Professional Documents
Culture Documents
Schaefer 2010
Schaefer 2010
To cite this article: Alexandre Schaefer , Frdric Nils , Xavier Sanchez & Pierre Philippot (2010) Assessing the
effectiveness of a large database of emotion-eliciting films: A new tool for emotion researchers, Cognition &
Emotion, 24:7, 1153-1172, DOI: 10.1080/02699930903274322
Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”)
contained in the publications on our platform. However, Taylor & Francis, our agents, and our
licensors make no representations or warranties whatsoever as to the accuracy, completeness, or
suitability for any purpose of the Content. Any opinions and views expressed in this publication
are the opinions and views of the authors, and are not the views of or endorsed by Taylor &
Francis. The accuracy of the Content should not be relied upon and should be independently
verified with primary sources of information. Taylor and Francis shall not be liable for any
losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities
whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or
arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial
or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or
distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use
can be found at http://www.tandfonline.com/page/terms-and-conditions
COGNITION AND EMOTION
2010, 24 (7), 11531172
Alexandre Schaefer
University of Leeds, Leeds, UK
Frédéric Nils
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
Xavier Sanchez
University of Chester, Chester, UK
Pierre Philippot
University of Louvain, Louvain-La-Neuve, Belgium
Using emotional film clips is one of the most popular and effective methods of emotion elicitation.
The main goal of the present study was to develop and test the effectiveness of a new and
comprehensive set of emotional film excerpts. Fifty film experts were asked to remember specific film
scenes that elicited fear, anger, sadness, disgust, amusement, tenderness, as well as emotionally
neutral scenes. For each emotion, the 10 most frequently mentioned scenes were selected and cut
into film clips. Next, 364 participants viewed the film clips in individual laboratory sessions and rated
each film on multiple dimensions. Results showed that the film clips were effective with regard to
several criteria such as emotional discreteness, arousal, positive and negative affect. Finally, ranking
scores were computed for 24 classification criteria: Subjective arousal, positive and negative affect
(derived from the PANAS; Watson & Tellegen, 1988), a positive and a negative affect scores derived
from the Differential Emotions Scale (DES; Izard et al., 1974), six emotional discreteness scores (for
anger, disgust, sadness, fear, amusement and tenderness), and 15 ‘‘mixed feelings’’ scores assessing
the effectiveness of each film excerpt to produce blends of specific emotions. In addition, a number of
emotionally neutral film clips were also validated. The database and editing instructions to construct
the film clips have been made freely available in a website.
Correspondence should be addressed to: Alexandre Schaefer, University of Leeds, Institute of Psychological Sciences, Leeds
LS2 9JT, UK. E-mail: a.schaefer@leeds.ac.uk
The writing of this article has been facilitated by grants from the ‘‘Fonds National de la Recherche Scientifique’’ (FNRS)
8.4510.99 and 8.4510.00.
The authors thank Deanna Barch, Todd Braver and Jeremy Gray for their helpful comments on an earlier draft of this
manuscript. The authors also wish to thank Sylvie De Coninck, Cécile Misonne and Olivier Beauclercq for their contribution to the
data collection, Pierre Mahau and Richard Robert for their help in recording and editing the film excerpts, and Adam Rickart for
helping with the data analysis.
The database of stimuli can be found at this URL: http://nemo.psp.ucl.ac.be/FilmStim/.
# 2009 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business 1153
http://www.psypress.com/cogemotion DOI:10.1080/02699930903274322
SCHAEFER ET AL.
The increasing interest in the effects of emotion emotional film scenes. The present article aims to
on cognitive, social and neural processes creates a describe the development of this database, includ-
constant need for efficient and reliable techniques ing the tests that attested its efficiency in eliciting
of emotion elicitation. Many emotion-elicitation emotions in a laboratory context.
