Perception and Production of StaccatoArticulation on the Piano
Bruno H. Repp
Haskins Laboratories
Unpublished manuscript (completed October 1998)
Bruno H. ReppHaskins Laboratories270 Crown StreetNew Haven, CT 06511-6695
Tel. (203) 865-6163, ext. 236FAX (203) 865-8963e-mail: [email protected]
Repp: Staccato articulation Page 2
Abstract
Skilled pianists were requested to interactively adjust note (i.e., key
depression) durations in simple sequences, played under computer control on a
digital piano, so as to sound optimally staccato (perception task) and to perform
analogous sequences with staccato articulation on the same instrument as well as on
a real piano (production task). The independent variables were tempo, register, and
pitch step size; also, in production, hand and sequential position. Note durations
increased significantly as tempo decreased in both perception and production, but
there were large individual differences in the magnitude of these effects. Note
durations were longer in production than in perception, and longer on the digital
than on the real piano. In perception but not in production, low notes were assigned
shorter durations than high notes. There were complex effects of sequential note
position, in part due to metrical structure. The results illustrate some of the factors
that influence staccato articulation, and they also reveal some discrepancies between
perception and production of staccato. In particular, it seems that produced staccato
is not necessarily perceptually staccato.
Repp: Staccato articulation Page 3
INTRODUCTION
Most musical instruments give the player control over the physical durations
of tones, be it by releasing keys, lifting the bow from strings, stopping airflow, or
damping vibrations with the hand. On these instruments, therefore, there are two
ways of articulating successive tones: When one tone is terminated near the
beginning of a following tone, the tones are perceived as connected or legato; but
when the first tone is terminated earlier, so that a noticeable gap occurs between the
tones, they are perceived as separated or non-legato.
The perceptual and kinematic distinction between legato and non-legato on
the piano has been the subject of several recent studies (Kuwano, Namba, Yamasaki,
& Nishiyama, 1994; Repp, 1995, 1997). On the piano it is possible to release the key
for one tone after the key for a following tone has been depressed, and optimal
legato is indeed characterized by a small overlap in key depressions (the key overlap
time or KOT). The acoustic overlap of the tones is even more substantial, due to the
post-release (damped) decay of piano tones, which may extend over several
hundreds of milliseconds (Repp, 1995, 1997).
Repp investigated legato articulation in simple ascending and descending
scales and arpeggi, and he hypothesized that the KOT would depend on certain
acoustic properties of piano tones. For example, high tones decay faster than low
tones while they are sustained (Martin, 1947; Repp, 1995, 1997), so that—given equal
durations of key depressions—they have a relatively lower amplitude at the time of
key release and hence a shorter post-release decay time than low tones. Also, a long
tone obviously decays further than a short tone of the same pitch while the key is
depressed, so that the former has a shorter post-release decay time than the latter.
Thus it was predicted that KOTs for optimal legato would be longer in sequences of
high or long tones than in sequences of low or short tones, on the assumption that a
Repp: Staccato articulation Page 4
constant degree of acoustic (or rather auditory) overlap would be aimed for. Both
predictions were confirmed in a perception task that required listeners to
interactively adjust the KOTs of a tone sequence until it sounded optimally legato.
When pianists played identical sequences legato on the same instrument, however,
KOTs varied only as a function of tone duration (tempo), possibly for kinematic
reasons unrelated to the decay characteristics of the tones. Thus, perception and
production were not perfectly matched. However, a tendency to assign longer KOTs
to relatively consonant than to dissonant successive tones was observed in both
perception and production. The production data also revealed a tendency of some
pianists to play more legato with the right than with the left hand and a highly
consistent pattern of variation in KOT as a function of position in the tone
sequence.
The present research used the same methods to investigate staccato
articulation, which apparently has not been the subject of any previous quantitative
studies. Staccato is not the same as non-legato, although the distinction vanishes
when the tempo is fast. Staccato articulation requires that successive tones be not
only separated by a gap but also short in duration; a sequence of long tones separated
by gaps does not sound staccato. A basic assumption underlying the present study
was that there is a perceptual category of staccato having the properties just stated. It
may be the case that notes marked as staccato in a score (by dots above the notes) are
sometimes executed in a way that is perceptually not staccato but rather non-legato.
This is especially likely with relatively long note values such as quarter notes or half
notes.
It was also assumed that for a given musical passage there is such a thing as a
typical or optimal staccato, in which the note durations are neither too long nor too
short. Of course, there are contextual, stylistic, expressive, and individual variations
in the degree of staccato in real music, and there is also a special notation for a very
Repp: Staccato articulation Page 5
sharp staccato (wedges above the notes). Within a given context, however, and
particularly in the stylistically and expressively neutral scales and arpeggi used here,
there may well be a typical or optimal staccato, at least for each individual pianist. It
might be asked why optimal staccato would not simply consist of the shortest
possible note durations. In a production task, however, mechanical and kinematic
constraints probably prevent the notes from getting shorter than a certain
minimum. Although such constraints do not operate in a perceptual adjustment
task, so that non-pianist participants might well select the shortest possible tone
duration as the optimal staccato, pianist participants were expected to rely on their
knowledge of typical staccato tone durations in performance and to refrain from
selecting minimal note durations. For this reason, skilled pianists were used as
participants in both perception and production tasks.
