f

symmetrical - - - - -
dynamic 22 24

46 4645
(b. 18)

tempo 55 33

55 88
accelerando Allegro
begins - - - - - - - - - - - begins
(b. 21) (b. 32)

(ii) Sonata for Two Pianos and Percussion: Finale

For three reasons Lendvai's proportional analysis of the Sonata's finale (1971:22-6) is
less convincing. First, many ratios described as golden section are only approximate:

76 MUSIC ANALYSIS 2:1, 1983

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

for example, in his Fig. 20, tracing bs 287 to 350, 22:16 and 38:26 (or worse, 38:271/4
if one counts the five bars of 5 each as 11/4 units). Second, his diagrams are selective
and often overlook important events, such as the coda's tonal sequence (see his Fig.
21). Third, his system of dividing the movement into neat formal sections, with a
diagram for each, interprets it explicitly and wrongly as a regular sonata form: for
example, his Fig. 17 calls bs 28-43 the B portion of the first subject; yet in the
recapitulation the same motive returns (inverted, at bs 287-300) appended to the
second subject. This emphasizes an ambiguity of formal function: Lendvai decides
that the first subject treatment in b. 247ff belongs to the recapitulation, but it can be
located more logically in the development, since the passage is tonally unstable and
thematically sequential. Only Lendvai's Figs 17 and 18, charting the first and third
thematic groups in the exposition (bs 1-43 and 103-33), are convincing; although the
musical descriptions can be challenged, the proportional calculations are correct.
(Had Lendvai cared to measure the music of his Fig. 17 more precisely, he could have
presented the movement's opening thematic sequence as one of 34, 55 and 88 crotchet
beats rather than of 17, 271/2 and 431/2 bars.) To see how comprehensively the logic
might apply to the movement, though, a thorough analysis would be necessary, taking
account of the movement's sonata-rondo elements (specified in Bart6k 1976a:418), its
formal ambiguities, the thematically re-ordered recapitulation and the strong tonal
sequences in the development and coda.
Lendvai provides two other calculations for the Sonata as a whole. He measures the
number of quaver beats of the whole Sonata, a total of 6,432, and identifies the golden
section of this, 3,975, as the beginning of the second movement (the central division
of the entire work's slow-fast + slow-fast sequence). Expressing the 6,432 quavers
as 804 semibreves, he then defines the 804 as rr x 256 - that is, as the product of the
closed circle and open symmetry (the latter represented by powers of two, or 16
squared), reflecting the closed versus open characteristics of the Sonata's first and last
movements - compound triple metre as against simple duple, chromatic as against
diatonic preponderence, and so on. (Lendvai might also have mentioned that 256 is
the approximate frequency of middle C, the tonic of the work.) Both Lendvai's
calculations are accurate, except that 3,975 quavers actually lead to the rests in the
first movement's last bar and not exactly to the beginning of the second movement.
However, the quaver has different tempi: Bart6k's metronome marks indicate rela-
tionships of dotted crotchet=quaver=minim linking the three movements. It is
unlikely, therefore, that these two relationships could proceed from a subconscious
feeling for musical time. The relationships were absent from the score until a late
stage of preparation: virtually the final revision to the Sonata's first movement was a
complete recasting of its development section from b. 229 onwards, reworking the
tonal structure and shortening it by nine bars."

(iii) Music for Strings, Percussion and Celeste

Lendvai's proportional plans for the first and third movements of the Music for
Strings (1971:28-9) are arresting because they follow Fibonacci numbers, and because
in each case one plan embraces the whole movement. In his Fig. 22 (1971:28) Lendvai
supplies a logical architectural model of the fugal first movement (Fig. 4a). The
principal point of golden section falls after 55 bars, corresponding to the dynamic
climax of the movement which is predominantly on Eb, the furthest tonal point from
the 'tonic' A. Intermediate divisions by Fibonacci numbers after 34 and 68 bars

MUSIC ANALYSIS 2:1, 1983 77

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

encompass the removal and replacement respectively of the mutes (68 marking the
13:21 division between the climax and the end), and the division after 21 refers to the
end of the fugal exposition. All this, though, is an oversimplified view. The first fugal
episode, at b. 21, begins after 20 bars, moving Lendvai's division after 21 forward by
one bar; the Bb timpani pedal and removal of string mutes take place after 33, not 34
bars; the important celeste entry after 77 bars goes uncharted (were it after 76 bars it
would produce an additional Fibonacci division in Fig. 4a); finally, there are 88 bars,
not 89, in the movement. Lendvai's footnoted explanation that the movement 'must
be completed by a whole-bar rest, in accordance with the Billow analyses of Beet-
hoven' (1971: 28) is unconvincing, since in this movement the final bar, marked 81' ,
contains only ten quaver beats to balance the anacrustic quaver at the start of the
movement, forming a closed system. Thus four adjustments must be made to Fig. 4a
to make it exact as shown in Fig. 4b. Since anacruses in this movement (including that
to b. 1) belong musically with their following bars, numbers of completed bars in the
diagram exclude any anacruses phrased over to the next bar; this also accords
proportionally with the presence of an anacrusis to b. 1 and the missing last quaver in
bar 88:

Fig. 4: Fugue from Music for Strings, Percussion and Celeste

PP fff PPP


21 3468 81 89

(a) _ _ ....
ideal proportions 21 13 21 13 13 8
by Lendvai

(b) 20 13 22 13 9 4 7
actual proportions

20 33 68 88
mutes off: mutes end
episode Bb timpani replaced 77
pedal 55 celeste
climax (coda)

Fennelly (1973) and Kramer (1973) also identify these discrepancies, and Fennelly
notes the opening four-bar sequences of the Fugue. Kramer concludes that the
Fugue's proportional tendencies are therefore 'a less significant structural force than
in other works . . .' (1973:120). Yet, as with the scheme seen in Fig. 3 above, the
clarity and force of this overall shape suggest the opposite, despite the inaccuracies.
Why the inaccuracies, then?
Two alternative methods of counting can be considered. First, because bar lengths
vary constantly, the movement can be counted by quaver beats. These total 705,
yielding a golden section of 436, which falls after the second quaver of b. 54. This is
almost two bars away from the climax, whereas, counting by bars, golden section of
88 gives 54.4 - less than two-thirds of a bar from the climax. Nor do the intermediate
divisions produce more accurate proportions when counted by quavers. The second
method is applied by Solomon (1973:140-53). He observes Lendvai's discrepancies