techniques have been used so far: Exposure to What tasks need to be achieved by film stimuli
emotional slides (Bradley & Lang, 2000; Schae- in emotion research? Traditionally, it has been
fer, Fletcher, Pottage, Alexander, & Brown, accepted that film stimuli must elicit differentiated
2009), autobiographical recollection (Schaefer emotional states such as anger or fear. This
& Philippot, 2005), mental imagery (Schaefer approach was linked to the ‘‘basic emotions’’
et al., 2003; Vrana, Cuthbert, & Lang, 1986), framework, i.e., the view that emotions are orga-
exposure to emotional film excerpts (Gray et al., nised in a finite set of basic categories. For instance,
2005; Philippot, Schaefer, & Herbette, 2003; Ekman (1984, 1992) suggested that there are six
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
Schaefer et al., 2006), Velten mood-induction basic emotions: happiness, surprise, fear, sadness,
technique (Velten, 1968), facial feedback (Matsu- anger and disgust. These views have led to the
moto, 1987), respiratory feedback (Philippot, creation of sets of emotional film stimuli able to
Chapelle, & Blairy, 2002), real-life techniques elicit differentiated experiential states correspond-
(Landis, 1924; Stemmler, Heldmann, Pauls, & ing to basic emotions. However, emotion research
Scherer, 2001), and many other techniques. often addresses questions beyond emotional dis-
Compared to other methods, exposure to emo- creteness and hence needs stimuli validated on
tional film excerpts has several advantages: First, it more diverse criteria. For instance, two films
is one of the easiest techniques to implement in inducing differentiated feelings of sadness and
a laboratory. Second, it has been widely observed happiness would not be useful for a study con-
that film excerpts can elicit strong subjective and cerned with the effects of different levels of
physiological changes (e.g., Frazier, Strauss, & emotional arousal (e.g., low vs. moderate vs. high)
Steinhauer, 2004; Gross, 1998; Palomba, Sarlo, on a cognitive task (e.g., Rottenstreich & Hsee,
Angrilli, Mini, & Stegagno, 2000). Third, the 2001). In fact, a significant proportion of emotion
dynamic nature of film scenes provides an optimal research uses parameters derived from a dimen-
artificial model of reality, without the ethical and sional approach of emotions, in which emotional
practical problems of real-life techniques. Fourth, states are thought to be organised along generic
it seems to be the most powerful technique to elicit continuous dimensions rather than discrete basic
emotion in a laboratory: Westermann, Spies, Stahl, emotions (e.g., Davidson, 1992, 1993; Lang,
and Hesse (1996) showed in a meta-analysis that Greenwald, Bradley, & Hamm, 1993; Plutchik,
among several other techniques, film clips were the 1980; Russell, 1980; Watson et al., 1988). Arousal
most potent inducers of both positive and negative and valence (i.e., the dissociation between positive
mood states. an negative states) tend to be very popular dimen-
Previous studies have already tested the relia- sions used in emotion research (Bradley & Lang,
bility of film clips to elicit emotions, leading to the 2000; Gray, 2001). Therefore, a battery of emo-
development of reliable sets of stimuli (e.g., Gross tional stimuli needs to provide contents allowing
& Levenson, 1995; Hagemann et al., 1999; the manipulation of these dimensions as well as the
McHugo, Smith, & Lanzetta, 1982; Philippot, manipulation of basic emotions. Moreover, there is
1993). The present validation study aimed to set also the need for some researchers to elicit
up a new battery of film stimuli that would expand ‘‘blends’’ of emotions rather than pure, discrete
the current choice of film stimuli and allow a more states. An emerging trend of research has
flexible selection of stimuli covering a larger array focused on how ‘‘mixed feelings’’ can arise, i.e.,
of potential research questions. Therefore, we the simultaneous presence of more than one
developed and tested the effectiveness of what is basic emotion (Hemenover & Schimmack, 2007;
to our knowledge the largest normative database of Larsen, McGraw, & Cacioppo, 2001; Schimmack,
2001). Therefore, an ideal database of emotional would be as adaptable as possible to a large array
stimuli would also provide the possibility of select- of research questions. Therefore it is important to
ing stimuli on the basis of their capacity to elicit stress that it was not the goal of this study to
specific ‘‘mixes’’ of emotional states. investigate mechanisms of emotion per se, but
In addition, a database of emotion stimuli needs instead to develop and assess the effectiveness of a
to offer the possibility of tapping a wide array of methodological tool for emotion research. The
emotional experiences. Regarding this issue, there present project had the following goals:
is a common bias in emotion research in favour
of negative emotions, which have traditionally 1. To select a large set of emotion-eliciting
received more attention than positive emotions in film scenes covering a wide range of emo-
the literature (Averill, 1982; Buck, 1999). This tional dimensions.
phenomenon is compounded by the fact that 2. To include subsets of positive films that can
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
emotional dimension (Lang et al., 1993; Russell, was chosen. The 70 selected excerpts were cut and
1980; Sonnemans & Frijda, 1995). recorded on VHS videotapes.
The duration of each film clip ranged from 1 to
7 minutes. Each clip was cut to make a coherent
METHOD segment in order to maximise the emotional
meaning of each clip. For instance, the excerpt
Preliminary survey and excerpt selections
of ‘‘American History X’’, in which a violent
The first step in the present study was the murder is shown, is cut just before the police
selection of a relatively large number of film sirens can be heard. The purpose was to prevent
excerpts corresponding to seven a priori emotional an attenuation of the meaning of injustice which
categories: anger, sadness, fear, disgust, amuse- is evoked by the scene. As participants were all
ment, tenderness, and neutral state. The choice of French-speaking students, the spoken language of
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
the first four categories was motivated by the all excerpts was French (dubbed versions were
assumptions: (a) that they cover the negative used for all non-French films).
