Beyond “happy, angry, or sad?”: Age-of-poser and age-of-rater

16 Pages·2012·0.33 MB·English

This article was downloaded by: [MPI Max-Planck-Institute Fur Bildungsforschung]
On: 24 January 2012, At: 00:18
Publisher: Psychology Press
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office:
Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Cognition & Emotion
Publication details, including instructions for authors and subscription
information:
http://www.tandfonline.com/loi/pcem20
Beyond “happy, angry, or sad?”: Age-of-poser
and age-of-rater effects on multi-dimensional
emotion perception
Michaela Riediger a , Manuel C. Voelkle b , Natalie C. Ebner c & Ulman
Lindenberger b
a Max Planck Research Group Affect Across the Lifespan, Max Planck
Institute (MPI) for Human Development, Berlin, Germany
b Center for Lifespan Psychology, Max Planck Institute (MPI) for Human
Development, Berlin, Germany
c Department of Psychology, Yale University, New Haven, CT, USA
Available online: 03 Mar 2011
To cite this article: Michaela Riediger, Manuel C. Voelkle, Natalie C. Ebner & Ulman Lindenberger (2011):
Beyond “happy, angry, or sad?”: Age-of-poser and age-of-rater effects on multi-dimensional emotion
perception, Cognition & Emotion, 25:6, 968-982
To link to this article: http://dx.doi.org/10.1080/02699931.2010.540812
PLEASE SCROLL DOWN FOR ARTICLE
Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions
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.
The publisher does not give any warranty express or implied or make any representation that the
contents will be complete or accurate or up to date. The accuracy of any instructions, formulae,
and drug doses should be independently verified with primary sources. The publisher shall not
be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or
howsoever caused arising directly or indirectly in connection with or arising out of the use of this
material.
COGNITIONANDEMOTION
2011,25(6),968"982
Beyond ‘‘happy, angry, or sad?’’: Age-of-poser and
age-of-rater effects on multi-dimensional
emotion perception
Michaela Riediger1, Manuel C. Voelkle2, Natalie C. Ebner3, and
2
1 Ulman Lindenberger2
0
2
y 1MaxPlanckResearchGroupAffectAcrosstheLifespan,MaxPlanckInstitute(MPI)forHuman
r
a
u Development, Berlin, Germany
n
a 2Center for LifespanPsychology, MaxPlanck Institute (MPI) for HumanDevelopment, Berlin,
J
4 Germany
2
8 3Department of Psychology, YaleUniversity, NewHaven, CT,USA
1
0:
0
at
] Young, middle-aged, and older raters (N!154) evaluated 1,026 prototypical facial poses of
g
n neutrality,happiness,anger,disgust,fear,andsadnessstemmingfrom171young,middle-aged,and
u
h older posers. The majority of poses were rated as multi-faceted, that is, to comprise several
c
rs expressionsofvaryingintensities.Consistentwiththenotionofage-relatedincreasesinnegativity"
o
sf avoidance/positivity effects, crossed-random effects analyses showed an age-related decrease in the
g
n attributions ofnegative, but not positive and neutral,target expressions (thatthe poserintended to
u
d show),andanage-relatedincreaseintheattributionsofpositiveandneutral,butnotnegative,non-
Bil targetexpressions(thattheposersdidnotintendtoshow).Expressionsweremoredifficulttoread
r
u theoldertheposers,particularlyformaleposers.Theseage-of-posereffectswereindependentofthe
F
e valenceoftheexpression,butpartlydifferedacrossagegroupsofraters.Thestudysupportstheidea
ut ofmulti-dimensionality and age-dependency ofemotion perception.
stit
n
-I Keywords: Facial expressions; Emotion recognition; Age of rater; Age of poser; Own-age effect;
k
c Multi-dimensional rating.
n
a
Pl
-
x
a
M
PI Emotional displays serve communicatory func- necessarily spontaneous expressions of an indi-
M
tions. For example, people may display anger to vidual’s experience, but are often intentionally
[
y signal the seriousness of their position in an used (or ‘‘posed’’) to convey social information
b
d interpersonal argument, or put on a sad face to or to influence interaction partners (Buck &
e
d
a discourage another person from behaving in an VanLear, 2002). Whether or not these commu-
o
nl undesired way. Thus, emotional displays are not nication goals are met depends, among other
w
o
D
Correspondenceshouldbeaddressedto:MichaelaRiediger,MPIforHumanDevelopment,Lentzeallee94,D-14195Berlin,
Germany.E-mail:[email protected]
WethankColinBauer,DulceErdt,LuaraFerreiradosSantos,GiorgiKobakhidze,PhilippMo¨ller,andtheCOGITO-Team
fortheirassistanceinconductingthisstudy.
