Transcript Stephens, C.L., Knepp, M.M. & Friedman, B.H. (May 2007). Empirical

Empirical Validation of Music Excerpts for
Investigations of Emotion and Emotion Elicitation
Chad L. Stephens, Michael M. Knepp, & Bruce H. Friedman
Introduction
•Music is an interesting and oft-explored method of
emotion elicitation in psychophysiological studies of
affect (e.g. Etzel et al, 2006; Krumhansl et al, 1997; Ng & Eich,
2007; Nyklicek et al., 1997)
• Music induction may increase the realism of an
emotional experience supplementary to, or independent
of, other elicitation approaches (Ng & Eich, 2007).
•Although music is an accepted form of emotion
elicitation, film is the predominant media used in emotion
research (for review see Rottenberg et al, 2007).
•Preference for film induction stems from the intrinsic
nature of film to create a dynamic, artificial version of
reality (Gross & Levenson, 1995).
•Films are advantageous due to the availability of
standardized film sets covering a broad range of emotions
(Gross & Levenson, 1995; Philippot 1993; Rottenberg et
al, 2007). However, no such standard exists for music
induction.
•Procedure: Subjects came to the Mind-Body Laboratory and were
presented with 37 excerpts of classical music/orchestral scores/white
noise/silence.
•The music excerpts/sounds (range = 70 to 240 seconds, average
length = 165 seconds) were presented on a desktop computer
through headphones (loudness = 50-90 dB) using a PowerPoint
presentation such that the procedure was self-paced.
•Dependent measures: Following the presentation of each musical
piece/sound the subject rated on a 7-point Likert scale, from not at
all (1) to very accurately (7) how much they felt each discrete
emotion (amusement, fear, anger, sadness, surprise, contentment,
neutral, and indifferent) as well as dimensional emotion categories
during the presentation.
Figure 3. Group centroids of emotion condition using dimensional ASR variables.
The horizontal axis (discriminant function 2) is interpreted as valence and the
vertical axis (discriminant function 1) as arousal.
•The excerpts were also rated on intensity and enjoyment using a 7point Likert-scale ranging from not at all (1) to very high (7) and
familiarity with musical piece (yes or no).
Results
•Although a standardized set of film excerpts covering a
wide-range of emotions has been created, no such
standard exists for music induction.
•The purpose of the current study was to empirically
validate a set of music excerpts such that there are at least
two musical pieces which reliably induce each of the six
discrete emotions, a neutral state, and a washout effect.
Method
Figure 4. An empirically validated set of music excerpts for the discrete emotions of
amusement, anger, contentment, fear, sadness, surprise, neutral, and washout.
Figure 1. The scores on all of the self-report variables differed significantly between the
discrete emotion categories: Hotelling’s F (df = 7,55) ranging from 11.08 for indifferent to
52.3 for amusement (all Ps <.001).
•Music Selection: All music clips were orchestral string or
piano compositions including film scores
Discussion
•Each emotion category (as represented by two music
excerpts) were distinct from all other emotion categories
based on discrete and dimensional self-report items (Figs. 1 &
2)
•Several excerpts had previously been used to elicit
happiness, sadness, serenity, agitation, anger, fear, or a
relatively neutral state
•Standardizing the self-reported dimensional scores of
valence and arousal allowed for mapping the discrete emotion
conditions into previously hypothesized state-space (Figure 3)
(cf. Berntson et al., 1991; Nyklicek et al., 1997).
•171 Virginia Tech undergraduates completed the
screening phase of this study.
•Ultimately, a set of empirically validated music excerpts
equivalent to the extant film set was created (Figure 4)
•72 undergraduates met the inclusion criteria of being
non-smokers, non-depressed, BDI-II < 19 (Beck et al.,
1996) and non-alexthymic, TAS-20 < 51 (Taylor et al,
1997).
•These results are preliminary to a larger project investigating
autonomic specificity of emotions elicited using both music
and film excerpts.
Correspondence to: [email protected]
Figure 2. The scores on all of the self-report variables differed significantly between the
dimensional emotion categories: Hotelling’s F (df = 7,55) ranging from 21.32 for bad to 59.82
for excited (all Ps <.001).
Presented at the 19th Annual Meeting of the Association for Psychological Science, Washington, DC, May 2007