78 MUSIC ANALYSIS 2:1, 1983

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

but accepts the formal divisions and measures them instead by clock time. Taking the
mean of Bart6k's metronome markings (Bart6k gives ranges of approximately ? 5%),
Solomon calculates that all the main points of division conform to golden section. In
fact they do so only to within 1%, virtually the same tolerance included in the bar
count, and even that disregards the additional ? 5% in Bart6k's metronome mark-
ings. Neither of these alternative methods therefore provides an immediate solution.
However, there is a more solid musical aspect to this. For all the logic of Fig. 4's
shape, only two of its divisions, after 20 and 55 bars, are associated with fugal events.
As Bart6k's preface to the score explains, the order of fugal entries follows a circle of
fifths, beginning on A and working simultaneously upwards via E, and downwards
via D, until the two diverging progressions converge on Eb/D$. The arrival at the Eb
counterpole, uncharted by Lendvai, occurs on the last quaver of b. 44, the halfway
division of the movement's 88 bars. Since the entire fugue forms a symmetrical tonal
system, this halfway division is logical. The fff climax after 55 bars is still concen-
trated on Eb; but this is now the turning point where the subject inverts (the Eb's
contradicted by the bass E and Bb) and begins the homeward sequence to the final A.
Thus the arrival at the symmetrically opposite pole of Eb is placed symmetrically at
the movement's halfway point, while the departure from it - the main formal and
dynamic turning point - is placed within a bar (at most) of the movement's golden

There is more structural intrigue in the Fugue's exposition. Five successive four-
bar entries take us to the end of b. 20, after which the first episode occupies six bars,
the first departure from four-bar regularity. But it is only after this that the first
violins enter for the first time, to complete the texture and close the exposition. This
last entry, set in relief by its delay, also introduces the Fugue's first stretto (with the
double-basses) across the C-Fr tritone that marks the halfway stage in the two-
pronged progression from A to Eb. With this ingenious containing of the first episode
and first stretto inside the fugal exposition, Bart6k marks three decisive structural
events: the beginning of the first episode after 20 bars, the first stretto and the tonal
'equinox' after 26 bars, and the exposition's proper completion after 30. The first
episode after 20 bars is further set in relief as the highest point in the 2nd violins'
melodic line (which then falls again until the end of b. 26); this marks the golden
section on the way to 51 (completed) bars, where the basses' larger melodic arch
reaches A (sounding A below middle C), their highest point in the Fugue. The first
stretto, after 26 bars, then marks the golden section on the way to its point of formal
reflection - the end of the inverted C-F? stretto after 68 bars (a ratio of 26:42=
13:21). Finally the completion of the exposition after 30 bars is within a bar of golden
section to the beginning of the coda after 77 bars, where the fugue subject, back on A,
sums up the movement's tonal journey by sounding with its inversion. Each of these
divisions, short-plus-long (which Lendvai calls 'negative,' though their associated
effect here is far from negative), reflects the movement's powerful expansion all
through its first part.
Fig. 5 combines these proportional relationships, showing also how this combi-
nation produces further symmetries on the one hand, and, on the other, additional
golden sections of 18:11, subdivided 7:11:7:4, leading to the focus of the climax.
From the arrival at the central Eb after 44 bars to the focus of the climax after 55,
there are 11 bars; correspondingly, 11 bars follow the final return to A to end the
movement, and these final 11 are divided in the golden section 7:4 (Fig. 5).

MUSIC ANALYSIS 2:1, 1983 79

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

Fig. 5: Fugue from Music for Strings, Percussion and Celeste

PP fff PPP

33 55
20 mutes climax 68 88
first off mutes end
episode on
13 20
dynamic arch


subject reaches

4 44 - - 44 -

S11 11 11

tonal symmetries 7 4 7 4

7 11 7 4

interactions I 11-
18 II I

20 44 55
0 I Eb fff
4 4 4 414 4

opening sequence 6 4
(2nd peak melodic
vlns) mo peak
p I
20 31

26 42 1
other I
connections C26 I en68 of I
C/F: end of
stretto inverted
C/F? stretto

30 47

30 77
end of coda begins

80 MUSIC ANALYSIS 2:1, 1983

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

As with Fig. 3 above, none of the one-bar discrepancies can be corrected without
introducing a greater discrepancy in another proportional relationship. At the same
time no discrepancy exceeds one bar, and they are restricted to the large proportions,
keeping the small-scale ones accurate to the nearest bar. The whole edifice grows
naturally out of the opening four-bar sequences, combining the nearest possible
approach to the Fibonacci series with other sequences based on the Lucas series (4, 7,
11, 18, 29, 47, 76. . . , stretched on one occasion here to 30 and 77) and on bijugate
numbers of the Fibonacci series (26, 42, 68. . .). Perhaps most formally subtle of all is
the sequence of turning points in the second half of the movement (the arrival on Eb,
the basses' highest point in the arch, the climax, the replacement of mutes, and the
coda): each marks a convergence of two separate proportional sequences, one from the
large-scale arches at the top of Fig. 5, and the other from lower in that diagram. This
process matches the sense of growth and tension in the outward part of the movement,
and of retraction and resolution in the homeward part. Parallel to this duality is
another between the two principles of golden section and symmetry, reflecting the
appearance of golden section mostly in organic nature, as against the characteristic
symmetry of inorganic forms. In this example golden section associates itself mostly
(though not uniquely) with points of growth and drama, and symmetry mostly with
points of arrival or resolution; and in the Sonata for Two Pianos and Percussion (Fig.
3 above) the duality appears as a symmetrical sequence of dynamic peaks against the
golden section ordering of the more organic relationships of tempo and developing
The one remaining anomaly in the Fugue is that its proportional system evidently
works best by counting bars and not quaver beats, despite the changing bar lengths.
But this movement also contains a feature that favours counting by bars: its quaver
pulse is not strongly rhythmic, whereas the bar sequence is very clearly articulated
throughout as phrase structure.

