Differences in Neural Activity When ProcessingEmotional Arousal and Valence in Autism

Spectrum Disorders

Angela Tseng,1 Zhishun Wang,1 Yuankai Huo,1 Suzanne Goh,1

James A. Russell,2 and Bradley S. Peterson1,3*

1Department of Psychiatry, Columbia University College of Physicians and Surgeons andNew York State Psychiatric Institute, New York, NY, USA

2Department of Psychology, Boston College, Chestnut Hill, MA, USA3Children’s Hospital Los Angeles and the Keck School of Medicine at the University of

Southern California, Institute for the Developing Mind, Children’s Hospital Los Angeles, KeckSchool of Medicine at the University of Southern California, Los Angeles, CA, USA

r r

Abstract: Individuals with autism spectrum disorders (ASD) often have difficulty recognizing andinterpreting facial expressions of emotion, which may impair their ability to navigate and communicatesuccessfully in their social, interpersonal environments. Characterizing specific differences betweenindividuals with ASD and their typically developing (TD) counterparts in the neural activity subserv-ing their experience of emotional faces may provide distinct targets for ASD interventions. Thus weused functional magnetic resonance imaging (fMRI) and a parametric experimental design to identifybrain regions in which neural activity correlated with ratings of arousal and valence for a broad rangeof emotional faces. Participants (51 ASD, 84 TD) were group-matched by age, sex, IQ, race, and socioe-conomic status. Using task-related change in blood-oxygen-level-dependent (BOLD) fMRI signal as ameasure, and covarying for age, sex, FSIQ, and ADOS scores, we detected significant differences acrossdiagnostic groups in the neural activity subserving the dimension of arousal but not valence. BOLD-signal in TD participants correlated inversely with ratings of arousal in regions associated primarilywith attentional functions, whereas BOLD-signal in ASD participants correlated positively with arousalratings in regions commonly associated with impulse control and default-mode activity. Only minordifferences were detected between groups in the BOLD signal correlates of valence ratings. Our find-ings provide unique insight into the emotional experiences of individuals with ASD. Although behav-ioral responses to face-stimuli were comparable across diagnostic groups, the corresponding neuralactivity for our ASD and TD groups differed dramatically. The near absence of group differences forvalence correlates and the presence of strong group differences for arousal correlates suggest that indi-viduals with ASD are not atypical in all aspects of emotion-processing. Studying these similarities anddifferences may help us to understand the origins of divergent interpersonal emotional experience inpersons with ASD. Hum Brain Mapp 00:000–000, 2015. VC 2015 Wiley Periodicals, Inc.

Additional Supporting Information may be found in the onlineversion of this article.

Contract grant sponsor: NIMH; Contract grant numbers: R01MH089582 (BSP), 2 T32 MH16434 (BSP)

*Correspondence to: Bradley S. Peterson, M.D., 4650 Sunset Blvd.MS# 135, Los Angeles, CA 90027, E-mail: bpeterson@chla.usc.edu

Received for publication 27 May 2015; Revised 21 September2015; Accepted 19 October 2015.

DOI: 10.1002/hbm.23041Published online 00 Month 2015 in Wiley Online Library(wileyonlinelibrary.com).

r Human Brain Mapping 00:00–00 (2015) r

VC 2015 Wiley Periodicals, Inc.

Key words: autism spectrum disorders; arousal; valence; facial emotion; fMRI

r r

INTRODUCTION

Autism spectrum disorders (ASD) are a set of complexneurodevelopmental disabilities that cause lifelong impair-ments in social ability, communication, and behavioralflexibility [American Psychiatric Association, 2000]. Indi-viduals with ASD often have difficulty recognizing andinterpreting facial expressions of emotions, which mayimpair their ability to understand the intentionality andminds of others, a capacity needed for successful socialcommunication [Golan et al., 2006; Grelotti et al., 2002].Despite having a general consensus that persons with ASDare atypical in their processing of human faces and emo-tional expressions [Harms et al., 2010; Sasson, 2006],researchers do not agree on the underlying brain andbehavioral mechanisms through which individuals withASD decode emotional faces. Some prior research suggests

that individuals with ASD rely more on cognitive-perceptual systems involving explicit cognitive or verballymediated processes to interpret facial expressions of emo-tions, in contrast to neurotypical individuals who processemotions more automatically [Harms et al., 2010; Pelphreyet al., 2007].

Although ASD is generally considered to involve deficitsin emotion recognition, prior studies have provided onlyinconsistent evidence for those deficits. For example, sev-eral studies have reported that adults and children withASD have more difficulty recognizing, responding to, andexpressing emotions than typically developing (TD) indi-viduals [Ashwin et al., 2006; Tantam et al., 1989; Uljarevicand Hamilton, 2013] and more than persons with otherneurodevelopmental disorders [Celani et al., 1999; Ribyet al., 2008]. However, other studies have reported typicallevels of facial emotion recognition in persons with ASD[Castelli, 2005; Harms et al., 2010; Ozonoff et al., 1990;Tseng et al., 2014].

Disparities in findings for the recognition and under-standing of emotions in individuals with ASD may, tosome extent, be due to fundamental differences in theunderlying model of emotion implicitly assumed whendesigning those studies. That underlying model has mostoften been the theory of basic emotions [Ekman, 1992; Pan-ksepp, 1992], which posits that each member of a core setof discrete, or “basic,” emotions (e.g., anger, sadness, orhappiness) are subserved by its own distinct and inde-pendent neural system [Ekman, 1992; Panksepp, 1992].Earlier reviews have documented the limitations andinconsistencies of this theory, including the absence ofone-to-one mappings of individual emotions with specificfacial expressions, motor behaviors, and autonomicresponses, as well as the absence of evidence for a core setof emotions from which other emotions derive [Hamann,2012; Posner et al., 2005; Russell, 1980; Vytal and Hamann,2010]. Moreover, interpreting findings from neuroimagingstudies based on the theory of basic emotions is compli-cated by the subtraction method employed in most func-tional imaging designs, in which brain activity ismeasured by comparing two tasks or stimuli that areassumed to differ only in the cognitive process of interest.Most functional imaging studies based on the theory ofbasic emotions have contrasted neural responses to indi-vidual emotions with neural responses to stimuli intendedto be emotion-neutral. Unfortunately, the use of “neutral”faces as control stimuli is an inherent confound in emotionresearch because of the difficulties involved in creatingtruly “neutral” stimuli [Killgore and Yurgelun-Todd, 2004;Klein et al., 2015; Posner et al., 2005; Thomas et al., 2001].

Additionally, most imaging studies of the basic emo-tions theory have focused only on a small number of

Abbreviation

ACC anterior cingulate cortexAMYG amygdalaBOLD blood oxygen level dependentBroca Broca’s areaCb cerebellumCN caudate nucleusCu cuneusDLPFC dorsolateral prefrontal cortexFG fusiform gyrusHIPP hippocampusILPFC inferolateral prefrontal cortexINS insulaIPC inferior parietal cortexIPS intraparietal sulcusM1 primary motor cortexMCC middle cingulate cortexMFG middle frontal gyrusMOG middle occipital gyrusMTG middle temporal gyrusOFC orbitofrontal cortexPC parietal cortexPCC posterior cingulate cortexPCu precuneusPreMC premotor cortexPrG precentral gyrusPUT putamenS1 primary somatosensory cortexS2 secondary somatosensory cortexSFG superior frontal gyrusSMA supplementary motor areaSPL superior parietal lobuleSTS superior temporal sulcusTHAL thalamusV1 primary visual cortex

r Tseng et al. r

r 2 r

emotions, generally those of high arousal and negativevalence (e.g., fear and anger), low arousal and negativevalence (e.g., sadness), or moderate arousal and positivevalence (e.g., happy). Consequently, researchers have hadtrouble disentangling the differing contributions of arousaland valence to the neural correlates of emotions. Forexample, “happy” stimuli are often the only positivevalence emotions included in study designs. Comparingthem to stimuli that are putatively neutral or even nega-tively valenced confounds the positive arousal componentwith the positive valence of the happy stimulus. In effect,even when comparing happy with putatively neutral faces,both of these types of stimuli have not only a valence, butalso an arousal component that is never considered as con-tributing to the reported fMRI activation [Fusar-Poli et al.,2009; Harms et al., 2010; Murphy et al., 2003].

An alternative theoretical framework to the theory ofbasic emotions is the “Circumplex Model of Affect,” whichholds that the subjective experience of all emotions arisesfrom the linear combination of two independent neuro-physiological systems, valence and arousal. Valence refersto hedonic tone, or the degree to which an emotion ispleasant or unpleasant, whereas arousal represents thedegree to which an emotion is associated with high or lowenergy. Under this model, a “happy” response to a stimu-lus arises from relatively intense activation of the neuralsystem associated with positive valence and moderate acti-vation of the neural system associated with positivearousal. Other emotional states thus arise from the sametwo underlying neurophysiological systems but differ indegree of activation of each. Because all emotions can berepresented as a linear combination of the dimensions ofarousal and valence, emotions shade imperceptibly fromone into another along the contour of the two-dimensionalcircumplex [Posner et al., 2005]. The subjective experienceof the neurophysiological signals for valence and arousalis determined by interpretations of the signals in relationto the experiential context of the stimuli and memories ofprior experiences of similar sensations [Posner et al., 2005;Russell, 2003]. Thus the labeling of our subjective experi-ence as one emotion rather than another nearby emotion isthe consequence, in part, of cognitive interpretation of theneurophysiological experiences of arousal and valencewithin the situational context [Russell, 2005].