968
#2011PsychologyPress,animprintoftheTaylor&FrancisGroup,anInformabusiness
http://www.psypress.com/cogemotion DOI:10.1080/02699931.2010.540812
MULTI-DIMENSIONALEMOTIONPERCEPTION
things, on how interaction partners perceive the Lindenberger, 2009). In addition, there is some
expression. evidence that people interpret facial emotional
In this article, we demonstrate that reading expressions of other individuals in more complex
emotional poses is among the interpersonal thancategoricaltermswhengiventheoptiontodo
phenomena at the intersection of emotion and so (Hall & Matsumoto, 2004; Phillips & Allen,
cognition that cannot be adequately understood 2003).
without taking the age-group memberships of both The possibility that facial expressions may
the perceiving and the expressing person into require complex interpretations is also acknowl-
account. We report a study that investigated how edged in theoretical frameworks on emotion
2
1 young, middle-aged, and older raters evaluated a expression and perception. Proponents of the
0
2 large number of prototypical facial poses from ‘‘basic-emotions approach’’ (e.g., Ekman, 1992),
y
ar young, middle-aged, and older posers on each of for example, suggest that emotional experiences
u
n thesixdimensionsofneutrality,happiness,anger, maygobeyondbasicemotionsandcomprisemore
a
4 J disgust, fear, and sadness. The purposes of this complexaffectiveexperiences,referredtoas‘‘emo-
2
8 studyweretodemonstrate:(i)thatraterstypically tionalplots’’and‘‘emotionalblends’’,whichshould
0:1 interpret emotional poses in more complex than also be reflected in facial displays. Similarly,
0 mere categorical terms; (ii) how such multi- proponents of the ‘‘dimensional"contextual ap-
at
] dimensional evaluations of emotional poses differ proach’’ (e.g., Russell & Bullock, 1986) maintain
g
n dependingontheageoftheposingandtherating that people may evaluate facial expressions in a
u
ch person; and (iii) that these effects are influenced complexmanner.Theyproposethatfacialexpres-
s
or by the valence of the attributed expressions. sions are initially perceived in terms of the extent
f
gs of pleasure and arousal expressed. Then, in order
n
u toverbalisethemeaningoftheexpression,theyare
d Reading emotional poses: Categorical or
Bil multi-dimensional? associated with specific emotion categories. This
r latter step may result in the attribution of one or
u
e F The most frequently used paradigm in the in- multipleemotioncategories(ofvaryingprototypi-
ut vestigation of emotional poses presents partici- cality) toanemotional expression.
stit pants with prototypical facial expressions of Based on these considerations, we predicted
n
-I intense basic emotions and asks them to select thatraterswouldnottypicallyinterpretemotional
k
nc the single best matching emotion from a list of poses categorically, but would rather interpret
a
Pl alternatives. This ‘‘forced-choice approach’’ as- poses in more complex terms when given the
x- sumes that people interpret emotional poses optiontodoso.Inaddition,wehypothesisedthat
a
M categorically.Weproposethatthismaynotalways suchmulti-dimensionalevaluationsoffacialposes
PI be the case, and that people may often interpret would differ between raters and posers from
M
[ emotional poses in more complex terms. This different adult age groups, and that the valence
by assumption is based on prior evidence that emo- of attributed expressions would play a role in this
ed tional experiences are often multi-faceted (e.g., respect, as will be elaborated next.
d
a Hemenover & Schimmack, 2007). Typical ap-
o
nl proaches to measuring emotional experiences ac-
w Age differences in the interpretation of
o count for this potential complexity by presenting
D emotional poses: How does valence matter?
participants with a list of emotion words. Partici-
pants arethen askedto indicate, for each of these Prior evidence suggests that older adults are less
emotions,howintenselytheyarecurrentlyfeeling accurate at decoding facial emotional pose than
this way. People from various age groups often young adults. However, this age-related decline
describe their momentary emotional experiences appearstobemoreconsistentandpronouncedfor
as comprising several affect facets of varying expressions of anger, fear, and sadness than for
intensity (e.g., Riediger, Schmiedek, Wagner, & expressionsofhappiness,surprise,anddisgust(see
969
COGNITIONANDEMOTION,2011,25(6)
RIEDIGERETAL.
