Professional Documents
Culture Documents
Lehmann 1998
Lehmann 1998
e ABSTRACT
This study investigated an expert pianist's nine-month preparation for a public
music performance (redtal) through the collection of practice diaries and MIDI
recordings of the eight scheduled pieces. Recordings were made under the
experimentally varied conditions of solitary performance and public performance.
The practice diaries revealed that the expert (an advanced student performer)
allocated practice time consistently across the entire preparation period and tended
to use mornings to practice the pieces perceived as being more difficult. Total
preparation time for each of the pieces could be predicted on the basis of the
pianist's subjective ratings of complexity and independent ratings of complexity
given by other experts. An analysis of the performance data showed that, near the
time of the recital, variability in performance tempo was large between pieces but
very small for multiple renditions of the same piece, even under the different
experimental conditions. Thus, to attain a highly reproducible public performance,
the expert allocated practice time in response to task demands and engaged in
specific preparations that would safeguard the performance against unexpected
problems.
Although the main task of musicians trained in the performance tradition of Western
classical music is to play extensively rehearsed music in front of an audience. the process
of acquiring such music has received little attention in the research literature. As we
document with this study. the usually covert extensive process of rehearsing music can be
empirically investigated under certain favorable real-life conditions. When studying
expert performers. linking performance to preparation is central to understanding the
The preparation for a public performance of a set of new solo pieces is an extended
process that may last many months or even years. Figure 1 illustrates this process which
is-at all levelsof proficiency-an integral part of learning to playa musical instrument.
At the .onset of learning a new piece. the student already has some previous knowledge
that will allow mastery of the piece. Then. each new piece is taken from its initially
unrehearsed performance to the final stages of a polished performance (see shaded oval
in Figure 0, providing an increase in general performance skills along with the mastery
of the individual piece. Overall, general performance increases faster in the beginning
and slows (asymptotes) at higher levels of performance. The performance preparation of
an expert is therefore likely to be distinct from skill acquisition, because the technical
challenges that foster large improvements in general skills have been mastered prior to
preparation of a specific piece of music.
There are many factors that influence the amount of necessary preparation time
needed to attain the polished level of performance of a public solo performance. The
most important factor concerns the relation between the difficulty level of the piece
(considering its length and recommended tempo) and the skill level of the performer.
There are subjective estimates of how students and teachers perceive the difficulty of a
piece. However. music experts have tried to develop more objective consensual ratings of
complexity for a given set of pieces. The result of such endeavors for the piano are
educational plans or syllabi such as those of the Associated Board Exam in the United
Kingdom. the syllabus of the Toronto conservatory (Piano Syllabus. 1986), the P-F-
Guide (Pierce & Fuszek, 1987). or annotated listings of piano literature (e.g.• Wolters &
lilI
Downloaded from msx.sagepub.com at CMU Libraries - library.cmich.edu on September 11, 2015
Relatlns practiceand performance
ANDREAS c. LEHMANN at K. ANDERS ERICSSON
unrehearsed polished
performance performance
Quality
General
performance Amount of ,~
preparation, , ,
skills ,,
,
,,
Duration of training
Figure 1. Overview of learning trajectories for individual pieces in the context of instrumental
music skill acquisition with points of interest for studies on pradice and performance (shaded
columns)
Goebels, 1967). Our analyses indicate high and significant correlations between such
lisrings.! Surprisingly, there are no empirical data validating the criteria used for rating
pieces according to difficulty, nor for the real-life validity of these ratings. As it turned
out, complexity ratings and preparation time for a given piece were significantly related
in this study.
An hour-long representative recital program can require several hundred hours of
preparation time. This fact poses a challenge to laboratory studies of this process. As a
result (see shaded columns in Figure I), laboratory studies have focused on either the
early acquisition stages or memorization (e.g., Rubin-Rabson, 1940; 1941; Gruson,
1988; Miklaszewski, 1989; 1995) or the analysis of the final performance (e.g., Seashore,
1938/1967; Shaffer, 1981; Repp, 1990; Clynes & Walker, 1986). To reduce the necessary
time to memorize or initially master the pieces, researchers have selected short pieces (or
brief excerpts of regular pieces) at relatively low difficulty levels (e.g., Rubin-Rabson,
1940; Krampe, 1994). The difficulty level of the music used in laboratory studies is often
well below the musicians' maximal level, especially for studies involving experts. In the
following review of research on the acquisition of new music and the structure of
(l) An analysis of works mentioned both in the Piano Syllabus (l986) and Pierce and Fuszc:k's (l987) guide
revealed a significant correlation of r(44) = .84. P < .001. Relating independent ratings of difficulty for
Beethoven's 32 piano sonatas listed in Pierce and Fuszek (l987) and Wolten and Goebels (l967) also results
in a significant correlation of r(30) = .83. p < .00 1.