Several studies have provided evidence for the existenceof distinct neural systems that subserve the experience ofemotional valence and arousal [Colibazzi et al., 2010;Gerber et al., 2008; Posner et al., 2009]. However, to ourknowledge, no other studies have examined whether neu-ral activity in circuits that subserve processing of the twodimensions of facial emotions differ between individualswith ASD and their TD counterparts. A prior publicationfrom our laboratory reported that, in the same sample, theASD group performed nearly as well as, and in a similarpattern to, the TD group when participants were asked torate emotional faces for arousal and valence [Tseng et al.,

2014]. However, without corresponding data on brainactivity, determining whether the TD and ASD groupsrecruited the same neural systems to appraise emotionalstimuli is impossible. Typical-level behavioral performanceon emotion-processing tasks does not exclude the possibil-ity of atypical neurocognitive processing of emotionalinformation. Rather, abnormalities in emotion-processingmight be obscured in some individuals because they havedeveloped compensatory strategies that yield “typical”levels of behavioral performance. Indeed, higher-functioning individuals with ASD might capitalize on theircognitive resources to identify facial expressions. Forexample, studies employing emotion-matching paradigms[Piggot et al., 2004; Rump et al., 2009] are more likely thanstudies using emotion-labeling paradigms [Katsyri et al.,2008; Piggot et al., 2004; Rutherford and Towns, 2008] toreveal differences in behavioral performance between TDand higher-functioning ASD groups. For some individualswith ASD, the use of emotion labels in a task may facili-tate recognition of facial expressions of emotions, espe-cially when they are trained to identify emotions as part ofan intervention program [Tanaka et al., 2010].

Although functional imaging studies of emotion-processing in ASD have yielded inconsistent findings, sev-eral have reported hypofunctioning in ASD in regionsassociated with socio-emotional processing (e.g., INS,AMYG) [Di Martino et al., 2009], in extrastriate cortices[Deeley et al., 2007], ventral PFC [Ashwin et al., 2007;Hadjikhani et al., 2006], medial-frontal and orbito-frontalcortices [Bachevalier and Loveland, 2006; Loveland et al.,2008; Ogai et al., 2003], ACC and FG [Hall et al., 2003],striatum, and IFG [Dapretto et al., 2006] compared to TDcontrols. Conversely, studies have found increased activityfor ASD compared to TD groups in STS, ACC [Ashwinet al., 2007; Hall et al., 2003; Pelphrey et al., 2007], andparieto-occipital regions [Dapretto et al., 2006; Hubl et al.,2003; Wang et al., 2004] when viewing facial emotions.These increases in neural activity may derive fromincreased visual and motor attention [Dapretto et al.,2006], more effortful processing of specific facial featureswithin the given social contexts [Ashwin et al., 2007], andincreased attentional load [Wang et al., 2004], supportingthe possibility that emotional processing is more effortfuland less automatic in individuals with ASD than in theirTD counterparts.

To address whether neural activity in circuits that sub-serve processing of arousal and valence differ betweenindividuals with ASD and TD individuals, we applied aparametric experimental design to identify brain regionsin which neural activity correlated with arousal andvalence ratings for a broad range of facial emotions. Theuse of a parametric design allows us to compare emotionalstimuli across multiple levels or through incrementalchanges along the affective dimensions of arousal andvalence. For example, parametric manipulation of emo-tional stimuli that change incrementally in the degree of

r FMRI of Arousal and Valence Dimensions in ASD r

r 3 r

arousal or valence that they generate can be mappedagainst concomitant variations in neural activity. Accord-ingly, activity in neural structures or pathways that corre-late with the degree of emotional arousal or valenceinduced by the emotional probes can be assessed in indi-vidual participants [Posner et al., 2005]. In the presentstudy, we used BOLD-signal intensity to index neuralactivity as participants viewed photographs depictingemotional faces. We identified brain regions in whichBOLD-signal systematically covaried with ratings ofarousal or valence in ASD and TD groups, and we deter-mined the areas in which these correlations differed statis-tically across emotional dimensions and diagnostic groups,indicating the differential associations of these regionswith processing arousal or valence within each group. Wesought to identify similarities and differences in neuralactivity when participants with ASD and TD participantsview and rate these experiences of facial emotions [Gerberet al., 2008; Russell et al., 1989]. Given the socio-emotionaldeficits associated with ASD, we hypothesize that the ASDgroup will show abnormal patterns of brain activationwhen compared to controls, particularly in brain regionsassociated with processing of emotional stimuli in personswith ASD, including AMY, INS, CN, PFC, OFC, and ACC.

MATERIALS AND METHODS

Study procedures were approved by New York StatePsychiatric Institute’s Institutional Review Board. All par-ticipants provided informed written consent or assent andreceived payment for participating (See Supporting Infor-mation for detailed consent procedures).

Participants

We recruited 51 individuals with ASD (6F, ages: 7–60years, Mean: 27.5 6 13.1 years) and 84 TD individuals (22F,ages: 7–60 yrs, Mean: 24.0 6 11.4 years) from the NewYork City area. A wide age-range was included in oursample in order to understand better the developmentaltrajectory of emotional processing in this under-studiedgroup. For example, if the child participants with ASDperformed similarly to our adult participants with ASD,then we might infer that any emotional deficits found arelikely a static, trait-like disturbance. We also hoped to usecross-sectional data from this investigation to generatehypotheses for future longitudinal research.

Participants were group-matched by age, sex, IQ(Wechsler Abbreviated Scale of Intelligence) [Wechsler,1999], handedness (Edinburgh Handedness Inventory)[Oldfield, 1971], race, and socioeconomic status (Hollings-head Index of Social Status) [Hollingshead, 1975]. Meanfull scale IQ (FSIQ) was 109.2 6 19.4 for the ASD groupand 115.9 6 12.4 for the TD group. Mean verbal IQ (VIQ)and mean performance IQ (PIQ) for both groups did not

differ significantly so we opted to conduct further analysesof IQ using only FSIQ as a covariate (Table I).

Additional individuals participated (N 5 4 ASD; N 5 1TD) but were not included in the final sample due to exces-sive head motion in the scanner.

TD participants were excluded if they met DSM-IV-TRcriteria for a current Axis-I-disorder, or had lifetime his-tory of developmental delay or other indicators of ASD,psychosis, substance abuse disorder, head trauma, seizuredisorder, or other neurological illness. None of the TD par-ticipants were taking prescription or over-the-countermedications; however, the use of dietary supplements wasnot assessed.

Participants with ASD were evaluated by an expert cli-nician and met Diagnostic and Statistical Manual of Men-tal Disorders, Fourth Edition, Text Revision (DSM-IV-TR)[American Psychiatric Association, 2000] criteria for autis-tic disorder, Asperger syndrome, or pervasive develop-mental disorder-not otherwise specified (PDD-NOS).Diagnoses were confirmed with the Autism DiagnosticInterview-Revised [Lord et al., 1994] and the Autism diag-nostic observation schedule (ADOS) [Lord et al., 1989]. Adetailed list of current medications was recorded for everyparticipant (available on request).

At the beginning of each study session, an experiencedmember of the study team explained verbally the natureof the research protocol, including potential risks and

TABLE I. Participant characteristics

ASD TD

Participants (N) 51 84ASD subtype

PDD-NOS 9 —Asperger’s syndrome 24 —Autistic disorder 18 —

Mean age (yrs) 27.5 6 13.1 24.0 6 11.4Children (<18 yrs) (N (%)) 12 (24%) 31 (37%)Males (N (%)) 45 (88%) 62 (74%)Caucasian (N (%)) 40 (78%) 60 (71%)Mean SESa 50 53Mean FSIQb 109.2 6 19.4 115.9 6 12.4Mean VIQ 110.9 6 20.9 115.7 6 13.2Mean PIQMean ADOS

(social affect 1 restrictive,repetitive behaviors)c

105.0 6 17.610.9 6 3.1

112.9 6 11.8—

Mean ADOS—calibratedseverity scores(modules 2 and 3)d

7.5 6 1.8 —

aSES scores for 7 TD and 14 ASD participants were unavailable.bFSIQ scores for 1 TD participant and 1 ASD participant wereunavailable.cADOS scores for 4 ASD participants were unavailable.dADOS CSS scores were calculated for participants tested usingModules 2 and 3 (N 5 10).

r Tseng et al. r

r 4 r

benefits, to all potential participants and (in the case of aminor) to their parents. For participants 8 to <18 yearsold, assent was discussed and obtained. For adult partici-pants (>18 years), the study member obtaining consentexplained the protocol and associated risks to the prospec-tive participant before asking participants to sign the con-sent form. For all adult participants with ASD, anindependent assessment of capacity to consent was con-ducted by a clinical monitor who was independent of thestudy team. When the clinical monitor determined that theparticipant lacked the capacity to consent (i.e., the partici-pant did not demonstrate an understanding of the proce-dures, alternatives, and potential risks and benefits of thestudy, and that participation was voluntary), an author-ized legal representative was designated to provideinformed consent.

Emotion Paradigm

We used a parametric experimental design to identifybrain regions in which neural activity correlated with par-ticipant ratings of arousal or valence (Fig. 1).

Neural activity was indexed using BOLD-signal inten-sity as participants viewed photographs of emotionalfaces. After viewing each face, participants rated arousaland valence simultaneously by selecting an individual boxon a 9 3 9 two-dimensional grid. Location on the x-axisindicated the participant’s rating of valence (left 5 negativevalence, right 5 positive valence), and location on the y-axis indicated the rating of arousal (top 5 high arousal,bottom 5 low arousal). We recorded the selected box astwo integer scores, each ranging from 24 to 14, represent-ing valence and arousal.

Each trial consisted of 3 sequential epochs: (1) Visualpresentation of a photograph of a human face for 18 s. Thephotographs were copied, with permission, from the 20distinct stimuli used by Russell and Bullock [1985] fortheir studies of the affective circumplex. Thirteen of these

20 images were taken from Ekman and Friesen’s [1976]“Pictures of Facial Affect” and depicted expressions of anumber of emotions (two pictures each of emotional facescommonly classified as expressing happiness, surprise,fear, anger, disgust, or sadness, and one commonly classi-fied as neutral). Russell and Bullock supplemented this setto represent better portions of the circumplex that theEkman series under-sampled (emotions associated withlow arousal but positive or neutral valence). Theseincluded two photographs each of actors and actressesexpressing boredom, contentment, and sleepiness, as wellas one expressing excitement. (2) Visual presentation of atwo-dimensional grid on which participants indicated theirratings of arousal and valence for each stimulus by mov-ing an arrow controlled by an MRI-compatible computermouse. This screen remained visible until the participantclicked the mouse button, up to a maximum of 20 s. (3)Visual presentation of a fixation point (1) at the center ofthe participant’s visual field. The fixation point was dis-played immediately following the rating of valence andarousal. The durations of rating and gaze fixation wereeach variable, but when summed together always equaled20 s. One imaging run consisted of 20 trials presented in apseudorandom order (but uniform from participant toparticipant).