states most commonly accepted as being ‘‘basic
emotions’’ (Ekman, 1984, 1992; Izard, 1991); (b)
Participants and design
that they are often studied in emotion research;
and (c) that they refer to verbal labels easy to Three hundred and sixty-four undergraduate
understand and commonly used in everyday students (294 females and 70 males, mean age:
language. Amusement and tenderness are not 19.6, SD3.11) from a Belgian French-speaking
generally regarded as canonical basic emotions university took part in the experiment for course
but they were chosen because they refer to two credits. Participants were pseudo-randomly as-
distinct positive categories evidenced in previous signed to 7 groups, such that each group had an
research on emotion categorisation (Shaver et al., equal proportion of male participants. Each group
1987). Further, and pragmatically, they can be had an average of 52 participants (max: 56 and
easily evoked by films. In addition, tenderness min: 44). Within each group, 10 subgroups were
is an attachment-related emotion (Hatfield & created, differing only according to the order of
Rapson, 2000) therefore filling a gap in the field presentation of the stimuli, to control for potential
of emotion elicitation. order effects. The order of presentation was also
In order to provide a fairly large number of set in such a way that (a) two films targeting the
film clips, 10 films were selected per emotional same emotion were not shown consecutively, and
category. These 70 film excerpts had to be (b) participants never watched more than two
selected from a large initial pool of possible film films of the same valence consecutively. A set of
scenes. For this purpose, 50 experts (i.e., inde- 10 different film excerpts was assigned to each
pendent film rental store managers) were asked to group. Within each 10-film set, 7 film categories
fill in a brief questionnaire in which they had to (anger, happiness, fear, disgust, sadness, tender-
recall, describe and rate for emotional arousal at ness and neutral) were represented. Consequently,
least three film excerpts for each emotional three film categories were represented twice in
category. Film rental store managers were chosen each group. The repetition of film categories was
because they were likely to have a significant fully counterbalanced across the seven groups.
knowledge about films, which could contribute to For subject-level analyses (in which participants’
creating a large preliminary list of film scenes. responses are the main unit of analyses, see results
A large initial set of 824 film excerpts was section), one film out of 2 was randomly selected
obtained with this method. For each emotional in each repeated category for each participant.
category, the 10 most frequently cited excerpts This method enabled us to compute analyses
were chosen. If two or more films were quoted of variance (ANOVAs) in which each observation
with the same frequency, the most intense film is comparable within each participant. For the
computation of ranking scores in film-level ana- displayed. Participants were instructed to watch
lyses all observations were included. the whole scene attentively, without diverting
their attention from the screen. The films were
displayed on 17ƒ colour screens. After each film
Procedure
excerpt, participants completed questionnaires
Participants came to the laboratory in subgroups about their emotional state (see measures section).
of 3 to 5 people. Upon their arrival, participants This procedure (including relaxation instructions)
were instructed that they would see several was repeated for each film excerpt.
emotional film excerpts. Participants were told
that they would go through a relaxation procedure Measures
before viewing each excerpt and that they would
have to report their emotional reactions to the Self-reported emotional arousal. The intensity of
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
film by filling out questionnaires after watching subjective emotional arousal was assessed using a
the excerpt. Following instructions first used by 7-point scale: ‘‘While I was watching the film . . .’’
Philippot (1993), participants were encouraged to (1)‘‘I felt no emotions at all’’ to (7)‘‘I felt very
report (1) what they had actually felt and not what intense emotions’’. The participants were encour-
they believed people should feel in reaction to the aged to answer this question according to what
movies (in order to strengthen this point, the they actually felt during the task and not accord-
importance of individual differences in reactions ing to what they thought they should feel when
to films was stressed), and (2) how they felt at the watching the film.
specific time they were watching the video
excerpt, and not their general mood of the day. Differential Emotions Scale. To assess discrete
This procedure of retrospective evaluation of emotional dimensions, we used a version of the
emotional feelings is known to be a good Differential Emotions Scale (DES; Izard et al.,
predictor of the actual state felt during the 1974) modified by McHugo et al. (1982) and
stimulation (i.e., exposure to a stimulus), espe- validated in French by Philippot (1993). This
cially the peak (most intense) state and the state choice was motivated by the fact that the DES is
felt by the end of the stimulation (Fredrickson & one of the most widely used self-report scales of
Kahneman, 1993). Participants were also in- discrete emotional feelings (Youngstrom &
formed that their responses would remain totally Green, 2003), and this particular version of the
anonymous. Moreover, they were told that they DES has been used in two previous validations of
could stop their participation in the experiment at emotional film stimuli (McHugo et al., 1982;
any time. All participants gave their consent to Philippot, 1993). Each item consisted of groups of
participate. emotional adjectives: (1) interested, concentrated,
An individual experimental context was created alert; (2) joyful, happy, amused; (3) sad, down-
for each participant: They were separated by hearted, blue; (4) angry, irritated, mad; (5) fearful,
partitions and used individual screens and head- scared, afraid; (6) anxious, tense, nervous; (7)
phones in order to watch the film clips without disgusted, turned off, repulsed; (8) disdainful, scorn-
being able to see or hear the other participants’ ful, contemptuous; (9) surprised, amazed, astonished;
reactions. When all participants were ready to (10) warm hearted, gleeful, elated (see Appendix).
start the experiment, the room’s lighting was In order to extend the range of emotional
dimmed, and pre-recorded relaxation instructions dimensions assessed, we used a modified version
were given: Participants had to close their eyes, to of the DES in which the following items were
relax every muscle (including the face), and breath added: (11) loving, affectionate, friendly; (12)
deeply and regularly for about 2.5 minutes. As guilty, remorseful; (13) moved; (14) satisfied,
soon as the relaxation was over, participants were pleased; (15) calm, serene, relaxed; (16) ashamed,
informed that the first film clip would be embarrassed. This modified version of the DES
has already been successfully used in previous order to provide a means of classifying and
studies (Schaefer & Philippot, 2005; Schaefer comparing individual film excerpts.
et al., 2003). For each group of adjectives,
participants used a 7-point scale (1‘‘not at all’’, 1. Which film categories yielded the stronger
7‘‘very intense’’) to rate the extent to which they levels of emotional arousal?
felt each state as they were watching the film clip.
