COGNITION AND EMOTION

2005, 19 (8), 1113±1139

Multidimensional scaling of emotional responses to

music: The effect of musical expertise and of the
duration of the excerpts
E. Bigand
LEAD-CNRS, Universite de Bourgogne, Dijon, France
S. Vieillard
IRCAM-CNRS-Paris, France
F. Madurell
LEAD-CNRS, Universite de Bourgogne, Dijon and UFR de musicologie, Paris
IV-Sorbonne, France
J. Marozeau
IRCAM-CNRS-Paris, France
A. Dacquet
UFR de musicologie, Paris IV-Sorbonne, France

Musically trained and untrained listeners were required to listen to 27 musical

excerpts and to group those that conveyed a similar emotional meaning (Experi-
ment 1). The groupings were transformed into a matrix of emotional dissimilarity
that was analysed through multidimensional scaling methods (MDS). A 3-
dimensional space was found to provide a good fit of the data, with arousal and
emotional valence as the primary dimensions. Experiments 2 and 3 confirmed the
consistency of this 3-dimensional space using excerpts of only 1 second duration.
The overall findings indicate that emotional responses to music are very stable
within and between participants, and are weakly influenced by musical expertise
and excerpt duration. These findings are discussed in light of a cognitive account
of musical emotion.

Correspondence should be addressed to E. Bigand, LEAD-CNRS, 2, esplanade Erasme, Pole

AAFE, Universite de Bourgogne. F-21000 Dijon, France; e-mail: [email protected]

# 2005 Psychology Press Ltd

http://www.psypress.com/cogemotion DOI:10.1080/02699930500204250
1114 BIGAND ET AL.

Music is a complex acoustic and temporal structure that induces a large variety
of emotional responses in listeners. Although the richness of these responses is
what motivates an engagement with music, relatively few studies have been
devoted to understanding their psychological foundation. This situation is cur-
rently changing, as attested by the publication of an increasing number of studies
investigating emotional responses to music at both behavioural level (Gab-
rielsson & Juslin, 1996, 2003; Gabrielson & Lindstrom, 2001; Gabrielsson,
2001; Robinson, 1997; Sloboda & Juslin, 2001) and neurophysiological level
(Bartlett, 1999; Peretz 2001; Scherer & Zentner, 2001, for reviews). Two main
issues of interest can be distinguished in this domain. The first concerns the
content of emotional responses to music, the second the nature of the process
involved in these responses. The present study deals with both issues.
The nature of emotions induced by music has been a matter of much debate.
Preliminary empirical investigations have demonstrated that basic emotions,
such as happiness, anger, fear, and sadness, can be recognised in and induced by
musical stimuli in adults and in young children (Cunningham & Sterling, 1988;
Dolgin & Adelson, 1990; Kastner & Crowder, 1990; Tergwogt & Van Grinsven,
1991). The experimental methods used consisted in matching emotional
responses with either linguistic labels or pictures of expressive faces, or involved
electrophysiological reactions to music. All of these methods converge to
demonstrate a strong consistency between participants, as long as musical
excerpts were chosen to convey very basic emotions. The conclusion that music
induces three or four basic emotions is encouraging for scientists but it is far
from compelling for music theorists, composers, and music lovers. Indeed, it is
likely to underestimate the richness of the emotional reactions to music that may
be experienced in real life. The question of whether emotional responses go
beyond four main categories is a central issue for theories of human emotion
(see Averill, 1994; Ekman, 1994; Smith & Ellsworth, 1985).
An alternative approach is to stipulate that musical emotions evolve in a
continuous way along two or three major psychological dimensions. A lot of
previous research has established a 2-dimensional structure of emotion for
facial expressions (Abelson & Sermat, 1962; Osgood, 1966), voice perception
(Green & Cliff, 1975), and affect words (Russell, 1978, 1980). According to
Russell (1980) our emotional knowledge are organised and summarised into a
cognitive structure that helps, in turn, to shape the perception and interpretation
of events. In his circumplex model of affect, Russell (1980) proposed a cogni-
tive structure of affective experience that relied on the interrelationship
between affective concepts (words with hedonic meaning). The affective space
he obtained by using three different scaling techniques revealed two orthogonal
and bipolar dimensions of pleasure-displeasure and degree of arousal. Beyond
arousal and pleasantness, the dimensions that emerge were often difficult to
interpret, and when interpretable, they vary from study to study (Smith &
Ellsworth, 1985).
 EMOTIONAL RESPONSES TO MUSIC 1115

In the music domain, an attempt to model the perception of expressive

content of Western music in multidimensional space has been developed
(Hevner, 1935; Juslin, 2001; Madsen, 1997; Russell, 1980; Schubert, 1996,
2001, 2004) and specific neurophysiological reactions associated with the
arousal and valence dimensions were also reported (Schmidt & Trainor, 2001;
Tsang, Trainor, Santesso, Tasker, & Schmidt, 2001). Only a few studies how-
ever attempted to confirm that arousal and valence were actually the most
important determinant of emotional responses to music. In Wedin (1969, 1972a,
1972b) musical excerpts were rated on a number of semantic scales defined by
emotionally coloured adjectives. Factor analysis then extracted the dimensions
accounting for the evaluation. In Imberty (1979, 1981), a 2-dimensional struc-
ture was found by analysing the semantic contents of words freely associated by
listeners to musical excerpts. In both groups of research, arousal and emotional
valence emerged as the main dimensions of emotional responses to music. In
each case, a third dimension was found but it corresponded to ``solemnity-
triviality'' in Weldin studies and was associated to figural gestures in Imberty
(1979). More recently, Canazza, De Poli, Vidolin, and Zanon (2001) presented
participants with an excerpt of Mozart's Clarinet Concerto (K 622) that was
interpreted and recorded with different levels of expressivity. Participants were
asked to use a set of sensorial adjectives (e.g., light, soft, dark, hard. . .) to
express the emotion they felt the excerpt possessed. Using multidimensional and
factorial analyses, they obtained a semantic 2-dimensional space of musical
expressivity in which two dimensions appeared: one called ``kinematics'' that
corresponded to the tempo parameter and another called ``energy'' related to the
intensity parameter. Such a kinematics-energy space provides useful information
on how expressiveness is organised in the listener's mind. Interestingly, this
study indicates that kinematics aspects could be a main determinant of partici-
pants' judgements on musical expressiveness.
A common feature of these studies was to investigate emotional responses to
music by using linguistic labels. The use of verbal labels is potentially proble-
matic since it can encourage participants to simplify what they actually
experience (Scherer, 1994). Some philosophers even argued that musical emo-
tions are definitely ineffable (JankeÂleÂvitch, 1974, 1983/2003). As nicely coined
by the composer, Felix Mendelssohn, the emotion induced by music could be so
rich that even language would not be able to account for it. If this was the case,
using linguistic labels in empirical research could result in missing important
aspects of musical emotion.1 The first goal of the present study was to inves-
tigate the emotion conveyed by musical pieces without using linguistic

1
In the place of linguistic responses, it is also possible to record electrophysiological responses to
musical stimuli (Khalfa, Peretz, Blondin, & Robert, 2002; Krumhansl, 1997). These methods are
promising, but it is uncertain whether they are refined enough to account for the subtlety of musical
emotions.
1116 BIGAND ET AL.

