Deficits in facial affect recognition among antisocial populations: A meta-analysis

13 Pages·2008·0.43 MB·English

This article was published in an Elsevier journal. The attached copy
is furnished to the author for non-commercial research and
education use, including for instruction at the author’s institution,
sharing with colleagues and providing to institution administration.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
ARTICLE IN PRESS
NeuroscienceandBiobehavioralReviews32(2008)454–465
www.elsevier.com/locate/neubiorev
Review
Deficits in facial affect recognition among antisocial populations:
A meta-analysis
(cid:2)
Abigail A. Marsh , R.J.R. Blair
MoodandAnxietyProgram,NationalInstituteofMentalHealth,NationalInstitutesofHealth,15kNorthDrive,MSC2670,Bethesda,MD20892,USA
Received15May2007;receivedinrevisedform22August2007;accepted23August2007
Abstract
Individualswithdisordersmarkedbyantisocialbehaviorfrequentlyshowdeficitsinrecognizingdisplaysoffacialaffect.Antisociality
may be associated with specific deficits in identifying fearful expressions, which would implicate dysfunction in neural structures that
subserve fearful expression processing. A meta-analysis of 20 studies was conducted to assess: (a) if antisocial populations show any
consistent deficits in recognizing six emotional expressions; (b) beyond any generalized impairment, whether specific fear recognition
deficitsareapparent;and(c)ifdeficitsinfearrecognitionareafunctionoftaskdifficulty.Resultsshowarobustlinkbetweenantisocial
behavior and specific deficits in recognizing fearful expressions. This impairment cannot be attributed solely to task difficulty. These
results suggest dysfunction among antisocial individuals in specified neural substrates, namely the amygdala, involved in processing
fearfulfacial affect.
r2007Elsevier Ltd. Allrights reserved.
Keywords:Facialaffect;Fearfulexpression;Antisocial;Amygdala;Psychopathy;Emotion;Meta-analysis
Contents
1. Introduction . .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. 455
2. Method .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. 456
2.1. Literature search.. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. . 456
2.2. Study characteristics .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. . 456
2.3. Statistical analyses . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. . 457
3. Results. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. 458
3.1. Doantisocial populations show deficits in recognizing the sixemotional expressions?. .. .. .. .. . .. .. .. .. .. .. . 458
3.1.1. Non-parametric analysis . .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. . 458
3.1.2. Parametric analysis:percent accuracy. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. . 459
3.1.3. Parametric analysis:effect size. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. . 459
3.2. Are observed deficits greater for thefear facialexpression thanfor otheremotionalexpressions?. . .. .. .. .. .. .. . 459
3.2.1. Non-parametric analysis . .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. . 459
3.2.2. Parametric analysis:percent accuracy. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. . 459
3.2.3. Parametric analysis:effect size. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. . 460
3.3. Are fear recognitiondeficits among antisocial populations attributableto task difficulty? .. .. .. . .. .. .. .. .. .. . 460
3.4. Assessments of moderatorvariables . . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. . 460
3.5. File drawer analysis .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. . 460
4. Discussion .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. 461
(cid:2)
Correspondingauthor.Tel.:[phone];fax:[phone].
E-mailaddress:[email protected](A.A.Marsh).
0149-7634/$-seefrontmatterr2007ElsevierLtd.Allrightsreserved.
doi:10.1016/j.neubiorev.2007.08.003
Author's personal copy
ARTICLE IN PRESS
A.A.Marsh,R.J.R.Blair/NeuroscienceandBiobehavioralReviews32(2008)454–465 455
5. Conclusions . . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. 463
Acknowledgments .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. 463
References .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. .. .. .. . .. .. .. .. .. .. .. . .. .. .. 463
1. Introduction mayinhibitoravertinappropriatebehaviorlikeaggression
(Blair,2003b;Blairetal.,1997;Marshetal.,2005;Walker
Processing facial affect is crucial for socialization and and Leister, 1994). Ethologists find similar aggression-
normal social interaction (Corden et al., 2006; Fridlund, inhibitingeffectsofdistresscuesinotherprimatesandhave
1991). Aggression and other maladaptive antisocial beha- speculated that distress cues evolved for this purpose
viors may result from failure to be appropriately guided by (Preuschoft,2000).Developmental,behavioral,andclinical
thesocialcuesofothers(Blair,2003a;Montagneetal.,2005; research shows that distress cues elicit empathy in those
Walker and Leister, 1994). Some have suggested that who see them (Hoffman, 1987; Marsh and Ambady, 2007;
distress-related cues, particularly fearful expressions, play Nichols, 2001; Preston and de Waal, 2002). Empathy is
an important role in inhibiting antisocial behavior (Blair, generally associated with decreased antisocial behavior
2001; Nichols, 2001; Price et al., 2004). Accordingly, many (Eisenberg, 2000). It has been proposed that distress cues
studies find impairments in processing distress-related cues possess perceptual properties that elicit empathy and
among antisocial populations. However, not all studies find inhibit aggression (Marsh et al., 2005). It has also been
this impairment. It is unclear whether these inconsistencies proposed that fearful and sad expressions serve as social
resultfromdisparatemethodologies,samplepopulations,or reinforcers that condition developing children to avoid
analytic techniques, or, alternately, from the absence of a engaging in the antisocial behaviors that elicit these
strong relationship between fearful affect processing and expressions. This process is specified by the Integrated
antisociality. Better understanding of facial affect recogni- Emotion Systems (IES) model (Blair, 2005).