(iv) Music for Strings, Percussion and Celeste: Third Movement

The last of Lendvai's analyses to be examined here, more briefly, is his Fig. 23
(1971:29), representing the third movement of the Music for Strings in terms of the
Fibonacci series (Fig. 6a). Again there are inaccuracies, as shown in Fig. 6b which
gives exact dimensions, counted here in units of 4 (following Lendvai). What can be
shown only roughly in Fig. 6b is the inherent ambiguity at transitional points of the
movement's arch form. For example, the opening A section ends clearly enough after
19 bars (not 21), but the following B section starts up only in stages over bs 20-22. Bar
75 is transitional, belonging to neither the recapitulated B section nor the recapitul-
ated A section. This formal scheme also incorporates smaller-scale details, such as the
xylophone interjections in various sections, which could affect proportional logic.
Another possible reason for the analytical inaccuracy here lies in symmetries con-
nected to dynamic shape: the movement's first climax, after 141/2 units of 4, deter-
mines the length of the B section and then that of the build-up to the climax in the C
section. 12
Unlike the Fugue, the third movement produces the most consistent proportional

pattern when counted

units. Rhythmically by iscrotchet
this logical: beat, taking the
the opening 3 bars and
xylophone solo the 4 as
could 1?/2 and
hardly 1?/4
ize the crotchet beat more clearly against the bar-line. Not only that, but it follows a
sequence of 1, 2, 3, 5, 8, 5, 3, 2 and 1 repeated F's per crotchet beat between the

MUSIC ANALYSIS 2:1, 1983 81

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

Fig. 6: third movement of Music for Strings, Percussion and Celeste

4 -89 -

4-34 5- 55-
version 21 13 13 21 13 8
(measured 1st theme 2nd theme climax 2nd theme 1st
in units theme
34 f -- 21 --

I (21) (34) (47) (68) (81) (89)


19 34 481/2 climax
601/4 I
691/4 811/4 891/4
(b) AlB C Allegretto D E
exact 203/4
proportions 19 111/2 141/2 20 12 7

221 122
2nd themeclimacticsttheme
493/4 1st /me

I I / I



version / - 21 - --- 13 --" - 8 - - 134 13 8

(measured 4 [14]
bybars) 3+5+5+3+5 5 121 8 5 4 2+8+3
4 r 4
metre: 4

one bar

beginnings of bs 2 and 4 and thus strongly presages the movement's overall arch form,
based on the Fibonacci series. Lowman (1971a:528) supplies an alternative pro-
portional reading measured by bars (Fig. 6c). This retains some of Lendvai's inaccur-
acies and introduces some new problems (for example, the selection of 'transitions'),
creating a less convincing overall form.
Despite the approximations, Fig. 6 (versions a and b) has a special relationship to
one of the most potent golden section forms in nature and art, the logarithmic spiral,
which underlies constructions ranging from snail shells to galaxies. Lendvai describes

82 MUSIC ANALYSIS 2:1, 1983

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

Fig 7: third movement of Music for Strings, Percussion and Celeste

/ 34

60'/48\ \

Alle rtto

~ climax-

/ \/
I -- 20?%-
I 48 269? /

I I 1'


segments of AB. C A'

arch form I---_ 2 -I

the logarithmic spiral (1971:31) and suggests that its cross-section corresponds closely
to the structures seen above in Figs 3a and 4a. In fact it fits neither of them well since
they are too symmetrical, but it very aptly matches the form and dynamic shape of the
third movement of the Music for Strings (again allowing for the approximations),
centering on the climax as shown in Fig. 7. In the process it shows a possible
explanation of why the two outermost transitions in the form - those uninvolved in
the spiral - are the most masked and ambiguous.
Because musical form cannot literally follow the spiral shape, this aspect of the
analysis is inevitably hypothetical. We can only guess that Bart6k may have intended
the spiral to be symbolic, and there is some circumstantial evidence of Bart6k's
interest in this shape as potentially musical: in 1909 he wrote out a copy of 'Seesaw'
(Seven Sketches, Op. 9b, No. 2) literally in spiral form (see the facsimile in Bart6k
1981, Series 1: Introduction).
Close analysis could also throw light on other golden section occurrences found by
George Perle in Bart6k's Fourth Quartet (reported by Antokoletz 1975: 146-50) and
Fifth Quartet (Perle 1967:7). The most immediately striking example is undoubtedly

MUSIC ANALYSIS 2:1, 1983 83

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

the third movement of the Fourth Quartet, with prominent structural transitions after
5, 123/4, 21, 34 and 54V/2 bars (all bars the same length), near to Fibonacci's 5, 13, 21,
34, 55. The drawback, as Antokoletz observes, is that the movement's central section
and end defy the sequence by coming after 41 and 71 bars. Perle's most detailed
golden section tracing, from the second movement of the Fifth Quartet, gives the
same central division of the ABCC'B'A' arch form measured by bars (345/8: 213/8)
that was identified through counting by beats (138:85 crotchets, p. 73 above): it also
accounts in golden section terms for some of the intermediate transitions. However,
not all the transitions are included, at least one of the divisions is questionably placed
(the final B'A' transition, spread across bs 49-50, is taken as the beginning of b. 49),
and the measurement follows a bar count despite the music's more clearly defined
crotchet pulse, with varying bar lengths. More detailed analysis, in addition to
tackling these problems, could seek specific musical coherence in the geometry, of the
kind now seen in the first movements of the Sonata for Two Pianos and Percussion
and the Music for Strings.

3: Conscious or subconscious?

The shapes and proportions found so far seem to be variously approximate, precise,
comprehensive or sporadic. Ironically the most exact proportions (as in the Third
Piano Concerto and the Divertimento) are often the least crucial musically, being the
least comprehensively implanted throughout the music's forms. In these cases the
porportions, if consciously applied, might be a matter of Bart6k's pride in profes-
sional craft, as a secret written into the score like the overall quaver counts applying to
the entire length of the Sonata for Two Pianos and Percussion. Another example
implies a humourous approach: the finale of Contrasts begins in markedly ribald
mood, with a well defined 34-bar opening section, after which Bart6k injects pro-
portional anarchy by allowing ad lib repetition of b. 34 while the violinist changes
Lendvai does not pursue the question of whether Bart6k used the golden section
consciously or unconsciously, but in summary of Lendvai's ideas, Laiszl6 Somfai takes
a more specific view:
It should be noted that the golden section proportions demonstrated in detail in
[the Music for Strings and the Sonata for Two Pianos and Percussion] by Lendvai,
based on his examination of Bart6k's sketches and manuscript corrections, were
not consciously planned but fortuitous (Somfai 1981: 215).13

'Fortuitous' here is an unfortunate complication. It is probable that Somfai means

'subconsciously formed' rather than 'random' (the literal sense of the word).
The strongest argument for the view that Bart6k's proportional systems were
subconsciously produced is that no definite proof is known of a conscious application.
Surviving manuscripts of the works discussed here show no signs of numerical
calculation, and Bart6k is not known to have spoken to anyone on the subject. Yet
secretiveness was one of Bart6k's strong traits: he habitually avoided describing his
compositional techniques in detail, except for some discussion of his use of modes and
chromaticism (Bart6k 1967a, techniques that could already be partly traced from his
published ethnomusical researches. Moreover, to have divulged the use of highly
abstract constructions in what many critics already regarded as over-cerebral music
would have been counterproductive in his constant battle to have his music accepted

