Acoustic intensity causes perceived changes in arousal levels in music: an experimental investigation.
ABSTRACT Listener perceptions of changes in the arousal expressed by classical music have been found to correlate with changes in sound intensity/loudness over time. This study manipulated the intensity profiles of different pieces of music in order to test the causal nature of this relationship. Listeners (N = 38) continuously rated their perceptions of the arousal expressed by each piece. An extract from Dvorak's Slavonic Dance Opus 46 No 1 was used to create a variant in which the direction of change in intensity was inverted, while other features were retained. Even though it was only intensity that was inverted, perceived arousal was also inverted. The original intensity profile was also superimposed on three new pieces of music. The time variation in the perceived arousal of all pieces was similar to their intensity profile. Time series analyses revealed that intensity variation was a major influence on the arousal perception in all pieces, in spite of their stylistic diversity.
- SourceAvailable from: Kirk N. Olsen[Show abstract] [Hide abstract]
ABSTRACT: In real-world listening domains such as speech and music, acoustic intensity and perceived loudness are dynamic and continuously changing through time. The percept of loudness change in response to continuous increases (up-ramps) and decreases (down-ramps) of intensity has received ongoing empirical and theoretical interest, the result of which has led to conflicting findings from a range of key paradigms. Therefore, the aim of this brief review is to: (a) describe key paradigms used to measure changes in loudness in response to continuous intensity change; (b) identify methodological issues associated with each paradigm; and (c) discuss the mechanisms proposed to explain differences in loudness change when methodological constraints and response biases are controlled. It is concluded that direct and indirect measures of loudness change reflect two distinct aspects of auditory perception. Specifically, magnitude estimation and continuous loudness paradigms reflect changes in perception associated with a ramp’s direction and magnitude of intensity change, and empirical evidence supports the conclusion that greater loudness change in response to down-ramps relative to up-ramps is the real-time perceptual outcome. On the other hand, retrospective global judgements of loudness change are disproportionally weighted on end-level intensity rather than magnitude of intensity change. However, an up-ramp-specific effect of duration on global loudness change is evident when end-level response bias is controlled, and this may be associated with end-point time-of-arrival responses to real and apparent looming auditory motion.Acoustics Australia / Australian Acoustical Society 12/2014; 42. · 0.36 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: The aim of this work was to investigate perceived loudness change in response to melodies that increase (up-ramp) or decrease (down-ramp) in acoustic intensity, and the interaction with other musical factors such as melodic contour, tempo, and tonality (tonal/atonal). A within-subjects design manipulated direction of linear intensity change (up-ramp, down-ramp), melodic contour (ascending, descending), tempo, and tonality, using single ramp trials and paired ramp trials, where single up-ramps and down-ramps were assembled to create continuous up-ramp/down-ramp or down-ramp/up-ramp pairs. Twenty-nine (Exp 1) and thirty-six (Exp 2) participants rated loudness continuously in response to trials with monophonic 13-note piano melodies lasting either 6.4 s or 12 s. Linear correlation coefficients >.89 between loudness and time show that time-series loudness responses to dynamic up-ramp and down-ramp melodies are essentially linear across all melodies. Therefore, ‘indirect’ loudness change derived from the difference in loudness at the beginning and end points of the continuous response was calculated. Down-ramps were perceived to change significantly more in loudness than up-ramps in both tonalities and at a relatively slow tempo. Loudness change was also greater for down-ramps presented with a congruent descending melodic contour, relative to an incongruent pairing (down-ramp and ascending melodic contour). No differential effect of intensity ramp/melodic contour congruency was observed for up-ramps. In paired ramp trials assessing the possible impact of ramp context, loudness change in response to up-ramps was significantly greater when preceded by down-ramps, than when not preceded by another ramp. Ramp context did not affect down-ramp perception. The contribution to the fields of music perception and psychoacoustics are discussed in the context of real-time perception of music, principles of music composition, and performance of musical dynamics.Acta Psychologica 05/2014; 149:117-128. · 2.26 Impact Factor
- [Show abstract] [Hide abstract]
