Emotional Reactions to Sounds without Meaning

doi:10.4236/psych.2012.38091

Paper Menu >>

Journal Menu >>

Psychology

2012. Vol.3, No.8, 606-609

Published Online August 2012 in SciRes (http://www.SciRP.org/journal/psych) http://dx.doi.org/10.4236/psych.2012.38091

606

Emotional Reactions to Sounds without Meaning

Daniel Västfjäll

Department of Behavioural Science and Learning, Linköping University, Linköping, Sweden

Email: [email protected]

Received May 19th, 2012; revised June 18th, 2012; accepted July 16th, 2012

The present research examined the relationship between emotional reactions to sounds without meaning

(tone and noise complexes) and objective sound descriptors. Two experiments showed that the core affect

dimensions valence and activation were related to perceived loudness (intensity) and sharpness (perceived

high frequency content), respectively. These results can be used as design criteria for emotion induction

with sounds, implementation of emotional sounds in products, as well as in research on environmental

noise perception.

Keywords: Emotion; Sound

Introduction

Early research by Wundt suggested that emotional reactions

to auditory events could be mapped onto a pleasantness-un-

pleasantness (Lust-Unlust) dimension (Wundt, 1924). Later stud-

ies on emotional reactions to non-musical, non-vocal sounds

have almost exclusively studied positive-negative responses

(Todd, 2001; Vitz, 1973), even though Wundt himself con-

cluded that emotional reactions to auditory rhythms needed to

be described by additional dimensions (strain-relaxation and

excitement-calmness). More recent studies on the relationship

between emotional responses and sound characteristics have

focused on a single affective state annoyance. Annoyance has

been shown to correlate moderately with descriptors of physical

characteristics such as equivalent dB(A) level for community

noise, and with other psychoacoustic dimensions such as per-

ceived sharpness and roughness for specific sound sources

(Berglund & Lindvall, 1995; Berglund, Hassmén, & Preis,

2002; Zwicker & Fastl, 1999).

Other research have demonstrated that two dimensions, va-

lence and activation, are suited to describe emotional reactions

to sounds (Björk, 1985). Bradley and Lang (2000) found that

self reported emotional reactions to 60 natural sounds were

scattered in a two-dimensional space of valence (pleasantness-

unpleasantness) and activation (arousal) (but see Stevenson &

Jameson, 2008 for a discrete emotional account of these

sounds). Moreover, the reactions were clustered along two axes,

one stretching from low activation and neutral valence to un-

pleasant high activation (avoidance), and the other one from

low activation and neutral valence to pleasant high activation

(approach). Importantly, Bradley and Lang found that valence

ratings was very weakly related to sound level (r = 0.07) and

activation ratings was moderately related (r = 0.38). However,

these correlations only accounted for 14% of the variance. Af-

fective reactions to these sounds must therefore be related to

other physical sound descriptors than sound level and/or other

psychological characteristics not captured by physical descrip-

tors (Asutay & Västfjäll, 2012; Asutay et al., 2012). Bradley

and Lang used recordings of a number of everyday sounds

sources such as recordings of a dog barking, cries from an

amusement park etc., why they could conclude that the ob-

served reactions were due to other aspects of the stimuli (i.e.

affective meaning1).

The present research complements the research by Bradley

and Lang by focusing on physical sound determinants of emo-

tional reactions. Results from such research are important for

many applications including emotion induction with sounds,

assessment of subjective noise reactions, prediction of subjective

noise experience, sound design, auditory interfaces, and devel-

opment of new sound abatement approaches (Desmet, 2002;

Picard, 1997; Västfjäll et al., 2002). A slightly different ap-

proach than that of Bradley and Lang’s is therefore used in the

present research. Rather than using everyday sounds with easily

identified meaning (such as recordings of people, animals, ac-

tivities), the present research focused sounds devoid of emo-

tional meaning (noise and tone complexes) induced through

activation of episodic memories or similar mechanisms (Juslin

& Västfjäll, 2008). Even though everyday emotional reactions

are related to both sound characteristics and the appraisal of the

sound/sound source (Asutay & Västfjäll, 2012; Tajadura, Väl-

jamäe, Asutay, & Västfjäll, 2010, Tajadura, Larsson, Väljamäe,

Västfjäll, & Kleiner, 2010), it may be desirable to first establish

a relationship between sound characteristics and emotional

reactions.