1i9
(2) Miklaszcwski (1989) reports rhat up to six pages were already memorized. implying that the subject was
in the final initial stages ofleaming the piece. The goal of the 4 practice sessions. totaling around 220 minutes,
was for the pianist to be able to "perform the composition from beginning to end without interruptions. and
with sketchy interpretation, but not necessarilyfrom memory" (p, 98).
70
Downloaded from msx.sagepub.com at CMU Libraries - library.cmich.edu on September 11, 2015
RelaUng practice and performance
ANDREAS C. LEHMANN & K ANDERS ERICSSON
found that the expert pianist identified the difficult sections and devoted more time to
these sections than to the rest of the piece. Miklaszewski (1995) also investigated the
preparation of four short pieces to performance level by three expert pianists. 3 These
results essentially replicated his earlier findings and. in addition, indicated that the length
of sections practiced increased over time. Chaffin and Imreh (1996) also found that their
subject made central decisions regarding final performance very early on in the
preparation process. Supporting evidence for experts' ability to anticipate possible
performance problems also was provided by a study with an organ player (Nielsen,
1997).
Laboratory studies and studies involving short pieces and practice times suggest that
increase in complexity is related to increase in practice time. Initial evidence exists
showing that experts may possess greater insight into possible future performance
problems. This study replicated and extended some of these results in the context of the
naturalistic preparation of an entire recital program. In contrast to other studies, we also
assessed the quality and adequacy of the final performance.
One of the primary challenges to the study of expert music performance is that the
polished performance of the same piece differs systematically among expert musicians as
a function of their intended interpretation of the piece (e.g.• Seashore. 1938/1967;
Gabrielsson, 1987; Repp, 1990). If the performer rehearses a single preferred
interpretation. the performance by a given musician should be very consistent when
observed on multiple occasions (at least in the laboratory). The most studied aspect of
repeated performances of expert musicians concern their tempo and expressive timing.
Large individual differences in the performance tempo (as high as 50%) have been
observed for recordings of the same piece by different professional pianists (Palmer. 1968;
Repp, 1990). In contrast. when pianists are asked to reproduce their own performances,
tempo is highly consistent across renditions (e.g.• Seashore. 1938/1967; Shaffer. 1981.
p. 358). Reduced ability to reproduce the tempo across performances by expert musicians
is usually attributed to inadequate rehearsal and preparation of the piece (Clynes &
Walker. 1986; Shaffer & Todd. 1987). Similar findings of high consistency of tempo for
different performances have been found for a string quartet (Clynes & Walker. 1986).
More detailed analysis of repeated music performance at the level of individual key
strokes has shown the high reliability of relative and absolute timing. Shaffer (1980)
found variation coefficients of 1 to 4% for beat length and bar length on repeated
performances. The remarkable reproducibility of the tempo for a given piece cannot be
explained by memory for the earlier performance, because Repp (1995) found a similar
(3) A1rhough no complexiry rarings are available for those pieces, they seem to have been easier rhan the piece
in rhe 1989 study.
71
METIlOD
• Subject. A pianist (GM) was preparing a program with eight pieces of music for her
degree recital (masters in piano pedagogy). keeping a detailed daily diary of her practice
times throughout the preparation phase. At the onset of this study. GM had already kept
a diary for six months and played her repertoire almost note-perfectly from memory.
Informed consent was obtained and the subject was paid for her participation in the
recording and interview sessions. Our subject was 25 years old at the time of the study.
She had started to play the piano at age 10 and had continuous instruction for 15 years.
• Procedure. Three interviews were conducted beginning six weeks prior to the recital;
the final one occurred two days before the public recital. The interviews covered
biographical information, estimates of difficulty of the repertoire. and retrospective
estimates of practice.
1. Interview. The subject gave biographical information regarding her musical
development. reported on her practice habits and practice times, and estimated amounts
of prior practice for the pieces in her program.
2. Interuieto. Information were requested about the difficulty of the recital music. and
new estimates of past practice for the recital program. During this interview. GM ranked
all pieces according to how difficult they were to learn.
3. Interuieu; The last interview occurred immediately after the second recording
session (audience condition. see below) and focused on GM's retrospective reports her
performance. The result section of this paper does not include a qualitative analysis of
interview data; however. the interesting details about the performance preparation that
GM provided in the interviews are shared in the discussion section to support our
statistical findings. On three occasions. the entire recital program was recorded.