ABSTRACT: Experience-based adaptation of emotional responses is an important faculty for cognitive and emotional functioning. Professional musicians represent an ideal model in which to elicit experience-driven changes in the emotional processing domain. The changes of the central representation of emotional arousal due to musical expertise are still largely unknown. The aim of the present study was to investigate the electroencephalogram (EEG) correlates of experience-driven changes in the domain of emotional arousal. Therefore, the differences in perceived (subjective arousal via ratings) and physiologically measured (EEG) arousal between amateur and professional musicians were examined. A total of 15 professional and 19 amateur musicians listened to the first movement of Ludwig van Beethoven's 5th symphony (duration=∼7.4 min), during which a continuous 76-channel EEG was recorded. In a second session, the participants evaluated their emotional arousal during the listening. In a tonic analysis, we examined the average EEG data over the time course of the music piece. For a phasic analysis, a fast Fourier transform was performed and covariance maps of spectral power were computed in association with the subjective arousal ratings. The subjective arousal ratings of the professional musicians were more consistent than those of the amateur musicians. In the tonic EEG analysis, a mid-frontal theta activity was observed in the professionals. In the phasic EEG, the professionals exhibited an increase of posterior alpha, central delta, and beta rhythm during high arousal. Professionals exhibited different and/or more intense patterns of emotional activation when they listened to the music. The results of the present study underscore the impact of music experience on emotional reactions.Neuroscience 03/2014; · 3.33 Impact Factor
Acoustic Intensity Causes Perceived Changes in Arousal
Levels in Music: An Experimental Investigation
Roger T. Dean1, Freya Bailes1*, Emery Schubert2
1MARCS Auditory Laboratories, University of Western Sydney, Milperra, New South Wales, Australia, 2Empirical Musicology Group, The University of New South Wales,
Sydney, New South Wales, Australia
Listener perceptions of changes in the arousal expressed by classical music have been found to correlate with changes in
sound intensity/loudness over time. This study manipulated the intensity profiles of different pieces of music in order to test
the causal nature of this relationship. Listeners (N=38) continuously rated their perceptions of the arousal expressed by
each piece. An extract from Dvorak’s Slavonic Dance Opus 46 No 1 was used to create a variant in which the direction of
change in intensity was inverted, while other features were retained. Even though it was only intensity that was inverted,
perceived arousal was also inverted. The original intensity profile was also superimposed on three new pieces of music. The
time variation in the perceived arousal of all pieces was similar to their intensity profile. Time series analyses revealed that
intensity variation was a major influence on the arousal perception in all pieces, in spite of their stylistic diversity.
Citation: Dean RT, Bailes F, Schubert E (2011) Acoustic Intensity Causes Perceived Changes in Arousal Levels in Music: An Experimental Investigation. PLoS
ONE 6(4): e18591. doi:10.1371/journal.pone.0018591
Editor: Mark W. Greenlee, University of Regensburg, Germany
Received September 14, 2010; Accepted March 13, 2011; Published April 20, 2011
Copyright: ? 2011 Dean et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits
unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The authors have no support or funding to report.
Competing Interests: The authors have declared that no competing interests exist.
* E-mail: firstname.lastname@example.org
A mounting body of scientific research confirms the intuitions of
many that music can be emotionally expressive [1–12]. The
literature suggests that in addition to extramusical associations,
and expectations arising from culturally familiar musical structures
such as harmonic relationships between notes , listeners find
music emotionally arousing. Arousal is one of the two dimensions
in Russell’s circumplex model of emotion , which has been
shown to be widely applicable to listeners’ perceptions of music
[1,3,4,9]. In common with most authors, we here study per-
ceptions of musical expression rather than an induced arousal
response (such as physiological arousal). Expressed arousal is
conceptualised along a scale from active to passive, where, for
example, ‘angry’ and ‘sleepy’ relate to more active and more
passive respectively. This scale can address the energetic and
tension components of arousal .
Listeners’ perceptions of arousal in music seem to be influenced
by variations in the basic acoustic property of sound intensity, with
its perceptual counterpart of loudness [4,5,11]. Bradley & Lang
 argue that appetitive and defensive systems underpin the
expression of emotion through sound. Increases in intensity might
well evoke an aversive response, signalling the approach of danger
. Temporal profiles of loudness correlate with the temporal
profiles of emotional arousal levels that listeners perceive while
listening to music . To test whether this relationship is causal,
we altered the intensity profiles (on which loudness largely depends
) of several pieces without perturbing other musical features: a
Dvorak Slavonic Dance previously studied by Schubert  was
chosen as one of these pieces. If a given piece presented in two
versions differing only in intensity profiles generates a perceived
arousal profile varying strongly with the intensity profile, this
provides direct evidence of a causal relationship specific to
intensity. Such an experiment was undertaken here with the
Dvorak Slavonic Dance. If arousal profiles were similar across very
different pieces displaying the same intensity profile, this would
also support this causality. This was achieved with several
stylistically diverse new compositions. We support the causal
hypothesis here in both respects.
