Between and Within Subject
Measures of Affect
William Revelle and Eshkol Rafaeli-Mor
Northwestern University
European Association of Personality Psychology
Krakow, Poland, July, 2000
http://www.personality-project.org
http://pmc.psych.nwu.edu
Between and Within Subject
Measures of Affect
• William Revelle and Eshkol Rafaeli-Mor
• With the collaboration of
–
–
–
–
–
–
Kris Anderson, GAO
Erin Baehr, Northwestern University
Douglas Billings, St. Marys College
Gregory Rogers, University of Chicago
Rishi Agrawal, Northwestern University
Neera Mehta, University of Illinois, Chicago
• Support from
– US ARI contract MDA903-93-K-0008
Between and Within Subject
Measures of Affect
• Personality traits and affective states
• Between versus within measures of mood
and affect
– Traditional measures of dimensionality,
stability and variability of affect
– Alternative within subject measures
– Studies of the “tides of emotion”
• Applications to cognitive performance
Personality traits and affective states
• Personality: a musical metaphor
– A tune may be recognizable even if played with different
notes with a different instrument
– A person is recognizable by the patterning of affective and
cognitive states even though specific behavioral acts vary
– Personality traits are coherent patterns of changes of states
• Multi-level modeling of parameters of affect
– Level
– Amplitude
– Phase
– Coherence
– Synchrony (of multiple affects)
The long and short term
predictability of affect
• How happy will you feel 12 years from
today?
– Are some people more likely to be happy than
others?
• How happy will you feel 12 hours from
now?
– Are some people more predictable over time?
• How do affective rhythms allow for a better
understanding of cognitive processes?
Multiple formulations of the
measurement of affect
• Two dimensional models
– Affective Valence and Arousal (Russell et al.)
– Positive and Negative Affect (Tellegen, Watson
& Clark)
– Energetic and Tense Arousal (Thayer)
• Multidimensional models
– Energetic and Tense Arousal, and Hedonic
Tone (Matthews)
– Hierarchical models (Watson and Tellegen)
Multiple sources of data-1
• Between subject “snap shots”
<--
– Adjective check lists
– Rating scales
• Within subject “diary” studies
– Very high frequency/continuous studies
– High frequency sampling
– Low frequency sampling
Between subject “snap shots”
• Adjective check lists (“I feel …”)
– Energetic
– Sleepy
– Tense
– Calm
• Rating scales (“I feel …”)
– very happy, happy, sad, very sad (Bipolar)
– not at all, somewhat, very happy (Unipolar)
Typical between subject structure
• Measures
– Motivational State Questionnaire (MSQ)
• 68-72 Item rating (0-3) scale
– Items taken from
• Thayer’s Activation-Deactivation ACL
• Watson and Clark’s PANAS
• Diener and Larson Circumplex measures
– Example Items:
• Alert
• Lively
Sleepy
Tired
Tense
Anxious
Calm
Relaxed
Typical between subject structure
• Subjects
– >2700 participants aggregated from > 40 studies
of personality and cognition at NU over 6 years
• Method
– Baseline measurements taken using the MSQ (R)
– Studies done from 5:30 am to 10:30 pm
– (additional analyses of effects of caffeine,
exercise and movies on affect-not reported here)
Typical between subject structure
• Results
– Factor extraction using PF and ML
– Factor number determined by Very Simple
Structure (VSS)
– Clear 2 factor solution
– Differential skew leads to suggestions of more
factors
• 4 cluster solution representing +/- ends of two
dimensions
Very Simple Structure => 2 Factors
2 Dimensions of Affect
1.0
FRUS TRATDIS TRES S
UPS ET
UNHAPPY S AD
TENS E
DEPRES S ED BLUE ANGRY CLUTCHED
NERVO US
GLO OMY S O RRY
AFRAID
S CARED
IRRITABL AS HAMED FEARFUL
ANXIOUS
GROUCHY HOS TILE
GUILTY
0.5
LO NELY
JITTERY INTENS E
AS TO NIS H
S URPRIS E
DULL
S LUGGIS H
S LEEPY
TIRED
DROWS Y
INACTIVE
0.0
DETERMIN
INSP IRED
ARO USED
US
STRONG VIGORO
EXC ITED
BO RED
Q UIET
ACTIVE
ELATED
Q UIES CEN FULL_O F_
IDLE
PLACID
S TILL
ALERT
ATTENTIV
ENERGETI
W IDEAW AK
LIVELY
ENTHUS IA INTEREST
W AKEFUL
PRO UD
DELIGHTE
C IABLE
CHEERFUL SOPLEASED
WARMHEAR
C ONFIDEN
HAPPY
SATISFIE
TRANQUIL AT_RES T
S ERENE
CONTENT
CALM
RELAXED
AT_EAS E
-0.5
-1.0
-0.5
0.0
0.5
Energetic Arousal/Positive Affect
1.0
2 Dimensions of Affect
Representative MSQ items
(arranged by angular location)
Item
energetic
elated
excited
anxious
tense
distressed
frustrated
sad
irritable
sleepy
tired
inactive
calm
relaxed
at ease
attentive
enthusiastic
lively
EA-PA
0.8
0.7
0.8
0.2
0.1
0.0
-0.1
-0.1
-0.3
-0.5
-0.5
-0.5
0.2
0.4
0.4
0.7
0.8
0.9
TA-NA Angle
0.0
1
0.0
2
0.1
6
0.6
70
0.7
85
0.8
93
0.8
98
0.7
101
0.6
114
0.1
164
0.2
164
0.0
177
-0.4
298
-0.5
307
-0.5
312
0.0
357
0.0
358
0.0
360
Multiple sources of data-2
• Between subject “snap shots”
– Adjective check lists
– Rating scales
• Within subject “diary” studies <-– Very high frequency/continuous studies
– High frequency sampling
– Low frequency sampling
Within subject diary studies-1
• Very High Frequency (continuous)
measurements
– Physiological assays
• Cortisol
• Body temperature
<--
– Core body temperature collected for ≈ 2 weeks
– Data taken by aggregating subjects from multiple studies
conducted by Eastman and Baehr on phase shifting by
light and exercise
Body Temperature as f(time of day)
(Baehr, Revelle & Eastman, 2000)
90
80
60
50
40
Temperature (C΅)
70
3 8 .0
S leep
3 7 .0
L a t est T
30
20
10
T em p .
