Detailed Exploration of Face-related Processing inCongenital Prosopagnosia 2 Functional

Neuroimaging Findings

Galia Avidan1 Uri Hasson2 Rafael Malach2 and Marlene Behrmann1

Abstract

amp Specific regions of the human occipito-temporal cortexare consistently activated in functional imaging studies offace processing To understand the contribution of theseregions to face processing we examined the pattern of fMRIactivation in four congenital prosopagnosic (CP) individualswho are markedly impaired at face processing despitenormal vision and intelligence and with no evidence ofbrain damage These individuals evinced a normal pattern offMRI activation in the fusiform gyrus (FFA) and in otherventral occipito-temporal areas in response to faces build-ings and other objects shown both as line drawings in

detection and discrimination tasks and under more natural-istic testing conditions when no task was required CPindividuals also showed normal adaptation levels in a block-design adaptation experiment and like control subjectsexhibited evidence of global face representation in the FFAThe absence of a BOLDndashbehavioral correlation (profoundbehavioral deficit normal face-related activation in theventral occipito-temporal cortex) challenges existing accountsof face representation and suggests that activation in thesecortical regions per se is not sufficient to ensure intact faceprocessing amp

INTRODUCTION

Findings from neuropsychology (Wada amp Yamamoto2001 De Renzi 1997 Farah 2004 Sergent amp Signoret1992 Damasio Tranel amp Damasio 1990) event-relatedpotential (ERP) studies (McCarthy Puce Belger ampAllison 1999 Bentin Allison Puce Perez amp McCarthy1996) and single-unit recording studies in monkeys(Gross Rocha-Miranda amp Bender 1972) implicate spe-cific regions within the occipito-temporal cortex inthe processing of faces Consistent with this func-tional imaging studies in humans (Halgren et al 1999Kanwisher McDermott amp Chun 1997 McCarthy PuceGore amp Allison 1997 Clark et al 1996) have docu-mented face-selective activation in a core set of regions(Rossion Caldara et al 2003 Haxby Hoffman ampGobbini 2000) including the fusiform gyrus (LevyHasson Avidan Hendler amp Malach 2001 KanwisherMcDermott et al 1997 McCarthy Puce et al 1997)the lateral occipital region in the vicinity of the lateraloccipital sulcus and inferior occipital gyrus (HassonHarel Levy amp Malach 2003 Gauthier et al 2000Hoffman amp Haxby 2000) and the superior temporalsulcus (STS) (Hoffman amp Haxby 2000 Puce AllisonBentin Gore amp McCarthy 1998) Because face-related

activation in the fusiform gyrus is the most robust(Haxby Hoffman et al 2000) this region is termedlsquolsquothe fusiform face arearsquorsquo (FFA) (Kanwisher McDermottet al 1997) in recognition of its putative modular face-dedicated property (Kanwisher 2000 McCarthy Puceet al 1997)

Although there is general consensus that the FFAis involved in face processing its precise role remainscontroversial One possibility is that it mediates facedetection giving rise to a signal that differentiates be-tween faces and all other visual objects Such a role isconsistent with findings showing that this region re-sponds equally well to a range of face stimuli includinghuman faces and cat and cartoon faces (Tong NakayamaMoscovitch Weinrib amp Kanwisher 2000) Additionallyface detection (but not face identification) is unaffectedby the orientation of the face and likewise the FFAresponse is minimally affected by face inversion (AguirreSingh amp DrsquoEsposito 1999 Haxby Ungerleider et al1999 Kanwisher Tong amp Nakayama 1998)

However such a minimal detection role for the FFAis inconsistent with the fact that individuals with ac-quired prosopagnosia an impairment in face pro-cessing following a lesion to the occipito-temporalcortex perform well in face detection tasks (Bruyeret al 1983) and rather are impaired at discriminatingbetween andor identifying faces (Mundel et al 2003Wada amp Yamamoto 2001) Moreover electrical stimula-

1Carnegie Mellon University Pittsburgh 2Weizmann Instituteof Science Rehovot

D 2005 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 177 pp 1150ndash1167

tion of face-specific ERP sites located in the fusiformgyrus most commonly results in a transient failure toidentify rather than detect faces (Mundel et al 2003Puce Allison amp McCarthy 1999) Finally a recent fMRIstudy showed that although a major component ofthe FFA BOLD activation was correlated with successfulface detection a significant additional signal increasewas correlated with successful face identification (Grill-Spector Knouf amp Kanwisher 2004) Note however thatdirect evidence supporting the contribution of the FFAto face identification is not clear-cut either Thus al-though some studies do not find a differential responseto familiar versus unfamiliar faces (Leveroni et al 2000Nakamura et al 2000) as might be expected of an areamediating identification others do report such differ-ences (Henson Shallice amp Dolan 2000 Rossion SchiltzRobaye Pirenne amp Crommelinck 2001)

To address the ongoing controversy concerning therole of the FFA in face processing we adopted adifferent approach in which we examined the extent towhich there is face-selective activation in the FFA in agroup of unusual individuals who are impaired at facerecognition despite normal visual acuity and intelli-gence a deficit termed lsquolsquocongenital prosopagnosiarsquorsquo(CP) In contrast with acquired prosopagnosic patientsin whom the ventral occipito-temporal cortex has beendamaged CP individuals exhibit an impairment in facerecognition that is present from early childhood in theabsence of any discernible cortical lesion or neurologicaldisease As such these individuals provide us with awindow onto the function of the fusiform gyrus andenable us to examine the neural mechanisms mediatingface recognition using fMRI unaffected by damage thatmight disrupt normal blood flow or neurovasculariza-tion (DrsquoEsposito et al 2003)

To date there have been relatively few studies of CPindividuals (eg Behrmann amp Avidan 2005 Duchaineamp Nakayuma 2005 Grueter et al in press Kress amp Daum2003a Duchaine 2000 De Haan amp Campbell 1991McConachie 1976) only two of which include ERP re-cordings (Kress amp Daum 2003b Bentin Deouell ampSoroker 1999) and two of which include functional imag-ing findings (Hasson Avidan Deouell Bentin amp Malach2003 Hadjikhani amp De Gelder 2002) The present studyhowever is the first to provide both detailed and concur-rent functional imaging and behavioral findings in a groupof four CP individuals To derive brainndashbehavior correla-tions fully we first assessed the behavioral performanceof CP individuals on faces common objects and novelobjects in a variety of tasks (see Behrmann AvidanMarotta amp Kimchi 2005) Having definitively establishedthat all CP subjects were markedly impaired relative tocontrol subjects we then carried out an extensive set offour different imaging experiments two of which havebeen used previously (Hasson Avidan et al 2003) andtwo of which involve new paradigms We used whole-brain scanning to explore the differences between CP

individuals and their controls across the entire cortexand not only in occipito-temporal regions as was donepreviously Finally and critically behavioral responsescollected during the scanning session enabled us toshow directly that even when CP subjects performedpoorly in a face-discrimination task their face-relatedactivation in the FFA was not differentiable from that ofthe normal control subjects The dissociation betweenthe apparently normal FFA BOLD activation and theimpairment in face processing in CP challenges existingaccounts of face processing and forces us to reconcep-tualize the role of the FFA in the representation of faces

RESULTS

Behavioral Profile of CP Subjects

To document the extent and specificity of the behav-ioral impairment CP subjects and 12 normal controlsfirst completed a set of behavioral experiments with facesand nonface objects (for a detailed description of thedifferent experiments and their results see Behrmannet al 2005) In brief relative to their controls all CPsubjects were significantly poorer in recognizing familiarfaces and in making samedifferent discrimination judg-ments on unfamiliar faces CP subjects were also im-paired although to a lesser extent and with greatervariability on tasks involving nonface stimuli All CPsubjects performed within the normal range on vari-ous low-level visual processing tasks Taken togetherthe behavioral data establish a clear deficit in face pro-cessing for all CP subjects

Imaging Experiments

Conventional Face and Object Mapping Experiment

fMRI responses for line drawings of faces buildingsobjects and patterns were first mapped for the CPsubjects and 10 control subjects Stimuli were presentedin a short block design and subjects performed a one-back memory task while fixating on a central dot(Figure 1A) Note that this one-back task is equivalentto a sequential face-discrimination task (lsquolsquoare consecu-tive images the same or differentrsquorsquo) We also mappedout the meridian borders of visual areas for each subjectto assess the location of face- and object-related activa-tion in relation to an objective definition of the visualretinotopic regions (Levy et al 2001)

Importantly even though the one-back memory taskperformed in the scanner was relatively easy as evi-dent from the high accuracy of the control subjects(Figure 1B) the CP group exhibited a significant andspecific behavioral impairment for faces reflected bothin accuracy and RT (Figure 1B) (ANOVA for accuracysignificant stimulus type group interaction p lt 05no significant main effect of stimulus type p = 09 norof group p = 24 ANOVA for RT significant stimulus

Avidan et al 1151

type group interaction p lt 05 significant main effectof stimulus type p lt 05 no significant main effect ofgroup p = 55) As evident from the figure howeverthere was some variability within the CP group withsome subjects trading off speed for accuracy or viceversa (compare subject MT with TM)

For each subject we first localized face-related activa-tion by searching for voxels that were selectively acti-vated by faces compared to buildings and objectsSimilar to the control group and in agreement with pre-

vious studies (Hasson Harel et al 2003 Gauthier et al2000 Halgren et al 1999 Levy et al 2001) face-relatedactivation in CP subjects was located anterior to theretinotopic regions within the fusiform gyrus in regionscorresponding to the previously defined FFA (KanwisherMcDermott et al 1997) and in the lateral occipitalregion in the vicinity of the lateral occipital sulcus andthe inferior occipital gyrus The average face-relatedactivation maps of the 4 CP subjects and 10 controlsare shown in Figure 1C (red patches) The maps are

Figure 1 Behavioral performance and functional maps of the conventional face and object mapping experiment (A) Examples of stimuli and

experimental design Subjects viewed short epochs containing line drawings of faces buildings common objects or geometrical patterns while

maintaining fixation and performing a one-back memory task (B) Accuracy (top graph) and reaction time on the one-back memory task performedin the scanner for the CP (red) and control subjects (light blue) error bars indicate standard error of the mean (SEM ) across subjects

in each group The graphs also show the behavioral data for individual CP subjects (C) Averaged face (red) and building (green) activation maps

of the CP subjects and control subjects projected on an inf lated brain representation shown from a ventral view The activation is projected on

the same brain and is shown in the same statistical threshold (multisubjects GLM statistics faces p lt 005 buildings p lt 05 random effects)to enable direct comparison between the two groups (D) Face and building activation maps are shown for each of the CP subjects (top row) and

for four representative control subjects (bottom row) Note that all CP subjects exhibit similar activation patterns to the control groups

Abbreviations FG = fusiform gyrus CoS = collateral sulcus Ant = anterior Post = posterior

1152 Journal of Cognitive Neuroscience Volume 17 Number 7

projected on an inflated brain of one individual subjectand shown from a ventral view To enable direct compar-isons between CP and control subjects both maps areshown using the same statistical threshold ( p lt 005 forfaces p lt 05 for buildings General Linear Model [GLM]multisubjects random-effects analysis) Note the similar-ity between the maps of the two groups in the locationand extent of the face-related activation Similar maps arealso shown in Figure 1D for each of the four CP subjectsand four representative controls Note that despite thevariability between subjects all CP subjects still clearlyexhibited face related activation similar to the controlsubjects

In order to assess further the activation in the fusiformgyrus we analyzed the spatial extent of the activation inthis region by counting the number of activated voxels inthe left and right hemispheres for each subject in eachgroup A repeated-measures ANOVA with group (CPcontrols) hemisphere (left right) as the repeated mea-sure and the number of voxels as the dependentmeasure did not reveal any significant effect ( p gt 1)thus suggesting that the spatial extent of the activa-tion in the fusiform gyrus did not differ between thetwo groups The similarity between the groups is alsoevident from the Talairach coordinates (Talairach ampTournoux 1988) of the face-related activation in thefusiform gyrus presented in Table 2 (Part 1)

Figure 1C also shows the average building-relatedactivation (building vs faces and objects) in the twogroups (green patches) Building-related activation istypically located outside the retinotopic borders in theparahippocampal gyrus and collateral sulcus medial tothe face-related activation in the FFA (Levy et al 2001Epstein amp Kanwisher 1998) and thus serves as animportant indicator for the reliability of the anatomicallocation of the face-related activation The control groupshows the same loci of activation as depicted in pre-vious studies as do the CP subjects Again this samepattern is also evident for each individual CP subject asshown in Figure 1D

To assess the selectivity for the different object cate-gories within the face-related regions the activationprofile from the fusiform gyrus and lateral occipitalregion was extracted for each subject using the lsquolsquointernallocalizerrsquorsquo approach (see Methods) Figure 2A showsthe percent signal change averaged across all the repe-titions of each condition and collapsed across bothhemispheres for the face-related regions for both CP(left column) and control subjects (right column) (seeTable 2 Part 2 for the left and right activation foci foreach group) A repeated-measures ANOVA was per-formed separately for the activation in the fusiformregion and lateral occipital region with group (CPcontrols) stimulus type (faces buildings objects pat-terns) and percent signal change as the dependentmeasure Only a main effect of stimulus type ( p lt0001) was found in both face-related regions Note that

the activation profiles were obtained using the internallocalizer approach and thus were not biased in any wayconfirming the clear face selectivity There was no maineffect of group nor a significant interaction betweengroup and stimulus type however the interaction effectin the fusiform gyrus was close to significant ( p lt 06activation for faces was slightly increased for CP com-pared with controls whereas activation for patterns wasdecreased) Note that although not statistically signifi-cant the face selectivity in the lateral occipital focus wassomewhat reduced in the CP subjects compared to theircontrols (percent signal change for faces compared tobuildings and patterns)

Face-related regions are part of a large distributedoccipito-temporal network of object representation(Hasson Harel et al 2003 Avidan Hasson HendlerZohary amp Malach 2002 Haxby Gobbini et al 2001) inwhich multiple regions even if not selective for facesstill carry substantial information about faces (HaxbyGobbini et al 2001) It is therefore possible that thesource of the behavioral deficit in CP might arise from adysfunction in such regions To explore this possibilitywe assessed the stimulus selectivity in regions that werepreferentially activated by the nonface stimuli used inthe experiment by applying the same procedure de-scribed above for faces (Figure 2B and C) This analysiswas first performed for the building-related activation inthe collateral sulcus and importantly the activationprofile was only sampled from the anterior nonretino-topic part of this region Three CP and all controlsubjects exhibited building-related activation bilaterallyand one CP subject (TM) showed right-lateralized acti-vation As in the face-related regions only a main effectof stimulus type ( p lt 0001) was observed with no maineffect of group nor an interaction between these factors

Using the same procedure we delineated foci withobject-specific activation (compared with faces andbuildings) in the lateral occipital sulcus the medial bankof the fusiform gyrus and the inferior temporal sulcus(ITS) However these activations were less consistentacross both groups and were mostly found in the lefthemisphere The most consistent activation was foundin the ITS in both groups (Figure 2C) Two CP subjectsexhibited bilateral activation and two others exhibitedleft-lateralized activation In the control group six sub-jects exhibited left-lateralized activation and two ex-hibited right-lateralized activation whereas two othersexhibited bilateral activation Again only the main effectof condition was significant ( p lt 001) the group effectwas close to significant ( p lt 06 higher percent signalchange in the control group for all conditions comparedto CP) and the interaction between the factors was notsignificant

The absence of group differences cannot be attribut-able to reduced statistical power although the patternsof activation were largely similar in occipito-temporalregions across the groups clear differences emerged in

Avidan et al 1153

prefrontal regions As is evident from Figure 3 the CPgroup exhibited strong face-related activation (faces vsbuilding and objects p lt 005 GLM multisubjectsrandom-effects analysis) in the precentral sulcus theinferior frontal sulcus and the anterior part of the lateral

sulcus in both hemispheres The control group ex-hibited some activation in the precentral sulcus but toa much lesser extent and only in the right hemisphere(Figure 3) Face-selective activation in prefrontal regionshas been previously reported when normal subjects

Figure 2 Activation profiles

in ROIs in the conventional

face and object mapping

experiment (A) Averagedactivation profiles of

face-related voxels in the

fusiform gyrus and lateraloccipital region of CP and

control subjects Activation

profiles were obtained

using the lsquolsquointernal localizerrsquorsquoapproach (see Methods)

and the graphs show the

averaged percent signal

change for each experimentalcondition (faces buildings

objects patterns) Error

bars indicate SEM acrosssubjects in each group (B)

Averaged activation profiles

in building-related voxels

in the collateral sulcus (C)Averaged activation profiles

in object-related voxels in

the inferior temporal sulcus

1154 Journal of Cognitive Neuroscience Volume 17 Number 7

perform working memory tasks using face stimuli(Druzgal amp DrsquoEsposito 2003 Sala Rama amp Courtney2003 Haxby Petit Ungerleider amp Courtney 2000) Theprefrontal face-related activation in the current studymight then reflect the increased difficulty for the CPsubjects in the one-back memory task for faces relativeto the other conditions and to the control subjects andthe consequent recruitment of prefrontal regions tomaintain face representations (see Figure 1B)

The results from the conventional face and objectmapping experiment indicate that individuals with CP re-veal normal face-related activation in the ventral occipito-temporal cortex and close-to-normal selectivity in thelateral occipital region in terms of the anatomical loca-tion of face-selective activation and its spatial extentthe signal strength and the stimulus selective responseCrucially this normal profile of activation is apparentdespite the significant face processing impairment ex-hibited at the very same time Of note also is that theCP subjects also exhibited normal object-related activa-tion (common objects and buildings) in the occipito-temporal cortex That group differences between theCP subjects and their controls can be detected (in pre-frontal regions) attests to the sufficiency of the statisticalpower of the current paradigm to detect differenceswhen they exist and therefore lends further weightto the conclusion that there are no detectable groupdifferences in the posterior areas of cortex

These results reveal a dissociation of a BOLDndashbehav-ior correlation in the ventral occipito-temporal cortexand particularly in the FFA To further confirm thisresult we utilized additional paradigms that could as-

sist in exploring the mechanisms giving rise to the ab-normal behavior

Motion Pictures Experiment

Although serving as a powerful tool for studying objectselectivity in the human visual cortex the conventionalface and object mapping experiment exploits viewingconditions and stimuli that are rather contrived Fur-thermore as evident from Figure 1B the one-backmemory task performed during this experiment wassignificantly more difficult for the CP subjects com-pared to their controls particularly for the face stimuliIt has previously been shown that activity in the FFAcan increase as a function of working memory load intasks involving faces (Druzgal amp DrsquoEsposito 2001)Thus the seemingly normal activation found in theconventional mapping experiment in the CP subjectscould actually reflect additional computation takingplace due to the relative difficulty of the behavioraltask for these individuals

To address these issues in this next experiment weused a motion picture paradigm first to test whetherthe face and object-related activation similarities ob-served in the conventional mapping experiment arereplicable with more natural stimuli and under morenatural viewing conditions and second to determinewhether normal face-related activation can be foundregardless of the behavioral task performed in thescanner In this experiment subjects freely viewed asequence of short video clips each containing stimuli

Figure 3 Activation maps in

prefrontal regions in the

conventional face and object

mapping experiment Sameactivation maps as in

Figure 1C but here the brain is

shown in a lateral view in orderto reveal the activation in

prefrontal regions CP subjects

are shown in the upper panel

and controls in the lower oneAbbreviations LS = lateral

sulcus CS = central sulcus

PreCS = precentral sulcus

IFS = inferior frontal sulcus

Avidan et al 1155

of faces and people buildings navigation or miscella-neous objects (Hasson Nir Levy Fuhrmann amp Malach2004) Importantly no specific task was required mak-ing this experiment more akin to naturally surveyingonersquos visual environment Figure 4A shows the activa-tion maps of a single representative CP and controlsubject projected on an inflated and unfolded brainrepresentation The white lines on the posterior partof the unfolded representation show the retinotopicborders mapped for each subject The face and build-ing activation maps obtained in this experiment largelyreplicate the results from the conventional mappingexperiment (compare Figure 4A and Figure 1C) ex-cept that the activation seems to be more widespreadin the motion pictures experiment selective activa-tion for faces was found in the fusiform gyrus andthe lateral occipital region when contrasting faces withbuildings and navigation scenes (red patches) andselective activation for buildings and navigation sceneswere found in the collateral sulcus when applyingthe opposite contrast (green patches) Interestinglyface-related activation was also found within the STSin regions typically activated in response to facialmovements (Hoffman amp Haxby 2000 Puce Allisonet al 1998) and biological motion (Puce and Perrett2003) That these regions are activated here is con-sistent with the richness of the faces and motions inthe displays

To assess the spatial extent of the face-related acti-vation obtained in this experiment in both CP andcontrols the number of face-related voxels within thelateral occipital region fusiform gyrus and STS wascounted for each subject in each group To allow di-rect comparisons all face-related foci for all subjectswere selected using the same statistical threshold ( p lt0005) (see Figure 4A for two representative subjects)A repeated-measures ANOVA with group (CPcontrols)hemisphere (left right) and face-related region as therepeated-measures within-subject factors and the num-ber of voxels as the dependent measure was usedThere were infrequent instances of control subjects forwhom not all face-related foci could be identified in theselected statistical threshold and these were excludedfrom the ANOVA Critically all face-related regions weredetected for the CP subjects using that threshold Thisanalysis revealed a significant main effect of hemisphere( p lt 001) confirming that overall there were moreface-related voxels in the right compared with the lefthemisphere In addition there was a significant maineffect of face-related region ( p lt 005) with moreactivated voxels in the lateral occipital region than inthe fusiform gyrus or STS Importantly however thisanalysis did not reveal a main effect of group or any two-or three-way interactions thus suggesting that therewere no differences in the spatial extent of the activationbetween the CP and control group in any of the face-related regions

To further quantify the similarity between the CP andcontrol groups we sampled the activation profile ofthree regions of interest (ROIs) (fusiform gyrus lateraloccipital region and collateral sulcus) defined indepen-dently for each subject on the basis of the conventionalface and object mapping experiment (see Methods)Percent signal change averaged across the entire exper-iment and collapsed across the left and right hemi-spheres for each group is shown in Figure 4B Asabove all ROIs exhibited clear stimulus selectivity butof even greater importance is the striking similaritybetween the activation profiles of the two groups (CP =black line controls = white line) This similarity is alsoevident from the high values of the correlation coeffi-cient (fusiform gyrus = 80 lateral occipital region =89 CoS = 81) calculated for each ROI between theaverage activation profile of each group Importantlyhowever as evident from the figure this similarity isnot only in the overall correlation but also in themoment-by-moment waxing and waning of the activityThese results confirm the normal profile of face- andbuilding-related activation in the CP subjects when morenatural viewing conditions and stimuli are employedMoreover because this normal activation was foundhere in the absence of any behavioral task it impliesthat the normal activation found in the conventionalmapping experiment was not confounded by the diffi-culty of the behavioral task

Adaptation Experiment

The behavioral results reveal a clear impairment inCP individuals in unfamiliar face discrimination in asimultaneous matching task (Behrmann et al 2005)and in a one-back memory (sequential discrimination)task (Figure 1B) These tasks are largely perceptualand do not require that a specific or individual facebe represented Indeed CP subjects might be evenmore impaired than already evident when more preciseface knowledge is required their own self-testimoniessuggest that lsquolsquoall faces look alikersquorsquo Using the fMR-adaptation paradigm for studying the neuronal proper-ties of higher-order visual areas at a subvoxel resolutionit has been shown that normal subjects typically exhibita reduction in BOLD signal with repeated presentationof a stimulus (Grill-Spector amp Malach 2001) To exam-ine whether the face-selective activation noted above isless sensitive to face repetition in CP than in controlsubjects we compared adaptation levels for face repeti-tion across the two groups We also examined the adap-tation effect for nonface stimuli to determine whetherthe entire occipito-temporal region is normally sensitiveto repetition (Avidan et al 2002)

Stimuli were presented in epochs that included either12 different stimuli (faces building cars) or 12 repeti-tions of the same identical stimulus (see Figure 5A) Theactivation profile was sampled from ROIs defined in-

1156 Journal of Cognitive Neuroscience Volume 17 Number 7

Figure 4 Activation maps and activation profiles from the motion pictures experiment (A) Activation maps for faces (red) as well as buildings

and navigation (green) are shown for one representative CP subject and one control subject The data are shown on the inf lated and unfoldedmap representation of each individual using the same statistical threshold (GLM statistics p lt 0005) The white dotted lines on the unfolded maps

indicate the borders of the retinotopic areas that were mapped for each subject in a separate experiment Abbreviations LOS = lateral occipital

sulcus IOG = inferior occipital gyrus IPS = intraparietal sulcus PCS = postcentral sulcus Other abbreviations as in Figures 1 and 3 (B) Averaged

activation profiles of the CP (black) and control subjects (white) are shown for face-related voxels in the fusiform gyrus and lateral occipital regionand for building-related voxels in the collateral sulcus The signal is shown along the time-axis of the experiment and the vertical colored bars

indicate the different experimental conditions The correlation coefficient values between the entire activation profile of CP and control subjects

are shown for each ROI Error bars indicate SEM across subjects in each group

Avidan et al 1157

Figure 5 Experimental design and activation profiles of the adaptation experiment (A) Examples of stimuli and experimental design Face

building and car line drawings were presented in epochs containing either 12 different or 12 identical images (B) Averaged activation profiles inface-related voxels in the fusiform gyrus (top) and the lateral occipital region (bottom) for the CP and control subjects Error bars indicate SEM

across subjects in each group (C) Averaged activation profiles in the building-related voxels in the collateral sulcus

1158 Journal of Cognitive Neuroscience Volume 17 Number 7

dependently for each subject using the conventionalmapping experiment Figure 5B shows the activationlevels within the face-related foci in the fusiform gyrusand lateral occipital region averaged across all repeti-tions of each experimental condition and across thetwo hemispheres for the two groups Again despitethe fact that the face-related regions were selected usingan external localizer CP individuals exhibited clear faceselectivity in the lsquolsquodifferentrsquorsquo conditions in these ROIsNote the similarity between the groups in the magni-tude of the adaptation that is the signal decrease fol-lowing face repetition (dark gray compared with lightgray bars) suggesting that face representations in CPsubjects are indeed sensitive to face repetition Theface-related regions also exhibit a strong adaptationeffect for the nonface stimuli (buildings and cars) in-dicating that like normal subjects these regions in theCP subjects contain neurons that are highly selectivefor those stimuli (Avidan et al 2002) These observa-tions were confirmed in a repeated-measure ANOVAin which the only significant effect in both regions wasa main effect of stimulus type ( p lt 0001) There wasno main effect of group nor an interaction betweenthe factors

Figure 5C shows a similar analysis for the building-related focus in the anterior collateral sulcus This regionexhibited strong adaptation for the building stimuli aswell as for the other object categories in both groupsOnce again the ANOVA revealed only a main effect ofstimulus type ( p lt 0001) The strong building selectiv-ity in the collateral sulcus combined with the findingthat this region exhibits similar adaptation levels forall stimulus categories including faces further suggeststhat the entire occipito-temporal object representationnetwork in CP subjects is intact and mirrors that of thenormal control subjects

Rubin FacendashVase Experiment

One possible interpretation of the apparently normalface-selective BOLD response in the CP subjects foundin the previous experiments is that it reflects a responseto the presence of face parts (eyes mouth and nose) inthe absence of an intact representation of the face as awhole To examine whether the activation in CP ismerely a product of a part-representation we employeda paradigm used previously which utilizes the famousRubin facendashvase illusion to demonstrate evidence forglobal face representations in normal subjects (HassonHendler Ben Bashat amp Malach 2001) Critically in theRubin facendashvase illusion similar local elements are al-ways present However depending on figurendashgroundsegmentation and global integration the local elementscan induce two different visual percepts either of a faceor of a vase (Figure 6A) If face-selective activation in thefusiform gyrus and other face-related regions depends

only on processing of the local elements then activa-tion in these regions should be similar in both theRubin-face and Rubin-vase conditions (same local ele-ments are present) However if face representation inthe fusiform gyrus and other face-related regions isaffected by the global integration of the face elementsinto the percept of either a face or a vase then increasedactivation should be found for the Rubin-face perceptualstates compared to the Rubin-vase perceptual states In-deed in normal subjects selective activation within face-related regions has been previously demonstrated forthe Rubin-face compared to the Rubin-vase stimuli sug-gesting that normal face-related activation is modulatedby global grouping processes and is not only induced bythe representation of local face parts (Hasson Hendleret al 2001)

Here face-selective regions were independently lo-calized initially for each subject by contrasting theepochs of frontal faces with those of household ob-jects (Figure 6A) Data from one control subject wereexcluded from this experiment as they were too noisyto localize face-related activation The activation profilewas extracted for the Rubin-face and Rubin-vase stimulifrom both the fusiform gyrus and lateral occipital foci

Figure 6 The Rubin facendashvase experiment (A) Examples of the

stimuli including the Rubin facendashvase stimuli (left) and frontal faces

and household objects (localizer stimuli) used to independently

localize face-related voxels (in the experiment the uniform surfacesshown here in gray were colored in light pink) (B) Activation profiles

for the Rubin facendashvase conditions in face-related voxels in the fusiform

gyrus and lateral occipital region Error bars indicate SEM acrosssubjects in each group

Avidan et al 1159

for each subject (see Table 2 Part 2 for foci) The av-eraged percent signal change across all repetitions ofeach condition and across the left and right hemi-spheres of each subject are presented in Figure 6BAs in the control group the average fusiform gyrusactivation in the CP subjects exhibited a selective re-sponse for the Rubin-face compared to Rubin-vaseindicating that this region in CP subjects is indeedinvolved in computing global information about facesand is sensitive to the context not simply to the localcomponents of the display The similarity between theactivation profiles of the two groups in the fusiformgyrus was confirmed by a repeated-measure ANOVAwhich revealed only a main effect of stimulus type( p lt 01)

The results in the lateral occipital region were onceagain more variable across subjects in both groups (seeTable 2 Part 2) The ANOVA revealed a marginallysignificant main effect of group ( p lt 06) but not ofstimulus type and no significant interaction Note thatalthough not statistically significant the selectivity forthe Rubin-face compared to the Rubin-vase conditionin the lateral occipital focus was somewhat reduced inthe CP subjects compared to their controls This resultis analogous to the somewhat reduced face selectivityfound in the lateral occipital region in the conventionalmapping experiment (Figure 2A)

DISCUSSION

In a series of four functional imaging experimentswe have shown that individuals with CP exhibit nor-mal face- and object-related fMRI activation patternsin the ventral occipito-temporal cortex particularly inthe face-related fusiform gyrus (FFA) and largely nor-mal activations in the lateral occipital region despitetheir profound behavioral impairment Importantly thispattern was evident across different types of stimuli(eg line drawings and movie clips) and across differ-ent paradigms

In the first study in a conventional mapping experi-ment a normal pattern of BOLD activation was foundwhen subjects viewed line drawings of face and nonfacestimuli and performed a one-back memory task Thisparadigm has been used extensively in the literature andserves as a rigorous baseline to demarcate areas of ac-tivation in the ventral occipito-temporal cortex Onepossible interpretation of the normal activation is thatthe task was not sufficiently taxing for the CP individ-uals Behavioral data collected concurrently with theimage acquisition rules out this possibilitymdashalthoughthe task was easy for the controls this was not so forthe CP individuals who were clearly impaired on thistask Thus the normal face-related activation found inCP cannot merely be explained by the claim that the taskwas too simplistic for the CP subjects

A second possible interpretation and one that isdiametrically opposed to the first one offered above isthat the task was excessively difficult for the CP and itis this difficulty that drove the face-related activationparticularly in the fusiform gyrus of the CP subjects In-deed previous studies have shown that activation in theFFA increases as a function of working memory load(Druzgal amp DrsquoEsposito 2001) This explanation is nottenable either as the CP subjects continued to exhibitnormal face-related activation even when they were notrequired to perform any behavioral task during themotion pictures experiment Further research is clearlyneeded to unequivocally determine the contribution oftask difficulty to face-related activation in CP individ-uals We are now starting to parametrically explore theeffects of task difficulty in these CP individuals usingnew paradigms The critical point at present is thatnormal face-related activation is obtained at the verysame time that the behavioral impairment is manifest