Materials and Methods
from all participants, and the study was approved by the Human
Research Ethics Committee of the University of New South Wales
(Approval No 09 2 006).
The measured intensity profile of part of Dvorak’s Slavonic Dance
Opus 46 No 1  was used to create a new version of the piece with
logarithmic intensities inverted with respect to median dB SPL
(increases become decreases and vice versa – sound file available
from the authors on request). Three extracts from compositions by
the first author were also studied: two minimal process music
computer-piano pieces , one completely tonal, one largely
atonal; and an electroacoustic piece comprising temporal waves of
filtered noise .
Written informed consent was obtained
The Dvorak Slavonic Dance No. 1 in C Major, Opus 46 was from
start to 29180, while the whole piece lasts 39520 in the recording by
the Slovak Philharmonic Orchestra, conducted by Zdenek Kosler
(Naxos CD 8.550008-09). The Dean tonal (Audio S1) and atonal
(Audio S2) extracts were from Mutase (2008), and comprised two
strands, with a repetitive isochronic (5+5+3 eighth notes, each
PLoS ONE | www.plosone.org1April 2011 | Volume 6 | Issue 4 | e18591
occupying 180 ms) melodic pattern together with progressive
probabilistic variation in individual pitches, in each case in keeping
with the tonal or atonal nature. The filtered noise piece was from
soundAffects , an audiovisual performance- and web-piece
(Audio S3). The diverse set of extracts was chosen to exemplify,
besides variations in intensity, rhythmically active tonal music
(Dvorak, Mutase tonal), rhythmically active atonal music (Mutase
atonal), and timbrally rich music (Dvorak, soundAffects). There was
also strong timbral contrast between the pieces: Dvorak being
orchestral (multi-instrument), Mutase being realised on a piano
(single-instrument), and soundAffects using complex noise textures.
The pieces also encompass aspects of both the 19thand 21st
centuries of Western music composition.
38 students (13 female) aged 19–26 (mean
21 yr) undertook the study. Participants had a median Ollen
Musical Sophistication Index  of 226 (range 17–956). They all
reported normal hearing.
Procedure.Sound intensity measures.
used for intensity analyses  and to manipulate intensity
profiles, using ‘intensity tiers’, with minimal concomitant change,
such as virtually unchanged spectral flatness profiles. Only the
Dvorak showed strong instrumental attack envelopes, and these
were slightly perturbed by the intensity transformation as judged
by trained listeners.
Perceived arousal measures.
all five stimuli over Sennheiser HD280 headphones, rating the
perceived arousal of each through time. Stimulus presentation
order was randomized, with the constraint that the two versions of
the Dvorak should not be presented in immediate succession.
Participants were tested individually, in an isolated space. A
modified version of the Schubert 2D-emotion space [3,21] was
installed on a Macintosh MacBook. This programme incorporates
a training phase presenting participants with detailed instructions
Praat v5 was
Participants listened once to
on-screen about the arousal scale, followed by exercises in which
participants rate the perceived arousal of practice stimuli, with
feedback provided. Having satisfied a given accuracy criterion in
rating the perceived arousal of verbal stimuli (e.g. the words
‘angry’, ‘sleepy’), participants preceded to the main listening task.
Listeners continuously rated their perception of the music’s arousal
levels (for the distinction between perceived and induced emotion,
see ) by moving a mouse . They initiated each of the five
trials by placing a cursor in a central box on the computer screen.
Each trial lasted for the duration of the stimulus, i.e. 29180.
The perceived arousal dimension (i.e. how passive or how active
the music seems) ranged from 2100 to +100. Data were sampled
every 250 ms, and subsequently averaged across all participants at
each sampled time point (N=557 time points) in order to produce
an arousal time series.