Ruffman, Henry, Livingstone, & Phillips, 2008, theattributionsofotherexpressionsthattheposer
for a meta-analysis). did not intend to show (non-target expressions).
The mechanisms underlying age differences in Specifically,wehypothesisedthatagedifferencesin
reading facial expressions are not yet well under- themotivationtoattendtoemotionalinformation
stood.Theyappeartoberelativelyindependentof ofdifferentvalenceleadtoanage-relateddeclinein
age-relateddeclinesinfluid-cognitivefunctioning the attribution of negative, but not positive or
(Phillips, MacLean, & Allen, 2002; Sullivan & neutral, target expressions, coupled with an age-
Ruffman, 2004; but see Keightley, Winocur, relatedincreaseinattributionsofpositive,butnot
Burianova, Hongwanishkul, & Grady, 2006). negativeorneutral,non-targetexpressions.
2
1 Some researchers have argued that age-related
0
2 gradual atrophy of brain structures involved in
y How does age of poser affect the
ar emotion processing, or reduced availability of
u interpretation of emotional poses?
n neurotransmitters leading to decreased activation
a
4 J oftheseareas,maycontributetoagedifferencesin Wefurtherhypothesisedthatage-relatedchanges
2
8 reading emotional expressions (e.g., Ruffman in facial features and skin texture of posers make
0:1 et al., 2008). emotional expressions of older as compared to
at 0 Another line of argument maintains that adult young posers more difficult to recognise. We
] age differences in the identification of emotional therefore expected raters to attribute less target
g
n posesreflectage-relatedshiftsinthemotivationto and more non-target expressions the older the
u
ch process emotional information (e.g., Williams posers, irrespective of the valence of the expres-
s
or et al., 2006). This position derived from evidence sion.Thispredictionisconsistentwithfindingsof
f
gs ofanage-relatedincreaseinpreferentialattention the few available studies involving posers of
n
u to positively, or away from negatively, valenced various ages (Borod, Yecker, & Brickman, 2004;
d
Bil information(seeCarstensen&Mikels,2005,fora Ebner, He, & Johnson, 2011 this issue; Ebner &
r review).Theseeffectshavebeeninterpretedwithin Johnson, 2009; Malatesta, 1987).
u
e F theframeworkofsocioemotionalselectivitytheory, We also expected these age-of-poser effects to
ut which proposes that increasing awareness of lim- be less pronounced for older raters. We derived
stit itedremaininglifetimeshiftsolderadults’motiva- this prediction from evidence that people are best
n
-I tiontowardswantingtomaximisetheiremotional at interpreting emotional expressions by indivi-
k
nc well-being. Age differences in the processing of duals of their own nationality, ethnicity, or
a
Pl positive and negative information are seen as cultural group (see Elfenbein & Ambady, 2002,
x- instrumental inthisrespect(Carstensen,Fung,& for a meta-analysis). Several mechanisms to
a
M Charles,2003). explainthesein-groupeffectshavebeendiscussed,
PI These age-differential cognitive styles in the such as a better knowledge base for interpreting
M
[ processingofvalencedinformationmightalsobeat facialexpressionsconveyedbyindividuals ofone’s
by work in the multi-dimensional interpretation of own culture, or a higher motivation to attend to,
ed emotional poses. Here, they might result in age- andprocess,expressionsofindividualsthatbelong
d
a related differences in attributing expressions that to a cultural group with which oneself-identifies.
o
nl the poser intended to show and/or in attributing These in-group advantages in expression identifi-
w
o expressionsthattheposerdidnotintendtoshow. cation also extend to other group memberships,
D
For example, when interpreting an angry pose, such as university affiliations or shared interests
negativity avoidance/positivity effects might be (Thibault, Bourgeois, & Hess, 2006). Thus, it
reflected in ascribing a lower intensity of anger stands to reason that age-group membership may
and/orinattributingsomehappiness.Wetherefore have similar effects.