The finding that face- and object-related activation inthe ventral visual cortex is apparently normal is replicat-ed in the motion picture experiment As mentionedabove this replication attests to the generalizability ofthe finding from the first conventional mapping exper-iment to more natural stimuli and viewing conditions

Importantly the normal signature of face-related ac-tivation in CP was also found when additional experi-mental manipulations were used CP subjects showednormal recovery from adaptation when different faceswere shown suggesting that neurons within these re-gions were sufficiently sensitive to differentiate betweenthe face exemplars used in this experiment Althoughone might think that normal adaptation is not thatsurprising given that the face stimuli used in this exper-iment were very different from each other it is impor-tant to bear in mind that CP subjects were actuallyimpaired in discriminating between such stimulimdashinthe conventional mapping experiment similar stimuliwere used and the CP subjects were impaired suggest-ing that the stimuli were confusing and not as percep-tually dissimilar for them as one might assume Of noteis that in addition to the adaptation for faces found inthe present experiment CP subjects also exhibitednormal adaptation levels in the fusiform gyrus and otherface-related regions for nonface stimuli suggestingthat similar to control subjects and replicating previousfindings these regions are involved in processing otherobject categories in addition to faces (Avidan et al 2002Haxby Gobbini et al 2001)

Although normal adaptation for faces is apparentdetermining its origin is less obvious Because the exactsame picture of an individual face was repeated duringthe adaptation condition we cannot determine at thisstage whether the adaptation found in the fusiformgyrus was due to the repetition of the exact same shape(ie based on perceptual geometry or structure) or therepetition of the same facial identity Differentiating

1160 Journal of Cognitive Neuroscience Volume 17 Number 7

between these alternatives and establishing whetherneural activity in the FFA is sufficient for normal rep-resentation of fine distinctions among individual faceswill require further experiments in which more subtlevariations will be made to face stimuli The use of anevent-related adaptation paradigm in such future ex-periments will also minimize possible attentional varia-tions that are inherent in the block-design adaptationparadigm employed here

Finally in the last experiment similar to control sub-jects once again CP subjects also exhibited a selectiveBOLD signal for Rubin-face compared to Rubin-vasestimuli in the fusiform gyrus Importantly these stimulicontain similar local elements but induce two differentvisual percepts either of a face or of a vase dependingon figurendashground segmentation and global integrationand shape processing Hence selectivity for the Rubin-face over the Rubin-vase stimuli implies that as in thecontrol group CP subjectsrsquo face-related activation ismodulated by global grouping processes and is notsimply induced by the representation of local face parts(Hasson Hendler et al 2001)

In sum the CP subjects exhibit normal face- andobject-related activation patterns in multiple regions ofthe ventral occipito-temporal cortex under various ex-perimental paradigms Of note however is that groupdifferences between CP subjects and their controls werefound in regions outside the occipito-temporal cortexindicating that the absence of group differences is notmerely a lack of statistical power Taken together thesefindings suggest that normal face-related activation inventral occipito-temporal regions as measured withfMRI is not sufficient to ensure intact face processingand does not necessarily reflect normal face represen-tation within those regions

Can the BehaviorndashBOLD Dissociationbe Reconciled

The central result is that the BOLD activation in CP inventral occipito-temporal regions is largely not differen-tiable from that of the control subjects notwithstandingtheir marked behavioral impairment How might thisdissociation be explained

It has been previously suggested that the FFA ismainly involved in face detection that is discriminatingbetween faces and stimuli from other object categories(Tong et al 2000) an ability that is largely preserved inCP The apparently normal FFA activation in the CPsubjects might then reflect the intact involvement ofthe FFA in deriving a rather coarse and rudimentarystructural representation which suffices for some tasks(ie detection) but not for more taxing tasks such asidentification Evidence compatible with this possibleinterpretation suggests that the final stage of face iden-tification is not completed at the level of high-order

visual face-related regions such as the FFA rathersuccessful recognition requires activation in more ante-rior regions of the temporal lobe where a more abstractor semantic representation of faces is derived (HaxbyHoffman et al 2000 Leveroni et al 2000 SergentOhta amp Macdonald 1992) Further support for the roleof anterior temporal regions in mediating face identifi-cation and particularly for being a site of facial memory(long-term representation) comes from lesion studiesFor example Barton and Cherkasova (2003) foundthat patients with anterior temporal lesions were themost impaired in a task that required generating long-term representations of famous faces as in mentalimagery compared to patients with more posterioroccipito-temporal lesions Similarly patients with ante-rior temporal lobectomy (following intractable epilepsy)were found to be impaired in identifying familiar faces(Glosser Salvucci amp Chiaravalloti 2003) The source ofthe behavioral problem in CP might then arise in theabnormal propagation of activation from the intact FFAto more anterior temporal regions

However the idea that the FFA only mediates facedetection is not universally accepted and the cleandivision of labor between the FFA and more anteriorregions might not be totally correct Although notdenying the contribution of more anterior regions toface processing and particularly to face identificationrecent experimental evidence suggests that the FFA isinvolved not only in face detection but also in faceidentification Thus normal subjects exhibited differen-tial responses for face detection and identification in thefusiform gyrus with the two processes resulting in eitherdifferential fMRI amplitude (Grill-Spector et al 2004) ortemporal scales (Puce Allison et al 1999 and seeSugase Yamane Ueno amp Kawano 1999 for similarfindings from single-unit recordings in monkeys) Thesedifferential responses in the fusiform gyrus may poten-tially be the outcome of feedback from more anteriorregions (as above) andor from additional computationtaking place within the fusiform gyrus itself Which ofthese is perturbed in CP remains unclear and they arenot mutually exclusive either Propagation to or fromanterior regions may be affected in CP However recallthat CP subjects exhibit a clear deficit not only in faceidentification (naming of famous faces) but also indiscrimination of unfamiliar faces (Figure 1B) (see alsoBehrmann et al 2005) This deficit in the more per-ceptual (not just memorial) aspects of face processingimplicates posterior occipito-temporal regions ratherthan exclusively the more anterior areas (Barton PressKeenan amp OrsquoConnor 2002 Wada amp Yamamoto 2001)It remains a possibility therefore that the fusiformactivation itself may be abnormal in CP Because thecurrent study employed detection and discriminationstasks the abnormality may still be uncovered undermore challenging conditions such as face identificationor other fine-grained face tasks

Avidan et al 1161

We also cannot rule out the possibility that differ-ences in the activation patterns in the fusiform gyrusandor other parts of the face processing networkbetween CP and normal control subjects might exist inthe temporal domain Such differences could underliethe behavioral dysfunction of CP subjects but may notbe evident with the temporal resolution afforded byfMRI (in the order of seconds) Some evidence sup-porting this view comes from ERP studies showing thatthe relatively early N170 potential which shows faceselectivity in normal subjects failed to show such se-lectivity in CP subjects (Kress amp Daum 2003b BentinDeouell et al 1999)

Finally it might be that the small differences inBOLD activation in CP subjects such as the slight re-duction in face selectivity in the lateral occipital region(eg Figures 2A and 6B) might be lsquolsquoamplifiedrsquorsquo whentranslated into behavioral performance If so the lateraloccipital region would serve as a critical link in the faceprocessing network Further support for the role ofthe occipital region in face processing comes fromstudies showing that acquired prosopagnosic patientsoften have lesions that involve not only the fusiformgyrus but also the lateral occipital region (Barton et al2002 Wada amp Yamamoto 2001) Moreover brain dam-age to this region but not to the FFA in some individ-uals may be associated with severe prosopagnosia(Rossion Caldara et al 2003)

As is evident there are a host of potential expla-nations to account for the dissociation between the im-paired behavioral profile and the apparently normalBOLD activation pattern We are currently exploringthese issues using new imaging experiments and hopethat these will uncover in greater detail the mecha-nisms giving rise to CP Suffice it to say that no one ex-planation captures the BOLDndashbehavior dissociationWhat is crucial however is that a simple assignmentof face processing to the FFA is no longer defensibleand a deeper examination of the computational prop-erties of the FFA and associated regions is demanded bythese findings

Face-related Activation Outside theOccipito-temporal Cortex

In the current study we acquired whole-brain functionalscans and thus were able to explore differences be-tween the CP and control groups in the responsepatterns in all regions of the brain not only in theoccipito-temporal regions A particularly novel findingwas that robust bilateral face-related activation wasevident in prefrontal regions (precentral sulcus inferiorfrontal sulcus anterior lateral sulcus) in the CP groupIn contrast the control group only exhibited right-lateralized activation in the precentral sulcus (Figure 3)Studies of normal individuals performing workingmemory tasks have revealed face-related activation in

prefrontal regions (Druzgal amp DrsquoEsposito 2003 Salaet al 2003 Haxby Petit et al 2000) and some ERPstudies have detected some inferior prefrontal cortexsites that generate small but specific face-specific re-sponses (Allison Puce Spencer amp McCarthy 1999)

The poor performance exhibited by CP subjectscompared to their controls on the one-back memorytask specifically for face stimuli is consistent with theinterpretation of this prefrontal activity being related toworking memory However given that the conven-tional face and object mapping experiment was notspecifically designed to look at the different componentsthat usually comprise a memory study (encoding mem-ory delay response) further research that specificallyand parametrically manipulates the memory load offaces and other object stimuli is required in order todetermine the role of this activation in face processing inCP subjects

This study has exploited the unique possibility ofexamining the activation of the fusiform gyrus andother cortical regions in an unusual group of individ-uals with CP Delineating the neural mechanism thatgives rise to abnormal behavior and by logical infer-ence normal behavior is a critical goal of neurosci-ence As might be predicted the correspondencesbetween the brain and behavior are not transparentand in this study are dissociated with the behavioralimpairment being evident concomitant with normalFFA activation Taken together however this studyhas indicated patterns of similarities and patterns ofpotential difference in this population of CP individ-uals relative to their normal counterparts The challengethat lies before us is to use these patterns to explainthe face processing deficit in detail

METHODS

Subjects

Four healthy CP subjects (2 men) aged between 29 and60 years with no discernable cortical lesion or anyhistory of neurological disease participated in all ex-periments (for details see Table 1) Another CP sub-

Table 1 Biographical Information about CP Subjects

Congenital Prosopagnosia (CP) Subjects

Subjectsrsquo Initials Sex Age

TM M 27

KM F 60

MT M 41

BE F 29

KM and TM are a mother and a son

1162 Journal of Cognitive Neuroscience Volume 17 Number 7

ject (NI male aged 40) was tested behaviorally (seeBehrmann et al 2005) but because we could not ac-quire his imaging data he was excluded from thepresent article Seven of the 12 control subjects whoparticipated in the behavioral studies along with threenew participants completed the imaging experimentsThe control group included at least two age- and sex-matched controls for each CP individual All CP andcontrol subjects were right-handed and had normal orcorrected-to-normal vision All subjects consented to par-ticipate in the experiments and the protocol was ap-proved by the Institutional Review Boards of CarnegieMellon University and the University of Pittsburgh

Imaging Experiments

Visual Stimulation

Visual stimuli were generated using the E-prime IFISsoftware (Psychology Software Tools Pittsburgh PAUSA) and projected via LCD to a screen located in theback of the scanner bore behind the subjectrsquos head Sub-jects viewed the stimuli through a tilted mirror mountedabove their eyes on the head coil Before the scan sub-

jects were familiarized with the visual stimuli and tasksof each of the imaging experiments

MRI Setup

All subjects were scanned in a 3-T Siemens Allegrascanner equipped with a standard head coil Subjectsparticipated in one scanning session that lasted 15 to2 hr in which all the functional and anatomical scanswere acquired The order of the functional experimentswithin the scanning session was generally constant forall subjects BOLD contrast was acquired using gradient-echo echo-planar imaging sequence Specific parame-ters TR = 3000 msec TE = 35 msec flip angle = 908FOV = 210 210 mm2 matrix size 64 64 35 axialslices 3 mm thickness no gap High-resolution anatom-ical scans (T1-weighted 3-D MPRAGE) were also acquiredto allow accurate cortical segmentation reconstructionand volume-based statistical analysis Specific parame-ters TE = 349 flip angle = 88 FOV = 256 256 mm2matrix size = 256 256 slice thickness = 1 mm numberof slices = 160ndash192 orientation of slices was eitherhorizontal or sagittal For three of the CP subjects the

Table 2 Laterality in force-selective activation

1 Talairach Coordinates Fusiform Gyrus

Left Hemisphere Right Hemisphere

x y z x y z

CP subjects 40 plusmn 2 53 plusmn 3 19 plusmn 2 38 plusmn 5 47 plusmn 8 17 plusmn 3

Control subjects 37 plusmn 4 41 plusmn 5 18 plusmn 4 36 plusmn 2 40 plusmn 7 17 plusmn 3

2 Activation Foci

Localizer Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 4 ndash ndash

Control subjects 9 1 ndash 4 2 1

Rubin FacendashVase Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 1 3 ndash

Control subjects 9 ndash ndash 3 2 4

Part 1 Talairach coordinates of the face-related activation in the fusiform gyrus of CP and control subjects extracted from the conventionalface and object mapping experiment Part 2 Number of CP and control subjects that showed bilateral right-lateralized or left-lateralizedface-related activation in the fusiform gyrus and in the lateral occipital region in the conventional face and object mapping experiment andin the Rubin facendashvase experiment

Avidan et al 1163

3-D brain reconstruction was performed on imagespreviously acquired on a 15-T GE scanner

Experiments

Conventional Face and Object Mapping Experiment

Stimuli included line drawings of faces buildings com-mon objects and geometric patterns presented in ashort block design (for details see Levy et al 2001Figure 1A) Each epoch lasted 9 sec and contained ninestimuli from the same category there were eight differ-ent images and one immediate repetition of one of thestimuli Each stimulus was presented for 800 msecfollowed by a 200-msec interval of blank screen Therewere seven repetitions of each condition and their orderwas pseudorandomized across subjects Visual epochswere interleaved with 6-sec blank epochs The experi-ment lasted 450 sec and started and ended with extend-ed blank epochs of 27 and 9 sec respectively Subjectswere instructed to fixate on a red central fixation dot toperform a one-back memory task and to press a key ona response glove to indicate the repeated image Bothaccuracy and reaction time were recorded

The Motion Pictures Experiment

The experiment lasted 10 minutes and was composed of32 epochs (15 sec long) of movie clips each containingimages from one of four possible categories people invarious situations navigation of the camera through citybuildings or through open fields and miscellaneousimages of objects from different categories (machinesfalling water etc for details see Hasson Nir et al2004 Figure 4) The experiment started with a 51-secblank period followed by 9 sec of pattern stimuli andended with a 1-minute blank Subjects were simplyinstructed to watch the movie clips

The Adaptation Experiment

The experiment included line drawing images fromthree different categories faces buildings and cars(for details see Avidan et al 2002 Figure 5A) Stimuliwere presented in 12-sec epochs that contained either12 different images (lsquolsquodifferentrsquorsquo condition) from thesame category or 12 repetitions of the same identicalstimulus (lsquolsquoidenticalrsquorsquo condition) Within epochs eachstimulus was presented for 800 msec followed by200 msec of fixation only The experiment lasted480 sec and started and ended with a long blank pe-riod of 21 and 15 sec respectively The first visualblock always contained geometric patterns and was ex-cluded from the analysis Visual epochs were interleavedwith 6-sec blank epochs and the order of visual epochswas pseudorandomized across subjects Subjects wereinstructed to continuously fixate a central red dot and

to covertly categorize the images as follows man wom-an or a child for the face stimuli apartment townhouseor public facility for the buildings and private car truckor bus for the vehicles

The Rubin FacendashVase Experiment

The experiment included modified versions of the Rubinfacendashvase stimuli as well as line drawings of front facesand household objects (for details see Hasson Hendleret al 2001 Figure 6A) The Rubin facendashvase stimuli weremodified so that subjectsrsquo perception was biased towardsone perceptual state or the other This was done by firstuniformly coloring the vase area within each of thestimuli and placing stripes over the profile Stimuli werethen placed either over a striped background (biastowards perception of the vase due to closure of thevase stimulus) or over a uniform-color background (biastowards perception of the face due to closure of theface stimulus) Critically the Rubin face and vase sharethe same local contours but give rise to different globalperceptions In order to ensure that subjectsrsquo perceptionremained biased to one perceptual interpretation (ei-ther the profile or vase) during the experiment stimuliwere presented briefly (200 msec) and were immediate-ly followed by a masking grid that was presented for800 msec Stimuli were presented in 9-sec blocks thatwere interleaved with 6-sec blanks Each condition wasrepeated eight times and the order of presentation waspseudorandomized across subjects Subjects maintainedcentral fixation throughout the experiment and per-formed a one-back memory task however due to atechnical failure their responses were not recorded Theexperiment lasted 507 sec and started and ended withextended blank periods

Data Analysis

Data analysis was performed using the BrainVoyager48 software package (Brain Innovation MaastrichtNetherlands) and complementary in-house softwareDetailed description of the data analysis is provided inHasson Harel et al (2003) Briefly for each subject thecortical surface was reconstructed and unfolded intothe flattened format Functional data for each subjectfrom each experiment were analyzed separately Pre-processing of functional scans included 3-D motioncorrection and filtering of low frequencies up to fivecycles per experiment (slow drift) Statistical analysiswas based on the GLM For each experiment each ofthe conditions (except for blank) was defined as a sepa-rate predictor a boxcar shape was used to model eachpredictor and a 3-sec lag was assumed Percent signalchange for each subject in each experiment (except forthe motion pictures experiment) was calculated as thepercent activation from a blank baseline For the mo-

1164 Journal of Cognitive Neuroscience Volume 17 Number 7

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

tion of face-specific ERP sites located in the fusiformgyrus most commonly results in a transient failure toidentify rather than detect faces (Mundel et al 2003Puce Allison amp McCarthy 1999) Finally a recent fMRIstudy showed that although a major component ofthe FFA BOLD activation was correlated with successfulface detection a significant additional signal increasewas correlated with successful face identification (Grill-Spector Knouf amp Kanwisher 2004) Note however thatdirect evidence supporting the contribution of the FFAto face identification is not clear-cut either Thus al-though some studies do not find a differential responseto familiar versus unfamiliar faces (Leveroni et al 2000Nakamura et al 2000) as might be expected of an areamediating identification others do report such differ-ences (Henson Shallice amp Dolan 2000 Rossion SchiltzRobaye Pirenne amp Crommelinck 2001)

To address the ongoing controversy concerning therole of the FFA in face processing we adopted adifferent approach in which we examined the extent towhich there is face-selective activation in the FFA in agroup of unusual individuals who are impaired at facerecognition despite normal visual acuity and intelli-gence a deficit termed lsquolsquocongenital prosopagnosiarsquorsquo(CP) In contrast with acquired prosopagnosic patientsin whom the ventral occipito-temporal cortex has beendamaged CP individuals exhibit an impairment in facerecognition that is present from early childhood in theabsence of any discernible cortical lesion or neurologicaldisease As such these individuals provide us with awindow onto the function of the fusiform gyrus andenable us to examine the neural mechanisms mediatingface recognition using fMRI unaffected by damage thatmight disrupt normal blood flow or neurovasculariza-tion (DrsquoEsposito et al 2003)

To date there have been relatively few studies of CPindividuals (eg Behrmann amp Avidan 2005 Duchaineamp Nakayuma 2005 Grueter et al in press Kress amp Daum2003a Duchaine 2000 De Haan amp Campbell 1991McConachie 1976) only two of which include ERP re-cordings (Kress amp Daum 2003b Bentin Deouell ampSoroker 1999) and two of which include functional imag-ing findings (Hasson Avidan Deouell Bentin amp Malach2003 Hadjikhani amp De Gelder 2002) The present studyhowever is the first to provide both detailed and concur-rent functional imaging and behavioral findings in a groupof four CP individuals To derive brainndashbehavior correla-tions fully we first assessed the behavioral performanceof CP individuals on faces common objects and novelobjects in a variety of tasks (see Behrmann AvidanMarotta amp Kimchi 2005) Having definitively establishedthat all CP subjects were markedly impaired relative tocontrol subjects we then carried out an extensive set offour different imaging experiments two of which havebeen used previously (Hasson Avidan et al 2003) andtwo of which involve new paradigms We used whole-brain scanning to explore the differences between CP

individuals and their controls across the entire cortexand not only in occipito-temporal regions as was donepreviously Finally and critically behavioral responsescollected during the scanning session enabled us toshow directly that even when CP subjects performedpoorly in a face-discrimination task their face-relatedactivation in the FFA was not differentiable from that ofthe normal control subjects The dissociation betweenthe apparently normal FFA BOLD activation and theimpairment in face processing in CP challenges existingaccounts of face processing and forces us to reconcep-tualize the role of the FFA in the representation of faces

RESULTS

Behavioral Profile of CP Subjects

To document the extent and specificity of the behav-ioral impairment CP subjects and 12 normal controlsfirst completed a set of behavioral experiments with facesand nonface objects (for a detailed description of thedifferent experiments and their results see Behrmannet al 2005) In brief relative to their controls all CPsubjects were significantly poorer in recognizing familiarfaces and in making samedifferent discrimination judg-ments on unfamiliar faces CP subjects were also im-paired although to a lesser extent and with greatervariability on tasks involving nonface stimuli All CPsubjects performed within the normal range on vari-ous low-level visual processing tasks Taken togetherthe behavioral data establish a clear deficit in face pro-cessing for all CP subjects

Imaging Experiments

Conventional Face and Object Mapping Experiment

fMRI responses for line drawings of faces buildingsobjects and patterns were first mapped for the CPsubjects and 10 control subjects Stimuli were presentedin a short block design and subjects performed a one-back memory task while fixating on a central dot(Figure 1A) Note that this one-back task is equivalentto a sequential face-discrimination task (lsquolsquoare consecu-tive images the same or differentrsquorsquo) We also mappedout the meridian borders of visual areas for each subjectto assess the location of face- and object-related activa-tion in relation to an objective definition of the visualretinotopic regions (Levy et al 2001)

Importantly even though the one-back memory taskperformed in the scanner was relatively easy as evi-dent from the high accuracy of the control subjects(Figure 1B) the CP group exhibited a significant andspecific behavioral impairment for faces reflected bothin accuracy and RT (Figure 1B) (ANOVA for accuracysignificant stimulus type group interaction p lt 05no significant main effect of stimulus type p = 09 norof group p = 24 ANOVA for RT significant stimulus

Avidan et al 1151

type group interaction p lt 05 significant main effectof stimulus type p lt 05 no significant main effect ofgroup p = 55) As evident from the figure howeverthere was some variability within the CP group withsome subjects trading off speed for accuracy or viceversa (compare subject MT with TM)

For each subject we first localized face-related activa-tion by searching for voxels that were selectively acti-vated by faces compared to buildings and objectsSimilar to the control group and in agreement with pre-

vious studies (Hasson Harel et al 2003 Gauthier et al2000 Halgren et al 1999 Levy et al 2001) face-relatedactivation in CP subjects was located anterior to theretinotopic regions within the fusiform gyrus in regionscorresponding to the previously defined FFA (KanwisherMcDermott et al 1997) and in the lateral occipitalregion in the vicinity of the lateral occipital sulcus andthe inferior occipital gyrus The average face-relatedactivation maps of the 4 CP subjects and 10 controlsare shown in Figure 1C (red patches) The maps are

Figure 1 Behavioral performance and functional maps of the conventional face and object mapping experiment (A) Examples of stimuli and

experimental design Subjects viewed short epochs containing line drawings of faces buildings common objects or geometrical patterns while

maintaining fixation and performing a one-back memory task (B) Accuracy (top graph) and reaction time on the one-back memory task performedin the scanner for the CP (red) and control subjects (light blue) error bars indicate standard error of the mean (SEM ) across subjects

in each group The graphs also show the behavioral data for individual CP subjects (C) Averaged face (red) and building (green) activation maps

of the CP subjects and control subjects projected on an inf lated brain representation shown from a ventral view The activation is projected on

the same brain and is shown in the same statistical threshold (multisubjects GLM statistics faces p lt 005 buildings p lt 05 random effects)to enable direct comparison between the two groups (D) Face and building activation maps are shown for each of the CP subjects (top row) and

for four representative control subjects (bottom row) Note that all CP subjects exhibit similar activation patterns to the control groups

Abbreviations FG = fusiform gyrus CoS = collateral sulcus Ant = anterior Post = posterior

1152 Journal of Cognitive Neuroscience Volume 17 Number 7

projected on an inflated brain of one individual subjectand shown from a ventral view To enable direct compar-isons between CP and control subjects both maps areshown using the same statistical threshold ( p lt 005 forfaces p lt 05 for buildings General Linear Model [GLM]multisubjects random-effects analysis) Note the similar-ity between the maps of the two groups in the locationand extent of the face-related activation Similar maps arealso shown in Figure 1D for each of the four CP subjectsand four representative controls Note that despite thevariability between subjects all CP subjects still clearlyexhibited face related activation similar to the controlsubjects

In order to assess further the activation in the fusiformgyrus we analyzed the spatial extent of the activation inthis region by counting the number of activated voxels inthe left and right hemispheres for each subject in eachgroup A repeated-measures ANOVA with group (CPcontrols) hemisphere (left right) as the repeated mea-sure and the number of voxels as the dependentmeasure did not reveal any significant effect ( p gt 1)thus suggesting that the spatial extent of the activa-tion in the fusiform gyrus did not differ between thetwo groups The similarity between the groups is alsoevident from the Talairach coordinates (Talairach ampTournoux 1988) of the face-related activation in thefusiform gyrus presented in Table 2 (Part 1)

Figure 1C also shows the average building-relatedactivation (building vs faces and objects) in the twogroups (green patches) Building-related activation istypically located outside the retinotopic borders in theparahippocampal gyrus and collateral sulcus medial tothe face-related activation in the FFA (Levy et al 2001Epstein amp Kanwisher 1998) and thus serves as animportant indicator for the reliability of the anatomicallocation of the face-related activation The control groupshows the same loci of activation as depicted in pre-vious studies as do the CP subjects Again this samepattern is also evident for each individual CP subject asshown in Figure 1D

To assess the selectivity for the different object cate-gories within the face-related regions the activationprofile from the fusiform gyrus and lateral occipitalregion was extracted for each subject using the lsquolsquointernallocalizerrsquorsquo approach (see Methods) Figure 2A showsthe percent signal change averaged across all the repe-titions of each condition and collapsed across bothhemispheres for the face-related regions for both CP(left column) and control subjects (right column) (seeTable 2 Part 2 for the left and right activation foci foreach group) A repeated-measures ANOVA was per-formed separately for the activation in the fusiformregion and lateral occipital region with group (CPcontrols) stimulus type (faces buildings objects pat-terns) and percent signal change as the dependentmeasure Only a main effect of stimulus type ( p lt0001) was found in both face-related regions Note that

the activation profiles were obtained using the internallocalizer approach and thus were not biased in any wayconfirming the clear face selectivity There was no maineffect of group nor a significant interaction betweengroup and stimulus type however the interaction effectin the fusiform gyrus was close to significant ( p lt 06activation for faces was slightly increased for CP com-pared with controls whereas activation for patterns wasdecreased) Note that although not statistically signifi-cant the face selectivity in the lateral occipital focus wassomewhat reduced in the CP subjects compared to theircontrols (percent signal change for faces compared tobuildings and patterns)

Face-related regions are part of a large distributedoccipito-temporal network of object representation(Hasson Harel et al 2003 Avidan Hasson HendlerZohary amp Malach 2002 Haxby Gobbini et al 2001) inwhich multiple regions even if not selective for facesstill carry substantial information about faces (HaxbyGobbini et al 2001) It is therefore possible that thesource of the behavioral deficit in CP might arise from adysfunction in such regions To explore this possibilitywe assessed the stimulus selectivity in regions that werepreferentially activated by the nonface stimuli used inthe experiment by applying the same procedure de-scribed above for faces (Figure 2B and C) This analysiswas first performed for the building-related activation inthe collateral sulcus and importantly the activationprofile was only sampled from the anterior nonretino-topic part of this region Three CP and all controlsubjects exhibited building-related activation bilaterallyand one CP subject (TM) showed right-lateralized acti-vation As in the face-related regions only a main effectof stimulus type ( p lt 0001) was observed with no maineffect of group nor an interaction between these factors

Using the same procedure we delineated foci withobject-specific activation (compared with faces andbuildings) in the lateral occipital sulcus the medial bankof the fusiform gyrus and the inferior temporal sulcus(ITS) However these activations were less consistentacross both groups and were mostly found in the lefthemisphere The most consistent activation was foundin the ITS in both groups (Figure 2C) Two CP subjectsexhibited bilateral activation and two others exhibitedleft-lateralized activation In the control group six sub-jects exhibited left-lateralized activation and two ex-hibited right-lateralized activation whereas two othersexhibited bilateral activation Again only the main effectof condition was significant ( p lt 001) the group effectwas close to significant ( p lt 06 higher percent signalchange in the control group for all conditions comparedto CP) and the interaction between the factors was notsignificant

The absence of group differences cannot be attribut-able to reduced statistical power although the patternsof activation were largely similar in occipito-temporalregions across the groups clear differences emerged in

Avidan et al 1153

prefrontal regions As is evident from Figure 3 the CPgroup exhibited strong face-related activation (faces vsbuilding and objects p lt 005 GLM multisubjectsrandom-effects analysis) in the precentral sulcus theinferior frontal sulcus and the anterior part of the lateral

sulcus in both hemispheres The control group ex-hibited some activation in the precentral sulcus but toa much lesser extent and only in the right hemisphere(Figure 3) Face-selective activation in prefrontal regionshas been previously reported when normal subjects

Figure 2 Activation profiles

in ROIs in the conventional

face and object mapping

experiment (A) Averagedactivation profiles of

face-related voxels in the

fusiform gyrus and lateraloccipital region of CP and

control subjects Activation

profiles were obtained

using the lsquolsquointernal localizerrsquorsquoapproach (see Methods)

and the graphs show the

averaged percent signal

change for each experimentalcondition (faces buildings

objects patterns) Error

bars indicate SEM acrosssubjects in each group (B)

Averaged activation profiles

in building-related voxels

in the collateral sulcus (C)Averaged activation profiles

in object-related voxels in

the inferior temporal sulcus

1154 Journal of Cognitive Neuroscience Volume 17 Number 7

perform working memory tasks using face stimuli(Druzgal amp DrsquoEsposito 2003 Sala Rama amp Courtney2003 Haxby Petit Ungerleider amp Courtney 2000) Theprefrontal face-related activation in the current studymight then reflect the increased difficulty for the CPsubjects in the one-back memory task for faces relativeto the other conditions and to the control subjects andthe consequent recruitment of prefrontal regions tomaintain face representations (see Figure 1B)

The results from the conventional face and objectmapping experiment indicate that individuals with CP re-veal normal face-related activation in the ventral occipito-temporal cortex and close-to-normal selectivity in thelateral occipital region in terms of the anatomical loca-tion of face-selective activation and its spatial extentthe signal strength and the stimulus selective responseCrucially this normal profile of activation is apparentdespite the significant face processing impairment ex-hibited at the very same time Of note also is that theCP subjects also exhibited normal object-related activa-tion (common objects and buildings) in the occipito-temporal cortex That group differences between theCP subjects and their controls can be detected (in pre-frontal regions) attests to the sufficiency of the statisticalpower of the current paradigm to detect differenceswhen they exist and therefore lends further weightto the conclusion that there are no detectable groupdifferences in the posterior areas of cortex

These results reveal a dissociation of a BOLDndashbehav-ior correlation in the ventral occipito-temporal cortexand particularly in the FFA To further confirm thisresult we utilized additional paradigms that could as-

sist in exploring the mechanisms giving rise to the ab-normal behavior

Motion Pictures Experiment

Although serving as a powerful tool for studying objectselectivity in the human visual cortex the conventionalface and object mapping experiment exploits viewingconditions and stimuli that are rather contrived Fur-thermore as evident from Figure 1B the one-backmemory task performed during this experiment wassignificantly more difficult for the CP subjects com-pared to their controls particularly for the face stimuliIt has previously been shown that activity in the FFAcan increase as a function of working memory load intasks involving faces (Druzgal amp DrsquoEsposito 2001)Thus the seemingly normal activation found in theconventional mapping experiment in the CP subjectscould actually reflect additional computation takingplace due to the relative difficulty of the behavioraltask for these individuals