at bed
East
West
T North
em p .
at w ak e
M I NS
(N = 11)
3 6 .0
16 :0 0
20 :0 0
00 :00
04 :00
08 :0 0
12 :00
T im e ( h ou r s)
0
1st Qtr
2nd Qtr
3rd Qtr
4th Qtr
16 :0 0
Morningness/Eveningness and BT
(Baehr, Revelle and Eastman, 2000)
90
80
70
Temperature (ΑC)
60
3 7.5
50
40
M -types
3 7.0
E-types
30
20
3 6.5
10
0
3 6.01st Qtr
2nd Qtr
16:00
20:00
00:00
3rd Qtr
4:00
4th Qtr
8:00
12:00
16:00
Tim e (ho u rs )
= Average Sleep
East
West
North
Multiple sources of data-3
• Between subject “snap shots”
– Adjective check lists
– Rating scales
• Within subject “diary” studies
– Very high frequency/continuous studies
– High frequency sampling
<-– Low frequency sampling
Within subject diary studies-2
• Measures
– Check lists
– Rating scales
• High frequency sampling <-– Multiple samples per day
• Low frequency sampling
– Once a day
– Sometimes at different times
High frequency measures of affect
• Measures taken every 3 hours during
waking day for 6-14 days
• Paper and pencil mood ratings
– Short form of the MSQ -- Visual Analog Scale
– Sampled every 3 hours
• Portable computer (Palm) mood ratings <-– Short form of the MSQ
– Sampled every 3 hours
Palm Affect Survey
Palm affect and activity survey
Traditional measures
• Mean level
–
–
–
–
Energetic arousal
Tense arousal
Positive affect
Negative affect
• Variability
• Correlation across measures (Synchrony)
Phasic measures of affect
• Fit 24 hour cosine to data
– Iterative fit for best fitting cosine
– Permutation test of significance of fit
• Measure
– Fit (coherence)
– Amplitude
– Phase
Affective rhythms can differ in phase
(simulation - double plotted to show rhythm)
William Revelle:
Should this come before
Phase differences
the 24 hour slide
of simulated daily data
Differences in coherence (fit) simulated daily data
Phase and Coherence differences
(simulated data -- double plotted)
Multi-level analysis of patterns of
affect across time-1: Method
• Within subject estimates of basic parameters
–
–
–
–
Level
Scatter (variability)
Phase
Coherence (fit)
• Between subject measures of reliability
– Week 1/Gap/Week 2
Multi-level analyses of affect-2:
1-2 week Test-Retest Reliability
VAS-1 VAS-2
Palm
Energetic Arousal
.67
.81
.82
Tense Arousal
.68
.57
.81
Fit EA
.55
.41
.07
Fit TA
.61
.25
.17
Phase EA
.69
.36
.58
Phase TA
.39
.25
.36
EA -TA Synchrony .63
.48
.35
Affective rhythms and
cognitive performance-1
• Design: High frequency diary study of
affect combined with a low frequency study
of reaction time
• Subjects: 28 NU undergraduate volunteers
• Method:
– 1 week diary study 5 times a day
– Simple reaction time once a day at 5 different
times using a Mac program at home
Affective rhythms and
cognitive performance-2
• Low negative correlations of RT with
concurrent measures of Energetic Arousal
• Stronger negative correlations of RT with
Cosine fitted Energetic Arousal
• => Diurnal variation in RT may be fitted by
immediate and patterns of arousal
Between and Within Subject
Measures of Affect
• Personality traits and affective states
• Between versus within measures of mood
and affect
• Alternative within subject measuresstudying the “tides of emotion”
• Applications to cognitive performance
• More information found on links from the personality
project -- http://www.personality-project.org and the
Personality-Motivation lab http://pmc.psych.nwu.edu