responses. Participants were encouraged to entirely focus on their own emo-

tional experience of the musical excerpts and to group those that conveyed
similar subjective emotions. A simple rule transformed the partition obtained
into a dissimilarity matrix: stimuli grouped together were set with a dissimilarity
of 0, while everything else was set to 1. Multidimensional scaling (MDS)
method was then used to analyse the psychological dimensions underlying this
matrix.
The instruction we used encouraged participants to focus on induced emo-
tions and not to recognise the emotional information encoded in the music. That
it to say, the task prevents using labels that can bias emotional judgement The
difference between induced and perceived emotion has been a matter of debate
in the literature of facial expression, and the debate has transpired in the music
domain as well (Gabrielsson, 2001; Scherer & Zentner 2001). Of course, asking
participants to focus on induced emotions does not guarantee that they actually
respond more on the basis of what they experienced than on the basis of the
expressions they recognise in the stimuli. It is even unclear whether both aspects
of emotional experience might be separable. In order to better specify the type of
emotion our task was tapping into, we ran a pilot experiment that was designed
to shed some light on this issue. Participants were first required to evaluate the
emotions they felt on bipolar semantic scales to while listening to musical
excerpts. They were then asked to evaluate on the same scale the emotions
presumably felt by a representative sample of listeners of the matched age and
sociocultural level. An extremely high correlation was found between both
evaluations with r(34) = .93, p < .001. This suggests that induced and perceived
emotions, if different, are likely to be two different faces of the same coin. We
thus will consider in the context of this article, that participants' responses
expressed emotions they perceived in the stimuli, either on basis of their ability
to recognise expression in music, or on the basis of the emotion they felt, or,
more probably, on a mixture of both phenomena. The first aim of the study was
to assess the degree of consistency in emotional responses observed with this
method, and to further specify the psychological dimensions that govern these
responses.
A second main broad issue of the study relates to the mechanisms that govern
emotion. Arguments for the contribution of cognitive processes to emotional
responses are clearly favoured by research in music cognition (see Bever, 1988;
Jackendoff, 1992; Meyer, 1956, 2001). Following Mandler's (1975) theory of
emotion, Meyer (1956) argued that emotional responses to music depend upon
whether the musical pieces fulfil perceptual and cognitive expectancies gener-
ated by the opening of the piece. Musical expectancies result from processing
several features, such as harmony, rhythm, melody, and thematic relationships
cognitively. This processing is believed to occur automatically at an implicit
level (for a review, see Tillmann, Bharucha, & Bigand, 2000). A cognitive
account of musical emotion is supported by empirical studies showing that
 EMOTIONAL RESPONSES TO MUSIC 1117

emotional responses are systematically associated with the presence of specific

musical features (Sloboda, 1991, 1992; for a review, see Gabrielsson, 2001; see
Sloboda & Juslin, 2001). For example, musical features such as modulation,
grace notes, and harmonic progressions, are often associated with emotional
responses in the verbal reports of participants (Sloboda, 1991). In a related vein,
Peretz, Gagnon, and Bouchard (1998) demonstrated that rhythm and modality
(major vs. minor) contribute to happiness or sadness. These studies pointed out
that part of the emotional response to music results from the cognitive proces-
sing of musical structures (key, modalities, rhythm), which are known to be
expressive in the context of the Western musical system. The precise organi-
sation of these cognitive processes, however, remains unspecified.
This cognitive account has recently been challenged by two sets of data.
Emotional responses were shown to rest on a very fast acting mechanism, so that
250 ms of music may suffice to distinguish happy from sad excerpts (Peretz et
al., 1998). This suggests that some basic emotions might function like sub-
cortical reflexes, without cortical mediation. The affective system is faster and
developed earlier than the cognitive system (Reiman, Lane, Ahern, Davidson, &
Schwarz, 2000; Zajonc, 1984). Independence between the affective and cogni-
tive systems has been demonstrated in the visual domain (LeDoux, 2000).
Cerebral lesions have been found to alter the recognition of facial expressions
while the ability to identify faces was spared (Adolphs, Tranel, Damasio, &
Damasio, 1994; Adolphs et al., 1999; Lane, Reiman, Ahern, Schwartz, &
Davidson, 1997). In the field of music, IR, a brain-damaged patient who suffers
from considerable deficits in music cognition, was shown to be able to dis-
criminate between happy and sad excerpts while being unable to determine
whether the music was familiar or not (Peretz et al., 1998; Peretz, Blood,
Penhume, & Zatorre, 2001). Moreover, IR performed similarly to controls when
differentiating sad from happy excerpts of extremely short duration. This
suggests that severe deficits in the cognitive processing of music may leave
emotional responses unimpaired.
The second main issue of the present study was to further investigate the
influence of excerpt duration on emotional response. The experimental method
used enabled us to specify which dimension of musical emotion would be
processed first. It was likely that the emotional valence of music, which seems to
be associated with the processing of Western harmony (major vs. minor), would
require more time to be processed than arousal, which is linked to dynamic
aspects of musical stimuli.
The final issue of the present study was to evaluate the effect of musical
expertise. It is usually believed that musical experts process musical structure in
a more refined way than nonmusicians. This issue has been challenged by a
large number of empirical studies showing that nonmusicians respond to subtle
musical structure in a comparable way to those with musical training (for a
review, see Bigand & Poulin-Charonnat, in press). It is thus of critical impor-
1118 BIGAND ET AL.

tance to further investigate whether the emotional experience of the two groups
is similar.
To summarise, the present study investigates three main issues for the study
of musical emotion. Experiment 1 was designed to outline through MDS the
psychological dimensions that govern emotional responses to Western music.
Experiments 2 and 3 assessed how a considerable shortening of excerpt duration
would modify the structure of emotional reactions. In addition, all experiments
were run with musicians and nonmusicians in order to evaluate the influence of
musical expertise on emotion.

EXPERIMENT 1

Method
Participants. A total of 19 participants performed the two experimental
sessions. Of these, 9 did not have any musical experience and are referred to as
nonmusicians and 10 were graduate music students and are referred to as
musicians.

Material. A sample of 27 musical excerpts of serious nonvocal music was

selected by music theorists and psychologists according to several constraints
(see footnote 1). All excerpts were expected to convey a strong emotional
experience. They were chosen to illustrate a large variety of emotions, and to be
representative of key musical periods of Western classical music (baroque,
classic, romantic, and modern) as well as of the most important instrumental
groups (solo, chamber music, orchestra). This final constraint is of methodo-
logical importance since it helps to neutralise any confound effect between the
structural surface similarity and the emotional similarity of excerpts. As much as
possible, we avoided selecting excerpts that were too famous (Vivaldi's Four
Seasons, Barber's or Albinoni's Adagio, Holst's Planets). Appendix A displays
a short description of the 27 excerpts used as stimuli. They corresponded either
to the beginning of a musical movement, or to the beginning of a musical theme
or idea. An average duration of 30 s duration sounded appropriate for the
purpose of Experiment 1.

Apparatus. All excerpts were real performances taken from commercial

compact disk (CDs). They were played by a mono file at 16 bits and 44 kHz.
The overall loudness of the excerpts was adjusted subjectively to ensure a
constant loudness across the experiment. The sound files were represented on
the computer screen by small loudspeaker icons in a Powerpoint file. The
position of each loudspeaker on the screen was chosen at random. Each
participant worked on a different Powerpoint file and listened to the excerpts by
clicking on the icons.
 EMOTIONAL RESPONSES TO MUSIC 1119

Procedure. Participants were presented with a random visual pattern of 27

loudspeakers, representing the 27 excerpts. They were required first to listen to
all of these excerpts and to focus their attention on their private emotional
experience. They were then asked to look for excerpts that induced similar
emotional experience (whatever that may be) and to drag the corresponding
icons in order to group these excerpts. They were allowed to listen to the
excerpts as many times as they wished, and to regroup as many excerpts as they
wished. A second experimental session was performed 2 weeks later. It followed
the same procedure. Each session lasted about 40±50 min, on average. After the
experiment, participants were also invited to describe verbally the emotions that
governed their grouping.

Results
The first analysis concerned the number of groups produced by participants.
Participants distinguished an average of 8 groups (7.68 for session 1, and 7.74
for session 2). Musicians did not produce more groups than nonmusicians. The
percentage of excerpts that were grouped differently in sessions 1 and 2 was low
(12%) but significantly higher for nonmusicians, (15%), than for musicians,
(9%), F(1, 14) = 8.34, p < .02.
The groupings of participants were then converted into a 27 6 27 matrix of
co-occurrence. Each cell of the matrix indicated the average number of times
that two excerpts were grouped together. The subtraction of the average matrix
of occurrence from 1 resulted in a matrix of dissimilarity. The matrices obtained
for sessions 1 and 2 were highly correlated for musicians, r(349) = .87, p < .001,
and for nonmusicians, r(349) = .78, p < .001. The correlation between the
matrices of both groups was r (349) = .83, p < .001. Given the strong consistency
of responses, the four matrices were fused. The resulting matrix was analysed
with MDS and cluster analysis methods.