84 MUSIC ANALYSIS 2:1, 1983

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

purely for its expressive power.14 One of his few recorded statements of compositional
method occurs in the third Harvard lecture (Bart6k 1976a: 376), in the context of
discussion of 20th-century chromaticism:

. .By the way, the working-out of bi-modality and modal chromaticism hap-
pened subconsciously and instinctively, as well. I never created new theories in
advance, I hated such ideas. I had, of course, a very definite feeling about certain
directions to take, but at the time of the work I did not care about the designations
which would apply to those directions or to their sources. This attitude does not
mean that I composed without . . . set plans and without sufficient control. The
plans were concerned with the spirit of the new work and with technical problems
(for instance, formal structure involved by the spirit of the work), all more or less
instinctively felt, but I never was concerned with general theories to be applied to
the work I was going to write. Now that the greatest part of my work has already
been written, certain general tendencies appear - general formulas from which
theories can be deduced. But even now I would prefer to try new ways and means
instead of deducing theories.

Evidently, instinctive working out of technical problems eventually reached a stage of

conscious recognition. Many of Bart6k's sketches contain, in fact, dimensional and
other variants, showing that such 'set plans' - proportional or other - were not
immune to subsequent modification, as has already been seen with the first movement
of the Sonata for Two Pianos and Percussion. Since formal structure was, like other
technical problems, 'more or less instinctively felt' to begin with, the process of
recognition could well have involved a subsequent refining in detail of proportional
qualities inherent in the forms originally intuited. This would be in keeping with
Bart6k's intense interest in science and natural history (see Bart6k fils 1966). Agatha
Fassett (1958:2) begins her reminiscences of Bart6k by recalling his fascination for
natural forms, relating that he would 'pull [a] pine cone apart and stare at each bit of
it,' or 'stare at an ordinary leaf on an ordinary tree just as if he'd never seen anything
like it before.' Lendvai (1971:29) also points out that the sunflower was Bart6k's
favourite plant, and that this and the fir-cone (which has the same structure as the
pine-cone) are two of nature's clearest manifestations of Fibonacci construction; and
that various types of leaves are clearly proportioned by golden section, as described in
various botanical treatises from the beginning of this century.
More positive manuscript evidence of proportional calculation exists in Bart6k's
transcriptions of folk music. Manuscript 80FSS 1 in the New York Bart6k Archive is a
sketchbook devoted mostly to Bart6k's first drafts of Turkish folk song transcrip-
tions, with early sketches for the Concerto for Orchestra inserted much later on spare
verso pages and in other gaps. Page 1 (recto) of the volume contains his exact metrical
plan for his transcription of the Turkish folk song"5 on the following recto page, with
a metrical sequence in each of its lines of 17, 1, 7, two bars of 1, . These two pages
are reproduced here as Ex. 2. The numbers are immediately recognizable as belonging
to the Lucas summation series; Bart6k's measurements show that each pair of 9 bars
in the song is simply a more manageable notation of what he understood as a group of
8. Thus, the song's metrical organization is analysed wholly in terms of the Lucas
sequence numbers.
Clearly, any argument that Bart6k had no conscious knowledge of the exact
proportions in some of his symphonic scores - an argument based mainly on the
absence of visible calculations in the surviving sketches - is weakened by this
documentary evidence of an interest in the exact proportions in folk song, especially

MUSIC ANALYSIS 2:1, 1983 85

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

Ex. 2: Facsimile of recto pages 1 and 2 from manuscript 80FSS1 in the New York
B61a Bart6k Archive, reproduced by kind permission of Dr Benjamin
Suchoff, Trustee of the Bart6k Estate.



f.( Jb2

IA -

of K ?/-/

~~?i~4it ji

_ g3 \

since in this example the calculations are written on a separate sheet from the music.
One might still argue that Bart6k was unaware of the significance of the numbering
here. In that case the same would have to apply to many other obvious or fundamental
relationships: the 1-2-3-5-8-5-3-2-1 sequence of the xylophone solo already men-
tioned in the Music for Strings; some of the 'Bulgarian' metres in pieces such as the

86 MUSIC ANALYSIS 2:1, 1983

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

finale of Contrasts ("3 divided principally 8+ 5, with smaller divisions of 3+2 + 3, etc.);
the intervallic 2-3-5-8 semitone aspects of pentatony; and some important motivic
progressions identified by Lendvai (1971: 36-9, 49), from the Miraculous Mandarin,
Dance Suite, Sonata for Two Pianos and Percussion and Divertimento, based on
progressive spans of 2, 3, 5, 8, 13 and then sometimes 21 semitones. This would also
have to apply to Bart6k's own numbering of the bars, with their Fibonacci divisions,
in the manuscripts of the Fugue from the Music for Strings and the third movement
of the Fourth Quartet. The numerical coincidences concerning the overall quaver
counts in the Sonata for Two Pianos and Percussion would have to be accepted as
completely fortuitous, as they do not follow the continuous temporal articulation of
the work across the three movements; and it would have to be accepted that intuition
achieved absolute golden section exactitude with just a single event in the first
movements of the Third Piano Concerto and Divertimento, but with no associated
proportional effect whatever on the rest of these movements. For Bart6k, a skilled
student of natural history, complete with curious preoccupations about the structure
of pine cones and sunflowers, to have repeatedly emulated their exact numerical
configurations, without ever being conscious of what he was doing, would suggest a
most unobservant nature - hardly posterity's documented view of him. On balance it
seems more probable that Bart6k quietly disposed of sheets of calculation connected
with his compositions.
A further line of enquiry is suggested by the possibility of a connection with the
proportional structuring of other composers. Lendvai (1983) includes some discussion
of Kodaily's music and lists correspondences between the two composers' tonal,
modal, intervallic and harmonic techniques, claiming that there are some occurrences
of golden section - though they are of dubious structural importance - in the
Psalmus Hungaricus and the prelude of Hdry Jdnos. The originality of these techniques
and especially their novelty in symphonic music suggest that Bart6k and Kodaily may
have shared such ideas early in their careers, probably during their years of folksong
research. Little correspondence or documentation of their work together is available
(Kodaily once stated that what Bart6k and he 'had to say to each other could be best
expressed by word of mouth': quoted in Crow 1976: 126), but a link is supplied by
Kodaily's early Miditation sur un motif de Claude Debussy (based on the cyclic motive of
Debussy's String Quartet, and also using the opening motive of Ravel's String
Quartet). Like the Fugue in Bart6k's Music for Strings, it begins pp and ends ppp,
and its form is also directed towards a central climax marked fff. Bar lengths vary but
the crotchet pulse is constant, so that the 98 bars of the ternary form produce a total of
508 crotchet beats. Golden section of 508 is 314; this point falls towards the end of b.
67, in the middle of the two climactic bars of fff (bs 67-8), forming one of the most
clearly-defined of all golden section dynamic arch forms.
The coincidence has a particular savour since this homage to Debussy was com-
posed in 1907, the year Kodaily drew Bart6k's attention to Debussy. In 1907 Debus-
sy's most recent publication was the first set of piano Images of 1905, which was soon
in Bart6k's concert repertoire (Demtny 1977:173). Two of the three pieces in that
collection, Reflets dans l'eau and Mouvement), similarly begin pp, and finish either ppp
(Reflets) or pp, presque rien (Mouvement), with their main climaxes-ff and fff
respectively - placed at their overall points of golden section. The remaining piece,
Hommage a Rameau, has different large-scale proportions but a similar dynamic
shape, and is built very clearly on Fibonacci numbers, as is Reflets (see Howat 1983:
23-9 and 140-4).