For convenience, the term “note duration” is used here to refer to the time
from key depression to key release.1 Two questions asked in the present study were
(1) what the typical duration of staccato notes might be in a neutral context, and (2)
whether that typical duration is constant or varies with tempo. When the tempo is
increased to a point where note inter-onset intervals (IOIs) are not much longer
than note durations, note duration would most likely have to covary with IOI
duration to prevent staccato from changing to non-legato or legato. At slower
tempi, however, where IOIs are substantially longer than the typical duration of
staccato notes, invariance of note duration seemed a theoretical possibility. It was
hypothesized, however, that the muscular relaxation and kinematic reorganization
brought about by a slowing of tempo might affect the criterion for an optimal
staccato, resulting in a relative lengthening of staccato notes at slower tempi in both
production and perception. This would be consistent with the finding in motor
control studies that the key contact time in rhythmic finger tapping increases as the
Repp: Staccato articulation Page 6
tapping rate is decreased (Billon & Semjen, 1995; Piek, Glencross, Barrett, & Love,
1993).
The effects of several additional variables on the duration of staccato notes
were investigated, in analogy to Repp’s previous studies of legato articulation. One
of these variables was pitch height or octave, referred to here as “register”. Since
higher piano tones decay more rapidly than lower tones, they may seem less loud
and/or subjectively shorter, and this may lead pianists to depress the key longer for
them to achieve an equal degree of staccato. (The highest piano keys, however, are
without dampers, so that the duration of the key depression does not affect the
acoustic tone duration.) Another independent variable was the degree of
consonance of the successive tones, defined in terms of their pitch separation (1, 2,
or 3 semitones). Since consonance is relevant primarily to the perceptual tolerance
of acoustic overlap, which is important in legato but minimal in staccato
articulation, this variable was not expected to have much of an effect on staccato
note durations.
In the perceptual adjustment tasks of the earlier legato studies, listeners were
asked to find the minimal and maximal as well as the optimallegato. In the present
study, only optimal staccato adjustments were requested, for rather obvious
reasons. Maximal staccato merely implies extremely short tone durations and
therefore is of little interest, unlike maximal legato, which defines an overlap
detection or overlap tolerance threshold. The boundary between minimal staccato
and non-legato (or rather non-staccato) is not well defined, unlike that between
minimal legato and non-legato, which represents a gap detection threshold. Instead
of varying the listeners’ perceptual criteria, then, it was decided to obtain several
replications of optimal staccato judgments.
In the staccato production task, two additional variables came into play. One
was the hand with which a tone sequence was played. Repp (1997) observed a
Repp: Staccato articulation Page 7
tendency for some pianists to play more legato with the right than with the left
hand. This makes sense because it is usually the right hand that plays the melodies
in which legato is important, and legato is a refined technique that requires much
practice. However, there was no similarly compelling reason to expect hand
differences in staccato playing. The other variable was the position of a tone within
the sequence or, alternatively, the finger with which a tone was played. In legato
playing, very consistent positional effects were observed that seemed to be related to
the melodic contour or metrical structure of the sequence. Here it was investigated
whether similar positional effects exist in staccato articulation.
One further variable of interest was the instrument used. The previous legato
studies compared perception and production, with different groups of participants
in the two tasks, on a digital piano (Repp, 1995) and on a real computer-controlled
piano, a Yamaha Disklavier (Repp, 1997). In the present perceptual task, only the
digital piano was used because the Disklavier was found to be too unreliable at the
very short tone durations required for the staccato adjustments. The production
experiment, however, was carried out on both instruments. Moreover, both tasks
were performed by the same pianists on the digital piano, and most of them played
on the Disklavier as well. This made possible a within-pianist comparison of
staccato perception and production, as well as of playing on two different
instruments. It was considered likely that the mechanical and acoustic properties of
an instrument would affect, respectively, the kinematics of staccato production and
the perceptual criterion for optimal staccato note duration.
Repp: Staccato articulation Page 8
PERCEPTION EXPERIMENT
Method
Participants. The participants were 7 pianists with advanced skills. Three of
them were graduate students of piano performance at the Yale School of Music, one
was an undergraduate studying there for a performance certificate, and three were
undergraduates at Yale University. They were paid for their participation.
Design. The stimuli were continuously ascending and descending 5-tone
sequences. There were 27 sequences resulting from the factorial combination of
three IOI durations, three pitch registers, and three pitch step sizes. The IOIs were
187, 375, and 750 ms. The registers were the second, fourth, and sixth octaves on the
piano. The step sizes were 1, 2, and 3 semitones (st). To keep the average pitch
constant within each register but at the same time make the keys lie comfortably
under the hand for the production study, all sequences were centered on G. The 1-st
(chromatic) sequence thus consisted of the notes F-F#-G-G#-A, the 2-st (whole-tone)
sequence of D#-F-G-A-B, and the 3-st (diminished-seventh arpeggio) sequence of
C#-E-G-A#-C#.