Multidimensional scaling. The matrix was processed by the EXSCAL MDS

program (Winsberg & Caroll, 1989). This model postulates that the distance dij,
between the ith and the jth stimuli is given by

" #12
X
R
Dij …Xir ÿ Xjr†2 ‡ Si ‡ Sj
rˆ1

where Xir is the coordinate of the ith stimuli on the rth dimension and r is the
number of dimensions. In this model, in addition to r common dimensions, the
stimuli can have a unique dimension, denoted Si, not shared by other stimuli. A
maximum likelihood criterion, named BIC and a study of the stress value (Borg
1120 BIGAND ET AL.

Figure 1. Geometrical solution for the 27 excerpts, resulting from MDS of Experiment 1 data. The
main outcome of the cluster analysis is superimposed on the MDS solution with dashed lines
representing the first two main clusters, and solid lines, the two subordinates clusters.

& Groenen, 1997) was used to determine the number of dimensions of the space.
Both methods indicated a 3-dimensional space with specificities. Adding the
third dimension to the model increases the explained variance from 73% to 83%.
The location of the 27 excerpts along the two principal dimensions is
represented in Figure 1.2 The vertical axis opposes musical excerpts that varied
obviously by their arousal level (with low arousal pieces at the bottom, and high
arousal pieces at the top). The horizontal axis presumably opposes musical
excerpts that differ by their emotional valence (with positive valence on the right
and negative valence on the left). The third axis tends to oppose pieces with
broad and regular melodic contours from those that proceed harmonically or by
broken arpeggios.

2
For the sake of convenience, the third dimension is not represented here.
 EMOTIONAL RESPONSES TO MUSIC 1121

A bootstrap analysis was performed to assess the instability of the location of

excerpts in this emotional space. A new MDS space was created by randomly
selecting 30 subjects, with replacement (a same subject can be used many times,
while others never). This technique was repeated 200 times. Each space was
rotated towards the same target. The greater the variability between subjects, the
more each space must be different, and vice versa. The 200 spaces were
superimposed on a single representation. The position of each excerpt in these
200 analyses defined a cloud of points on this single representation. The size of
the ``cloud'' expresses the stability of each excerpt in the three dimensional
space. As illustrated by Diagram 1 (Appendix A), the surfaces covered by the
ellipses were fairly small, providing evidence for the reliability of the geome-
trical representation.
Finally, the MDS solution was compared with a cluster analysis performed on
the emotional matrix. The main clusters isolated at the top of the dendogram
were superimposed on the MDS solution in Figure 1. The two largest groups
opposed pieces of different arousal level. The next two groups contrasted pieces
of different emotional valence. This analysis delineated four main groups of
pieces varying by their dynamic and emotional valence, and roughly corre-
sponding to ``happiness'' (dynamic and positive valence), ``serenity '' (low
dynamics, and positive valence), ``fear and anger'' (high dynamics and negative
valence), and sadness (low dynamics and negative valence). Subtler categor-
isation was found within each of these groups.
To further investigate the psychological meaning of these dimensions, a
control experiment was run with 19 musically untrained listeners from the
University of Burgundy. They were asked to order the 27 excerpts on a single
continuous dimension (varying from 7100 to +100) that corresponded either to
an ``arousal dimension'', an ``emotional valence'' dimension, and a ``plea-
santness'' dimension. The average coordinates obtained for each of these
dimensions separately with the control group was then correlated with the
coordinates of the excerpts revealed by the MDS and illustrated in Figure 1.
Correlation was r(25) = .94, p < .001 for the arousal dimensions, and r(25) = .82,
p < .001 for the emotional valence dimensions. The pleasantness judgements
were not correlated to the emotional valence dimension, r(25) = .08, but were
moderately related to the arousal dimension, r(25) = .51, p < .01. None of these
judgements correlated with the third dimension revealed by the MDS.

Discussion
Experiment 1 provides some evidence that music induces straightforward
emotions. Strong consistency within listeners was found between sessions 1 and
2 and between the two groups of participants, irrespective of their musical
expertise. The data also confirmed that two main psychological dimensions
govern emotional responses: arousal and emotional valence. The control
1122 BIGAND ET AL.

condition contributed an important finding: The emotional valence dimension

was not correlated with pleasantness judgements. This suggests that sad music is
not systematically associated with unpleasant emotion and vice versa. This
apparent paradox distinguishes the present finding from those usually reported
in multidimensional scaling of emotion (see Smith & Ellsworth, 1985) and
emphasises a specific characteristic of musical emotions. This apparent paradox
is consistent with recent brain imagery studies showing that sad music may
activate neural pathways involved in biological reward (Blood & Zatorre, 2001).
Experiment 1 also revealed that a third dimension contributes to emotional
responses to music. This dimension contrasted pieces with different melodic
gestures. A possible interpretation is that this dimension express the influence
body gestures evoked by musical excerpts might have on perceived emotions.
The link between music and movement is well established, and several authors
have emphasised that musical affects arise in large part from their relation to
physical patterns of posture and gesture (Damasio, 2003; FranceÁs, 1956; Jack-
endoff & Lerdahl, in press). A large melodic arch should thus evoke slow, large
and continuous gestures, while broken arpeggios may evoke discontinuous and
more saccadic movements. Given that a considerable amount of emotional
experience is presumably embedded in cortical sensory motor maps (Damasio,
2003), the evocation of gestures by music could have relevant influence on
listeners' emotional experience. However, given the difficulty to describe with
linguistic terms these kinetic emotions, it remains difficult to further investigate
the emotional qualities associated with this third dimension.
The last important finding of Experiment 1 is that emotional responses to
music were only weakly influenced by musical expertise. Both groups of
musicians and nonmusicians produced an equal number of clusters, and their
matrices of emotional similarity were highly correlated. This weak difference is
all the more surprising given the complexity of the musical stimuli used.
All of these findings (i.e., test-retest reliability, strong consistency
between two independent groups) suggest that processing musical emotions
relies on robust processes that do not depend upon an explicit knowledge
of Western music. Experiment 2 was designed to further investigate the nat-
ure of these processes by manipulating the time duration of the excerpts.
According to Peretz et al, (2001), emotional responses can arise from a
very fast process that does not necessarily involve complex cognitive media-
tion. This conclusion might, however, be qualified since arousal and emo-
tional valence were confounded in her study, with all positive excerpts
being of high arousal (``happy'' category) and all negative excerpts being
of low dynamics (``sadness'' category). Accordingly, participants may have
correctly performed the ``happy''-``sad'' distinction at 250 ms, by consider-
ing only the dynamics of the excerpts. The purpose of Experiment 2 was
to evaluate how reducing the duration of the music to 1 s would affect the
dimensionality of emotional response, and the location of the 27 excerpts
 EMOTIONAL RESPONSES TO MUSIC 1123

in this space. We expected that only one dimension (arousal) would then con-
tribute to the emotional response.

EXPERIMENT 2
Method
Participants. Twenty participants without musical training (referred to as
nonmusicians) and 20 with an average of 10 years of musical training and
instrumental practice participated in this experiment. None of them had taken
part in Experiment 1.

Material and apparatus. The material and apparatus were identical to those
described for Experiment 1, with the exception that musical excerpts were
reduced to a duration of 1 s in the 1 s condition. The sections of 1 s were the
beginnings of the original excerpts. Each 1 s excerpt began with a 50 ms fade-in
and ended with a 50 ms fade-out to remove any acoustic clicks. A short
description of the musical structure found in the first second of each of these 27
excerpts is provided in Appendix B.