To address these issues in this next experiment weused a motion picture paradigm first to test whetherthe face and object-related activation similarities ob-served in the conventional mapping experiment arereplicable with more natural stimuli and under morenatural viewing conditions and second to determinewhether normal face-related activation can be foundregardless of the behavioral task performed in thescanner In this experiment subjects freely viewed asequence of short video clips each containing stimuli

Figure 3 Activation maps in

prefrontal regions in the

conventional face and object

mapping experiment Sameactivation maps as in

Figure 1C but here the brain is

shown in a lateral view in orderto reveal the activation in

prefrontal regions CP subjects

are shown in the upper panel

and controls in the lower oneAbbreviations LS = lateral

sulcus CS = central sulcus

PreCS = precentral sulcus

IFS = inferior frontal sulcus

Avidan et al 1155

of faces and people buildings navigation or miscella-neous objects (Hasson Nir Levy Fuhrmann amp Malach2004) Importantly no specific task was required mak-ing this experiment more akin to naturally surveyingonersquos visual environment Figure 4A shows the activa-tion maps of a single representative CP and controlsubject projected on an inflated and unfolded brainrepresentation The white lines on the posterior partof the unfolded representation show the retinotopicborders mapped for each subject The face and build-ing activation maps obtained in this experiment largelyreplicate the results from the conventional mappingexperiment (compare Figure 4A and Figure 1C) ex-cept that the activation seems to be more widespreadin the motion pictures experiment selective activa-tion for faces was found in the fusiform gyrus andthe lateral occipital region when contrasting faces withbuildings and navigation scenes (red patches) andselective activation for buildings and navigation sceneswere found in the collateral sulcus when applyingthe opposite contrast (green patches) Interestinglyface-related activation was also found within the STSin regions typically activated in response to facialmovements (Hoffman amp Haxby 2000 Puce Allisonet al 1998) and biological motion (Puce and Perrett2003) That these regions are activated here is con-sistent with the richness of the faces and motions inthe displays

To assess the spatial extent of the face-related acti-vation obtained in this experiment in both CP andcontrols the number of face-related voxels within thelateral occipital region fusiform gyrus and STS wascounted for each subject in each group To allow di-rect comparisons all face-related foci for all subjectswere selected using the same statistical threshold ( p lt0005) (see Figure 4A for two representative subjects)A repeated-measures ANOVA with group (CPcontrols)hemisphere (left right) and face-related region as therepeated-measures within-subject factors and the num-ber of voxels as the dependent measure was usedThere were infrequent instances of control subjects forwhom not all face-related foci could be identified in theselected statistical threshold and these were excludedfrom the ANOVA Critically all face-related regions weredetected for the CP subjects using that threshold Thisanalysis revealed a significant main effect of hemisphere( p lt 001) confirming that overall there were moreface-related voxels in the right compared with the lefthemisphere In addition there was a significant maineffect of face-related region ( p lt 005) with moreactivated voxels in the lateral occipital region than inthe fusiform gyrus or STS Importantly however thisanalysis did not reveal a main effect of group or any two-or three-way interactions thus suggesting that therewere no differences in the spatial extent of the activationbetween the CP and control group in any of the face-related regions

To further quantify the similarity between the CP andcontrol groups we sampled the activation profile ofthree regions of interest (ROIs) (fusiform gyrus lateraloccipital region and collateral sulcus) defined indepen-dently for each subject on the basis of the conventionalface and object mapping experiment (see Methods)Percent signal change averaged across the entire exper-iment and collapsed across the left and right hemi-spheres for each group is shown in Figure 4B Asabove all ROIs exhibited clear stimulus selectivity butof even greater importance is the striking similaritybetween the activation profiles of the two groups (CP =black line controls = white line) This similarity is alsoevident from the high values of the correlation coeffi-cient (fusiform gyrus = 80 lateral occipital region =89 CoS = 81) calculated for each ROI between theaverage activation profile of each group Importantlyhowever as evident from the figure this similarity isnot only in the overall correlation but also in themoment-by-moment waxing and waning of the activityThese results confirm the normal profile of face- andbuilding-related activation in the CP subjects when morenatural viewing conditions and stimuli are employedMoreover because this normal activation was foundhere in the absence of any behavioral task it impliesthat the normal activation found in the conventionalmapping experiment was not confounded by the diffi-culty of the behavioral task

Adaptation Experiment

The behavioral results reveal a clear impairment inCP individuals in unfamiliar face discrimination in asimultaneous matching task (Behrmann et al 2005)and in a one-back memory (sequential discrimination)task (Figure 1B) These tasks are largely perceptualand do not require that a specific or individual facebe represented Indeed CP subjects might be evenmore impaired than already evident when more preciseface knowledge is required their own self-testimoniessuggest that lsquolsquoall faces look alikersquorsquo Using the fMR-adaptation paradigm for studying the neuronal proper-ties of higher-order visual areas at a subvoxel resolutionit has been shown that normal subjects typically exhibita reduction in BOLD signal with repeated presentationof a stimulus (Grill-Spector amp Malach 2001) To exam-ine whether the face-selective activation noted above isless sensitive to face repetition in CP than in controlsubjects we compared adaptation levels for face repeti-tion across the two groups We also examined the adap-tation effect for nonface stimuli to determine whetherthe entire occipito-temporal region is normally sensitiveto repetition (Avidan et al 2002)

Stimuli were presented in epochs that included either12 different stimuli (faces building cars) or 12 repeti-tions of the same identical stimulus (see Figure 5A) Theactivation profile was sampled from ROIs defined in-

1156 Journal of Cognitive Neuroscience Volume 17 Number 7

Figure 4 Activation maps and activation profiles from the motion pictures experiment (A) Activation maps for faces (red) as well as buildings

and navigation (green) are shown for one representative CP subject and one control subject The data are shown on the inf lated and unfoldedmap representation of each individual using the same statistical threshold (GLM statistics p lt 0005) The white dotted lines on the unfolded maps

indicate the borders of the retinotopic areas that were mapped for each subject in a separate experiment Abbreviations LOS = lateral occipital

sulcus IOG = inferior occipital gyrus IPS = intraparietal sulcus PCS = postcentral sulcus Other abbreviations as in Figures 1 and 3 (B) Averaged

activation profiles of the CP (black) and control subjects (white) are shown for face-related voxels in the fusiform gyrus and lateral occipital regionand for building-related voxels in the collateral sulcus The signal is shown along the time-axis of the experiment and the vertical colored bars

indicate the different experimental conditions The correlation coefficient values between the entire activation profile of CP and control subjects

are shown for each ROI Error bars indicate SEM across subjects in each group

Avidan et al 1157

Figure 5 Experimental design and activation profiles of the adaptation experiment (A) Examples of stimuli and experimental design Face

building and car line drawings were presented in epochs containing either 12 different or 12 identical images (B) Averaged activation profiles inface-related voxels in the fusiform gyrus (top) and the lateral occipital region (bottom) for the CP and control subjects Error bars indicate SEM

across subjects in each group (C) Averaged activation profiles in the building-related voxels in the collateral sulcus

1158 Journal of Cognitive Neuroscience Volume 17 Number 7

dependently for each subject using the conventionalmapping experiment Figure 5B shows the activationlevels within the face-related foci in the fusiform gyrusand lateral occipital region averaged across all repeti-tions of each experimental condition and across thetwo hemispheres for the two groups Again despitethe fact that the face-related regions were selected usingan external localizer CP individuals exhibited clear faceselectivity in the lsquolsquodifferentrsquorsquo conditions in these ROIsNote the similarity between the groups in the magni-tude of the adaptation that is the signal decrease fol-lowing face repetition (dark gray compared with lightgray bars) suggesting that face representations in CPsubjects are indeed sensitive to face repetition Theface-related regions also exhibit a strong adaptationeffect for the nonface stimuli (buildings and cars) in-dicating that like normal subjects these regions in theCP subjects contain neurons that are highly selectivefor those stimuli (Avidan et al 2002) These observa-tions were confirmed in a repeated-measure ANOVAin which the only significant effect in both regions wasa main effect of stimulus type ( p lt 0001) There wasno main effect of group nor an interaction betweenthe factors

Figure 5C shows a similar analysis for the building-related focus in the anterior collateral sulcus This regionexhibited strong adaptation for the building stimuli aswell as for the other object categories in both groupsOnce again the ANOVA revealed only a main effect ofstimulus type ( p lt 0001) The strong building selectiv-ity in the collateral sulcus combined with the findingthat this region exhibits similar adaptation levels forall stimulus categories including faces further suggeststhat the entire occipito-temporal object representationnetwork in CP subjects is intact and mirrors that of thenormal control subjects

Rubin FacendashVase Experiment

One possible interpretation of the apparently normalface-selective BOLD response in the CP subjects foundin the previous experiments is that it reflects a responseto the presence of face parts (eyes mouth and nose) inthe absence of an intact representation of the face as awhole To examine whether the activation in CP ismerely a product of a part-representation we employeda paradigm used previously which utilizes the famousRubin facendashvase illusion to demonstrate evidence forglobal face representations in normal subjects (HassonHendler Ben Bashat amp Malach 2001) Critically in theRubin facendashvase illusion similar local elements are al-ways present However depending on figurendashgroundsegmentation and global integration the local elementscan induce two different visual percepts either of a faceor of a vase (Figure 6A) If face-selective activation in thefusiform gyrus and other face-related regions depends

only on processing of the local elements then activa-tion in these regions should be similar in both theRubin-face and Rubin-vase conditions (same local ele-ments are present) However if face representation inthe fusiform gyrus and other face-related regions isaffected by the global integration of the face elementsinto the percept of either a face or a vase then increasedactivation should be found for the Rubin-face perceptualstates compared to the Rubin-vase perceptual states In-deed in normal subjects selective activation within face-related regions has been previously demonstrated forthe Rubin-face compared to the Rubin-vase stimuli sug-gesting that normal face-related activation is modulatedby global grouping processes and is not only induced bythe representation of local face parts (Hasson Hendleret al 2001)

Here face-selective regions were independently lo-calized initially for each subject by contrasting theepochs of frontal faces with those of household ob-jects (Figure 6A) Data from one control subject wereexcluded from this experiment as they were too noisyto localize face-related activation The activation profilewas extracted for the Rubin-face and Rubin-vase stimulifrom both the fusiform gyrus and lateral occipital foci

Figure 6 The Rubin facendashvase experiment (A) Examples of the

stimuli including the Rubin facendashvase stimuli (left) and frontal faces

and household objects (localizer stimuli) used to independently

localize face-related voxels (in the experiment the uniform surfacesshown here in gray were colored in light pink) (B) Activation profiles

for the Rubin facendashvase conditions in face-related voxels in the fusiform

gyrus and lateral occipital region Error bars indicate SEM acrosssubjects in each group

Avidan et al 1159

for each subject (see Table 2 Part 2 for foci) The av-eraged percent signal change across all repetitions ofeach condition and across the left and right hemi-spheres of each subject are presented in Figure 6BAs in the control group the average fusiform gyrusactivation in the CP subjects exhibited a selective re-sponse for the Rubin-face compared to Rubin-vaseindicating that this region in CP subjects is indeedinvolved in computing global information about facesand is sensitive to the context not simply to the localcomponents of the display The similarity between theactivation profiles of the two groups in the fusiformgyrus was confirmed by a repeated-measure ANOVAwhich revealed only a main effect of stimulus type( p lt 01)

The results in the lateral occipital region were onceagain more variable across subjects in both groups (seeTable 2 Part 2) The ANOVA revealed a marginallysignificant main effect of group ( p lt 06) but not ofstimulus type and no significant interaction Note thatalthough not statistically significant the selectivity forthe Rubin-face compared to the Rubin-vase conditionin the lateral occipital focus was somewhat reduced inthe CP subjects compared to their controls This resultis analogous to the somewhat reduced face selectivityfound in the lateral occipital region in the conventionalmapping experiment (Figure 2A)

DISCUSSION

In a series of four functional imaging experimentswe have shown that individuals with CP exhibit nor-mal face- and object-related fMRI activation patternsin the ventral occipito-temporal cortex particularly inthe face-related fusiform gyrus (FFA) and largely nor-mal activations in the lateral occipital region despitetheir profound behavioral impairment Importantly thispattern was evident across different types of stimuli(eg line drawings and movie clips) and across differ-ent paradigms

In the first study in a conventional mapping experi-ment a normal pattern of BOLD activation was foundwhen subjects viewed line drawings of face and nonfacestimuli and performed a one-back memory task Thisparadigm has been used extensively in the literature andserves as a rigorous baseline to demarcate areas of ac-tivation in the ventral occipito-temporal cortex Onepossible interpretation of the normal activation is thatthe task was not sufficiently taxing for the CP individ-uals Behavioral data collected concurrently with theimage acquisition rules out this possibilitymdashalthoughthe task was easy for the controls this was not so forthe CP individuals who were clearly impaired on thistask Thus the normal face-related activation found inCP cannot merely be explained by the claim that the taskwas too simplistic for the CP subjects

A second possible interpretation and one that isdiametrically opposed to the first one offered above isthat the task was excessively difficult for the CP and itis this difficulty that drove the face-related activationparticularly in the fusiform gyrus of the CP subjects In-deed previous studies have shown that activation in theFFA increases as a function of working memory load(Druzgal amp DrsquoEsposito 2001) This explanation is nottenable either as the CP subjects continued to exhibitnormal face-related activation even when they were notrequired to perform any behavioral task during themotion pictures experiment Further research is clearlyneeded to unequivocally determine the contribution oftask difficulty to face-related activation in CP individ-uals We are now starting to parametrically explore theeffects of task difficulty in these CP individuals usingnew paradigms The critical point at present is thatnormal face-related activation is obtained at the verysame time that the behavioral impairment is manifest

The finding that face- and object-related activation inthe ventral visual cortex is apparently normal is replicat-ed in the motion picture experiment As mentionedabove this replication attests to the generalizability ofthe finding from the first conventional mapping exper-iment to more natural stimuli and viewing conditions

Importantly the normal signature of face-related ac-tivation in CP was also found when additional experi-mental manipulations were used CP subjects showednormal recovery from adaptation when different faceswere shown suggesting that neurons within these re-gions were sufficiently sensitive to differentiate betweenthe face exemplars used in this experiment Althoughone might think that normal adaptation is not thatsurprising given that the face stimuli used in this exper-iment were very different from each other it is impor-tant to bear in mind that CP subjects were actuallyimpaired in discriminating between such stimulimdashinthe conventional mapping experiment similar stimuliwere used and the CP subjects were impaired suggest-ing that the stimuli were confusing and not as percep-tually dissimilar for them as one might assume Of noteis that in addition to the adaptation for faces found inthe present experiment CP subjects also exhibitednormal adaptation levels in the fusiform gyrus and otherface-related regions for nonface stimuli suggestingthat similar to control subjects and replicating previousfindings these regions are involved in processing otherobject categories in addition to faces (Avidan et al 2002Haxby Gobbini et al 2001)

Although normal adaptation for faces is apparentdetermining its origin is less obvious Because the exactsame picture of an individual face was repeated duringthe adaptation condition we cannot determine at thisstage whether the adaptation found in the fusiformgyrus was due to the repetition of the exact same shape(ie based on perceptual geometry or structure) or therepetition of the same facial identity Differentiating

1160 Journal of Cognitive Neuroscience Volume 17 Number 7

between these alternatives and establishing whetherneural activity in the FFA is sufficient for normal rep-resentation of fine distinctions among individual faceswill require further experiments in which more subtlevariations will be made to face stimuli The use of anevent-related adaptation paradigm in such future ex-periments will also minimize possible attentional varia-tions that are inherent in the block-design adaptationparadigm employed here

Finally in the last experiment similar to control sub-jects once again CP subjects also exhibited a selectiveBOLD signal for Rubin-face compared to Rubin-vasestimuli in the fusiform gyrus Importantly these stimulicontain similar local elements but induce two differentvisual percepts either of a face or of a vase dependingon figurendashground segmentation and global integrationand shape processing Hence selectivity for the Rubin-face over the Rubin-vase stimuli implies that as in thecontrol group CP subjectsrsquo face-related activation ismodulated by global grouping processes and is notsimply induced by the representation of local face parts(Hasson Hendler et al 2001)

In sum the CP subjects exhibit normal face- andobject-related activation patterns in multiple regions ofthe ventral occipito-temporal cortex under various ex-perimental paradigms Of note however is that groupdifferences between CP subjects and their controls werefound in regions outside the occipito-temporal cortexindicating that the absence of group differences is notmerely a lack of statistical power Taken together thesefindings suggest that normal face-related activation inventral occipito-temporal regions as measured withfMRI is not sufficient to ensure intact face processingand does not necessarily reflect normal face represen-tation within those regions

Can the BehaviorndashBOLD Dissociationbe Reconciled

The central result is that the BOLD activation in CP inventral occipito-temporal regions is largely not differen-tiable from that of the control subjects notwithstandingtheir marked behavioral impairment How might thisdissociation be explained

It has been previously suggested that the FFA ismainly involved in face detection that is discriminatingbetween faces and stimuli from other object categories(Tong et al 2000) an ability that is largely preserved inCP The apparently normal FFA activation in the CPsubjects might then reflect the intact involvement ofthe FFA in deriving a rather coarse and rudimentarystructural representation which suffices for some tasks(ie detection) but not for more taxing tasks such asidentification Evidence compatible with this possibleinterpretation suggests that the final stage of face iden-tification is not completed at the level of high-order

visual face-related regions such as the FFA rathersuccessful recognition requires activation in more ante-rior regions of the temporal lobe where a more abstractor semantic representation of faces is derived (HaxbyHoffman et al 2000 Leveroni et al 2000 SergentOhta amp Macdonald 1992) Further support for the roleof anterior temporal regions in mediating face identifi-cation and particularly for being a site of facial memory(long-term representation) comes from lesion studiesFor example Barton and Cherkasova (2003) foundthat patients with anterior temporal lesions were themost impaired in a task that required generating long-term representations of famous faces as in mentalimagery compared to patients with more posterioroccipito-temporal lesions Similarly patients with ante-rior temporal lobectomy (following intractable epilepsy)were found to be impaired in identifying familiar faces(Glosser Salvucci amp Chiaravalloti 2003) The source ofthe behavioral problem in CP might then arise in theabnormal propagation of activation from the intact FFAto more anterior temporal regions

However the idea that the FFA only mediates facedetection is not universally accepted and the cleandivision of labor between the FFA and more anteriorregions might not be totally correct Although notdenying the contribution of more anterior regions toface processing and particularly to face identificationrecent experimental evidence suggests that the FFA isinvolved not only in face detection but also in faceidentification Thus normal subjects exhibited differen-tial responses for face detection and identification in thefusiform gyrus with the two processes resulting in eitherdifferential fMRI amplitude (Grill-Spector et al 2004) ortemporal scales (Puce Allison et al 1999 and seeSugase Yamane Ueno amp Kawano 1999 for similarfindings from single-unit recordings in monkeys) Thesedifferential responses in the fusiform gyrus may poten-tially be the outcome of feedback from more anteriorregions (as above) andor from additional computationtaking place within the fusiform gyrus itself Which ofthese is perturbed in CP remains unclear and they arenot mutually exclusive either Propagation to or fromanterior regions may be affected in CP However recallthat CP subjects exhibit a clear deficit not only in faceidentification (naming of famous faces) but also indiscrimination of unfamiliar faces (Figure 1B) (see alsoBehrmann et al 2005) This deficit in the more per-ceptual (not just memorial) aspects of face processingimplicates posterior occipito-temporal regions ratherthan exclusively the more anterior areas (Barton PressKeenan amp OrsquoConnor 2002 Wada amp Yamamoto 2001)It remains a possibility therefore that the fusiformactivation itself may be abnormal in CP Because thecurrent study employed detection and discriminationstasks the abnormality may still be uncovered undermore challenging conditions such as face identificationor other fine-grained face tasks

Avidan et al 1161

We also cannot rule out the possibility that differ-ences in the activation patterns in the fusiform gyrusandor other parts of the face processing networkbetween CP and normal control subjects might exist inthe temporal domain Such differences could underliethe behavioral dysfunction of CP subjects but may notbe evident with the temporal resolution afforded byfMRI (in the order of seconds) Some evidence sup-porting this view comes from ERP studies showing thatthe relatively early N170 potential which shows faceselectivity in normal subjects failed to show such se-lectivity in CP subjects (Kress amp Daum 2003b BentinDeouell et al 1999)

Finally it might be that the small differences inBOLD activation in CP subjects such as the slight re-duction in face selectivity in the lateral occipital region(eg Figures 2A and 6B) might be lsquolsquoamplifiedrsquorsquo whentranslated into behavioral performance If so the lateraloccipital region would serve as a critical link in the faceprocessing network Further support for the role ofthe occipital region in face processing comes fromstudies showing that acquired prosopagnosic patientsoften have lesions that involve not only the fusiformgyrus but also the lateral occipital region (Barton et al2002 Wada amp Yamamoto 2001) Moreover brain dam-age to this region but not to the FFA in some individ-uals may be associated with severe prosopagnosia(Rossion Caldara et al 2003)

As is evident there are a host of potential expla-nations to account for the dissociation between the im-paired behavioral profile and the apparently normalBOLD activation pattern We are currently exploringthese issues using new imaging experiments and hopethat these will uncover in greater detail the mecha-nisms giving rise to CP Suffice it to say that no one ex-planation captures the BOLDndashbehavior dissociationWhat is crucial however is that a simple assignmentof face processing to the FFA is no longer defensibleand a deeper examination of the computational prop-erties of the FFA and associated regions is demanded bythese findings

Face-related Activation Outside theOccipito-temporal Cortex

In the current study we acquired whole-brain functionalscans and thus were able to explore differences be-tween the CP and control groups in the responsepatterns in all regions of the brain not only in theoccipito-temporal regions A particularly novel findingwas that robust bilateral face-related activation wasevident in prefrontal regions (precentral sulcus inferiorfrontal sulcus anterior lateral sulcus) in the CP groupIn contrast the control group only exhibited right-lateralized activation in the precentral sulcus (Figure 3)Studies of normal individuals performing workingmemory tasks have revealed face-related activation in

prefrontal regions (Druzgal amp DrsquoEsposito 2003 Salaet al 2003 Haxby Petit et al 2000) and some ERPstudies have detected some inferior prefrontal cortexsites that generate small but specific face-specific re-sponses (Allison Puce Spencer amp McCarthy 1999)

The poor performance exhibited by CP subjectscompared to their controls on the one-back memorytask specifically for face stimuli is consistent with theinterpretation of this prefrontal activity being related toworking memory However given that the conven-tional face and object mapping experiment was notspecifically designed to look at the different componentsthat usually comprise a memory study (encoding mem-ory delay response) further research that specificallyand parametrically manipulates the memory load offaces and other object stimuli is required in order todetermine the role of this activation in face processing inCP subjects

This study has exploited the unique possibility ofexamining the activation of the fusiform gyrus andother cortical regions in an unusual group of individ-uals with CP Delineating the neural mechanism thatgives rise to abnormal behavior and by logical infer-ence normal behavior is a critical goal of neurosci-ence As might be predicted the correspondencesbetween the brain and behavior are not transparentand in this study are dissociated with the behavioralimpairment being evident concomitant with normalFFA activation Taken together however this studyhas indicated patterns of similarities and patterns ofpotential difference in this population of CP individ-uals relative to their normal counterparts The challengethat lies before us is to use these patterns to explainthe face processing deficit in detail

METHODS

Subjects

Four healthy CP subjects (2 men) aged between 29 and60 years with no discernable cortical lesion or anyhistory of neurological disease participated in all ex-periments (for details see Table 1) Another CP sub-

Table 1 Biographical Information about CP Subjects

Congenital Prosopagnosia (CP) Subjects

Subjectsrsquo Initials Sex Age

TM M 27

KM F 60

MT M 41

BE F 29

KM and TM are a mother and a son

1162 Journal of Cognitive Neuroscience Volume 17 Number 7

ject (NI male aged 40) was tested behaviorally (seeBehrmann et al 2005) but because we could not ac-quire his imaging data he was excluded from thepresent article Seven of the 12 control subjects whoparticipated in the behavioral studies along with threenew participants completed the imaging experimentsThe control group included at least two age- and sex-matched controls for each CP individual All CP andcontrol subjects were right-handed and had normal orcorrected-to-normal vision All subjects consented to par-ticipate in the experiments and the protocol was ap-proved by the Institutional Review Boards of CarnegieMellon University and the University of Pittsburgh

Imaging Experiments

Visual Stimulation

Visual stimuli were generated using the E-prime IFISsoftware (Psychology Software Tools Pittsburgh PAUSA) and projected via LCD to a screen located in theback of the scanner bore behind the subjectrsquos head Sub-jects viewed the stimuli through a tilted mirror mountedabove their eyes on the head coil Before the scan sub-

jects were familiarized with the visual stimuli and tasksof each of the imaging experiments

MRI Setup

All subjects were scanned in a 3-T Siemens Allegrascanner equipped with a standard head coil Subjectsparticipated in one scanning session that lasted 15 to2 hr in which all the functional and anatomical scanswere acquired The order of the functional experimentswithin the scanning session was generally constant forall subjects BOLD contrast was acquired using gradient-echo echo-planar imaging sequence Specific parame-ters TR = 3000 msec TE = 35 msec flip angle = 908FOV = 210 210 mm2 matrix size 64 64 35 axialslices 3 mm thickness no gap High-resolution anatom-ical scans (T1-weighted 3-D MPRAGE) were also acquiredto allow accurate cortical segmentation reconstructionand volume-based statistical analysis Specific parame-ters TE = 349 flip angle = 88 FOV = 256 256 mm2matrix size = 256 256 slice thickness = 1 mm numberof slices = 160ndash192 orientation of slices was eitherhorizontal or sagittal For three of the CP subjects the

Table 2 Laterality in force-selective activation

1 Talairach Coordinates Fusiform Gyrus

Left Hemisphere Right Hemisphere

x y z x y z

CP subjects 40 plusmn 2 53 plusmn 3 19 plusmn 2 38 plusmn 5 47 plusmn 8 17 plusmn 3

Control subjects 37 plusmn 4 41 plusmn 5 18 plusmn 4 36 plusmn 2 40 plusmn 7 17 plusmn 3

2 Activation Foci

Localizer Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 4 ndash ndash

Control subjects 9 1 ndash 4 2 1

Rubin FacendashVase Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 1 3 ndash

Control subjects 9 ndash ndash 3 2 4

Part 1 Talairach coordinates of the face-related activation in the fusiform gyrus of CP and control subjects extracted from the conventionalface and object mapping experiment Part 2 Number of CP and control subjects that showed bilateral right-lateralized or left-lateralizedface-related activation in the fusiform gyrus and in the lateral occipital region in the conventional face and object mapping experiment andin the Rubin facendashvase experiment

Avidan et al 1163

3-D brain reconstruction was performed on imagespreviously acquired on a 15-T GE scanner

Experiments

Conventional Face and Object Mapping Experiment

Stimuli included line drawings of faces buildings com-mon objects and geometric patterns presented in ashort block design (for details see Levy et al 2001Figure 1A) Each epoch lasted 9 sec and contained ninestimuli from the same category there were eight differ-ent images and one immediate repetition of one of thestimuli Each stimulus was presented for 800 msecfollowed by a 200-msec interval of blank screen Therewere seven repetitions of each condition and their orderwas pseudorandomized across subjects Visual epochswere interleaved with 6-sec blank epochs The experi-ment lasted 450 sec and started and ended with extend-ed blank epochs of 27 and 9 sec respectively Subjectswere instructed to fixate on a red central fixation dot toperform a one-back memory task and to press a key ona response glove to indicate the repeated image Bothaccuracy and reaction time were recorded

The Motion Pictures Experiment

The experiment lasted 10 minutes and was composed of32 epochs (15 sec long) of movie clips each containingimages from one of four possible categories people invarious situations navigation of the camera through citybuildings or through open fields and miscellaneousimages of objects from different categories (machinesfalling water etc for details see Hasson Nir et al2004 Figure 4) The experiment started with a 51-secblank period followed by 9 sec of pattern stimuli andended with a 1-minute blank Subjects were simplyinstructed to watch the movie clips

The Adaptation Experiment

The experiment included line drawing images fromthree different categories faces buildings and cars(for details see Avidan et al 2002 Figure 5A) Stimuliwere presented in 12-sec epochs that contained either12 different images (lsquolsquodifferentrsquorsquo condition) from thesame category or 12 repetitions of the same identicalstimulus (lsquolsquoidenticalrsquorsquo condition) Within epochs eachstimulus was presented for 800 msec followed by200 msec of fixation only The experiment lasted480 sec and started and ended with a long blank pe-riod of 21 and 15 sec respectively The first visualblock always contained geometric patterns and was ex-cluded from the analysis Visual epochs were interleavedwith 6-sec blank epochs and the order of visual epochswas pseudorandomized across subjects Subjects wereinstructed to continuously fixate a central red dot and

to covertly categorize the images as follows man wom-an or a child for the face stimuli apartment townhouseor public facility for the buildings and private car truckor bus for the vehicles

The Rubin FacendashVase Experiment

The experiment included modified versions of the Rubinfacendashvase stimuli as well as line drawings of front facesand household objects (for details see Hasson Hendleret al 2001 Figure 6A) The Rubin facendashvase stimuli weremodified so that subjectsrsquo perception was biased towardsone perceptual state or the other This was done by firstuniformly coloring the vase area within each of thestimuli and placing stripes over the profile Stimuli werethen placed either over a striped background (biastowards perception of the vase due to closure of thevase stimulus) or over a uniform-color background (biastowards perception of the face due to closure of theface stimulus) Critically the Rubin face and vase sharethe same local contours but give rise to different globalperceptions In order to ensure that subjectsrsquo perceptionremained biased to one perceptual interpretation (ei-ther the profile or vase) during the experiment stimuliwere presented briefly (200 msec) and were immediate-ly followed by a masking grid that was presented for800 msec Stimuli were presented in 9-sec blocks thatwere interleaved with 6-sec blanks Each condition wasrepeated eight times and the order of presentation waspseudorandomized across subjects Subjects maintainedcentral fixation throughout the experiment and per-formed a one-back memory task however due to atechnical failure their responses were not recorded Theexperiment lasted 507 sec and started and ended withextended blank periods

Data Analysis

Data analysis was performed using the BrainVoyager48 software package (Brain Innovation MaastrichtNetherlands) and complementary in-house softwareDetailed description of the data analysis is provided inHasson Harel et al (2003) Briefly for each subject thecortical surface was reconstructed and unfolded intothe flattened format Functional data for each subjectfrom each experiment were analyzed separately Pre-processing of functional scans included 3-D motioncorrection and filtering of low frequencies up to fivecycles per experiment (slow drift) Statistical analysiswas based on the GLM For each experiment each ofthe conditions (except for blank) was defined as a sepa-rate predictor a boxcar shape was used to model eachpredictor and a 3-sec lag was assumed Percent signalchange for each subject in each experiment (except forthe motion pictures experiment) was calculated as thepercent activation from a blank baseline For the mo-

1164 Journal of Cognitive Neuroscience Volume 17 Number 7

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

type group interaction p lt 05 significant main effectof stimulus type p lt 05 no significant main effect ofgroup p = 55) As evident from the figure howeverthere was some variability within the CP group withsome subjects trading off speed for accuracy or viceversa (compare subject MT with TM)

For each subject we first localized face-related activa-tion by searching for voxels that were selectively acti-vated by faces compared to buildings and objectsSimilar to the control group and in agreement with pre-

vious studies (Hasson Harel et al 2003 Gauthier et al2000 Halgren et al 1999 Levy et al 2001) face-relatedactivation in CP subjects was located anterior to theretinotopic regions within the fusiform gyrus in regionscorresponding to the previously defined FFA (KanwisherMcDermott et al 1997) and in the lateral occipitalregion in the vicinity of the lateral occipital sulcus andthe inferior occipital gyrus The average face-relatedactivation maps of the 4 CP subjects and 10 controlsare shown in Figure 1C (red patches) The maps are