Procedure. A Roland RD-250s digital piano was controlled via a MIDI
interface by a MAX patcher running on a Macintosh Quadra 660AV computer.2 All
tones were played with “Piano 1” sound at a constant MIDI key velocity of 60. The
participant sat in front of the computer, listened to the sequences over Sennheiser
HD540 II earphones, and started and stopped sequences by clicking “buttons” in a
panel displayed on the monitor. Also displayed was a “slider” with a “handle” that
could be dragged with the mouse to vary note duration.3 The slider had a fixed
range from 10 to 160 ms. This range was based on the author’s judgments in pilot
runs, according to which note durations beyond 160 ms seemed definitely
unacceptable as staccato articulations. The slider was unlabeled and was
Repp: Staccato articulation Page 9
automatically reset to the left-most (10 ms) position at the beginning of each trial, so
that the tones in each sequence were extremely short initially. The participant’s task
was to start a sequence (which always began with the lowest tone), then move the
slider to the right and slowly back and forth until the tones in the sequence sounded
optimally staccato, then to stop the sequence and start the next one. The MAX
patcher presented the 27 sequences in random order and stored the participant’s
final slider settings. After some initial practice, each participant completed three
blocks of 27 trials in different random orders. Blocks were separated by short rests.
Analysis. Average note durations and standard deviations across blocks were
analyzed in separate repeated-measure ANOVAs with register (3), step size (3), and
IOI (3) as fixed variables, and pianists (7) as the random variable. ANOVAs on
pianists’ individual data were also conducted, with blocks (3) as the random
variable. Because of the many significance tests and the smaller error variance
within individuals, only effects with p < .01 were considered significant in the
individual analyses.
Results
The results are summarized in Figure 1 in terms of average note durations
and average within-pianist standard deviations, each with standard error bars
representing the variability among pianists. Average adjusted note durations fell
between about 20 and 80 ms, which suggests a stable perceptual criterion that was
not constrained by the upper limit of the MAX slider (160 ms). There were two
significant main effects in the ANOVA. First, there was the predicted increase of
adjusted note duration with increases in pitch register [F(2,12) = 14.1, p < .002]. The
increase occurred mainly between the low register and the two higher registers.
Second, note duration clearly increased with IOI duration [F(2,12) = 18.4, p < .001],
across all three values. The within-pianist standard deviation of the adjustments
Repp: Staccato articulation Page 10
likewise increased with IOI duration [F(2,12) = 13.9, p < .002]. Step size did not have
a significant effect, and there were no significant interactions.
-----------------------------
Insert Figure 1 here
-----------------------------
There were considerable individual differences among the seven pianists
with regard to the preferred absolute note durations and the sizes of the main effects
(see Fig. 3 below). Five pianists showed a significant effect of register and five one of
IOI. Two pianists showed a significant interaction between register and IOI, such that
effects of register were largest at the longest IOI, and effects of IOI were largest in the
high register. A trend in this direction can also be seen in Figure 1, though it was not
significant overall.
In summary, these results indicate that low tones are perceived as less staccato
than high tones having the same key depression duration. This could be due to
their slower pre- and post-release acoustic decay (Repp, 1995), which may make
them seem longer in duration than high tones. The results also suggest that
acoustically identical tones are perceived as less staccato when the tempo is
increased. This effect occurred at IOIs that were 6 to 24 times as long as the preferred
note durations and therefore cannot easily be attributed to a psychoacoustic
interaction between successive tones. More likely, a perceptual interaction between
rate and duration is involved which parallels a similar interaction in motor
kinematics (cf. Billon & Semjen, 1995; Piek et al., 1993).
Repp: Staccato articulation Page 11
PRODUCTION EXPERIMENT: DIGITAL PIANO
Method
Participants. The same pianists as in the perception study participated, during
the same experimental session.4
Procedure. The sequences were the same as in the perception study, but each
was restricted to three ascending-descending cycles and ended with a long note. The
sequences were notated in common meter on separate music sheets, and the three
IOI durations were represented by sixteenth, eighth, and quarter notes, respectively.
The sheets for the middle register were duplicated and marked “left hand” or “right
hand”, so that there were 36 sheets in all. The low-register sequences (notated in bass
clef) were to be played with the left hand only, and the high-register sequences with
the right hand only. The staccato articulation was not indicated in the notation but
was requested verbally. The music sheets were shuffled randomly for each pianist. A
metronome flashing silently at a rate of 80 per minute (IOI = 750 ms, understood to
correspond to quarter-note beats) was in view, and the “Piano 1” sound was audible
over Sennheiser HD540 II earphones. The pianists were instructed to play each
sequence at the prescribed tempo, but not necessarily in strict synchrony with the
metronome. The resulting IOIs were thus approximately the same as in the
perception study (i.e., 187, 375, and 750 ms). The fingering for each cycle was
prescribed to be 1-2-3-4-5-4-3-2 for the right hand and 5-4-3-2-1-2-3-4 for the left hand.
The performances were recorded by a MAX patcher, and the order of the sequences
was recorded by the experimenter. The last step was accidentally omitted for one
pianist, so that it could not be determined later which middle-register sequences had
been played with which hand. That pianist’s middle-register data were averaged
across the two hands and then duplicated to fill all cells of the design for the
ANOVAs. The note durations were calculated from the recorded MIDI data.