Procedure. The experimental task was similar to that of Experiment 1. In

session 1, participants were asked to group the 1 s excerpts that induced similar
emotional experiences. A second session was performed 1 week later with the
same participants who performed the task with the excerpts used in Experiment
1 (referred to below as the 30 s condition).

Results
The first analysis compared the number of groups produced by participants in
sessions 1 and 2. Participants distinguished 6.73 and 7.44 groups on average, in
the 1 s and 30 s conditions, respectively, and there was no influence of musical
expertise. The number of changes in grouping between the 1 second and the 30 s
conditions was 16.6% and 19.6%, on average, for musicians and nonmusicians,
respectively. This difference did not reach significance.
The dissimilarity matrices obtained with musicians for the 1 s and 30 s
conditions were highly correlated r(349) = .76, p < .001. Interestingly, the
matrix of the 30 s condition was also correlated with the matrix obtained in
Experiment 1, r(349) = .71, p < .001 for session 1, and r(349) = .70, p < .001 for
session 2. This demonstrates a strong stability of the emotional response
between two independent groups of participants. For nonmusicians, the corre-
lation between the 1 s and the 30 s matrices was r(349) = .70, p < .001. In
addition, musically untrained participants responded very similarly to musicians
in the 1 s duration, r(349) = .80, p < .001 and in the 30 s duration, r(349) = .87, p
< .001. Once again, the 30 s similarity matrix of Experiment 2 correlated
1124 BIGAND ET AL.

significantly with the nonmusicians' matrix found in Experiment 1 for session 1:

r(349) = .87, p < .001; and for session 2; r(349) = .89, p < .001.
Multidimensional scaling analyses were performed separately for each group
and in each duration condition. In each case, a 3-dimensional solution provided a
very good fit with the data. For musicians, the addition of the third dimension to
the model increased the percentage of explained variance from 68% to 85% and
from 67% to 80% in the 1 s and 30 s conditions, respectively. For nonmusicians,
this percentage of variance increased from 77% to 87% for the 1 s and 30 s
conditions. The location of the 27 excerpts along the first two spatial dimensions
is represented in Figure 2 (see note 2). The first dimension contrasts musical
excerpts of different arousal. The second dimension contrasts musical excerpts
of different emotional valence. As already reported in Experiment 1, the third
dimension tended to oppose excerpts beginning with a melodic interval (small
melodic gesture) against those that begin with a chord (no melodic gesture). The
geometrical solution found for the 30 s duration was very similar to that reported
in Experiment 1.3 An interesting new issue is the geometrical solution for the 1 s
condition. On the whole, the location of the excerpts did not change con-
siderably, except that the excerpts were further apart than in the 30 s condition.
Cluster analysis helped to highlight some other noticeable differences
between the 1 s and 30 s conditions. The main outcome of this analysis was
superimposed on the geometrical solution in Figure 2. With musicians, three
main groups of equal size were found in the 1 s condition, with all the excerpts
of low arousal forming a single cluster. In the 30 s condition, the cluster of low
arousal excerpts split into two clusters: one of low dynamic and negative valence
(excerpts 9,7,8,3), the other of low dynamic and positive valence (excerpts 19, 2,
20, 1, 6, 4, 21, 5). This difference suggests that musicians encountered some
difficulties perceiving the emotional valence when the 1 s excerpts induced a
low arousal. In other words, the distinction between ``sad'' and ``serene'' was
rendered more difficult with short duration excerpts, as illustrated by the mis-
allocation of excerpt 7.
The geometrical solution found for nonmusicians in the 1 s and 30 s con-
ditions was globally similar to that of musicians. As summarised in Figure 2, the
first three main groups of the 1 s condition differentiated excerpts of different
arousal, irrespective of their emotional valence. When the excerpts are 30 s long,
high arousal excerpts were split into two groups: one of negative emotional
valence (25, 27, 18, 16, 26, 17, 12) and one of positive valence (10, 11, 13, 15,
24, 14, 22, 23). A similar phenomenon occurred for the excerpts of low arousal
that split into a cluster of negative valence (9, 7, 8, 3) and a cluster of positive
valence (19, 2, 5, 20, 21, 1, 4, 6), in the 30 s condition. This finding indicates a

3
Bootstrap analyses were also performed in order to evaluate the degree of instability of these
geometrical solutions. They confirmed a generally high stability for the MDS geometrical solutions.
 EMOTIONAL RESPONSES TO MUSIC 1125

slight difference between nonmusicians and musicians. In the 1 s condition, the

former encountered some difficulty in distinguishing excerpts according to their
emotional valence, while the latter encountered this difficulty mostly with
excerpts of low arousal.

Discussion

Experiment 2 demonstrated that a considerable shortening of musical excerpts

(30 s to 1 s) had only a weak effect on emotional responses. This finding is
consistent with the conclusion of Peretz et al. (2001). It is possible however, that
this weak effect of duration was due to a lack of sensitivity of the experimental
method. Given that participants were not asked to compare all pairs of musical
excerpts before responding, the resulting matrix of dissimilarity might capture
only the most robust features of emotional experience. If time duration was
influencing response in a more subtle way, the current method would not be
refined enough to capture this effect. A third experiment was run to address this
issue. This time, participants were presented with all possible pairs of 1 s
excerpts (i.e., 351) and were asked to evaluate the emotional dissimilarity of the
excerpts of each pair on a subjective scale. The geometrical solution produced
by this method was compared with those found in the 1 s condition of
Experiment 2.

EXPERIMENT 3
Method
Participants. Fifteen participants without musical training and 15 with
musical training and instrumental practice performed this experiment. None had
participated in the previous experiments.

Material and apparatus. These were identical to Experiment 2 except that

the 1 s excerpts were presented in pairs, with a 400 ms silence between the two
excerpts.

Procedure. In the first part of the experiment, participants were familiarised

with the 27 excerpts ± they could listen to them as many times as they wished. In
the second part, they were presented with 351 pairs of excerpts, played in a
random order. After each pair, they were required to evaluate on a continuous
scale how the emotion experienced for each excerpt differed. The scale included
two extremities ranging from ``different'' on the left to ``similar'' on the right,
and participants had to move a cursor along this continuous scale. Listeners
could locate the cursor anywhere on this scale between the two extremities. The
experiment lasted about 1 hour.
1126
 Figure 2. Geometrical solution for the 27 excerpts, resulting from MDS as a function of musical expertise and duration for the
musicians (Left) and nonmusicians (Right) in the 1 s (Top) and 30 s condition (Bottom). The main outcomes of the cluster analysis are
superimposed on the MDS space, with dashed lines representing the first two main clusters, and solid lines, the two subordinates
clusters.

1127
1128 BIGAND ET AL.

Results

For each listener, the emotional judgements were mapped on to an individual

dissimilarity matrix. The average matrix for musicians and nonmusicians was
significantly correlated, r(349) = .86, p < .001. Moreover, the matrices obtained
in Experiment 2 in the 1 s condition were correlated with those of Experiment 3
for both musicians, r(349) = .71, p < .001, and nonmusicians, r(349) = .73, p <
.001. The dissimilarity matrices of Experiment 3 were also significantly corre-
lated with the matrices found in the 30 s condition of Experiment 2 for both
musicians, r(349) = .65, p < .001, and nonmusicians, r(349) = .65, p < .001.
The dissimilarity matrix of each group was analysed using MDS and cluster
analysis methods. The resulting geometrical solutions are illustrated in Figure 3,
with the main outcome of the cluster analysis superimposed on the MDS
representation. The MDS replicates the 3-dimensional solution previously
found. As already reported, the first dimension contrasts pieces of different
arousal, the second dimension those of different emotional values. The third
dimension tended to contrast pieces that started with a melodic interval from
those that started with a chord.
The positions of excerpts in the geometrical space were generally consistent
with those reported in Experiment 2. Some differences were nevertheless
highlighted by the cluster analysis. A careful comparison of Figures 2 and 3
shows that four main clusters of excerpts, rather than three, could now be
distinguished for both musicians and nonmusicians. For musicians, the cluster
comprising excerpts 2, 11, 10, 13, 14, 15, 22, 23, and 24 (i.e., high arousal and
positive valence) in Figure 2 was split into two clusters in Figure 3: one cluster
with the excerpts number (11, 12, 13, 15); and another with the excerpts number
(10, 14, 22, 23, 24). This suggests that the ability of musicians to differentiate
the valence of excerpts high in arousal was better expressed in Experiment 3
than in Experiment 2. The analysis of nonmusicians' data leads to a similar
conclusion. In Experiment 3, three main clusters of excerpts with high arousal
were revealed by the cluster analysis. They contrasted excerpts of negative,
neutral, and positive emotional valence. Interestingly, the cluster of low arousal
excerpts was unchanged for nonmusicians between Experiments 3 and 2.