expected systematic valence-specific age differ- Empirical evidence regarding own-age advan-
encesin(i)theattributionsofexpressionsthatthe tages in expression identification is still scarce.
poserintendedtoshow(targetexpressions)and(ii) Malatesta and colleagues (1987) found that
970
COGNITIONANDEMOTION,2011,25(6)
MULTI-DIMENSIONALEMOTIONPERCEPTION
expression recognition rates were lower when majority of studies compared young and older
decoding spontaneous facial expressions of older adults only. Little is known about how middle-
adultsascomparedtodecodingspontaneousfacial aged adults interpret emotional poses.
expressions of young adults. This difference was The present study sought to address these
lesspronouncedinolderthaninyoungraters,but methodological limitations of earlier investiga-
older raters did not perform better than young tions. Specifically, we tested the following hy-
raters when decoding emotional expressions of potheses: Participants would attribute more
older adults. However, the numbers of expressers complexemotionalexperiencestoemotionalposes
and raters per age group in this study were small. (insteadofinterpretingthemincategoricalterms)
2
1 Two other investigations found nosupport for an when given the option to do so in a multi-
0
2 own-age advantage in decoding posed facial dimensional response format (Hypothesis 1). In
y
ar expressions (Ebner et al., 2011 this issue; Ebner line with assumptions of age-related increases in
u
n &Johnson,2009).Bothofthesestudies,however, negativity-avoidance/positivityeffects,wefurther-
a
4 J used forced-choice paradigms, and primarily fo- more expected systematic valence-specific age
2
8 cusedontheidentificationoftarget,butnotnon- differences in the attribution of both target and
0:1 target, expressions. Furthermore, only young and non-targetexpressions.Inparticular,wepredicted
0 older,butnotmiddle-aged,adultswerecompared. an age-related decrease in correct attributions of
at
] The present study sought to extend this previous negative expressions (anger, disgust, fear, and
g
n research, as summarised next. sadness), but not in correct attributions of happi-
u
ch ness or neutrality (Hypothesis 2a). In addition, we
s
or expectedanage-relatedincreaseintheattribution
f The present study
gs ofhappinesstoexpressionsthatwerenotintended
n
u Taken together, older adults, as compared to to show happiness, while we expected no age-
d
Bil young adults, have been found to be less accurate related increase in the attribution of negative
r at identifying emotional poses. In most previous expressions (anger, fear, disgust, and sadness) or
u
e F studies, however, participants evaluated a rela- neutrality to expressions that did not target these
ut tively small number of poses, and were limited to emotions (Hypothesis 2b). Furthermore, we hy-
stit choosing one single best matching emotion from pothesised that raters would attribute less target,
n
-I several response options. This forced-choice ap- and more non-target, expressions to poses from
k
nc proach assumes that people interpret emotional middle-aged or older as compared to young
a
Pl poses in categorical terms, when in fact they may posers, irrespective of the valence of the expres-
x- attribute several simultaneous emotions to one sions (Hypothesis 3). Finally, we expected that
a
M facialemotionexpression.Therelativelysmallsets these latter age-of-poser effects would be attenu-
PI of stimuli may also limit the generalisability of ated the older the raters (Hypothesis 4).
M
[ these prior findings. Moreover, previous studies
y
b have typically analysed the ‘‘hit rate’’ of responses
ed (i.e., correct recognition of the expression the METHOD
d
a poserintendedtoshow).However,agedifferences
o Participants
nl may also be evident in systematic differences in
w
o attributing expressions that were not intended by The sample consisted of 154 (n!76 female)
D
theposer.Furthermore,whileearlierstudiesoften raters from three age groups: Young (20 to 31
varied the age of the persons who rated the years, n!52), middle-aged (44 to 55 years,
expressions, they did not typically vary the age n!51), and older (70 to 81 years, n!51) adults.
of the persons who posed the expressions. Inter- Men and women were approximately equally
pretations of emotional poses, however, may be distributed within each age group. All raters
influenced not only by the age of the perceiver, were German speaking and Caucasian. The latter
but also by the age of the poser. Finally, the criterion was included because the to-be-rated
971
COGNITIONANDEMOTION,2011,25(6)
RIEDIGERETAL.