tiondeficitsassociatedwithantisocialitywouldpermitmore Facial affect processing relieson a distributednetwork of
precise hypotheses to be formulated regarding neurocogni- structuresthatincludesoccipitotemporalcortex(particularly
tivecorrelatesofantisocialbehavior.Wethusconductedthis fusiform gyrus and superior temporal gyrus), anterior
meta-analysis to assess associations between antisociality cingulate cortex, amygdala, and ventromedial prefrontal
and facial affect recognition deficits. We hypothesized that, cortex (Adolphs, 2006; Murphy et al., 2003). Evidence is
beyond any general facial affect recognition deficits, accumulating, however, that beyond this general network
antisocial individuals show specific deficits in processing partially separable neural systems process different expres-
fearful expressions. We also hypothesized that such deficits sions. Detection of fearful expressions relies disproportio-
are not solely attributable to task difficulty. nately on the amygdala, and the detection of disguston the
Emotional facial expressions play an important role in insulaandbasalganglia(Adolphs,2002;Phillipsetal.,2004;
modulating interpersonal behavior. Given this, extensive Murphyetal.,2003).Evidencethatantisocialindividualsare
research has assessed the relationship between facial affect impaired in processing specific expressions like fear or
recognition and psychiatric disorders characterized by disgustcouldfacilitatetargetedresearchintoneurocognitive
interpersonal deficits. Generalized impairments in proces- deficits underlying antisocial behavior.
singfacialaffecthavebeenfoundinantisocialpopulations Antisocial behaviors are those that violate the rights or
(Kropp and Haynes, 1987; Woodbury-Smith et al., 2005; welfare of other individuals. Multiple studies have shown
Zabel, 1979). But such impairments have also been found specific impairments in fearful expression processing in
in disorders like autism, schizophrenia and social anxiety populations who engage in these behaviors (e.g., Blair
disorder (Gross, 2004; Singh et al., 1998; Tremeau, 2006; etal.,2004;BlairandCipolotti,2000;CarrandLutjemeier,
Easter et al., 2005). Myriad factors, including general 2005; Walker and Leister, 1994; Woodbury-Smith et al.,
intelligence, age, attention, verbal ability, and task-specific 2005). These populations include those primarily classified
motivation can be associated with reductions in facial by the presence of antisocial behaviors (e.g., aggressive,
affectrecognitionscores(HerbaandPhillips,2004;Moore, criminal,externalizing,abusive)andthoseclassified onthe
2001). It is thus unsurprising that many clinical popula- basis of both antisocial behaviors and personality traits
tions suffer facial affect recognition deficits. However, it suchasalackofempathyandremorse(e.g.,psychopaths).
means that little clinically or neuropsychologically specific However, not all of these studies find fearful expression
information regarding a particular population’s impair- recognition impairments (Kosson et al., 2002). Many
mentscanbeextractedfromgeneralfacialaffectprocessing studies that find fear recognition deficits assess children
deficits. (Blair and Coles, 2000; Blair et al., 2001; Stevens et al.,
More informative would be evidence for an association 2001). This leaves questions remaining, such as whether
betweenantisocialbehaviorandadeficitinrecognizingone apparentlyspecificdeficitsinrecognizingfearfulexpressions
or more specific expressions. Likely candidates include stem from task difficulty rather than fear-specific neurop-
distresscueslikefearandsadnessbecausetheseexpressions sychological deficits.
Author's personal copy
ARTICLE IN PRESS
456 A.A.Marsh,R.J.R.Blair/NeuroscienceandBiobehavioralReviews32(2008)454–465
Even in healthy populations, fearful expressions are psychopath, callous, aggression, violence, empathy, and
more often misidentified than other expressions. It is sympathy.
important to determine whether task difficulty can fully 2. All articles referenced by usable articles acquired via
account for the link between fear recognition deficits and these searches were examined.
antisociality. If it can, it suggests the deficits can be 3. The Social Science Citation Index was used to check
accounted for by general facial affect processing impair- citations for any usable article that had been found
ments. Such deficits may simply be more obvious for through other methods.
expressions that are more difficult to recognize. If, 4. Unpublished manuscripts were directly solicited. Re-
however, fear-recognition deficits cannot be explained by quests for such manuscripts were placed via the e-mail
task difficulty, this suggests dysfunction in neural struc- list for a psychology research listserve and were made
tures or systems, such as the amygdala, that are particu- directly to several investigators with published research
larly important to fearful expression processing (Adolphs, pertainingtothetopic.Ourownfileswerealsoreviewed
2002; Rapcsak et al., 2000). for preprints and unpublished manuscripts.
Prior research supports the contention that fearful
expression recognition deficits may be associated with both Studiesthatwereincludedsatisfiedthefollowingcriteria:
amygdaladysfunctionandantisocialbehavior.Patientswith
amygdala lesions show deficits in recognizing fearful 1. Participants must have been objectively selected, classi-
expressions (Adolphs et al., 1999; Papps et al., 2003). fied, or assessed on the basis of a pathology, trait, or
Neuroimagingstudiesalsosupporttheroleoftheamygdala behavior that is primarily defined by antisocial traits or
in processing fearful expressions (Corden et al., 2006; behaviors. Under this criteria, participants clinically
Murphy et al., 2003; Phillips et al., 2004; Whalen et al., diagnosed with antisocial personality disorder, conduct
1998). Considerable recent work has implicated impaired disorder, or classified as having externalizing behavior
amygdala functioning in antisocial populations during disorders or psychopathy were included. Also included
affective tasks (Levenston et al., 2000; Patrick, 1994; Blair, were participants selected for showing high levels of
2003a,b). And recent neuroimaging research has supported violenceand/oraggression.Theseincludedparticipantsin
the existence of amygdala dysfunction in antisocial indivi- forensic settings. Participants were also included who
duals(Birbaumeretal.,2005;Kiehletal.,2001;Sterzeretal., were assessed or classed using trait measures of antisoci-
2005).Thisbodyofevidenceisconsistentwiththehypothesis ality, such as psychopathy scales. Not included were
that across studies, antisocial individuals show a consistent participants primarily classed or assessed on traits or
and specific deficit in the recognition of fearful expressions. disordersthatarenotprimarilydefinedbythepresenceof
However, this hypothesis has not yet been tested. antisociality,suchasADHD,autism,orbipolardisorder.
