Anda di halaman 1dari 6



 
           !"  # $% & $#' ( )*+ -,. /  &   ,
0 1!2 3 4%5*67 2 8 41

Dirk Moelants
IPEM-Dept. of Musicology
Ghent University, Belgium
Dirk.Moelants@rug.ac.be

The American composer Morton Feldman (1926-1987) was one of the key figures in the New-York school of
experimental music of the 1950ies. Most typical for his style is the use of soft, long sounding chords, composing
from sound to sound, each time creating a new color field. The "Last Pieces" (dated March 1959), are typical for
Feldmans works of the late 50ies. In the beginning of the decade he mainly used graphic notations, creating
freedom of both rhythm and pitch. But here we see an evolution towards a notation in which the pitches are
exactly determined, while the rhythmic freedom from the graphic scores is kept (Sani, 2000). Later, in the 70ies,
he will go further on the path of determination, specifying rhythmic values, dynamics and timbral elements. The
work consists of four parts, each with their own tempo-indication: 1.
; 2.
; 3.
; 4.
. The fourth movement stands a bit apart. It uses an asynchronous temporal process, where every
voice (hand) can have its own timing, independent of the vertical alignment of the notes in the score. Since this
introduces a different approach for the performers and extra variants in the analysis, the fourth part has been
omitted from the analyses below.
In the first three parts, the vertical position of the chords is fixed, but still the durations are free. The
determination of the time intervals is left to the performer, and only a general indication of the tempo is given.
This is also reflected in the notation, where we only find black note-heads, and some occasional grace notes and
fermata signs. An example of the notation can be seen in Figure 1. Yet, the apparent freedom of interpretation
does not mean that a merely random timing should be applied. On the contrary: Feldman gives the performer
responsibility rather than freedom. The pianist has to determine the timing of every single chord, starting from
the characteristics of the sound in the given acoustic circumstances. Thus he forces the performer to start a
search: which time interval fits best with this sound (Borio, 1996). An additional limitation is given by the tempo
indication, which forces the performer to choose for certain idea of speed, changing from slow to fast and back
to very slow between the three pieces.

M N O K L PQ R I K S N T JI R U O R U U V W X YZ [ \ ]!^ _ \ ` a ^ bc X d a e \ f Zd X W ` X W W g h i j!k 9 l : m ; n <!o p =q 9 k ;n> l ? =m @r!q A m B m C s ?t D u; h E o FvC w B i Gl > n sB q G x G ml i h Hk I i Jh K h L
k q ih l y zz l x {

|} ~  | ~      |%   ! * }   }  


The focus in this paper will be on the length of the time intervals chosen by performers. What do they consider
as fast and slow, how much variation can be found and can the variation be connected with specific musical
parameters, are there significant differences between performers These aspects will then be interpreted within
a general framework of rhythm and tempo perception. The fundamental methodological goal is an interchange
between the study of music performance and perception theory. To show how the study of a specific artistic
product can give information about perception, and how perception theory allows us to clarify aspects of music
performance.
For this study the timing in four performances of the first three parts of Feldman's "Last Pieces" has been
analyzed. Each of the performances was of a different nature, and was made by a specialist of contemporary
classical music: 1. Steffen Schleiermacher's (SS) recording, found on the CD "Morton Feldman: Works for Piano
II" [HAT HUT ART CD 6143]; 2. A live performance by Anne Piret (AP), recorded on the 7th of may 1992 at
the Logos Thetraeder in Ghent, Belgium; 3. A 'home-recording' by Geert Callaert (GC) made on the 18th of
august 2000; 4. A realization in MIDI made by the composer/guitarist Dave Philips (DP)
[http://www.bright.net/~dlphilp/dp_music2.html].

* !
!
! * !

Mean
Minimum
Maximum

5093
2299
9521

1327
356
2955

2109
1180
3950

4522
1330
8700

Mean
Minimum
Maximum

986
596
2816

560
150
2269

1081
230
2580

511
110
1460

Mean
Minimum
Maximum

5429
2715
8914

1528
727
2751

2700
1650
5040

4381
1840
6780

*          %     %  * 
%  
! %
 ! ! 