emotional poses also were from Caucasian weretoldthattheywouldseefaces,oneatatime,
individuals, who represent the vast majority of andbeaskedtoindicate(amongotherratings)the
the population in Germany. Participants were extenttowhicheachfaceexpressed:(a)neutrality;
recruited through the Max Planck Institute for (b) happiness; (c) anger; (d) disgust; (e) fear; and
Human Development’s subject pool and adver- (f) sadness.
tisements in local newspapers. Self-reported phy- Facial expressions and rating dimensions were
sical functioning in the sample was good (single presented in randomised order. Response options
item, ‘‘How would you describe your current rangedfrom0(doesnotapplyatall)to100(applies
general physical functioning?’’, scale 1 to 8, with completely),andwereselectedbyadjustingaslider.
2
1 8!excellent;M!5.5,SD!1.5)anddidnotvary Having rated a given expression on all six
0
2 significantly between age and gender groups (all expression dimensions prompted the next facial
y
ar ps!.05). Participants’ visual-motor processing expression to appear. Rating sessions were termi-
u
n speed as assessed with the Digit-Symbol-Sub- nated after 100minutes each.Participants rated a
a
4 J stitution Test (Wechsler, 1981) was comparable median number of 1,024 faces in an average of
8 2 to typical performance levels (with young women 11.28 (SD!4.7) test sessions. All participants
0:1 and men, MYW!66.3, SD!11.1; MYM!64.0, were reimbursed for study participation. The
0 SD!9.6, scoring higher than middle-aged, study was approved by the ethics committee of
g] at MMW!46.0,SD!9.1;MMM!48.5,SD!14.4, the Max Planck Institute for Human Develop-
n and older women and men, M !44.8, ment.
u OW
ch SD!10.7; MOM!47.7, SD!12.1).
s
r
o
f Crossed-random effects analyses
s
ng Posed facial expressions
u For the majority of our analyses we used crossed-
d
Bil Facial expressions were taken from the FACES random effects models; an approach that has
r database (Ebner, Riediger, & Lindenberger, recently been receiving increased attention (e.g.,
u
e F 2010). In two parallel sets, the database contains Baayen, Davidson, & Bates, 2008; Hoffman &
ut 2,052 pictures of 171 Caucasian posers, each Rovine, 2007). Crossed-random effects models
stit displayingneutral,happy,angry,disgusted,fearful, offer a number of advantages over traditional
n
-I and sad facial expressions. Posers were young (19 analysis of variance models, such as the examina-
k
nc to 31 years, n!58), middle-aged (39 to 55 years, tion of associations among study variables within
Pla n!56), or older (69 to 80 years, n!57) adults, (rather than between) raters, the possibility of
x- with approximately equal numbers of men and estimating the extent to which effects vary
a
M womenineachofthesegroups.Thecreationofthe between posers and/or raters, and the maximisa-
PI database involved a standardised production and tionofpowerthrough theuse offull-information
M
[ selection procedure to obtain maximally prototy- maximum likelihood (Baayen et al., 2008;
by pical facial expression from each poser (see Ebner Hoffman & Rovine, 2007). The conceptual
ed etal.,2010,fordetails).Figure1presentsexamples approach of crossed-random effects models is
d
a of facialexpressions in the three age groups. similar to that of multilevel models. In contrast
o
nl to multilevel models, which assume a nested
w
o (hierarchical) data structure, however, crossed-
D Rating procedure
random effects models permit independent
Eachraterwasassignedtooneofthetwoparallel sources of variance, which in our case refer to
sets, each containing 1,026 pictures.1 First, parti- the posers and raters.
cipantswereinformedaboutthetestingprocedure Using the lme4(Bates & Maechler,2009) and
and signed a consent form. Next, participants languageR packages (Baayen, 2009), we specified
1Stimulussetdidnotexplainanyvarianceintheanalysesreported.Wethereforecollapsedacrossthisfactorinallanalyses.