Figure 1 shows the original and inverted intensity patterns. The
three newly created piece profiles were extremely similar to the
original. Figure 2 shows the arousal profiles for the Dvorak
original and its inverted-intensity versions. It is clear that the
arousal profile is also inverted when the intensity profile is
inverted. Procrustes distances (d: where 0 is superimposable and 1
is maximum separation) were measured in order to quantify the
distances between the various time series data points studied here
(see Gower  for a discussion of Procrustes transformations,
with examples from many scientific fields). We used unrestricted
transformations (which give the minimal distance estimate), and
standardised the variables. As Table 1 shows, after the exchange of
increase and decrease in the inverted version, the Procrustes
distance between the two arousal profiles was small (0.006),
consistent with the mirroring effect of inversion of intensity. The d
Figure 1. Original and Inverted intensity profiles of the Dvorak Slavonic Dance Opus 46 No 1. The dotted blue line represents the intensity
in decibels (Sound Pressure Level) as a function of time in seconds of the original recording of the piece. The solid red line shows the intensity profile
inverted with respect to median decibels (Sound Pressure Level) through time.
Acoustic Intensity and Musical Arousal
PLoS ONE | www.plosone.org2April 2011 | Volume 6 | Issue 4 | e18591
values in Table 1 confirm the close relation of arousal profiles to
intensity profiles both in the original and its intensity inversion.
Given this indication that the intensity profile of the Dvorak
strongly influenced the perceived arousal profile, we next
investigated whether the original profile would similarly influence
perceived arousal in diverse unrelated pieces. Figure 3 shows mean
arousal ratings for the four pieces sharing the original intensity
profile. The arousal profiles were remarkably similar, and all
peaks/troughs corresponded to the intensity profile. Arousal
perceived in the original Dvorak showed larger peaks where the
largest crescendi occur, and it had a larger coefficient of variation
(CV) than the others : 0.14 for the Dvorak original vs. 0.07 for
both the tonal and atonal pieces. (For determining arousal
coefficients of variation, where CV=SD/M, 100 was added to
each mean time series value to make them all positive; determining
CV of the modified series then permits comparisons between
different pieces.) The arousal CV difference may reflect familiarity
with the Dvorak genre, and associated perceptual fluency . Yet
the CV was only 0.08 for the ‘inverted’ arousal profile, possibly
reflecting the introduction of incongruence between intensity and
structural features of the music. It may also be relevant that in the
inverted version, decrescendi occupy more time than crescendi,
and vice versa for the original . Procrustes distances between
the arousal curves, which only share the original intensity profile,
are shown in Table 1. These small d values confirm the similarity
of the four temporal profiles of perceived arousal.
The distances for the arousal/intensity Dvorak relationships
were also small (Table 1). Thus the Procrustes distances between
input and output response profiles were in all cases small and
similar. However, Procrustes calculations disregard the fact that
time series showed autocorrelation: so we undertook elaborate
time series analyses  and confirmed the large predictive power
of the intensity series for the perceived arousal. The analyses were
done in Stata 10. Relationships between the arousal and intensity
time series were assessed with stationarized series, in each case
achieved by taking the first difference series (termed dintensity and
darousal respectively). Vector autoregression (VAR) was used to
test for Granger causality, really an index of correlation: there was
highly significant Granger causality (p,.01) of dintensity upon
darousal in each case. ARIMAX (autoregressive integrated
Figure 2. Mean arousal profiles for the Dvorak in its Original and Inverted forms. Perceived arousal through time was measured on a scale
from 2100 (very passive) to +100 (very active), and averaged across participants at each time point to produce an aggregate arousal response per
piece. Arousal ratings for the original form of the piece are indicated in blue, with the perceived arousal of the inverted intensity form of the Dvorak
marked in red.
Table 1. Procrustes values (d) between the Dvorak Original arousal time series and other stimulus arousal and Dvorak intensity
Dvorak Original Arousal 0.00010.0002 0.00030.0006 0.00030.0004
Acoustic Intensity and Musical Arousal
PLoS ONE | www.plosone.org3April 2011 | Volume 6 | Issue 4 | e18591
moving average analysis with an exogenous variable, intensity) was
undertaken, and in both VAR and ARIMAX highly significant
(p,.001) models with white noise residuals free of autocorrelation
were accepted; model refinement was based on parsimony and the
Akaike Information Criterion. Figure 4 shows the accurate
prediction of arousal in a time series model based solely on
intensity and autoregressive properties.