The goal of this meta-analysis was to assess the existing 2. Judgmentsoffacialaffectmusthavebeenmadewherein
literature that has measured facial expression recognition there was an objective criterion for accuracy (i.e., the
abilities in individuals with disorders or behaviors that are stimuli must have been appropriately normed or
defined by antisociality. Such individuals include those validated). Only judgments of the six emotions most
classifiedaspsychopathic,aggressive,criminal,delinquent, frequently assessed in facial affect studies (anger,
or externalizing. We investigated whether (a) individuals disgust, fear, happiness, sadness, and surprise) were
with antisocial behavior show impairments in recognition included in the analyses.
of each of the six expressions; (b) whether this deficit is 3. Participants in the study must otherwise have been
greater for fearful expressions than for other emotional relatively well matched in terms of other relevant
expressions;and(c)whetherdeficitsinfearrecognitionare variables (e.g., age, gender). When multiple comparison
attributable to task difficulty. groupswereincludedinastudy,themosthomogeneous
sample or samples were selected in order to be
2. Method conservative (e.g., in a study comparing psychopathic
inmates with non-psychopathic inmates and healthy
Avarietyofstrategieswereusedtofindrelevantarticles. adults from the community, comparisons between only
Once located, several criteria were applied to determine the former two groups were used).
whether a study could be included in the meta-analysis: 4. Information regarding the accuracy with which indivi-
dual expressions were identified must be available.
2.1. Literature search 5. Participants presenting with brain injuries or axis I
psychiatric disorders were excluded.
Methods used to locate relevant studies included:
1. First, initial computer searches of PsycINFO and 2.2. Study characteristics
PubMed were conducted to retrieve articles; searches
contained terms such as facial expression, facial affect, Twenty articles were identified that met our criteria
emotion recognition, face, antisocial, conduct disorder, (Table 1). These articles included 38 independent samples
Author's personal copy
ARTICLE IN PRESS
A.A.Marsh,R.J.R.Blair/NeuroscienceandBiobehavioralReviews32(2008)454–465 457
of groups differing in antisocial behavior, incorporating observed deficits are greater for the fearful expressions or
1244 research participants. The mean number of partici- extend equally to other expressions; and whether fear
pants per study was 62.2 (SD¼97.3), with a median of recognition deficits can be attributed to task difficulty.
38.5. Fourteen of the 20 studies used the Pictures of Toallowallstudiestofactorintoacommonanalysis,we
FacialAffectstimulusset(EkmanandFriesen,1976).Two first computed non-parametric tests addressing questions
studies (Carr and Lutjemeier, 2005; Stevens et al., 2001) (a)and(b).Althoughlower powered,non-parametrictests
used the Diagnostic Assessment of Nonverbal Accuracy are not disrupted by non-normality of variance and can
(DANVA) (Nowicki and Duke, 1994). The remaining compare different types of data. We next conducted
studies (Dadds et al., 2006; Kropp and Haynes, 1987; parametric analyses to compare group differences in
Montagne et al., 2005; Walker, 1981) used other validated percent accuracy and to compare effect sizes (r) for those
stimulus sets. Eighteen studies employed a multiple-choice studies that provided data sufficient to calculate each.
response format and two (Matheson and Jahoda, 2005; (Most of the studies were included in all analyses.