Repp: Staccato articulation Page 12
Analysis. An overall repeated-measures ANOVA was conducted on the note
durations averaged across the three cycles within each sequence, with IOI (3), step
size (3), register (2), hand (2), and position (8) as fixed variables, and pianists (7) as
the random variable. Register and hand were treated as 2-level orthogonal variables
in this analysis, even though hand was confounded with register: The register
variable contrasted low-left and mid-right with mid-left and high-right, whereas the
hand variable contrasted low-left and mid-left with mid-right and high-right. The
position variable represented the 8 notes in a cycle, which were played by different
fingers in the two hands. Subsequently, separate ANOVAs were conducted on the
left-hand and right-hand data, and on the mid-register data. Finally, separate
ANOVAs were conducted on each pianist’s individual data, with the same design as
the overall ANOVA but with cycles (3) as the random variable. Effects in the
individual ANOVAs had to be significant at p < .01 to be considered reliable. Even
so, due to the small error variances, numerous effects reached significance at the
individual level, so that they will be mentioned only where they are of special
interest. F values reported below are from the overall ANOVA, unless specified
otherwise.
Results
The effects of register, step size, and IOI are summarized in Figure 2, which
has the same format as Figure 1. However, the standard deviations here represent
the average within-pianist variabilities across the 8 positions in a sequence, after
averaging across the three cycles. (Variabilities across cycles were even smaller.) The
standard deviations did not show any systematic trends and were not analyzed
statistically.-----------------------------Insert Figure 2 here
-----------------------------
Repp: Staccato articulation Page 13
It can be seen right away in Figure 2 that the average produced note durations,
which ranged roughly from 80 to 170 ms, were substantially longer than the
perceptually adjusted ones (Fig. 1). Note duration clearly increased with IOI duration
[F(2,12) = 6.9, p < .02], although the significance level was not very high, due to large
individual differences in the magnitude of the IOI effect (see below). These
individual differences are evident in the large standard errors at the longest IOI
duration. By contrast, standard errors were very small at the shortest IOI duration,
indicating that the pianists played with similar note durations at this fast tempo. As
in the perceptual adjustment data, step size did not have a significant main effect.
Register did have a significant main effect [F(1,6) = 10.1, p < .02], but it was very
small (only 5 ms) and surprisingly indicated longer note durations in the lower
registers, contrary to the perceptual results. There were no significant effects of
register in the separate ANOVAs of the left- and right-hand data. The only
significant interaction among the three variables represented in Figure 2 was that
between IOI and step size [F(4,24) = 3.8, p < .02]. It was due to a somewhat reduced
lengthening of notes in 3-st sequences at the 750-ms IOI duration, compared to 1-
and 2-st sequences at that IOI. It was also significant in the left-hand data [F(4,24) =
3.4, p < .03], but not in the right-hand data.
The large individual differences in the magnitude of the IOI effect deserve
attention; the individual F(2,4) values ranged from 15.8 (p < .02) to 2959. To
determine whether these differences showed any relationship to individual
differences in the effect of IOI on the perceptual adjustments, Figure 3 plots the
results of the two tasks against each other, with the note durations at the three IOIs
being shown separately for each pianist. There was indeed a relationship, though
not a very strong one. Three pianists (M.A., D.G., T.C.) showed large effects of IOI in
both perception and production, and one (C.M.) moderate ones. Two pianists (P.C.,
H.S.), however, showed sizeable IOI effects only in perception, not in production.
Repp: Staccato articulation Page 14
The seventh pianist (P.W.) produced very short durations and showed only very
small effects of IOI in both tasks. Apparently, different pianists used different
kinematic strategies in staccato production.
-----------------------------
Insert Figure 3 here
-----------------------------
There was no significant main effect of hand in the overall ANOVA. The
separate ANOVA on the mid-register data (omitting the pianist for whom no hand
information was available) likewise did not reveal any overall effect of hand.
Moreover, no individual pianist showed a significant hand difference, and hand did
not interact significantly with IOI, register, or step size.
More complex results were obtained for the remaining variable, position in
the sequence. Position had a significant main effect in the overall ANOVA [F(7,42) =
2.9, p < .02] and was also involved in four significant interactions: with step size
[F(14,84) = 3.3, p < .0005], with register and hand [F(7,42) = 7.0, p < .0001], with hand
and step size [F(14,84) = 2.6, p < .005], and with all three of these variables [F(14,84) =
2.8, p < .002]. The position main effect was significant for the left hand alone [F(7,42)
= 3.8, p < .004], but not for the right hand alone. It reached significance in the
separate mid-register ANOVA [F(7,42) = 2.7, p < .03], but its interaction with hand
fell short of significance. Each of these subsidiary ANOVAs also showed significant
interactions involving position. For the left hand alone, position interacted with
register [F(7,42) = 3.4, p < .006], with register and step size [F(14,84) = 2.2, p < .02], and
with register and IOI [F(14,84) = 2.0, p < .04]. For the right hand alone, it interacted
with register [F(7,42) = 6.2, p < .0001] and with step size [F(14,42) = 4.9, p < .0001]. In
the mid register, it interacted with step size [F(14,42) = 2.4, p < .008] and with hand
and step size [F(14,42) = 2.1, p < .03]. In summary, there was significant positional
Repp: Staccato articulation Page 15
variation in note duration, and this variation depended on hand, register, and step
size, but not on IOI.