MUSIC ANALYSIS 2:1, 1983 87

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

Mouvement must have interested Bart6k particularly since it is based on a polarity

between C and F?, a relationship subsequently central to Bart6k's musical imagery,
and especially important in Duke Bluebeard's Castle (1911), the most substantial work
Bart6k produced in the six years following his encounter with Debussy's music. Both
the plot and musical language of the opera are infused with the aesthetic and the
musical techniques of Debussy's Pellias - on the symbolic plane, M6lisande's fear of
invasion matched by Bluebeard's, and in the music itself, its speech-like setting of the
text and its synthesis of pentatonic, whole-tone and 'acoustic' scales (analysed in
Lendvai 1983). Indeed, Bluebeard begins by quoting the opening motive of Pellias
(with an octave transposition of the first note, as Debussy has it in Act 3 of Pelkas):16

Example 3. Andante

(low r s.trings)
Opening of PP
Bluebeard M ste,1ooso

Tre's modrer (Yniold's motif)

(lower strin (cYnodsmtf

Opening of ACT3 f
Pelleas ppsc. 4 rs expressp
U %I r -jI

Bluebeard also begins pp, ends ppp and reaches a lyrical fff climax, at the 'opening of
the fifth door' (six bars before fig. 75). This dynamic sequence matches both the
large-scale key sequence, F$-C-F$, and the progress on the stage from darkness to
brilliant light (at the opening of the fifth door) followed by the return of darkness that
closes the work. Since the drama is articulated on a large scale into eight portions by
the opening of the seven doors, the opening of the fifth door is placed exactly
five-eighths of the way through (surprisingly, Lendvai does not mention this), and
this design recurs later, over the same tonal and dynamic span and at precisely the
same ratio of five-eighths (55/88 bars), in the Fugue of the Music for Strings.
The most speculative aspect of Lendvai's work is the attempt to project analytical
findings back a stage in the creative process in order to explore the symbolic potential
of the structures and techniques used by the composer. Lendvai (1983) examines the
relation of the explicit imagery in Bart6k's stage and choral works to what he regards
as the implicit symbolic value of parallel musical structures in the symphonic and
chamber works. He traces relationships also with symbolism of shapes and motion in
sources ranging from Dante to Faust, and he specifically relates the form and imagery
of Bluebeard to Vor6smarty's fairy play Csongor and Tiinde. Similar processes traced

88 MUSIC ANALYSIS 2:1, 1983

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

from other stage works, the Quartets, the Sonata for Two Pianos and Percussion and
the Music for Strings, relate them to myths and creation archetypes - in the
last-named work through details like the Fugue's use of celeste and polarities of
rhythm and metre elsewhere. For all the risks involved, Lendvai's conclusions make
some fascinating reading and are less vulnerable to the oversimplifications of his
technical analyses. Szabolcsi's ideas (1965) on the Miraculous Mandarin and on
Bart6k's choice of stage works corroborate this synthesis of ideas, at least in principle,
and the formal correspondences seen above between Bluebeard and the Music for
Strings lend further support to similar interpretations of the symphonic works. Louis
Galanffy, a Bart6k pupil, recounted that Bart6k sometimes talked of creation arche-
types informing works like the Music for Strings and the Sonata for Two Pianos and
Percussion: for example, Bart6k reportedly suggested that the opening of the latter
was an evocation of the world evolving through primeval gaseous and molten states
and the beginning of measured time.17 This last aspect suggests specific corroboration
of the way in which the Sonata for Two Pianos and Percussion, unusually, inaugurates
its rhythmic and its proportional articulation only in b. 2 (see Fig. 3).
A similar interpretative clue comes from Kairpaiti's anecdote (1975:238) of Bart6k's
remark to Endre Gertler during a performance of the third movement of the Music for
Strings: 'Listen! The sea!' This movement is cast in Fibonacci spiral dimensions (Fig.
7) and comes between the fast second movement and the finale; in the overall form of
Debussy's La mer the nearest parallel section - linking the work's fast second
movement to its finale proper - is an introductory passage (up to fig. 46) consisting
of 55 bars and laid out as a Fibonacci sequence in an identical spiral configuration (see
Howat 1983:96-7). In the Debussy passage the music's imagery suggests wind and
storm - Lockspeiser was reminded of 'vortices and whirlpools' (1963:60) - and
Lendvai's tag for the lead into the central structural vortex of the corresponding
Bart6k movement is the 'roaring of the wind' (1971:29; see also Fig. 6a above).
Combining that with the spirals, we have here standard symbols of mystery, magic
and initiation, prominent in literature of that genre from all epochs.18
Does all this imply that Bart6k had an interest - as Debussy certainly did - in the
esoteric? The lack of external evidence in Bart6k's case should not negate the
question. In Debussy's case, research into the literary influences surrounding his
formative years (Howat 1983: 163-78) traces some firm preoccupations on the part of
his friends and colleagues with esotericism, numerology, cabbala, proportions and
geometry as applied to art. Some of these mainly Symbolist figures were also highly
knowledgeable scientists, reminding us that - an important point with Bart6k - the
avant-garde milieux of those days did not know the wide gulf now customary
(especially in the Anglo-Saxon mind) between science and parascience. To investigate
Bart6k from this angle invites exploration of the literary sources and influences
around him, and around colleagues of his such as Bdla Balizs (the librettist of
Bluebeard), during and after their formative years, as well as of the relationships
between Hungarian literary movements and the French ones of the fin-de-sicle and
after. Naturally, number manipulation and the symbolism of the noumenal are not a
monopoly of esotericism, and the most fruitful routes of exploration may eventually
lead elsewhere; but the music poses these questions, and it is as unscientific to rebuff
them as to jump to rash conclusions.
Finally, proportional analysis can sometimes be editorially helpful, though its use
there must be very circumspect. Brian Trowell's analyses (1979) of highly sophistic-
ated numerology in music by Dunstable have proved themselves practical by helping