Recordings. The three recording sessions used a Yamaha Disklavier interfaced with a
computer. At each session. GM played through the entire program twice with each piece
being separated from its repetition by all other pieces. The pieces were always performed
in the order planned for the recital. During the first and second recording sessions. only
the experimenter and an assistant were present. We will refer to these private
performances as the practicecondition. The first session took place when we met GM six
weeks before the recital; the second occurred a week before the recital. For the third
recording session, which took place three days after the second. we simulated a
performance situation (henceforth audience condition). GM played in front of a new
audience in the first and second rendition. and a video camera was installed to increase
realism. When not otherwise stated. only the last two recording sessions were analyzed.
• Analysis of the data. In her daily diaries. GM listed the date. starting and ending times
of her practice separately for each of the eight pieces. and occasionally added metronome
markings and personal observations and evaluations. Typical entries after the date read as
follows: "9 - 9:35 Haydn 2" or "10:55 - 11:05 Break" or "11:10 - 11:55 Haydn 2. 6
rhythms." Dates. starting times. ending times. and piece information were entered into
a statistics program. Next. the computed practice durations were log-transformed prior
to further statistical analysis as recommended in earlier studies of skill acquisition and
expert performance (Newell & Rosenbloom. 1981; Ericsson et al., 1993).
Ratings ofcomplexity. Independent ratings of difficulty were obtained for the Haydn
and Debussy pieces by consulting a list of graded repertoire (Wolters & Goebels, 1967).
No ratings were available for the Prokofiev pieces. and two expert pianists were asked to
generate ratings based on the score for the Prokofiev pieces. the rating criteria (Wolters
& Goebels, 1967. p. 13). and the other graded pieces as a reference. The experts' ratings
were very similar and therefore averaged for use in the analyses. The final difficulty
ratings on a 15 point scale were as follows: Hayl: 11.5; Hay2: 11.5; Pro1: 7; Pro2: 9.5;
Pro3: 10.25; Deb l: 12.5; Deb2: 11.5; Deb3: 13.5 (x = 10.9; SO = 2).
Durations and tempo. Performance durations were measured for each piece for four
renditions (two in the second practice condition and two in the audience condition).
First. each piece was divided into a number of musically meaningful segments. Next. we
corrected for a small number oferrors and omissions (e.g.• omitted repeats) made by GM
7J
The result section is organized imo..ftve parts. First we show that GM's level of
accumulated practice is representative for music students at her school of music. Then
we describe GM's allocation of practice time on the repertoire pieces over the entire
period of preparation. In the third section we analyze the relation between difficulty
ratings of pieces and practice times. The fourth section assesses the quality of GM's final
performance, and the last part reports results from analyses regarding consistency of
starting tempo and average tempo.
(4) A parallel analysis to all the foUowing ones was done without deleting the first six lOis. The pattern of
results remained the same.
(5) Conducting the analyses on the raw data instead of the log-transformed data yielded similar patterns of
results, Unless otherwise indicated. the following sections only report the analyses done on log-transformed
data.
74
Downloaded from msx.sagepub.com at CMU Libraries - library.cmich.edu on September 11, 2015
Relating practice and performance
ANDREAS c. LEHMANN ll< K. ANDERS ERICSSON
=104
~:S
CJ-
4000
CJ~
<~0 2000
0
0 2 4 6 8 10 12 14
Years of training
Figure 2. GM's amount of deliberate practice accumulated during her years of training compared
to a sample from a previous study (Lehmann & Ericsson, 1993).
Note. Error bars indicate standard deviation.
75
6000
j
3a
5000
--- Hayl
---0--
.e ~
Hay2
j 4000
~
:::l
e=- 3000
I
2000
1000
-e
0
0 50 100 150 200 250
Diary day
6000
3b
J1:1 5000 Prol
--- Pro2
---0---
'i Pro3
.S 4000
~
j
~ 3000
:a
u
f
=-2000
J
=" 1000
~
u
u
-e 0
0 50 100 ISO 200 250
Diaryday
71
Downloaded from msx.sagepub.com at CMU Libraries - library.cmich.edu on September 11, 2015
Relatlns practice and performance
ANDREAS c. LEHMANN l!< K. ANDERS ERICSSON
6000
3c
i 5000 ---0--
Debl
'i Deb2
-~=
CU 4000
----.-- Deb3
CU
CJ
;:I 3000
CJ
f
Co
2000
]
as
:I
§ 1000
-e 0
0 50 100 150 200 250
Diary day
Figure 3. a-c. Accumulated practice times in minutes for all eight pieces, grouped by composers
(Fig. 3a: Haydn; Fig. 3b: Prokofiev; Fig. 3c: Debussy), showing relatively stable increasesacross
the entire diary period except for Deb3, Pro3, and Hay2.