Visual stimuli were presented to each participant viaMRI-compatible LCD goggles (Resonance Technology,Northridge, CA) using E-Prime software, version 1.138running on a Dell IBM-compatible computer. Measures ofstimulus durations and reaction times were accurate to 20ms. Stimuli were presented at the center of the partici-pant’s visual field, subtending 198 of the vertical and 158

of the horizontal visual field.Prior to the study session, all participants were given a

practice session with the task so they could familiarizethemselves with task instructions, the types of stimuli theywould be seeing (practice stimuli were not shown duringthe study session), the grid on which they would be rating

Figure 1.

Emotion paradigm.

r FMRI of Arousal and Valence Dimensions in ASD r

r 5 r

arousal and valence, and the computer mouse they wouldbe clicking to indicate their ratings. Researchers wereavailable to review the practice responses in detail, toexplain the instructions further, or to answer any ques-tions about the task during this practice round to ensurefull comprehension. Each participant was told, “You willbe shown a face that expresses a certain feeling. You willbe asked to assess the feeling on the chart shownbelow. . ..On the chart, the vertical dimension representsdegree of arousal. Arousal has to do with how awake,alert, or energetic a person is. . .. The right half of the chartrepresents pleasant feelings—the farther to the right, themore pleasant. The left half represents unpleasant feel-ings—the farther to the left, the more unpleasant. . .. Dur-ing the experiment, you will first be shown a face. Thiswill appear on the screen for 18s. Then you will be shownthe grid. When the grid appears, you will click on the areayou think best describes the face. . .Try to think about thefeeling expressed by the face during the 18s shown. It willnot be on the screen when you are shown the grid.” Atthe time of instruction and during the experiment itself,the words “High Pleasure” appeared to the right of thegrid, and “High Energy” above the grid. The shortenedpractice version consisted of three faces—one eachexpressing sadness, happiness, and anger. To minimizethe possibility of habituation, none of the practice faceswere identical to actual experimental stimuli. During thescan, researchers monitored on-line behavioral responsesin real-time so that we could ensure attention to the task.

Behavioral Data Analyses

Behavioral data gathered from the present study wereanalyzed and reported in detail in a prior publicationfrom our laboratory [Tseng et al., 2014]. The participantswere identical across the two studies, with the exceptionthat several participants were eliminated from fMRI analy-sis because of excessive head motion in the scanner. Inaddition, we collected data for an additional 3 TD adults,1 TD child, and 4 adults with ASD after the publication ofour prior study. We did not find any significant differen-ces between our findings with or without the additionaleight participants, so we included them in our fMRI sam-ple. As described in our previous report, we divided par-ticipants into four groups by diagnosis and age tocompare behavioral performances across groups: AdultASD (N 5 39, 5F, ages: 18–61 years, Mean: 31.9 6 11.8years), Adult TD (N 5 53, 8F, ages: 18–60 years, Mean:30.1 6 10.2 years), Child ASD (N 5 12, 1F, ages: 7–17 years,mean: 13.2 6 3.1 years), and Child TD (N 5 31,14F, ages: 7–17 years, Mean: 13.7 6 2.7 years). Mean FSIQ scores inthese groups were: Adult ASD (110.2 6 18.4), Adult TD(116.42 6 1.9), Child ASD (105.6 6 22.2), and Child TD(115.0 6 13.4). We also divided participants by diagnosisalone to compare the entire ASD and TD groups.

Multivariate ANCOVAs were conducted with arousaland valence ratings as dependent variables, group as theindependent variable, and age and gender as covariates toassess emotion-specific differences between groups. Weused hierarchical multiple regressions for ASD and TDgroups (controlling for age and sex) with arousal andvalence ratings as dependent variables and FSIQ scores asthe independent variable to assess whether IQ was signifi-cantly correlated with how participants rated eachemotion-type. Similar analyses were conducted with totalADOS scores (Social Affect (SA) 1 Restrictive, RepetitiveBehaviors (RRB), Mean 5 10.9 6 3.1 [Gotham et al., 2007].CSS conversion algorithms are not available for partici-pants over the age of 16 or who were assessed with mod-ule 4 of the ADOS.

To assess whether severity of diagnosis significantlycorrelated with how participants rated each emotion-type,we used hierarchical multiple regressions for analyses inthe ASD group (controlling for age and sex) in whicharousal or valence ratings were entered separately as thedependent variable and total ADOS score was the inde-pendent variable. These regressions were applied sepa-rately to each facial stimulus. We also conducted theseanalyses with only the social affect scores from the ADOSas the independent variable, because we expected thesocial affect measure alone might correlate more stronglywith how participants with ASD rated these affectivestimuli.

Finally, we conducted multivariate ANCOVAs witharousal or valence ratings entered separately as thedependent variable, ASD subtype (PDD-NOS, Asperger’sSyndrome, Autistic Disorder) entered as the independentvariable, and age and gender entered as covariates toassess whether participant responses varied according tospecific ASD subtype.

Task Performance

So that we could be as confident as possible that all par-ticipants were performing the task as instructed and toensure the face validity of their responses, we first visuallycompared each individual’s arousal and valence ratingsqualitatively against the canonical circumplex to ensurethat the responses seemed reasonable. Then, assuming thatthe responses of the healthy adults represent the end prod-uct of development, we used the arousal and valencescores from typically-developing adults reported by Rus-sell and Bullock [1985] as reference ratings for “correct”performance by assessing quantitatively the correlations ofeach individual participant’s data with the reference rat-ings. Our rationale was that an individual responding atrandom to the stimuli or who was not understanding orfollowing instructions would be unlikely to produce a sim-ilar response pattern to the reference ratings. Then, as asubset analysis, we removed participants (N 5 13: 4 ChildASD, 4 Adult ASD, 5 Child TD) whose correlations

r Tseng et al. r

r 6 r

between arousal or valence ratings with the reference val-ues were significant at P> 0.2 (corresponding to a Pear-son’s r> 0.42). Similar to findings from our originalanalysis with the entire sample (N 5 135), we detectedwith this smaller sample (N 5 122) a main effect of diagno-sis (P< 0.05). Thus, although we were unable to measuretask comprehension directly during the scan, the use ofprescan practice trials and the similarity of results in oursubset analysis with those of the original analysis showthat the vast majority of our participants were able tounderstand and perform the task as instructed.

Image Acquisition

Imaging was performed on a GE Signa 3T whole bodyscanner (Milwaukee, WI) using a GE single channel quad-rature head-coil. A 3D spoiled gradient recall (SPGR)image was acquired for coregistration with axial functionalimages and for coregistration with a standard referenceimage (Montreal Neurological Institute (MNI)). Functionalimages were acquired using a single shot gradient echoplanar (EPI) pulse sequence in groups of 43 axial slices pervolume and 273 volumes per run (preceded by six“dummy” volumes to ensure scanner stability). Parametersfor the EPI images were: repetition time 5 2,800 ms, echotime 5 25 ms, flip angle 5 908, acquisition matrix 5 64 3

64, field of view 5 24 cm 3 24 cm, slice thickness 5 3 mm,skip 5 0.5 mm, receiver bandwidth 5 62.5 kHz, in-planeresolution 5 3.75 mm 3 3.75 mm. Each run lasted 13 min1 s, for a total EPI scan time of 39 min 3s.

Image Preprocessing

Prior to statistical analyses, we used SPM8 (http://www.fil.ion.ucl.ac.uk/spm/, run under MATLAB2009b) to pre-process the fMRI data. Slice timing was corrected using themiddle slice (22 of 43) as the timing reference. Slice timingcorrected functional images were then realigned to the mid-dle image of the middle run for motion correction for threetranslational directions and rotations. Images with motiongreater than one voxel were excluded from all subsequentanalyses. Motion corrected images of each participant werecoregistered to the corresponding T1-weighted high-resolu-tion anatomical image, which in turn was spatially normal-ized to the standard MNI template with voxel dimensions of3 mm3. These participant-specific normalization parameterswere then used to warp the functional images into the sameMNI template. A spatial smoother with a Gaussian kernel of8-mm Full Width at Half Maximum was applied to the func-tional images, which were then temporally filtered using aDiscrete Cosine Transform high-pass filter with a cutoff fre-quency of 1/128 Hz to remove low frequency noise such asscanner drift.

We then assessed data quality by plotting motion param-eters and mean intensity values for raw, normalized, andsmoothed images for each run in each participant. Visual

inspection allowed us to identify scans for which averageintensity values across voxels were significantly outside themean and which occurred at the same moment as a largehead movement. We also used histogram plots for eachcontrast image in each participant to help identify outliersfor mean intensity that might have been missed by thebatch preprocessing procedure. We used the ArtRepairalgorithm (http://cibsr.stanford.edu/tools/human-brain-project/artrepair-software.html) to detect and repair thoseimage volumes that were contaminated by spiking motionartifacts and outliers [Mazaika et al., 2009]. Volumes withmotion larger than 1mm were repaired. Participantsfor whom motion affected more than 15% of their data(>41/273 volumes per run) were excluded from furtheranalyses; based on this criterion, we eliminated from ourfinal analysis 1 TD and 4 ASD participants (from the origi-nal 140 participants).

Statistical Analyses

We analyzed fMRI data at the individual (first) levelusing a general linear model (GLM) to detect BOLD-signal correlates of arousal or valence within each indi-vidual participant and at the group (second) level usingBayesian posterior inference [Neumann and Lohmann,2003] at a posterior probability threshold of 98.75%, todetect random effects of arousal or valence correlateswithin and between diagnostic groups. We covaried forage and sex of the participants. We also conducted addi-tional analyses covarying for FSIQ in all participants andfor ADOS scores in the ASD group. We assessed the maineffects of arousal and valence ratings on BOLD-signal foreach diagnostic group (TD, ASD). We also assessedBOLD-signal correlates with quadratics of arousal andvalence ratings, allowing us to assess at each voxelwhether the correlation of ratings with BOLD-signal hada significant curvilinear component. We included simulta-neously in our model the main effects and quadratic val-ues of arousal and valence ratings (including themseparately yielded identical findings). Finally, we assessedwhether the within-group valence and arousal correlatesdiffered significantly across ASD and TD participants byassessing the interactions of the correlations with diagnos-tic group. We included simultaneously in our model themain effects and their interactions with diagnostic groupto ensure that the models were hierarchically well formu-lated. We plotted the scatters for the linear and quadraticassociations of BOLD-signal with ratings of arousal andvalence to assess the distribution of data around theregressions and to determine the group contributions tosignificant interactions.