All the best time series models were nested within that for the
Dvorak original intensity arousal response: where darousal was
well modelled by lags 1–12 of dintensity (where 1 lag=250 ms),
with lags 1–5 of the autoregression (ar), and a moving average (ma)
window of 19. No constant was needed. This model gave a
correlation of forecast with observed data of .86 (Fig. 4 illustrates
this). All of the models for the other series used the same dintensity
lags, and at least lag 1 of the autoregression (ar), while none
required the ma term, and in summary, beyond this their
additional features are shown in Table 2. These models confirmed
the substantial impact of intensity upon perceived arousal for all
the pieces (as did impulse response function analysis, a technique
which is discussed below).
Possible roles of musical expertise.
further aspects of the impact of intensity on perceived arousal,
judged above both by controlled experimental manipulation (the
Dvorak intensity inversion experiment) and by comparative
response pattern analysis (across four very different musical
entities which solely shared an intensity profile). First we
considered the possible influence of musical expertise. For
example, perhaps musicians, or even simply people familiar with
a piece, might have learned responses to structural components
other than intensity. More than 97.2% of participants indicated
that they had some level of familiarity with both the Dvorak
original and its inverted transform. Thus they were familiar at least
We investigated two
with the genre it exemplifies, and so for this piece familiarity
differences seem unlikely to have had an impact.
We investigated a possible role of musical expertise on the basis
of participant OMSI scores. We first divided our participants into
two equal-sized groups, split at the median OMSI. We assessed the
relationship between intensity and arousal separately for the two
groups. The mean time series of the arousal response for the two
groups were virtually superimposable upon each other and upon
the whole group grand average discussed above, both for the
original Dvorak and for its intensity inverted transform. From this,
the influence of intensity on the perception of arousal does not
seem to depend on musical expertise.
The median split in our participant group was at an OMSI of
226. Given that the OMSI is a probability (61000) that an
individual would be judged by a group of musical experts as having
a musical expertise beyond 5 on a scale from zero to 10, a ‘more
musical’ group could be defined as one with OMSI values .500.
Only seven of our participants were in this group (OMSI range
801–956), and so only brief comments about them are warranted,
and the issue of expertise requires further investigation. First the
average Dvorak arousal profile of the musicians was clearlydifferent
from the grand average, showing different levels (and a CV of 0.08),
but the mirroring effect of the inversion was still apparent. The key
parameters of the high OMSI group are shown in Table 3.
The d values confirm that even for the high OMSI group, the
arousal responses were similar to the intensity profiles engendering
them, and to each other, as for the grand average group. Music-
structural features beyond intensity may well influence these more
musical participants, and their response lag structure is another
possible distinctiveness. Time series analysis by ARIMAX, as
above, showed that a significant influence of intensity remained in
the best Dvorak original darousal model for the high OMSI group.
Figure 3. Perceived arousal of the stimuli that feature the Original intensity profile. Mean arousal profiles for the four pieces studied, each
bearing the original intensity profile of the Dvorak. Perceived arousal through time was measured on a scale from 2100 (very passive) to +100 (very
active), and averaged across participants at each time point to produce an aggregate arousal response per piece.
Acoustic Intensity and Musical Arousal
PLoS ONE | www.plosone.org4April 2011 | Volume 6 | Issue 4 | e18591
Differencing gave stationarity, and the model was again nested
within that of the whole participant group grand average described
above, and contained lags 1–4, 7, 8, 10 and 12 of dintensity, and
autoregressive lags 1, 3, and 4. It had a lower correlation of
forecast with observed data (0.43) than that of the grand average
model (0.86). Similarly, the best model for predicting the arousal
response of the high OMSI group with the inverted Dvorak
contained only lags 1, 6, 8, 10 and 12 of intensity, and
autoregressive lags 1, 2, and 4: it had a correlation of forecast
with observed data of 0.35. Thus intensity is still influential upon
the arousal perceived by the high OMSI subgroup, though their
performance is clearly different from the majority of our
Possible roles of loudness perception in the effect of
acoustic intensity on perceived arousal.
literature relating perceived loudness of music (as opposed to
computed intensity or computed
There is little
perceptions such as arousal. Loudness is defined as the
perceptual counterpart of acoustic intensity, and computational
models of loudness have intensity as the key quantitative influence
[26,27]. With few exceptions [28,29] these models relate to
constant sounds, or to short and simple if inconstant sounds (up to
about 30 seconds). Thus we here assessed continuous perceptual
loudness in relation to intensity, and considered its relation to
perceived arousal. As part of a larger study, the loudness responses
of 24 listeners (8 female, mean age 21 years) to the first 65 sec
(only) of the Dvorak original and of the intensity-inverted version
were determined. The temporal resolution of this study was lower
(2 Hz) than used above, in part because loudness perception
improves with increasing duration up to about 300 msec . The
corresponding intensity profiles (down-sampled to 2 Hz), both
for the original and inverted versions of the Dvorak.