Note. No practice times were available between day number 130 and 147, when GM was on
winter break.
For many of the pieces, GM appeared to allocate approximately the same amount of
practice every week during the preparation period. To identify possible deviations from
this pattern, the amount of recorded practice for each week was correlated with the
consecutively numbered index for the diary week. In the case of similar amounts of
weekly practice throughout the diary period, no significant correlations would be
expected. The correlations for three pieces were significant, indicating for those pieces an
increase in practice time over the preparation period (Deb3, r[29] = .79, P < .001; Pro3,
r[29] = .65, p < .001; and Hay2, r[29] = .49, p < .Ol). Follow-up r-tests comparing the
average weekly practice time before and after the winter break were only significant for
Pro3 and Deb3, with higher means after the break, t(28) = 5.38, P < .001 and t(28) =
4.14, P < .00 I, respectively. This finding suggests that GM selectively increased her
practice on these two specific pieces rather than to increase it overall.
How did GM distribute her practice time across weeks and days?The average length
of GM's individual practice sessions, of which there were often more than one each day,
was 1.82 hours (SD = 1.43). An ANaVA revealed no significant difference in the length
of practice sessions for the different days of the week, suggesting a comparable length and
variability of individual sessions across the week. As a consequence, the average number
of sessions for each day of the week correlated highly with the average amount of practice
for that day, r(5) = .96, p < .001. Average practice times for all seven days across the diary
77
3.5
3.0 a
2.5
0.5
0.0
Mon Tue Wed Thu Fri Sat Sun
Days of the week
Figure 4. Number of practice sessions averaged across the entire diary period for each day of the
week.
Note. Error bars show standard deviations; (a) Thursday differs reliably from the rest of the week;
(b)Sunday differs reliably from the rest of the week.
71
Downloaded from msx.sagepub.com at CMU Libraries - library.cmich.edu on September 11, 2015
Relating practice and performance
ANDREAS c. LEHMANN ll< K. ANDERS ERICSSON
on total practice time (R2 = .85, F(2,5) = 20.87, P < .01), both complexity ratings
contributed unique variance (p < .05). This suggested, that each type of complexity
rating explained a different part of the variance in practice time.
The amount of information to be learned is one primary factor that theoretically
influences the total amount of preparation time, and the next analysis controlled for this
variable. To arrive at an independent measure of the information contained in each piece,
we averaged the number of keystrokes performed for each piece across all four renditions.
(Admittedly. this measure contains some redundancy since pieces may contain repeats
and repetitive patterns.) A hierarchical regression analysis was performed using practice
time as the dependent variable and complexity ratings and number of keystrokes as
independent variables. First, the number of keystrokes was forced into the equation to
control for the amount of information contained in the pieces (adj. R2 = .56, Fine[l, 6]
= 9.82, P < .05). Then, both difficulty ratings were entered such that the one with the
highest partial correlation would enter first. GM's subjective complexity rating entered
the equation first, f1R2 = .29, F(1,6) = 16.58, P < .01. The experts' ratings did not enter
after that, most likely due to the high correlation between number of keystrokes and
experts' ratings, r(6) = .87, P < .01. Together, the amount of information and GM's own
ratings of difficulty accounted for 88% of adjusted variance in total practice times (F[2,
5] = 25.95. P < .OI), leaving little room for additional explanation.
Based on their review, Ericsson et al. (1993) have suggested that mornings are
especially effective for improvements of performance. To evaluate this claim, the two
difficulty ratings were related to the number of sessions starting before (AM) or after
noon (PM) for each piece. As can be seen in Table 1, GM's own ratings predicted practice
before noon, while afternoon practice was reliably correlated with the experts' ratings.
Hence. GM preferred to practice those pieces in the morning that she considered to be
the most difficult and challenging (see Discussion section for explanation of correlation
between PM practice and expert ratings).
79
10
Downloaded from msx.sagepub.com at CMU Libraries - library.cmich.edu on September 11, 2015
Relating pradlce and performance
ANDREAS c. LEHMANN II< K. ANDERS ERICSSON
Performance tempo was increased throughout the diary period, and alrhough there
were not enough data available from the diary to trace the increase for every piece.
plotting the available points revealed an increase in mastered tempo over time (see Figure
5). Not all pieces showed the same dramatic increase in tempo; for example. the initial
tempo for Hay1 approached its final tempo. The final tempo obtained under the practice
condition in the laboratory was indicated to the right of the diary data in Figure 5. The
spontaneously initiated tempo was very dose to the practiced tempo, differing from it
between 0.4 and 4 beats per minute. This validates the diary entries.