First-level analysis

We used GLM in SPM8 for the analyses of data at theindividual level. We modeled preprocessed BOLD time

r FMRI of Arousal and Valence Dimensions in ASD r

r 7 r

series data at each voxel, using 8 independent functions(Fn) or regressors that consisted of:

Fn(1): the canonical hemodynamic response function(HRF) convolved with a box car function (BCF)derived from the onsets and durations of the pre-sentation of facial stimuli

Fn(2): Fn(1) modulated by the linear arousal rating foreach stimulus

Fn(3): Fn(1) modulated by the linear valence rating foreach stimulus

Fn(4): Fn(1) modulated by the quadratic arousal ratingfor each stimulus

Fn(5): Fn(1) modulated by the quadratic valence ratingfor each stimulus

Fn(6): the canonical HRF convolved with a BCF index-ing the manual responses of each participant tothe task stimuli

Fn(7): the canonical HRF convolved with a BCF index-ing the presentation of a fixation cross

Fn(8): a constant

Our model, which included the main effects and quad-ratic values of arousal and valence ratings on a 24 to 4scale for each participant, was estimated using theRestricted Maximum Likelihood (ReML) algorithm. Task-related T-contrast images were generated using SPM8 con-trast manager. We ran our models for valence and arousalseparately (i.e., with functions 1, 2, 4, 6, 7, and 8 forarousal and with functions 1, 3, 5, 7, and 8 for valence)and both with and without the quadratic terms (functions4 and 5) to ensure that the model was not over-specified;the findings for the linear arousal and valence terms inthese reduced models were unchanged from findings forthe model that included all eight functions. We thuselected to present findings for the full model so that wecould account for every event that occurs during the task,allowing us to control for signal variability in each trial.Also, by including both linear and quadratic componentswe were able to assess whether the response is truly linearacross the range of ratings or whether it is curvilinear[Acton and Friston, 1998; Buchel et al., 1996, 1998; Fracko-wiak, 2004]. We also ran our model using both the SPMdefault that orthogonalizes parametric variables, as well aswithout orthogonalization, because we were concernedthat if our regressors were inter-correlated and we didorthogonalize our modulators, then the explained variancein BOLD-signal would not be assigned to any of theregressors and our power would be reduced for statisticaltesting. We also wanted to ensure that our findings wererobust with respect to orthogonalization. Our findingswere nearly identical with or without orthogonalization,so we elected to present our findings using the orthogonal-ized analyses. Finally, we also ran the GLMs with motionparameters as regressors and found that they had no sig-nificant effect on our findings, so we elected to presentour findings without motion regressors in the model.

Second-level analysis

We used Bayesian inference to detect random effects byassessing the posterior probability of detecting within orbetween group difference, b, given the activation map thatwe attained in a particular contrast. We used a posteriorprobability of greater than 98.75% as the threshold for sta-tistical significance in each of the contrast maps and, inaddition, required a spatial extent of at least eight contigu-ous voxels to further strengthen the biological validity ofour findings. Unlike a more conventional second-levelanalysis that uses classical parametric inference to detect agroup effect in a statistical parametric map by disprovingthe null hypothesis (b 5 0) at each voxel of the image, agroup effect using the Bayesian method infers the poste-rior probability of detecting the observed group effects(b 6¼ 0), given the data in a posterior probability map [Neu-mann and Lohmann, 2003]. Whereas the voxelwise tests ina statistical parametric map require correction for thenumber of statistical comparisons performed, the Bayesianmethod, because it infers posterior probability, by defini-tion, does not generate false positives and does not requireadjustment of its P values based on stringent P valuethresholding (a feature of these analyses that has been con-firmed in numerous simulations and empirical studies)[Friston and Penny, 2003; Friston et al., 2002].

Post-Hoc Analyses

Several additional analyses ensured that possible con-founding effects did not unduly influence our findings. Weconducted post-hoc analyses while covarying for FSIQ in allparticipants and ADOS scores in analyses involving onlyparticipants with ASD. Results did not differ significantlywhen we covaried for age, sex, FSIQ, or ADOS scores in par-ticipants with ASD. Additionally, we assessed the age-by-diagnosis interaction but found none; restricting our ASDsample to participants who were older than 18-years did notchange our findings from those for our overall sample(ASD: 24% of group (12/51) and TD: 37% of group (31/84)).Additionally, we analyzed our dataset with only male par-ticipants (45 ASD, 62 TD) and found the patterns of activa-tion to be similar to those for our main model (SupportingInformation Fig. S4). We also assessed age correlationswithin each group and detected none that were significantfor valence or arousal. Finally, restricting analyses to partici-pants who were medication-na€ıve (ASD: 68% (34/51)yielded the same results as for our overall sample.

RESULTS

Behavioral Data

On the whole, our behavioral findings suggest thatwhile participants in the ASD group rated arousal andvalence for a wide range of emotions similar to individu-als in the TD group, emotion ratings for the ASD groups

r Tseng et al. r

r 8 r

along both valence and arousal dimensions were some-what constricted in their ranges relative to those of the TDgroups. These findings did not change when we covariedfor overall intelligence. Also, for the ASD group, correla-tions of ADOS scores with arousal and valence yieldedonly one marginally significant finding: ADOS scores cor-related with valence ratings for surprise faces (b 5 0.32,t42 5 2.1, P 5 0.05). Results did not vary by ASD subtype.

Emotion-specific Analyses

Given that emotional processing in typically developingadults is presumably the desired outcome of emotionalprocessing in typical and atypical development, we usedour average Adult TD data as a point of reference for visu-ally comparing data from the other three groups, eventhough our primary analyses of the behavioral data treatedage as a continuous variable. We report subtle but signifi-cant differences between groups for specific emotions,although the overall assignment of valence and arousalscores across the all emotion-types were similar for theASD and TD groups (Fig. 2).

Ratings for both valence and arousal dimensions of emo-tions in the child ASD group were somewhat constricted intheir ranges relative to those of the Adult TD group: the childASD group reported significantly lower arousal ratings forhigh arousal emotions such as Excited (F3,63 5 3.53, P 5 0.0008)and surprised (F3,63 5 3.38, P 5 0.0013), and higher arousal rat-ings for Sleepy, a low arousal emotion, (F3,63 5 2.02, P 5 0.048).They also reported significantly less negative valence ratingsfor negatively valenced emotions, including Disgusted(F3,63 5 2.01, P 5 0.049) and Sad (F3,63 5 2.83, P 5 0.006), and atrend for less positive valence ratings than the adult TD groupfor the positively valenced excited (F3,63 5 1.96, P 5 0.055) andhappy (F3,63 5 1.90, P 5 0.061) (Fig. 2A).

Emotions for the adult ASD group relative to the adultTD group also showed a trend for constriction in theirranges; they reported significantly less negative valenceratings for the negatively valenced sad faces (F3,90 5 2.33,P 5 0.022) (Fig. 2B).

No significant age differences were detected within theASD groups. However, adult TD participants did reporthigher arousal ratings than Child TD participants for thenegatively valenced angry (F3,82 5 2.64, P 5 0.01), disgusted(F3,82 5 2.46, P 5 0.016), sad (F3,82 5 2.82, P 5 0.006), andscared faces (F3,82 5 2.14, P 5 0.036) and less positive valenceratings for excited faces (F3,47 5 2.818, P 5 0.045) (Fig. 2C).

fMRI Data

Main effects

Linear and quadratic correlates of arousal. These analy-ses revealed significant inverse linear and quadratic correla-tions of BOLD-signal with arousal ratings for our TDparticipants in ILPFC and DLPFC, dorsal ACC, inferoposte-rior PC, dorsal PC, CN, and PUT. For ASD participants, wedetected significant positive linear associations of BOLD-signal with arousal ratings in the posterior temporal/infe-rior PC, mesial wall (pregenual and dorsal portions of SFGand ACC), premotor and supplementary motor regions, Cuand PCu, subcortical regions (all basal ganglia nuclei,THAL), and dorsal Cb (Fig. 3, Tables II and IV).

Conjunction maps of the linear and quadratic effects ofarousal in each group show regions where both linear andquadratic effects were detected (e.g., Cb, Broca’s, CN,DLPFC, PCu), and scatterplots show the combined effect ofthe linear and curvilinear components of the correlation(Fig. 4; Supporting Information Fig. S-2, Tables S-1A, S-1C).

Linear and quadratic correlates of valence. BOLD-signalcorrelated with ratings of linear valence and quadraticvalence similarly for both diagnostic groups. In both TDand ASD participants, valence ratings correlated positivelywith BOLD-signal in ACC, FG, and, and inversely withBOLD-signal in posterior PC, S1, M1 and SMA. We didfind regions where the correlation of valence ratings withBOLD-signal differed significantly across the ASD and TDgroups but they were very small in spatial extent and of

Figure 2.

Emotion-specific group comparisons of behavioral findings.

r FMRI of Arousal and Valence Dimensions in ASD r

r 9 r

questionable biological significance (Fig. 5, Tables IIIand V; Supporting Information Tables S-1B, S-1D, S-3).

Interactions

Diagnosis-by-arousal. Our analyses revealed significantdifferences in the correlation of arousal ratings withBOLD-signal across ASD and TD participants (ASD>TD)in AMYG, HIPP, CN, ACC, and SFG (Fig. 3, Table VI).

Diagnosis-by-valence. We only detected minor differen-ces across groups in the correlation of valence ratings with

BOLD-signal, suggesting that brain activity in individualswith ASD is typical for some components of emotion-processing (Fig. 5).

DISCUSSION

Our goal was to assess whether activity in neural systemsunderlying the processing of arousal and valence of facialemotions in individuals with ASD differ from their TDcounterparts. Although participants from both diagnosticgroups performed with a comparable level of accuracy on

Figure 3.