A repeated-measures ANOVA tested the effect of
film category (anger vs. happiness vs. fear vs.
PANAS. A validated French translation (Gau-
disgust vs. sadness vs. affection vs. neutral) on
dreau, Sanchez, & Blondin, 2006) of the Positive
the arousal scale. Due to the large sample size
and Negative Affect Schedule (PANAS; Watson
(N364), we decided to consider as significant
et al., 1988) was used. The PANAS is a self-
only the effects that fulfilled two criteria: a
reported adjective checklist that contains two
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
8
7
6
5
4
3
2
1
0
ANG FEAR DISG SAD AM TEND NEUT
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
Figure 1. Subject-level mean self-reported emotional arousal as a function of emotional category. Error bars represent91 standard
deviation of the mean. AMAmusement; TENDTenderness; DISGDisgust, ANGAnger; SADSadness; NEUTNeutral.
comparisons in order to test whether negative and of the PA scale (e.g., ‘‘determined’’, ‘‘interested’’
positive films would be different in terms of NA and ‘‘excited’’).
and PA scores. For NA scores, all negative films Taken as a whole, these results indicate that
had significantly higher scores than positive films NA scores were more successful than PA scores at
(pB.001). PA scores were less successful at differentiating between positive and negative dis-
differentiating between positive and negative films, crete emotions. The apparent failure of PA scores
as AmusementAnger, AmusementFear and to differentiate negative and positive films seems
TendernessFear contrasts on PA scores were not to be driven by a subset of items (e.g., attentive,
significant. In order to investigate this issue, we alert, active) reflecting a general ‘‘arousal’’ and
performed the same three contrasts (Amusement ‘‘alertness’’ component that might not necessarily
Anger, AmusementFear and TendernessFear) be valence specific. This result argues for the need
for every individual item contained in the PA of other alternative measures to test the validity of
composite score. We found that several items the positivenegative differentiation of our a priori
were producing unexpected results in the form of categories of emotional films. Therefore, we
significantly higher PA ratings for negative than created two alternative negative and positive
for positive films (pB.001). For the FearTen- valence scores averaging classes of DES items.
derness comparison, this was found for ‘‘active’’ The positive composite score included 5 items
(M1.8, SD1.0; M1.6, SD0.9), ‘‘alert’’ (‘‘joyful, happy, amused ’’; ‘‘warm hearted, gleeful,
(M3.3, SD1.2; M2.4, SD1.2) and ‘‘ex- elated ’’; ‘‘loving, affectionate, friendly’’; ‘‘moved ’’;
cited’’ (M2.0, SD1.1; M1.6, SD0.9). ‘‘satisfied, pleased ’’), and the negative composite
For the AngerAmusement comparison, this was score included 8 items (‘‘sad, downhearted, blue’’;
found for ‘‘alert’’ (M3.3, SD1.1; M3.1, ‘‘angry, irritated, mad ’’; ‘‘fearful, scared, afraid ’’;
SD1.1), and for FearAmusement, this pattern ‘‘anxious, tense, nervous’’; ‘‘disgusted, turned off,
was found for ‘‘alert’’ (M3.3, SD1.2; M2.2, repulsed ’’; ‘‘disdainful, scornful, contemptuous’’;
SD1.2), ‘‘attentive’’ (M3.7, SD1.0; M ‘‘guilty, remorseful ’’; ‘‘Ashamed, embarrassed ’’). All
3.2, SD1.0) and ‘‘excited’’ (M2.0, SD1.1; Cronbach’s alphas were greater than .601. A one-
M1.8, SD1.0). Moreover, the same contrasts way ANOVA on the positive composite score
were not significantly different for other sub-items testing for differences between emotional films
1
We used the criterion proposed by Schmitt (1996), according to which a.50 is satisfactory if the composite measure has a
meaningful content coverage of some domain and reasonable unidimensionality.
2.5
1.5
0.5
ANG FEAR DISG SAD AM TEND NEUT
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
Figure 2. Positive affect (Watson et al., 1988): Subject-level means and standard deviations.
revealed a significant main effect of Emotional DES items, F(90, 32,670)228.3, p1
Category F(5, 1810)289.5, pB.001, h2 .44. 1015, h2 .39, indicating that the self-reported
Bonferroni post hoc comparisons revealed signifi- emotional profile was modulated by film category.