In Table 1, an overview is given of the mean inter onset intervals (iois) and the extreme values, this for each of
the three parts examined, and for each of the four performances separately. Note that elements written down as
grace notes or with a fermata sign were omitted for the numbers in this table. They will not be included in any of
the calculations below, mainly because their number is too small to enable statistically significant results.
Already at first sight we see very big differences between the four performances. In other words, there is clearly
no agreement about what exactly is slow, fast or very slow.
For the first (slow) and third (very slow) parts we find similar results, but with large differences between the
performers: averages around 5 seconds for SS and DP against averages around 1.5 seconds for AP, while GC
takes a middle position. A plot of the results is given in Figure 2a, the visualization shows both the striking
differences between the four interpretations and the similarity of performed iois between parts 1 and 3, within
each of the performances. An ANOVA shows a significant difference between the iois of parts 1 and 3 only for
GC (F(1,87) = 21.482, p < 0.001), for the three other interpretations, the difference does not reach significance.
3000

10000
9000

2500

8000
7000

2000
6000
5000

1500

4000
SS

2000

AP

1000

GC

0
N=

DP
40

40

40

40

49

49

49

49

SS
AP

ioi in ms

ioi in ms

1000
3000

500
GC
0
N=

DP
74

74

74

74

     
        
    

  
 
 
   
     
PART 1

PART 3

PART 2

In Figure 2b the spread of the iois in part two (fast) is represented. Also here we find clear differences between
the four interpretations. Two groups can be distinguished: SS and GC, with averages around 1 second, and AP
and DP, with averages around half a second. In any case the distinction between the fast second part and the
other two parts is clear, ANOVAs indicating significant differences for each performer.

After we gave a general overview of the timing characteristics we can explore some specific elements. First we
take a look at the amount of variation in the iois used in every performance. To neutralize the considerable
differences in tempo, the iois were normalized, expressing them as a percentage of the mean in each piece and
performance. For these normalized data the standard deviations were computed, an overview is given in Table 2.
Comparing the four performers we can distinguish two groups: SS and GC, with average standard deviations

around 32%, and AP and DP with average st.dev. around 45%. SS is the only one who keeps a similar spread
through each of the three pieces. Comparing the three pieces, the standard deviations seem to rise together with
the indicated tempo. However, the differences between performers are very large in the first two pieces, the third
being the only one that is executed fairly regularly in each of the four interpretations.
Another aspect of variance is the evolution in time. Is there spread of longer and shorter elements throughout
each piece, or do we find linear evolutions from shorter to longer iois or vice-versa. To investigate this aspect,
the correlations between the iois and their rank number were computed. Only one significant correlation was
found: DP in the second part. However it could be reduced to the use of a kind of final retard, lengthening the
last two iois considerably. When the two last elements were removed, the correlation dropped far below
significance. So, we can conclude that the performers dont make gradual changes in speed, but spread out the
longer and shorter iois more or less equally.
SS

AP

GC

DP

LP1

33,46

47,27

26,45

50,5

mean

LP2

30,41

65,59

48,81

49,51

48,58

LP3

30,89

22,2

23,32

33,68

27,5225

mean

31,58667

45,02

32,86

44,56333

38,5075

39,42

   !#" $ %  & '  ( '*)+ , -  % - . & /. 0 % 1 & . ( 2  - 3 '- . - 4 /
5 6 7+6 8 9*: 6 7+6 ; < = 6 5 >
Despite the differences in tempo, the different interpretations might show a similar timing profile, making the
same elements relatively long or short. To investigate this, the correlations between the iois of each
interpretation were computed for each of the three pieces. This analysis shows only one highly significant
correlation: between GC and AP in the first part (r = .536, p < 0.001). In the second part no significant
correlations could be found, and in the third part marginally significant correlations were found between the
timing of GC and those of SS and AP (for both r = .346, p < 0.05), no significant correlation was found between
SS and AP. So despite the total freedom of timing for the performers, there are some indications for similar
timing strategies independent of the tempo. Yet there is clearly not a uniform performance strategy shared by all
of the performers or in each of the three parts.