The aim of the present research is therefore to study 1) if

emotional reactions to tone and noise complexes that vary in

the two-dimensional emotional experiential space; 2) to find

self-reported and physical correlates to valence and activation.

A related aim is to show that auditory dimensions other than

loudness or sound level influence emotional reactions to sounds

(Zwicker & Fastl, 1999).

1Affective meaning is defined as an evaluation of auditory events where

the object/activity creating the sound is easily identified (Ballas, 1993),

and that this activity/object is perceived as positive-negative. Furthermore

it is the object/activity, rather that the acoustic information, that creates

the affective reaction

In Experiment 1, participants either rated their emotional re-

actions to 16 tone and noise complexes or rated the perceptual

or psychoacoustic qualities of the same sounds. On basis of the

results of Experiment 1, Experiment 2 used experimentally

D. VÄSTFJÄLL

manipulated sounds to further investigate physical determinants

of valence and activation.

Experiment 1

The first experiment aimed at investigating the relationship

between emotional reactions and sound characteristics. An

exploratory approach was taken where 40 participants rated

their emotional reactions to stationary sounds. Twenty additional

participants rated the sounds in terms of their perceptual prop-

erties.

Participants. Sixty undergraduates at Chalmers University

of Technology, Göteborg, Sweden, an equal number of men

and women, participated on a voluntary basis. They were com-

pensated with the equivalence of US$10. Their mean age was

26.1 years (SD 4.1). All participants had normal hearing as

determined by an audiogram.

Measures. The affect measures consisted of two bipolar

scales each defined by three adjective pairs found in previous

research to tap valence and activation, respectively (Västfjäll et

al., 2002; Västfjäll & Gärling, 2007). Sleepy-awake, dull-peppy,

and passive-active were used to define the activation scale,

displeased-pleased, sad-glad, and depressed-happy were used to

define the valence scale. Numbers ranging from –4 over 0 to 4

were typed beneath the three adjective pairs defining a scale.

Participants were requested to circle the number that cor-

responded to their feeling.

From previous research on auditory event evaluation (Björk,

1985; von Bismark, 1974; Solomon, 1958), a set of sensory

adjectives were identified and selected. These included (trans-

lated to English) hard, strong, low, loud, clear, tonal, even, high

in frequency, low in frequency, rough, soft, regular, irregular,

weak in tone, harsh, natural, artificial, tiring, sharp, edgy, blunt,

strong in tone, presence of extraneous sounds, and balance

between left and right ear.

Stimuli and presentation. Sixteen binaural sounds varying

in psychoacoustical qualities were used. The sounds were tone

and noise complexes and a pilot rating experiment suggested

that they were not systematically identified as a having a specific

meaning. Psychoacoustic metrics were calculated for all sounds

using a HEAD Acoustics Artemis analysis system on a PC. The

sounds varied in psychoacoustical properties such as loudness

(intensity), roughness (frequency or amplitude modulations

between 20 - 70 Hz), sharpness (high frequency components),

fluctuation strength (amplitude and frequency modulations

below 20 Hz), and tonal content (tone-to-noise ratio; see

Zwicker & Fastl, 1999 for an overview of these metrics). The

sounds were presented in an acoustically well-damped room

over Alpine loudspeakers, using cross cancellation technique to

maintain binaural information.

Procedure

Emotion ratings. Participants (n = 40) arrived individually

to the laboratory. After having been seated participants listened

to the first of in all 16 sounds sequentially presented. Each

sound was presented for 2 minutes. Participants were asked to

rate their affective reactions on the adjective scales by checking

each scale to indicate the degree to which the adjective de-

scribed how they felt at the time of listening to the sound.

Following the procedure devised by Bradley and Lang

(2000), participants were instructed to refer how they felt when

listening to the sound. The moment participants were asked to

refer to when rating their affective reactions was indicated by a

“rating probe” consisting of a blinking arrow on a computer

screen. The rating probe was displayed approximately in the

middle of the duration of each sound.

After listening to a sound, participants were instructed relax

and try to return to a neutral affective state. When participants

felt relaxed, they were instructed to rate how they felt on the

two adjective scales. After this they continued with the next

sound by clicking a button on the computer.

Sensory ratings. 20 separate participants rated the sensory

characteristics of the sounds. The procedure was similar to the

affective ratings condition. Sounds were presented in different

random orders for each participant. One minute was allowed

between each sound. Participants could indicate any number

from 0 (not at all) to 8 (very much) for each of the adjectives.