cant differences between each of the two emo- In order to further specify whether discrete
tional categories and all of the negative categories emotional states were induced for each film
(pB.001). A one-way ANOVA on the negative category, paired t-test comparisons between the
composite score also revealed a main effect of DES items were computed separately for each
Emotional Category F(5, 1810)121.1, pB.001, emotional state using individual participant’s self-
h2 .25, and Bonferroni post hoc comparisons report scores as unit of analysis. More specifically,
also showed a full significant differentiation for each category of emotional films (anger,
between positive and negative films. These results sadness, fear, disgust, amusement, tenderness,
are depicted in Figures 25. and neutral), a set of 6 predefined paired t-tests
contrasting the target state and each non-target
state were computed for every participant. For
3. Can differentiated emotional feeling
instance: for the category of films evoking fear,
states be elicited with the film set?
one target DES item was chosen (fearful, scared,
A 716 repeated-measures ANOVA found a afraid) and compared to 6 non-target items,
significant interaction between film category and which are targets to the other 6 emotional
2.5
1.5
0.5
ANG FEAR DISG SAD AM TEND NEUT
Figure 3. Negative affect (Watson et al., 1988): Subject-level means and standard deviations.
5
4.5
4
3.5
3
2.5
2
1.5
1
0.5
ANG FEAR DISG SAD AM TEND NEUT
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
Figure 4. Positive composite score (DES-based): Subject-level means and standard deviations.
categories (namely: anger, sadness, amusement, discreteness (see the ranking scores section for
tenderness, neutral, disgust). Due to the large the calculation of this coefficient), yielded sig-
sample size (N364) and to the large number of nificant AngerFear comparisons (pB.005) for
planned comparisons, we decided to (a) apply all but one of them (American History X),
Bonferroni corrections for multiple comparisons, indicating that a fairly good number of anger
and (b) to set a stringent significance threshold of excerpts yielded a high level of differentiation
pB.0001. Therefore, only t-values4.3 were between anger and fear when they were analysed
considered significant. Results revealed that individually. Individual film identities can be
most comparisons were significant according to found in Table 1.
these criteria, confirming the expected differentia-
tion between target states. Only one exception 4. Were there any gender differences?
was found: The difference between fear and anger In order to assess gender differences, ANOVAs
did not reach the significance criterion during described above were run with gender as an
anger induction (although the same difference is additional factor. We found a main effect of
highly significant during fear induction). To Gender in the arousal scale, F(1, 362)30.1, pB
further analyse this issue, we computed discrete- .0001, h2 .08, suggesting that women reported
ness analyses separately for every anger film. The higher levels of emotional arousal than men when
five film excerpts with the higher level of anger viewing video clips (Women: M4.4, SE0.04,
4.5
4
3.5
3
2.5
2
1.5
1
0.5
ANG FEAR DISG SAD AM TEND NEUT
Figure 5. Negative composite score (DES-based): Subject-level means and standard deviations.
Table 1. Sets of ten first-ranked film excerpts (by arousal, positive and negative affect)
1 Misery (6.12) The Dead Poets Society [2] (2.82) American History X (2.73)
Annie (Kathy Bates) breaks Paul’s legs All the students climb on their desks A neo-Nazi (played by Edward
(James Caan) to express their solidarity with Mr Norton) kills an African-American
Keating (Robin William), who has just man, smashing his head on the curb
been fired
2 Schindler’s List [1] (5.88) The Dinner Game (2.50) Schindler’s List [1] (2.62)
Dead bodies are being carried away Complex humoristic scene Dead bodies are being carried away
in a concentration camp in a concentration camp
3 Dead Man Walking (5.87) Life is beautiful [4] (2.49) The Piano (2.49)
The main character is put to death In a concentration camp, a father One of the characters gets her finger
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
Notes: In each column, the films listed correspond to the 10 movies that obtained the highest score on a given criterion. The ranks are
in descending order (e.g., Misery is the most intense film excerpt) with mean scores in parentheses. Numbers in square brackets
correspond to different scenes of the same film. Mixed feelings coefficients are available online at: http://nemo.psp.ucl.ac.be/
FilmStim/
Table 1 (Continued)
Rank Amusement (d) Tenderness (d) Anger (d) Sadness (d) Fear (d) Disgust (d)
1 ‘‘Les trois fre`res’’ Forrest Gump Schindler’s List [2] City of Angels The Blair Witch Trainspotting
(4.55) (4.20) (2.19) (2.32) Project (2.93) [2] (4.07)
One of the Father and son are A concentration Maggie (Meg Final scene in The main
characters takes reunited camp commander Ryan) dies in which the character dives
part in a TV randomly shoots Seth’s (Nicolas characters are into a filthy
game prisoners from his Cage) arms apparently killed toilet
balcony
2 The Dinner Game Life is Beautiful Sleepers (2.14) Dangerous Mind The Shining (2.77) Seven [3]
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
Table 1 (Continued)
Rank Amusement (d) Tenderness (d) Anger (d) Sadness (d) Fear (d) Disgust (d)
6 Benny and Joon The Dead Poets Schindler’s List [3] Dead Man Child’s Play 2: The Dentist
(3.73) Society [2] (3.24) (1.33) Walking (1.02) Chucky’s Back (2.02) (2.84)
Benny (Johnny By the end of the Killing of Jews in The main Chucky beats A man finds a
Depp) plays movie, all the a ghetto during character is put to Andy’s teacher woman whose
the fool in a students climb on WWII death by lethal with a ruler tongue has
coffee shop their desks to injection been savagely
express their cut off
solidarity with
Mr Keating
(Robin William),
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
Notes: In each column, the films listed correspond to the 10 movies that obtained the highest score on a given criterion. The ranks
are in a descending order (e.g., Misery is the most intense film excerpt) with mean scores in parentheses. (d) indicates that movies
are ranked according to their score of discreteness in a given emotional dimension. Numbers in square brackets correspond to
different scenes of the same film. Mixed feelings coefficients are available online at: http://nemo.psp.ucl.ac.be/FilmStim/
Table 1 (Continued)
Sets of ten first-ranked film excerpts (by positive and negative scores)
Notes: In each column, the films listed correspond to the 10 movies that obtained the highest score on a given criterion. The ranks are in a
descending order (e.g., Misery is the most intense film excerpt) with mean scores in parentheses. Numbers in square brackets correspond
to different scenes of the same film. Mixed feelings coefficients are available online http://nemo.psp.ucl.ac.be/FilmStim/.