MUSIC ANALYSIS 2:1, 1983 89

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

to solve paleographic problems of note values in manuscript sources. Analysis of

Debussy's La cathidrale engloutie, yielding Fibonacci proportions, corroborates the
evidence from a piano roll recording of the piece made by Debussy, indicating tempo
doublings not marked in the score (Howat 1983: 159-62).19 A further example is
guardedly offered from the finale of Bart6k's Third Piano Concerto. Since Bart6k
died before editing the movement, Tibor Serly and others added performing instruc-
tions, including the indication quaver = quaver across the transition from 3 to 2 at the
printed b. 427 (the finale's printed bar numbers include the 137 bars of the slow
movement). This tempo relationship is problematic in that it either rushes the gently
bucolic preceding section (bs 396-426) or else labours the following nimble fugato (bs
427-72). The strong rhythmic impulse of the bar sequence in the rest of the move-
ment suggests that at this transition the basic pulse of the bar should be maintained -
that is, minim = dotted crotchet at b. 427 and vice versa at b. 473 - and this indeed
solves the problem in performance. (The change to crotchet notation at the printed b.
644 is similarly taken as continuing the basic bar pulse, though probably at a slightly
attenuated pace, and not at the impossible editorial suggestion of dotted minim = 96.)
If one takes this bar-to-bar basis for measuring the movement, the golden section of
its 631 bars (remembering to omit the slow movement from the count) occurs after
390, within a bar, or three quaver beats, of the movement's recapitulation at the
printed b. 527; this matches the golden section positioning of the first movement's
recapitulation in the same concerto, mentioned above.
Remarkable proportions can be found in a surprisingly large range of music, in
some cases suggesting only subconscious application by the composer, and in some
cases telling us little about the music - though these two categories are by no means
completely coincident. Some accurate golden section constructions by Schubert,
Ravel20 and Faure are listed in Howat (1983: 187-93). At least two of these composers
are hardly to be normally associated with that type of musical thinking. But if these
occurrences were subconsciously intuited, then an urgent sense of proportional order
must have taken the composers powerfully by stealth, for the proportional systems are
accurate, highly detailed, and make logical sense of otherwise very idiosyncratic


Antokoletz, Elliott (1975): 'Principles of pitch organization in Bart6k's Fourth String

Quartet', unpub. Ph.D. dissertation, City University of New York.
-(1983): The music of Bela Bart6k: a study of tonality and progression in 20th-century
music, Berkeley, University of California.
Bachmann, Tibor, and Bachmann, Peter J. (1979): 'An analysis of B&la Bart6k's
music through Fibonacci numbers and the golden mean', Musical Quarterly, Vol.
65, No. 1, pp. 72-82.
Bart6k, Bela (1976): Turkish folk music from Asia Minor, edited by Benjamin Suchoff
with an afterword by Kurt Reinhard (New York Bart6k Archive, Studies in
Musicology, 7), Princeton, Princeton University.
-(1976a): Essays, selected and edited by Benjamin Suchoff (New York Bart6k
Archive, Studies in Musicology, 8), London, Faber.
(1981): Piano music of B&la Bart6k, The Archive Edition, edited by Benjamin
Suchoff, 2 vols, New York, Dover.
Bart6k fils, Bela (1966): 'Remembering my father, B61a Bartok', New Hungarian
Quarterly, Vol. 7 No, 22, pp. 201-3.

90 MUSIC ANALYSIS 2:1, 1983

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

Benjafield, John (1976): 'The golden rectangle: some new data', American Journal of
Psychology, Vol. 89, No. 4, pp. 737-43.
Camp, E. J. P. (1968): 'Temporal proportion: a study of sonata form in the piano
sonatas of Mozart', unpub. Ph.D. dissertation, Florida State University.
Church, A. H. (1904): On the relation of phyllotaxis to mechanical laws, 3 vols, London,
Williams and Norgate.
Colman, Samuel (1912): Nature's harmonic unity. A treatise on its relation to proportional
form, ed. C. A. Coan, New York, G. P. Putnam's Sons.
-(1920): Proportional form. Further studies in the science of beauty (with C. A. Coan),
New York, G. P. Putnam's Sons.
Cook, Theodore A. (1903): Spirals in nature and art, London, John Murray.
-(1914): The curves of life, London, Constable.
Crevel, Marcus van, ed. (1959; 1964): Jacobus Obrecht: opera omnia, Vols 6 and 7
(Missae Sub Tuum Presidium and Maria Zart), Amsterdam, 1959 (Vol. 6), 1964
(Vol. 7).
Crow, Todd (1973): Review of Lendvai (1971) in Notes, Vol. 29, No. 4, pp. 722-4.
-(ed.) (1976): Bart6k studies (articles reprinted from the New Hungarian
Quarterly), Detroit, Information Coodinators.
Demeny, Jinos (1977): 'Zeitgenossische Musik in Bart6k's Konzert-repertoire',
Documenta Bart6kiana, Vol. 5, Budapest 1977, pp. 169-76.
Fassett, Agatha (1958): The naked face of genius: Bela Bart6k's last years, London,
Gollancz, reprinted as Bila Bart6k: the American years, New york, Dover, 1970.
Fechner, Gustav (1876): Vorschule der Aesthetik, 2 vols, Leipzig, Breitkopf & Hartel.
Fennelly, Brian (1973): Review of Lendvai (1971) in Journal of Music Theory, Vol. 17,
No. 2 (Fall), pp. 330-4.
Franz, Marie-Louise von (1975): Numbers and time, London, Aldus Books.
Ghyka, Matila (1927): Esthetique des proportions dans la nature et dans les arts, Paris,
-(1931): Le nombre d'or, 2 vols, Paris, Gallimard.
-(1946): The geometry of art and life, New York; reprinted New York, Dover,
(1952): Philosophie et mystique du nombre, Paris, Payot.
Graue, Jerald C. (1981): 'Novel symmetries in Bart6k's piano music', unpub. paper
read at the annual meeting of the American Musicological Society, Boston, 13
November 1981.
Hambidge, Jay (1920): Dynamic symmetry: the Greek vase, New Haven, Yale
-(1924): The Parthenon and other Greek temples: their dynamic symmetry, New
Haven, Yale University.
-(1948): The elements of dynamic symmetry, New Haven, Yale University.
Howat, Roy (1983): Debussy in proportion: a musical analysis, Cambridge, CUP.
Kairpaiti, Jainos (1975): Bart6k's string quartets, trans. Fred Macnicol, Budapest, Cor-
vina; original Hungarian edition Bart6k von6sn'ysei, Budapest, Zenemikiad6,
Kramer, Jonathan (1973): 'The Fibonacci series in 20th-century music', Journal of
Music Theory, Vol. 17, No. 1, pp. 110-48.
Larson, Paul (1978): 'The golden section in the earliest notated Western music',
Fibonacci Quarterly, Vol. 16, No. 6, pp. 513-5.
Le Corbusier [pseudonym of Jeanneret, Charles Edouard] (1954): The modulor. A