Walz and Benson, 1996) employed a free-response format However, some studies report percent accuracy or means
(Table 2). thatcanbeconvertedtopercentaccuracy,butnomeasures
of variance, e.g., standard deviation. Other studies report
2.3. Statistical analyses Pearson’s r or values that could be converted to r, e.g.,
F-values, but not mean or percent accuracy scores.) It will
The resultsofthese studies were analyzed witha goal of be noted that degrees of freedom differ across expressions
determining for which expressions recognition deficits are withinsomeanalyseswhenrelevantdatawerenotprovided
associated with antisocial behavior tendencies; whether foreveryexpression.Inallcases,thresholdsignificancewas
Table1
Characteristicsandresultsofofstudiesincludedinthemeta-analysis
Study N Mage %Female Recognitionaccuracy
Anti C Anti C Anti C A D F H Sa Su
Proportioncorrect(C–Anti)
Percentaccuracydata
Psychopathic
BlairandCipolotti(2000) 5 5 48 49 0 0 (cid:2)0.20 0.01 0.17 (cid:2)0.02 0.01 0.01
BlairandColes(2000) 11 10 12.4 12.6 18 40 0.26 0.09 0.22 0.00 0.35 0.04
Blairetal.(2001) 20 31 12.9 12.8 0 0 0.06 0.03 0.26 0 0.12 (cid:2)0.03
Blairetal.(2004) 19 19 33.6 30.6 0 0 (cid:2)0.02 0.06 0.25 0.02 0 0.11
DolanandFullam(2006) 22 27 35.2 35.2 0 0 0.06 (cid:2)0.01 0.11 0.06 0.19 0.08
Kossonetal.(2002) 27 27 34 33 0 0 (cid:2)0.11 0.20 0 0.01 0.01 0.03
Montagneetal.(2005) 16 16 – – 50 44 (cid:2)0.07 0.11 0.24 0 0.06 (cid:2)0.02
Stevensetal.(2001) 9 9 11.7 11.4 0 0 0.16 – 0.28 0.00 0.30 –
Antisocial,notpsychopathic
KroppandHaynes(1987) 20 20 29.3 29.3 100 100 0.33 – 0.36 0.02 (cid:2)0.01 0.23
Loughetal.(2006) 18 13 61 57 11 31 0.38 0.34 0.26 0.06 0.26 0.14
MathesonandJahoda(2005) 19 15 40.1 32.1 42 53 0.08 (cid:2)0.05 0.01 0.07 0.18 0.21
McCownetal.(1986) 40 40 15.4 14.9 0 0 0.11 0.11 0.03 0.07 0.13 0.18
WalkerandLeister(1994) 191 273 15.1 15.2 15 21 0.08 0.02 0.29 0.09 0.07 0.05
Walker(1981) 15 15 11.6 11.7 40 53 0.02 (cid:2)0.04 0.11 0.03 0.10 0.04
WalzandBenson(1996) 18 21 38.6 40.3 0 0 0 (cid:2)0.09 0.15 (cid:2)0.02 0.09 0.10
Effectsizer
Dataavailableonlyaseffectsize(r)
Psychopathic
Daddsetal.(2006)a 76 – 12.1 – 43 – 0.29 0.01 0.28 (cid:2)0.11 (cid:2)0.07 –
Mitchelletal.(2006) 5 5 41.4 43.6 0 0 0.3 0.13 0.74 0.12 0 0.35
Antisocial,notpsychopathic
CarrandLutjemeier(2005)b 19 – 15.3 – 0 – (cid:2)0.10 – 0.43 (cid:2)0.01 0.19 –
Woodbury-Smithetal.(2005) 21 23 35.4 29.7 14 13 (cid:2)0.06 (cid:2)0.21 0.31 (cid:2)0.04 0.06 (cid:2)0.11
Zabel(1979) 40 51 11.5 14.1 33 18 – – 0.34 0.34 – 0.22
Anti¼Antisocial;C¼Comparison.
aMeanrepresentmeansofchild-ratedantisocialbehaviorandcallous-unemotionalsubscales.
br’srepresentmeansofthefollowingcorrelations:(a)childDANVAemotionrecognitionandempathy;(b)childDANVAemotionrecognitionand
violentbehavior;(c)adultDANVAemotionrecognitionandempathy;and(d)adultDANVAemotionrecognitionandviolentbehavior.
Author's personal copy
ARTICLE IN PRESS
458 A.A.Marsh,R.J.R.Blair/NeuroscienceandBiobehavioralReviews32(2008)454–465
Table2
Subjectselectioncriteriaandtaskvariablesacrossstudies
Study Subjectselection Taskvariables
Selectionbasis Criterion Stimulus Expression Stimuli Multiple Responsetime
duration intensity randomized choice limit
BlairandCipolotti PCL-R p20/X30 3s Morphs O O O
(2000)
BlairandColes(2000) PSD p4/X13 3s Hexagon O O O
Blairetal.(2001) PSD o20/428 3s Morphs O O O
Blairetal.(2004) PCL-R o20/X30 3s Morphs O O O
CarrandLutjemeier SDRQ, All Nolimit Fullintensity O
(2005) empathy
Daddsetal.(2006) APSD All 2s Morphs O O
DolanandFullam PCL-SV o17/X17 Nolimit Morphs O O
(2006)
Kossonetal.(2002) PCL-R p20/X30 1s Fullintensity O O
KroppandHaynes Abusiveness History/nohistoryofchild 30s Fullintensity O O
(1987) abuse
Loughetal.(2006) FTDdiagnosis, Diagnosis/healthy 23s Morphs O
IRI
MathesonandJahoda CCB 0/X4incidents Nolimit Fullintensity O
(2005
McCownetal.(1986) Incarceration Incarcerated/community 5s Fullintensity O
Mitchelletal.(2006) PCL-R p20/X30 3s Morphs O O O
Montagneetal.(2005) BIS/BAS o10th/490thpercentile Notspecified Morphs O O O
Stevensetal.(2001) PSD o20/425 2s Fullintensity O
WalkerandLeister Externalizing Diagnosis/healthy 20s Fullintensity O O O
(1994)
Walker(1981) Diagnosis Aggressive-unsocialized/ Nolimit Fullintensity O O
healthy
WalzandBenson(1996) RBPCCD p2/X13 10s Fullintensity
subscale
Woodbury-Smithetal. Criminalrecord 0/X1conviction 5s Hexagon O O
(2005)
Zabel(1979) Schoolsetting Emotionallydisturbed/ Nolimit Fullintensity O
normal
PCL-R¼PsychopathyChecklist-Revised;PSD¼PsychopathyScreeningDevice,analternateabbreviationforthisscaleisAPSD;PCL-SV¼Psycho-
pathyChecklist:ScreeningVersion;FTD¼Frontotemporaldementia;IRI¼InterpersonalReactivityIndex;CCB¼ChecklistofChallengingBehaviors;
BIS/BAS¼Behavioral Approach System/Behavioral Inhibition System; RBPC¼Revised Behavior Problem Checklist; CD¼conduct disorder;
SRDQ¼Self-ReportedDelinquencyQuestionnaire.
setatPo0.05.AllP-valuesaretwo-tailedunlessotherwise 3. Results
indicated.