? @ A B C D @ E D F AHG I DJ+C K ? E L B E F @ G D @ G
Is the choice of the time-intervals completely random, or are there any elements in the score that have a uniform
influence on the produced iois. To investigate this the text of the score was transformed into numerical values
and the correlations between these variables and the performed iois were computed (Moelants, 2000). Since the
rhythm is not specified and timbre, dynamics and location are presumed to be constant, only aspects of pitch
were retained for this statistical analysis. For each element, the highest and lowest note, the ambitus (difference
between highest and lowest note), the mean pitch, the number of pitches and the (theoretical) dissonance were
collected. It must be noted that in all three pieces an almost perfect match between the last two parameters was
found: the more notes in the chord, the more dissonant. Therefore, the two can be regarded as one parameter: the
complexity of the chord, below only the numbers for the dissonance will be given.
In the first part, the complexity has a significant influence on the performed iois in three of the four
interpretations. We found positive correlations for SS (r(diss, ioiSS) = .362, p < 0.05) and DP (r(diss, ioiDP) =
.322, p < 0.05), but a negative correlation for AP (r(diss, ioiAP) = -.313, p < 0.05). For SS the positive correlation
between the iois and the complexity of their content is also found in the third part (r(diss, ioiSS) = .384, p <
0.01). AP on the other hand shows a similar behavior in the second part (r(diss, ioiAP) = -.276, p < 0.05).
Additionally significant positive correlations at the p < .05-level were found between the iois for AP in the
second part and the lowest, highest and mean pitch. Apart from these, no other significant effects were found.
The results indicate that two different timing strategies are possible: SS gives more complex sounds more time to
resonate, while AP gives more time to relatively consonant elements. None of the performers, however, shows a
truly consequent timing behavior in relation to the complexity. The influence of the other score-based parameters
is marginal, only a lengthening of relatively high-pitched elements by AP in the second part was found.

MON P N Q R SUTWVXN SUV Y Z\[]^ [UT`_+N Q a N b^ c V P


In Figure 3 an overview of the most important thresholds in rhythm and tempo perception is given. Please note
that the numbers, indicating iois in ms, are approximate. Hence, they should be considered as a rough indication,
as symbols for a zone. A thorough discussion of the experimental data justifying the choice of these numbers
and the theoretical background of the different concepts goes beyond the scope of this paper. Therefore I will
only give a short characterization of the relevant thresholds and the classes they define.

The numbers at the left of the figure are not so important for the present paper. They indicate the basic resolution
of our temporal perception (acuity), the possibility to perceive distinct elements or the theoretical lower border of
rhythm (fusion threshold) and the smallest ioi used in music or the practical lower limit of rhythm. Of course if
we are talking about slowness, the most important perceptual zones are those at the middle and right parts of the
figure, representing longer time-spans.
We distinguish two perceptual functions, the more general notion of rhythm perception and the more specific
notion of tempo perception. Under rhythm perception we want to treat everything that concerns the perception of
the succession of (musical) elements in time. This includes the position of the different elements in time, as well
as the way in which these elements give weight to specific points in time (accent), the different ways in which
these elements and time-points are grouped, and the movements (gestures) that can be associated with them. In
our definition, rhythm is only present if there is a perceptible and quantifiable succession. Thus, when two
elements are so close in time that their points of occurrence can not be distinguished (fusion), or when two
elements stand so far from each other that actively linking them is not possible anymore, we dont consider this
succession as rhythmic. For the upper limit of rhythm our thresholds are based on memory capacity. We
distinguish two zones: one up to 3 seconds, coinciding with the capacity of our working (or echoic) memory,
within which we can process different elements as a whole, compare and group them. A second zone coincides
with our short-term memory capacity and goes up to 10 or 20 seconds. In this zone, the important notion of
grouping is lost, the different elements appear as isolated, but they will still be remembered and recognized as
part of one stream.
One aspect of rhythm is the beat or pulse, a fairly regular movement underlying and structuring rhythmic
sequences, often associated with quasi-periodic movements like dancing and foot-tapping. The rate of the beat is
called the tempo. At one end also tempo perception is limited by the capacity of our working memory, since one
must be able to keep at least two successive beats within this memory buffer to create a naturally floating pulse.
This puts the maximum tempo at 3000/2=1500 ms. At the other end, the possibility to perceive a beat is lost
around 200 ms. Elements within this time-span are automatically grouped into perceptual units, that can no
longer be divided, even if such a fast regular movement would be possible. Between these points there is a zone
centered around 500 ms, in which tempo perception is optimal, the zone of preferred tempo, that can be
associated with the natural (resonance) frequency of human movement (Van Noorden & Moelants, 1999).
In the piece under investigation, the intention of the composer is to destroy the perceptibility of a (regular) beat,
by making the durations free, and letting the performer decide on the iois from one sound to the next. Therefore
tempo and beat as defined above can not be applied in this study. Nevertheless, the time zones normally
associated with tempo can be applied to elements in free succession to give them a characterization in terms of
speed. Here we will distinguish between iois associated with slow tempo (500-1500 ms), iois associated with
fast tempo (200-500 ms), and elements that will be automatically grouped into rhythmic Gestalts since they are
separated by a time interval smaller than 200 ms.