Figure 1 Behavioral performance and functional maps of the conventional face and object mapping experiment (A) Examples of stimuli and

experimental design Subjects viewed short epochs containing line drawings of faces buildings common objects or geometrical patterns while

maintaining fixation and performing a one-back memory task (B) Accuracy (top graph) and reaction time on the one-back memory task performedin the scanner for the CP (red) and control subjects (light blue) error bars indicate standard error of the mean (SEM ) across subjects

in each group The graphs also show the behavioral data for individual CP subjects (C) Averaged face (red) and building (green) activation maps

of the CP subjects and control subjects projected on an inf lated brain representation shown from a ventral view The activation is projected on

the same brain and is shown in the same statistical threshold (multisubjects GLM statistics faces p lt 005 buildings p lt 05 random effects)to enable direct comparison between the two groups (D) Face and building activation maps are shown for each of the CP subjects (top row) and

for four representative control subjects (bottom row) Note that all CP subjects exhibit similar activation patterns to the control groups

Abbreviations FG = fusiform gyrus CoS = collateral sulcus Ant = anterior Post = posterior

1152 Journal of Cognitive Neuroscience Volume 17 Number 7

projected on an inflated brain of one individual subjectand shown from a ventral view To enable direct compar-isons between CP and control subjects both maps areshown using the same statistical threshold ( p lt 005 forfaces p lt 05 for buildings General Linear Model [GLM]multisubjects random-effects analysis) Note the similar-ity between the maps of the two groups in the locationand extent of the face-related activation Similar maps arealso shown in Figure 1D for each of the four CP subjectsand four representative controls Note that despite thevariability between subjects all CP subjects still clearlyexhibited face related activation similar to the controlsubjects

In order to assess further the activation in the fusiformgyrus we analyzed the spatial extent of the activation inthis region by counting the number of activated voxels inthe left and right hemispheres for each subject in eachgroup A repeated-measures ANOVA with group (CPcontrols) hemisphere (left right) as the repeated mea-sure and the number of voxels as the dependentmeasure did not reveal any significant effect ( p gt 1)thus suggesting that the spatial extent of the activa-tion in the fusiform gyrus did not differ between thetwo groups The similarity between the groups is alsoevident from the Talairach coordinates (Talairach ampTournoux 1988) of the face-related activation in thefusiform gyrus presented in Table 2 (Part 1)

Figure 1C also shows the average building-relatedactivation (building vs faces and objects) in the twogroups (green patches) Building-related activation istypically located outside the retinotopic borders in theparahippocampal gyrus and collateral sulcus medial tothe face-related activation in the FFA (Levy et al 2001Epstein amp Kanwisher 1998) and thus serves as animportant indicator for the reliability of the anatomicallocation of the face-related activation The control groupshows the same loci of activation as depicted in pre-vious studies as do the CP subjects Again this samepattern is also evident for each individual CP subject asshown in Figure 1D

To assess the selectivity for the different object cate-gories within the face-related regions the activationprofile from the fusiform gyrus and lateral occipitalregion was extracted for each subject using the lsquolsquointernallocalizerrsquorsquo approach (see Methods) Figure 2A showsthe percent signal change averaged across all the repe-titions of each condition and collapsed across bothhemispheres for the face-related regions for both CP(left column) and control subjects (right column) (seeTable 2 Part 2 for the left and right activation foci foreach group) A repeated-measures ANOVA was per-formed separately for the activation in the fusiformregion and lateral occipital region with group (CPcontrols) stimulus type (faces buildings objects pat-terns) and percent signal change as the dependentmeasure Only a main effect of stimulus type ( p lt0001) was found in both face-related regions Note that

the activation profiles were obtained using the internallocalizer approach and thus were not biased in any wayconfirming the clear face selectivity There was no maineffect of group nor a significant interaction betweengroup and stimulus type however the interaction effectin the fusiform gyrus was close to significant ( p lt 06activation for faces was slightly increased for CP com-pared with controls whereas activation for patterns wasdecreased) Note that although not statistically signifi-cant the face selectivity in the lateral occipital focus wassomewhat reduced in the CP subjects compared to theircontrols (percent signal change for faces compared tobuildings and patterns)

Face-related regions are part of a large distributedoccipito-temporal network of object representation(Hasson Harel et al 2003 Avidan Hasson HendlerZohary amp Malach 2002 Haxby Gobbini et al 2001) inwhich multiple regions even if not selective for facesstill carry substantial information about faces (HaxbyGobbini et al 2001) It is therefore possible that thesource of the behavioral deficit in CP might arise from adysfunction in such regions To explore this possibilitywe assessed the stimulus selectivity in regions that werepreferentially activated by the nonface stimuli used inthe experiment by applying the same procedure de-scribed above for faces (Figure 2B and C) This analysiswas first performed for the building-related activation inthe collateral sulcus and importantly the activationprofile was only sampled from the anterior nonretino-topic part of this region Three CP and all controlsubjects exhibited building-related activation bilaterallyand one CP subject (TM) showed right-lateralized acti-vation As in the face-related regions only a main effectof stimulus type ( p lt 0001) was observed with no maineffect of group nor an interaction between these factors

Using the same procedure we delineated foci withobject-specific activation (compared with faces andbuildings) in the lateral occipital sulcus the medial bankof the fusiform gyrus and the inferior temporal sulcus(ITS) However these activations were less consistentacross both groups and were mostly found in the lefthemisphere The most consistent activation was foundin the ITS in both groups (Figure 2C) Two CP subjectsexhibited bilateral activation and two others exhibitedleft-lateralized activation In the control group six sub-jects exhibited left-lateralized activation and two ex-hibited right-lateralized activation whereas two othersexhibited bilateral activation Again only the main effectof condition was significant ( p lt 001) the group effectwas close to significant ( p lt 06 higher percent signalchange in the control group for all conditions comparedto CP) and the interaction between the factors was notsignificant

The absence of group differences cannot be attribut-able to reduced statistical power although the patternsof activation were largely similar in occipito-temporalregions across the groups clear differences emerged in

Avidan et al 1153

prefrontal regions As is evident from Figure 3 the CPgroup exhibited strong face-related activation (faces vsbuilding and objects p lt 005 GLM multisubjectsrandom-effects analysis) in the precentral sulcus theinferior frontal sulcus and the anterior part of the lateral

sulcus in both hemispheres The control group ex-hibited some activation in the precentral sulcus but toa much lesser extent and only in the right hemisphere(Figure 3) Face-selective activation in prefrontal regionshas been previously reported when normal subjects

Figure 2 Activation profiles

in ROIs in the conventional

face and object mapping

experiment (A) Averagedactivation profiles of

face-related voxels in the

fusiform gyrus and lateraloccipital region of CP and

control subjects Activation

profiles were obtained

using the lsquolsquointernal localizerrsquorsquoapproach (see Methods)

and the graphs show the

averaged percent signal

change for each experimentalcondition (faces buildings

objects patterns) Error

bars indicate SEM acrosssubjects in each group (B)

Averaged activation profiles

in building-related voxels

in the collateral sulcus (C)Averaged activation profiles

in object-related voxels in

the inferior temporal sulcus

1154 Journal of Cognitive Neuroscience Volume 17 Number 7

perform working memory tasks using face stimuli(Druzgal amp DrsquoEsposito 2003 Sala Rama amp Courtney2003 Haxby Petit Ungerleider amp Courtney 2000) Theprefrontal face-related activation in the current studymight then reflect the increased difficulty for the CPsubjects in the one-back memory task for faces relativeto the other conditions and to the control subjects andthe consequent recruitment of prefrontal regions tomaintain face representations (see Figure 1B)

The results from the conventional face and objectmapping experiment indicate that individuals with CP re-veal normal face-related activation in the ventral occipito-temporal cortex and close-to-normal selectivity in thelateral occipital region in terms of the anatomical loca-tion of face-selective activation and its spatial extentthe signal strength and the stimulus selective responseCrucially this normal profile of activation is apparentdespite the significant face processing impairment ex-hibited at the very same time Of note also is that theCP subjects also exhibited normal object-related activa-tion (common objects and buildings) in the occipito-temporal cortex That group differences between theCP subjects and their controls can be detected (in pre-frontal regions) attests to the sufficiency of the statisticalpower of the current paradigm to detect differenceswhen they exist and therefore lends further weightto the conclusion that there are no detectable groupdifferences in the posterior areas of cortex

These results reveal a dissociation of a BOLDndashbehav-ior correlation in the ventral occipito-temporal cortexand particularly in the FFA To further confirm thisresult we utilized additional paradigms that could as-

sist in exploring the mechanisms giving rise to the ab-normal behavior

Motion Pictures Experiment

Although serving as a powerful tool for studying objectselectivity in the human visual cortex the conventionalface and object mapping experiment exploits viewingconditions and stimuli that are rather contrived Fur-thermore as evident from Figure 1B the one-backmemory task performed during this experiment wassignificantly more difficult for the CP subjects com-pared to their controls particularly for the face stimuliIt has previously been shown that activity in the FFAcan increase as a function of working memory load intasks involving faces (Druzgal amp DrsquoEsposito 2001)Thus the seemingly normal activation found in theconventional mapping experiment in the CP subjectscould actually reflect additional computation takingplace due to the relative difficulty of the behavioraltask for these individuals

To address these issues in this next experiment weused a motion picture paradigm first to test whetherthe face and object-related activation similarities ob-served in the conventional mapping experiment arereplicable with more natural stimuli and under morenatural viewing conditions and second to determinewhether normal face-related activation can be foundregardless of the behavioral task performed in thescanner In this experiment subjects freely viewed asequence of short video clips each containing stimuli

Figure 3 Activation maps in

prefrontal regions in the

conventional face and object

mapping experiment Sameactivation maps as in

Figure 1C but here the brain is

shown in a lateral view in orderto reveal the activation in

prefrontal regions CP subjects

are shown in the upper panel

and controls in the lower oneAbbreviations LS = lateral

sulcus CS = central sulcus

PreCS = precentral sulcus

IFS = inferior frontal sulcus

Avidan et al 1155

of faces and people buildings navigation or miscella-neous objects (Hasson Nir Levy Fuhrmann amp Malach2004) Importantly no specific task was required mak-ing this experiment more akin to naturally surveyingonersquos visual environment Figure 4A shows the activa-tion maps of a single representative CP and controlsubject projected on an inflated and unfolded brainrepresentation The white lines on the posterior partof the unfolded representation show the retinotopicborders mapped for each subject The face and build-ing activation maps obtained in this experiment largelyreplicate the results from the conventional mappingexperiment (compare Figure 4A and Figure 1C) ex-cept that the activation seems to be more widespreadin the motion pictures experiment selective activa-tion for faces was found in the fusiform gyrus andthe lateral occipital region when contrasting faces withbuildings and navigation scenes (red patches) andselective activation for buildings and navigation sceneswere found in the collateral sulcus when applyingthe opposite contrast (green patches) Interestinglyface-related activation was also found within the STSin regions typically activated in response to facialmovements (Hoffman amp Haxby 2000 Puce Allisonet al 1998) and biological motion (Puce and Perrett2003) That these regions are activated here is con-sistent with the richness of the faces and motions inthe displays

To assess the spatial extent of the face-related acti-vation obtained in this experiment in both CP andcontrols the number of face-related voxels within thelateral occipital region fusiform gyrus and STS wascounted for each subject in each group To allow di-rect comparisons all face-related foci for all subjectswere selected using the same statistical threshold ( p lt0005) (see Figure 4A for two representative subjects)A repeated-measures ANOVA with group (CPcontrols)hemisphere (left right) and face-related region as therepeated-measures within-subject factors and the num-ber of voxels as the dependent measure was usedThere were infrequent instances of control subjects forwhom not all face-related foci could be identified in theselected statistical threshold and these were excludedfrom the ANOVA Critically all face-related regions weredetected for the CP subjects using that threshold Thisanalysis revealed a significant main effect of hemisphere( p lt 001) confirming that overall there were moreface-related voxels in the right compared with the lefthemisphere In addition there was a significant maineffect of face-related region ( p lt 005) with moreactivated voxels in the lateral occipital region than inthe fusiform gyrus or STS Importantly however thisanalysis did not reveal a main effect of group or any two-or three-way interactions thus suggesting that therewere no differences in the spatial extent of the activationbetween the CP and control group in any of the face-related regions

To further quantify the similarity between the CP andcontrol groups we sampled the activation profile ofthree regions of interest (ROIs) (fusiform gyrus lateraloccipital region and collateral sulcus) defined indepen-dently for each subject on the basis of the conventionalface and object mapping experiment (see Methods)Percent signal change averaged across the entire exper-iment and collapsed across the left and right hemi-spheres for each group is shown in Figure 4B Asabove all ROIs exhibited clear stimulus selectivity butof even greater importance is the striking similaritybetween the activation profiles of the two groups (CP =black line controls = white line) This similarity is alsoevident from the high values of the correlation coeffi-cient (fusiform gyrus = 80 lateral occipital region =89 CoS = 81) calculated for each ROI between theaverage activation profile of each group Importantlyhowever as evident from the figure this similarity isnot only in the overall correlation but also in themoment-by-moment waxing and waning of the activityThese results confirm the normal profile of face- andbuilding-related activation in the CP subjects when morenatural viewing conditions and stimuli are employedMoreover because this normal activation was foundhere in the absence of any behavioral task it impliesthat the normal activation found in the conventionalmapping experiment was not confounded by the diffi-culty of the behavioral task

Adaptation Experiment

The behavioral results reveal a clear impairment inCP individuals in unfamiliar face discrimination in asimultaneous matching task (Behrmann et al 2005)and in a one-back memory (sequential discrimination)task (Figure 1B) These tasks are largely perceptualand do not require that a specific or individual facebe represented Indeed CP subjects might be evenmore impaired than already evident when more preciseface knowledge is required their own self-testimoniessuggest that lsquolsquoall faces look alikersquorsquo Using the fMR-adaptation paradigm for studying the neuronal proper-ties of higher-order visual areas at a subvoxel resolutionit has been shown that normal subjects typically exhibita reduction in BOLD signal with repeated presentationof a stimulus (Grill-Spector amp Malach 2001) To exam-ine whether the face-selective activation noted above isless sensitive to face repetition in CP than in controlsubjects we compared adaptation levels for face repeti-tion across the two groups We also examined the adap-tation effect for nonface stimuli to determine whetherthe entire occipito-temporal region is normally sensitiveto repetition (Avidan et al 2002)

Stimuli were presented in epochs that included either12 different stimuli (faces building cars) or 12 repeti-tions of the same identical stimulus (see Figure 5A) Theactivation profile was sampled from ROIs defined in-

1156 Journal of Cognitive Neuroscience Volume 17 Number 7

Figure 4 Activation maps and activation profiles from the motion pictures experiment (A) Activation maps for faces (red) as well as buildings

and navigation (green) are shown for one representative CP subject and one control subject The data are shown on the inf lated and unfoldedmap representation of each individual using the same statistical threshold (GLM statistics p lt 0005) The white dotted lines on the unfolded maps

indicate the borders of the retinotopic areas that were mapped for each subject in a separate experiment Abbreviations LOS = lateral occipital

sulcus IOG = inferior occipital gyrus IPS = intraparietal sulcus PCS = postcentral sulcus Other abbreviations as in Figures 1 and 3 (B) Averaged

activation profiles of the CP (black) and control subjects (white) are shown for face-related voxels in the fusiform gyrus and lateral occipital regionand for building-related voxels in the collateral sulcus The signal is shown along the time-axis of the experiment and the vertical colored bars

indicate the different experimental conditions The correlation coefficient values between the entire activation profile of CP and control subjects

are shown for each ROI Error bars indicate SEM across subjects in each group

Avidan et al 1157

Figure 5 Experimental design and activation profiles of the adaptation experiment (A) Examples of stimuli and experimental design Face

building and car line drawings were presented in epochs containing either 12 different or 12 identical images (B) Averaged activation profiles inface-related voxels in the fusiform gyrus (top) and the lateral occipital region (bottom) for the CP and control subjects Error bars indicate SEM

across subjects in each group (C) Averaged activation profiles in the building-related voxels in the collateral sulcus

1158 Journal of Cognitive Neuroscience Volume 17 Number 7

dependently for each subject using the conventionalmapping experiment Figure 5B shows the activationlevels within the face-related foci in the fusiform gyrusand lateral occipital region averaged across all repeti-tions of each experimental condition and across thetwo hemispheres for the two groups Again despitethe fact that the face-related regions were selected usingan external localizer CP individuals exhibited clear faceselectivity in the lsquolsquodifferentrsquorsquo conditions in these ROIsNote the similarity between the groups in the magni-tude of the adaptation that is the signal decrease fol-lowing face repetition (dark gray compared with lightgray bars) suggesting that face representations in CPsubjects are indeed sensitive to face repetition Theface-related regions also exhibit a strong adaptationeffect for the nonface stimuli (buildings and cars) in-dicating that like normal subjects these regions in theCP subjects contain neurons that are highly selectivefor those stimuli (Avidan et al 2002) These observa-tions were confirmed in a repeated-measure ANOVAin which the only significant effect in both regions wasa main effect of stimulus type ( p lt 0001) There wasno main effect of group nor an interaction betweenthe factors

Figure 5C shows a similar analysis for the building-related focus in the anterior collateral sulcus This regionexhibited strong adaptation for the building stimuli aswell as for the other object categories in both groupsOnce again the ANOVA revealed only a main effect ofstimulus type ( p lt 0001) The strong building selectiv-ity in the collateral sulcus combined with the findingthat this region exhibits similar adaptation levels forall stimulus categories including faces further suggeststhat the entire occipito-temporal object representationnetwork in CP subjects is intact and mirrors that of thenormal control subjects

Rubin FacendashVase Experiment

One possible interpretation of the apparently normalface-selective BOLD response in the CP subjects foundin the previous experiments is that it reflects a responseto the presence of face parts (eyes mouth and nose) inthe absence of an intact representation of the face as awhole To examine whether the activation in CP ismerely a product of a part-representation we employeda paradigm used previously which utilizes the famousRubin facendashvase illusion to demonstrate evidence forglobal face representations in normal subjects (HassonHendler Ben Bashat amp Malach 2001) Critically in theRubin facendashvase illusion similar local elements are al-ways present However depending on figurendashgroundsegmentation and global integration the local elementscan induce two different visual percepts either of a faceor of a vase (Figure 6A) If face-selective activation in thefusiform gyrus and other face-related regions depends

only on processing of the local elements then activa-tion in these regions should be similar in both theRubin-face and Rubin-vase conditions (same local ele-ments are present) However if face representation inthe fusiform gyrus and other face-related regions isaffected by the global integration of the face elementsinto the percept of either a face or a vase then increasedactivation should be found for the Rubin-face perceptualstates compared to the Rubin-vase perceptual states In-deed in normal subjects selective activation within face-related regions has been previously demonstrated forthe Rubin-face compared to the Rubin-vase stimuli sug-gesting that normal face-related activation is modulatedby global grouping processes and is not only induced bythe representation of local face parts (Hasson Hendleret al 2001)

Here face-selective regions were independently lo-calized initially for each subject by contrasting theepochs of frontal faces with those of household ob-jects (Figure 6A) Data from one control subject wereexcluded from this experiment as they were too noisyto localize face-related activation The activation profilewas extracted for the Rubin-face and Rubin-vase stimulifrom both the fusiform gyrus and lateral occipital foci

Figure 6 The Rubin facendashvase experiment (A) Examples of the

stimuli including the Rubin facendashvase stimuli (left) and frontal faces

and household objects (localizer stimuli) used to independently

localize face-related voxels (in the experiment the uniform surfacesshown here in gray were colored in light pink) (B) Activation profiles

for the Rubin facendashvase conditions in face-related voxels in the fusiform

gyrus and lateral occipital region Error bars indicate SEM acrosssubjects in each group

Avidan et al 1159

for each subject (see Table 2 Part 2 for foci) The av-eraged percent signal change across all repetitions ofeach condition and across the left and right hemi-spheres of each subject are presented in Figure 6BAs in the control group the average fusiform gyrusactivation in the CP subjects exhibited a selective re-sponse for the Rubin-face compared to Rubin-vaseindicating that this region in CP subjects is indeedinvolved in computing global information about facesand is sensitive to the context not simply to the localcomponents of the display The similarity between theactivation profiles of the two groups in the fusiformgyrus was confirmed by a repeated-measure ANOVAwhich revealed only a main effect of stimulus type( p lt 01)

The results in the lateral occipital region were onceagain more variable across subjects in both groups (seeTable 2 Part 2) The ANOVA revealed a marginallysignificant main effect of group ( p lt 06) but not ofstimulus type and no significant interaction Note thatalthough not statistically significant the selectivity forthe Rubin-face compared to the Rubin-vase conditionin the lateral occipital focus was somewhat reduced inthe CP subjects compared to their controls This resultis analogous to the somewhat reduced face selectivityfound in the lateral occipital region in the conventionalmapping experiment (Figure 2A)

DISCUSSION

In a series of four functional imaging experimentswe have shown that individuals with CP exhibit nor-mal face- and object-related fMRI activation patternsin the ventral occipito-temporal cortex particularly inthe face-related fusiform gyrus (FFA) and largely nor-mal activations in the lateral occipital region despitetheir profound behavioral impairment Importantly thispattern was evident across different types of stimuli(eg line drawings and movie clips) and across differ-ent paradigms

In the first study in a conventional mapping experi-ment a normal pattern of BOLD activation was foundwhen subjects viewed line drawings of face and nonfacestimuli and performed a one-back memory task Thisparadigm has been used extensively in the literature andserves as a rigorous baseline to demarcate areas of ac-tivation in the ventral occipito-temporal cortex Onepossible interpretation of the normal activation is thatthe task was not sufficiently taxing for the CP individ-uals Behavioral data collected concurrently with theimage acquisition rules out this possibilitymdashalthoughthe task was easy for the controls this was not so forthe CP individuals who were clearly impaired on thistask Thus the normal face-related activation found inCP cannot merely be explained by the claim that the taskwas too simplistic for the CP subjects

A second possible interpretation and one that isdiametrically opposed to the first one offered above isthat the task was excessively difficult for the CP and itis this difficulty that drove the face-related activationparticularly in the fusiform gyrus of the CP subjects In-deed previous studies have shown that activation in theFFA increases as a function of working memory load(Druzgal amp DrsquoEsposito 2001) This explanation is nottenable either as the CP subjects continued to exhibitnormal face-related activation even when they were notrequired to perform any behavioral task during themotion pictures experiment Further research is clearlyneeded to unequivocally determine the contribution oftask difficulty to face-related activation in CP individ-uals We are now starting to parametrically explore theeffects of task difficulty in these CP individuals usingnew paradigms The critical point at present is thatnormal face-related activation is obtained at the verysame time that the behavioral impairment is manifest

The finding that face- and object-related activation inthe ventral visual cortex is apparently normal is replicat-ed in the motion picture experiment As mentionedabove this replication attests to the generalizability ofthe finding from the first conventional mapping exper-iment to more natural stimuli and viewing conditions

Importantly the normal signature of face-related ac-tivation in CP was also found when additional experi-mental manipulations were used CP subjects showednormal recovery from adaptation when different faceswere shown suggesting that neurons within these re-gions were sufficiently sensitive to differentiate betweenthe face exemplars used in this experiment Althoughone might think that normal adaptation is not thatsurprising given that the face stimuli used in this exper-iment were very different from each other it is impor-tant to bear in mind that CP subjects were actuallyimpaired in discriminating between such stimulimdashinthe conventional mapping experiment similar stimuliwere used and the CP subjects were impaired suggest-ing that the stimuli were confusing and not as percep-tually dissimilar for them as one might assume Of noteis that in addition to the adaptation for faces found inthe present experiment CP subjects also exhibitednormal adaptation levels in the fusiform gyrus and otherface-related regions for nonface stimuli suggestingthat similar to control subjects and replicating previousfindings these regions are involved in processing otherobject categories in addition to faces (Avidan et al 2002Haxby Gobbini et al 2001)

Although normal adaptation for faces is apparentdetermining its origin is less obvious Because the exactsame picture of an individual face was repeated duringthe adaptation condition we cannot determine at thisstage whether the adaptation found in the fusiformgyrus was due to the repetition of the exact same shape(ie based on perceptual geometry or structure) or therepetition of the same facial identity Differentiating

1160 Journal of Cognitive Neuroscience Volume 17 Number 7

between these alternatives and establishing whetherneural activity in the FFA is sufficient for normal rep-resentation of fine distinctions among individual faceswill require further experiments in which more subtlevariations will be made to face stimuli The use of anevent-related adaptation paradigm in such future ex-periments will also minimize possible attentional varia-tions that are inherent in the block-design adaptationparadigm employed here

Finally in the last experiment similar to control sub-jects once again CP subjects also exhibited a selectiveBOLD signal for Rubin-face compared to Rubin-vasestimuli in the fusiform gyrus Importantly these stimulicontain similar local elements but induce two differentvisual percepts either of a face or of a vase dependingon figurendashground segmentation and global integrationand shape processing Hence selectivity for the Rubin-face over the Rubin-vase stimuli implies that as in thecontrol group CP subjectsrsquo face-related activation ismodulated by global grouping processes and is notsimply induced by the representation of local face parts(Hasson Hendler et al 2001)

In sum the CP subjects exhibit normal face- andobject-related activation patterns in multiple regions ofthe ventral occipito-temporal cortex under various ex-perimental paradigms Of note however is that groupdifferences between CP subjects and their controls werefound in regions outside the occipito-temporal cortexindicating that the absence of group differences is notmerely a lack of statistical power Taken together thesefindings suggest that normal face-related activation inventral occipito-temporal regions as measured withfMRI is not sufficient to ensure intact face processingand does not necessarily reflect normal face represen-tation within those regions

Can the BehaviorndashBOLD Dissociationbe Reconciled

The central result is that the BOLD activation in CP inventral occipito-temporal regions is largely not differen-tiable from that of the control subjects notwithstandingtheir marked behavioral impairment How might thisdissociation be explained

It has been previously suggested that the FFA ismainly involved in face detection that is discriminatingbetween faces and stimuli from other object categories(Tong et al 2000) an ability that is largely preserved inCP The apparently normal FFA activation in the CPsubjects might then reflect the intact involvement ofthe FFA in deriving a rather coarse and rudimentarystructural representation which suffices for some tasks(ie detection) but not for more taxing tasks such asidentification Evidence compatible with this possibleinterpretation suggests that the final stage of face iden-tification is not completed at the level of high-order

visual face-related regions such as the FFA rathersuccessful recognition requires activation in more ante-rior regions of the temporal lobe where a more abstractor semantic representation of faces is derived (HaxbyHoffman et al 2000 Leveroni et al 2000 SergentOhta amp Macdonald 1992) Further support for the roleof anterior temporal regions in mediating face identifi-cation and particularly for being a site of facial memory(long-term representation) comes from lesion studiesFor example Barton and Cherkasova (2003) foundthat patients with anterior temporal lesions were themost impaired in a task that required generating long-term representations of famous faces as in mentalimagery compared to patients with more posterioroccipito-temporal lesions Similarly patients with ante-rior temporal lobectomy (following intractable epilepsy)were found to be impaired in identifying familiar faces(Glosser Salvucci amp Chiaravalloti 2003) The source ofthe behavioral problem in CP might then arise in theabnormal propagation of activation from the intact FFAto more anterior temporal regions

However the idea that the FFA only mediates facedetection is not universally accepted and the cleandivision of labor between the FFA and more anteriorregions might not be totally correct Although notdenying the contribution of more anterior regions toface processing and particularly to face identificationrecent experimental evidence suggests that the FFA isinvolved not only in face detection but also in faceidentification Thus normal subjects exhibited differen-tial responses for face detection and identification in thefusiform gyrus with the two processes resulting in eitherdifferential fMRI amplitude (Grill-Spector et al 2004) ortemporal scales (Puce Allison et al 1999 and seeSugase Yamane Ueno amp Kawano 1999 for similarfindings from single-unit recordings in monkeys) Thesedifferential responses in the fusiform gyrus may poten-tially be the outcome of feedback from more anteriorregions (as above) andor from additional computationtaking place within the fusiform gyrus itself Which ofthese is perturbed in CP remains unclear and they arenot mutually exclusive either Propagation to or fromanterior regions may be affected in CP However recallthat CP subjects exhibit a clear deficit not only in faceidentification (naming of famous faces) but also indiscrimination of unfamiliar faces (Figure 1B) (see alsoBehrmann et al 2005) This deficit in the more per-ceptual (not just memorial) aspects of face processingimplicates posterior occipito-temporal regions ratherthan exclusively the more anterior areas (Barton PressKeenan amp OrsquoConnor 2002 Wada amp Yamamoto 2001)It remains a possibility therefore that the fusiformactivation itself may be abnormal in CP Because thecurrent study employed detection and discriminationstasks the abnormality may still be uncovered undermore challenging conditions such as face identificationor other fine-grained face tasks

Avidan et al 1161

We also cannot rule out the possibility that differ-ences in the activation patterns in the fusiform gyrusandor other parts of the face processing networkbetween CP and normal control subjects might exist inthe temporal domain Such differences could underliethe behavioral dysfunction of CP subjects but may notbe evident with the temporal resolution afforded byfMRI (in the order of seconds) Some evidence sup-porting this view comes from ERP studies showing thatthe relatively early N170 potential which shows faceselectivity in normal subjects failed to show such se-lectivity in CP subjects (Kress amp Daum 2003b BentinDeouell et al 1999)

Finally it might be that the small differences inBOLD activation in CP subjects such as the slight re-duction in face selectivity in the lateral occipital region(eg Figures 2A and 6B) might be lsquolsquoamplifiedrsquorsquo whentranslated into behavioral performance If so the lateraloccipital region would serve as a critical link in the faceprocessing network Further support for the role ofthe occipital region in face processing comes fromstudies showing that acquired prosopagnosic patientsoften have lesions that involve not only the fusiformgyrus but also the lateral occipital region (Barton et al2002 Wada amp Yamamoto 2001) Moreover brain dam-age to this region but not to the FFA in some individ-uals may be associated with severe prosopagnosia(Rossion Caldara et al 2003)

As is evident there are a host of potential expla-nations to account for the dissociation between the im-paired behavioral profile and the apparently normalBOLD activation pattern We are currently exploringthese issues using new imaging experiments and hopethat these will uncover in greater detail the mecha-nisms giving rise to CP Suffice it to say that no one ex-planation captures the BOLDndashbehavior dissociationWhat is crucial however is that a simple assignmentof face processing to the FFA is no longer defensibleand a deeper examination of the computational prop-erties of the FFA and associated regions is demanded bythese findings

Face-related Activation Outside theOccipito-temporal Cortex

In the current study we acquired whole-brain functionalscans and thus were able to explore differences be-tween the CP and control groups in the responsepatterns in all regions of the brain not only in theoccipito-temporal regions A particularly novel findingwas that robust bilateral face-related activation wasevident in prefrontal regions (precentral sulcus inferiorfrontal sulcus anterior lateral sulcus) in the CP groupIn contrast the control group only exhibited right-lateralized activation in the precentral sulcus (Figure 3)Studies of normal individuals performing workingmemory tasks have revealed face-related activation in

prefrontal regions (Druzgal amp DrsquoEsposito 2003 Salaet al 2003 Haxby Petit et al 2000) and some ERPstudies have detected some inferior prefrontal cortexsites that generate small but specific face-specific re-sponses (Allison Puce Spencer amp McCarthy 1999)

The poor performance exhibited by CP subjectscompared to their controls on the one-back memorytask specifically for face stimuli is consistent with theinterpretation of this prefrontal activity being related toworking memory However given that the conven-tional face and object mapping experiment was notspecifically designed to look at the different componentsthat usually comprise a memory study (encoding mem-ory delay response) further research that specificallyand parametrically manipulates the memory load offaces and other object stimuli is required in order todetermine the role of this activation in face processing inCP subjects

This study has exploited the unique possibility ofexamining the activation of the fusiform gyrus andother cortical regions in an unusual group of individ-uals with CP Delineating the neural mechanism thatgives rise to abnormal behavior and by logical infer-ence normal behavior is a critical goal of neurosci-ence As might be predicted the correspondencesbetween the brain and behavior are not transparentand in this study are dissociated with the behavioralimpairment being evident concomitant with normalFFA activation Taken together however this studyhas indicated patterns of similarities and patterns ofpotential difference in this population of CP individ-uals relative to their normal counterparts The challengethat lies before us is to use these patterns to explainthe face processing deficit in detail

METHODS

Subjects

Four healthy CP subjects (2 men) aged between 29 and60 years with no discernable cortical lesion or anyhistory of neurological disease participated in all ex-periments (for details see Table 1) Another CP sub-