140 140
iB 120
-
-
--0-
-
DolI1
!'nlII2
Dob3
Hay1
•
120
~
I
J;~ !.~
j.~ 100 100
1·1
•
-ell
Il c
~
11:1 80 80
~! • 11
! 60
40
60
40
~
S
rI.l
llII
0 5 10 15 20 Z5 30 35 MIDI
DIary week recording
Figure 5. Diary data indicating increases in tempo over time for four pieces. The observed starting
tempo from the second recording under pradice conditions is plotted at the right hand side of the
graph for reference. It was calculated using lOIs.
Note. Enough data points were available for only four of the eight pieces.
8'
Downloaded from msx.sagepub.com at CMU Libraries - library.cmich.edu on September 11, 2015
To assessthe detailed mechanism that allowed GM to achieve consistent performance
durations, her ability to reproduce the starting tempo was analyzed. Our first analysis
assessed the effect of individual piece, condition, and rendition on the starting tempo
(quarrernote lOis). A MANOVA was conducted using lOIs (l through 18) as the
dependent variables, and the between-subjects factors were condition (audience
condition, practice condition), rendition (first, second rendition), and piece (the eight
pieces performed). Not surprisingly, a significant main effect of piece was found and an
effect of condition, F(7,119) = 144.84, P < .0001 and F(l, 119) = 7.73, P < .025,
respectively. The effect of condition indicated a reliably faster performance under the
audience condition (x = -.311, SO = .464) than under the practice condition (x = -.327,
SO = .430). This finding is consistent with the one mentioned above concerning the
duration of the pieces. Figure 6 shows the corresponding means for each piece for each
condition. The differences between conditions are indeed small, with the possible
exception of piece 6 (Deb I: Pagodes). In fact, a series of post-hoc tests corrected for
multiple comparisons revealed significant differences in starting tempi only for piece 6,
which had reliably shorter quarternote lOis (i.e., faster starting tempo) in the audience
condition compared to the practice condition (see Figure 6). This was already implied by
a marginally significant interaction of piece and condition, F(7, 119) = 2.44, P < .059.
This is an interesting finding that cannot be due to an order effect, because the pieces
were always played in the same sequence. For the other seven pieces, the differences
between conditions were not significant.
II 0.0
J~ -0.1
je
9 -0.4
1 -0.6 Conditions
- - Audience
.. -0.8 - - 0 - - Practice
i
roli" -1.0
Bayt Bay2 Prokll"rala I'k'ukJ Debt Dcb2 DcbJ
Pieces
Figure 6. Means illustrating the significant effect of piece and condition. Error bars indicate
standard deviations.
A similar analysis was conducted to compare the recordings under practice
conditions, the first ofwhich was made around the time of the first interview; the second
1:1
dated from shortly before the recital. Again, the between-subjects factors were recording
session (first, second), adjacent rendition (first, second rendition), and piece (8). There
were no reliable differences in starting tempo across time, suggesting that the tempo had
already stabilized when GM was recorded for the first time.
The next analysis tried to explain differences in performance under the two
conditions of practice and performance. This analysis predicted average tempo from
starting tempo, piece, condition, and rendition in order to establish how much variability
in overall duration remained unexplained after controlling for the measures we had
collected. Using average tempo rather than duration controlled for the length of the
pieces.
Both tempo measures, log starting tempo and log average tempo for each of the 32
performances (8 pieces x 2 conditions x 2 renditions), were included in a regression
analysis. The dependent variable was average tempo, which served as an index for the
performance duration of a piece. In the first step of our hierarchical regression analysis
(see Table 2), we entered the starting tempo; in the second step, we forced our piece
variable in the equation; in the third step, we entered the condition and rendition
variables. Taken together, starting tempo, piece, and condition accounted for virtually all
of the variance in average tempo, R2 = .99. All three blocks showed significant regression
weights and the final F was significant (F[lO, 21] = 11392.04, P < .OOOI).
Table 2
Step 1
Starring tempo .685 .089 .599 59.336 .0001
Step 2
Piece 2815.76 .0001
(dummy coded)
Step 3
Condition .006 .001 .013 15.203 .001
(practice. audience)
Rendition .001 .001 .002 .484 .494
(firsr. second)
Note. R2; .821 for Step 1; ~R2 ; .173 for Step 2 (p < .001); ~R2 ; 00013 for Step 3 (p < .01).