de f g h ijkl m n i op q e m r s i h s e i tur vh n w q n oxp y zq i o{ r+{ i h m i { q e r y


|\}~ }UW     W  }+  W \ 
Based on the model of rhythm and tempo perception outlined above, the iois found in the four performances of
Feldmans "Last Pieces" were divided into five categories: shorter than 200 ms, 200-500 ms, 500-1500 ms, 15003000 ms and longer than 3 seconds. Iois shorter than 50 ms or longer than 10 seconds were not found.
First lets take a look at the distribution for the first (slow) and third (very slow) parts. In figure 4 the
distributions of the iois in each of the four performances are represented. We see that SS definitely chooses for
iois that go beyond the capacity of our working memory, creating a succession of isolated events. This timing

strategy is followed in a less radical way by DP, who creates more variation by including clearly shorter iois
within a majority of long ones. AP, on the contrary, does not include any elements with a duration beyond 3
seconds. Her time intervals are centered around the limit of tempo perception, clearly preserving the possibility
of linking successive elements, using iois with a connotation of slowness. GC, finally takes kind of a middle
position, mainly choosing for time intervals between 1.5 and 3 seconds, but also including a fair amount of
longer intervals. It seems as if he wants to explore the limits of our memory capacity and to challenge our linking
possibilities.
1500-3000

500-1500

200-500

500-1500
> 3000

1500-3000
19,1%

20,2%

1500-3000
46,1%
51,7%
500-1500
77,5%

94,4%

75,3%

SS

DP

> 3000

AP

GC

> 3000

1500-3000

      H *
In Figure 5 we see the distributions of the iois in the four performances of the second part (fast). Also here SS
makes the clearest choice for one specific time category, in this case for the time intervals associated with a slow
tempo (500-1500 ms). GC roughly follows this interpretation, but uses a larger ranges of values, including
relatively large amounts of iois in the neighboring classes (200-500 and 1500-3000). AP and DP, on the other
hand, choose for faster values, centered around the zone of preferred tempo and including elements below 200
ms.
1500-3000

200-500

1500-3000

< 200

< 200

1500-3000

14,9%

9,5
%

13,5%

29,7%

500-1500
500-1500

40,5%

48,6%
37,8%

97,3%

47,3%

55,4%

SS
500-1500

GC

200-500

DP

500-1500

 U    U 
#  
A first remarkable finding is the considerable difference found between interpretations of the same piece with
the same tempo indication by different performers. Clearly there is no uniform interpretation of, at first sight,
simple concepts like slow, fast, and very slow. Also the difference between slow and very slow does not
seems to be very relevant to the interpreters, yielding similar results in each of the four interpretations.
It has been shown that the choice for different magnitudes of iois can be interpreted following a general model
of rhythm and tempo perception that divides the area in which we can perceive rhythm into five categories. In
the two slow parts, two performers choose for the slowest category of iois, mainly using intervals above three
seconds. This leads to the performance of isolated elements that can not be perceptually grouped with their
neighbors. The other two performers want to preserve the possibility to link one element to the next one, mainly
using time intervals below the 3000 ms threshold, but above the area of preferred tempo (500 ms), and thus
associated with slow movement. In the fast second movement mainly intervals that fall within the limits of
tempo perception (200-1500 ms) are used. But also here we can distinguish two groups, two performers having
an average ioi around the preferred tempo and including elements shorter than the 200 ms limit, and two
performers with averages around 1000 ms that mainly use values associated with a slow tempo.
All of the performances are rather consistent. Sometimes a large majority of the iois falls within one time
category, sometimes they are grouped around a border value. Despite the total freedom relatively few extremes
occur, there is no tempo drift, and the difference between fast and slow is clear. The performers seem to
approach the work as a whole, with a clear vision on the tempo characteristics of each of the movements.