Men: M3.8, SE0.09). The EmotionGen- cance criteria in analyses involving the PA factor
der interaction failed to pass the significance or DES measures.
criteria for arousal scores. For NA, a main effect
of Gender was found, F(1, 362)28.7, pB.001, 5. How to select efficient films from the
h2 .07, indicating that women reported a higher
NA level than men when viewing video clips
database?
(Women: M1.7, SE0.02, Men: M1.4, A film-level analysis was computed in order to
SE0.04). The EmotionGender interaction provide a means to classify and compare indivi-
failed to pass the significance criteria for NA dual films according to specific criteria. We
scores. No effects of Gender passed the signifi- therefore computed for each film a mean of each
measured variable, obtaining a final data matrix scores (DES based); (4) the six discreteness scores;
including the scores of each film for 19 variables and (5) the 15 mixed-feelings scores. We defined
corresponding to the arousal scale, the positive subgroups of high-ranking films for the 26 final
and negative affect scores, and the 16 extended- criteria in order to identify individual films with
DES items. Next, we computed two film-level high levels of efficiency (see ‘‘high-scoring sub-
coefficients for positive and negative composite groups’’ section). Within each subgroup, an out-
scores averaging classes of DES items as explained lier analysis was performed in order to detect
previously (Question 2 of the results section). whether specific films were less or more efficient
Finally, we computed two classes of coefficients than the other films of the subgroup. Films
aimed at classifying film stimuli according to the included in high-scoring subgroups for Arousal,
levels of differentiation of emotional states elicited PA, NA, composite scores and discreteness scores
by them. First, we computed discreteness coeffi-
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
Figure 6. Distribution boxplots of subgroups of high-scoring films and outliers for each validity criterion. The vertical axis represents a scale
of standardised scores (Z-scores) for all the classification criteria. Each grey box represents the interquartile range of the subgroup; the dark
horizontal line represents the median of the subgroup; the whiskers indicate the farthest scores that are not outliers and the dots designate
outliers. An outlier is a score at 1.5 interquartile ranges from the lower and upper edges of the box. A star represents scores beyond
3 interquartile ranges. The identity of the outliers is listed in the text. Numbers in the horizontal axis refer to specific classification criteria:
1subjective arousal; 2NA: PANAS negative affect; 3PA: PANAS positive affect; 4Positive composite score (DES-based);
5Negative composite score. The following numbers refer to discreteness coefficients: 6Amusement; 7Tenderness; 8Anger;
9Sadness, 10Fear; 11Disgust. The following numbers represent pairwise mixed feelings coefficients: 12AmusementTenderness,
13AmusementAnger, 14 AmusementSadness, 15AmusementFear, 16AmusementDisgust, 17TendernessAnger,
18TendernessSadness, 19TendernessFear, 20TendernessDisgust, 21AngerSadness, 22AngerFear, 23AngerDisgust,
24SadnessFear, 25 SadnessDisgust, 26FearDisgust.
inefficient film. Outliers were defined as scores for both TendernessAnger and Tenderness
beyond 1.5 interquartile ranges from the 1st and Disgust, and Schindler’s List [1] was an outlier
3rd quartile of each subgroup (Tukey, 1977). for SadnessFear.
No subgroup was found to include films signifi-
cantly below the subgroup median score, and
the distribution within subgroups appeared to be
homogeneous (Figure 6). Observed outliers were
DISCUSSION
for films obtaining scores well above the median
The present study tested the effectiveness of a set
of their subgroups, indicating that their efficiency
of 70 film clips to elicit emotions in an experi-
is even higher than the efficiency of the high-
scoring films. The following films: The Dead Poets mental context according to different validity
Society [2], Trainspotting [2] and American History criteria. Several questions were addressed by
X, were found to be outliers for PA, NA and the data analysis. First, ANOVAs showed that
disgust discreteness, respectively. For mixed-feel- the films were capable of eliciting significantly
ings scores, Life is Beautiful [4] was an outlier for intense emotions and differentiated responses in
AmusementTenderness, Man Bites Dog [2] was terms of positive and negative affect. Second,
an outlier for AmusementAnger, Trainspotting Bonferroni comparisons yielded fairly high levels
[3] was an outlier for AmusementDisgust, A of discreteness for the 7 target states for each
Perfect World and Life is Beautiful [1] were outliers category of film.