Table 1 Biographical Information about CP Subjects

Congenital Prosopagnosia (CP) Subjects

Subjectsrsquo Initials Sex Age

TM M 27

KM F 60

MT M 41

BE F 29

KM and TM are a mother and a son

1162 Journal of Cognitive Neuroscience Volume 17 Number 7

ject (NI male aged 40) was tested behaviorally (seeBehrmann et al 2005) but because we could not ac-quire his imaging data he was excluded from thepresent article Seven of the 12 control subjects whoparticipated in the behavioral studies along with threenew participants completed the imaging experimentsThe control group included at least two age- and sex-matched controls for each CP individual All CP andcontrol subjects were right-handed and had normal orcorrected-to-normal vision All subjects consented to par-ticipate in the experiments and the protocol was ap-proved by the Institutional Review Boards of CarnegieMellon University and the University of Pittsburgh

Imaging Experiments

Visual Stimulation

Visual stimuli were generated using the E-prime IFISsoftware (Psychology Software Tools Pittsburgh PAUSA) and projected via LCD to a screen located in theback of the scanner bore behind the subjectrsquos head Sub-jects viewed the stimuli through a tilted mirror mountedabove their eyes on the head coil Before the scan sub-

jects were familiarized with the visual stimuli and tasksof each of the imaging experiments

MRI Setup

All subjects were scanned in a 3-T Siemens Allegrascanner equipped with a standard head coil Subjectsparticipated in one scanning session that lasted 15 to2 hr in which all the functional and anatomical scanswere acquired The order of the functional experimentswithin the scanning session was generally constant forall subjects BOLD contrast was acquired using gradient-echo echo-planar imaging sequence Specific parame-ters TR = 3000 msec TE = 35 msec flip angle = 908FOV = 210 210 mm2 matrix size 64 64 35 axialslices 3 mm thickness no gap High-resolution anatom-ical scans (T1-weighted 3-D MPRAGE) were also acquiredto allow accurate cortical segmentation reconstructionand volume-based statistical analysis Specific parame-ters TE = 349 flip angle = 88 FOV = 256 256 mm2matrix size = 256 256 slice thickness = 1 mm numberof slices = 160ndash192 orientation of slices was eitherhorizontal or sagittal For three of the CP subjects the

Table 2 Laterality in force-selective activation

1 Talairach Coordinates Fusiform Gyrus

Left Hemisphere Right Hemisphere

x y z x y z

CP subjects 40 plusmn 2 53 plusmn 3 19 plusmn 2 38 plusmn 5 47 plusmn 8 17 plusmn 3

Control subjects 37 plusmn 4 41 plusmn 5 18 plusmn 4 36 plusmn 2 40 plusmn 7 17 plusmn 3

2 Activation Foci

Localizer Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 4 ndash ndash

Control subjects 9 1 ndash 4 2 1

Rubin FacendashVase Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 1 3 ndash

Control subjects 9 ndash ndash 3 2 4

Part 1 Talairach coordinates of the face-related activation in the fusiform gyrus of CP and control subjects extracted from the conventionalface and object mapping experiment Part 2 Number of CP and control subjects that showed bilateral right-lateralized or left-lateralizedface-related activation in the fusiform gyrus and in the lateral occipital region in the conventional face and object mapping experiment andin the Rubin facendashvase experiment

Avidan et al 1163

3-D brain reconstruction was performed on imagespreviously acquired on a 15-T GE scanner

Experiments

Conventional Face and Object Mapping Experiment

Stimuli included line drawings of faces buildings com-mon objects and geometric patterns presented in ashort block design (for details see Levy et al 2001Figure 1A) Each epoch lasted 9 sec and contained ninestimuli from the same category there were eight differ-ent images and one immediate repetition of one of thestimuli Each stimulus was presented for 800 msecfollowed by a 200-msec interval of blank screen Therewere seven repetitions of each condition and their orderwas pseudorandomized across subjects Visual epochswere interleaved with 6-sec blank epochs The experi-ment lasted 450 sec and started and ended with extend-ed blank epochs of 27 and 9 sec respectively Subjectswere instructed to fixate on a red central fixation dot toperform a one-back memory task and to press a key ona response glove to indicate the repeated image Bothaccuracy and reaction time were recorded

The Motion Pictures Experiment

The experiment lasted 10 minutes and was composed of32 epochs (15 sec long) of movie clips each containingimages from one of four possible categories people invarious situations navigation of the camera through citybuildings or through open fields and miscellaneousimages of objects from different categories (machinesfalling water etc for details see Hasson Nir et al2004 Figure 4) The experiment started with a 51-secblank period followed by 9 sec of pattern stimuli andended with a 1-minute blank Subjects were simplyinstructed to watch the movie clips

The Adaptation Experiment

The experiment included line drawing images fromthree different categories faces buildings and cars(for details see Avidan et al 2002 Figure 5A) Stimuliwere presented in 12-sec epochs that contained either12 different images (lsquolsquodifferentrsquorsquo condition) from thesame category or 12 repetitions of the same identicalstimulus (lsquolsquoidenticalrsquorsquo condition) Within epochs eachstimulus was presented for 800 msec followed by200 msec of fixation only The experiment lasted480 sec and started and ended with a long blank pe-riod of 21 and 15 sec respectively The first visualblock always contained geometric patterns and was ex-cluded from the analysis Visual epochs were interleavedwith 6-sec blank epochs and the order of visual epochswas pseudorandomized across subjects Subjects wereinstructed to continuously fixate a central red dot and

to covertly categorize the images as follows man wom-an or a child for the face stimuli apartment townhouseor public facility for the buildings and private car truckor bus for the vehicles

The Rubin FacendashVase Experiment

The experiment included modified versions of the Rubinfacendashvase stimuli as well as line drawings of front facesand household objects (for details see Hasson Hendleret al 2001 Figure 6A) The Rubin facendashvase stimuli weremodified so that subjectsrsquo perception was biased towardsone perceptual state or the other This was done by firstuniformly coloring the vase area within each of thestimuli and placing stripes over the profile Stimuli werethen placed either over a striped background (biastowards perception of the vase due to closure of thevase stimulus) or over a uniform-color background (biastowards perception of the face due to closure of theface stimulus) Critically the Rubin face and vase sharethe same local contours but give rise to different globalperceptions In order to ensure that subjectsrsquo perceptionremained biased to one perceptual interpretation (ei-ther the profile or vase) during the experiment stimuliwere presented briefly (200 msec) and were immediate-ly followed by a masking grid that was presented for800 msec Stimuli were presented in 9-sec blocks thatwere interleaved with 6-sec blanks Each condition wasrepeated eight times and the order of presentation waspseudorandomized across subjects Subjects maintainedcentral fixation throughout the experiment and per-formed a one-back memory task however due to atechnical failure their responses were not recorded Theexperiment lasted 507 sec and started and ended withextended blank periods

Data Analysis

Data analysis was performed using the BrainVoyager48 software package (Brain Innovation MaastrichtNetherlands) and complementary in-house softwareDetailed description of the data analysis is provided inHasson Harel et al (2003) Briefly for each subject thecortical surface was reconstructed and unfolded intothe flattened format Functional data for each subjectfrom each experiment were analyzed separately Pre-processing of functional scans included 3-D motioncorrection and filtering of low frequencies up to fivecycles per experiment (slow drift) Statistical analysiswas based on the GLM For each experiment each ofthe conditions (except for blank) was defined as a sepa-rate predictor a boxcar shape was used to model eachpredictor and a 3-sec lag was assumed Percent signalchange for each subject in each experiment (except forthe motion pictures experiment) was calculated as thepercent activation from a blank baseline For the mo-

1164 Journal of Cognitive Neuroscience Volume 17 Number 7

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

projected on an inflated brain of one individual subjectand shown from a ventral view To enable direct compar-isons between CP and control subjects both maps areshown using the same statistical threshold ( p lt 005 forfaces p lt 05 for buildings General Linear Model [GLM]multisubjects random-effects analysis) Note the similar-ity between the maps of the two groups in the locationand extent of the face-related activation Similar maps arealso shown in Figure 1D for each of the four CP subjectsand four representative controls Note that despite thevariability between subjects all CP subjects still clearlyexhibited face related activation similar to the controlsubjects

In order to assess further the activation in the fusiformgyrus we analyzed the spatial extent of the activation inthis region by counting the number of activated voxels inthe left and right hemispheres for each subject in eachgroup A repeated-measures ANOVA with group (CPcontrols) hemisphere (left right) as the repeated mea-sure and the number of voxels as the dependentmeasure did not reveal any significant effect ( p gt 1)thus suggesting that the spatial extent of the activa-tion in the fusiform gyrus did not differ between thetwo groups The similarity between the groups is alsoevident from the Talairach coordinates (Talairach ampTournoux 1988) of the face-related activation in thefusiform gyrus presented in Table 2 (Part 1)

Figure 1C also shows the average building-relatedactivation (building vs faces and objects) in the twogroups (green patches) Building-related activation istypically located outside the retinotopic borders in theparahippocampal gyrus and collateral sulcus medial tothe face-related activation in the FFA (Levy et al 2001Epstein amp Kanwisher 1998) and thus serves as animportant indicator for the reliability of the anatomicallocation of the face-related activation The control groupshows the same loci of activation as depicted in pre-vious studies as do the CP subjects Again this samepattern is also evident for each individual CP subject asshown in Figure 1D

To assess the selectivity for the different object cate-gories within the face-related regions the activationprofile from the fusiform gyrus and lateral occipitalregion was extracted for each subject using the lsquolsquointernallocalizerrsquorsquo approach (see Methods) Figure 2A showsthe percent signal change averaged across all the repe-titions of each condition and collapsed across bothhemispheres for the face-related regions for both CP(left column) and control subjects (right column) (seeTable 2 Part 2 for the left and right activation foci foreach group) A repeated-measures ANOVA was per-formed separately for the activation in the fusiformregion and lateral occipital region with group (CPcontrols) stimulus type (faces buildings objects pat-terns) and percent signal change as the dependentmeasure Only a main effect of stimulus type ( p lt0001) was found in both face-related regions Note that

the activation profiles were obtained using the internallocalizer approach and thus were not biased in any wayconfirming the clear face selectivity There was no maineffect of group nor a significant interaction betweengroup and stimulus type however the interaction effectin the fusiform gyrus was close to significant ( p lt 06activation for faces was slightly increased for CP com-pared with controls whereas activation for patterns wasdecreased) Note that although not statistically signifi-cant the face selectivity in the lateral occipital focus wassomewhat reduced in the CP subjects compared to theircontrols (percent signal change for faces compared tobuildings and patterns)

Face-related regions are part of a large distributedoccipito-temporal network of object representation(Hasson Harel et al 2003 Avidan Hasson HendlerZohary amp Malach 2002 Haxby Gobbini et al 2001) inwhich multiple regions even if not selective for facesstill carry substantial information about faces (HaxbyGobbini et al 2001) It is therefore possible that thesource of the behavioral deficit in CP might arise from adysfunction in such regions To explore this possibilitywe assessed the stimulus selectivity in regions that werepreferentially activated by the nonface stimuli used inthe experiment by applying the same procedure de-scribed above for faces (Figure 2B and C) This analysiswas first performed for the building-related activation inthe collateral sulcus and importantly the activationprofile was only sampled from the anterior nonretino-topic part of this region Three CP and all controlsubjects exhibited building-related activation bilaterallyand one CP subject (TM) showed right-lateralized acti-vation As in the face-related regions only a main effectof stimulus type ( p lt 0001) was observed with no maineffect of group nor an interaction between these factors

Using the same procedure we delineated foci withobject-specific activation (compared with faces andbuildings) in the lateral occipital sulcus the medial bankof the fusiform gyrus and the inferior temporal sulcus(ITS) However these activations were less consistentacross both groups and were mostly found in the lefthemisphere The most consistent activation was foundin the ITS in both groups (Figure 2C) Two CP subjectsexhibited bilateral activation and two others exhibitedleft-lateralized activation In the control group six sub-jects exhibited left-lateralized activation and two ex-hibited right-lateralized activation whereas two othersexhibited bilateral activation Again only the main effectof condition was significant ( p lt 001) the group effectwas close to significant ( p lt 06 higher percent signalchange in the control group for all conditions comparedto CP) and the interaction between the factors was notsignificant

The absence of group differences cannot be attribut-able to reduced statistical power although the patternsof activation were largely similar in occipito-temporalregions across the groups clear differences emerged in

Avidan et al 1153

prefrontal regions As is evident from Figure 3 the CPgroup exhibited strong face-related activation (faces vsbuilding and objects p lt 005 GLM multisubjectsrandom-effects analysis) in the precentral sulcus theinferior frontal sulcus and the anterior part of the lateral

sulcus in both hemispheres The control group ex-hibited some activation in the precentral sulcus but toa much lesser extent and only in the right hemisphere(Figure 3) Face-selective activation in prefrontal regionshas been previously reported when normal subjects

Figure 2 Activation profiles

in ROIs in the conventional

face and object mapping

experiment (A) Averagedactivation profiles of

face-related voxels in the

fusiform gyrus and lateraloccipital region of CP and

control subjects Activation

profiles were obtained

using the lsquolsquointernal localizerrsquorsquoapproach (see Methods)

and the graphs show the

averaged percent signal

change for each experimentalcondition (faces buildings

objects patterns) Error

bars indicate SEM acrosssubjects in each group (B)

Averaged activation profiles

in building-related voxels

in the collateral sulcus (C)Averaged activation profiles

in object-related voxels in

the inferior temporal sulcus

1154 Journal of Cognitive Neuroscience Volume 17 Number 7

perform working memory tasks using face stimuli(Druzgal amp DrsquoEsposito 2003 Sala Rama amp Courtney2003 Haxby Petit Ungerleider amp Courtney 2000) Theprefrontal face-related activation in the current studymight then reflect the increased difficulty for the CPsubjects in the one-back memory task for faces relativeto the other conditions and to the control subjects andthe consequent recruitment of prefrontal regions tomaintain face representations (see Figure 1B)

The results from the conventional face and objectmapping experiment indicate that individuals with CP re-veal normal face-related activation in the ventral occipito-temporal cortex and close-to-normal selectivity in thelateral occipital region in terms of the anatomical loca-tion of face-selective activation and its spatial extentthe signal strength and the stimulus selective responseCrucially this normal profile of activation is apparentdespite the significant face processing impairment ex-hibited at the very same time Of note also is that theCP subjects also exhibited normal object-related activa-tion (common objects and buildings) in the occipito-temporal cortex That group differences between theCP subjects and their controls can be detected (in pre-frontal regions) attests to the sufficiency of the statisticalpower of the current paradigm to detect differenceswhen they exist and therefore lends further weightto the conclusion that there are no detectable groupdifferences in the posterior areas of cortex

These results reveal a dissociation of a BOLDndashbehav-ior correlation in the ventral occipito-temporal cortexand particularly in the FFA To further confirm thisresult we utilized additional paradigms that could as-

sist in exploring the mechanisms giving rise to the ab-normal behavior

Motion Pictures Experiment

Although serving as a powerful tool for studying objectselectivity in the human visual cortex the conventionalface and object mapping experiment exploits viewingconditions and stimuli that are rather contrived Fur-thermore as evident from Figure 1B the one-backmemory task performed during this experiment wassignificantly more difficult for the CP subjects com-pared to their controls particularly for the face stimuliIt has previously been shown that activity in the FFAcan increase as a function of working memory load intasks involving faces (Druzgal amp DrsquoEsposito 2001)Thus the seemingly normal activation found in theconventional mapping experiment in the CP subjectscould actually reflect additional computation takingplace due to the relative difficulty of the behavioraltask for these individuals

To address these issues in this next experiment weused a motion picture paradigm first to test whetherthe face and object-related activation similarities ob-served in the conventional mapping experiment arereplicable with more natural stimuli and under morenatural viewing conditions and second to determinewhether normal face-related activation can be foundregardless of the behavioral task performed in thescanner In this experiment subjects freely viewed asequence of short video clips each containing stimuli

Figure 3 Activation maps in

prefrontal regions in the

conventional face and object

mapping experiment Sameactivation maps as in

Figure 1C but here the brain is

shown in a lateral view in orderto reveal the activation in

prefrontal regions CP subjects

are shown in the upper panel

and controls in the lower oneAbbreviations LS = lateral

sulcus CS = central sulcus

PreCS = precentral sulcus

IFS = inferior frontal sulcus

Avidan et al 1155

of faces and people buildings navigation or miscella-neous objects (Hasson Nir Levy Fuhrmann amp Malach2004) Importantly no specific task was required mak-ing this experiment more akin to naturally surveyingonersquos visual environment Figure 4A shows the activa-tion maps of a single representative CP and controlsubject projected on an inflated and unfolded brainrepresentation The white lines on the posterior partof the unfolded representation show the retinotopicborders mapped for each subject The face and build-ing activation maps obtained in this experiment largelyreplicate the results from the conventional mappingexperiment (compare Figure 4A and Figure 1C) ex-cept that the activation seems to be more widespreadin the motion pictures experiment selective activa-tion for faces was found in the fusiform gyrus andthe lateral occipital region when contrasting faces withbuildings and navigation scenes (red patches) andselective activation for buildings and navigation sceneswere found in the collateral sulcus when applyingthe opposite contrast (green patches) Interestinglyface-related activation was also found within the STSin regions typically activated in response to facialmovements (Hoffman amp Haxby 2000 Puce Allisonet al 1998) and biological motion (Puce and Perrett2003) That these regions are activated here is con-sistent with the richness of the faces and motions inthe displays

To assess the spatial extent of the face-related acti-vation obtained in this experiment in both CP andcontrols the number of face-related voxels within thelateral occipital region fusiform gyrus and STS wascounted for each subject in each group To allow di-rect comparisons all face-related foci for all subjectswere selected using the same statistical threshold ( p lt0005) (see Figure 4A for two representative subjects)A repeated-measures ANOVA with group (CPcontrols)hemisphere (left right) and face-related region as therepeated-measures within-subject factors and the num-ber of voxels as the dependent measure was usedThere were infrequent instances of control subjects forwhom not all face-related foci could be identified in theselected statistical threshold and these were excludedfrom the ANOVA Critically all face-related regions weredetected for the CP subjects using that threshold Thisanalysis revealed a significant main effect of hemisphere( p lt 001) confirming that overall there were moreface-related voxels in the right compared with the lefthemisphere In addition there was a significant maineffect of face-related region ( p lt 005) with moreactivated voxels in the lateral occipital region than inthe fusiform gyrus or STS Importantly however thisanalysis did not reveal a main effect of group or any two-or three-way interactions thus suggesting that therewere no differences in the spatial extent of the activationbetween the CP and control group in any of the face-related regions

To further quantify the similarity between the CP andcontrol groups we sampled the activation profile ofthree regions of interest (ROIs) (fusiform gyrus lateraloccipital region and collateral sulcus) defined indepen-dently for each subject on the basis of the conventionalface and object mapping experiment (see Methods)Percent signal change averaged across the entire exper-iment and collapsed across the left and right hemi-spheres for each group is shown in Figure 4B Asabove all ROIs exhibited clear stimulus selectivity butof even greater importance is the striking similaritybetween the activation profiles of the two groups (CP =black line controls = white line) This similarity is alsoevident from the high values of the correlation coeffi-cient (fusiform gyrus = 80 lateral occipital region =89 CoS = 81) calculated for each ROI between theaverage activation profile of each group Importantlyhowever as evident from the figure this similarity isnot only in the overall correlation but also in themoment-by-moment waxing and waning of the activityThese results confirm the normal profile of face- andbuilding-related activation in the CP subjects when morenatural viewing conditions and stimuli are employedMoreover because this normal activation was foundhere in the absence of any behavioral task it impliesthat the normal activation found in the conventionalmapping experiment was not confounded by the diffi-culty of the behavioral task

Adaptation Experiment

The behavioral results reveal a clear impairment inCP individuals in unfamiliar face discrimination in asimultaneous matching task (Behrmann et al 2005)and in a one-back memory (sequential discrimination)task (Figure 1B) These tasks are largely perceptualand do not require that a specific or individual facebe represented Indeed CP subjects might be evenmore impaired than already evident when more preciseface knowledge is required their own self-testimoniessuggest that lsquolsquoall faces look alikersquorsquo Using the fMR-adaptation paradigm for studying the neuronal proper-ties of higher-order visual areas at a subvoxel resolutionit has been shown that normal subjects typically exhibita reduction in BOLD signal with repeated presentationof a stimulus (Grill-Spector amp Malach 2001) To exam-ine whether the face-selective activation noted above isless sensitive to face repetition in CP than in controlsubjects we compared adaptation levels for face repeti-tion across the two groups We also examined the adap-tation effect for nonface stimuli to determine whetherthe entire occipito-temporal region is normally sensitiveto repetition (Avidan et al 2002)

Stimuli were presented in epochs that included either12 different stimuli (faces building cars) or 12 repeti-tions of the same identical stimulus (see Figure 5A) Theactivation profile was sampled from ROIs defined in-

1156 Journal of Cognitive Neuroscience Volume 17 Number 7

Figure 4 Activation maps and activation profiles from the motion pictures experiment (A) Activation maps for faces (red) as well as buildings

and navigation (green) are shown for one representative CP subject and one control subject The data are shown on the inf lated and unfoldedmap representation of each individual using the same statistical threshold (GLM statistics p lt 0005) The white dotted lines on the unfolded maps

indicate the borders of the retinotopic areas that were mapped for each subject in a separate experiment Abbreviations LOS = lateral occipital

sulcus IOG = inferior occipital gyrus IPS = intraparietal sulcus PCS = postcentral sulcus Other abbreviations as in Figures 1 and 3 (B) Averaged

activation profiles of the CP (black) and control subjects (white) are shown for face-related voxels in the fusiform gyrus and lateral occipital regionand for building-related voxels in the collateral sulcus The signal is shown along the time-axis of the experiment and the vertical colored bars

indicate the different experimental conditions The correlation coefficient values between the entire activation profile of CP and control subjects

are shown for each ROI Error bars indicate SEM across subjects in each group

Avidan et al 1157

Figure 5 Experimental design and activation profiles of the adaptation experiment (A) Examples of stimuli and experimental design Face

building and car line drawings were presented in epochs containing either 12 different or 12 identical images (B) Averaged activation profiles inface-related voxels in the fusiform gyrus (top) and the lateral occipital region (bottom) for the CP and control subjects Error bars indicate SEM

across subjects in each group (C) Averaged activation profiles in the building-related voxels in the collateral sulcus

1158 Journal of Cognitive Neuroscience Volume 17 Number 7

dependently for each subject using the conventionalmapping experiment Figure 5B shows the activationlevels within the face-related foci in the fusiform gyrusand lateral occipital region averaged across all repeti-tions of each experimental condition and across thetwo hemispheres for the two groups Again despitethe fact that the face-related regions were selected usingan external localizer CP individuals exhibited clear faceselectivity in the lsquolsquodifferentrsquorsquo conditions in these ROIsNote the similarity between the groups in the magni-tude of the adaptation that is the signal decrease fol-lowing face repetition (dark gray compared with lightgray bars) suggesting that face representations in CPsubjects are indeed sensitive to face repetition Theface-related regions also exhibit a strong adaptationeffect for the nonface stimuli (buildings and cars) in-dicating that like normal subjects these regions in theCP subjects contain neurons that are highly selectivefor those stimuli (Avidan et al 2002) These observa-tions were confirmed in a repeated-measure ANOVAin which the only significant effect in both regions wasa main effect of stimulus type ( p lt 0001) There wasno main effect of group nor an interaction betweenthe factors

Figure 5C shows a similar analysis for the building-related focus in the anterior collateral sulcus This regionexhibited strong adaptation for the building stimuli aswell as for the other object categories in both groupsOnce again the ANOVA revealed only a main effect ofstimulus type ( p lt 0001) The strong building selectiv-ity in the collateral sulcus combined with the findingthat this region exhibits similar adaptation levels forall stimulus categories including faces further suggeststhat the entire occipito-temporal object representationnetwork in CP subjects is intact and mirrors that of thenormal control subjects

Rubin FacendashVase Experiment

One possible interpretation of the apparently normalface-selective BOLD response in the CP subjects foundin the previous experiments is that it reflects a responseto the presence of face parts (eyes mouth and nose) inthe absence of an intact representation of the face as awhole To examine whether the activation in CP ismerely a product of a part-representation we employeda paradigm used previously which utilizes the famousRubin facendashvase illusion to demonstrate evidence forglobal face representations in normal subjects (HassonHendler Ben Bashat amp Malach 2001) Critically in theRubin facendashvase illusion similar local elements are al-ways present However depending on figurendashgroundsegmentation and global integration the local elementscan induce two different visual percepts either of a faceor of a vase (Figure 6A) If face-selective activation in thefusiform gyrus and other face-related regions depends

only on processing of the local elements then activa-tion in these regions should be similar in both theRubin-face and Rubin-vase conditions (same local ele-ments are present) However if face representation inthe fusiform gyrus and other face-related regions isaffected by the global integration of the face elementsinto the percept of either a face or a vase then increasedactivation should be found for the Rubin-face perceptualstates compared to the Rubin-vase perceptual states In-deed in normal subjects selective activation within face-related regions has been previously demonstrated forthe Rubin-face compared to the Rubin-vase stimuli sug-gesting that normal face-related activation is modulatedby global grouping processes and is not only induced bythe representation of local face parts (Hasson Hendleret al 2001)

Here face-selective regions were independently lo-calized initially for each subject by contrasting theepochs of frontal faces with those of household ob-jects (Figure 6A) Data from one control subject wereexcluded from this experiment as they were too noisyto localize face-related activation The activation profilewas extracted for the Rubin-face and Rubin-vase stimulifrom both the fusiform gyrus and lateral occipital foci

Figure 6 The Rubin facendashvase experiment (A) Examples of the

stimuli including the Rubin facendashvase stimuli (left) and frontal faces

and household objects (localizer stimuli) used to independently

localize face-related voxels (in the experiment the uniform surfacesshown here in gray were colored in light pink) (B) Activation profiles

for the Rubin facendashvase conditions in face-related voxels in the fusiform

gyrus and lateral occipital region Error bars indicate SEM acrosssubjects in each group

Avidan et al 1159

for each subject (see Table 2 Part 2 for foci) The av-eraged percent signal change across all repetitions ofeach condition and across the left and right hemi-spheres of each subject are presented in Figure 6BAs in the control group the average fusiform gyrusactivation in the CP subjects exhibited a selective re-sponse for the Rubin-face compared to Rubin-vaseindicating that this region in CP subjects is indeedinvolved in computing global information about facesand is sensitive to the context not simply to the localcomponents of the display The similarity between theactivation profiles of the two groups in the fusiformgyrus was confirmed by a repeated-measure ANOVAwhich revealed only a main effect of stimulus type( p lt 01)

The results in the lateral occipital region were onceagain more variable across subjects in both groups (seeTable 2 Part 2) The ANOVA revealed a marginallysignificant main effect of group ( p lt 06) but not ofstimulus type and no significant interaction Note thatalthough not statistically significant the selectivity forthe Rubin-face compared to the Rubin-vase conditionin the lateral occipital focus was somewhat reduced inthe CP subjects compared to their controls This resultis analogous to the somewhat reduced face selectivityfound in the lateral occipital region in the conventionalmapping experiment (Figure 2A)

DISCUSSION

In a series of four functional imaging experimentswe have shown that individuals with CP exhibit nor-mal face- and object-related fMRI activation patternsin the ventral occipito-temporal cortex particularly inthe face-related fusiform gyrus (FFA) and largely nor-mal activations in the lateral occipital region despitetheir profound behavioral impairment Importantly thispattern was evident across different types of stimuli(eg line drawings and movie clips) and across differ-ent paradigms

In the first study in a conventional mapping experi-ment a normal pattern of BOLD activation was foundwhen subjects viewed line drawings of face and nonfacestimuli and performed a one-back memory task Thisparadigm has been used extensively in the literature andserves as a rigorous baseline to demarcate areas of ac-tivation in the ventral occipito-temporal cortex Onepossible interpretation of the normal activation is thatthe task was not sufficiently taxing for the CP individ-uals Behavioral data collected concurrently with theimage acquisition rules out this possibilitymdashalthoughthe task was easy for the controls this was not so forthe CP individuals who were clearly impaired on thistask Thus the normal face-related activation found inCP cannot merely be explained by the claim that the taskwas too simplistic for the CP subjects

A second possible interpretation and one that isdiametrically opposed to the first one offered above isthat the task was excessively difficult for the CP and itis this difficulty that drove the face-related activationparticularly in the fusiform gyrus of the CP subjects In-deed previous studies have shown that activation in theFFA increases as a function of working memory load(Druzgal amp DrsquoEsposito 2001) This explanation is nottenable either as the CP subjects continued to exhibitnormal face-related activation even when they were notrequired to perform any behavioral task during themotion pictures experiment Further research is clearlyneeded to unequivocally determine the contribution oftask difficulty to face-related activation in CP individ-uals We are now starting to parametrically explore theeffects of task difficulty in these CP individuals usingnew paradigms The critical point at present is thatnormal face-related activation is obtained at the verysame time that the behavioral impairment is manifest

The finding that face- and object-related activation inthe ventral visual cortex is apparently normal is replicat-ed in the motion picture experiment As mentionedabove this replication attests to the generalizability ofthe finding from the first conventional mapping exper-iment to more natural stimuli and viewing conditions

Importantly the normal signature of face-related ac-tivation in CP was also found when additional experi-mental manipulations were used CP subjects showednormal recovery from adaptation when different faceswere shown suggesting that neurons within these re-gions were sufficiently sensitive to differentiate betweenthe face exemplars used in this experiment Althoughone might think that normal adaptation is not thatsurprising given that the face stimuli used in this exper-iment were very different from each other it is impor-tant to bear in mind that CP subjects were actuallyimpaired in discriminating between such stimulimdashinthe conventional mapping experiment similar stimuliwere used and the CP subjects were impaired suggest-ing that the stimuli were confusing and not as percep-tually dissimilar for them as one might assume Of noteis that in addition to the adaptation for faces found inthe present experiment CP subjects also exhibitednormal adaptation levels in the fusiform gyrus and otherface-related regions for nonface stimuli suggestingthat similar to control subjects and replicating previousfindings these regions are involved in processing otherobject categories in addition to faces (Avidan et al 2002Haxby Gobbini et al 2001)

Although normal adaptation for faces is apparentdetermining its origin is less obvious Because the exactsame picture of an individual face was repeated duringthe adaptation condition we cannot determine at thisstage whether the adaptation found in the fusiformgyrus was due to the repetition of the exact same shape(ie based on perceptual geometry or structure) or therepetition of the same facial identity Differentiating

1160 Journal of Cognitive Neuroscience Volume 17 Number 7

between these alternatives and establishing whetherneural activity in the FFA is sufficient for normal rep-resentation of fine distinctions among individual faceswill require further experiments in which more subtlevariations will be made to face stimuli The use of anevent-related adaptation paradigm in such future ex-periments will also minimize possible attentional varia-tions that are inherent in the block-design adaptationparadigm employed here

Finally in the last experiment similar to control sub-jects once again CP subjects also exhibited a selectiveBOLD signal for Rubin-face compared to Rubin-vasestimuli in the fusiform gyrus Importantly these stimulicontain similar local elements but induce two differentvisual percepts either of a face or of a vase dependingon figurendashground segmentation and global integrationand shape processing Hence selectivity for the Rubin-face over the Rubin-vase stimuli implies that as in thecontrol group CP subjectsrsquo face-related activation ismodulated by global grouping processes and is notsimply induced by the representation of local face parts(Hasson Hendler et al 2001)

In sum the CP subjects exhibit normal face- andobject-related activation patterns in multiple regions ofthe ventral occipito-temporal cortex under various ex-perimental paradigms Of note however is that groupdifferences between CP subjects and their controls werefound in regions outside the occipito-temporal cortexindicating that the absence of group differences is notmerely a lack of statistical power Taken together thesefindings suggest that normal face-related activation inventral occipito-temporal regions as measured withfMRI is not sufficient to ensure intact face processingand does not necessarily reflect normal face represen-tation within those regions

Can the BehaviorndashBOLD Dissociationbe Reconciled

The central result is that the BOLD activation in CP inventral occipito-temporal regions is largely not differen-tiable from that of the control subjects notwithstandingtheir marked behavioral impairment How might thisdissociation be explained