The procedure took in total approximately one hour.

For each participant a different random order of the sounds

was generated. The participants needed about sixty minutes to

complete the ratings. After listening to the sixteen sounds, par-

ticipants were debriefed, compensated and thanked for their

participation.

Results and Discussion

First, emotional reactions were examined. Valence and acti-

vation were relatively independent (r = 0.09, p > 0.05). Figure

1 shows that the sounds induced variations in emotional

reactions in terms of activation and valence. This was sub-

stantiated by a within-subject ANOVA on activation, F(2.12,

49.66) = 12.44, p < 0.001, and on valence, F(2.77, 46.82) =

20.10, p < 0.001.

To assess the relationship between the psychoacoustic met-

rics and emotional reactions multiple regression analyses with

either the mean valence or mean activation index as dependent

variable was performed. The analysis for the activation index

showed that sharpness contributed significantly (β = 0.68, p <

0.01) for an R2adj of 0.42, F(2, 14) = 11.98, p < 0.01.

Figure 1.

Emotional reactions to 16 tone and noise complexes along

the valence and activation dimensions (Experiment 1).

D. VÄSTFJÄLL

For the regression analysis of the valence index the loudness

(β = –0.89, p < 0.001) and roughness (β = 0.38, p < 0.05) met-

rics contributed significantly giving an R2adj of 0.81, F(3, 13) =

33.42, p < 0.01.

To further corroborate these results, sensory ratings from the

separate sample were investigated. A PCA was first performed

on the correlations between the mean ratings over individuals

for each sound to determine the underlying perceptual dimen-

sions. The PCA resulted in five factors with eigenvalues larger

than 1.0, that together accounted for 76% of the variance. Ad-

jectives denoting loudness loaded on the first factor, adjectives

describing mainly sensory sharpness loaded on the second fac-

tor, adjectives describing fluctuation and modulation loaded on

the third factor, adjectives describing tonal content loaded on

the fourth factor, and finally adjectives concerning the natural-

ness vs. artificiality of the sounds loaded on the fifth factor.

From the PCA five indices of loudness, sharpness, fluctuation

strength, tonal content, and naturalness were formed by sum-

ming with the appropriate sign.

Next, multiple regression analyses were performed with each

of the affect indices as dependent variables. Sharpness (β =

0.47, p < 0.05) and tonal content (β = –0.43, p < 0.05) indices

were reliably related to the activation index, R2adj = 0.53, F(3,

13) = 12.85, p < 0.01, indicating that activation increases with

increasing sharpness and decreases with increasing tonal con-

tent. Loudness (β = –0.51, p < 0.01) and naturalness (β = 0.32,

p < 0.05) were reliably related to valence, R2adj = 0.66, F(3, 13)

= 18.83, p < 0.01.

The results of the first experiment showed that tone and noise

complexes varied in both valence and activation. More impor-

tantly, the results suggested that valence and activation reac-

tions differed in their determinants. Activation was related to

rated or perceived tonal content and sharpness, whereas valence

was associated with perceived loudness, roughness, and natu-

ralness.

To further study these relationships, Experiment 2 employed

experimentally manipulated sounds.

Experiment 2

Method

Participants. 16 undergraduates, 7 female and 9 males, vol-

untarily participated in the experiment. Their mean age was

22.3 (SD, 2.12). All reported having normal hearing.

Measures. The affect rating scales from Experiment 1 was

used.

Stimuli and Presentation

Activation manipulation. Since the first experiment showed

that activation was related to perceived sharpness of the sound a

set of five sounds varying from strong low-frequency tonal

content strong to high frequency content was created using

signal processing. A tone-noise complex was used as the base

stimulus (reference sound). From this sound a number of modi-

fied versions were created where the fundamental frequency

and/or harmonics or the noise spectrum were changed. The

modifications for strong tonal low frequency content were an

amplification of the fundamental frequency (100 Hz) of 6 and

12 dB, respectively. To increase the high frequency content of

the sound, a high-pass filter was used to amplify noise and

tones above 3000 Hz with 6 and 12 db, respectively. Finally, all

sounds were equalized to the same loudness level and were

replayed to participants at 60 dBA2.