972
COGNITIONANDEMOTION,2011,25(6)
MULTI-DIMENSIONALEMOTIONPERCEPTION
2
1
0
2
y
r
a
u
n
a
J
4
2
8
1
0:
0
at
]
g
n
u
h
c
s
r
o
f
s
g
n
u
d
Bil Figure1. Samplefacialexpressionsofposersfromthethreeagegroups.[Toviewthisfigureincolour.pleasevisittheonlineversionofthis
r Journal].
u
F
e
ut
nstit ssiixxseexpparreastseiomnoddiemlsewnsitihonpsaratsicdipepanentsd’eranttinvgasrioafbltehse. nnoalr.mLalikdeiwstirsieb,uttihoen aenrrdorweterermtrewataesdaasssuomrthedogoto-
I
k- Predictorsincludedthetargetexpression(dummy be normally distributed with zero mean and
c
n codedwith1indicatingthatthedependentvariable variance s2. Moreover, we included the rater’s
a r
Pl correspondedtothetargetexpression),theposer’s age(dummycodedwithyoungadultsasreference
-
x
a age(dummycodedwithyoungadultsasreference group) as explanatory variable to predict the
M
I group), and their interactions. The intercept was random effects (intercept and slopes).2 We also
P
M allowedtovarybetweenposersandraters,andthe conducted follow-up analyses to explore whether
y [ slopesofthedummyvariablesfortargetexpression the reported effects differed between men and
b
d andposeragegroupwereallowedtovarybetween women. Apart from few exceptions, which are
e
d raters.Allrandomeffectswereassumedtofollowa reportedintheresultssection,thiswasnotthecase.
a
o
nl
ow 2Crossed-random effects model equations: yijk!b0j#b1j (Target expression)#b2j (Middle poser)#b3j (Older poser)#b4j
D (Target expression$Middle poser)#b5j (Target expression$Older poser)#Wk#rijk. Predictors of random coefficients:
b0j!g00#g01 (Middle rater)#g02 (Older rater)#u0j; b1j!g10#g11 (Middle rater)#g12 (Older rater)#u1j; b2j!g20#g21
(Middlerater)#g22(Olderrater)#u2j;b3j!g30#g31(Middlerater)#g32(Olderrater)#u3j;b4j!g40#g41(Middlerater)#g42
(Olderrater);b5j!g50#g51(Middlerater)#g52(Olderrater).Whereyijkistheratingoftheithexpressionstimulusfromthekth
poserbythejthrater,W istherandomeffectforposers,andr istherandomresidualterm.Thefixedintercept(i.e.,theexpected
k ijk
ratingofayoungraterofanon-targetexpressionposedbyayoungposer)isdenotedbyg00,thefixedinterceptofthemthrandom
coefficient (b0j to b5j), by gm0; gml represents the fixed slope for age-of-rater dummy codes as predictors of the mth random
coefficient,andu istherandomresidualtermassociatedwiththemthrandomcoefficientforthejthrater(i.e.,therater-specific
mj
deviationfromthefixedeffectexamined).
973
COGNITIONANDEMOTION,2011,25(6)
RIEDIGERETAL.
RESULTS considerable variations of intercepts between
posers and raters, respectively.
Multi-dimensional expression ratings
In all six models, fixed slopes of target expres-
(Hypothesis 1)
sion were of substantial size (i.e.,!64) and
For M!35.9% (SD!26.0) of rated expressions, significantly different from zero. This indicates
ratersusedexactlyoneratingdimensionandchose that young raters, on average, differentiated well
zeros on all other dimensions, with no significant between expressions that were targeted and those
differences between age groups of raters, F(2, that were not targeted in facial expressions of
2 151)!0.84, p!.434. Thus, consistent with young posers. For example, young raters attrib-
1 Hypothesis1,themajorityofevaluationsreflected uted69.23scalepointsmoreangertoyounganger
0
y 2 more complex, multi-faceted attributions rather expressions than to young non-anger expressions.
ar than mere categorisations into discrete emotions. The variance components for these slopes were
u
n significant in all models, and the random effects
a On average, raters chose values above zero for
4 J M!2.46(SD!1.15)ofthesixratingdimensions coverageratesrevealedsubstantialvariationinthe
2
8 for a given expression, with no significant differ- extent to which raters differentiated between
0:1 ences between age groups, F(2, 151)!1.04, target and non-target expressions.
at 0 p!.357. However, while on average the highest Hypothesis 2a predicted an age-related de-
g] rating on one of the six rating dimensions was crease in the extent to which raters attributed
n negative affect (i.e., anger, disgust, fear, sadness),
u 82.93 (SD!11.35), the average second highest
sch rating was only 21.82 (SD!16.30). Again, raters but not neutrality and happiness, to poses that
or wereintendedtoshowtheseexpressions.Relevant
f from different age groups did not differ in this
ngs respect for the highest, F(2, 151)!0.98, parameter estimates are represented in the inter-
u actions between target expression and age group
d p!.377, and the second highest ratings, F(2,
Bil 151)!0.70, p!.499, respectively. That is, as of raters, indicating whether ratings of target
r expressions differed between middle-aged (or
u expected, raters interpreted facial poses in more
F older) and young raters. In line with our predic-
e complexthanmerelycategoricalterms,butnever-
ut tion, the interactions involving older raters were
stit theless typically made a clear attribution to a significant for anger, disgust, fear, and sadness,
n primary expression.