Accordingly, Figure 4 shows the results separately for the two hands, broken
down according to register and step size.5 Error bars are omitted for clarity. The
results are not easy to interpret. There are three factors that, hypothetically, could
have played a role. One is metrical structure. This hypothesis predicts a main effect
of position, with longer note durations in the metrically strong first and fifth
positions.6 The first note was indeed lengthened in some conditions, but by no
means in all; and there was little evidence for lengthening in the fifth position. The
second factor is finger. This hypothesis predicts a position by hand interaction, such
that note durations in the second, third, and fourth positions are similar to those in
the eighth, seventh, and sixth position of the same hand, and also to those in the
sixth, seventh, and eighth positions of the other hand (see the abscissa labels of Fig.
4). The data lend some support to this hypothesis within hands (mainly within
conditions), but not between hands. The third possible factor is black versus white
keys. This hypothesis predicts an interaction between position and step size. The
black keys were in positions 2, 4, 6, and 8 in the 1-st sequence, in position 1 only in
the 2-st sequence, and in positions 1, 4, 5, and 6 in the 3-st sequence. No related
patterns are evident in Figure 4. None of these three hypotheses predicts the
observed interaction of position with register. Some similarities between the mid-
register patterns for the two hands should be noted: Longer notes played with the
third finger in 1-st sequences, shorter notes played with the fourth finger in 2-st
sequences. However, the pattern of positional variation remains largely
unexplained.
-----------------------------
Insert Figure 4 here
-----------------------------
Repp: Staccato articulation Page 16
In summary, the main results of this experiment are that produced staccato
note durations are longer than perceptually adjusted ones, and that they increase as
the tempo decreases. The first result suggests a lower limit to produced staccato note
durations, perhaps specific to the mechanical properties of the instrument. In other
words, the pianists may have wanted to produce shorter note durations but
somehow were not able to, either because they could not move their hands/fingers
fast enough or because the digital piano keys did not bounce back quickly enough.
The effect of tempo suggests a small, obligatory effect on hand/finger kinematics,
perhaps due to general muscular tension/relaxation, plus a much larger effect for
some pianists that presumably was under conscious control and may reflect
common performance practice. In other words, these pianists played non-legato
rather than staccato at slower tempi, as may be common when there are dots above
longer note values in printed music.
PRODUCTION EXPERIMENT: REAL PIANO
Method
Participants. Ten pianists participated in a separate experimental session that
preceded the digital piano session and during which also the legato productions
reported in Repp (1997) were recorded. Six of the participants were the same as in
the digital piano session. The additional four participants were one graduate student
of piano performance, two undergraduate pianists, and the author.
Procedure and analysis. The procedure was exactly the same as with the
digital piano. The instrument was a Yamaha Disklavier grand piano. Due to an
unfortunate oversight, the order of the sequences was not recorded for five pianists,
so that hand information was unavailable for them. Their middle-register data were
averaged across the two hands and then duplicated to fill the cells of the overall
Repp: Staccato articulation Page 17
design. Their data were omitted in the separate ANOVA on the mid-register
condition.
Results
The effects of register, step size, and IOI are summarized in Figure 5. It is
evident that the average note durations, ranging from about 40 to 120 ms, were
much shorter than on the digital piano (Fig. 2), though still not as short as the
perceptual adjustments (Fig. 1). Otherwise, their pattern was rather similar to the
digital piano data. Again, there was a clear increase in note durations with IOI
duration [F(2,18) = 6.1, p < .01], but the relatively low significance level and the large
standard errors at the longer IOIs again indicate pronounced individual differences.
There was no significant effect of step size. Register had a marginally significant
main effect [F(2,18) = 6.1, p < .04], again indicating slighly longer note durations in
the lower registers. There were no significant interactions among these three factors.
-----------------------------
Insert Figure 5 here
-----------------------------
The main effect of IOI fell short of significance for two pianists, neither of
whom participated in the digital piano session. Figure 6 compares the note
durations on the digital (Roland) and real (Yamaha) pianos at the three IOIs for the
six pianists who participated in both sessions. There was considerable agreement.
Two pianists (T.C., D.G.) showed similarly large effects of IOI on both instruments,
two (M.A., C.M.) showed smaller effects on the Yamaha than on the Roland, and
two (P.W., P.C.) showed very small effects on both.
-----------------------------
Insert Figure 6 here
-----------------------------
Repp: Staccato articulation Page 18
Unexpectedly, there was a significant main effect of hand in the overall
analysis [F(1,9) = 14.8, p < .004], indicating slightly longer note durations in the left
than in the right hand (an average difference of 8 ms). The effect was also significant
in the ANOVA on the mid-register data of the five pianists for whom hand
information was available [F(1,4) = 11.3, p < .03]. However, it reached significance
for only two of these five pianists individually. Hand did not interact with IOI,
register, or step size, except for a weak quadruple interaction that may be
disregarded.