Prior to calculating parametric tests, percent accuracy 3.1. Do antisocial populations show deficits in recognizing
and r scores were weighted by the sample size of the the six emotional expressions?
studies. Skewness (Z¼7.97, Po0.001) in the distribution
of sample sizes was corrected via log transformation. This Foreachanalysisaddressingthisquestion,wecompared
transformation successfully reduced skewness to non- accuracy for recognizing the six basic expressions among
significance (Z¼1.89, P¼0.06). Transformed sample antisocial individuals to accuracy for controls. We used
sizes werethenmultiplied bytherelevantpercent accuracy bothsingle-samplet-testsandtheirnon-parametricequiva-
scores and effect sizes, and the products were divided by lent to calculate these comparisons. In all cases, threshold
the total of all log-transformed sample sizes to create significance was set at Po0.05; only results that survive
weighted estimates of effect size (Doria, 2005). correction for multiple comparisons via the Bonferroni
We also conducted follow-up analyses to determine method are presented.
whether any detected deficits varied as a function of the
samples’classification,age,orgenderdistribution.Finally, 3.1.1. Non-parametric analysis
we conducted a file-drawer analysis to determine the The binomial distribution test is the non-parametric
number of null results required to counteract our equivalentof a single-sample t-test. We calculated six such
findings. tests; each assessed whether the proportion of studies in
Author's personal copy
ARTICLE IN PRESS
A.A.Marsh,R.J.R.Blair/NeuroscienceandBiobehavioralReviews32(2008)454–465 459
whichantisocialindividualsshoweddeficitsforrecognizing Thus, the results of the three types of analysis showed
an expression was greater than the proportion of studies consistent deficits in the recognition fear and sadness in
that did not show deficits for that expression. These antisocial samples (two of the three analyses found
analyses incorporated all 20 studies. The results (Ps one- differences for surprise). No consistent deficits were found
tailed) indicated deficits in recognizing fear, Z¼4.25, in the recognition of the remaining three expressions.
Po0.0000001 and sadness, Z¼2.46, Po0.006. No sig- Deficitsweregreaterforfear thanforall other expressions
nificant group differences for anger, disgust, happiness, or across all analyses (Table 3).
surprise were found (Table 2).
3.2. Are observed deficits greater for the fear facial
3.1.2. Parametric analysis: percent accuracy expression than for other emotional expressions?
Sixteenofthe20studiesreportedpercentageaccuracyor
data that could be converted to percentage accuracy and We next assessed whether antisocial samples’ fear
were included in this analysis. Of these, for the 14 studies recognition deficits were significantly greater than deficits
incorporating a multiple-choice format, we corrected for for the remaining five expressions. For each analysis, we
thedegreeofaccuracytobeexpectedfromchanceguessing assessed whether accuracy deficits among antisocial in-
to compare studies with different numbers of response dividuals were greater for fear expressions than for other
options using the formula: (proportion correct–(1/number expressions.Weusedrepeated-measuresANOVAsandthe
of choices))/(1–(1/number of choices)) (Elfenbein and non-parametric equivalent of this analysis (the Friedman
Ambady, 2002). Then for each expression, we calculated test) to calculate these comparisons.
asingle-samplet-test tocompare averagepercentaccuracy
in antisocial and comparison groups. The results indicated 3.2.1. Non-parametric analysis
lower percent accuracy scores inantisocial populations for TheFriedmantestisanalogoustotheone-wayANOVA.
fear, t(15)¼5.65, Po0.001; sadness, t(14)¼4.77, We computed the rank order of expression recognition
Po0.001; and surprise t(13)¼3.81, Po0.005. No group deficits for each study in our analysis. For example, if in a
differences for anger, disgust, or happiness were found. study antisocial individuals showed the greatest deficit for
fear,fearwouldberanked1forthatstudy.Ifdeficitswere
3.1.3. Parametric analysis: effect size second-greatest for sadness, sadness would be ranked 2.
Wecalculatedeffectsizesforthe17studiesthatprovided Theresult oftheFriedmantestcomparingthedistribution
sufficient data. This calculation indexed the magnitude of of ranks across expressions was significant, w2(5)¼16.67,
the effect (r) of antisocial tendencies on the recognition of P¼0.005, suggesting that recognition deficits were not
various emotional expressions. For 15 studies, r represents evenly distributed among the six expressions. With five
the magnitude of the difference between two samples: paired-samples Wilcoxon’s tests, we compared fear recog-
r¼square root (t2/(t2+df)). For two studies (Carr and nition deficits to deficits for each other expression.
Lutjemeier, 2005; Dadds et al., 2006), r represents the These tests were again computed on the rank ordered
correlation between accuracy and antisocial scores in a data.Theresultsshowedthatfearrecognitiondeficitswere
single sample: r¼(P(Z *Z ))/N. All scores significantly greater than deficits for happiness, Z¼3.479,
accuracy antisociality
were normalized using a Fisher’s Z transformation (prior Po0.001;disgust,Z¼2.592,P¼0.005;anger,Z¼2.371,
to weighting by sample size). Then we calculated single- P¼0.009; surprise, Z¼1.694, P¼0.045; and sadness,
sample t-tests to compare Zr scores in antisocial and Z¼1.666, P¼0.048 expressions (all Ps one-tailed).