Discussion

Experiment 3 replicated the data found in the 1 s condition of Experiment 2

using a different experimental procedure. Compared to Experiment 2, the only
noticeable difference was that participants were better able to differentiate the
emotional valence of high arousal excerpts. This suggests that the method used
in Experiment 3 was, at best, marginally more sensitive than that used in
Experiments 1 and 2.
Figure 3. Geometrical solution for the 27 excerpts, resulting from MDS as a function of musical
expertise in Experiment 3, with musicians at the top and nonmusicians at the bottom. The main
outcomes from the cluster analysis are superimposed on the MDS space with dashed lines repre-
senting the first two main clusters, and solid lines, the two subordinates clusters.

1129
1130 BIGAND ET AL.

GENERAL DISCUSSION

The present study investigated emotional responses to musical stimuli. First, the
data confirm the strong consistency of emotional response between participants
musicians and nonmusicians (Experiment 1) as well as between participants
(Experiments 1 and 2). This stability was replicated even when the method
involved a substantial change. The fact that this finding was obtained in an
experimental task that explicitly encouraged participants to focus on their own
emotional experience suggests that emotional responses to music are not subject
to strong individual differences, and are reproducible within and between
participants.
A second important feature of the current study is the provision of evidence
that categorising musical excerpts on the basis of emotional experience trans-
cends the usual criteria of musical categorisation, such as surface similarities,
musical instrumentation, and musical style. As illustrated in Figure 1, some
excerpts that were very close in the emotional space differ considerably in
instrumentation and musical style. This is the case, for example, for excerpts 4
and 20 that correspond to a flute duo by W.-F. Bach and an orchestral piece by J.
Brahms. Examples of this sort are numerous. They suggest that emotional
experience may correspond to the most abstract level of musical categorisation.
The next important issue of the study has been to highlight the structure of the
emotions induced by serious Western music. Two main approaches have been
distinguished for musical emotions (Sloboda & Juslin, 2001). According to the
categorical approach, listeners experience emotions as basic categories, such as
``happiness'', ``serenity'' ``sadness'', and ``anger or fear''. The present data
confirm the relevance of these categories and further demonstrate that they are
unlikely to be a by-product of linguistic labelling. However, the data also
provide evidence showing that the emotional experience of participants was not
restricted to these four basic categories. First of all, participants systematically
grouped the 27 excerpts into more than four groups, and this grouping was very
consistent within participants (Experiment 1). Second, the geometrical solutions
displayed in Figures 1 and 2 underline that musical emotion cannot be fully
accounted for by categories, and argue in favour of a multidimensional
approach. The present data notably suggest that it is possible to manipulate
emotions perceived in music in a continuous way along the one or the other
dimension while keeping the other constant. For example, playing the excerpts
9, 7, 8, 3, 1, and 6, results in a continuous shift from depressed towards serene
emotions, while maintaining an almost constant rate of arousal. In a same way,
playing excerpts 1, 6, 5, 20, 23, 22, 14, 24, and 15, results in a continuous shift
from serenity to jubilation, without changing the emotional valence. Finally, the
present study revealed a third dimension that may link perceived emotions with
body posture and gestures. This finding is consistent with the kinematics
dimension found by Canazza et al. (2001), and more generally with the
 EMOTIONAL RESPONSES TO MUSIC 1131

importance of movement for human emotions (see Damasio, 2003; Heider &
Simmel, 1994).
A further intriguing focus of the present data is the time course of emotional
responses. A previous study by Watt and Ash (1998) showed that 3±5 s extracts
of orchestral interludes from Wagner operas were quite sufficient to generate
consistent, and emotionally relevant, categorisations in listeners. Peretz et al.
(2001) went one step further by showing that 250 ms music was enough to
differentiate sadness from happiness. This provocative finding can be partly
explained by the fact that participants were required to perform a forced-choice
decision, contrasting musical pieces that differ by both arousal and valence. The
important finding of the present study was that the emotions perceived in 1 s
musical excerpts were highly similar to those experienced with longer excerpts.
The present data thus extends the findings of Peretz et al. (2001) to a larger set
of complex musical stimuli. They further suggest that extremely short duration
is likely to impair the processing of the emotional valence of musical excerpts of
a low dynamic. At first glance, this finding may sound extremely surprising. In
half of the cases, the first second of our excerpts contained a single chord or
pitch interval (excerpts 1, 3, 4, 7, 8, 9, 10, 17, 21) and sometimes even a single
tone (excerpts 5, 6, 19). It may be argued that participants may have been able to
recognize the emotion contents of stimuli as short as one second without being
actually ``moved'' by such short excerpts (see Gabrielsson, 2001; Scherer &
Zentner 2001, for a discussion of induced vs. perceived emotions). But even in
this case, it remains remarkable that such a short duration suffices for refined
emotional judgements.
The fact that these excerpts generally elicited an accurate response highlights
that 1 second of music contains enough cues for an emotional reaction.
Appendix B reveals that these cues may relate to structural factors, such as key
and mode (major vs. minor), pitch intervals (consonance vs. dissonance),
chordal disposition, orchestration, rhythm, pulsation, metre, and timbre, as well
as performance features (notably amplitude envelopes). Hearing the 1 s excerpts
would convince readers of the considerable importance of performers' cues.
This importance explains why the location of excerpts 5, 6, 19, which contained
a single tone, may have been so appropriate. Performing one event expressively,
even a single note, is clearly enough to prime in listeners the musical expression
of the remainder of the excerpt. The importance of performance for musical
expressivity is well established (see Juslin, 2001; for a review, see Gabrielson
and LindstroÈm, 2001). It is well known that each event of a piece should be
performed as a function of the musical expressivity of the context and as a
function of the musical style of the piece. The constraints of performing Western
classical music are such that highly expert performers follow an extremely long
and intensive training. It seems likely that Western listeners have internalised,
through mere exposure to Western music, the way in which these cues are
usually combined to convey specific emotions. This combination is so codified
1132 BIGAND ET AL.

in Western culture that a very short slice of expressive music could be sufficient
to induce a specific emotion. Emotional responses to short musical excerpts thus
presumably arise from a rapid process that consists in matching available cues
found in musical structure and performance with emotional responses previously
experienced from these cues. Great music lovers are indeed often emotionally
engaged from the very first sound of a performance.
The last striking issue of the present study concerns the role of musical
expertise. The present study confirmed that emotional responses to music do not
strongly depend upon the extent of musical training (Lynchner, 1998). This
finding is somewhat remarkable since the music stimuli of the present study
were of great complexity and were voluntarily chosen by a music theorist to
induce refined emotions. The weak difference between the two groups can be
understood in the light of several current studies on music cognition that
demonstrate the great ability of nonmusicians to process subtle musical struc-
tures of crucial importance in Western music (for a review, see Bigand &
Poulin-Charonnat, in press). As long as the experimental task involves an
implicit level of processing, nonmusicians perform similarly to musical experts
exactly as they did in the present study.
Manuscript received 3 October 2003
Revised manuscript received 25 May 2005