200-500

AP

Another aspect is the apparent freedom, the durations are free, but bound by tempo categories, moreover the
freedom should be interpreted as responsibility, rather than as randomness. Feldman wanted the performer to
search for the way in which the piece sounded best (Kopstick Ames, 1996). Despite this, we dont find many
similarities between our four interpretations. Sometimes clearly different strategies were found, e.g. regarding
the treatment of more and less complex elements. Partly this dissimilarity might be explained by the different
nature of the recordings. All four were realized in completely different circumstances that might necessitate a
different interpretation of the same text in order to obtain the ideal sound.
Anyway this type of freedom and its notation give the composer a very simple way to destroy the beat. While
most composers of that period, especially those of the serial school, tried to destroy the sense of tempo by
writing extremely complex and detailed rhythms, Feldman just neglects the existence of a pulse. In this way not
only the listener, but also the performer can leave metrical structure behind and go straight to the sound of the
sounds.
Clearly two answers can be given to our initial question "what is slow?". One possibility is to choose for
intervals in the range 500-3000 ms, creating a rhythmical chain of events that follow each other at a slow pace,
but can still be actively linked and form groups together. The other possibility is a choice for intervals above the
3000 ms border, exceeding the normal buffer of our working memory. This leads to a succession of isolated
events, in which grouping and hierarchical perception disappear.
The different approaches can be related to the fundamental interpretation of the composition in its relation to
time. In the first approach the perception of horizontal relations, and thus a sense of movement, is preserved,
while in the second, everything is focussed on the moment. When we take a look at the writings of Feldman
himself, we find that he wants to destroy the memory, erase every reference to previous elements and thus
concentrate the listeners attention on the moment (Feldman, 1985). This philosophy would suggest a choice for
intervals over 3 seconds. But on the other hand, we can never really destroy the content, we never really hear a
sound in a composition just as we hear it out of the composition (Costello Hirata, 1996). Moreover analyses of
the pieces (DeLio, 1996) show some systematics in the horizontal relations, which can only be revealed to the
listener by keeping the iois below the 3 seconds threshold. These elements are in favor of the first type of
interpretation.
The main issue in this paper was to show how interpretation of a (somewhat enigmatic) musical score can be
clarified by an understanding of the basics of music perception. From the other side these performances can be
approached as an experiment in which the subjects are asked to give an interpretation of tempo categories. It
shows us that even a simple concept like slow can be interpreted in at least two clearly different ways,
depending on whether the performer wants to keep a link between successive elements or tries to isolate each of
them.

\
Borio, G. (1996). Morton Feldman e lespressionismo astratto: La costruzione di tempo e suono nelle miniature
pianistiche degli anni cinquanta e sessanta. In G. Borio & G. Taglietti (Eds.), H O+
+ (pp. 119-134). Lucca: una cosa rara-Lim.
Costello Hirata, C. (1996). The Sounds of the Sounds Themselves: Analyzing the Early Music of Morton
Feldman.    + \\ , (1), 6-22.
DeLio, T. (1996). Last Pieces #3 (1959). In T. Delio (Ed.) U H + + (pp. 39-68). Westport:
Greenwood Press.
Feldman, M. (1985).Conversation
Morton Feldman and Walter Zimmermann. In W. Zimmermann
    


between
(Ed.),
(pp. 229-244). Kerpen: Beginner Press.
Kopstick Ames, P. (1996). Piano (1977). In T. Delio (Ed.),            !"  . Westport:
Greenwood Press.
Moelants, D. (2000). Statistical Analysis of Written and Performed Music. A Study of Compositional Principles
Problems of Coordination and Expression in Punctual Serial Music, # $ % & ' ( )*$ +-,/. 012% 3 4 576. 3 . ( & 5 8 ,
9and
: (1),
37-60.
;-< = > ?@ ACB DFE GIH J @ K L M A DON E PO@Q> L D R H D D L M A . [Online]. Available:
Sani,
F.
(2000).
http://www.cnvill.demon.co.uk/mfsani1.htm [2000, September 18].
Van Noorden,
and Moelants, D. (1999). Resonance in the Perception of Musical Tempo, S T U V W X YT Z/[/\ ]
^ U _ ` a bL.,
\ _ \ X V a c , d e (1), 43-66.

Anda mungkin juga menyukai