comparison groups. The results indicated significant effect
sizes associated with deficits in recognizing fear, 3.2.2. Parametric analysis: percent accuracy
t(16)¼7.15, Po0.001; sadness, t(15)¼4.39, P¼0.001; We next calculated a six-level (emotion) repeated-
and surpriset(11)¼3.24, P¼0.008. No group differences measures ANOVA to assess whether percent accuracy
for anger, disgust, or happiness were found. deficits in antisocial individuals significantly vary among
Table3
Comparisonsofdeficitsinrecognizingsixemotionalfacialexpressionsinantisocialpopulations
Expression Corrected%accuracy(meanandSD) Mdiff(antisocial(cid:2)comparison) Mdifffromfear
Antisocial Control %Accuracy r %Accuracy r
Fear 48.4(18.6) 63.8(20.7) 18.09** 0.34** – –
Sadness 63.9(14.5) 74.4(8.7) 12.38** 0.19** 6.05** 0.16*
Surprise 67.1(25.0) 74.5(20.5) 8.37* 0.15* 9.34* 0.20*
Disgust 57.8(24.7) 61.6(28.2) 4.97 0.12* 11.09** 0.23*
Anger 65.7(21.0) 72.1(13.7) 7.61 0.10 10.81* 0.28*
Happiness 90.5(10.1) 92.7(8.7) 2.60 0.11 15.82** 0.25**
Author's personal copy
ARTICLE IN PRESS
460 A.A.Marsh,R.J.R.Blair/NeuroscienceandBiobehavioralReviews32(2008)454–465
the six emotions. This ANOVA, calculated on weighted
percentaccuracydifferencescores,revealedamaineffectof
expressioncategory,F(5, 60)¼3.86,Po0.01.Fivepaired-
samplet-testscomparingfearrecognitiondeficitstodeficits
for each of the remaining emotional expressions showed
greater percent accuracy deficits for fear than for happi-
ness, t(14)¼4.68, Po0.001; anger, t(14)¼2.82,
P¼0.014; disgust, t(13)¼2.58, P¼0.023; and surprise,
t(13)¼2.20, P¼ 0.046. Deficitsfor fear were also greater
than for sadness, but this difference was not significant,
t(14)¼1.53, P¼0.15.
3.2.3. Parametric analysis: effect size
We conducted a second six-level (emotion) repeated-
measures ANOVA on the weighted Fisher’s Zr scores to
assesswhethereffectsizesassociatedwithrecognitiondeficits
varied among the six expressions. The results indicated a
significant main effect of emotion category, F(5, 55)¼2.44,
Po.05. Five follow-up paired-sample t-tests showed that
recognition deficits in antisocial individuals were associated
with greater effect sizes for fear than for happiness,
t(13)¼3.14, P¼0.007; anger, t(14)¼3.11, P¼0.008;
disgust, t(13)¼2.60, P¼0.022; surprise, t(11)¼2.40,
P¼0.035; and sadness, t(15)¼2.35 P¼0.033.
These results indicated that antisocial individuals show
consistently greater deficits in recognizing fear expressions
than in recognizing the remaining five expressions (two of Fig. 1. Median differences in recognizing six facial expressions in
the three analyses found greater deficits for fear than antisocialandcomparisonpopulations. Differences expressedaspercent
accuracy(a)andr(b).Errorbarsshowstandarderrorofthemean(SEM).
sadness).Thissuggestsaspecificdeficitratherthanaglobal
expression processing deficit distributed evenly across the
six expressions (Fig. 1). We next calculated whether this
specific deficit is attributable to the general difficulty of classification. We thus computed follow-up correlations
recognizing fearful expressions. to assess whether any of these variables moderated the
relationship between antisociality, measured as either
3.3. Are fear recognition deficits among antisocial percent accuracy difference scores (comparison minus
populations attributable to task difficulty? antisocial group) or Zr scores. The results showed no
consistent relationship between deficits for recognition of
WeusedthemethoddescribedbyRapcsaketal.(2000)to any of the six expressions and whether the antisocial
assesswhetherfearrecognitiondeficitsremainedsignificantly populationwasclassified aspsychopathicornot(binomial
greater in antisocial populations than comparisons after coding),theaverageageofparticipants,andthepercentage
accounting for task difficulty. We created Z-scores for each of female participants in the sample (all Ps40.05).
antisocialandcomparisonsample.Thesewerecreatedusing
each sample’s percent accuracy scores for fear recognition 3.5. File drawer analysis
and the means and standard deviation of percent accuracy
scores for all six expressions. Using a paired-samples t-test, We calculated six file drawer analyses to determine, for
we found Z scores for fear expressions to be significantly eachexpression,howmanyunpublishedstudieswitheffect
lower in antisocial samples (Z¼(cid:2)0.94) than comparison sizes of zero would be required to render the obtained
samples(Z¼(cid:2)0.54),t(16)¼2.19,Po0.05.Thisshowsthat effects nonsignificant. These analyses were based on effect
antisocial samples show proportionally more difficulty sizes (r) for each expression in each study for which effect
recognizing fearful expressions than do comparisons, in- sizes were available. Using the formula provided by
dicating that fear recognition deficits among antisocial Rosenthal (1979), we compared the values to a standard
individuals do not simply result from task difficulty. tolerance level (5k+15), the value of which indicates the
minimum acceptable number of file drawer studies, with k
3.4. Assessments of moderator variables equal to the number of studies included in the meta-
analysis. This equation sets the minimum possible toler-
The studies we assessed varied widely in terms of their ance level at 20. Only the file drawer statistics for fear and
samples’ age and gender distributions and primary sadness expressions exceeded the tolerance level, with,
Author's personal copy
ARTICLE IN PRESS
A.A.Marsh,R.J.R.Blair/NeuroscienceandBiobehavioralReviews32(2008)454–465 461
respectively, 241 and 106 null findings required to render Murphy et al., 2003). Clinical studies find that amygdala
the results of the meta-analysis non-significant. lesions lead to specific impairments in recognizing fearful
expressions (Adolphs et al., 1999; Papps et al., 2003).