Arousal correlates (A) Regions of significant correlations of

BOLD-signal with ratings of arousal for TD participants. (Posi-

tive correlations are coded in red to yellow, and inverse correla-

tions are coded in green to blue.). (B) Scatterplots of

correlations for BOLD-signal with ratings of arousal for TD par-

ticipants (green) or participants with ASD (purple). (C) Regions

of significant correlations of BOLD-signal with ratings of arousal

for participants with ASD. (D) The regions where the correla-

tion of arousal ratings with BOLD-signal for participants with

ASD differs significantly from the correlation of arousal ratings

with BOLD-signal for TD participants. (ASD>TD coded in red

to yellow, TD >ASD coded in blue to green). (E) Scatterplots

of correlations for BOLD-signal with ratings of arousal for TD

participants (green) and participants with ASD (purple) in

regions where the correlation of arousal ratings with BOLD for

participants with ASD differs significantly from the correlation of

arousal ratings with BOLD for TD participants.

r Tseng et al. r

r 10 r

TABLE II. Regions of significant linear correlations of BOLD signal with ratings of AROUSAL for participants with ASD

Anatomical regionsa

Location MNI coordinates

ZSide BA x y z Correlation

Linear arousal

Premotor cortex R 6, 1 39 228 67 8.21 PositiveDorsolateral prefrontal cortex L 8 26 47 25 8.13 PositiveCerebellum L 26 261 220 8.04 PositiveDorsolateral prefrontal cortex L 9 221 20 58 7.97 PositiveDorsolateral prefrontal cortex R 21 41 49 7.66 PositiveBroca’s area L 251 20 13 4.23 Positive

aRegions were significant at P< 0.001 or posterior probability> 99.0%. (All of the other tables in this document show significance atP< 0.0125 or posterior probability> 98.75%).

TABLE III. Regions of significant linear correlations of BOLD signal with ratings of VALENCE for participants with ASD

Location MNI coordinates

Anatomical regions Side BA x Y z Z Correlation

Linear Valence

Cerebellum L 23 252 238 4.94 PositiveCerebellum R 9 255 238 4.69 PositivePrimary visual cortex R 17/18 30 258 4 4.58 PositiveLingual gyrus R 18 12 288 25 4.28 PositiveDorsolateral prefrontal cortex L 218 62 10 4.01 PositiveDorsolateral prefrontal cortex R 24 62 10 3.97 PositiveHippocampus R 24 216 214 3.79 PositiveDorsolateral prefrontal cortex R 15 53 1 3.79 PositiveFusiform gyrus R 39 222 223 3.75 PositivePrimary visual cortex L 18 215 255 10 3.74 PositiveAnterior cingulate cortex L 29 35 25 3.55 PositiveHippocampus R 27 240 1 3.54 PositiveInferotemporal visual area L 239 231 214 3.29 PositiveCaudate nucleus L 23 20 22 3.2 PositiveDorsolateral prefrontal cortex R 30 44 46 3.2 PositiveOrbitofrontal cortex R 27 32 211 3.04 PositiveBroca’s area L 44 254 20 13 6.93 NegativePrimary somatosensory cortex L 3b 260 24 19 6 NegativePrimary motor cortex L 254 5 28 5.93 NegativeSecondary somatosensory cortex L 260 27 13 5.7 NegativeInsula L 233 23 10 5.33 NegativePosterior parietal cortex R 24 270 49 5.28 NegativeParietal-temporal cortex L 248 240 43 4.85 NegativePrimary somatosensory cortex R 1 63 213 31 4.85 NegativeDorsolateral prefrontal cortex R 27 5 55 4.58 NegativeDorsolateral prefrontal cortex L 236 53 13 4.56 NegativeVisual association cortex R 36 282 34 4.52 NegativeSupplementary motor area R 6 12 11 46 4.48 NegativeParietal-temporal cortex R 45 240 40 4.45 NegativeDorsolateral prefrontal cortex R 44 39 5 43 4.39 NegativeBroca’s area R 44 45 14 34 4.23 NegativeDorsolateral prefrontal cortex L 0 29 43 4.18 NegativeInferotemporal visual area R 54 267 28 4.14 NegativeInferotemporal visual area R 63 255 22 3.77 NegativePrimary visual cortex L 18 29 288 28 3.2 NegativeBroca’s area R 45 54 20 1 3.09 Negative

r FMRI of Arousal and Valence Dimensions in ASD r

r 11 r

the emotion-rating task [Tseng et al., 2014], we detected dra-matic differences across diagnostic groups in the neural activ-ity subserving the dimensions of emotion, particularly arousal.BOLD-signal correlated linearly with ratings of emotionalarousal, but in opposite directions and in differing locationsfor the two groups. BOLD-signal in TD participants correlatedinversely with ratings of arousal in regions associated primar-ily with attentional functions, whereas BOLD-signal in ASDparticipants correlated positively with arousal ratings inregions most commonly associated with impulse control andwith default-mode activity. In contrast, we found that BOLD-signal correlated with ratings of valence similarly across theTD and ASD groups, positively in regions associated previ-ously with processing of emotional faces, and inversely in sen-sorimotor regions. Only minor group differences weredetected in the BOLD-signal correlates with valence ratings.

Arousal-related Activity in TD Participants

Brain regions that correlated inversely with ratings ofarousal in our TD group included the ILPFC, DLPFC, dor-sal ACC, and inferoposterior and dorsal PC, as well as CNand PUT, regions thought to support attentional functions[Petersen and Posner, 2012]. Attention is the means bywhich brains optimize the flexible use of limited cognitiveresources to process prioritized stimuli (i.e., to select task-relevant and ignore task-irrelevant information) [Mansouriet al., 2009], plan for future contingencies, inhibit compet-ing responses, and pursue long-term goals [Penningtonand Ozonoff, 1996]. The neural substrates of attention aregenerally thought to include ACC, inferolateral and pre-frontal cortices, and basal ganglia [Fan et al., 2005; Razand Buhle, 2006], regions that correlated inversely witharousal ratings in the TD group.

The inverse linear correlations of arousal ratings withBOLD-signal in attentional networks suggests that TD partici-pants may have had to allocate progressively more attentionalresources to processing incrementally less-arousing face-stim-uli. Numerous studies have reported that TD participantsview low-arousal faces as ambiguous—i.e., as stimuli thatmay be construed in more than one way, or whose contentrequires greater examination of contextual cues to decipher[Kryklywy et al., 2013; Rosen and Donley, 2006]. Althoughearlier studies of ambiguity in emotional faces focused onfear-related, high-arousal stimuli (such as a fearful face signal-ing potential threat in the environment) [Adolphs, 2010;Whalen, 1998; Whalen et al., 2001], more recent studies havehighlighted the inherent ambiguity in many low-arousal faces(e.g., faces with neutral, bored, or calm expressions) [Thomaset al., 2001; Tottenham et al., 2013]. These low-arousal facesprovide neither clear safety nor danger signals about theimmediate vicinity and so individuals must allocate moreattentional resources to gauge the potential danger of the faces[Adolphs, 2010]. Thus, we posit that TD participants enlistedmore attentional resources when viewing low-arousal facesbecause they were attempting to assess and classify low-arousal, ambiguous stimuli [Petersen and Posner, 2012; Posnerand Petersen, 1990; Reiman et al., 1997]. Consistent with thisinterpretation, prior studies have reported that attentional net-works activate more strongly when processing ambiguousstimuli [Mushtaq et al., 2011; Raz et al., 2007; Volz et al., 2004].

Stimulus Specificity in Activating Attention

Networks during Emotional Tasks

Our current findings and their interpretations, whencombined with those in our prior studies using differing

TABLE IV. Regions of significant linear correlations of BOLD signal with ratings of AROUSAL for TD participants

Anatomical regions

Location MNI coordinates

ZSide BA x y z Correlation

Linear arousal

Inferolateral prefrontal cortex R 27 62 19 7.29 NegativeBroca’s area R 51 41 16 6.96 NegativePosterior parietal cortex R 30 276 52 6.8 NegativeDorsolateral prefrontal cortex R 39 56 4 6.74 NegativePrimary motor cortex R 57 11 34 6.74 NegativeIntraparietal sulcus L 244 246 48 6.73 NegativeParietal-occipital-temporal cortex L 242 255 61 6.73 NegativePrecuneus R 12 270 64 6.34 NegativeCuneus R 21 285 46 6.31 NegativeBroca’s area R 48 20 34 5.69 NegativeInsula L 239 213 10 5.04 NegativeInsula R 36 26 22 4.53 NegativeCaudate nucleus R 15 8 16 4.3 NegativeVisual association cortex L 23 288 25 4.21 NegativeVisual association cortex L 260 210 217 4.09 NegativeHippocampus R 18 234 10 4.09 NegativePutamen L 221 2 7 4.03 Negative

r Tseng et al. r

r 12 r

types of emotional stimuli, suggest that the allocation ofattentional resources may depend critically upon thenature of the emotional stimulus [Colibazzi et al., 2010;Gerber et al., 2008; Landa et al., 2013; Posner et al., 2009].In the present study, for example, we found that neuralactivity within attention networks of the TD group corre-lated inversely with ratings of arousal in emotional faces.Yet in our prior study using a mood-induction task, neuralactivity in attention networks of a different TD group cor-related inversely with ratings of emotional valence, notarousal [Colibazzi et al., 2010; Landa et al., 2013] (i.e., neu-

ral activity during mood-induction increased in attentionnetworks as the valence of mood-inducing stimuli becamemore unpleasant and aversive). Moreover, these correla-tions during mood-induction were significantly strongerfor interpersonal than for non-interpersonal stimuli [Landaet al., 2013]. The inherently interpersonal nature of facesmay account for the similar neural responses of potentiallythreatening faces with aversive interpersonal, mood-inducing situations.

The differential activation of attention circuits inresponse to differing stimuli (emotional faces or induced

Figure 4.