Several validity criteria were defined to classify and negative films. These alternative scales were
and compare films: A general arousal criterion, also provided as classification criteria.
seven criteria for emotional discreteness (Anger, Next, gender differences in reactivity to the
Sadness, Fear, Disgust, Amusement, Tenderness), films were observed. More specifically, women
a general positive affect dimension (PANAS seemed to report higher levels of subjective
positive affect scale), a general negative affect arousal, and higher levels of negative affect than
dimension (PANAS negative affect), two positive men. No significant gender differences were
and negative composite scores based on the DES, found in positive affect as defined by the PANAS
and 15 ‘‘mixed-feelings’’ scores. For each criterion, scales. A higher level of reported emotional
a subgroup was created. Results indicated that the feelings for female participants is consistent with
films contained in these subgroups are those most Gross and Levenson’s (1995) and Hagemann
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
likely to efficiently elicit expected affective states et al.’s (1999) results and coherent with many
in laboratory sessions. These films are presented studies showing that women usually report more
in Table 1, and extensive additional data on these intense emotions than men (Brody & Hall, 2000).
films and on all the films of the database are These results may reflect either neurobiological
available on the Internet along with descriptions differences in emotional systems across genders
of the film excerpts and editing instructions.3 This (Cahill, Uncapher, Kilpatrick, Alkire, & Turner,
allows emotion researchers to select the most 2004) or the effects of socio-cultural stereotypes
appropriate stimuli for their research and to edit about gender roles (Brody & Hall, 2000; Feld-
the excerpts in versions of the movies dubbed in man-Barrett, Robin, Pietromonaco, & Eyssel,
their study participants’ language. 1998; Kelly & Hutson-Corneaux, 1999).
A notable unexpected result was the failure Two caveats concerning this study have to be
of the PA scale (Watson et al., 1988) to considered. First, this validation study was con-
differentiate negative and positive films. This ducted in a specific cultural context (French-
phenomenon has already been observed with speaking European) that may have influenced
emotional picture stimuli (Patrick & Lavoro, the choice and evaluation of films for some
1997) in a study showing that negative pictures emotional categories. In order to address this
generated high PA scores. Our analysis of issue, we included films from different cultural
individual items showed that this effect was backgrounds (French, Italian, British and US
driven by specific items that appear to reflect cinema) to allow some culture-specific choice of
arousal/alertness. Arousal/alertness items seem to stimuli. For instance, even though some amuse-
be valence independent because they increase in ment films are derived from the French cinema,
response to both positive and negative, especially half of the ten highest ranked amusement excerpts
fear-inducing and anger-inducing, film clips. The were taken from US and British cinema: When
same items also drove Patrick and Lavoro’s Harry Met Sally, A Fish Called Wanda, Benny and
(1997) finding that PA scores increased in Joon and Something About Mary. Further research
response to negative pictures. These findings will be needed to expand the choice of stimuli to
suggest that caution should be exercised in non-Western cultures. Second, in line with all
the interpretation of PANAS PA scores as previous film-validation experiments, the present
indicators of the valence of affective experiences. study relied solely on self-reports to assess emo-
We have calculated alternative positive and tional responses. Future research is needed to
negative valence scores (based on the DES), validate stimuli on expressive and psychophysio-
which were successful in differentiating positive logical criteria.
3
Detailed information about the film excerpts used in this study is available at the following URL: http://
nemo.psp.ucl.ac.be/FilmStim/. This website contains the database used for this study, free of charge.
Beyond using films as elicitors of affective including the film-level data matrix so that any-
states that can modulate cognition and social one can download and analyse the data in order to
behaviour, we believe that several pathways are flexibly select a set of stimuli. It is worth noting
open for future research on emotional responsive- that data from additional film excerpts can be
ness to films per se. For instance, in combination added by researchers who wish to validate other
with psychophysiological and neuroimaging tech- film excerpts using similar methods and measures.
niques (e.g., fMRI, ERP, MEG, eye-tracking The comparison with the original database could
techniques) research using film stimuli may yield provide an index of external validity to the newly
interesting insights about the perceptual and added film excerpts.
attentional processes involved in emotional pro-
cessing, as well as the neuroanatomical and Manuscript received 10 March 2009
neurochemical correlates of emotional processes. Revised manuscript received 14 August 2009
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
Likewise, using film excerpts could help further Manuscript accepted 14 August 2009
First publised online 26 November 2009
our understanding of the time course of an
emotional state, i.e., how an emotional reaction
unfolds over a given period, and what factors
determine its maintenance or decline. This ques-
REFERENCES
tion is linked to the duration of an emotional
Averill, J. R. (1982). On the paucity of positive
state, a feature important for conceptual distinc-
emotions. In K. R. Blankstein, P. Pliner, & J. Polivy
tions between mood and emotion (Ekman & (Eds.), Assessment and modification of emotional
Davidson, 1994, pp. 4996; Frijda, 1993). behavior (pp. 745). New York: Plenum Press.