As expected, intensity seemed to strongly determine loudness
perception. In other work we assess whether additional aspects of
musical structure can perturb loudness perception, or conversely
whether there is a largely ‘bottom up’ influence of intensity alone.
Our evidence suggests the latter, and so here we considered
Table 2. Piece Specific Features of ARIMAX models.
forecast and observed data
Dvorak inverted intensity
All models shared the following ARIMAX terms: lags 1–12 of dintensity, and lag
1 of autoregression.
Figure 4. Perceived and modelled arousal in the Original Dvorak composition. Time series analysis (ARIMAX) prediction of the arousal
profile of the Dvorak (original form), compared with the measured arousal. The model is statistically highly significant (p=.0000). Arousal ratings are
Table 3. Procrustes values (d) for the Dvorak extracts from
the high OMSI group.
Dvorak Original Arousal 0.00030.0003
Dvorak Inverted Intensity 0.0005
Acoustic Intensity and Musical Arousal
PLoS ONE | www.plosone.org5April 2011 | Volume 6 | Issue 4 | e18591
whether this necessarily means that loudness is the mediator of the
effect of acoustic intensity on arousal. Intensity is experienced
through auditory and proprioceptive routes , even when music
is played through headphones, so it might act directly on the
perception of arousal in a piece, and/or via the influence of
loudness. Intensity may also act more quickly on or in parallel with
perceptions of arousal than on those of loudness; in either case,
apparently direct impacts of intensity on arousal would occur, with
lesser impact of loudness. Time series analysis can suggest which
possibilities are pertinent for empirical investigation in the future.
We therefore investigated Granger causality relationships
between arousal, intensity and loudness. VAR may be done with
the conservative assumption that all variables potentially influence
each other: they are ‘endogenous’ in statistical terms (like
psychological dependent variables). When this is done with the
Dvorak original and inversion series (standardised but undiffer-
enced), intensity is Granger-causal on arousal (p,.000 and
p=.020, respectively). However, loudness is also weakly Grang-
er-causal on arousal in the original only (p=.048). Results were
similar when the assessment was run on the stationarized first
difference series. Thus VAR suggests that intensity acts directly on
perception of the music’s arousal level, with only secondary
mediation by loudness perception.
This was investigated further by conducting a more realistic, less
conservative, VARX in which intensity (lags 1–4) was taken as an
exogenous (independent) variable (X) with loudness and arousal as
endogenous variables. Assessment of the impulse response function
reveals the statistical impact of unitary change in the endogenous
variables on each other, and of the exogenous on the endogenous
variables. With standardised but undifferenced variables, satisfac-
tory models (i.e. with white noise residuals for the model of
arousal, and a very close fit of predicted and data) could be
obtained. Errors in the impulse response function are boot-
strapped. Figure 5 shows that during the first 8 steps (lags) after
any particular starting point in the arousal response, as expected
the arousal at the starting point is an important predictor of itself.
But more importantly, at step one a significant impulse effect of
intensity is observed declining progressively thereafter. The impact
of an exogenous variable is measured as a Dynamic Multiplier, the
effect of a one unit change in an exogenous variable on the
Figure 5. Impulse Response Functions for the dependence of darousal on dintensity and dloudness - Dvorak Original. Vector
Autoregression of the stationarized darousal series for the original Dvorak melody (first 65 sec), with dintensity as an exogenous variable. Variables
were standardised, and the figure shows the effect of a unit increase in each of the predictor variables, for 8 lags (each 0.5 sec) after the increase.