Linear and quadratic correlations of arousal with BOLD-signal in

ASD (A) Regions of significant linear correlations of BOLD-

signal with ratings of arousal for participants with ASD. (Positive

correlations are coded in red to yellow, and inverse correlations

are coded in green to blue.). (B) Scatterplots of linear correla-

tions for BOLD-signal with ratings of arousal for participants

with ASD (orange). (C) Regions of significant linear correlations

of BOLD-signal with ratings of arousal combined with regions of

significant quadratic correlations of arousal ratings for partici-

pants with ASD. Positive linear correlations with arousal are

shown alone (orange), quadratic correlations with arousal alone

(fuschia), and linear1quadratic correlations with arousal (red).

(D) Scatterplots of correlations between BOLD-signal change

and ratings of linear1quadratic correlations with arousal for par-

ticipants with ASD (red) in regions where BOLD correlated

positively for both linear and quadratic correlations with arousal

ratings.

r FMRI of Arousal and Valence Dimensions in ASD r

r 13 r

mood) suggests that we cannot easily distinguish or disen-tangle the allocation of attentional resources from the twomajor dimensions of emotion, valence and arousal. It alsodemonstrates that instead of identifying neural activity thatsubserves or produces the arousal and valence componentsof emotional experience, our fMRI paradigm may be detect-ing varying activity in attention circuits associated with vary-ing individual experiences of emotion. Our findings showthat attention-related activation can dominate overall neuralactivity during emotional tasks, overwhelming the variancein neural activity that produces arousal and valence.

Arousal in Participants with ASD

Regions where BOLD-signal correlated positively witharousal ratings in the ASD group included posterior tem-

poral and inferior PC, the mesial wall (pregenual and dor-sal portions of the superior frontal and anterior cingulategyri), PreMC and supplementary motor cortices, Cu andPCu, all basal ganglia nuclei, THAL, and dorsal Cb. Muchof the pattern of neural activity associated with arousal inparticipants with ASD was similar to the pattern ofregional activation in tasks that require suppression of anautomatic response [Peterson et al., 2002; Viviani, 2013]and in preparing for action. The function of arousal is toprepare an organism for action, and the restraint of thatmotor preparedness from execution during the fMRIexperiment would require activation of an inhibitoryresponse. Alternatively, participants with ASD may findmore-arousing faces to be aversive, thereby requiring acti-vation of control networks to suppress their motoric urgeto withdraw from the putatively noxious stimulus [Peter-son, 2003; Tabu et al., 2012]. Consistent with both

Figure 5.

Valence correlates (A) Regions of significant correlations of

BOLD-signal with ratings of valence for TD participants. (Posi-

tive correlations are coded in red to yellow, and inverse correla-

tions are coded in green to blue.). (B) Scatterplots of

correlations for BOLD-signal with ratings of valence for TD par-

ticipants (green) or participants with ASD (purple). (C) Regions

of significant correlations of BOLD-signal with ratings of valence

for participants with ASD. (D) The regions where the correla-

tion of valence ratings with BOLD for participants with ASD dif-

fers significantly from the correlation of valence ratings with

BOLD for TD participants. (ASD>TD coded in red to yellow,

TD >ASD coded in blue to green). (E) Scatterplots of correla-

tions for BOLD-signal with ratings of valence for TD partici-

pants (green) and participants with ASD (purple) in regions

where the correlation of valence ratings with BOLD for partici-

pants with ASD differs significantly from the correlation of

valence ratings with BOLD for TD participants.

r Tseng et al. r

r 14 r

interpretations, children with ASD reportedly have astronger skin conductance response, and therefore aremore aroused, when viewing faces with a direct comparedwith averted gaze, and compared to TD children in eithercondition [Kylliainen et al., 2012]. Also consistent withboth interpretations are the maps of curvilinear (quadratic)responses of neural activity related to arousal ratings,which showed disproportionately stronger BOLDresponses to more arousing stimuli in regions that supportinhibitory responses, including dorsal frontal, ACC, andbasal ganglia regions (Fig. 4).

Several regions that correlated positively with arousalratings in the ASD group, including medial prefrontal,anterior cingulate, posterior cingulate, and precuneus cor-tices, are also components of the default mode network(DMN). Activity in the DMN is greatest at rest anddeclines during cognitively demanding tasks [Raichleet al., 2001]. In TD individuals, the DMN activates during

performance of various social, emotional, and introspectivetasks [Gusnard et al., 2001; Raichle et al., 2001]. Prior stud-ies suggest that persons with ASD may not deactivate theDMN sufficiently when performing cognitively demandingtasks [Kennedy and Courchesne, 2008]. In participantswith ASD, however, we found proportionately greaterneural activity in DMN regions, presumably representinga more strongly felt social and emotional experience withprogressively more-arousing faces.

Group Similarities and Differences in Arousal

Correlates

We were surprised to find the strikingly different pat-terns of activation to arousal ratings across the TD andASD groups. Participants in our TD group increased neu-ral activity in their attention networks when rating less-

TABLE V. Regions of significant linear correlations of BOLD signal with ratings of VALENCE for TD participants

Location MNI coordinates

ZAnatomical regions Side BA x y z Correlation

Linear valence

Primary visual cortex R 17 12 297 4 7.87 PositiveCuneus R 18 21 2100 10 6.44 PositiveOrbitofrontal cortex L 0 56 28 4.65 PositivePrecuneus L 0 258 19 3.42 PositiveVisual association cortex L 18 215 2103 1 3.23 PositiveDorsolateral prefrontal cortex R 3 59 10 3.2 PositiveInferotemporal visual area R 45 273 25 3.15 PositiveAnterior cingulate cortex L 23 26 211 3.05 PositiveVisual association cortex L 242 273 25 2.91 PositiveFusiform gyrus L 230 231 226 2.89 PositivePosterior parietal cortex R 18 276 58 4.51 NegativeParietal-temporal cortex L 245 255 58 4.25 NegativeSupplementary motor area L 0 20 58 4.16 NegativePrecuneus R 9 261 49 4.11 NegativeDorsolateral prefrontal cortex R 18 8 70 3.98 NegativeParietal-occipital-temporal cortex L 230 252 37 3.97 NegativeRolandic operculum R 39 219 22 3.95 NegativeParietal-temporal cortex L 251 249 55 3.85 NegativePrecuneus L 26 279 43 3.84 NegativeMiddle cingulate cortex L 29 11 40 3.72 NegativePrimary motor cortex R 6 51 210 52 3.68 NegativePrimary somatosensory cortex R 6 30 231 67 3.67 NegativeDorsolateral prefrontal cortex L 26 17 43 3.6 NegativeLingual gyrus L 215 285 28 3.57 NegativeInferotemporal visual area R 63 255 7 3.56 NegativeCuneus L 26 285 34 3.54 NegativePrimary motor cortex L 239 5 37 3.51 NegativeBroca’s area L 248 41 1 3.33 NegativeDorsolateral prefrontal cortex L 236 44 28 3.23 NegativeAuditory association cortex R 54 240 13 3.23 NegativeDorsolateral prefrontal cortex R 36 35 22 3.03 NegativeMiddle cingulate cortex L 26 219 34 2.93 NegativeAnterior cingulate cortex L 29 26 28 2.72 NegativeMiddle cingulate cortex R 12 222 34 2.6 Negative

r FMRI of Arousal and Valence Dimensions in ASD r

r 15 r

arousing faces, a pattern absent in the ASD group. Thesestrong inverse correlations of BOLD signal with arousal rat-ings in attentional networks of the TD group is counter tothe more intuitive expectation that highly arousing or emo-tionally engaging stimuli would be more attention-grabbingthan less arousing and less emotionally engaging stimuli.

In contrast, individuals in the ASD group demonstratedstrong positive correlations with arousal ratings in regula-tory systems (IPC, SFG, THAL, CN, Cu) and the DMN(ACC, PCC), a pattern similar to that shown in TD controlsin a prior mood-induction study from our laboratory [Coli-bazzi et al., 2010]. This similarity suggests that participantswith ASD, more than the TD controls, may have beenincreasingly immersed in the emotions expressed by themore arousing face stimuli of the present study. In otherwords, activation of the regulatory systems and DMN mayrequire the absorbing, vitalizing experience of interpersonalstimuli, whether they are more highly arousing faces forpersons with ASD in the present study or more arousinginterpersonal moods for TD individuals in our prior study.

Thus TD participants in the present study may haveinstead approached the rating of facial emotions (asopposed to the mood induction task of our prior study) asmore of a cognitive than a mood-inducing task, with the rat-ing of progressively less arousing, more ambiguous, andpotentially threatening faces requiring a progressivelygreater allocation of attentional resources. Individuals withASD, in contrast, seem either not to have experienced low-arousal faces as threatening, or they found those faces lesssalient and less attention-grabbing than did the TD controls.

Valence

Only minor differences were detected between diagnos-tic groups in the BOLD signal correlates of valence ratings.In ASD and TD groups, BOLD-signal correlated positivelywith valence ratings in ACC, FG, and OFC, regions impli-

cated in facial emotion processing. Studies of adults sug-gest, for example, that FG supports the perceptualidentification of faces [Haxby et al., 2000] and more specif-ically the coding of fearful (high-arousal, high-valence)faces [Pessoa et al., 2002; Vuilleumier et al., 2001]. Thenear absence of significant group differences suggests thatindividuals with ASD do not activate the brain atypicallyfor all components of emotion-processing.

Limitations

One prominent limitation is the absence of eye-trackingdata during participants’ viewing of emotional faces. Priorstudies have shown that individuals with ASD do notspontaneously attend to, and they may even avoid, theeyes of face-stimuli, even though the eyes are a rich sourceof information about emotional states [Klin et al., 2002].Although less attention to the eyes of our face-stimulicould have impaired the ability of participants with ASDto recognize and rate accurately both valence and arousal[Kliemann et al., 2010], if participants with ASD had notattended to task-stimuli, their valence and arousal ratingswould have differed from the ratings of TD participants[Tseng et al., 2014]. Nevertheless, we cannot exclude thepossibility that subtle group differences in attention to spe-cific facial features influenced our findings.

Additionally, as is often the case in research on ASD,we struggled when designing our experiment with thetrade-offs between task difficulty, selection of a task thatcan provide scientifically important information, and thegeneralizability of the study and its findings to the entireautism spectrum. We considered multiple issues simulta-neously. In particular, we needed to include in our studyindividuals who would and did understand a task thataddressed meaningfully our fundamental research ques-tions. Nonverbal individuals, for example, would beunlikely to understand or perform our task adequately.