Again, more complex but reliable results were obtained for position. The
main effect of position was highly significant [F(7,63) = 8.8, p < .0001], and it
interacted strongly with register and hand [F(7,63) = 10.2, p < .0001] and with register,
hand, and step size [F(14,126) = 3.2, p < .0004], and weakly with register and step size
[F(14,126) = 1.8, p < .05] and with hand and IOI [F(14,126) = 2.2, p < .02]. In the left-
hand data, it interacted mainly with register [F(7,63) = 3.9, p < .002], though there
were three other weak interactions. In the right-hand data, it interacted strongly
with register [F(7,63) = 7.1, p < .0001], with register and step size [F(14,126) = 3.2, p <
.0003], and weakly with IOI. In the mid-register data (for 5 pianists only), only the
main effect reached significance. So, it was again mainly the three variables of hand,
register, and step size that affected positional differences.
The patterns of note durations are shown in Figure 7, which is analogous to
Figure 4 for the Roland data. The data are more systematic here in that they show
lengthenings in the first and fifth positions in nearly all conditions. This supports a
metrical interpretation. The finger hypothesis is less clearly supported, and there
seems to be no support at all for the black key hypothesis. The data for the two hands
show some considerable similarities, even between the low and high registers. The
right hand played shorter note durations in the high than in the mid register,
Repp: Staccato articulation Page 19
especially at the end of each cycle, which may account for some of the interactions in
the ANOVA.
In summary, these results largely replicate those for the digital piano. The
shorter note durations presumably reflect the mechanical properties of the real
piano, whose keys offered greater resistance and therefore bounced back more
quickly. However, they evidently still imposed a lower limit to the duration of
staccato notes.
GENERAL DISCUSSION
The results of the adjustment task support the assumption that pianists, at
least, have a perceptual category of staccato according to which a certain range of
note durations is considered optimal. On the digital piano, that range was from 20 to
80 ms. Of course, this finding may be of limited generality. First, there were
considerable individual differences and also considerable variability within
individuals. Second, the results were obtained on a specific instrument whose
sounds were heard over earphones. Although the digital piano tones resembled
natural piano tones, their post-release decay times were relatively short (Repp, 1995,
1997), which means they simulated a relatively nonreverberant environment. Note,
however, that this should have biased participants to choose longer note durations,
not shorter ones. Third, the adjusted note durations increased strongly with IOI, and
it seems possible that even longer note durations might have been chosen, had
even slower tempi been included. Thus the perceptual category of staccato is highly
flexible and context sensitive, as is staccato production.
There were several striking discrepancies between the perceptual adjustments
and the played note durations. The latter were much longer than the former,
especially on the Roland digital piano, but also on the Yamaha Disklavier. The note
Repp: Staccato articulation Page 20
durations at the fastest tempo, which varied little between pianists, probably
represent a lower limit due to the mechanical properties of the instrument. On the
Roland, this lower limit was about 80 ms, whereas on the Yamaha it was 40–50 ms,
on the average. (Some individual pianists achieved shorter durations.) This
difference between instruments most likely reflects the lower resistance and weaker
resilience of the digital piano keys as compared to the real piano keys. These results
suggest that the duration of short staccato notes may depend critically on the piano
action, even within the range of real pianos. In theory, the difference could also be
due to the faster post-release decay of the digital piano tones (see Repp, 1995, 1997),
for which pianists may have compensated by holding the keys down longer, but this
psychoacoustic explanation would not account for the different criteria in perception
and production of the digital piano tones.
At the slower tempi in production, large individual differences came into
play. While some pianists changed their produced note durations relatively little as
IOIs increased, others showed dramatic increases that went way beyond the
acceptable range for staccato suggested by the perceptual adjustment data. Thus these
tones presumably would not be perceived as staccato, even though they were
intended as staccato. This apparent paradox probably reflects a discrepancy between
perception and performance practice. It may have been caused by the musical
notation that was used to elicit the productions, specifically by the fact that IOI
durations were coded as different note values. Staccato articulations of long note
values may be perceptually non-legato (or rather, non-staccato). An alternative way
of notating the music would have been to keep the note values constant (sixteenth
notes) and to code IOI durations in terms of following rests. In that case, less of an
increase in note duration with IOI duration might have been observed.
To the extent that a small increase in note duration with IOI is obligatory or at
least natural, it may be attributed to the kinematics of executing cyclical wrist
Repp: Staccato articulation Page 21
movements at different tempi (Billon & Semjen, 1995; Piek et al., 1993). It seems
reasonable that a slow tempo relaxes somewhat the rapid and tense downward
movement that is required to produce a staccato tone. It is less easy to explain the
increase in perceptually adjusted note durations with IOI. There is no obvious
psychoacoustic reason why short tones spaced further apart should sound
subjectively shorter than tones that occur in close succession. Yet, this is what the
perceptual data suggest. In the absence of a psychoacoustic explanation, it can only be
speculated that staccato tones sound subjectively shorter at a slow tempo because
they tend to be played longer at such a tempo. In other words, the finding may
represent a perceptual-motor interaction (cf. Viviani & Stucchi, 1992a, 1992b).
Another difference between perception and production concerned the effect
of register. Perceptually adjusted note durations were clearly shorter in the low
register than in the mid and high registers. This may be explained by the slower pre-
and post-release decay of low tones, which is likely to increase their subjective
duration and perhaps also their loudness. In staccato playing on either instrument,
however, there was no such effect of register; in fact, there was a slight trend in the
opposite direction. This implies that notes played in the low register would be
perceived as less staccato than those in higher registers. It appears that the pianists
maintained a similar kinematic regime in different registers, regardless of the
perceptual result.