MUSIC ANALYSIS 2:1, 1983 91

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

harmonious measure to the human scale universally applicable to architecture and

mechanics, trans. P. de Francia and A. Bostock, London, Faber.
Lendvai, Ern6 (1955): Bart6k stilusa [Bart6k's style], Budapest, Zenemikiad6.
-(1962): 'Duality and synthesis in the music of Bel61a Bart6k', New Hungarian
Quarterly, Vol. 3, No. 7, pp. 91-114.
-(1971): Bdla Bart6k: an analysis of his music, London, Kahn & Averill.
- (1971A): Bart6k k6ltoi vilaga [Bart6k's poetic world], Budapest, Szepirodalmi
-(1983): Bart6k and Kodaly, Budapest, Editio Musica; originally issued privately,
4 vols, by the Zoltin Kodily Pedagogical Institute, Kecskemet, 1976-80.
Lockspeiser, Edward (1963): 'Debussy's concept of the dream', Proceedings of the
Royal Musical Association, Vol. 89, pp. 49-61.
Lowman, Edward (1971): 'An example of Fibonacci numbers used to generate
rhythmic values in modern music', Fibonacci Quarterly, Vol. 9, No. 4, pp. 423-6,
- (1971a): 'Some striking proportions in the music of Bela Bart6k', Fibonacci
Quarterly, Vol. 9, No. 5, pp. 527-8, 537.
McManus, I. C. (1980): 'The aesthetics of simple figures', British Journal of Psy-
chology, Vol. 71, Part 4, pp. 505-24.
Maxwell, Judith Shepherd (1975): 'An investigation of axis-based symmetrical struc-
tures in two compositions of Bela Bart6k', unpub. D.M.E. dissertation, Univer-
sity of Oklahoma.
Norden, Hugo (1972): 'Proportions and the composer', Fibonacci Quarterly, Vol. 10,
No. 3, pp. 319-23.
Pascoe, Clive (1973): 'Golden proportion in musical design', unpub. D.M.E. disser-
tation, University of Cincinnati.
Perle, George (1967): 'The string quartets of Bela Bart6k', essay accompanying the
recording by the Taitrai String Quartet of the six Bart6k quartets, New York,
Dover; reprinted in A musical offering: essays in honor of Martin Bernstein, New
York, Pendragon, 1977, pp. 193-210.
Pousseur, Henri (n.d.): 'Haydn et Fibonacci (a propos de la Sonate N' 30 en Re
majeur)', unpublished.
Powell, Newman W. (1979): 'Fibonacci and the Golden Mean; rabbits, rumbas and
rondeaux', Journal of Music Theory, Vol. 23, No. 2, pp. 227-73.
Purce, Jill (1974): The mystic spiral: journey of the soul, London, Thames and Hudson.
Reich, Willi (1965): The life and work of Alban Berg, trans. Cornelius Cardew,
London, Thames and Hudson.
Rogers, Michael R. (1977): 'The Golden Section in musical time: speculations on
temporal proportion', unpub. Ph.D. dissertation, University of Iowa.
Rutter, John (1977): 'The sonata principle', Open University Course A 241 (Elements
of music), Milton Keynes, Open University.
Sallis, F. (1983): ' "K6rusmivei [Choral education], studium in vivo", contribution ai
l'6tude de la musique chorale de Bela Bart6k', unpub. M.Mus. dissertation,
Universit6 Laval, Quebec.
Saygun, A. Adnan (1976): Bila Bart6k's folk music researches in Turkey (ed. Liszl6
Vikir), Budapest, Akad~miai Kiad6.
Solomon, Larry J. (1973): 'Symmetry as a determinant of musical composition',
unpub. Ph.D. dissertation, University of West Virginia.
Somfai, L~izl6 (1981): 'Bart6k, B6la' (sections 6-10), in The New Grove Dictionary of
Music and Musicians, Vol. 2, pp. 205-18.

92 MUSIC ANALYSIS 2:1, 1983

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

Souvtchinsky, Pierre (1939): 'La notion du temps et la musique', La Revue Musicale,

No. 191, pp. 70-80.
Stockhausen, Karlheinz (1959):' . . . How time passes . ..', Die Reihe, Vol. 3, Bryn
Mawr, Pennsylvania, Theodore Presser, pp. 10-40.
Stravinsky, Igor (1947): Poetics of music, London, OUP.
Szabolcsi, Bence (1965): 'Bart6k's Miraculous Mandarin', New Hungarian Quarterly,
Vol. 6, No. 20, pp. 108-24.
Szentkirailyi, Andris (1976): 'Bart6k's Second Sonata for Violin and Piano (1922)',
unpub. Ph.D. dissertation, Princeton University.
Thompson, D'Arcy Wentworth (1917): On growth and form, Cambridge; new editions
1942 & 1961, Cambridge, CUP.
Trowell, Brian (1979): 'Proportion in the music of Dunstable', Proceedings of the
Royal Music Association, Vol. 105, pp. 100-41.
Wittkower, Rudolf (1960): 'The changing concept of proportion', Daedalus, Vol. 89,
pp. 199-215.
Zeising, Adolf (1854): Neue Lehre von den Proportionen des menschlichen Korpers,
Leipzig, Rudolph Weigel.
-(1884): Der goldne Schnitt, Halle, W. Engelmann.