It has been previously suggested that the FFA ismainly involved in face detection that is discriminatingbetween faces and stimuli from other object categories(Tong et al 2000) an ability that is largely preserved inCP The apparently normal FFA activation in the CPsubjects might then reflect the intact involvement ofthe FFA in deriving a rather coarse and rudimentarystructural representation which suffices for some tasks(ie detection) but not for more taxing tasks such asidentification Evidence compatible with this possibleinterpretation suggests that the final stage of face iden-tification is not completed at the level of high-order

visual face-related regions such as the FFA rathersuccessful recognition requires activation in more ante-rior regions of the temporal lobe where a more abstractor semantic representation of faces is derived (HaxbyHoffman et al 2000 Leveroni et al 2000 SergentOhta amp Macdonald 1992) Further support for the roleof anterior temporal regions in mediating face identifi-cation and particularly for being a site of facial memory(long-term representation) comes from lesion studiesFor example Barton and Cherkasova (2003) foundthat patients with anterior temporal lesions were themost impaired in a task that required generating long-term representations of famous faces as in mentalimagery compared to patients with more posterioroccipito-temporal lesions Similarly patients with ante-rior temporal lobectomy (following intractable epilepsy)were found to be impaired in identifying familiar faces(Glosser Salvucci amp Chiaravalloti 2003) The source ofthe behavioral problem in CP might then arise in theabnormal propagation of activation from the intact FFAto more anterior temporal regions

However the idea that the FFA only mediates facedetection is not universally accepted and the cleandivision of labor between the FFA and more anteriorregions might not be totally correct Although notdenying the contribution of more anterior regions toface processing and particularly to face identificationrecent experimental evidence suggests that the FFA isinvolved not only in face detection but also in faceidentification Thus normal subjects exhibited differen-tial responses for face detection and identification in thefusiform gyrus with the two processes resulting in eitherdifferential fMRI amplitude (Grill-Spector et al 2004) ortemporal scales (Puce Allison et al 1999 and seeSugase Yamane Ueno amp Kawano 1999 for similarfindings from single-unit recordings in monkeys) Thesedifferential responses in the fusiform gyrus may poten-tially be the outcome of feedback from more anteriorregions (as above) andor from additional computationtaking place within the fusiform gyrus itself Which ofthese is perturbed in CP remains unclear and they arenot mutually exclusive either Propagation to or fromanterior regions may be affected in CP However recallthat CP subjects exhibit a clear deficit not only in faceidentification (naming of famous faces) but also indiscrimination of unfamiliar faces (Figure 1B) (see alsoBehrmann et al 2005) This deficit in the more per-ceptual (not just memorial) aspects of face processingimplicates posterior occipito-temporal regions ratherthan exclusively the more anterior areas (Barton PressKeenan amp OrsquoConnor 2002 Wada amp Yamamoto 2001)It remains a possibility therefore that the fusiformactivation itself may be abnormal in CP Because thecurrent study employed detection and discriminationstasks the abnormality may still be uncovered undermore challenging conditions such as face identificationor other fine-grained face tasks

Avidan et al 1161

We also cannot rule out the possibility that differ-ences in the activation patterns in the fusiform gyrusandor other parts of the face processing networkbetween CP and normal control subjects might exist inthe temporal domain Such differences could underliethe behavioral dysfunction of CP subjects but may notbe evident with the temporal resolution afforded byfMRI (in the order of seconds) Some evidence sup-porting this view comes from ERP studies showing thatthe relatively early N170 potential which shows faceselectivity in normal subjects failed to show such se-lectivity in CP subjects (Kress amp Daum 2003b BentinDeouell et al 1999)

Finally it might be that the small differences inBOLD activation in CP subjects such as the slight re-duction in face selectivity in the lateral occipital region(eg Figures 2A and 6B) might be lsquolsquoamplifiedrsquorsquo whentranslated into behavioral performance If so the lateraloccipital region would serve as a critical link in the faceprocessing network Further support for the role ofthe occipital region in face processing comes fromstudies showing that acquired prosopagnosic patientsoften have lesions that involve not only the fusiformgyrus but also the lateral occipital region (Barton et al2002 Wada amp Yamamoto 2001) Moreover brain dam-age to this region but not to the FFA in some individ-uals may be associated with severe prosopagnosia(Rossion Caldara et al 2003)

As is evident there are a host of potential expla-nations to account for the dissociation between the im-paired behavioral profile and the apparently normalBOLD activation pattern We are currently exploringthese issues using new imaging experiments and hopethat these will uncover in greater detail the mecha-nisms giving rise to CP Suffice it to say that no one ex-planation captures the BOLDndashbehavior dissociationWhat is crucial however is that a simple assignmentof face processing to the FFA is no longer defensibleand a deeper examination of the computational prop-erties of the FFA and associated regions is demanded bythese findings

Face-related Activation Outside theOccipito-temporal Cortex

In the current study we acquired whole-brain functionalscans and thus were able to explore differences be-tween the CP and control groups in the responsepatterns in all regions of the brain not only in theoccipito-temporal regions A particularly novel findingwas that robust bilateral face-related activation wasevident in prefrontal regions (precentral sulcus inferiorfrontal sulcus anterior lateral sulcus) in the CP groupIn contrast the control group only exhibited right-lateralized activation in the precentral sulcus (Figure 3)Studies of normal individuals performing workingmemory tasks have revealed face-related activation in

prefrontal regions (Druzgal amp DrsquoEsposito 2003 Salaet al 2003 Haxby Petit et al 2000) and some ERPstudies have detected some inferior prefrontal cortexsites that generate small but specific face-specific re-sponses (Allison Puce Spencer amp McCarthy 1999)

The poor performance exhibited by CP subjectscompared to their controls on the one-back memorytask specifically for face stimuli is consistent with theinterpretation of this prefrontal activity being related toworking memory However given that the conven-tional face and object mapping experiment was notspecifically designed to look at the different componentsthat usually comprise a memory study (encoding mem-ory delay response) further research that specificallyand parametrically manipulates the memory load offaces and other object stimuli is required in order todetermine the role of this activation in face processing inCP subjects

This study has exploited the unique possibility ofexamining the activation of the fusiform gyrus andother cortical regions in an unusual group of individ-uals with CP Delineating the neural mechanism thatgives rise to abnormal behavior and by logical infer-ence normal behavior is a critical goal of neurosci-ence As might be predicted the correspondencesbetween the brain and behavior are not transparentand in this study are dissociated with the behavioralimpairment being evident concomitant with normalFFA activation Taken together however this studyhas indicated patterns of similarities and patterns ofpotential difference in this population of CP individ-uals relative to their normal counterparts The challengethat lies before us is to use these patterns to explainthe face processing deficit in detail

METHODS

Subjects

Four healthy CP subjects (2 men) aged between 29 and60 years with no discernable cortical lesion or anyhistory of neurological disease participated in all ex-periments (for details see Table 1) Another CP sub-

Table 1 Biographical Information about CP Subjects

Congenital Prosopagnosia (CP) Subjects

Subjectsrsquo Initials Sex Age

TM M 27

KM F 60

MT M 41

BE F 29

KM and TM are a mother and a son

1162 Journal of Cognitive Neuroscience Volume 17 Number 7

ject (NI male aged 40) was tested behaviorally (seeBehrmann et al 2005) but because we could not ac-quire his imaging data he was excluded from thepresent article Seven of the 12 control subjects whoparticipated in the behavioral studies along with threenew participants completed the imaging experimentsThe control group included at least two age- and sex-matched controls for each CP individual All CP andcontrol subjects were right-handed and had normal orcorrected-to-normal vision All subjects consented to par-ticipate in the experiments and the protocol was ap-proved by the Institutional Review Boards of CarnegieMellon University and the University of Pittsburgh

Imaging Experiments

Visual Stimulation

Visual stimuli were generated using the E-prime IFISsoftware (Psychology Software Tools Pittsburgh PAUSA) and projected via LCD to a screen located in theback of the scanner bore behind the subjectrsquos head Sub-jects viewed the stimuli through a tilted mirror mountedabove their eyes on the head coil Before the scan sub-

jects were familiarized with the visual stimuli and tasksof each of the imaging experiments

MRI Setup

All subjects were scanned in a 3-T Siemens Allegrascanner equipped with a standard head coil Subjectsparticipated in one scanning session that lasted 15 to2 hr in which all the functional and anatomical scanswere acquired The order of the functional experimentswithin the scanning session was generally constant forall subjects BOLD contrast was acquired using gradient-echo echo-planar imaging sequence Specific parame-ters TR = 3000 msec TE = 35 msec flip angle = 908FOV = 210 210 mm2 matrix size 64 64 35 axialslices 3 mm thickness no gap High-resolution anatom-ical scans (T1-weighted 3-D MPRAGE) were also acquiredto allow accurate cortical segmentation reconstructionand volume-based statistical analysis Specific parame-ters TE = 349 flip angle = 88 FOV = 256 256 mm2matrix size = 256 256 slice thickness = 1 mm numberof slices = 160ndash192 orientation of slices was eitherhorizontal or sagittal For three of the CP subjects the

Table 2 Laterality in force-selective activation

1 Talairach Coordinates Fusiform Gyrus

Left Hemisphere Right Hemisphere

x y z x y z

CP subjects 40 plusmn 2 53 plusmn 3 19 plusmn 2 38 plusmn 5 47 plusmn 8 17 plusmn 3

Control subjects 37 plusmn 4 41 plusmn 5 18 plusmn 4 36 plusmn 2 40 plusmn 7 17 plusmn 3

2 Activation Foci

Localizer Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 4 ndash ndash

Control subjects 9 1 ndash 4 2 1

Rubin FacendashVase Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 1 3 ndash

Control subjects 9 ndash ndash 3 2 4

Part 1 Talairach coordinates of the face-related activation in the fusiform gyrus of CP and control subjects extracted from the conventionalface and object mapping experiment Part 2 Number of CP and control subjects that showed bilateral right-lateralized or left-lateralizedface-related activation in the fusiform gyrus and in the lateral occipital region in the conventional face and object mapping experiment andin the Rubin facendashvase experiment

Avidan et al 1163

3-D brain reconstruction was performed on imagespreviously acquired on a 15-T GE scanner

Experiments

Conventional Face and Object Mapping Experiment

Stimuli included line drawings of faces buildings com-mon objects and geometric patterns presented in ashort block design (for details see Levy et al 2001Figure 1A) Each epoch lasted 9 sec and contained ninestimuli from the same category there were eight differ-ent images and one immediate repetition of one of thestimuli Each stimulus was presented for 800 msecfollowed by a 200-msec interval of blank screen Therewere seven repetitions of each condition and their orderwas pseudorandomized across subjects Visual epochswere interleaved with 6-sec blank epochs The experi-ment lasted 450 sec and started and ended with extend-ed blank epochs of 27 and 9 sec respectively Subjectswere instructed to fixate on a red central fixation dot toperform a one-back memory task and to press a key ona response glove to indicate the repeated image Bothaccuracy and reaction time were recorded

The Motion Pictures Experiment

The experiment lasted 10 minutes and was composed of32 epochs (15 sec long) of movie clips each containingimages from one of four possible categories people invarious situations navigation of the camera through citybuildings or through open fields and miscellaneousimages of objects from different categories (machinesfalling water etc for details see Hasson Nir et al2004 Figure 4) The experiment started with a 51-secblank period followed by 9 sec of pattern stimuli andended with a 1-minute blank Subjects were simplyinstructed to watch the movie clips

The Adaptation Experiment

The experiment included line drawing images fromthree different categories faces buildings and cars(for details see Avidan et al 2002 Figure 5A) Stimuliwere presented in 12-sec epochs that contained either12 different images (lsquolsquodifferentrsquorsquo condition) from thesame category or 12 repetitions of the same identicalstimulus (lsquolsquoidenticalrsquorsquo condition) Within epochs eachstimulus was presented for 800 msec followed by200 msec of fixation only The experiment lasted480 sec and started and ended with a long blank pe-riod of 21 and 15 sec respectively The first visualblock always contained geometric patterns and was ex-cluded from the analysis Visual epochs were interleavedwith 6-sec blank epochs and the order of visual epochswas pseudorandomized across subjects Subjects wereinstructed to continuously fixate a central red dot and

to covertly categorize the images as follows man wom-an or a child for the face stimuli apartment townhouseor public facility for the buildings and private car truckor bus for the vehicles

The Rubin FacendashVase Experiment

The experiment included modified versions of the Rubinfacendashvase stimuli as well as line drawings of front facesand household objects (for details see Hasson Hendleret al 2001 Figure 6A) The Rubin facendashvase stimuli weremodified so that subjectsrsquo perception was biased towardsone perceptual state or the other This was done by firstuniformly coloring the vase area within each of thestimuli and placing stripes over the profile Stimuli werethen placed either over a striped background (biastowards perception of the vase due to closure of thevase stimulus) or over a uniform-color background (biastowards perception of the face due to closure of theface stimulus) Critically the Rubin face and vase sharethe same local contours but give rise to different globalperceptions In order to ensure that subjectsrsquo perceptionremained biased to one perceptual interpretation (ei-ther the profile or vase) during the experiment stimuliwere presented briefly (200 msec) and were immediate-ly followed by a masking grid that was presented for800 msec Stimuli were presented in 9-sec blocks thatwere interleaved with 6-sec blanks Each condition wasrepeated eight times and the order of presentation waspseudorandomized across subjects Subjects maintainedcentral fixation throughout the experiment and per-formed a one-back memory task however due to atechnical failure their responses were not recorded Theexperiment lasted 507 sec and started and ended withextended blank periods

Data Analysis

Data analysis was performed using the BrainVoyager48 software package (Brain Innovation MaastrichtNetherlands) and complementary in-house softwareDetailed description of the data analysis is provided inHasson Harel et al (2003) Briefly for each subject thecortical surface was reconstructed and unfolded intothe flattened format Functional data for each subjectfrom each experiment were analyzed separately Pre-processing of functional scans included 3-D motioncorrection and filtering of low frequencies up to fivecycles per experiment (slow drift) Statistical analysiswas based on the GLM For each experiment each ofthe conditions (except for blank) was defined as a sepa-rate predictor a boxcar shape was used to model eachpredictor and a 3-sec lag was assumed Percent signalchange for each subject in each experiment (except forthe motion pictures experiment) was calculated as thepercent activation from a blank baseline For the mo-

1164 Journal of Cognitive Neuroscience Volume 17 Number 7

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

prefrontal regions As is evident from Figure 3 the CPgroup exhibited strong face-related activation (faces vsbuilding and objects p lt 005 GLM multisubjectsrandom-effects analysis) in the precentral sulcus theinferior frontal sulcus and the anterior part of the lateral

sulcus in both hemispheres The control group ex-hibited some activation in the precentral sulcus but toa much lesser extent and only in the right hemisphere(Figure 3) Face-selective activation in prefrontal regionshas been previously reported when normal subjects

Figure 2 Activation profiles

in ROIs in the conventional

face and object mapping

experiment (A) Averagedactivation profiles of

face-related voxels in the

fusiform gyrus and lateraloccipital region of CP and

control subjects Activation

profiles were obtained

using the lsquolsquointernal localizerrsquorsquoapproach (see Methods)

and the graphs show the

averaged percent signal

change for each experimentalcondition (faces buildings

objects patterns) Error

bars indicate SEM acrosssubjects in each group (B)

Averaged activation profiles

in building-related voxels

in the collateral sulcus (C)Averaged activation profiles

in object-related voxels in

the inferior temporal sulcus

1154 Journal of Cognitive Neuroscience Volume 17 Number 7

perform working memory tasks using face stimuli(Druzgal amp DrsquoEsposito 2003 Sala Rama amp Courtney2003 Haxby Petit Ungerleider amp Courtney 2000) Theprefrontal face-related activation in the current studymight then reflect the increased difficulty for the CPsubjects in the one-back memory task for faces relativeto the other conditions and to the control subjects andthe consequent recruitment of prefrontal regions tomaintain face representations (see Figure 1B)

The results from the conventional face and objectmapping experiment indicate that individuals with CP re-veal normal face-related activation in the ventral occipito-temporal cortex and close-to-normal selectivity in thelateral occipital region in terms of the anatomical loca-tion of face-selective activation and its spatial extentthe signal strength and the stimulus selective responseCrucially this normal profile of activation is apparentdespite the significant face processing impairment ex-hibited at the very same time Of note also is that theCP subjects also exhibited normal object-related activa-tion (common objects and buildings) in the occipito-temporal cortex That group differences between theCP subjects and their controls can be detected (in pre-frontal regions) attests to the sufficiency of the statisticalpower of the current paradigm to detect differenceswhen they exist and therefore lends further weightto the conclusion that there are no detectable groupdifferences in the posterior areas of cortex

These results reveal a dissociation of a BOLDndashbehav-ior correlation in the ventral occipito-temporal cortexand particularly in the FFA To further confirm thisresult we utilized additional paradigms that could as-

sist in exploring the mechanisms giving rise to the ab-normal behavior

Motion Pictures Experiment

Although serving as a powerful tool for studying objectselectivity in the human visual cortex the conventionalface and object mapping experiment exploits viewingconditions and stimuli that are rather contrived Fur-thermore as evident from Figure 1B the one-backmemory task performed during this experiment wassignificantly more difficult for the CP subjects com-pared to their controls particularly for the face stimuliIt has previously been shown that activity in the FFAcan increase as a function of working memory load intasks involving faces (Druzgal amp DrsquoEsposito 2001)Thus the seemingly normal activation found in theconventional mapping experiment in the CP subjectscould actually reflect additional computation takingplace due to the relative difficulty of the behavioraltask for these individuals

To address these issues in this next experiment weused a motion picture paradigm first to test whetherthe face and object-related activation similarities ob-served in the conventional mapping experiment arereplicable with more natural stimuli and under morenatural viewing conditions and second to determinewhether normal face-related activation can be foundregardless of the behavioral task performed in thescanner In this experiment subjects freely viewed asequence of short video clips each containing stimuli

Figure 3 Activation maps in

prefrontal regions in the

conventional face and object

mapping experiment Sameactivation maps as in

Figure 1C but here the brain is

shown in a lateral view in orderto reveal the activation in

prefrontal regions CP subjects

are shown in the upper panel

and controls in the lower oneAbbreviations LS = lateral

sulcus CS = central sulcus

PreCS = precentral sulcus

IFS = inferior frontal sulcus

Avidan et al 1155

of faces and people buildings navigation or miscella-neous objects (Hasson Nir Levy Fuhrmann amp Malach2004) Importantly no specific task was required mak-ing this experiment more akin to naturally surveyingonersquos visual environment Figure 4A shows the activa-tion maps of a single representative CP and controlsubject projected on an inflated and unfolded brainrepresentation The white lines on the posterior partof the unfolded representation show the retinotopicborders mapped for each subject The face and build-ing activation maps obtained in this experiment largelyreplicate the results from the conventional mappingexperiment (compare Figure 4A and Figure 1C) ex-cept that the activation seems to be more widespreadin the motion pictures experiment selective activa-tion for faces was found in the fusiform gyrus andthe lateral occipital region when contrasting faces withbuildings and navigation scenes (red patches) andselective activation for buildings and navigation sceneswere found in the collateral sulcus when applyingthe opposite contrast (green patches) Interestinglyface-related activation was also found within the STSin regions typically activated in response to facialmovements (Hoffman amp Haxby 2000 Puce Allisonet al 1998) and biological motion (Puce and Perrett2003) That these regions are activated here is con-sistent with the richness of the faces and motions inthe displays

To assess the spatial extent of the face-related acti-vation obtained in this experiment in both CP andcontrols the number of face-related voxels within thelateral occipital region fusiform gyrus and STS wascounted for each subject in each group To allow di-rect comparisons all face-related foci for all subjectswere selected using the same statistical threshold ( p lt0005) (see Figure 4A for two representative subjects)A repeated-measures ANOVA with group (CPcontrols)hemisphere (left right) and face-related region as therepeated-measures within-subject factors and the num-ber of voxels as the dependent measure was usedThere were infrequent instances of control subjects forwhom not all face-related foci could be identified in theselected statistical threshold and these were excludedfrom the ANOVA Critically all face-related regions weredetected for the CP subjects using that threshold Thisanalysis revealed a significant main effect of hemisphere( p lt 001) confirming that overall there were moreface-related voxels in the right compared with the lefthemisphere In addition there was a significant maineffect of face-related region ( p lt 005) with moreactivated voxels in the lateral occipital region than inthe fusiform gyrus or STS Importantly however thisanalysis did not reveal a main effect of group or any two-or three-way interactions thus suggesting that therewere no differences in the spatial extent of the activationbetween the CP and control group in any of the face-related regions

To further quantify the similarity between the CP andcontrol groups we sampled the activation profile ofthree regions of interest (ROIs) (fusiform gyrus lateraloccipital region and collateral sulcus) defined indepen-dently for each subject on the basis of the conventionalface and object mapping experiment (see Methods)Percent signal change averaged across the entire exper-iment and collapsed across the left and right hemi-spheres for each group is shown in Figure 4B Asabove all ROIs exhibited clear stimulus selectivity butof even greater importance is the striking similaritybetween the activation profiles of the two groups (CP =black line controls = white line) This similarity is alsoevident from the high values of the correlation coeffi-cient (fusiform gyrus = 80 lateral occipital region =89 CoS = 81) calculated for each ROI between theaverage activation profile of each group Importantlyhowever as evident from the figure this similarity isnot only in the overall correlation but also in themoment-by-moment waxing and waning of the activityThese results confirm the normal profile of face- andbuilding-related activation in the CP subjects when morenatural viewing conditions and stimuli are employedMoreover because this normal activation was foundhere in the absence of any behavioral task it impliesthat the normal activation found in the conventionalmapping experiment was not confounded by the diffi-culty of the behavioral task

Adaptation Experiment

The behavioral results reveal a clear impairment inCP individuals in unfamiliar face discrimination in asimultaneous matching task (Behrmann et al 2005)and in a one-back memory (sequential discrimination)task (Figure 1B) These tasks are largely perceptualand do not require that a specific or individual facebe represented Indeed CP subjects might be evenmore impaired than already evident when more preciseface knowledge is required their own self-testimoniessuggest that lsquolsquoall faces look alikersquorsquo Using the fMR-adaptation paradigm for studying the neuronal proper-ties of higher-order visual areas at a subvoxel resolutionit has been shown that normal subjects typically exhibita reduction in BOLD signal with repeated presentationof a stimulus (Grill-Spector amp Malach 2001) To exam-ine whether the face-selective activation noted above isless sensitive to face repetition in CP than in controlsubjects we compared adaptation levels for face repeti-tion across the two groups We also examined the adap-tation effect for nonface stimuli to determine whetherthe entire occipito-temporal region is normally sensitiveto repetition (Avidan et al 2002)

Stimuli were presented in epochs that included either12 different stimuli (faces building cars) or 12 repeti-tions of the same identical stimulus (see Figure 5A) Theactivation profile was sampled from ROIs defined in-

1156 Journal of Cognitive Neuroscience Volume 17 Number 7

Figure 4 Activation maps and activation profiles from the motion pictures experiment (A) Activation maps for faces (red) as well as buildings

and navigation (green) are shown for one representative CP subject and one control subject The data are shown on the inf lated and unfoldedmap representation of each individual using the same statistical threshold (GLM statistics p lt 0005) The white dotted lines on the unfolded maps

indicate the borders of the retinotopic areas that were mapped for each subject in a separate experiment Abbreviations LOS = lateral occipital

sulcus IOG = inferior occipital gyrus IPS = intraparietal sulcus PCS = postcentral sulcus Other abbreviations as in Figures 1 and 3 (B) Averaged

activation profiles of the CP (black) and control subjects (white) are shown for face-related voxels in the fusiform gyrus and lateral occipital regionand for building-related voxels in the collateral sulcus The signal is shown along the time-axis of the experiment and the vertical colored bars

indicate the different experimental conditions The correlation coefficient values between the entire activation profile of CP and control subjects

are shown for each ROI Error bars indicate SEM across subjects in each group

Avidan et al 1157

Figure 5 Experimental design and activation profiles of the adaptation experiment (A) Examples of stimuli and experimental design Face

building and car line drawings were presented in epochs containing either 12 different or 12 identical images (B) Averaged activation profiles inface-related voxels in the fusiform gyrus (top) and the lateral occipital region (bottom) for the CP and control subjects Error bars indicate SEM

across subjects in each group (C) Averaged activation profiles in the building-related voxels in the collateral sulcus

1158 Journal of Cognitive Neuroscience Volume 17 Number 7

dependently for each subject using the conventionalmapping experiment Figure 5B shows the activationlevels within the face-related foci in the fusiform gyrusand lateral occipital region averaged across all repeti-tions of each experimental condition and across thetwo hemispheres for the two groups Again despitethe fact that the face-related regions were selected usingan external localizer CP individuals exhibited clear faceselectivity in the lsquolsquodifferentrsquorsquo conditions in these ROIsNote the similarity between the groups in the magni-tude of the adaptation that is the signal decrease fol-lowing face repetition (dark gray compared with lightgray bars) suggesting that face representations in CPsubjects are indeed sensitive to face repetition Theface-related regions also exhibit a strong adaptationeffect for the nonface stimuli (buildings and cars) in-dicating that like normal subjects these regions in theCP subjects contain neurons that are highly selectivefor those stimuli (Avidan et al 2002) These observa-tions were confirmed in a repeated-measure ANOVAin which the only significant effect in both regions wasa main effect of stimulus type ( p lt 0001) There wasno main effect of group nor an interaction betweenthe factors

Figure 5C shows a similar analysis for the building-related focus in the anterior collateral sulcus This regionexhibited strong adaptation for the building stimuli aswell as for the other object categories in both groupsOnce again the ANOVA revealed only a main effect ofstimulus type ( p lt 0001) The strong building selectiv-ity in the collateral sulcus combined with the findingthat this region exhibits similar adaptation levels forall stimulus categories including faces further suggeststhat the entire occipito-temporal object representationnetwork in CP subjects is intact and mirrors that of thenormal control subjects

Rubin FacendashVase Experiment

One possible interpretation of the apparently normalface-selective BOLD response in the CP subjects foundin the previous experiments is that it reflects a responseto the presence of face parts (eyes mouth and nose) inthe absence of an intact representation of the face as awhole To examine whether the activation in CP ismerely a product of a part-representation we employeda paradigm used previously which utilizes the famousRubin facendashvase illusion to demonstrate evidence forglobal face representations in normal subjects (HassonHendler Ben Bashat amp Malach 2001) Critically in theRubin facendashvase illusion similar local elements are al-ways present However depending on figurendashgroundsegmentation and global integration the local elementscan induce two different visual percepts either of a faceor of a vase (Figure 6A) If face-selective activation in thefusiform gyrus and other face-related regions depends

only on processing of the local elements then activa-tion in these regions should be similar in both theRubin-face and Rubin-vase conditions (same local ele-ments are present) However if face representation inthe fusiform gyrus and other face-related regions isaffected by the global integration of the face elementsinto the percept of either a face or a vase then increasedactivation should be found for the Rubin-face perceptualstates compared to the Rubin-vase perceptual states In-deed in normal subjects selective activation within face-related regions has been previously demonstrated forthe Rubin-face compared to the Rubin-vase stimuli sug-gesting that normal face-related activation is modulatedby global grouping processes and is not only induced bythe representation of local face parts (Hasson Hendleret al 2001)

Here face-selective regions were independently lo-calized initially for each subject by contrasting theepochs of frontal faces with those of household ob-jects (Figure 6A) Data from one control subject wereexcluded from this experiment as they were too noisyto localize face-related activation The activation profilewas extracted for the Rubin-face and Rubin-vase stimulifrom both the fusiform gyrus and lateral occipital foci

Figure 6 The Rubin facendashvase experiment (A) Examples of the

stimuli including the Rubin facendashvase stimuli (left) and frontal faces

and household objects (localizer stimuli) used to independently

localize face-related voxels (in the experiment the uniform surfacesshown here in gray were colored in light pink) (B) Activation profiles

for the Rubin facendashvase conditions in face-related voxels in the fusiform

gyrus and lateral occipital region Error bars indicate SEM acrosssubjects in each group

Avidan et al 1159

for each subject (see Table 2 Part 2 for foci) The av-eraged percent signal change across all repetitions ofeach condition and across the left and right hemi-spheres of each subject are presented in Figure 6BAs in the control group the average fusiform gyrusactivation in the CP subjects exhibited a selective re-sponse for the Rubin-face compared to Rubin-vaseindicating that this region in CP subjects is indeedinvolved in computing global information about facesand is sensitive to the context not simply to the localcomponents of the display The similarity between theactivation profiles of the two groups in the fusiformgyrus was confirmed by a repeated-measure ANOVAwhich revealed only a main effect of stimulus type( p lt 01)

The results in the lateral occipital region were onceagain more variable across subjects in both groups (seeTable 2 Part 2) The ANOVA revealed a marginallysignificant main effect of group ( p lt 06) but not ofstimulus type and no significant interaction Note thatalthough not statistically significant the selectivity forthe Rubin-face compared to the Rubin-vase conditionin the lateral occipital focus was somewhat reduced inthe CP subjects compared to their controls This resultis analogous to the somewhat reduced face selectivityfound in the lateral occipital region in the conventionalmapping experiment (Figure 2A)

DISCUSSION

In a series of four functional imaging experimentswe have shown that individuals with CP exhibit nor-mal face- and object-related fMRI activation patternsin the ventral occipito-temporal cortex particularly inthe face-related fusiform gyrus (FFA) and largely nor-mal activations in the lateral occipital region despitetheir profound behavioral impairment Importantly thispattern was evident across different types of stimuli(eg line drawings and movie clips) and across differ-ent paradigms

In the first study in a conventional mapping experi-ment a normal pattern of BOLD activation was foundwhen subjects viewed line drawings of face and nonfacestimuli and performed a one-back memory task Thisparadigm has been used extensively in the literature andserves as a rigorous baseline to demarcate areas of ac-tivation in the ventral occipito-temporal cortex Onepossible interpretation of the normal activation is thatthe task was not sufficiently taxing for the CP individ-uals Behavioral data collected concurrently with theimage acquisition rules out this possibilitymdashalthoughthe task was easy for the controls this was not so forthe CP individuals who were clearly impaired on thistask Thus the normal face-related activation found inCP cannot merely be explained by the claim that the taskwas too simplistic for the CP subjects

A second possible interpretation and one that isdiametrically opposed to the first one offered above isthat the task was excessively difficult for the CP and itis this difficulty that drove the face-related activationparticularly in the fusiform gyrus of the CP subjects In-deed previous studies have shown that activation in theFFA increases as a function of working memory load(Druzgal amp DrsquoEsposito 2001) This explanation is nottenable either as the CP subjects continued to exhibitnormal face-related activation even when they were notrequired to perform any behavioral task during themotion pictures experiment Further research is clearlyneeded to unequivocally determine the contribution oftask difficulty to face-related activation in CP individ-uals We are now starting to parametrically explore theeffects of task difficulty in these CP individuals usingnew paradigms The critical point at present is thatnormal face-related activation is obtained at the verysame time that the behavioral impairment is manifest

The finding that face- and object-related activation inthe ventral visual cortex is apparently normal is replicat-ed in the motion picture experiment As mentionedabove this replication attests to the generalizability ofthe finding from the first conventional mapping exper-iment to more natural stimuli and viewing conditions

Importantly the normal signature of face-related ac-tivation in CP was also found when additional experi-mental manipulations were used CP subjects showednormal recovery from adaptation when different faceswere shown suggesting that neurons within these re-gions were sufficiently sensitive to differentiate betweenthe face exemplars used in this experiment Althoughone might think that normal adaptation is not thatsurprising given that the face stimuli used in this exper-iment were very different from each other it is impor-tant to bear in mind that CP subjects were actuallyimpaired in discriminating between such stimulimdashinthe conventional mapping experiment similar stimuliwere used and the CP subjects were impaired suggest-ing that the stimuli were confusing and not as percep-tually dissimilar for them as one might assume Of noteis that in addition to the adaptation for faces found inthe present experiment CP subjects also exhibitednormal adaptation levels in the fusiform gyrus and otherface-related regions for nonface stimuli suggestingthat similar to control subjects and replicating previousfindings these regions are involved in processing otherobject categories in addition to faces (Avidan et al 2002Haxby Gobbini et al 2001)

Although normal adaptation for faces is apparentdetermining its origin is less obvious Because the exactsame picture of an individual face was repeated duringthe adaptation condition we cannot determine at thisstage whether the adaptation found in the fusiformgyrus was due to the repetition of the exact same shape(ie based on perceptual geometry or structure) or therepetition of the same facial identity Differentiating