-I butnotforneutralityandhappiness(seePartAof
k
c Table 1). That is, young and older raters did not
n
a
Pl Valence-specific variations in age-of-rater differ in their attributions of neutrality and
x- effects (Hypothesis 2) happinesstoneutralandhappytargetexpressions,
a
M respectively. Older adults, however, attributed
PI The following results were derived from the significantly less anger, disgust, fear, and sadness
M crossed-random effects models previously intro-
to poses that intended to show these expressions
[
y duced.AsshowninTable1,withtheexceptionof
b thandidyoungraters.Middle-agedratersshowed
ed happiness, the fixed intercepts in these models a similar pattern for two of the four negative
ad were significantly different from zero, but rela- expressions, that is, they attributed significantly
o
nl tively small (B 12 on a scale from 0 to 100). less disgust and sadness to poses that intended to
w
o Thus, on average and with the exception of show these expressions than did young raters.
D
happiness, young raters attributed significant but None of these results differed between male and
small intensities of non-target expressions to female raters (i.e., interactions with gender of
young adults’ poses. Variance components and raters: p!.05).
randomeffectscoverageranges(thelatterindicat- Hypothesis 2b predicted an age-related in-
ing the range of individual intercepts for 95% of crease in the extent to which raters attributed
the young posers and the range of individual positive (i.e., happiness) but not negative affect
intercepts for 95% of the young raters) indicate (i.e., anger, disgust, fear, sadness) or neutrality to
974
COGNITIONANDEMOTION,2011,25(6)
Table1. Crossed-randomeffectsanalysespredictingratingsoffacialexpressionsfromtargetexpression,ageofrater,andageofposer.
2
1
20 Parameters Neutral Happy Angry Disgusted Fearful Sad
y
r
a Fixedeffects
u
n Intercept 4.10**(1.26) 1.27(0.70) 11.72**(1.72) 10.46**(1.39) 10.12**(1.73) 9.12**(1.45)
a
4 J 95%CR(posers) 0to9.79 0to4.33 5.21to18.23 0to16.35 5.59to14.65 0.52to17.72
2 95%CR(raters) 0to21.35 0to10.91 0to35.60 0to29.50 0to34.65 0to28.22
18 Expressiona
0: ! Target 80.90**(2.38) 89.83**(1.65) 69.23**(2.45) 64.44**(2.24) 67.04**(2.60) 67.90**(2.12)
0
at 95%CR 47.08to100 66.28to100 34.62to100 32.79to96.09 30.08to100 38.12to97.68
g] PartA(Hypothesis2a)
n
u Target$
h
c ! Middlerater %2.37(3.38) %2.65(2.34) %4.36(3.49) !6.76*(3.19) %4.91(3.69) !5.97*(3.01)
s
or ! Olderrater %4.72(3.38) %2.66(2.34) !14.10**(3.49) !11.91**(3.19) %7.79*(3.69) !7.16*(3.01)
f
s
g PartB(Hypothesis2b)
un Ageofraterb
d
Bil ! Middlerater 1.20(1.72) 1.15(1.20) 0.71(2.37) %0.84(1.90) %0.51(2.43) %1.27(1.90)
r ! Olderrater 4.40*(1.72) 2.15*(0.96) 2.96(2.37) 1.60(1.90) 2.38(2.43) 1.54(1.90)
u
F PartC-1(Hypothesis3)
ute C Target$
