On Dimensions in Emotion Psychology

On Dimensions
...Basic Emotions
....Appraisal Theories
.....Dynamical Systems Approach
in Emotion Psychology
Christian Kaernbach
Institut für Psychologie
Christian-Albrechts-Universität zu Kiel
Dimensional Theories of Emotion
• A limited set of orthogonal dimensions
• Emotional experience is described by
coordinates in a Euclidean space
• Number and character controversial
– Wundt (1896): 3 dimensions of emotional experience
• Lust – Unlust
• Erregung – Beruhigung
• Spannung – Lösung
– First two dimensions relatively undisputed
• Lust – Unlust  Wohlgefallen  pleasure  valence
• Erregung – Beruhigung  Aktivierung  arousal
– Further dimensions in dispute
• Is a third dimension needed? If so: What does it represent?
– Spannung-Lösung  dominance  control  social nearness
• Or do we need more than three dimensions?
– Scherer: 4th dimension “novelty”
IAPS
International Affective Picture System
Bradley & Lang (1994)
• more than 800 pictures with ratings of
valence, arousal, and dominance
9
arousal ratings
8
7
6
5
4
3
2
1
1
2
3
4
5
6
7
8
9
valence ratings (5 = neutral)
V-shape
• Valence-Arousal scatter plots
show characteristic V-shape
• Consider, e.g., a 3 x 2 design with
– valence:
– arousal:
low / medium / high
low / high
 „Holes“ in the dimensional space
– alternative interpretation:
two independent (mutual exclusive)
processes
• positive affect
A+
• negative affect
A–
Watson & Tellegen, 1985
8
arousal ratings
– contradictory to the concept
of a dimensional space
9
A–
7
A+
6
5
4
3
2
1
1
2
3
4
5
6
7
8
9
valence ratings (5 = neutral)
Metamerism
Metamerism
• Dimensions of color space are based on metamerism
– Colors with the same hue/saturation/brightness are indistinguishable
• Does metamerism work for emotions?
V 5.8
A 4.3
D 5.9
V 5.8
A 4.3
D 6.0
8
arousal ratings
– IAPS slides occupying the same point
in valence/arousal/dominance space:
9
7
6
5
4
3
2
1
1
2
3
4
5
6
7
8
9
valence ratings (5 = neutral)
Slide 4531: “erotic male”
Slide 7351: “pizza”
Rating
A note on rating data
• Dimensional theories of emotion are based on ratings
• The rating scales follow from theory, not from data
• It remains thus unclear:
– How many scales should be used?
– Which scales should be used?
– Do emotions even map on a dimensional space?
• How to confirm ratings
– Factor analysis
• Morris & Bradley (1994): PCA of 135 emotionally loaded adjectives:
three main factors correlate with valence / arousal / dominance
• Factor analysis does not test underlying model of space
– Multidimensional scaling (MDS)
• tests, whether dimensional arrangement is internally consistent
MDS
Multidimensional scaling (MDS)
• Given N items
• Collect dissimilarity data for
N ∙ (N–1) / 2 pairs of items
• Establish configuration
– Arrange items in multidimensional space
such that similar items are close to each other
and dissimilar items are far from each other
– Distance should be a monotonically
increasing function of dissimilarity
– Move items in space so as to minimize
deviation from monotonic function
• Stress = portion of variance
not explained by monotonic function
J: Joy
W: Anger
G: Anticipation
D: Acceptance
Dimensional Analysis:
Why it
How
one
should
should
benot
done
do it
A–
Arousal
Example data from a classroom experiment
disproving the “Wheel of Emotions”
(Plutchick, 1980)
P: Sadness
M: Fear
T: Surprise
A: Disgust
A+
Valence
• Collect 8∙7/2 = 28 dissimilarity data
– 90 participants, each sending 28 votes per SMS
– calculate stress for
– original “wheel” comes of badly
– decide for two-dimensional configuration
• Interpretation (post hoc)
– valence, arousal
– A+/A–
Stress
• “Wheel of Emotions” (red circle)
• random data (gray lines)
• optimal 1-dim / 2-dim / 3-dim configuration
0,5
0,4
Stress of
wheel model:
0.235
0,311
* p  0.05
0,3
0,2
0,125
** p  0.005
0,1
0,088
n.s. p > 0.05
0
0
1
2
3
Number of assumed dimensions
Dimensional Analysis:
Why one should not do it
• Question
– What is the opposite of sweet?