1160 Journal of Cognitive Neuroscience Volume 17 Number 7

between these alternatives and establishing whetherneural activity in the FFA is sufficient for normal rep-resentation of fine distinctions among individual faceswill require further experiments in which more subtlevariations will be made to face stimuli The use of anevent-related adaptation paradigm in such future ex-periments will also minimize possible attentional varia-tions that are inherent in the block-design adaptationparadigm employed here

Finally in the last experiment similar to control sub-jects once again CP subjects also exhibited a selectiveBOLD signal for Rubin-face compared to Rubin-vasestimuli in the fusiform gyrus Importantly these stimulicontain similar local elements but induce two differentvisual percepts either of a face or of a vase dependingon figurendashground segmentation and global integrationand shape processing Hence selectivity for the Rubin-face over the Rubin-vase stimuli implies that as in thecontrol group CP subjectsrsquo face-related activation ismodulated by global grouping processes and is notsimply induced by the representation of local face parts(Hasson Hendler et al 2001)

In sum the CP subjects exhibit normal face- andobject-related activation patterns in multiple regions ofthe ventral occipito-temporal cortex under various ex-perimental paradigms Of note however is that groupdifferences between CP subjects and their controls werefound in regions outside the occipito-temporal cortexindicating that the absence of group differences is notmerely a lack of statistical power Taken together thesefindings suggest that normal face-related activation inventral occipito-temporal regions as measured withfMRI is not sufficient to ensure intact face processingand does not necessarily reflect normal face represen-tation within those regions

Can the BehaviorndashBOLD Dissociationbe Reconciled

The central result is that the BOLD activation in CP inventral occipito-temporal regions is largely not differen-tiable from that of the control subjects notwithstandingtheir marked behavioral impairment How might thisdissociation be explained

It has been previously suggested that the FFA ismainly involved in face detection that is discriminatingbetween faces and stimuli from other object categories(Tong et al 2000) an ability that is largely preserved inCP The apparently normal FFA activation in the CPsubjects might then reflect the intact involvement ofthe FFA in deriving a rather coarse and rudimentarystructural representation which suffices for some tasks(ie detection) but not for more taxing tasks such asidentification Evidence compatible with this possibleinterpretation suggests that the final stage of face iden-tification is not completed at the level of high-order

visual face-related regions such as the FFA rathersuccessful recognition requires activation in more ante-rior regions of the temporal lobe where a more abstractor semantic representation of faces is derived (HaxbyHoffman et al 2000 Leveroni et al 2000 SergentOhta amp Macdonald 1992) Further support for the roleof anterior temporal regions in mediating face identifi-cation and particularly for being a site of facial memory(long-term representation) comes from lesion studiesFor example Barton and Cherkasova (2003) foundthat patients with anterior temporal lesions were themost impaired in a task that required generating long-term representations of famous faces as in mentalimagery compared to patients with more posterioroccipito-temporal lesions Similarly patients with ante-rior temporal lobectomy (following intractable epilepsy)were found to be impaired in identifying familiar faces(Glosser Salvucci amp Chiaravalloti 2003) The source ofthe behavioral problem in CP might then arise in theabnormal propagation of activation from the intact FFAto more anterior temporal regions

However the idea that the FFA only mediates facedetection is not universally accepted and the cleandivision of labor between the FFA and more anteriorregions might not be totally correct Although notdenying the contribution of more anterior regions toface processing and particularly to face identificationrecent experimental evidence suggests that the FFA isinvolved not only in face detection but also in faceidentification Thus normal subjects exhibited differen-tial responses for face detection and identification in thefusiform gyrus with the two processes resulting in eitherdifferential fMRI amplitude (Grill-Spector et al 2004) ortemporal scales (Puce Allison et al 1999 and seeSugase Yamane Ueno amp Kawano 1999 for similarfindings from single-unit recordings in monkeys) Thesedifferential responses in the fusiform gyrus may poten-tially be the outcome of feedback from more anteriorregions (as above) andor from additional computationtaking place within the fusiform gyrus itself Which ofthese is perturbed in CP remains unclear and they arenot mutually exclusive either Propagation to or fromanterior regions may be affected in CP However recallthat CP subjects exhibit a clear deficit not only in faceidentification (naming of famous faces) but also indiscrimination of unfamiliar faces (Figure 1B) (see alsoBehrmann et al 2005) This deficit in the more per-ceptual (not just memorial) aspects of face processingimplicates posterior occipito-temporal regions ratherthan exclusively the more anterior areas (Barton PressKeenan amp OrsquoConnor 2002 Wada amp Yamamoto 2001)It remains a possibility therefore that the fusiformactivation itself may be abnormal in CP Because thecurrent study employed detection and discriminationstasks the abnormality may still be uncovered undermore challenging conditions such as face identificationor other fine-grained face tasks

Avidan et al 1161

We also cannot rule out the possibility that differ-ences in the activation patterns in the fusiform gyrusandor other parts of the face processing networkbetween CP and normal control subjects might exist inthe temporal domain Such differences could underliethe behavioral dysfunction of CP subjects but may notbe evident with the temporal resolution afforded byfMRI (in the order of seconds) Some evidence sup-porting this view comes from ERP studies showing thatthe relatively early N170 potential which shows faceselectivity in normal subjects failed to show such se-lectivity in CP subjects (Kress amp Daum 2003b BentinDeouell et al 1999)

Finally it might be that the small differences inBOLD activation in CP subjects such as the slight re-duction in face selectivity in the lateral occipital region(eg Figures 2A and 6B) might be lsquolsquoamplifiedrsquorsquo whentranslated into behavioral performance If so the lateraloccipital region would serve as a critical link in the faceprocessing network Further support for the role ofthe occipital region in face processing comes fromstudies showing that acquired prosopagnosic patientsoften have lesions that involve not only the fusiformgyrus but also the lateral occipital region (Barton et al2002 Wada amp Yamamoto 2001) Moreover brain dam-age to this region but not to the FFA in some individ-uals may be associated with severe prosopagnosia(Rossion Caldara et al 2003)

As is evident there are a host of potential expla-nations to account for the dissociation between the im-paired behavioral profile and the apparently normalBOLD activation pattern We are currently exploringthese issues using new imaging experiments and hopethat these will uncover in greater detail the mecha-nisms giving rise to CP Suffice it to say that no one ex-planation captures the BOLDndashbehavior dissociationWhat is crucial however is that a simple assignmentof face processing to the FFA is no longer defensibleand a deeper examination of the computational prop-erties of the FFA and associated regions is demanded bythese findings

Face-related Activation Outside theOccipito-temporal Cortex

In the current study we acquired whole-brain functionalscans and thus were able to explore differences be-tween the CP and control groups in the responsepatterns in all regions of the brain not only in theoccipito-temporal regions A particularly novel findingwas that robust bilateral face-related activation wasevident in prefrontal regions (precentral sulcus inferiorfrontal sulcus anterior lateral sulcus) in the CP groupIn contrast the control group only exhibited right-lateralized activation in the precentral sulcus (Figure 3)Studies of normal individuals performing workingmemory tasks have revealed face-related activation in

prefrontal regions (Druzgal amp DrsquoEsposito 2003 Salaet al 2003 Haxby Petit et al 2000) and some ERPstudies have detected some inferior prefrontal cortexsites that generate small but specific face-specific re-sponses (Allison Puce Spencer amp McCarthy 1999)

The poor performance exhibited by CP subjectscompared to their controls on the one-back memorytask specifically for face stimuli is consistent with theinterpretation of this prefrontal activity being related toworking memory However given that the conven-tional face and object mapping experiment was notspecifically designed to look at the different componentsthat usually comprise a memory study (encoding mem-ory delay response) further research that specificallyand parametrically manipulates the memory load offaces and other object stimuli is required in order todetermine the role of this activation in face processing inCP subjects

This study has exploited the unique possibility ofexamining the activation of the fusiform gyrus andother cortical regions in an unusual group of individ-uals with CP Delineating the neural mechanism thatgives rise to abnormal behavior and by logical infer-ence normal behavior is a critical goal of neurosci-ence As might be predicted the correspondencesbetween the brain and behavior are not transparentand in this study are dissociated with the behavioralimpairment being evident concomitant with normalFFA activation Taken together however this studyhas indicated patterns of similarities and patterns ofpotential difference in this population of CP individ-uals relative to their normal counterparts The challengethat lies before us is to use these patterns to explainthe face processing deficit in detail

METHODS

Subjects

Four healthy CP subjects (2 men) aged between 29 and60 years with no discernable cortical lesion or anyhistory of neurological disease participated in all ex-periments (for details see Table 1) Another CP sub-

Table 1 Biographical Information about CP Subjects

Congenital Prosopagnosia (CP) Subjects

Subjectsrsquo Initials Sex Age

TM M 27

KM F 60

MT M 41

BE F 29

KM and TM are a mother and a son

1162 Journal of Cognitive Neuroscience Volume 17 Number 7

ject (NI male aged 40) was tested behaviorally (seeBehrmann et al 2005) but because we could not ac-quire his imaging data he was excluded from thepresent article Seven of the 12 control subjects whoparticipated in the behavioral studies along with threenew participants completed the imaging experimentsThe control group included at least two age- and sex-matched controls for each CP individual All CP andcontrol subjects were right-handed and had normal orcorrected-to-normal vision All subjects consented to par-ticipate in the experiments and the protocol was ap-proved by the Institutional Review Boards of CarnegieMellon University and the University of Pittsburgh

Imaging Experiments

Visual Stimulation

Visual stimuli were generated using the E-prime IFISsoftware (Psychology Software Tools Pittsburgh PAUSA) and projected via LCD to a screen located in theback of the scanner bore behind the subjectrsquos head Sub-jects viewed the stimuli through a tilted mirror mountedabove their eyes on the head coil Before the scan sub-

jects were familiarized with the visual stimuli and tasksof each of the imaging experiments

MRI Setup

All subjects were scanned in a 3-T Siemens Allegrascanner equipped with a standard head coil Subjectsparticipated in one scanning session that lasted 15 to2 hr in which all the functional and anatomical scanswere acquired The order of the functional experimentswithin the scanning session was generally constant forall subjects BOLD contrast was acquired using gradient-echo echo-planar imaging sequence Specific parame-ters TR = 3000 msec TE = 35 msec flip angle = 908FOV = 210 210 mm2 matrix size 64 64 35 axialslices 3 mm thickness no gap High-resolution anatom-ical scans (T1-weighted 3-D MPRAGE) were also acquiredto allow accurate cortical segmentation reconstructionand volume-based statistical analysis Specific parame-ters TE = 349 flip angle = 88 FOV = 256 256 mm2matrix size = 256 256 slice thickness = 1 mm numberof slices = 160ndash192 orientation of slices was eitherhorizontal or sagittal For three of the CP subjects the

Table 2 Laterality in force-selective activation

1 Talairach Coordinates Fusiform Gyrus

Left Hemisphere Right Hemisphere

x y z x y z

CP subjects 40 plusmn 2 53 plusmn 3 19 plusmn 2 38 plusmn 5 47 plusmn 8 17 plusmn 3

Control subjects 37 plusmn 4 41 plusmn 5 18 plusmn 4 36 plusmn 2 40 plusmn 7 17 plusmn 3

2 Activation Foci

Localizer Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 4 ndash ndash

Control subjects 9 1 ndash 4 2 1

Rubin FacendashVase Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 1 3 ndash

Control subjects 9 ndash ndash 3 2 4

Part 1 Talairach coordinates of the face-related activation in the fusiform gyrus of CP and control subjects extracted from the conventionalface and object mapping experiment Part 2 Number of CP and control subjects that showed bilateral right-lateralized or left-lateralizedface-related activation in the fusiform gyrus and in the lateral occipital region in the conventional face and object mapping experiment andin the Rubin facendashvase experiment

Avidan et al 1163

3-D brain reconstruction was performed on imagespreviously acquired on a 15-T GE scanner

Experiments

Conventional Face and Object Mapping Experiment

Stimuli included line drawings of faces buildings com-mon objects and geometric patterns presented in ashort block design (for details see Levy et al 2001Figure 1A) Each epoch lasted 9 sec and contained ninestimuli from the same category there were eight differ-ent images and one immediate repetition of one of thestimuli Each stimulus was presented for 800 msecfollowed by a 200-msec interval of blank screen Therewere seven repetitions of each condition and their orderwas pseudorandomized across subjects Visual epochswere interleaved with 6-sec blank epochs The experi-ment lasted 450 sec and started and ended with extend-ed blank epochs of 27 and 9 sec respectively Subjectswere instructed to fixate on a red central fixation dot toperform a one-back memory task and to press a key ona response glove to indicate the repeated image Bothaccuracy and reaction time were recorded

The Motion Pictures Experiment

The experiment lasted 10 minutes and was composed of32 epochs (15 sec long) of movie clips each containingimages from one of four possible categories people invarious situations navigation of the camera through citybuildings or through open fields and miscellaneousimages of objects from different categories (machinesfalling water etc for details see Hasson Nir et al2004 Figure 4) The experiment started with a 51-secblank period followed by 9 sec of pattern stimuli andended with a 1-minute blank Subjects were simplyinstructed to watch the movie clips

The Adaptation Experiment

The experiment included line drawing images fromthree different categories faces buildings and cars(for details see Avidan et al 2002 Figure 5A) Stimuliwere presented in 12-sec epochs that contained either12 different images (lsquolsquodifferentrsquorsquo condition) from thesame category or 12 repetitions of the same identicalstimulus (lsquolsquoidenticalrsquorsquo condition) Within epochs eachstimulus was presented for 800 msec followed by200 msec of fixation only The experiment lasted480 sec and started and ended with a long blank pe-riod of 21 and 15 sec respectively The first visualblock always contained geometric patterns and was ex-cluded from the analysis Visual epochs were interleavedwith 6-sec blank epochs and the order of visual epochswas pseudorandomized across subjects Subjects wereinstructed to continuously fixate a central red dot and

to covertly categorize the images as follows man wom-an or a child for the face stimuli apartment townhouseor public facility for the buildings and private car truckor bus for the vehicles

The Rubin FacendashVase Experiment

The experiment included modified versions of the Rubinfacendashvase stimuli as well as line drawings of front facesand household objects (for details see Hasson Hendleret al 2001 Figure 6A) The Rubin facendashvase stimuli weremodified so that subjectsrsquo perception was biased towardsone perceptual state or the other This was done by firstuniformly coloring the vase area within each of thestimuli and placing stripes over the profile Stimuli werethen placed either over a striped background (biastowards perception of the vase due to closure of thevase stimulus) or over a uniform-color background (biastowards perception of the face due to closure of theface stimulus) Critically the Rubin face and vase sharethe same local contours but give rise to different globalperceptions In order to ensure that subjectsrsquo perceptionremained biased to one perceptual interpretation (ei-ther the profile or vase) during the experiment stimuliwere presented briefly (200 msec) and were immediate-ly followed by a masking grid that was presented for800 msec Stimuli were presented in 9-sec blocks thatwere interleaved with 6-sec blanks Each condition wasrepeated eight times and the order of presentation waspseudorandomized across subjects Subjects maintainedcentral fixation throughout the experiment and per-formed a one-back memory task however due to atechnical failure their responses were not recorded Theexperiment lasted 507 sec and started and ended withextended blank periods

Data Analysis

Data analysis was performed using the BrainVoyager48 software package (Brain Innovation MaastrichtNetherlands) and complementary in-house softwareDetailed description of the data analysis is provided inHasson Harel et al (2003) Briefly for each subject thecortical surface was reconstructed and unfolded intothe flattened format Functional data for each subjectfrom each experiment were analyzed separately Pre-processing of functional scans included 3-D motioncorrection and filtering of low frequencies up to fivecycles per experiment (slow drift) Statistical analysiswas based on the GLM For each experiment each ofthe conditions (except for blank) was defined as a sepa-rate predictor a boxcar shape was used to model eachpredictor and a 3-sec lag was assumed Percent signalchange for each subject in each experiment (except forthe motion pictures experiment) was calculated as thepercent activation from a blank baseline For the mo-

1164 Journal of Cognitive Neuroscience Volume 17 Number 7

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

perform working memory tasks using face stimuli(Druzgal amp DrsquoEsposito 2003 Sala Rama amp Courtney2003 Haxby Petit Ungerleider amp Courtney 2000) Theprefrontal face-related activation in the current studymight then reflect the increased difficulty for the CPsubjects in the one-back memory task for faces relativeto the other conditions and to the control subjects andthe consequent recruitment of prefrontal regions tomaintain face representations (see Figure 1B)

The results from the conventional face and objectmapping experiment indicate that individuals with CP re-veal normal face-related activation in the ventral occipito-temporal cortex and close-to-normal selectivity in thelateral occipital region in terms of the anatomical loca-tion of face-selective activation and its spatial extentthe signal strength and the stimulus selective responseCrucially this normal profile of activation is apparentdespite the significant face processing impairment ex-hibited at the very same time Of note also is that theCP subjects also exhibited normal object-related activa-tion (common objects and buildings) in the occipito-temporal cortex That group differences between theCP subjects and their controls can be detected (in pre-frontal regions) attests to the sufficiency of the statisticalpower of the current paradigm to detect differenceswhen they exist and therefore lends further weightto the conclusion that there are no detectable groupdifferences in the posterior areas of cortex

These results reveal a dissociation of a BOLDndashbehav-ior correlation in the ventral occipito-temporal cortexand particularly in the FFA To further confirm thisresult we utilized additional paradigms that could as-

sist in exploring the mechanisms giving rise to the ab-normal behavior

Motion Pictures Experiment

Although serving as a powerful tool for studying objectselectivity in the human visual cortex the conventionalface and object mapping experiment exploits viewingconditions and stimuli that are rather contrived Fur-thermore as evident from Figure 1B the one-backmemory task performed during this experiment wassignificantly more difficult for the CP subjects com-pared to their controls particularly for the face stimuliIt has previously been shown that activity in the FFAcan increase as a function of working memory load intasks involving faces (Druzgal amp DrsquoEsposito 2001)Thus the seemingly normal activation found in theconventional mapping experiment in the CP subjectscould actually reflect additional computation takingplace due to the relative difficulty of the behavioraltask for these individuals

To address these issues in this next experiment weused a motion picture paradigm first to test whetherthe face and object-related activation similarities ob-served in the conventional mapping experiment arereplicable with more natural stimuli and under morenatural viewing conditions and second to determinewhether normal face-related activation can be foundregardless of the behavioral task performed in thescanner In this experiment subjects freely viewed asequence of short video clips each containing stimuli

Figure 3 Activation maps in

prefrontal regions in the

conventional face and object

mapping experiment Sameactivation maps as in

Figure 1C but here the brain is

shown in a lateral view in orderto reveal the activation in

prefrontal regions CP subjects

are shown in the upper panel

and controls in the lower oneAbbreviations LS = lateral

sulcus CS = central sulcus

PreCS = precentral sulcus

IFS = inferior frontal sulcus

Avidan et al 1155

of faces and people buildings navigation or miscella-neous objects (Hasson Nir Levy Fuhrmann amp Malach2004) Importantly no specific task was required mak-ing this experiment more akin to naturally surveyingonersquos visual environment Figure 4A shows the activa-tion maps of a single representative CP and controlsubject projected on an inflated and unfolded brainrepresentation The white lines on the posterior partof the unfolded representation show the retinotopicborders mapped for each subject The face and build-ing activation maps obtained in this experiment largelyreplicate the results from the conventional mappingexperiment (compare Figure 4A and Figure 1C) ex-cept that the activation seems to be more widespreadin the motion pictures experiment selective activa-tion for faces was found in the fusiform gyrus andthe lateral occipital region when contrasting faces withbuildings and navigation scenes (red patches) andselective activation for buildings and navigation sceneswere found in the collateral sulcus when applyingthe opposite contrast (green patches) Interestinglyface-related activation was also found within the STSin regions typically activated in response to facialmovements (Hoffman amp Haxby 2000 Puce Allisonet al 1998) and biological motion (Puce and Perrett2003) That these regions are activated here is con-sistent with the richness of the faces and motions inthe displays

To assess the spatial extent of the face-related acti-vation obtained in this experiment in both CP andcontrols the number of face-related voxels within thelateral occipital region fusiform gyrus and STS wascounted for each subject in each group To allow di-rect comparisons all face-related foci for all subjectswere selected using the same statistical threshold ( p lt0005) (see Figure 4A for two representative subjects)A repeated-measures ANOVA with group (CPcontrols)hemisphere (left right) and face-related region as therepeated-measures within-subject factors and the num-ber of voxels as the dependent measure was usedThere were infrequent instances of control subjects forwhom not all face-related foci could be identified in theselected statistical threshold and these were excludedfrom the ANOVA Critically all face-related regions weredetected for the CP subjects using that threshold Thisanalysis revealed a significant main effect of hemisphere( p lt 001) confirming that overall there were moreface-related voxels in the right compared with the lefthemisphere In addition there was a significant maineffect of face-related region ( p lt 005) with moreactivated voxels in the lateral occipital region than inthe fusiform gyrus or STS Importantly however thisanalysis did not reveal a main effect of group or any two-or three-way interactions thus suggesting that therewere no differences in the spatial extent of the activationbetween the CP and control group in any of the face-related regions

To further quantify the similarity between the CP andcontrol groups we sampled the activation profile ofthree regions of interest (ROIs) (fusiform gyrus lateraloccipital region and collateral sulcus) defined indepen-dently for each subject on the basis of the conventionalface and object mapping experiment (see Methods)Percent signal change averaged across the entire exper-iment and collapsed across the left and right hemi-spheres for each group is shown in Figure 4B Asabove all ROIs exhibited clear stimulus selectivity butof even greater importance is the striking similaritybetween the activation profiles of the two groups (CP =black line controls = white line) This similarity is alsoevident from the high values of the correlation coeffi-cient (fusiform gyrus = 80 lateral occipital region =89 CoS = 81) calculated for each ROI between theaverage activation profile of each group Importantlyhowever as evident from the figure this similarity isnot only in the overall correlation but also in themoment-by-moment waxing and waning of the activityThese results confirm the normal profile of face- andbuilding-related activation in the CP subjects when morenatural viewing conditions and stimuli are employedMoreover because this normal activation was foundhere in the absence of any behavioral task it impliesthat the normal activation found in the conventionalmapping experiment was not confounded by the diffi-culty of the behavioral task

Adaptation Experiment

The behavioral results reveal a clear impairment inCP individuals in unfamiliar face discrimination in asimultaneous matching task (Behrmann et al 2005)and in a one-back memory (sequential discrimination)task (Figure 1B) These tasks are largely perceptualand do not require that a specific or individual facebe represented Indeed CP subjects might be evenmore impaired than already evident when more preciseface knowledge is required their own self-testimoniessuggest that lsquolsquoall faces look alikersquorsquo Using the fMR-adaptation paradigm for studying the neuronal proper-ties of higher-order visual areas at a subvoxel resolutionit has been shown that normal subjects typically exhibita reduction in BOLD signal with repeated presentationof a stimulus (Grill-Spector amp Malach 2001) To exam-ine whether the face-selective activation noted above isless sensitive to face repetition in CP than in controlsubjects we compared adaptation levels for face repeti-tion across the two groups We also examined the adap-tation effect for nonface stimuli to determine whetherthe entire occipito-temporal region is normally sensitiveto repetition (Avidan et al 2002)

Stimuli were presented in epochs that included either12 different stimuli (faces building cars) or 12 repeti-tions of the same identical stimulus (see Figure 5A) Theactivation profile was sampled from ROIs defined in-

1156 Journal of Cognitive Neuroscience Volume 17 Number 7

Figure 4 Activation maps and activation profiles from the motion pictures experiment (A) Activation maps for faces (red) as well as buildings

and navigation (green) are shown for one representative CP subject and one control subject The data are shown on the inf lated and unfoldedmap representation of each individual using the same statistical threshold (GLM statistics p lt 0005) The white dotted lines on the unfolded maps

indicate the borders of the retinotopic areas that were mapped for each subject in a separate experiment Abbreviations LOS = lateral occipital

sulcus IOG = inferior occipital gyrus IPS = intraparietal sulcus PCS = postcentral sulcus Other abbreviations as in Figures 1 and 3 (B) Averaged

activation profiles of the CP (black) and control subjects (white) are shown for face-related voxels in the fusiform gyrus and lateral occipital regionand for building-related voxels in the collateral sulcus The signal is shown along the time-axis of the experiment and the vertical colored bars

indicate the different experimental conditions The correlation coefficient values between the entire activation profile of CP and control subjects

are shown for each ROI Error bars indicate SEM across subjects in each group

Avidan et al 1157

Figure 5 Experimental design and activation profiles of the adaptation experiment (A) Examples of stimuli and experimental design Face

building and car line drawings were presented in epochs containing either 12 different or 12 identical images (B) Averaged activation profiles inface-related voxels in the fusiform gyrus (top) and the lateral occipital region (bottom) for the CP and control subjects Error bars indicate SEM

across subjects in each group (C) Averaged activation profiles in the building-related voxels in the collateral sulcus

1158 Journal of Cognitive Neuroscience Volume 17 Number 7

dependently for each subject using the conventionalmapping experiment Figure 5B shows the activationlevels within the face-related foci in the fusiform gyrusand lateral occipital region averaged across all repeti-tions of each experimental condition and across thetwo hemispheres for the two groups Again despitethe fact that the face-related regions were selected usingan external localizer CP individuals exhibited clear faceselectivity in the lsquolsquodifferentrsquorsquo conditions in these ROIsNote the similarity between the groups in the magni-tude of the adaptation that is the signal decrease fol-lowing face repetition (dark gray compared with lightgray bars) suggesting that face representations in CPsubjects are indeed sensitive to face repetition Theface-related regions also exhibit a strong adaptationeffect for the nonface stimuli (buildings and cars) in-dicating that like normal subjects these regions in theCP subjects contain neurons that are highly selectivefor those stimuli (Avidan et al 2002) These observa-tions were confirmed in a repeated-measure ANOVAin which the only significant effect in both regions wasa main effect of stimulus type ( p lt 0001) There wasno main effect of group nor an interaction betweenthe factors

Figure 5C shows a similar analysis for the building-related focus in the anterior collateral sulcus This regionexhibited strong adaptation for the building stimuli aswell as for the other object categories in both groupsOnce again the ANOVA revealed only a main effect ofstimulus type ( p lt 0001) The strong building selectiv-ity in the collateral sulcus combined with the findingthat this region exhibits similar adaptation levels forall stimulus categories including faces further suggeststhat the entire occipito-temporal object representationnetwork in CP subjects is intact and mirrors that of thenormal control subjects

Rubin FacendashVase Experiment

One possible interpretation of the apparently normalface-selective BOLD response in the CP subjects foundin the previous experiments is that it reflects a responseto the presence of face parts (eyes mouth and nose) inthe absence of an intact representation of the face as awhole To examine whether the activation in CP ismerely a product of a part-representation we employeda paradigm used previously which utilizes the famousRubin facendashvase illusion to demonstrate evidence forglobal face representations in normal subjects (HassonHendler Ben Bashat amp Malach 2001) Critically in theRubin facendashvase illusion similar local elements are al-ways present However depending on figurendashgroundsegmentation and global integration the local elementscan induce two different visual percepts either of a faceor of a vase (Figure 6A) If face-selective activation in thefusiform gyrus and other face-related regions depends

only on processing of the local elements then activa-tion in these regions should be similar in both theRubin-face and Rubin-vase conditions (same local ele-ments are present) However if face representation inthe fusiform gyrus and other face-related regions isaffected by the global integration of the face elementsinto the percept of either a face or a vase then increasedactivation should be found for the Rubin-face perceptualstates compared to the Rubin-vase perceptual states In-deed in normal subjects selective activation within face-related regions has been previously demonstrated forthe Rubin-face compared to the Rubin-vase stimuli sug-gesting that normal face-related activation is modulatedby global grouping processes and is not only induced bythe representation of local face parts (Hasson Hendleret al 2001)

Here face-selective regions were independently lo-calized initially for each subject by contrasting theepochs of frontal faces with those of household ob-jects (Figure 6A) Data from one control subject wereexcluded from this experiment as they were too noisyto localize face-related activation The activation profilewas extracted for the Rubin-face and Rubin-vase stimulifrom both the fusiform gyrus and lateral occipital foci

Figure 6 The Rubin facendashvase experiment (A) Examples of the

stimuli including the Rubin facendashvase stimuli (left) and frontal faces

and household objects (localizer stimuli) used to independently

localize face-related voxels (in the experiment the uniform surfacesshown here in gray were colored in light pink) (B) Activation profiles

for the Rubin facendashvase conditions in face-related voxels in the fusiform

gyrus and lateral occipital region Error bars indicate SEM acrosssubjects in each group

Avidan et al 1159

for each subject (see Table 2 Part 2 for foci) The av-eraged percent signal change across all repetitions ofeach condition and across the left and right hemi-spheres of each subject are presented in Figure 6BAs in the control group the average fusiform gyrusactivation in the CP subjects exhibited a selective re-sponse for the Rubin-face compared to Rubin-vaseindicating that this region in CP subjects is indeedinvolved in computing global information about facesand is sensitive to the context not simply to the localcomponents of the display The similarity between theactivation profiles of the two groups in the fusiformgyrus was confirmed by a repeated-measure ANOVAwhich revealed only a main effect of stimulus type( p lt 01)

The results in the lateral occipital region were onceagain more variable across subjects in both groups (seeTable 2 Part 2) The ANOVA revealed a marginallysignificant main effect of group ( p lt 06) but not ofstimulus type and no significant interaction Note thatalthough not statistically significant the selectivity forthe Rubin-face compared to the Rubin-vase conditionin the lateral occipital focus was somewhat reduced inthe CP subjects compared to their controls This resultis analogous to the somewhat reduced face selectivityfound in the lateral occipital region in the conventionalmapping experiment (Figure 2A)

DISCUSSION

In a series of four functional imaging experimentswe have shown that individuals with CP exhibit nor-mal face- and object-related fMRI activation patternsin the ventral occipito-temporal cortex particularly inthe face-related fusiform gyrus (FFA) and largely nor-mal activations in the lateral occipital region despitetheir profound behavioral impairment Importantly thispattern was evident across different types of stimuli(eg line drawings and movie clips) and across differ-ent paradigms

In the first study in a conventional mapping experi-ment a normal pattern of BOLD activation was foundwhen subjects viewed line drawings of face and nonfacestimuli and performed a one-back memory task Thisparadigm has been used extensively in the literature andserves as a rigorous baseline to demarcate areas of ac-tivation in the ventral occipito-temporal cortex Onepossible interpretation of the normal activation is thatthe task was not sufficiently taxing for the CP individ-uals Behavioral data collected concurrently with theimage acquisition rules out this possibilitymdashalthoughthe task was easy for the controls this was not so forthe CP individuals who were clearly impaired on thistask Thus the normal face-related activation found inCP cannot merely be explained by the claim that the taskwas too simplistic for the CP subjects

A second possible interpretation and one that isdiametrically opposed to the first one offered above isthat the task was excessively difficult for the CP and itis this difficulty that drove the face-related activationparticularly in the fusiform gyrus of the CP subjects In-deed previous studies have shown that activation in theFFA increases as a function of working memory load(Druzgal amp DrsquoEsposito 2001) This explanation is nottenable either as the CP subjects continued to exhibitnormal face-related activation even when they were notrequired to perform any behavioral task during themotion pictures experiment Further research is clearlyneeded to unequivocally determine the contribution oftask difficulty to face-related activation in CP individ-uals We are now starting to parametrically explore theeffects of task difficulty in these CP individuals usingnew paradigms The critical point at present is thatnormal face-related activation is obtained at the verysame time that the behavioral impairment is manifest

The finding that face- and object-related activation inthe ventral visual cortex is apparently normal is replicat-ed in the motion picture experiment As mentionedabove this replication attests to the generalizability ofthe finding from the first conventional mapping exper-iment to more natural stimuli and viewing conditions