Final F statistics: F(l0.21) = 11392. P < .0001; adj. R2 ; .99.
Thus, after controlling for starting tempo, systematic variability associated with a given
piece, and the condition under which it is performed, there remained almost no variance
8J
0.25
•
§ 0.20
.=
=0.15
~
CU
•
•
"d
'E=0.10
] 0.05 •
rI..l
This case study monitored the nine month preparation of a recital program by an expert
musician (advanced student performer) and also analyzed the attained music
performance observed in a situation simulating public performance. Case studies have
previously led the way for research on music performance (Shaffer, 1981) and analysis of
the detailed structure of practice (Miklaszewski, 1989; Chaffin & Irnreh, 1996). Unlike
experts in the domain of sports, it is unusual for musicians to keep detailed records of
their practice and improvements in performance. Therefore. GM offered the unique
opportunity to investigate a usually undocumented process in more detail. The fact that
GM kept her diary voluntarily adds to the external validity of our findings. Finally, her
84
program of eight pieces was sufficiently varied in terms of complexity and tempo to allow
meaningful statistical analyses.
The data analysis showed that
1. Accumulated practice times for a specific piece could be predicted largely by subjective
and objective complexity ratings, and complexity ratings were highly correlated with the
time of day when a piece was practiced.
2. A listener evaluation by experts of our pianist's quality of final performance found it
to be comparable to that of other established expert pianists.
3. The pianist showed a remarkable consistency of tempo across repeated renditions
(non-adjacent) of the same piece on a given occasion as well as under different
performance conditions. Only one out of eight pieces was played reliably faster under
performance conditions.
4. Starting tempo, piece played, and performance condition explained almost all of the
variability in average performance tempo, which was used as an indicator of duration.
5. There is a tendency for faster pieces to show less variability at the quarrernote inter-
onset level than slower pieces.
In our discussion we will review the results of this case study in light of general
principles proposed in prior research on expert performance and point out how future
research on the relation between practice and attained performance in public recitals can
be extended. We will first discuss GM's practice and then the analyses of the consistency
of her performance.
• The structure and optimization of practice during the preparation of a solo recital
GM's history of practice during the development of her piano performance was
representative of other students at the School of Music at FSU which we have
investigated (Lehmann & Ericsson, 1996). Also, her level of weekly practice was
comparable to that of other expert musicians from a previous study (Ericsson et al.,
1993). Therefore her practice diary offered a unique data set to evaluate the claims of
consistency of practice that previously have only been based on interviews and one-week-
long diaries.
The most striking result is the consistency of GM's weekly amount of practice across
35 weeks of preparation. GM was found to practice, on the average, the same amount
every day of the week, except for Thursdays and Sundays. Her practice was irregular on
Sundays and was most extensive on Thursday, the day following her weekly lesson. One
might speculate that the lesson provided her with new ideas that she did not want to
forget or new goals that had to be met. Practice on the individual pieces tended to be
initiated more frequently in the morning compared to the afternoons, yet this finding
was not statistically significant. The pattern of results regarding GM's practice habits is
consistent with results by Ericsson et al. (1993) for academy students preparing to
become music teachers (see Sloboda et al., 1996, for younger music students). Future
85
Ii
her stable practice habits, GM seems to have responded to encountered task demands
and perceived difficulties by differentially allocating effort and time.
87
II
keystrokes were included in the tempo analysis.Thus, the correct tempo was not adjusted
very much as the piece unfolded. but was produced at the onset. or even before. This is
conceptually important, because one might have hypothesized that the performer starts
out with an approximate tempo and adjusts the tempo to match an internal
representation only after hearing herself. As casual observations reveal. this does not seem
to be the case with our expen or expert musicians in general. However, it happens to
novices who start out at some tempo and then adjust. Often, novices are anxious during
public performances and stan out too quickly. Incidentally, a somewhat faster speed was
also noticeable for our subject. which provides evidence that even experts have to cope
with and compensate for higher arousal levels.
Are consistent tempi related to practice times? In our case. there was no reliable
correlation between the standard deviations of the performance durations or the lOIs
and the practice times extracted from the diary. But we cannot rule out that there may
have been a relation during earlier phases of the acquisition. At some point in the
acquisition process. tempo may become part of the performance plan and a stable
parameter. As Shaffer and Todd (l987) succinctly put it: "... an expressive form can have
a precise mental representation and can be precisely executed" (p, 142). The authors
argued for a stable and accurate timekeeper, and Shaffer (l981) has demonstrated
experts' superior skills with regard to the timing of motor programs.