TABLE VI. Centers of activation in regions where the linear correlation of BOLD signal intensity with ratings of

AROUSAL for participants with ASD differs significantly from the linear correlation of BOLD with ratings of

AROUSAL for TD participants

Anatomical regionsa

Location MNI coordinates

ZSide BA x Y z Correlation

Linear arousal

Broca’s area L 45 248 41 25 5.68 ASD>TDDorsolateral prefrontal cortex L 9 224 38 46 5.12 ASD>TDDorsolateral prefrontal cortex R 8 12 50 49 5.09 ASD>TDBroca’s area R 45 57 26 16 4.95 ASD>TDCaudate nucleus R 18 14 13 4.68 ASD>TDHippocampus R 21 228 25 4.56 ASD>TDAmygdala R 27 21 217 4.41 ASD>TDAnterior cingulate cortex R 32 12 44 16 4.19 ASD>TD

aRegions were significant at P< 0.001 or posterior probability> 99.0%. (All of the other tables in this document show significance atP< 0.0125 or posterior probability> 98.75%).

r Tseng et al. r

r 16 r

Also, if we had included lower functioning persons withASD (i.e., those with lower IQs), we would have had toinclude control participants with comparable levels ofintelligence, which in turn would introduce a host of con-founding variables and sample heterogeneity that wouldmake interpretation of findings difficult. We were carefulto also covary for full-scale IQ, as well as for age and sex,and detected no significant differences in our findings fordiagnosis. Additionally, the individuals with ASD in oursample ranged in ASD diagnosis from PDD-NOS toAsperger’s to Autism (mean ADOS Score 5 10.9 6 3.1),suggesting that we can extrapolate our findings to individ-uals with moderate to high-functioning ASD.

CONCLUSIONS

Our findings provide unique insight into the emotionalexperiences of individuals with ASD. Although behavioralresponses to face-stimuli were similar across diagnosticgroups, the corresponding neural activity of the behavioralresponses for arousal differed prominently across groups.TD individuals and persons with ASD seem to find differ-ing aspects of emotional stimuli to be salient and relevant.Studying these differences may help us to understand theorigins of atypical interpersonal emotional experiences inindividuals with ASD.

REFERENCES

Acton PD, Friston KJ (1998): Statistical parametric mapping infunctional neuroimaging: Beyond PET and fMRI activationstudies. Eur J Nucl Med 25:663–667.

Adolphs R (2010): What does the amygdala contribute to socialcognition? Ann N Y Acad Sci 1191:42–61.

American Psychiatric Association (2000): Diagnostic and StatisticalManual of Mental Disorders, DSM-IV-TR. Arlington: AmericanPsychiatric Publishing, Incorporated.

Ashwin C, Chapman E, Colle L, Baron-Cohen S (2006): Impairedrecognition of negative basic emotions in autism: A test of theamygdala theory. Soc Neurosci 1:349–363.

Ashwin C, Baron-Cohen S, Wheelwright S, O’riordan M, BullmoreET (2007): Differential activation of the amygdala and the“social brain” during fearful face-processing in Asperger syn-drome. Neuropsychologia 45:2–14.

Bachevalier J, Loveland KA (2006): The orbitofrontal-amygdalacircuit and self-regulation of social-emotional behavior inautism. Neurosci Biobehav Rev 30:97–117.

Buchel C, Wise RJ, Mummery CJ, Poline JB, Friston KJ (1996):Nonlinear regression in parametric activation studies. Neuro-image 4:60–66.

Buchel C, Holmes AP, Rees G, Friston KJ (1998): Characterizingstimulus-response functions using nonlinear regressors inparametric fMRI experiments. Neuroimage 8:140–148.

Castelli F (2005): Understanding emotions from standardizedfacial expressions in autism and normal development. Autism9:428–449.

Celani G, Battacchi MW, Arcidiacono L (1999): The understandingof the emotional meaning of facial expressions in people withautism. J Autism Dev Disord 29:57–66.

Colibazzi T, Posner J, Wang Z, Gorman D, Gerber A, Yu S, Zhu

H, Kangarlu A, Duan Y, Russell JA, Peterson BS (2010): Neural

systems subserving valence and arousal during the experience

of induced emotions. Emotion 10:377–389.Dapretto M, Davies MS, Pfeifer JH, Scott AA, Sigman M,

Bookheimer SY, Iacoboni M (2006): Understanding emotions in

others: Mirror neuron dysfunction in children with autism

spectrum disorders. Nat Neurosci 9:28–30.Deeley Q, Daly EM, Surguladze S, Page L, Toal F, Robertson D,

Curran S, Giampietro V, Seal M, Brammer MJ, Andrew C,

Murphy K, Phillips ML, Murphy DG (2007): An event related

functional magnetic resonance imaging study of facial emotion

processing in Asperger syndrome. Biol Psychiatry 62:207–217.Di Martino A, Ross K, Uddin LQ, Sklar AB, Castellanos FX,

Milham MP (2009): Functional brain correlates of social and

nonsocial processes in autism spectrum disorders: An activa-

tion likelihood estimation meta-analysis. Biol Psychiatry 65:63–

74.Ekman P (1992): Are there basic emotions? Psychol Rev 99:550–

553.Ekman P, Friesen WV (1976): Pictures of Facial Affect. Palo Alto,

CA: Consulting Psychologists Press.Fan J, Mccandliss BD, Fossella J, Flombaum JI, Posner MI (2005):

The activation of attentional networks. Neuroimage 26:471–479.Frackowiak RSJ (2004): Human Brain Function. Amsterdam, Bos-

ton: Elsevier Academic Press.Friston KJ, Glaser DE, Henson RN, Kiebel S, Phillips C,

Ashburner J (2002): Classical and Bayesian inference in neuroi-

maging: Applications. Neuroimage 16:484–512.Friston KJ, Penny W (2003): Posterior probability maps and SPMs.

Neuroimage 19:1240–1249.Fusar-Poli P, Placentino A, Carletti F, Landi P, Allen P,

Surguladze S, Benedetti F, Abbamonte M, Gasparotti R, Barale

F, Perez J, Mcguire P, Politi P (2009): Functional atlas of emo-

tional faces processing: A voxel-based meta-analysis of 105

functional magnetic resonance imaging studies. J Psychiatry

Neurosci 34:418–432.Gerber AJ, Posner J, Gorman D, Colibazzi T, Yu S, Wang Z,

Kangarlu A, Zhu H, Russell J, Peterson BS (2008): An affective

circumplex model of neural systems subserving valence,

arousal, and cognitive overlay during the appraisal of emo-

tional faces. Neuropsychologia 46:2129–2139.Golan O, Baron-Cohen S, Hill J (2006): The Cambridge Mindread-

ing (CAM) Face-Voice battery: Testing complex emotion recog-nition in adults with and without Asperger syndrome.

J Autism Dev Disord 36:169–183.Gotham K, Risi S, Pickles A, Lord C (2007): The Autism diagnostic

observation schedule: Revised algorithms for improved diag-

nostic validity. J Autism Dev Disord 37:613–627.Grelotti DJ, Gauthier I, Schultz RT (2002): Social interest and the

development of cortical face specialization: What autism

teaches us about face processing. Dev Psychobiol 40:213–225.Gusnard DA, Akbudak E, Shulman GL, Raichle ME (2001): Medial

prefrontal cortex and self-referential mental activity: Relation

to a default mode of brain function. Proc Natl Acad Sci USA

98:4259–4264.Hadjikhani N, Joseph RM, Snyder J, Tager-Flusberg H (2006):

Anatomical differences in the mirror neuron system and social

cognition network in autism. Cereb Cortex 16:1276–1282.Hall GB, Szechtman H, Nahmias C (2003): Enhanced salience and

emotion recognition in Autism: A PET study. Am J Psychiatry

160:1439–1441.

r FMRI of Arousal and Valence Dimensions in ASD r

r 17 r

Hamann S (2012): Mapping discrete and dimensional emotions

onto the brain: Controversies and consensus. Trends Cogn Sci

16:458–466.Harms MB, Martin A, Wallace GL (2010): Facial emotion recogni-

tion in autism spectrum disorders: A review of behavioral and

neuroimaging studies. Neuropsychol Rev 20:290–322.Haxby JV, Hoffman EA, Gobbini MI (2000): The distributed

human neural system for face perception. Trends Cogn Sci 4:

223–233.Hollingshead A (1975): Four-factor Index of Social Status. New

Haven, CT: Yale University Press.Hubl D, Bolte S, Feineis-Matthews S, Lanfermann H, Federspiel

A, Strik W, Poustka F, Dierks T (2003): Functional imbalance

of visual pathways indicates alternative face processing strat-

egies in autism. Neurology 61:1232–1237.Katsyri J, Saalasti S, Tiippana K, Von Wendt L, Sams M (2008):

Impaired recognition of facial emotions from low-spatial fre-

quencies in Asperger syndrome. Neuropsychologia 46:1888–

1897.Kennedy DP, Courchesne E (2008): Functional abnormalities of

the default network during self- and other-reflection in autism.

Soc Cogn Affect Neurosci 3:177–190.Killgore WD, Yurgelun-Todd DA (2004): Activation of the amyg-

dala and anterior cingulate during nonconscious processing of

sad versus happy faces. Neuroimage 21:1215–1223.Klein F, Iffland B, Schindler S, Wabnitz P, Neuner F (2015): This

person is saying bad things about you: The influence of physi-

cally and socially threatening context information on the proc-

essing of inherently neutral faces. Cogn Affect Behav Neurosci.Kliemann D, Dziobek I, Hatri A, Steimke R, Heekeren HR (2010):

Atypical reflexive gaze patterns on emotional faces in autism

spectrum disorders. J Neurosci 30:12281–12287.Klin A, Jones W, Schultz R, Volkmar F, Cohen D (2002): Visual

fixation patterns during viewing of naturalistic social situations

as predictors of social competence in individuals with autism.