There were no hand differences in performance on the digital piano. On the
Disklavier, however, several pianists played staccato notes somewhat longer with
the left than with the right hand. This may be due to a relative weakness of the left
hand in these pianists and thus is not of great theoretical interest.
Effects of sequential position in production were complex, especially on the
digital piano, and only partially explainable by fingering. On the Disklavier, notes in
metrically strong positions tended to be held longer. Additional data not reported
Repp: Staccato articulation Page 22
here (performance of a short piece by Schumann) suggest that staccato articulation is
less variable than legato articulation as a function of musical factors. This is not
surprising because staccato is essentially a more primitive form of articulation than
legato, requiring only a relatively stereotyped movement and bringing to the fore
the percussive nature of the piano, whereas legato attempts to overcome this
mundane aspect by simulating the singing voice, however imperfectly.
Repp: Staccato articulation Page 23
ACKNOWLEDGMENTS
This research was supported by NIH Grant MH-51230. It would not have been
accomplished without the assiduous assistance of Lisa Robinson and Paul Buechler.
Address correspondence to Bruno H. Repp, Haskins Laboratories, 270 Crown Street,
New Haven, CT 065112-6695. Electronic mail: [email protected]. Internet:
http://www.haskins.yale.edu/haskins/STAFF/repp.html.
Repp: Staccato articulation Page 24
FOOTNOTES
1“Note duration” is MIDI jargon for the time between “note on” (i.e., key
depression) and “note off” (i.e., key release) commands. It is to be distinguished
from acoustic tone duration, which is longer than note duration due to post-release
decay of acoustic vibrations, from auditory (subjective) tone duration, which is a
perceptual magnitude, and from note value, which is a symbol in a score and has no
duration at all.
2 A MAX patcher is a program written in MAX, which is a graphic
programming language for MIDI applications. One peculiarity of MAX is that the
tempo of the MIDI output and input does not correspond exactly to the time values
specified in and recorded by the program. With the exception of the IOI durations of
the stimuli, which were specified in MAX so as to yield the real-time values stated
(187, 325, and 750 ms), all data (perceptual adjustments and production durations)
will be reported in uncorrected form, as they appeared inside MAX. The real-time
values were uniformly longer by about 4%, as determined by control measurements.
(This was with a setup in which the OMS software was disabled.)
3The decay characteristics of the digital piano tones are described in Repp
(1995). It should be noted that strings above F#6 do not have dampers on a real
piano, and this was simulated on the digital piano as well, so that varying note
duration had no effect on the sound of the upper three notes of the high-register
sequences.
4Due to a long delay between data collection and analysis, the order of the
perception and production tasks could no longer be determined with certainty.
Repp: Staccato articulation Page 25
However, the author believes that the production task always preceded the
perception task.
5The data of the pianist whose mid-register data were averaged across left and
right hands are included here.
6An interaction of meter with IOI might also be predicted, due to changes in
metrical structure with tempo, but no such interaction was evident in the statistical
analyses.
Repp: Staccato articulation Page 26
REFERENCES
Billon, M., & Semjen, A. (1995). The timing effects of accent production in
synchronization and continuation tasks performed by musicians and
nonmusicians. Psychological Research, 58, 206–217.
Kuwano, S., Namba, S., Yamasaki, T., and Nishiyama, K. (1994). Impression of
smoothness of a sound stream in relation to legato in musical performance.
Perception & Psychophysics, 56, 173–182.
Martin, D. W. (1947). Decay rates of piano tones. Journal of the Acoustical Society of
America , 19, 535–541.
Piek, J. P., Glencross, D. J., Barrett, N. C., & Love, G. L. (1993). The effect of temporal
and force changes on the patterning of sequential movements. Psychological
Research, 55, 116–123.
Repp, B. H. (1995). Acoustics, perception, and production of legato articulation on a
digital piano. Journal of the Acoustical Society of America, 97, 3862–3874.
Repp, B. H. (1997). Acoustics, perception, and production of legato articulation on a
computer-controlled grand piano. Journal of the Acoustical Society of
America , 102, 1878-1890.
Viviani, P., & Stucchi, N. (1992a). Biological movements look uniform: Evidence of
motor-perceptual interactions. Journal of Experimental Psychology: Human
Perception and Performance, 18, 603–623.
Viviani, P., & Stucchi, N. (1992b). Motor-perceptual interactions. In G. E. Stelmach &
J. Requin (eds.), Tutorials in motor behavior II (pp. 229–248). Amsterdam:
Elsevier.
Repp: Staccato articulation Page 27
FIGURE CAPTIONS
Fig. 1. Results of perceptual adjustment experiment: Average note duration
and within-pianist standard deviation as a function of IOI, step size, and register,
with standard error bars.
Fig. 2. Results of production experiment on the Roland digital piano: Average
note duration and within-pianist standard deviation (across sequential note
positions) as a function of IOI, step size, and register, with standard error bars.
Fig. 3. Perceptually adjusted note durations plotted against note durations
played on the Roland digital piano, separately for 7 pianists, as a function of IOI
(increasing from lower left to upper right).
Fig. 4. Average note durations on the Roland digital piano as a function of
sequential position, register, and step size, separately for (a) the left hand and (b) the
right hand.