4. Discussion Neuroimaging studies find larger increases in amygdala
activation to fearful expressions than other expressions.
We found a consistent, robust link between antisocial The resultsof a recent meta-analysis (Murphy etal., 2003)
behavior and impaired recognition of fearful facial affect. indicated that amygdala activity is ‘‘remarkably selective’’
Relative to comparison groups, antisocial populations (p. 225) for fear-related processing. This conclusion was
showed significant impairments in recognizing fearful, basedonresultsshowingthatover60%ofstudiesassessing
sad, and surprised expressions. They were not reliably fear expression processing show enhanced amygdala
impaired in recognizing happiness, anger, or disgust activation, compared to fewer than 20% of studies
expressions. Deficits for recognizing fear were significantly assessing emotional processing relevant to disgust, anger,
greater than deficits for any of the other expressions. This happiness, or sadness. Several studies have shown en-
specificity suggests that fear recognition deficits in anti- hanced amygdala activity across multiple expressions;
social samples do not simply reflect differences in factors however, the largest increase in amygdala activity still
that alter facial affect recognition broadly, such as general may be found to fearful expressions (Winston et al., 2003;
intelligence, attention, task-specific motivation, or percep- Fitzgerald et al., 2006). Finally, sampling participants on
tualprocessingdeficits.Rather,theysuggestthatantisocial the basis of their fear recognition ability shows that
behavior may be associated with deficits in neurocognitive reduced ability to identify fearful expressions is associated
mechanisms that specifically underlie processing of fearful with amygdala hyporesponsivity (Corden et al., 2006).
expressions. These results reinforce that meta-analyses can Together, this evidence supports the association between
reveal latent but consistent data patterns across multiple specificfearrecognitiondeficitsandamygdaladysfunction.
studies (Rosenthal and DiMatteo, 2001). Although the Extensive indirect and direct evidence supports the
results of many of the studies we assessed showed fear existence of amygdala dysfunction in antisocial popula-
recognitiondeficits,oftentheauthorsdidnotemphasizeor tions.Patrick(1994)wasthefirsttoconsiderthatamygdala
discuss these deficits. And in aggregate, they yielded a dysfunction might be linked to the development of
consistent relationship between fear deficits and antisoci- psychopathy. Subsequent neurocognitive testing to probe
ality, although they contained samples with heterogeneous functions subserved by the amygdala has found consistent
age and gender distributions and diverse specific classifica- deficits in antisocial individuals. Individuals with psycho-
tions. pathyshowimpairmentontasksthatassessthegeneration
Fear is indisputably more difficult for even healthy of autonomic responses to anticipated threat (Hare, 1982;
individualstorecognizethanexpressionssuchashappiness Ogloff and Wong, 1990), the augmentation of the startle
and sadness (Russell, 1994; Elfenbein and Ambady, 2002). reflex following visually presented threat primes (Leven-
Accuracy rates for this expression are typically around ston et al., 2000), passive avoidance learning (Lykken,
60–70% in healthy populations. It has been suggested that 1957; Newman and Kosson, 1986), and affective priming
fear recognition deficits in some clinical populations may (Blair et al., 2006). Individuals with antisocial personality
result from the general difficulty of recognizing fearful disorderand children withconductdisorder have similarly
expressions. Rapcsak et al. (2000) found that fear recogni- shown reduced autonomic responding to aversive stimuli
tion deficits in individuals with large temporal or parietal (Dinn and Harris, 2000; Herpertz et al., 2001). Children
cortex lesions were statistically related to healthy indivi- who show antisocial behavior report reduced arousal to
duals’ difficulty in processing fearful expressions. When aversive stimuli (Sharp et al., 2006). The results of more
taskdifficultywasaccountedfor,subjectswithlesionswere recent imaging studies have confirmed that amygdala
not disproportionately impaired in fear recognition. By activation is reduced during affective processing tasks in
contrast, our results show that fear recognition deficits in psychopathic adults (Birbaumer et al., 2005; Kiehl et al.,
antisocial populations are proportionally greater than in 2001) and children with conduct disorder (Sterzer et al.,
comparison populations, indicating that these deficits are 2005).Behavioraldatacannotconclusivelysettlequestions
not attributable solely to task difficulty (see Adolphs, relevant to neuroanatomical substrates. However, the
2002).Itshouldalsobenotedthatourcomparisonsamples current data are consistent with existing neuroanatomical
showed nearly equivalent accuracy for fear and disgust data that suggest amygdala impairment to be associated
expression recognition, as has been found previously with both antisociality and specific impairments in fearful
(Calder et al., 2003; Camras and Allison, 1985; Ekman et expression processing.
al., 1987). However, antisocial individuals showed no The role that the amygdala plays in the correct
recognition deficits for disgust. identificationoffearfulexpressionsisnotyetclear.Several
Fear recognition deficits may instead indicate consistent possibilities have been proposed. One is that perception of
neurocognitivedysfunctionsamongantisocialpopulations. an emotion causes the viewer to simulate the perceived
Recent reviews indicate that fear recognition relies emotional experience. A viewer may draw on information
disproportionately on the amygdala (Adolphs, 2006; fromanamygdala-basedfearsimulationtoreconstructand
Author's personal copy
ARTICLE IN PRESS
462 A.A.Marsh,R.J.R.Blair/NeuroscienceandBiobehavioralReviews32(2008)454–465
identify the expresser’s emotional state (Goldman and suggests that dysfunction in one or more structures in this
Sripada, 2005; Lawrence and Calder, 2004). This explana- network may be associated with antisocial behavior.