REFERENCES
Abelson, R. P., & Sermat, V. (1962). Multidimensional scaling of facial expressions. Journal of
Experimental Psychology, 63, 546±554.
Adolphs, R., Tranel, D., Damasio, H., & Damasio, A. R. (1994). Impaired recognition of emotion in
facial expressions following bilateral damage to the human amygdala, Nature, 372, 669±672.
Adolphs, R., Tranel, D., Hamann, S., Young, A. W., Calder, A. J., Phelps, E. A., Anderson, A., Lee,
G. P., & Damasio, A. R. (1999). Recognition of facial emotion in nine individuals with bilateral
amygdala damage. Neuropsychologia, 37, 1111±1117.
Averill, J. (1994). In the eyes of the beholder. In P. Ekman & R. Davidson (Eds.), The nature of
emotion: Fundamental questions (pp. 7±14). Oxford, UK: Oxford University Press.
Bartlett, D. (1999). Physiological responses to music and sound stimuli. In D. Hodges (Ed.),
Handbook of music psychology (pp. 343±85). San Antonio, TX: IMR Press.
Bever, T. (1988). A cognitive theory of emotion and aesthetics in music. Psychomusicology, 7,
165±175.
Bigand, E. & Poulin-Charonnat, B. (in press). Are we all ``experienced listeners''? Cognition.
Blood, A. J., & R. J. Zatorre (2001). ``Intensely pleasurable responses to music correlate with activity
in brain regions implicated in reward and emotion.'' Proceedings of the National Academy of
Sciences, 98, 11818±11823.
Borg, I., & Groenen, P. (1997). Modern multidimensional scaling. New York: Springer.
Canazza, S., De Poli, G., RodaÁ, A., Vidolin, A., & Zanon, P. (2001). Kinematics-Energy space for
expressive interaction in music performance. In Proceedings of MOSART. Workshop on current
research directions in computer music (pp. 35±40). Barcelona, 15±17 November. Retrieved 20
October, 2005, from http://www.iua.upf.es/mtg/mosart/papers/p16.pdf
 EMOTIONAL RESPONSES TO MUSIC 1133
Cunningham, J. G., & Sterling, R. S. (1988). Developmental change in the understanding of affective
meaning in music. Motivation and emotion, 12, 399±413.
Damasio, A. R. (1994). Descartes' error: Emotion, reason and the human brain. Grosset-Putman.
Damasio, A. R. (2003). Looking for Spinoza: Joy, sorrow and the feeling brain. New York:
Harcourt.
Dolgin, K. G., & Adelson, E. H. (1990). Age changes in the ability to interpret affect in sung and
instrumentally-presented melodies. Psychology of Music, 18, 87±98.
Ekman, P. (1994). The nature of emotion: Fundamental questions. New York: Oxford University
Press.
FranceÁs, R. (1988). The perception of music (J. Dolwing, Trans.). Hillsdale, NJ: Erlbaum.
Gabrielsson, A., & Juslin, P. N. (1996). Emotional expression in music performances: Between the
performer's intention and the listener's experience. Psychology of Music, 24, 68±91.
Gabrielsson, A., & Juslin, P. N. (2003). Emotional expression in music. In R. J. Davidson, K.R.
Scherer & H.H. Goldsmith (Eds.), Handbook of affective sciences (pp 503±535). New York:
Oxford University Press.
Gabrielsson, A., & Lindstrom, E. (2001). The influence of musical structure on emotional expression.
In J. A. Sloboda & P. N. Juslin (Eds.), Music and emotion: Theory and research (pp. 223±249).
New York: Oxford University Press.
Gabrielsson, A. (2001). Emotions in strong experiences with music. In J. A. Sloboda & P. N. Juslin
(Eds.), Music and emotion: theory and research (pp. 431±449). New York: Oxford University
Press.
Green, R. S., & Cliff, N. (1975). Multidimensional comparisons of structures of vocally and facially
expressed emotions. Perception and Psychophysics, 17, 429±438.
Heider, F., & Simmel, M. (1994). An experimental study of apparent behavior. American Journal of
Psychology, 57, 243±259.
Hevner, K. (1935). The affective character of the major and minor modes in music. American
Journal of Psychology, 47, 103±118.
Imberty, M. (1979). Entendre la musique; semantique psychologique de la musique. Paris: Dunod.
Imberty, M. (1981). Les eÂcritures du temps, seÂmantique psychologique de la musique: Tome 2.
Psychisme. Paris: Dunod.
Jackendoff, R. (1992). Musical parsing and musical affect. Music Perception, 9, 199±230.
Jackendoff, R., & Lerdahl, F. (in press). The capacity for music: what is it, and what's special about
it? Cognition
JankeÂleÂvitch, V. (1974). Faure et l'inexprimable. Paris: Plon.
JankeÂleÂvitch, V. (2003). Music and the ineffable (C. Abbate, Trans.). Princeton, NJ: Princeton
University Press. (Original work published 1983)
Juslin, P. N. (2001). Communicating emotion in music performance: A review and a theoretical
framework. In J. A. Sloboda & P. N. Juslin (Eds.), Music and emotion: Theory and research (pp.
309±337). New York: Oxford University Press.
Kastner, M. P., & Crowder, R. G. (1990). Perception of the major/minor distinction: Emotional
communication in young children. Music Perception, 8, 189±202.
Khalfa, S., Peretz, I., Blondin, J. P., & Robert, M. (2002). Event-related skin conductance responses
to musical emotions in humans. Neuroscience Letters, 328, 145±149.
Krumhansl, C. L. (1997). An exploratory study of musical emotions and psychophysiology. Cana-
dian Journal of Experimental Psychology, 51, 336±352.
Lane R. D., Reiman E. M., Ahern G. L., Schwartz , G. E., & Davidson, R. J. (1997). Neuroanato-
mical correlates of happiness, sadness, and disgust. American Journal of Psychiatry, 154,
926±933.
LeDoux, J. E. (2000). Emotion circuits in the brain. Annual Review of Neuroscience, 23, 155±184.
Lynchner, J. (1998). An empirical study concerning terminology relating to aesthetic response to
music. Journal of Research in Music Education, 46, 303±320.
1134 BIGAND ET AL.

Madsen, C. K. (1997). Emotional response to music as measured by the two-dimensional CRDI.