nstit OGN !! MOliddedrlepopsoesrer !!164..8638****((00..5554)) !!13..5732****((00..3300)) !!165..3130****((00..7700)) !!29..2128****((00..6644)) !!11..1989***((00.5.598)) !!126..2355****((00..6655)) M
ck-I ITIO PartC-2(Hypothesis3) ULT
an N Ageofposerc I-D
MPI Max-Pl ANDEMOTIO !! MO9955li%%ddedrCCleRR %%0233....90832788*(tt*oo0(.0631.2.079).716)4 %%0011....27908807*(tt*oo0(.0314.2..330)744) 014...585484***t**o((008...675834)) %%003...038452*((to00..66326.1))2 %002...709*956*(to0(0.53.43.69)3) %231...706605***t**o((009...890536)) IMENSIONAL
ded by [ N,2011,2 PT!!aarrtgMOeDtli$-dd1edMrl(eHriadrytadeptrloeetrhpeosisser4)$ 10..1374((00..7787)) 00..3692((00..4433)) 24..6831****((00..9999)) %%11..0247((00..9911)) %10..0164((00..8833)) %%10..7942((00..9933)) EMOTIO
a 5 N
wnlo (6) T!argMeti$ddOleldrearteproser $ %0.66(0.77) 0.23(0.43) 1.42(0.99) %0.06(0.91) 0.43(0.83) !2.82**(0.93) PER
Do 9 ! Olderrater %0.99(0.77) 1.92**(0.43) 4.70**(0.98) 1.29(0.91) %0.06(0.82) !2.38**(0.92) CEPT
7 IO
5 N
9 R
7 IE
6 D
2 IG
1 E
0 R
y 2 C ET
nuar OGN AL.
a IT
4 J IO
2 N Table1(Continued)
8 A
1 N
0: D Parameters Neutral Happy Angry Disgusted Fearful Sad
0 E
ng] at MOTIO PMairdtdDle-2po(sHery$pothesis4)
hu N, ! Middlerater %0.06(0.44) %0.24(0.21) %0.35(0.42) %0.40(0.40) %0.15(0.34) !0.78*(0.40)
sc 20 ! Olderrater !1.15**(0.44) %0.13(0.21) %0.34(0.42) %0.26(0.40) 0.03(0.34) %0.16(0.40)
r 1
gsfo 1,25 O!ldeMrpidodseler$rater 1.02(0.79) %0.37(0.35) !1.13*(0.57) %0.60(0.51) 0.29(0.45) %0.46(0.66)
un (6) ! Olderrater %0.32(0.79) !0.76*(0.34) !2.35**(0.57) %0.36(0.51) 0.78(0.44) 0.61(0.66)
d
Bil Variancecomponents
r Interceptd
u
F ! Posers 8.08** 2.34** 10.60** 8.66** 5.12** 18.51**
ute ! Raters 74.39** 23.24** 142.61** 90.64** 150.46** 91.16**
nstit S!lopTesaerget 285.97** 138.81** 299.54** 250.44** 341.57** 221.67**
-I ! Middleposer 2.24** 0.28** 0.26 0.55 B.01 0.57
k
nc ! Olderposer 13.18** 2.20** 4.00** 3.01** 2.02** 7.32**
Pla ! Residualf 314.31** 96.65** 510.02** 432.47** 359.35** 448.98**
-
x
a Notes:Resultsfromcrossed-randomeffectsmodelsestimatedusingthelme4(Bates&Maechler,2009)andlanguageRpackages(Baayen,2009).Boldfontindicatessignificantresults.
M
Valuesforfixedeffectsreportedintable:parameterestimates(standarderrors).SignificancelevelsforfixedeffectswereestimatedusingthepvalsfunctiondefinedinthelanguageR
I
P package.Significancelevelsforvariancecomponentsweredeterminedbyfittingthemodelwithandwithouttherespectivevariancecomponentsandcomparingthequalityoffitsusing
M
likelihoodratiotests(Baayenetal.,2008).95%CR!95%coveragerangefortherandomeffect(variation)aroundthefixedeffect(fixedparameterestimate92"randomparameter
[
y estimate;Hoffman&Rovine,2007).Middlerater/poser!middle-agedrater/poser.aDummycoded,referencegroupnon-targetexpression.bDummycoded,referencegroupyoung
b
d rater.cDummycoded,referencegroupyoungposer.dEstimatedvariationofinterceptsbetweenposersandbetweenraters.eBetween-ratervarianceslopes.fRemainingwithin-person
de variance.*p 5.05;**pB.01.
a
o
nl
w
o
D