tion specifies that reduced fear responding will be Dysfunction in superior temporal sulcus has been impli-
associated with impaired fearful expression recognition. cated in antisocial populations (Kiehl, 2006). In addition,
As has been discussed, some antisocial populations regions involved in general emotional processing that are
(particularly psychopaths) show just this pattern: both functionally connected to the amygdala may be associated
limited fearful responding (Patrick, 1994) and limited with deficits in processing facial expressions, particularly
fearful expression recognition (Blair et al., 2004). Another fearful expressions (Murphy et al., 2003). Such regions
proposed mechanism is that activity in the amygdala include the dorsal anterior cingulate cortex (Stein et al.,
facilitates retrieval ofsemanticlabelsfor facial expressions 2007)andtheventromedialprefrontalcortex,aregionthat
like fear. In this view, expression recognition impairments has extensive functional connections with the amygdala
represent impairments in the ability to generate a verbal and that is often found to be impaired in antisocial
labelforanexpressionlikefear(Adolphs,2006;Blairetal., individuals (Birbaumer et al., 2005; Kiehl et al., 2001).
2005).Empiricalsupportexistsforbothofthesemodelsof Lesionstotheseregionsanddisruptedconnectionsbetween
facial affect recognition; this meta-analysis cannot provide theseregionsandtheamygdalaappeartoaffectprocessing
conclusive evidence in support of one over the other. of multiple emotional expressions rather than of fearful
Thismeta-analysisalsocannotclarifywhetherdeficitsin expression specifically (Hornak et al., 2003). Finally,
therecognitionoffearexpressionsarespecifictoantisocial subcortical routes that involve the thalamus and superior
populations or extend to other clinical populations. colliculus appear to be involved in processing emotional
However, the available evidence does not suggest that expressions, particularly fearful expressions (Luo et al.,
specific deficits in fear expression recognition are char- 2007; Morris et al., 1999). Dysfunction in these regions
acteristic of all clinical samples. There is no theoretical could also exist among antisocial populations.
supportofwhichwe areawareforfear recognition deficits Most of the populations in the studies included in this
in disorders such as depression, anxiety, or borderline meta-analysis had clinical diagnoses, were incarcerated, or
personality disorder. Some data even suggest that indivi- were in other secure settings (e.g., camps or schools for
duals with these disorders show enhanced sensitivity to children with behavior problems). These populations
fearfulexpressions(Bhagwagaretal.,2004;LeMasurieret presumablyposessedseriousfunctional impairments.Only
al., 2007; Surcinelli et al., 2006; Wagner and Linehan, three studies (Blair and Coles, 2000; Dadds et al., 2006;
1999).Indicationsofamygdaladysfunctioninautismhave Montagneetal.,2005)drewfromcommunitysamples.Itis
led to suggestions of fear recognition impairment in this thus difficult to infer how the observed relationship
population (Howard et al., 2000). However, individuals between antisociality and fear recognition would vary as
with autism are typically found to show equivalent a function of functional impairments. We speculate that
impairments in the recognition of multiple expressions increasingly severe antisocial behaviors and traits such as
(Castelli,2005;Humphreysetal.,2007;Sasson,2006).This callousness would be associated with increasingly impo-
suggests broader face processing deficits in autism in verished fear recognition abilities. Thus, any two samples
accordancewiththebroadsocialdysfunctioncharacteristic thatdifferinantisocialitywillalsodifferinfearrecognition
of this disorder. skills, but the greater the disparity in antisocial traits and
It should be emphasized that identifying any emotional behavior, the greater the disparity in facial affect recogni-
expression,includingfearfulexpressions,isacomplextask tion.
that requires visual scanning, perceptual processing, Thismeta-analysisincludedpopulationscharacterizedas
effortfulattention, workingmemory,andsemanticproces- psychopathic,conductdisordered,aggressive,unsocialized,
sing. Accordingly, such processing relies on a large, abusive,andcriminal.Theseclassificationsallareprimarily
distributednetworkofneuralstructures.Atthemostbasic defined by persistent behavior that violates the rights and
level,intactfunctioningofoccipitotemporalvisualcortexis welfare of others or breaks important normative rules—
required to process the geometric configuration of the i.e., antisocial behavior. Unlike the other classifications, a
features of the face (Allison et al., 1999). The superior classificationofpsychopathyindicatesparticularantisocial
temporal gyrus and fusiform gyrus (part of inferior personality traits (e.g., lack of empathy and remorse) in
temporal cortex) play central roles in processing faces addition to antisocial behaviors such as criminal offenses.
(LaBar et al., 2003). Once configural features have been The remaining classifications do not specify what person-
assessed, intact functioning of structures in the temporal ality or environmental variables (such as impulsiveness,
lobes is required to link the configural properties of facial substance abuse, physical abuse, or trauma)are associated
expressions with stored knowledge about what those with observed antisocial behavior. Psychopathic indivi-
properties represent (Haxby et al., 2002). Dysfunction in duals thus represent a more homogeneous group, and
any of these structures may characterize individuals who evidenceofimpairedamygdalafunctioninginthisgroupis
showgeneralizedimpairmentsinprocessingfacialemotion. accordingly more consistent (Birbaumer et al., 2005; Blair
Lowlevels ofimpairmentwerefound acrossexpressions et al., 2006; Herpertz et al., 2001; Kiehl et al., 2001;
in the antisocial samples in this meta-analysis. This Mitchell et al., 2006). This may be because amygdala