Arch Gen Psychiatry 59:809–816.Kryklywy JH, Nantes SG, Mitchell DG (2013): The amygdala enco-

des level of perceived fear but not emotional ambiguity in vis-

ual scenes. Behav Brain Res 252:396–404.Kylliainen A, Wallace S, Coutanche MN, Leppanen JM, Cusack J,

Bailey AJ, Hietanen JK (2012): Affective-motivational brain

responses to direct gaze in children with autism spectrum dis-

order. J Child Psychol Psychiatry 53:790–797.Landa A, Wang Z, Russell JA, Posner J, Duan Y, Kangarlu A,

Huo Y, Fallon BA, Peterson BS (2013): Distinct neural circuits

subserve interpersonal and non-interpersonal emotions. Soc

Neurosci 8:474–488.Lord C, Rutter M, Goode S, Heemsbergen J, Jordan H, Mawhood

L, Schopler E (1989): Autism diagnostic observation schedule:

A standardized observation of communicative and social

behavior. J Autism Dev Disord 19:185–212.Lord C, Rutter M, Le Couteur A (1994): Autism diagnostic inter-

view-revised: A revised version of a diagnostic interview for

caregivers of individuals with possible pervasive developmen-

tal disorders. J Autism Dev Disord 24:659–685.Loveland KA, Steinberg JL, Pearson DA, Mansour R, Reddoch S

(2008): Judgments of auditory-visual affective congruence inadolescents with and without autism: A pilot study of a new

task using fMRI. Percept Mot Skills 107:557–575.Mansouri FA, Tanaka K, Buckley MJ (2009): Conflict-induced

behavioural adjustment: A clue to the executive functions of

the prefrontal cortex. Nat Rev Neurosci 10:141–152.

Mazaika P, Hoeft F, Glover GH, Reiss AL (2009): Methods and

software for fMRI analysis for clinical subjects. Hum Brain

Mapp.Murphy FC, Nimmo-Smith I, Lawrence AD (2003): Functional

neuroanatomy of emotions: A meta-analysis. Cogn Affect

Behav Neurosci 3:207–233.Mushtaq F, Bland AR, Schaefer A (2011): Uncertainty and cogni-

tive control. Front Psychol 2:249.Neumann J, Lohmann G (2003): Bayesian second-level analysis of

functional magnetic resonance images. Neuroimage 20:1346–

1355.Ogai M, Matsumoto H, Suzuki K, Ozawa F, Fukuda R, Uchiyama

I, Suckling J, Isoda H, Mori N, Takei N (2003): fMRI study of

recognition of facial expressions in high-functioning autistic

patients. Neuroreport 14:559–563.Oldfield RC (1971): The assessment and analysis of handedness:

The Edinburgh inventory. Neuropsychologia 9:97–113.Ozonoff S, Pennington BF, Rogers SJ (1990): Are there emotion

perception deficits in young autistic children? J Child Psychol

Psychiatry 31:343–361.Panksepp J (1992): A critical role for “affective neuroscience” in

resolving what is basic about basic emotions. Psychol Rev 99:

554–560.Pelphrey KA, Morris JP, Mccarthy G, Labar KS (2007): Perception

of dynamic changes in facial affect and identity in autism. Soc

Cogn Affect Neurosci 2:140–149.Pennington BF, Ozonoff S (1996): Executive functions and devel-

opmental psychopathology. J Child Psychol Psychiatry 37:

51–87.Pessoa L, Mckenna M, Gutierrez E, Ungerleider LG (2002): Neural

processing of emotional faces requires attention. Proc Natl

Acad Sci USA 99:11458–11463.Peterson BS (2003): Conceptual, methodological, and statistical

challenges in brain imaging studies of developmentally based

psychopathologies. Dev Psychopathol 15:811–832.Petersen SE, Posner MI (2012): The attention system of the human

brain: 20 years after. Annu Rev Neurosci 35:73–89.Peterson BS, Kane MJ, Alexander GM, Lacadie C, Skudlarski P,

Leung HC, May J, Gore JC (2002): An event-related functional

MRI study comparing interference effects in the Simon and

Stroop tasks. Brain Res Cogn Brain Res 13:427–440.Piggot J, Kwon H, Mobbs D, Blasey C, Lotspeich L, Menon V,

Bookheimer S, Reiss AL (2004): Emotional attribution in high-

functioning individuals with autistic spectrum disorder: A

functional imaging study. J Am Acad Child Adolesc Psychia-

try 43:473–480.Posner MI, Petersen SE (1990): The attention system of the human

brain. Annu Rev Neurosci 13:25–42.Posner J, Russell JA, Peterson BS (2005): The circumplex model of

affect: An integrative approach to affective neuroscience, cog-

nitive development, and psychopathology. Dev Psychopathol

17:715–734.Posner J, Russell JA, Gerber A, Gorman D, Colibazzi T, Yu S,

Wang Z, Kangarlu A, Zhu H, Peterson BS (2009): The neuro-

physiological bases of emotion: An fMRI study of the affective

circumplex using emotion-denoting words. Hum Brain Mapp

30:883–895.Raichle ME, Macleod AM, Snyder AZ, Powers WJ, Gusnard DA,

Shulman GL (2001): A default mode of brain function. Proc

Natl Acad Sci USA 98:676–682.Raz A, Buhle J (2006): Typologies of attentional networks. Nat

Rev Neurosci 7:367–379.

r Tseng et al. r

r 18 r

Raz A, Lamar M, Buhle JT, Kane MJ, Peterson BS (2007): Selectivebiasing of a specific bistable-figure percept involves fMRI sig-nal changes in frontostriatal circuits: A step toward unlockingthe neural correlates of top-down control and self-regulation.Am J Clin Hypn 50:137–156.

Reiman EM, Lane RD, Ahern GL, Schwartz GE, Davidson RJ,Friston KJ, Yun LS, Chen K (1997): Neuroanatomical correlatesof externally and internally generated human emotion. Am JPsychiatry 154:918–925.

Riby DM, Doherty-Sneddon G, Bruce V (2008): Exploring face per-ception in disorders of development: Evidence from Williamssyndrome and autism. J Neuropsychol 2:47–64.

Rosen JB, Donley MP (2006): Animal studies of amygdala functionin fear and uncertainty: Relevance to human research. BiolPsychol 73:49–60.

Rump KM, Giovannelli JL, Minshew NJ, Strauss MS (2009): Thedevelopment of emotion recognition in individuals withautism. Child Dev 80:1434–1447.

Russell JA (1980): A circumplex model of affect. J Pers Social Psy-chol 39:1161–1178.

Russell JA (2003): Core affect and the psychological construction ofemotion. Psychol Rev 110:145–172.

Russell JA (2005): Emotion in human consciousness is built oncore affect. J Consciousness Stud 12:26–42.

Russell J, Bullock M (1985): Multidimensional scaling of emotionalfacial expressions: Similarity from preschoolers to adults.J Pers Social Psychol 48:1290–1298.

Russell J, Weiss A, Mendelsohn GA (1989): Affect grid: A single-item scale of pleasure and arousal. J Pers Social Psychol 57:493–502.

Rutherford MD, Towns AM (2008): Scan path differences and sim-ilarities during emotion perception in those with and withoutautism spectrum disorders. J Autism Dev Disord 38:1371–1381.

Sasson NJ (2006): The development of face processing in autism.J Autism Dev Disord 36:381–394.

Tabu H, Mima T, Aso T, Takahashi R, Fukuyama H (2012): Com-mon inhibitory prefrontal activation during inhibition of handand foot responses. Neuroimage 59:3373–3378.

Tanaka JW, Wolf JM, Klaiman C, Koenig K, Cockburn J, HerlihyL, Brown C, Stahl S, Kaiser MD, Schultz RT (2010): Using com-puterized games to teach face recognition skills to childrenwith autism spectrum disorder: The Let’s Face It! program.J Child Psychol Psychiatry 51:944–952.

Tantam D, Monaghan L, Nicholson H, Stirling J (1989): Autisticchildren’s ability to interpret faces: A research note. J ChildPsychol Psychiatry 30:623–630.

Thomas KM, Drevets WC, Whalen PJ, Eccard CH, Dahl RE, RyanND, Casey BJ (2001): Amygdala response to facial expressionsin children and adults. Biol Psychiatry 49:309–316.

Tottenham N, Phuong J, Flannery J, Gabard-Durnam L, Goff B(2013): A negativity bias for ambiguous facial-expressionvalence during childhood: Converging evidence from behaviorand facial corrugator muscle responses. Emotion 13:92–103.

Tseng A, Bansal R, Liu J, Gerber AJ, Goh S, Posner J, Colibazzi T,Algermissen M, Chiang IC, Russell JA, Peterson BS (2014):Using the circumplex model of affect to study valence andarousal ratings of emotional faces by children and adults withautism spectrum disorders. J Autism Dev Disord 44:1332–1346.

Uljarevic M, Hamilton A (2013): Recognition of emotions inautism: A formal meta-analysis. J Autism Dev Disord 43:1517–1526.

Viviani R (2013): Emotion regulation, attention to emotion, andthe ventral attentional network. Front Hum Neurosci 7:746.

Volz KG, Schubotz RI, Von Cramon DY (2004): Why am I unsure?Internal and external attributions of uncertainty dissociated byfMRI. Neuroimage 21:848–857.

Vuilleumier P, Armony JL, Driver J, Dolan RJ (2001): Effects ofattention and emotion on face processing in the human brain:An event-related fMRI study. Neuron 30:829–841.

Vytal K, Hamann S (2010): Neuroimaging support for discreteneural correlates of basic emotions: A voxel-based meta-analy-sis. J Cogn Neurosci 22:2864–2885.

Wang AT, Dapretto M, Hariri AR, Sigman M, Bookheimer SY(2004): Neural correlates of facial affect processing in childrenand adolescents with autism spectrum disorder. J Am AcadChild Adolesc Psychiatry 43:481–490.

Wechsler D (1999): Wechsler Abbreviated Scale of Intelligence(WASI). San Antonio, TX: Harcourt Assessment.

Whalen PJ (1998): Fear, vigilance and ambiguity: Initial neuroi-maging studies of the human amygdala. Curr Dir Psychol Sci7:177–188.

Whalen PJ, Shin LM, Mcinerney SC, Fischer H, Wright CI, RauchSL (2001): A functional MRI study of human amygdalaresponses to facial expressions of fear versus anger. Emotion 1:70–83.

r FMRI of Arousal and Valence Dimensions in ASD r

r 19 r