The present paper has extended the groundbreaking case studies by Shaffer (l98l)
on temporal aspects of rehearsed music performance by describing its lengthy
preparation in real-life and examining its relation to the consistency of performance
under experimentally varied conditions. Contrary to most other studies that have looked
at performance of an individual piece. this study has investigated a collection of pieces.
It also complemented and validated existing work on practice by relating performance
under representative conditions to practice as it occurred in a largely unconstrained real-
life situation. Our subject knew what preparatory activities to engage in and for how
long in order to generate a high quality performance. The results are in general agreement
with research in other domains of expertise that show that experts are maximally adapted
to the representative conditions under which they perform (Ericsson & Charness, 1994;
Ericsson & Lehmann. 1996). I
(I) We wish to thank GM, whose artistic efforts are the topic of this paper, for her patient and friendly
cooperation. The Center for Music Research at Florida State University generously offered use of their facilities.
Ben Pringle extracted the many lOIs. Helpful and critical comments on previous drafts were provided by
Bruno Repp, John Sloboda, Robert Woody, and two anonymous reviewers. This study was in part presented
in June 1995 at the SMPC95 in Berkeley, CA.
Address for correspondence:
Andreas C. Lehmann,
Department of Psychology;
Florida State University,
Tallahassee, FL 32306-1270;
phone: (850) 644-9850; fax: (850) 644·6100;
email: Lehmann@psy.fsu.edu
I'
Downloaded from msx.sagepub.com at CMU Libraries - library.cmich.edu on September 11, 2015
• REFERENCES
Chaffin, R., & Imreh, G. (l996). Effects of difficulty on expert practice: A case study of
a concert pianist. Poster presented at the 4th International Confnmct: onMusic
Perception and Cognition, McGill University, Montreal, Canada, August 11-
15.
Clynes, M., & Walker, J. (1986). Music as time's measure. Music Perception, 4 (l), 85-
120.
Ericsson, K. A. (l997). Deliberate practice and the acquisition of expert performance:
An overview. In H. jergensen & A. Lehmann (Eds.) , Dots practice makt
perftct? Currmt thtory anti rmarch on instrumental music practice (pp, 9-52).
Oslo, Norway: Norges musikkhegskole,
Ericsson, K. A., & Charness, N. (1994). Expert performance. Its structure and
acquisition. American Psychologist, 49, 725-747.
Ericsson, K. A., & Lehmann, A. C. (1996). Expert and exceptional performance:
Evidence of maximal adaptation to task constraints. Annual Review of
Psychology, 47, 273-305.
Ericsson, K. A., Smith, J. (1991). Prospects and limits in the empirical study ofexpertise:
An introduction. In K. A. Ericsson & J. Smith (Eds.), Toward a gmeralthtory
ofexpertist: Prospects and limits (pp, 1-38). Cambridge: Cambridge University
Press.
Ericsson, K. A., Krampe, R. T., & Tesch-Romer, C. (1993). The role of deliberate
practice in the acquisition of expert performance. Psychological Review; 100,
363-406.
Gabrielsson, A. (l987). Once again: The theme from Mozart's piano sonata in a major
(K. 331). In A. Gabrielsson (Ed.), Action andperception in rhythm anti music,
(pp, 81-103). Stockholm, Sweden: Royal Swedish Academy of Music.
Gruson, L. M. (1988). Rehearsal skill and musical competence: does practice make
perfect? In J. A. Sloboda (Ed.), Gmerativt proctsstS in music (pp. 91-112).
Oxford, England: Clarendon.
Harnischrnacher, C. (l993). Instrumentales Obm und Asptktt der Personlichktit. Eine
Grundlagtnstudit zur Erfassung physischer und psychischtr Abwtichungm durch
InstrummtalspitL Frankfurt, Germany: Lang.
Jones, A. R. (l990). The role of analytical prestudy in the memorization and retention
of piano music with subjects of varied aural/kinesthetic ability. Dissertation
Abstracts Intnnational, 51(7), 2307A (UMI no. 9026216).
Kopiez, R. (1996). Aspekte der Performanceforschung. In H. Mette-Haber (Ed.),
Handbuch der Musikpsychologit (pp, 505 - 587). Laaber, Germany: Laaber.
!to
Downloaded from msx.sagepub.com at CMU Libraries - library.cmich.edu on September 11, 2015
Rel.tlns practice and performance
ANDREAS c. LEHMANN I!c K. ANDERS ERICSSON
'1
Downloaded from msx.sagepub.com at CMU Libraries - library.cmich.edu on September 11, 2015
Seashore. C. E. (1938/1967). Tb« psychology of music. New York: Dover. Originally
published 1938.