Valence manipulation. For the valence manipulation only

loudness was changed. The same reference sound that was used

for the activation manipulation was again used, but replayed at

five different sound levels (40, 50, 60, 70, and 80 dBA).

All stimuli were generated using digital signal processing

software and were digitally stored on a computer. Stimulus

presentation was made on computers using an experiment pro-

gram. The sounds were delivered via Stax electrostatic head-

phones.

Procedure. Participants performed the experiment individu-

ally or in groups of maximum three persons at a time. Upon

arrival to the laboratory, participants were first instructed how

to use the scales and equipment. They were instructed that they

would perform ratings of their reactions to various sounds.

Participants then listened to and rated five trial sounds (other

than the test stimuli). After that they rated the five trial sounds,

participants continued with remaining two sound blocks. The

blocks and order within blocks were counterbalanced across

participants. When participants had rated all the sounds, they

were debriefed, compensated, and thanked for their participation.

Results and Discussion

Activation manipulation. The within-subjects ANOVAs for

activation, F(2.31, 34.66) = 46.01, p < 0.001, was as expected

significant. As may be seen in Figure 2, activation increases

with increasing sharpness. The ANOVA for valence was not

significant F(1.44, 21.66) = 1.09, p > 0.05.

Valence manipulation. The within-subjects ANOVA for

valence was significant, F(2.86, 42.96) = 24.12, p < 0.001

(Figure 3). The ANOVA for activation was also significant,

F(1.99, 37.97) = 4.01, p < 0.05. A contrast however showed

that only the 80 dBA vs 40 dBA was significantly different.

In line with experiment 1 and previous findings (Västfjäll et

al., 2002), Experiment 2 yielded support for the idea that va-

lence reactions are mainly affected by loudness and the activa-

tion dimension by perceived sharpness of the sound.

Figure 2.

2A pilot experiment indicated that the perceived sharpness/tonal con-

tent varied as predicted, whereas perceived loudness remained con-

stant.

Valence and activation ratings for sharpness modifications (Experiment

2).

608

D. VÄSTFJÄLL

REFERENCES

Asutay, E., & Västfjäll, D. (2012). Perception of loudness is influenced

by emotion. PLoS ONE, 7, e38660.

doi:10.1371/journal.pone.0038660

Asutay, E., Västfjäll, D., Tajadura-Jimenez, A., Genell, A., Bergman, P.,

& Kleiner, M. (2012). Emoacoustics: A study of the psychoacoustical

and psychological dimensions of emotional sound design. Journal of

the Audio Engineering Society, 60 , 21-28.

Ballas, J. A. (1993). Common factors in the identification of an assort-

ment of brief everyday sounds. Journal of Experimental Psychology:

Human Perception and Performance, 19, 250-267.

doi:10.1037/0096-1523.19.2.250

Berglund, B., Hassmén, P., & Preis, A. (2002). Annoyance and spectral

contrast are cues for similarity and preference of sounds. Journal of

Sound and Vibratio n, 250, 53-64. doi:10.1006/jsvi.2001.3889

Figure 3. Berglund, B., & Lindvall, T. (1995). Community noise. Archives of the

Center for Sensory Res ea rc h, 2. Stockholm: Stockholm University.

Valence and activation ratings for loudness changes (Experiment 2).

Björk, E. A. (1985). The perceived quality of natural sounds. Acustica,

57, 185-188.

General Discussion

Bradley, M. M., & Lang, P. J. (2000). Affective reactions to acoustic

stimuli. Psychophysiology, 37, 204-215.

doi:10.1111/1469-8986.3720204

The present research showed that the valence and activation

dimensions of auditory-induced emotions are related to differ-

ent physical characteristics. Valence is primarily determined by

the perceived loudness and activation by the perceived sharpness

of the sound. These findings are line with other research show-

ing that tonal sounds decrease wakefulness and that sound level

is related to annoyance (see Berglund & Lindvall, 1995 for an

overview). The present research however goes beyond most

research on noise annoyance in showing that it is useful to de-

compose emotional reactions to sounds into the two core affect

dimensions valence and activation (see also Asutay et al., 2012).

Desmet, P. (2002). Designing emotions. Delft: Delft University of

Technology.

Juslin, P. N., & Västfjäll, D. (2008). Emotional responses to music: The

need to consider underlying mechanisms. Behavioral Brain Science,

31, 559-621.