Journal of Music Therapy, 34, 187±199.
Mandler, G. (1975). Mind and Emotion. New York: Wiley.
Meyer, L. B. (1956). Emotion and meaning in music. Chicago: Chicago University Press.
Meyer, L. B. (2001). Music and emotion: distinctions and uncertainties. In J. A. Sloboda & P. N.
Juslin (Eds.), Music and emotion: Theory and research (pp. 341±361). New York: Oxford
University Press.
Osgood, C. E. (1966). Dimensionality of the semantic space for communication via facial expres-
sions. Scandinavian Journal of Psychology, 7, 1±30.
Peretz, I. (2001). Listen to the brain: a biological perspective on music and emotion. In J. A. Sloboda
& P. N. Juslin (Eds.), Music and emotion: Theory and research (pp. 105±135). New York:
Oxford University Press.
Peretz, I., Blood, A. J., Penhune, V., & Zatorre, R. (2001). Cortical deafness to dissonance. Brain,
124, 928±940.
Peretz, I., Gagnon, L., & Bouchard, B. (1998). Music and emotion: Perceptual determinants,
immediacy and isolation after brain damage. Cognition, 68, 111±141.
Reiman E., Lane R., Ahern G., Davidson R., & Schwartz G. (2000). Positron emission tomography in
the study of emotion, anxiety and anxiety disorders. In R. Lane (Ed.), Cognitive neuroscience of
emotion (pp. 389±406). New York: Oxford University Press.
Robinson, J. (1997). Music and meaning. Ithaca, NY: Cornell University Press.
Russell, J. A. (1978). Evidence of convergent validity on the dimensions of affect. Journal of
Personality and Social Psychology, 36, 1152±1168.
Russell, J. A. (1980). A circumplex model of affect. Journal of Personality and Social Psychology,
39, 1161±1178.
Scherer, K. R. (1994). Affect bursts. In S. van Goozen, N. E. van de Poll, & J. A. Sergeant (Eds.),
Emotions: Essays on emotion theory (pp. 161±196). Hillsdale, NJ: Erlbaum.
Scherer, K. R., & Zentner, M.R. (2001). Emotional effects of music productions rules. In J. A.
Sloboda & P. N. Juslin (Eds.), Music and emotion: Theory and research (pp. 361±392). New
York: Oxford University Press.
Schmidt, L. A., & Trainor, L. J. (2001). Frontal brain electrical activity distinguishes valence and
intensity of musical emotions. Cognition and Emotion, 25, 487±500.
Schubert, E. (1996). Enjoyment of negative emotions in music: an associative network explanation.
Psychology of music, 24, 18±28.
Schubert, E. (2001). Continuous measurement of self-report emotional response to music. In J. A.
Sloboda & P. N. Juslin (Eds.), Music and emotion: Theory and research (pp. 393±414). New
York: Oxford University Press.
Schubert, E. (2004). Modeling perceived emotion with continuous musical features. Music Per-
ception, 21, 561±585.
Sloboda, J. A. (1991). Music structure and emotional response: Some empirical findings. Psychology
of music, 19, 110-120.
Sloboda, J. A. (1992). Empirical studies of emotional response to music. In M. Riess Jones (Ed.),
Cognitive bases of musical communication (pp. 33±46). Washington: American Psychological
Association.
Sloboda, J. A., & Juslin, P. N. (2001). Music and emotion: Theory and Research. New York: Oxford
University Press.
Smith,C. A., & Ellsworth, P. C. (1985). Patterns of cognitive appraisal in emotion. Journal of
Personality and Social Psychology, 48, 813±838.
Terwogt, M. M., & Van Grinsven, F. (1991). Musical expression of moodstates. Psychology of
music, 19, 99±109.
Tillmann, B., Bharucha, J., & Bigand, E. (2000). Implicit learning of tonality: A self-organizing
approach. Psychological Review, 107, 885±913.
 EMOTIONAL RESPONSES TO MUSIC 1135
Tsang, C., Trainor, L., Santesso, D., Tasker, S., & Schmidt, L. (2001). Frontal EEG responses as a
function of affective musical features. In R. Zatorre & I. Peretz (Eds.) The biological foundations
of music (pp. 439±442). New York: Annals of the New York Academy of Sciences.
Trainor, L. J., & Schmidt L. A. (2003). Processing emotion induced by music. In Peretz, I. & Zatorre,
R. (Eds.), The cognitive neuroscience of music (pp. 310±324). Oxford, UK: Oxford University
Press.
Watt, R. J, & Ash, R. L. (1998). A psychological investigation of meaning in music, Musicae
Scientae, 2, 33±54.
Wedin, L. (1969). Dimension analysis of emotional expression in music. Swedish Journal Musi-
cology, 51, 119±140.
Wedin, L. (1972a). Multidimensional scaling of emotional expression in music. Swedish Journal of
Musicology, 51, 1±43.
Wedin, L. (1972b). A multidimensional study of perceptual-emotional qualities in music, Scandi-
navian Journal of Psychology, 13, 241±257.
Winsberg S., & Carroll, D. (1989). A quasi non-metric method for multidimensional scaling via an
extended Euclidian model. Psychometrika, 54, 217±229.
Zajonc, R. B. (1984). On the primacy of affect. American Psychologist, 39, 117±123.

APPENDIX A
Experiment 1: 27 music excerpts used as stimuli

1. R. Strauss. Also sprach Zarathustra, op. 30. Long, slow, ascending melody in a major key with
consonant harmonies, characterised by a warm string timbre and a full sound resulting from the
arrangement of parts. Some local tensions which are immediately resolved.
2. J.S.Bach. Violin Sonata 3, Fuga. Long melody in a major key for solo violin, played at a moderate
tempo, with a polyphonic feel. Dissonances are neutralised by great rhythmic stability and pre-
dictable patterns of voice leading.
3. W.A.Mozart. Piano Concerto Adagio, K 488. Theme in a minor key, played at a very slow tempo.
Melancholic trochaic rhythm characterised by a large intervallic distance between sounds grouped by
the left hand, and the melody in the high register of the right hand, creating avoid in the middle of the
range which reinforces the desolate aspect of the theme.
4. J. Brahms. Violin Concerto, Adagio. Thematic exposition on the oboe of a slow, pure melodic
line, built on the tonic major chord, and standing apart above a timbrally rich, sustained orchestra.
Harmonic expanses of perfect major chords are spread across the orchestra. The doubling of lines
serves to reinforce the fullness of sound of the whole.
5. D. Scarlatti. Sonata A for Harpsichord, K208. Ascending theme comprising melodic contour with
a very simple rhythm, played at a slow tempo with the soft and intimate timbre of the guitar, above
strongly anticipated major chord harmonies that resonate fully in the instrument.
6. R. Schumann. TraÈumerei, op. 15, no. 7. Theme with a very pure melodic contour, built on a
perfect major chord, played tenderly at a slow tempo by the right hand of the piano, accompanied in
the left hand by calm major chords, sometimes arpeggiated, which fill out the tessitura and whose
consonance resonates fully. Softness of parallel movements a tenth apart.
7. D, Shostakovitch. Symphony 15, Adagio. Brass chorale on a very slow homorhythmic sequence of
chords in a minor key at the bottom of the register, animated by some chromatic embellishment and
melodic movement. Dark atmosphere.
1136 BIGAND ET AL.