This study attempted to validate emotional Bradley, M. M., & Lang, P. J. (2000). Measuring
film stimuli taking into account the theoretical emotion: Behavior, feeling and physiology. In R. D.
diversity of the field. Specifically, we defined Lane & L. Nadel (Eds.), Cognitive neuroscience of
several validity criteria derived from two cate- emotion (pp. 242276). New York: Oxford Univer-
gories of theoretical models of emotion: the basic sity Press.
emotions approach and the dimensional approach. Brody, L. R., & Hall, J. A. (2000). Gender, emotion
Although these two theories are prominent, they and expression. In M. Lewis & J. M. Haviland-
do not exhaust the diversity of theories of Jones (Eds.), Handbook of emotions (pp. 338349).
New York: Guilford Press.
emotion. Of particular interest is the componen-
Buck, R. (1999). The biological affects: A typology.
tial approach*or appraisal models*that defines
Psychological Review, 106, 301336.
emotion in terms of different combinations and Cahill, L., Uncapher, M., Kilpatrick, C., Alkire, M. T.,
blends of discrete cognitive and motivational & Turner, J. (2004). Sex-related hemispheric later-
components such as goal relevance, novelty and alization of amygdala function in emotionally influ-
expectancy (Scherer, 1999, 2000). Future devel- enced memory: An fMRI investigation. Learning &
opments in emotion elicitation techniques would Memory, 11, 261266.
potentially benefit from using validity criteria Davidson, R. J. (1992). A prolegomenon to the
derived from these constructs. structure of emotion: Gleanings from neuropsychol-
In sum, the main goal of this study was to build ogy. Cognition and Emotion, 6, 245268.
a flexible tool for emotion research, that is, a set of Davidson, R. J. (1993). Parsing affective space:
Perspectives from neuropsychology and psychophy-
film clips from which different subsets of stimuli
siology. Special section: Neuropsychological per-
can be chosen according to specific criteria. For
spectives on components of emotional processing.
this purpose, we gathered a comprehensive set of Neuropsychology, 7, 464475.
films, which successfully fulfilled different validity Depue, R. A., & Morrone-Strupinsky, J. V. (2005). A
criteria. Next, we created several different indices neurobehavioral model of affiliative bonding: impli-
along which films were ranked. The entire cations for conceptualizing a human trait of affilia-
database was made freely available on Internet, tion. Behavioral and Brain Sciences, 28, 313350.
Ekman, P. (1984). Expression and the nature of Gross, J. J., & Levenson, R. W. (1995). Emotion
emotion. In K. R. Scherer & P. Ekman (Eds.), elicitation using films. Cognition and Emotion, 9,
Approaches to emotion (pp. 319344). Hillsdale, NJ: 87108.
Lawrence Erlbaum Associates, Inc. Hagemann, D., Naumann, E., Maier, S., Becker, G.,
Ekman, P. (1992). An argument for basic emotions. Lurken, A., & Bartussek, D. (1999). The assessment
Cognition and Emotion, 6, 169200. of affective reactivity using films: Validity, reliability
Ekman, P., & Davidson, R. J. (1994). The nature of and sex differences. Personality and Individual
emotion: Fundamental questions. Oxford, UK: Oxford Differences, 26, 627639.
University Press. Hatfield, E., & Rapson, R. L. (2000). Love
Feldman-Barrett, L., Robin, L., Pietromonaco, P. R., and attachment processes. In M. Lewis & J. M.
& Eyssel, K. M. (1998). Are women the more Haviland-Jones (Eds.), Handbook of emotions
emotional sex? Evidence from emotional experiences (pp. 595604). New York: Guilford Press.
in social context. Cognition and Emotion, 12,
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
Panksepp, J. (1998). Affective neuroscience. New York: Scherer, K. R. (2000). Psychological models of emo-
Oxford University Press. tion. In J. C. Borod (Ed.), The neuropsychology of
Patrick, C. J., & Lavoro, S. A. (1997). Ratings of emotion (pp. 137162). Oxford, UK: Oxford Uni-
emotional response to pictorial stimuli: Positive and versity Press.
negative affect dimensions. Motivation and Emotion, Schimmack, U. (2001). Pleasure, displeasure, and
21(4), 297321. mixed feelings: Are semantic opposites mutually
Philippot, P. (1993). Inducing and assessing differen- exclusive? Cognition and Emotion, 15(1), 8197.
tiated emotional feeling states in the laboratory. Schmitt, N. (1996). Uses and abuses of coefficient
Cognition and Emotion, 7, 171193. alpha. Psychological Assessment, 8, 350353.
Philippot, P., Chapelle, C., & Blairy, S. (2002). Shaver, P., Schwartz, J., Kirson, D., & O’Connor, C.
Respiratory feedback in the generation of emotion. (1987). Emotion knowledge: Further exploration of
Cognition and Emotion, 16, 605627. a prototype approach. Journal of Personality and
Philippot, P., Schaefer, A., & Herbette, G. (2003).
Downloaded by [University of California Santa Barbara] at 08:55 07 July 2013
APPENDIX
Original French version of the modified
Differential Emotions Scale used in the
study