Fig. 5. Results of production experiment on the Yamaha Disklavier: Average
note duration and within-pianist standard deviation (across sequential note
positions) as a function of IOI, step size, and register, with standard error bars.
Fig. 6. Note durations played on the Roland digital piano plotted against note
durations played on the Yamaha Disklavier, separately for 6 pianists, as a function of
IOI (increasing from lower left to upper right).
Fig. 7. Average note durations on the Yamaha Disklavier as a function of
sequential position, register, and step size, separately for (a) the left hand and (b) the
right hand.
Repp: Staccato articulation Page 28
J
J
J
J
J
J
J
JJ
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
XX
X
XX
X
XXX
XXX
XXX
XX
X
XX
X
XXX
XXX
187
375
750
187
375
750
187
375
750
187
375
750
187
375
750
187
375
750
187
375
750
187
375
750
187
375
750
0
10
20
30
40
50
60
70
80
90
100A
vera
ge n
ote
dura
tion
and
s.d.
(m
s) ±
1 s.
e.
IOI (ms)
J Average dur.
X Average s.d.
Low register Middle register High register
1 st 2 st 3 st 1 st 2 st 3 st 1 st 2 st 3 st
Fig. 1
Repp: Staccato articulation Page 29
J
J
J
J
J
J
JJ
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
XXX XXX XXX XXX XXX XXX XXX XXX XXX
187
375
750
187
375
750
187
375
750
187
375
750
187
375
750
187
375
750
187
375
750
187
375
750
187
375
750
0
50
100
150
200
250A
vera
ge n
ote
dura
tion
and
s.d.
(m
s) ±
1 s.
e.
IOI (ms)
J Average dur.
X Average s.d.
Low register Middle register High register
1 st 2 st 3 st 1 st 2 st 3 st 1 st 2 st 3 st
Fig. 2
Repp: Staccato articulation Page 30
B
B
B
J
J
J
HHH
FF
F
Ñ
Ñ
Ñ
É ÉÉÇ Ç Ç
0
50
100
150
200
250
300
0 10 20 30 40 50 60 70 80
Pla
yed
note
dur
atio
n on
Rol
and
(ms)
Adjusted note duration (ms)
B M.A.
J T.C.
H P.W.
F C.M.
Ñ D.G.
É P.C.
Ç H.S.
Fig. 3
Repp: Staccato articulation Page 31
B
BB
BB
BB
B
J J
JJ J
J
J
J
H
H
H
H
H
H
H H
GG
G
G G G
G
G
EE
E
E
E
E
E
EC
CC
C CC
C
C
5 4 3 2 1 2 3 4100
110
120
130
140N
ote
dura
tion
(ms)
Left-hand fingers
B 1 st low
J 2 st low
H 3 st low
G 1 st mid
E 2 st mid
C 3 st mid
B
BB
B B
BB
B
J
J J
J
J
J
JJ
H
H
H
H
H
H
HH
G
G
G
G
G
GG
G
E
E E
E
E
E
E
EC
CC
C CC
C C
1 2 3 4 5 4 3 2100
110
120
130
Not
e du
ratio
n (m
s)
Right-hand fingers
B 1 st high
J 2 st high
H 3 st high
G 1 st mid
E 2 st mid
C 3 st mid
(a)
(b)
Fig. 4
Repp: Staccato articulation Page 32
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
XXX XXX XXX XXX XXX XXX XXX XXX XXX
187
375
750
187
375
750
187
375
750
187
375
750
187
375
750
187
375
750
187
375
750
187
375
750
187
375
750
0
20
40
60
80
100
120
140
160A
vera
ge n
ote
dura
tion
and
s.d.
(m
s) ±
1 s.
e.
IOI (ms)
J Average dur.
X Average s.d.
Low register Middle register High register
1 st 2 st 3 st 1 st 2 st 3 st 1 st 2 st 3 st
Fig. 5
Repp: Staccato articulation Page 33
B
B
B
J
J
J
HHH
FF
FÑ
Ñ
Ñ
ÉÉÉ
0
50
100
150
200
250
300
0 50 100 150 200 250 300
Pla
yed
note
dur
atio
n on
Yam
aha
(ms)
Played note duration on Roland (ms)
B M.A.
J T.C.
H P.W.
F C.M.
Ñ D.G.
É P.C.
Fig. 6
Repp: Staccato articulation Page 34
B
B
B B
B
BB
B
J
J
J
J
J
JJ
J
H
H
HH
H
H
H
HG
G G
G
G
G
G G
E
E E
E
E
E
EE
C
CC
C
C
C
C
C
5 4 3 2 1 2 3 470
80
90
100N
ote
dura
tion
(ms)
Left-hand fingers
B 1 st low
J 2 st low
H 3 st low
G 1 st mid
E 2 st mid
C 3 st mid
(a)
B
BB
B
B
B BB
J
JJ
J
J
J
JJ
H
H HH
H
H
H
H
G
G
GG
G
GG
G
EE
E
E
E
E
E
E
CC
C
C
C
C
C
C
1 2 3 4 5 4 3 260
70
80
90
Not
e du
ratio
n (m
s)
Right-hand fingers
B 1 st high
J 2 st high
H 3 st high
G 1 st mid
E 2 st mid
C 3 st mid
(b)
Fig. 7