Norman, D. A. (2002). Emotion and design: Attractive things work

better. Interactions Magazine, 9, 36-42.

Picard, R. (1997). Affecti ve computing. Cambridge: MIT Press.

Solomon, L. N. (1958). Semantic approach to the perception of com-

plex sounds. Journal of the Acoustical Society of America, 30, 421-

425. doi:10.1121/1.1909632

The sounds used in the present research were all static, sta-

tionary sounds devoid of meaning that may modulate emotional

experience. Thus, the present research complement previous

research on emotional reactions to everyday natural sounds

(Bradley & Lang, 2000) by showing that physical characteris-

tics may be a good predictor of emotional reactions to sounds

when they carry little affective meaning, but still induce affect

in the listener. It is also possible that physical characteristics

uncovered here may be good predictors of emotional reactions

to sounds that initially carry emotional meaning, but that

through habituation is reduced (like traffic noise). Other re-

search also suggests that objective measures may predict emo-

tional reactions to any set of sounds that do not vary drastically

in meaning (Västfjäll et al., 2002; Asutay et al., 2012). It should

however be noted that for sounds that do vary in emotional

meaning, physical characteristics will likely be much less pre-

dictive of the emotional reaction. In music, for instance, musi-

cal structure is very important to convey emotions, but much

less important for inducing emotion (Juslin & Västfjäll, 2008).

Stevenson, R. A., & James, T. W. (2008). Affective auditory stimuli:

Characterization of the International Affective Digitized Sounds

(IADS) by discrete emotional categories. Behavior Research Methods,

40, 315-321. doi:10.3758/BRM.40.1.315

Tajadura, A., & Västfjäll, D. (2008). Auditory induced-emotion: A

neglected channel for communication in HCI. In Affect and Emotion

in HCI. In C. Peter, & B. Russell (Eds.), Berlin: Springer Verlag.

doi:10.1007/978-3-540-85099-1_6

Tajadura-Jimenez, A., Väljamäe, A., Asutay, E., & Västfjäll, D. (2010).

Embodied auditory perception: The emotional Impact of approaching

and receding sound sources. Emotion, 10, 216-229.

doi:10.1037/a0018422

Tajadura-Jimenez, A., Larsson, P., Väljamäe, A., Västfjäll, D., &

Kleiner, M. (2010). The influence of auditory space on emotional

response to sounds. E motion.

Todd, N. (2001). Evidence for a behavioral significance of saccular

acoustic sensitivity in humans. Journal of the Acoustical Society of

America, 110, 380-390. doi:10.1121/1.1373662

Västfjäll, D. (2002). Emotion induction through music: A review of the

musical mood induction procedure. Musicae Scientiae, 6, 171-203.

Västfjäll, D., & Gärling, T. (2007). Validation of short self-report

measure of core affect. Scandinavian Journal of Psychology, 48,

233-238. doi:10.1111/j.1467-9450.2007.00595.x

The present results can be used as design criteria for emotion

induction studies using sound (Västfjäll, 2002), design of emo-

tive sounds for various applications (Norman, 2002; Tajadura

& Västfjäll, 2008), affective computing (Picard, 1997), as well

as a base for future studies on emotions in noise perception

(Västfjäll et al., 2002). An important task for future research is

to investigate if the relationship between experienced affect and

objective sound characterization holds for a wider range of

sounds and situations (e.g. natural soundscapes).

Västfjäll, D., Gulbol, M.-A., Kleiner, M., & Gärling, T. (2002). Affective

reactions to- and evaluations of interior and exterior vehicle auditory

quality. Journal of Sound and Vibration , 255, 501-518.

doi:10.1006/jsvi.2001.4166

Västfjäll, D., Friman, M., Gärling, T., & Kleiner, M. (2002). The meas-

urement of core affect: A Swedish self-report measure. Scandinavian

Journal of Psychology, 43, 19-32. doi:10.1111/1467-9450.00265

Vitz, P. C. (1973). Preference for tones as a function of frequency and

intensity. Perception and Psychophysics, 11, 84-88.

doi:10.3758/BF03212689

Acknowledgements Wundt, W. (1924). An introduction to psychology. London: Allen &

Unwin.

This research was financially supported by Swedish Council

or Working Life and Social Research. Zwicker, E., & Fastl, H. (1999). Psychoacoustics—Facts and models.

(2nd ed.). Heidelberg: Springer-Verlag.