dm=dynamic multiplier, the change in darousal due to one unit change in the exogenous variable. fevd=fractional error variance decomposition,
the proportion of variance of darousal that might be explained by unit change in the predictor endogenous variable. The shaded areas reveal the
95% confidence limits of the responses (estimated by bootstrapping). In an autoregressive system it is to be expected that the response variable,
darousal, will be a good predictor of itself, as shown. dintensity is also a significant influence, while dloudness is not; the overall model is highly
Acoustic Intensity and Musical Arousal
PLoS ONE | www.plosone.org6April 2011 | Volume 6 | Issue 4 | e18591
endogenous variables: so by step 1 a unit change in intensity
creates 0.34 units of change in perceived arousal. The 95%
confidence limits for this impulse do not breach zero until step 4,
and hence it is highly significant. In contrast, the response to
loudness changes is never significantly different from zero. Results
for the corresponding VARX on stationarized (differenced)
variables were closely similar. In the case of the intensity inverted
Dvorak, again it was intensity that provided a statistically
significant impulse on arousal (and it cumulated over more lags
than with the original Dvorak), while loudness did not. The
impulse response functions also confirm clearly the strong impact
of intensity changes on perceived loudness changes.
The inversion of perceived arousal in response to the inversion
of intensity in the Dvorak, with no other changes to the piece,
demonstrates directly that intensity is a powerful influence on
perceived arousal in this case. Furthermore, the shared arousal
pattern of the original Dvorak, and the constructed tonal, atonal,
and noise pieces confirms that intensity powerfully influences
perceived arousal in a wide range of musical contexts. This does
not exclude the possibility that other time-varying musical features
(e.g. tempo ) are also powerful influences on arousal. It will be
interesting to study how perceived arousal varies within pieces that
have very limited intensity change, such as some ambient music,
and what factors influence such perceived arousal.
Our ‘more musical’ participants also show an influence of
intensity on perceptions of arousal, although their performance is
different from that of the majority of participants. Such differences
may also include their learned responses to musical structure.
Perceptual loudness clearly mirrors acoustic intensity very closely,
and hence does not show evidence of learned responses. But it is
intensity rather than perceived loudness that seems to more
directly influence perceived arousal. This requires further focused
Thus our data build upon the existing knowledge of a
correlation between the timing of intensity/computed loudness
and perceived arousal , providing strong evidence that intensity
profiles are a major causal factor upon continuously perceived
arousal in music. Future work might explore whether changes in
musical sound intensity are equally important in causing listeners
to experience or feel arousal, i.e. induce an emotional response .
Arousal is a key dimension in many theories of emotional response
to music , but it is only one, albeit important, component.
Other aspects of musical affect and meaning are relatively poorly
understood, and it is here that other musical features may be most
important . Future research can extend the findings and
method outlined here, to experimentally manipulate, for example,
the spectral profile of music to determine its role in shaping listener
perceptions of the valence, or positive/negative emotions, of music
Mutase (2008), with superimposed intensity profile.
Stimulus based on extract from tonal version of Dean’s
Dean’s Mutase (2008), with superimposed intensity profile.
Stimulus based on extract from atonal version of
(2003) with superimposed intensity profile.
Stimulus based on sound extract from soundAffects
Thank you to Dr. Sam Ferguson and Vivian Shek for data collection and
preparation, and to Johnson Chen for programming assistance.
Conceived and designed the experiments: RTD FB ES. Performed the
experiments: ES. Analyzed the data: RTD FB. Contributed reagents/
materials/analysis tools: RTD. Wrote the paper: FB RTD ES.
1. Witvliet CVO, Vrana SR (2007) Play it again Sam: repeated exposure to
emotionally evocative music polarises liking and smiling responses, and
influences other affective reports, facial EMG, and heart rate. Cognition and
Emotion 21: 3–25.
2. Sloboda JA (1991) Music structure and emotional response: some empirical
findings. Psychology of Music 19: 110–120.
3. Schubert E (2004) Modelling Perceived Emotion with Continuous Musical
Features. Music Perception 21: 561–585.
4. Leman M, Vermeulen V, de Voogdt T, Moelants D, Lesaffre M (2005)
Prediction of musical affect using a combination of acoustic structural cues. J of
New Music Research 34: 39–67.
5. Krumhansl CL (1997) An exploratory study of musical emotions and
psychophysiology. Can J Exp Psychol 51: 336–353.
6. Khalfa S, Peretz I, Blondin J-P, Manon R (2002) Event-related skin conductance
responses to musical emotions in humans. Neuroscience Lett 328: 145–149.