1. Lendvai's publications are listed in the References.

2. Crow (1973) and Fenelly (1973) - apparently the only two complete reviews of
Lendvai (1971) - Bachmann and Bachmann (1979), Kramer (1973), Lowman
(1971a), Maxwell (1975), Rogers (1977), Sallis (1983), Solomon (1973), Szent-
kiraily (1976), Antokoletz (1975) and Perle (1967); the last two trace additional
proportional occurrences from Bart6k's Fourth and Fifth Quartets. Graue (1981)
extends the field to a number of proportional ratios in Bart6k's earlier piano
music. I am grateful to Dr Antokoletz, to the late Dr Jerald Graue, and to Dr
Marie Rolf for information useful to this article, and to Geoffrey Walker for
helpful comments.
3. Antokoletz (1975:100-6) is critical of Lendvai's theories of Bart6k's intervallic
structure; see also Antokoletz (1983). Kairpaiti (1975:155-6) also provides a
thorough critique of the intervallic aspects of Lendvai's theories.
4. The role of golden section in nature is documented by Church (1904), Colman
and Coan (1912, 1920), Cook (1903, 1914) and Thompson (1917). Its role in the
visual arts is traced by Wittkower (1960); one of its most recent applications was
Le Corbusier's 'modulor,' an architectural grid based on golden proportions
derived from the human body. See also Ghyka (1927, 1931, 1946, 1952) and
Hambidge (1920, 1924, 1948).
5. This sequence is named after the 19th-century French mathematician Edouard
Lucas, who himself gave the name 'Fibonacci sequence' to the 1, 2, 3, 5 .
series. 'Fibonacci' (= Figlio [son of] Bonaccio) was Leonardo of Pisa, the 12th-
century mathematician who was largely responsible for introducing arabic num-
bers to Europe and who wrote about the series later named after him. The
extensive manifestation of the series in science, art and nature is studied in the
Fibonacci Quarterly, which includes at least four musical articles (Lowman 1971
& 1971a, Norden 1972 and Larson 1978).

MUSIC ANALYSIS 2:1, 1983 93

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

6. The reader interested in the associated questions of how we may experience time
either generally or more specifically in music can find discussions in Bergson
(1910), Carter (1977), Erickson (1967), von Franz (1975), Rogers (1977), Souvt-
chinsky (1939), Stockhausen (1959) and Stravinsky (1947).
7. One criterion for measurement should be clarified here. If a piece ends after, say,
60 bars (assuming the bar as unit), the halfway point and two-thirds point come
after 30 and 40 bars, that is, at the beginning of bs 31 and 41, not of bs 30 and 40.
This arithmetic thus refers to bars (or other units) completed, and is adhered to in
the following analyses. This might seem to be labouring the obvious, were it not
that it has already tripped up some of the investigators of proportion mentioned
in this article, some cases of which are examined later.
8. Lendvai (1971) is a considerably curtailed version of his investigations. A com-
plete presentation in English (Lendvai 1983) is forthcoming; this was originally
issued privately (4 vols, 1976-80) by the Zoltin Kodily Pedagogical Institute,
Kecskemet, and I am grateful to Dr Mihaily Ittzes of that Institute for providing
me with a copy. References in the following pages are all to Lendvai (1971), but
all Lendvai's interpretations discussed here occur in the same form in the other
sources: Lendvai 1955, 1971a (an updated version), 1962 (the first appearance in
English of his theory) and 1983.
9. Antokoletz (1975:102-3) and Kairpaiti (1975:155-6) expose some weaknesses and
oversimplifications in this aspect of Lendvai.
10. Antokoletz (1975:100-1) observes that 'dominant' and 'subdominant' are
inappropriate terms for the functions these key relationships sometimes perform
in Bart6k's music.
11. The earlier version, which leads to the recapitulation via 15 bars of dominant
pedal on G, survives in the intermediate draft 75SID2 in the New York Bart6k
Archive. Bart6k's final autograph copy, 75FSID1, is an amended copy from the
same transparent master as the earlier draft (by Ozalid process; Bart6k normally
worked thus to save redundant copying by hand and to guard against postal loss),
with the obsolete passage indicated by 'vi--de' and the new version inserted on
separate sheets. I am grateful to Dr Benjamin Suchoff, the Trustee of the Bart6k
Estate, for kindly allowing me to consult material belonging to the Estate.
12. Judith Maxwell (1975:81-2, 109-12 and 117-19) identifies other intricacies in the
organization of the moment, including symmetries of tempi, metre, dynamics
and timbre.
13. Lendvai's published writings in fact contain no reference to the consultation of
manuscript sources.
14. Cf. Alban Berg's letter to Schoenberg, describing the construction of his Cham-
ber Concerto:

I realize that-in so far as I make this [use of numbers] generally known-my

reputation as a mathematician will grow in proportion ( . . . to the square of
the distance) as my reputation as a composer sinks (Reich 1965: 147).
15. This song was not included in Bart6k's final drafts of Turkish folk songs (Bart6k
1976, Saygun 1976); it was No. 33 of the Columbia/HMV collection recorded in
the early 1930s in cooperation with the city of Istanbul. Suchoff documents this
fully in his preface to Bart6k 1976 (pp. 5, 15 with n. 33).
16. The same quotation opens the second of Bart6k's Four Dirges, Op. 9a (1909-10).
Kirpiti (1975:30) reports, from several sources, Bart6k's allusion to M61isande's

94 MUSIC ANALYSIS 2:1, 1983

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms

death music at the same pitch in the last movement of the Second Quartet
(around fig. 7 in the score), which was composed four or five years after
Bluebeard. Many aspects of Bluebeard represent a synthesis of elements from
Pellias with elements from Paul Dukas's Ariane et Barbe-bleue of 1907 (text also
by Maeterlinck; and one of Barbe-bleue's wives is named M6lisande).
17. This information comes from Galanffy's former pupil John Aielli of KUT Radio,
University of Texas at Austin.
18. Purce (1974) documents the symbolism of the spiral in various periods of art
19. Bars 7-12 and 22-83 are played at twice the speed of the rest of the piece, thus
solving practical problems of tempo. It is not known if Bart6k, who played the
piece, was aware of this.
20. 'Oiseaux tristes' from Miroirs (1904-5), built on the Fibonacci series; it was
probably the first Ravel piece in Bart6k's concert repertoire (Demeny 1977).

MUSIC ANALYSIS 2:1, 1983 95

This content downloaded from on Fri, 03 Jun 2016 18:21:27 UTC
All use subject to http://about.jstor.org/terms