Shaffer. L. H. (1980). Analysing piano performance: A study of concert pianists. In G.
E. Stelmach & J. Requin (Eds.), Tutorials in motor b~havior, (pp, 443-455).
Amsterdam: North Holland.
Shaffer. L. H. (1981). Performances of Chopin. Bach. and Bartok: Studies in motor
programming. Cognitiv~ Psychology. 13.326-376.
Shaffer. L. H.. & Todd. N. P. (1987). The interpretive component in musical
performance. In A Gabrielsson (Ed.), Action and perception in rhythm and
music(pp. 137-152). Stockholm. Sweden: Royal Swedish Academy of Music.
Sloboda. J. A (1994). Music performance: Expression and the development of musical
excellence. In R. Aiello & J. A Sloboda (Eds.), Musical perc1Jtions (pp, 152-
172). New York: Oxford University Press.
Sloboda. J. A. Davidson. J. w..
Howe. M. J. A. & Moore, D. G. (1996). The role of
practice in the development of performing musicians. British journal of
Psychology, 87. 287-309.
The Pianist [Computer software]. (1993). Buffalo. NY: PG Music.. Inc.
Wagner, C. (1974). Experimmtel/e Untersuchungm uber das Tempo. Osterreichische
Musikzeitschrift, 29(12). 589-604.
Weissenberg, A (I 985). Estamp~s. On Piano Music. &/~ctions [compact disk]. Hamburg.
Germany: Deutsche Grammophon.
Wing, AM .• & Kristofferson, A B. (1973). The timing of interresponse intervals.
Perception and Psychophysics, 13.455-460.
Wolters, K.. & Goebels, F. (1967). Handbuch tier K/avierliteratur. Zurich. Switzerland:
Atlantis.
• Preparazlone dl un recital:
la relazlone tra pratlca ed esecuzlone
Sono stati studiati i nove mesi che un pianista esperto ha dedicato alia preparazione
di un recital, basandosi sui diario da lui scritto relativo alia propria pratica, e sulla
registrazione (M.I.D.I.) di otto pezzi da lui messi in programma, registrati in diverse
condizioni sperimentali -individuali 0 pubbliche. Dal diario risulta che I'interprete -
uno studente di grade avanzato- si e dedicato a tempi di pratica regolari e che i
pezzi giudicati piu difficili tendevano ad essere studiati di mattina. II tempo di
preparazione dedicato ad ogni pezzo e dipeso dall'apprezzamento soggettivo della
sua complesslta: il pianista, cioe, non ha tenuto assolutamente in considerazione i
giudizi di complessita emessi da altri interpreti. L'analisi dei dati registrati dimostra
che, in prosslmita dell'esecuzione pubblica e indipendentemente dalla condizione
sperimentale, Ie variazioni di tempo sono maggiori da un pezzo all'altro e minori tra
Ie diverse interpretazioni delle stesso brano. Se ne deduce che per arrivare a
un'esecuzione pubblica perfettamente riproducibile, I'interprete esperto si dedica a
tempi di pratica scelti in funzione del compito richlesto, assicurandosi una
preparazione che salvaguardi I'esecuzione da ogni dlfftcolta inaspettata.
On etudle ici les neuf mois de preparation qu'un pianiste experirnente a consacres
a la mise au point d'un recital, en se fondant sur Ie journal qu'i1 a tenu de sa
pratique et sur I'enregistrement (M.I.D.I.), sous diverses conditions experirnentales
• travail solitaire et execution publique -, des huit pieces qu'il avait rnlses a son
programme. II ressort du journal que l'lnterprete en question - un etudiant avance
- s'est astreint a un temps de pratique regulier tout au long de la periode
preparatoire et que les pieces jugees subjectivement les plus difficiles tendaient a
etre travaillees Ie matin. Le temps de preparation alloue a chacune des pieces est
fonction de l'appredatlon subjective de sa complexite: c'est dire qu'i1 ne tient
aucunement compte des jugements de complexite emis par d'autres interpretes
experlmentes, lndependamment de la condition experimentale, I'analyse des
donnees enreglstrees montre que, a I'approche du recital, les variations de tempo
sont importantes d'une piece a I'autre, alors qu'elles sont infimes s'agissant des
multiples interpretations d'une meme piece. II en decoule que, pour parvenir a une
execution publique parfaitement reproductible, l'lnterprete experimente se soumet
a un temps de pratique qui est fonction des exigences de la tache et qu'll securise
en outre son execution par une preparation visant a pallier toute dlfflculte
inattendue.
'4
Downloaded from msx.sagepub.com at CMU Libraries - library.cmich.edu on September 11, 2015