7. Juslin P, Laukka P (2004) Expression, perception and induction of musical
emotions: A review and a questionnaire study of everyday listening. J of New
Music Research 33: 217–238.
8. Dubnov S, McAdams S, Reynolds R (2006) Structural and affective aspects of
music from statistical audio signal analysis. J American Society for Information
Science and Technology 57: 1526–1536.
9. Bigand E, Vieillard S, Madurell F, Marozeau J, Dacquet A (2005) Multidimen-
sional scaling of emotional responses to music: the effect of musical expertise and
of the duration of the excerpts. Cognition and Emotion 19: 1113–1139.
10. Balkwill LL, Thompson WF (1999) A cross-cultural investigation of the
perception of emotion in music: psychophysical and cultural cues. Music
Perception 17: 43–84.
11. Gabrielsson A, Lindstro ¨m E (2010) The role of structure in the musical expression
of emotions. In: Juslin PN, Sloboda J, eds. Handbook of Music and Emotion:
Theory, Research, Applications. Oxford: Oxford University Press. pp 367–
12. Juslin PN, Va ¨stfja ¨ll D (2008) Emotional responses to music: The need to consider
underlying mechanisms. Behavioral and Brain Sciences 31: 559–621.
13. Russell JA (1980) A circumplex model of affect. Journal of Personality and Social
Psychology 39: 1161–1178.
14. Russell JA (2003) Core affect and the psychological construction of emotion.
Psychological Review 110: 145–172.
15. Bradley MM, Lang PJ (2000) Affective reactions to acoustic stimuli.
Psychophysiology 37: 204–215.
16. Neuhoff JG (2001) An adaptive bias in the perception of looming auditory
motion. Ecological Psychology 132: 87–110.
17. Geringer JM (1993) Loudness estimations of noise, synthesizer and music
excerpts by musicians and nonmusicians. Psychomusicology 12: 22–30.
19. Brewster A, Smith H, Dean RT (2003) AFFECTions: friendship, community,
bodies. Text 7: http://www.textjournal.com.au/oct04/smith02.htm.
20. Ollen J (2006) A criterion-related validity test of selected indicators of musical
sophistication using expert ratings. Columbus: Ohio State University. 246 p.
21. Bailes F, Dean RT (2009) Listeners discern affective variation in computer-
generated musical sounds. Perception 38: 1386–1404.
22. Evans P, Schubert E (2008) Relationships between expressed and felt emotions
in music. Musicae Scientiae 12: 75–99.
23. Gower JC, Dijksterhuis GB (2004) Procrustes Problems. Oxford: Oxford
University Press. 248 p.
24. Dean RT, Bailes F (2010) A rise-fall temporal asymmetry of intensity in
composed and improvised electroacoustic music. Organised Sound 15: 147–158.
25. Dean RT, Bailes F (in press) Time series analysis as a method to examine
acoustical influences on real-time perception of music. Empirical Musicology
26. Glasberg BR, Moore BCJ (2002) A model of loudness applicable to time-varying
sounds. Journal of the Audio Engineering Society 50: 331–342.
Acoustic Intensity and Musical Arousal
PLoS ONE | www.plosone.org7April 2011 | Volume 6 | Issue 4 | e18591
27. Zwicker E, Fastl H (1999) Psychoacoustics. Berlin: Springer.
28. Namba S, Kuwano S, Fastl H (2008) Loudness of non-steady-state sounds.
Japanese Psychological Research 50: 154–166.
29. Cabrera D, Ferguson S, Schubert E (2010) Comparing Continuous Subjective
Loudness Responses and Computational Models of Loudness for Temporally
Varying Sounds. 129th Convention of the Audio Engineering Society (AES).
San Francisco, CA, USA.
30. Viemeister NF, Wakefield GH (1991) Temporal integration and multiple looks.
J Acoust Soc Am 90: 858–865.
31. Todd NPM, Cody FW (2000) Vestibular responses to loud dance music: A
physiological basis of the ‘‘rock and roll threshold’’? J Acoust Soc Am 107:
32. Schubert E (2001) Continuous measurement of self-report emotional response to
music. In: Sloboda JA, Juslin PN, eds. Music and Emotion. Oxford: Oxford
University Press. pp 393–414.
Acoustic Intensity and Musical Arousal
PLoS ONE | www.plosone.org8 April 2011 | Volume 6 | Issue 4 | e18591