8. D. Shostakovitch. Trio 2 for piano, violin, and cello, Largo. Long melody for violin, played at a
slow tempo above low notes on the piano, to which the cello replies with false imitations that are
temporally out of phase, and create strong expressive dissonances. The heartbreaking character of
these dissonances is reinforced by the feeling of emptiness produced by the intervallic distance
between the two parts.
9. R. Wagner. Tristan, Act 3. Declamation in the low register of the strings of the orchestra. Very
strong harmonic tension within a minor key with on the 6th chord against a dissonant second. Slow
and dilated tempo. The upper parts ascend in pitch by chromatic movement, with unresolved
intervallic tensions. The absence of a bass creates a feeling of vertigo and of ascension into infinity.
10. J. Brahms. Trio, piano, violon, and horn, mvt 2. Repetition of a thematic rhythmic motif, above
major key harmony, punctuated by brass effects, at a rapid tempo and with a very rich sound. The
sonority of the French horn enriches the timbral quality of the ensemble, and the structure of the
piece is reinforced by the presence of transposed harmonic progressions.
11. F. Liszt. Les PreÂludes. Powerful orchestral tutti presents consonant harmonies with a martial
rhythm at a solemn tempo, proceding by movements of a third and fifth, played with full sonorities,
doubled by the brass in the bass.
12. S. Prokovief. Sonata for piano, no. 3, op. 28. Highly agitated piano line, with aggressive attacks
and wild sonorities, pounded in the left hand, developing at a rapid tempo above an irregular rhythm,
with juxtaposed harmonies that are neither prepared nor resolved.
13. L. Beethoven. Symphony 7, Vivace. Powerful orchestral tutti develops a rhythmic ostinato
doubled by the timpani, based on a consonant harmony above which a melody is constructed from
the tonic major. Great timbral richness. Joy, colossal energy.
14. L. Beethoven. Piano, Sonata 32, mvt 2. Dynamic motif, presented at a fast tempo on a syncopated
rhythm evocative of jazz, and constructed around major tonic and dominant chords. Great energy.
15. F. Mendelssohn. Italian Symphony, mvt 1. Long melodic line stated by the violins with vigour
and lightness. Major key harmonies built on tonic, dominant and subdominant chords, played at a
high tempo with a pulsating rhythm, supported by a semiquaver ostinato in the rest of the orchestra.
16. F. Chopin Prelude 22. Motif in the low register of the piano repeated obsessively and char-
acterised by pounding octaves in the left hand, dissonant harmonies, and accompanied in the right
hand by a panting rhythm, accentuating the weak part of the beat, and breaking up the violent and
hopeless discourse of the left hand.
17. F. Liszt. Tasso Lamento & Triomfo. Powerful orchestral line develops tense minor harmonies on
a choppy rhythm and at a rapid tempo, supported by the entry of the percussion.
18. R. Strauss. Tod and VerklaÈrung, 7'-7'30. Powerful orchestral line, presenting dissonant harmony
above a long pedal, and unresolved chromatic movements starting from the extremes of the register
and progressing in contrary motion to result in an out of phase rhythmic pounding in the bass drum
and brass.
19. J. S Bach Violin Sonata. Minor key melodic theme at a moderate tempo. Great metric stability,
based on part imitation and dissonances which are always resolved.
20. W.-F. Bach. Duetto for two flutes in F, Lamentabile. Repeated melodic figures built on the tonic
major chord, containing tensions and dissonances that are not always resolved immediately. Slow
tempo and very soft timbre.
21. J. Haydn. Symphony Bdur, Hob I 105, Andante. Thematic motif presented in thirds on the violin
and bassoon, played at a moderate tempo, centred on notes from the tonic and dominant major
chords.
 EMOTIONAL RESPONSES TO MUSIC 1137
22. W.-F. Bach. Duetto for two flutes in G, Allegro. Duo of flutes imitate each other over broken
arpeggios of tonic and dominant major chords. Very soft timbre though a leaping rhythm and a lively
tempo.
23. F. Schubert. Valse no. 3, op. 50, D779. Dynamic melodic motif played in the top register of the
piano. Regular waltz rhythm and a lively tempo over simple tonic/dominant harmonies.
24. I. Stravinsky. Petrouchka. Repeated melodic pattern of ascending 4ths presented with great
momentum and harshness by the flute, to which the strings reply in modal harmonies, all above the
incessant swarming of the orchestra, evoking a jubilant crowd.
25. A. Schoenberg. Erwartung. Dissonant rhythmic and melodic ostinato in the strings, flutes and
bassoons dominated by the strident entry of the piccolo, clarinet, and muted trumpet. Acceleration of
the already lively tempo, and large crescendo culminating in a violent punctuation from the
orchestra.
26. D. Shostakovitch. Trio 2 for violin, cello, and piano, moderato. Series of snatched intervals in the
strings, very dissonant with a grating sonority, punctuated over an agitated and irregular rhythm by
tense chords in the piano. Continued by a dissonant chromatic ascent in the violin and cello over a
harmonic suspension in the high register of the piano.
27. F. Liszt. Totentanz. Aggressive march; pounding dissonant diminished seventh chords in the low
register of the piano, above which the dies irae theme is presented, strongly out of metrical phase,
supported by the bass drum.

Diagram 1. Bootstrap analysis of Experiment 1

1138 BIGAND ET AL.

APPENDIX B
Experiment 2: Features in the first second of the 27 music excerpts

1. R. Strauss. Also sprach Zarathustra. Long sustained texture in the strings with the fifth apparent,
suggesting the beginning of a very slow melodic movement. Ambiguous emotional valence.
2. J. S. Bach. Violin Sonata 3, Fuga. Single high note repeated three times giving a bouncing
character marked by differences of attack and timbre to create a metrical pulsation.
3. W. A. Mozart. Piano Concerto, Adagio, K 488. A high, perfect minor chord with a soft attack and
wide spacing between the third and the tonic. The long resonance heralds a very slow movement.
4. J. Brahms. Violin Concerto, Adagio. A perfect major chord played by the orchestra over which a
high oboe note stands apart. This is ``sung'' and the attack indicates the start of a long, slow, and
serene melody.
5. D. Scarlatti, Sonata A for Harpsichord, K208. A single guitar note played softly and with a
sonorous resonance in the middle of the tessitura, foreshadowing a broad melodic movement with a
regular rhythm.
6. R. Schumann. TraÈumerei, op. 15, no. 7. A single piano note in the middle high register, with a
very soft attack and with full resonance predicting the start of a very slow movement.
7. D. Shostakovitch. Symphonie 15, Adagio. An interval of a minor third held by the brass in a dark,
low register, with an amplitude enveloppe that anticipates a slow movement.
8. D. Shostakovitch, trio 2 for piano, violin, and cello, Largo. A single, held, dissonant harmony
resonant in the low register of the strings, with an amplitude enveloppe that foreshadows a long
musical phrase.
9. R. Wagner. Tristan, Act 3. Declamation of a minor chord in the low register of the strings of the
orchestra.
10. J. Brahms. Trio, piano, violin and horn, mvt 2. A strident major chord, with the arrangement of
the parts and the timbre of the French horn confering a very full sonority.
11. F. Liszt. Poeme symphonique. Powerful orchestral tutti on a major chord, with a triumphant
character marked by the strong trumpet presence, and an amplitude envelope which grows drama-
tically.
12. S. Prokovief. Sonata for piano, no. 3, op. 28. Complex melodic line of nine notes in contrary
motion at a lively tempo, marked by aggressive attacks and many dissonances.
13. L. Beethoven. Symphony. 7, Vivace. Strong ascending melodic motif in the strings leading
towards a powerful orchestral tutti on a perfect major chord.
14. L. Beethoven. Piano, Sonata 32, mvt 2. Energetic melodic line of seven arpeggiated notes of a
major chord in a repeated compound rhythm announcing a rapid metrical beat.
15. F. Mendelssohn. Italian Symphony, mvt 1. Strong major third motif played in the violins with
accents that indicate a rapid rhythmic pulse, confirmed by the semiquaver ostinato played on a tonic
chord by the rest of the orchestra.
16. F. Chopin. Prelude 22. Melody of four piano notes centred on a minor third, strengthened in the
low register of the piano. Strong attack on a rhythm that indicates a fast tempo.
17. F. Liszt. Tasso Lamento & Triomfo. Biting explosion of an orchestral tutti on a dissonant chord,
strengthened by the percussion in a clearly romantic style of orchestration.
 EMOTIONAL RESPONSES TO MUSIC 1139
18. R. Strauss. Tod and VerklaÈrung, 7'-7'30. Two minor key chords played with force in the low
registers of the orchestra, the first reinforced by low percussion instruments and the second by a
bright brass entry.
19. J. S Bach. Violin Sonata. A single high note on the violin, supported but with little vibrato and a
very bright timbre, whose amplitude envelope announces a determined musical gesture.
20. W.-F. Bach. Duetto for two flutes in F, Lamentabile. Two detached flute notes with slight
sustain, and a very soft timbre. Short long rhythm announces a moderate beat and rhythm.
21. J. Haydn. Symphony Bdur, Hob I 105, Andante. Minor third interval from a major chord, played
by the violin and bassoon, which gives a soft, pastoral feel. The dynamic evolution announces a
moderate tempo.
22. W.-F. Bach. Duetto for two flutes in G, Allegro. Five arpeggiated flute notes from a perfect major
chord with a soft timbre and a swift, regular metre.
23. F. Schubert. Valse no. 3, op. 50, D779. Five high piano notes pick out a perfect major chord in a
triple rhythm, whose articulation indicates the style and character of the rest of the piece.
24. I. Stravinsky. Petrouchka. Four high flute notes, in ascending fourths, separate above a dynamic
string texture with accents pointing to a fast rhythmic pulse.
25 A. Schoenberg. Erwartung. Orchestral ostinato on 5 notes marks a swift tempo, played with
shortened attack and resonance to create a breathless character with very tense pitch intervals.
26. D. Shostakovitch. Trio 2 for violin, cello, and piano, moderato. Three snatched intervals in the
strings, containing many high and very dissonant harmonics, separating themselves from a low piano
note with a rapid tempo.
27. F. Liszt. Totentanz. Aggressive march played in the low register of the piano, in a minor
harmony, with a mechanical character indicating a very stable, rapid tempo.