• Polemics
– Demonstration of the dimensional approach
in the science of taste
Taste perception
for the
moment
let us feign
ignorance
• Definition: A basic quality of taste perception is
related to a specific receptor on the tongue
• Receptors for basic qualities operate independently
• Five (six) basic qualities
–
–
–
–
–
–
sweet
sour
salty
bitter
umami
(fat)
• Classroom experiment
log10(lemon juice)
Taste space
2
7
8
9
4
5
6
1
2
3
1
0,7
– Nine liquids
• 3 concentrations of sugar
(10 30 60 lumps / l)
 3 concentrations of lemon juice (20 60 120 ml / l)
– 52 participants, 4 test per person
– dissimilarity ratings on a scale reaching from 0 to 4
– each pair (9 ∙ 8 / 2 = 36) was tested about six times
1,7
log10(sugar)
log10(lemon juice)
Results
7
8
9
4
5
6
1
2
3
1
• Stress analysis:
0,7
– 1 dimension very good
– 2 dimensions significant
1,7
log10(sugar)
0,5
2-dimensional
Stress
0,4
• Configurations:
1-dimensional
2
0,3
0,249
** p  0.005
0,2
0,126
* p  0.05
0,1
0,076
n.s. p > 0.05
sour
sweet
sour
• Interpretation:
Intensity
0
0
sweet
– 1st dimension “sour / sweet”
– 2nd dimension “intensity”
1
2
Number of assumed dimensions
(valence?)
(arousal?)
3
A little Learning is a dang'rous Thing;
Drink deep, or taste not the Pierian Spring.
• Ignoramus says:
sour
113°
Intensity
Discussion
sweet
log10(lemon juice)
sour
sweet
– The first dimension of taste space is a sour – sweet axis.
– A second dimension describes the intensity of a taste.
2
7
8
9
4
5
6
1
2
3
1
0,7
1,7
log10(sugar)
• This interpretation obfuscates the mechanisms of taste perception
– Sweet and sour are detected by independent receptors
• Good knowledge about the underlying mechanisms
is a precondition for a good interpretation of dimensional analysis
– Taste perception: The underlying mechanisms are well known. The discrepancy
between dimensional analysis and underlying mechanisms is revealing.
• The cognitive representation of sweet and sour tastes is not quite orthogonal.
– Color perception: Elementary Three-Color Theory (Helmholtz/Young, receptors),
perceptually Opponent-Color Theory (Hering, bipolar cells).
– Emotion psychology: The underlying mechanisms are unclear. Are there basic
emotions? How many of them are there? Are they independent from each other? ...
• At present, a dimensional analysis of perceived emotions
is mere phenomenology without explanatory power.
Basic emotions
Basic emotions
• A limited set of discrete emotions
– universal
– corresponding to neurophysiological/anatomical substrates
– building blocks for other, nonbasic emotions
• Number and character controversial
– Mowrer (1960): 2 basic emotions
– Arnold (1960): 11 basic emotions
Stimuli from KASPAR, the
Kiel Affective SPeech ARchive
• Ekman, 1984: 6 emotions
•
•
•
•
•
•
•
Sadness
Fear
negative
Anger
Disgust
positive
Joy
Surprise
(neutral)
arousal
Basic emotions
valence
• slightly more differentiated account of emotions
– more differentiation is needed
• Static account, emotions are described as states
– No account for “burst emotions” (disgust)
• difficult to speak an entire sentence disgustedly
Appraisal theories
Appraisal theories
• Appraisal theories explain the process
how a stimulus builds up an emotion
– Appraisal may stay subconscious, may happen very fast
and may be automatic
• Appraisal is feed forward only
– Stimulus  Appraisal  Emotion  Response
– No account for body loop
• Cartoons are rated more amusing by participants holding a pencil
between their teeth (grinning) than by participants holding it with their
lips (frowning) Strack, Martin & Stepper, 1988
• In preparation in our lab: Participants rate heart-moving stimuli (movies,
music) while being submitted to high static voltage, making their hairs
stand up (artificial goose bumps)
Dynamical Systems Approach
Dynamical systems approach
• Emotions are both cause and effect of appraisals
• Recurrent interactions
• Stable emotional states result from interplay of
amplifying effects and negative feedback loops
– “The basic emotional systems may act as „strange attractors‟
within widespread neural networks that exert a certain type of
„neurogravitational‟ force on many ongoing activities of the
brain, from physiological to cognitive.” (Panksepp, 1998)
Lewis, M.D.: Bridging emotion theory and neurobiology
through dynamic systems modeling.
Behavioral and Brain Sciences 28 (2005) 169–245.
Conclusions
Conclusions
• Don‘t oversimplify
Thank you for your attention!