Importantly the normal signature of face-related ac-tivation in CP was also found when additional experi-mental manipulations were used CP subjects showednormal recovery from adaptation when different faceswere shown suggesting that neurons within these re-gions were sufficiently sensitive to differentiate betweenthe face exemplars used in this experiment Althoughone might think that normal adaptation is not thatsurprising given that the face stimuli used in this exper-iment were very different from each other it is impor-tant to bear in mind that CP subjects were actuallyimpaired in discriminating between such stimulimdashinthe conventional mapping experiment similar stimuliwere used and the CP subjects were impaired suggest-ing that the stimuli were confusing and not as percep-tually dissimilar for them as one might assume Of noteis that in addition to the adaptation for faces found inthe present experiment CP subjects also exhibitednormal adaptation levels in the fusiform gyrus and otherface-related regions for nonface stimuli suggestingthat similar to control subjects and replicating previousfindings these regions are involved in processing otherobject categories in addition to faces (Avidan et al 2002Haxby Gobbini et al 2001)

Although normal adaptation for faces is apparentdetermining its origin is less obvious Because the exactsame picture of an individual face was repeated duringthe adaptation condition we cannot determine at thisstage whether the adaptation found in the fusiformgyrus was due to the repetition of the exact same shape(ie based on perceptual geometry or structure) or therepetition of the same facial identity Differentiating

1160 Journal of Cognitive Neuroscience Volume 17 Number 7

between these alternatives and establishing whetherneural activity in the FFA is sufficient for normal rep-resentation of fine distinctions among individual faceswill require further experiments in which more subtlevariations will be made to face stimuli The use of anevent-related adaptation paradigm in such future ex-periments will also minimize possible attentional varia-tions that are inherent in the block-design adaptationparadigm employed here

Finally in the last experiment similar to control sub-jects once again CP subjects also exhibited a selectiveBOLD signal for Rubin-face compared to Rubin-vasestimuli in the fusiform gyrus Importantly these stimulicontain similar local elements but induce two differentvisual percepts either of a face or of a vase dependingon figurendashground segmentation and global integrationand shape processing Hence selectivity for the Rubin-face over the Rubin-vase stimuli implies that as in thecontrol group CP subjectsrsquo face-related activation ismodulated by global grouping processes and is notsimply induced by the representation of local face parts(Hasson Hendler et al 2001)

In sum the CP subjects exhibit normal face- andobject-related activation patterns in multiple regions ofthe ventral occipito-temporal cortex under various ex-perimental paradigms Of note however is that groupdifferences between CP subjects and their controls werefound in regions outside the occipito-temporal cortexindicating that the absence of group differences is notmerely a lack of statistical power Taken together thesefindings suggest that normal face-related activation inventral occipito-temporal regions as measured withfMRI is not sufficient to ensure intact face processingand does not necessarily reflect normal face represen-tation within those regions

Can the BehaviorndashBOLD Dissociationbe Reconciled

The central result is that the BOLD activation in CP inventral occipito-temporal regions is largely not differen-tiable from that of the control subjects notwithstandingtheir marked behavioral impairment How might thisdissociation be explained

It has been previously suggested that the FFA ismainly involved in face detection that is discriminatingbetween faces and stimuli from other object categories(Tong et al 2000) an ability that is largely preserved inCP The apparently normal FFA activation in the CPsubjects might then reflect the intact involvement ofthe FFA in deriving a rather coarse and rudimentarystructural representation which suffices for some tasks(ie detection) but not for more taxing tasks such asidentification Evidence compatible with this possibleinterpretation suggests that the final stage of face iden-tification is not completed at the level of high-order

visual face-related regions such as the FFA rathersuccessful recognition requires activation in more ante-rior regions of the temporal lobe where a more abstractor semantic representation of faces is derived (HaxbyHoffman et al 2000 Leveroni et al 2000 SergentOhta amp Macdonald 1992) Further support for the roleof anterior temporal regions in mediating face identifi-cation and particularly for being a site of facial memory(long-term representation) comes from lesion studiesFor example Barton and Cherkasova (2003) foundthat patients with anterior temporal lesions were themost impaired in a task that required generating long-term representations of famous faces as in mentalimagery compared to patients with more posterioroccipito-temporal lesions Similarly patients with ante-rior temporal lobectomy (following intractable epilepsy)were found to be impaired in identifying familiar faces(Glosser Salvucci amp Chiaravalloti 2003) The source ofthe behavioral problem in CP might then arise in theabnormal propagation of activation from the intact FFAto more anterior temporal regions

However the idea that the FFA only mediates facedetection is not universally accepted and the cleandivision of labor between the FFA and more anteriorregions might not be totally correct Although notdenying the contribution of more anterior regions toface processing and particularly to face identificationrecent experimental evidence suggests that the FFA isinvolved not only in face detection but also in faceidentification Thus normal subjects exhibited differen-tial responses for face detection and identification in thefusiform gyrus with the two processes resulting in eitherdifferential fMRI amplitude (Grill-Spector et al 2004) ortemporal scales (Puce Allison et al 1999 and seeSugase Yamane Ueno amp Kawano 1999 for similarfindings from single-unit recordings in monkeys) Thesedifferential responses in the fusiform gyrus may poten-tially be the outcome of feedback from more anteriorregions (as above) andor from additional computationtaking place within the fusiform gyrus itself Which ofthese is perturbed in CP remains unclear and they arenot mutually exclusive either Propagation to or fromanterior regions may be affected in CP However recallthat CP subjects exhibit a clear deficit not only in faceidentification (naming of famous faces) but also indiscrimination of unfamiliar faces (Figure 1B) (see alsoBehrmann et al 2005) This deficit in the more per-ceptual (not just memorial) aspects of face processingimplicates posterior occipito-temporal regions ratherthan exclusively the more anterior areas (Barton PressKeenan amp OrsquoConnor 2002 Wada amp Yamamoto 2001)It remains a possibility therefore that the fusiformactivation itself may be abnormal in CP Because thecurrent study employed detection and discriminationstasks the abnormality may still be uncovered undermore challenging conditions such as face identificationor other fine-grained face tasks

Avidan et al 1161

We also cannot rule out the possibility that differ-ences in the activation patterns in the fusiform gyrusandor other parts of the face processing networkbetween CP and normal control subjects might exist inthe temporal domain Such differences could underliethe behavioral dysfunction of CP subjects but may notbe evident with the temporal resolution afforded byfMRI (in the order of seconds) Some evidence sup-porting this view comes from ERP studies showing thatthe relatively early N170 potential which shows faceselectivity in normal subjects failed to show such se-lectivity in CP subjects (Kress amp Daum 2003b BentinDeouell et al 1999)

Finally it might be that the small differences inBOLD activation in CP subjects such as the slight re-duction in face selectivity in the lateral occipital region(eg Figures 2A and 6B) might be lsquolsquoamplifiedrsquorsquo whentranslated into behavioral performance If so the lateraloccipital region would serve as a critical link in the faceprocessing network Further support for the role ofthe occipital region in face processing comes fromstudies showing that acquired prosopagnosic patientsoften have lesions that involve not only the fusiformgyrus but also the lateral occipital region (Barton et al2002 Wada amp Yamamoto 2001) Moreover brain dam-age to this region but not to the FFA in some individ-uals may be associated with severe prosopagnosia(Rossion Caldara et al 2003)

As is evident there are a host of potential expla-nations to account for the dissociation between the im-paired behavioral profile and the apparently normalBOLD activation pattern We are currently exploringthese issues using new imaging experiments and hopethat these will uncover in greater detail the mecha-nisms giving rise to CP Suffice it to say that no one ex-planation captures the BOLDndashbehavior dissociationWhat is crucial however is that a simple assignmentof face processing to the FFA is no longer defensibleand a deeper examination of the computational prop-erties of the FFA and associated regions is demanded bythese findings

Face-related Activation Outside theOccipito-temporal Cortex

In the current study we acquired whole-brain functionalscans and thus were able to explore differences be-tween the CP and control groups in the responsepatterns in all regions of the brain not only in theoccipito-temporal regions A particularly novel findingwas that robust bilateral face-related activation wasevident in prefrontal regions (precentral sulcus inferiorfrontal sulcus anterior lateral sulcus) in the CP groupIn contrast the control group only exhibited right-lateralized activation in the precentral sulcus (Figure 3)Studies of normal individuals performing workingmemory tasks have revealed face-related activation in

prefrontal regions (Druzgal amp DrsquoEsposito 2003 Salaet al 2003 Haxby Petit et al 2000) and some ERPstudies have detected some inferior prefrontal cortexsites that generate small but specific face-specific re-sponses (Allison Puce Spencer amp McCarthy 1999)

The poor performance exhibited by CP subjectscompared to their controls on the one-back memorytask specifically for face stimuli is consistent with theinterpretation of this prefrontal activity being related toworking memory However given that the conven-tional face and object mapping experiment was notspecifically designed to look at the different componentsthat usually comprise a memory study (encoding mem-ory delay response) further research that specificallyand parametrically manipulates the memory load offaces and other object stimuli is required in order todetermine the role of this activation in face processing inCP subjects

This study has exploited the unique possibility ofexamining the activation of the fusiform gyrus andother cortical regions in an unusual group of individ-uals with CP Delineating the neural mechanism thatgives rise to abnormal behavior and by logical infer-ence normal behavior is a critical goal of neurosci-ence As might be predicted the correspondencesbetween the brain and behavior are not transparentand in this study are dissociated with the behavioralimpairment being evident concomitant with normalFFA activation Taken together however this studyhas indicated patterns of similarities and patterns ofpotential difference in this population of CP individ-uals relative to their normal counterparts The challengethat lies before us is to use these patterns to explainthe face processing deficit in detail

METHODS

Subjects

Four healthy CP subjects (2 men) aged between 29 and60 years with no discernable cortical lesion or anyhistory of neurological disease participated in all ex-periments (for details see Table 1) Another CP sub-

Table 1 Biographical Information about CP Subjects

Congenital Prosopagnosia (CP) Subjects

Subjectsrsquo Initials Sex Age

TM M 27

KM F 60

MT M 41

BE F 29

KM and TM are a mother and a son

1162 Journal of Cognitive Neuroscience Volume 17 Number 7

ject (NI male aged 40) was tested behaviorally (seeBehrmann et al 2005) but because we could not ac-quire his imaging data he was excluded from thepresent article Seven of the 12 control subjects whoparticipated in the behavioral studies along with threenew participants completed the imaging experimentsThe control group included at least two age- and sex-matched controls for each CP individual All CP andcontrol subjects were right-handed and had normal orcorrected-to-normal vision All subjects consented to par-ticipate in the experiments and the protocol was ap-proved by the Institutional Review Boards of CarnegieMellon University and the University of Pittsburgh

Imaging Experiments

Visual Stimulation

Visual stimuli were generated using the E-prime IFISsoftware (Psychology Software Tools Pittsburgh PAUSA) and projected via LCD to a screen located in theback of the scanner bore behind the subjectrsquos head Sub-jects viewed the stimuli through a tilted mirror mountedabove their eyes on the head coil Before the scan sub-

jects were familiarized with the visual stimuli and tasksof each of the imaging experiments

MRI Setup

All subjects were scanned in a 3-T Siemens Allegrascanner equipped with a standard head coil Subjectsparticipated in one scanning session that lasted 15 to2 hr in which all the functional and anatomical scanswere acquired The order of the functional experimentswithin the scanning session was generally constant forall subjects BOLD contrast was acquired using gradient-echo echo-planar imaging sequence Specific parame-ters TR = 3000 msec TE = 35 msec flip angle = 908FOV = 210 210 mm2 matrix size 64 64 35 axialslices 3 mm thickness no gap High-resolution anatom-ical scans (T1-weighted 3-D MPRAGE) were also acquiredto allow accurate cortical segmentation reconstructionand volume-based statistical analysis Specific parame-ters TE = 349 flip angle = 88 FOV = 256 256 mm2matrix size = 256 256 slice thickness = 1 mm numberof slices = 160ndash192 orientation of slices was eitherhorizontal or sagittal For three of the CP subjects the

Table 2 Laterality in force-selective activation

1 Talairach Coordinates Fusiform Gyrus

Left Hemisphere Right Hemisphere

x y z x y z

CP subjects 40 plusmn 2 53 plusmn 3 19 plusmn 2 38 plusmn 5 47 plusmn 8 17 plusmn 3

Control subjects 37 plusmn 4 41 plusmn 5 18 plusmn 4 36 plusmn 2 40 plusmn 7 17 plusmn 3

2 Activation Foci

Localizer Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 4 ndash ndash

Control subjects 9 1 ndash 4 2 1

Rubin FacendashVase Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 1 3 ndash

Control subjects 9 ndash ndash 3 2 4

Part 1 Talairach coordinates of the face-related activation in the fusiform gyrus of CP and control subjects extracted from the conventionalface and object mapping experiment Part 2 Number of CP and control subjects that showed bilateral right-lateralized or left-lateralizedface-related activation in the fusiform gyrus and in the lateral occipital region in the conventional face and object mapping experiment andin the Rubin facendashvase experiment

Avidan et al 1163

3-D brain reconstruction was performed on imagespreviously acquired on a 15-T GE scanner

Experiments

Conventional Face and Object Mapping Experiment

Stimuli included line drawings of faces buildings com-mon objects and geometric patterns presented in ashort block design (for details see Levy et al 2001Figure 1A) Each epoch lasted 9 sec and contained ninestimuli from the same category there were eight differ-ent images and one immediate repetition of one of thestimuli Each stimulus was presented for 800 msecfollowed by a 200-msec interval of blank screen Therewere seven repetitions of each condition and their orderwas pseudorandomized across subjects Visual epochswere interleaved with 6-sec blank epochs The experi-ment lasted 450 sec and started and ended with extend-ed blank epochs of 27 and 9 sec respectively Subjectswere instructed to fixate on a red central fixation dot toperform a one-back memory task and to press a key ona response glove to indicate the repeated image Bothaccuracy and reaction time were recorded

The Motion Pictures Experiment

The experiment lasted 10 minutes and was composed of32 epochs (15 sec long) of movie clips each containingimages from one of four possible categories people invarious situations navigation of the camera through citybuildings or through open fields and miscellaneousimages of objects from different categories (machinesfalling water etc for details see Hasson Nir et al2004 Figure 4) The experiment started with a 51-secblank period followed by 9 sec of pattern stimuli andended with a 1-minute blank Subjects were simplyinstructed to watch the movie clips

The Adaptation Experiment

The experiment included line drawing images fromthree different categories faces buildings and cars(for details see Avidan et al 2002 Figure 5A) Stimuliwere presented in 12-sec epochs that contained either12 different images (lsquolsquodifferentrsquorsquo condition) from thesame category or 12 repetitions of the same identicalstimulus (lsquolsquoidenticalrsquorsquo condition) Within epochs eachstimulus was presented for 800 msec followed by200 msec of fixation only The experiment lasted480 sec and started and ended with a long blank pe-riod of 21 and 15 sec respectively The first visualblock always contained geometric patterns and was ex-cluded from the analysis Visual epochs were interleavedwith 6-sec blank epochs and the order of visual epochswas pseudorandomized across subjects Subjects wereinstructed to continuously fixate a central red dot and

to covertly categorize the images as follows man wom-an or a child for the face stimuli apartment townhouseor public facility for the buildings and private car truckor bus for the vehicles

The Rubin FacendashVase Experiment

The experiment included modified versions of the Rubinfacendashvase stimuli as well as line drawings of front facesand household objects (for details see Hasson Hendleret al 2001 Figure 6A) The Rubin facendashvase stimuli weremodified so that subjectsrsquo perception was biased towardsone perceptual state or the other This was done by firstuniformly coloring the vase area within each of thestimuli and placing stripes over the profile Stimuli werethen placed either over a striped background (biastowards perception of the vase due to closure of thevase stimulus) or over a uniform-color background (biastowards perception of the face due to closure of theface stimulus) Critically the Rubin face and vase sharethe same local contours but give rise to different globalperceptions In order to ensure that subjectsrsquo perceptionremained biased to one perceptual interpretation (ei-ther the profile or vase) during the experiment stimuliwere presented briefly (200 msec) and were immediate-ly followed by a masking grid that was presented for800 msec Stimuli were presented in 9-sec blocks thatwere interleaved with 6-sec blanks Each condition wasrepeated eight times and the order of presentation waspseudorandomized across subjects Subjects maintainedcentral fixation throughout the experiment and per-formed a one-back memory task however due to atechnical failure their responses were not recorded Theexperiment lasted 507 sec and started and ended withextended blank periods

Data Analysis

Data analysis was performed using the BrainVoyager48 software package (Brain Innovation MaastrichtNetherlands) and complementary in-house softwareDetailed description of the data analysis is provided inHasson Harel et al (2003) Briefly for each subject thecortical surface was reconstructed and unfolded intothe flattened format Functional data for each subjectfrom each experiment were analyzed separately Pre-processing of functional scans included 3-D motioncorrection and filtering of low frequencies up to fivecycles per experiment (slow drift) Statistical analysiswas based on the GLM For each experiment each ofthe conditions (except for blank) was defined as a sepa-rate predictor a boxcar shape was used to model eachpredictor and a 3-sec lag was assumed Percent signalchange for each subject in each experiment (except forthe motion pictures experiment) was calculated as thepercent activation from a blank baseline For the mo-

1164 Journal of Cognitive Neuroscience Volume 17 Number 7

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

of faces and people buildings navigation or miscella-neous objects (Hasson Nir Levy Fuhrmann amp Malach2004) Importantly no specific task was required mak-ing this experiment more akin to naturally surveyingonersquos visual environment Figure 4A shows the activa-tion maps of a single representative CP and controlsubject projected on an inflated and unfolded brainrepresentation The white lines on the posterior partof the unfolded representation show the retinotopicborders mapped for each subject The face and build-ing activation maps obtained in this experiment largelyreplicate the results from the conventional mappingexperiment (compare Figure 4A and Figure 1C) ex-cept that the activation seems to be more widespreadin the motion pictures experiment selective activa-tion for faces was found in the fusiform gyrus andthe lateral occipital region when contrasting faces withbuildings and navigation scenes (red patches) andselective activation for buildings and navigation sceneswere found in the collateral sulcus when applyingthe opposite contrast (green patches) Interestinglyface-related activation was also found within the STSin regions typically activated in response to facialmovements (Hoffman amp Haxby 2000 Puce Allisonet al 1998) and biological motion (Puce and Perrett2003) That these regions are activated here is con-sistent with the richness of the faces and motions inthe displays

To assess the spatial extent of the face-related acti-vation obtained in this experiment in both CP andcontrols the number of face-related voxels within thelateral occipital region fusiform gyrus and STS wascounted for each subject in each group To allow di-rect comparisons all face-related foci for all subjectswere selected using the same statistical threshold ( p lt0005) (see Figure 4A for two representative subjects)A repeated-measures ANOVA with group (CPcontrols)hemisphere (left right) and face-related region as therepeated-measures within-subject factors and the num-ber of voxels as the dependent measure was usedThere were infrequent instances of control subjects forwhom not all face-related foci could be identified in theselected statistical threshold and these were excludedfrom the ANOVA Critically all face-related regions weredetected for the CP subjects using that threshold Thisanalysis revealed a significant main effect of hemisphere( p lt 001) confirming that overall there were moreface-related voxels in the right compared with the lefthemisphere In addition there was a significant maineffect of face-related region ( p lt 005) with moreactivated voxels in the lateral occipital region than inthe fusiform gyrus or STS Importantly however thisanalysis did not reveal a main effect of group or any two-or three-way interactions thus suggesting that therewere no differences in the spatial extent of the activationbetween the CP and control group in any of the face-related regions

To further quantify the similarity between the CP andcontrol groups we sampled the activation profile ofthree regions of interest (ROIs) (fusiform gyrus lateraloccipital region and collateral sulcus) defined indepen-dently for each subject on the basis of the conventionalface and object mapping experiment (see Methods)Percent signal change averaged across the entire exper-iment and collapsed across the left and right hemi-spheres for each group is shown in Figure 4B Asabove all ROIs exhibited clear stimulus selectivity butof even greater importance is the striking similaritybetween the activation profiles of the two groups (CP =black line controls = white line) This similarity is alsoevident from the high values of the correlation coeffi-cient (fusiform gyrus = 80 lateral occipital region =89 CoS = 81) calculated for each ROI between theaverage activation profile of each group Importantlyhowever as evident from the figure this similarity isnot only in the overall correlation but also in themoment-by-moment waxing and waning of the activityThese results confirm the normal profile of face- andbuilding-related activation in the CP subjects when morenatural viewing conditions and stimuli are employedMoreover because this normal activation was foundhere in the absence of any behavioral task it impliesthat the normal activation found in the conventionalmapping experiment was not confounded by the diffi-culty of the behavioral task

Adaptation Experiment

The behavioral results reveal a clear impairment inCP individuals in unfamiliar face discrimination in asimultaneous matching task (Behrmann et al 2005)and in a one-back memory (sequential discrimination)task (Figure 1B) These tasks are largely perceptualand do not require that a specific or individual facebe represented Indeed CP subjects might be evenmore impaired than already evident when more preciseface knowledge is required their own self-testimoniessuggest that lsquolsquoall faces look alikersquorsquo Using the fMR-adaptation paradigm for studying the neuronal proper-ties of higher-order visual areas at a subvoxel resolutionit has been shown that normal subjects typically exhibita reduction in BOLD signal with repeated presentationof a stimulus (Grill-Spector amp Malach 2001) To exam-ine whether the face-selective activation noted above isless sensitive to face repetition in CP than in controlsubjects we compared adaptation levels for face repeti-tion across the two groups We also examined the adap-tation effect for nonface stimuli to determine whetherthe entire occipito-temporal region is normally sensitiveto repetition (Avidan et al 2002)

Stimuli were presented in epochs that included either12 different stimuli (faces building cars) or 12 repeti-tions of the same identical stimulus (see Figure 5A) Theactivation profile was sampled from ROIs defined in-

1156 Journal of Cognitive Neuroscience Volume 17 Number 7

Figure 4 Activation maps and activation profiles from the motion pictures experiment (A) Activation maps for faces (red) as well as buildings

and navigation (green) are shown for one representative CP subject and one control subject The data are shown on the inf lated and unfoldedmap representation of each individual using the same statistical threshold (GLM statistics p lt 0005) The white dotted lines on the unfolded maps

indicate the borders of the retinotopic areas that were mapped for each subject in a separate experiment Abbreviations LOS = lateral occipital

sulcus IOG = inferior occipital gyrus IPS = intraparietal sulcus PCS = postcentral sulcus Other abbreviations as in Figures 1 and 3 (B) Averaged

activation profiles of the CP (black) and control subjects (white) are shown for face-related voxels in the fusiform gyrus and lateral occipital regionand for building-related voxels in the collateral sulcus The signal is shown along the time-axis of the experiment and the vertical colored bars

indicate the different experimental conditions The correlation coefficient values between the entire activation profile of CP and control subjects

are shown for each ROI Error bars indicate SEM across subjects in each group

Avidan et al 1157

Figure 5 Experimental design and activation profiles of the adaptation experiment (A) Examples of stimuli and experimental design Face

building and car line drawings were presented in epochs containing either 12 different or 12 identical images (B) Averaged activation profiles inface-related voxels in the fusiform gyrus (top) and the lateral occipital region (bottom) for the CP and control subjects Error bars indicate SEM

across subjects in each group (C) Averaged activation profiles in the building-related voxels in the collateral sulcus

1158 Journal of Cognitive Neuroscience Volume 17 Number 7

dependently for each subject using the conventionalmapping experiment Figure 5B shows the activationlevels within the face-related foci in the fusiform gyrusand lateral occipital region averaged across all repeti-tions of each experimental condition and across thetwo hemispheres for the two groups Again despitethe fact that the face-related regions were selected usingan external localizer CP individuals exhibited clear faceselectivity in the lsquolsquodifferentrsquorsquo conditions in these ROIsNote the similarity between the groups in the magni-tude of the adaptation that is the signal decrease fol-lowing face repetition (dark gray compared with lightgray bars) suggesting that face representations in CPsubjects are indeed sensitive to face repetition Theface-related regions also exhibit a strong adaptationeffect for the nonface stimuli (buildings and cars) in-dicating that like normal subjects these regions in theCP subjects contain neurons that are highly selectivefor those stimuli (Avidan et al 2002) These observa-tions were confirmed in a repeated-measure ANOVAin which the only significant effect in both regions wasa main effect of stimulus type ( p lt 0001) There wasno main effect of group nor an interaction betweenthe factors

Figure 5C shows a similar analysis for the building-related focus in the anterior collateral sulcus This regionexhibited strong adaptation for the building stimuli aswell as for the other object categories in both groupsOnce again the ANOVA revealed only a main effect ofstimulus type ( p lt 0001) The strong building selectiv-ity in the collateral sulcus combined with the findingthat this region exhibits similar adaptation levels forall stimulus categories including faces further suggeststhat the entire occipito-temporal object representationnetwork in CP subjects is intact and mirrors that of thenormal control subjects

Rubin FacendashVase Experiment

One possible interpretation of the apparently normalface-selective BOLD response in the CP subjects foundin the previous experiments is that it reflects a responseto the presence of face parts (eyes mouth and nose) inthe absence of an intact representation of the face as awhole To examine whether the activation in CP ismerely a product of a part-representation we employeda paradigm used previously which utilizes the famousRubin facendashvase illusion to demonstrate evidence forglobal face representations in normal subjects (HassonHendler Ben Bashat amp Malach 2001) Critically in theRubin facendashvase illusion similar local elements are al-ways present However depending on figurendashgroundsegmentation and global integration the local elementscan induce two different visual percepts either of a faceor of a vase (Figure 6A) If face-selective activation in thefusiform gyrus and other face-related regions depends

only on processing of the local elements then activa-tion in these regions should be similar in both theRubin-face and Rubin-vase conditions (same local ele-ments are present) However if face representation inthe fusiform gyrus and other face-related regions isaffected by the global integration of the face elementsinto the percept of either a face or a vase then increasedactivation should be found for the Rubin-face perceptualstates compared to the Rubin-vase perceptual states In-deed in normal subjects selective activation within face-related regions has been previously demonstrated forthe Rubin-face compared to the Rubin-vase stimuli sug-gesting that normal face-related activation is modulatedby global grouping processes and is not only induced bythe representation of local face parts (Hasson Hendleret al 2001)

Here face-selective regions were independently lo-calized initially for each subject by contrasting theepochs of frontal faces with those of household ob-jects (Figure 6A) Data from one control subject wereexcluded from this experiment as they were too noisyto localize face-related activation The activation profilewas extracted for the Rubin-face and Rubin-vase stimulifrom both the fusiform gyrus and lateral occipital foci

Figure 6 The Rubin facendashvase experiment (A) Examples of the

stimuli including the Rubin facendashvase stimuli (left) and frontal faces

and household objects (localizer stimuli) used to independently

localize face-related voxels (in the experiment the uniform surfacesshown here in gray were colored in light pink) (B) Activation profiles

for the Rubin facendashvase conditions in face-related voxels in the fusiform

gyrus and lateral occipital region Error bars indicate SEM acrosssubjects in each group

Avidan et al 1159

for each subject (see Table 2 Part 2 for foci) The av-eraged percent signal change across all repetitions ofeach condition and across the left and right hemi-spheres of each subject are presented in Figure 6BAs in the control group the average fusiform gyrusactivation in the CP subjects exhibited a selective re-sponse for the Rubin-face compared to Rubin-vaseindicating that this region in CP subjects is indeedinvolved in computing global information about facesand is sensitive to the context not simply to the localcomponents of the display The similarity between theactivation profiles of the two groups in the fusiformgyrus was confirmed by a repeated-measure ANOVAwhich revealed only a main effect of stimulus type( p lt 01)

The results in the lateral occipital region were onceagain more variable across subjects in both groups (seeTable 2 Part 2) The ANOVA revealed a marginallysignificant main effect of group ( p lt 06) but not ofstimulus type and no significant interaction Note thatalthough not statistically significant the selectivity forthe Rubin-face compared to the Rubin-vase conditionin the lateral occipital focus was somewhat reduced inthe CP subjects compared to their controls This resultis analogous to the somewhat reduced face selectivityfound in the lateral occipital region in the conventionalmapping experiment (Figure 2A)

DISCUSSION

In a series of four functional imaging experimentswe have shown that individuals with CP exhibit nor-mal face- and object-related fMRI activation patternsin the ventral occipito-temporal cortex particularly inthe face-related fusiform gyrus (FFA) and largely nor-mal activations in the lateral occipital region despitetheir profound behavioral impairment Importantly thispattern was evident across different types of stimuli(eg line drawings and movie clips) and across differ-ent paradigms

In the first study in a conventional mapping experi-ment a normal pattern of BOLD activation was foundwhen subjects viewed line drawings of face and nonfacestimuli and performed a one-back memory task Thisparadigm has been used extensively in the literature andserves as a rigorous baseline to demarcate areas of ac-tivation in the ventral occipito-temporal cortex Onepossible interpretation of the normal activation is thatthe task was not sufficiently taxing for the CP individ-uals Behavioral data collected concurrently with theimage acquisition rules out this possibilitymdashalthoughthe task was easy for the controls this was not so forthe CP individuals who were clearly impaired on thistask Thus the normal face-related activation found inCP cannot merely be explained by the claim that the taskwas too simplistic for the CP subjects

A second possible interpretation and one that isdiametrically opposed to the first one offered above isthat the task was excessively difficult for the CP and itis this difficulty that drove the face-related activationparticularly in the fusiform gyrus of the CP subjects In-deed previous studies have shown that activation in theFFA increases as a function of working memory load(Druzgal amp DrsquoEsposito 2001) This explanation is nottenable either as the CP subjects continued to exhibitnormal face-related activation even when they were notrequired to perform any behavioral task during themotion pictures experiment Further research is clearlyneeded to unequivocally determine the contribution oftask difficulty to face-related activation in CP individ-uals We are now starting to parametrically explore theeffects of task difficulty in these CP individuals usingnew paradigms The critical point at present is thatnormal face-related activation is obtained at the verysame time that the behavioral impairment is manifest

The finding that face- and object-related activation inthe ventral visual cortex is apparently normal is replicat-ed in the motion picture experiment As mentionedabove this replication attests to the generalizability ofthe finding from the first conventional mapping exper-iment to more natural stimuli and viewing conditions

Importantly the normal signature of face-related ac-tivation in CP was also found when additional experi-mental manipulations were used CP subjects showednormal recovery from adaptation when different faceswere shown suggesting that neurons within these re-gions were sufficiently sensitive to differentiate betweenthe face exemplars used in this experiment Althoughone might think that normal adaptation is not thatsurprising given that the face stimuli used in this exper-iment were very different from each other it is impor-tant to bear in mind that CP subjects were actuallyimpaired in discriminating between such stimulimdashinthe conventional mapping experiment similar stimuliwere used and the CP subjects were impaired suggest-ing that the stimuli were confusing and not as percep-tually dissimilar for them as one might assume Of noteis that in addition to the adaptation for faces found inthe present experiment CP subjects also exhibitednormal adaptation levels in the fusiform gyrus and otherface-related regions for nonface stimuli suggestingthat similar to control subjects and replicating previousfindings these regions are involved in processing otherobject categories in addition to faces (Avidan et al 2002Haxby Gobbini et al 2001)

Although normal adaptation for faces is apparentdetermining its origin is less obvious Because the exactsame picture of an individual face was repeated duringthe adaptation condition we cannot determine at thisstage whether the adaptation found in the fusiformgyrus was due to the repetition of the exact same shape(ie based on perceptual geometry or structure) or therepetition of the same facial identity Differentiating

1160 Journal of Cognitive Neuroscience Volume 17 Number 7

between these alternatives and establishing whetherneural activity in the FFA is sufficient for normal rep-resentation of fine distinctions among individual faceswill require further experiments in which more subtlevariations will be made to face stimuli The use of anevent-related adaptation paradigm in such future ex-periments will also minimize possible attentional varia-tions that are inherent in the block-design adaptationparadigm employed here

Finally in the last experiment similar to control sub-jects once again CP subjects also exhibited a selectiveBOLD signal for Rubin-face compared to Rubin-vasestimuli in the fusiform gyrus Importantly these stimulicontain similar local elements but induce two differentvisual percepts either of a face or of a vase dependingon figurendashground segmentation and global integrationand shape processing Hence selectivity for the Rubin-face over the Rubin-vase stimuli implies that as in thecontrol group CP subjectsrsquo face-related activation ismodulated by global grouping processes and is notsimply induced by the representation of local face parts(Hasson Hendler et al 2001)

In sum the CP subjects exhibit normal face- andobject-related activation patterns in multiple regions ofthe ventral occipito-temporal cortex under various ex-perimental paradigms Of note however is that groupdifferences between CP subjects and their controls werefound in regions outside the occipito-temporal cortexindicating that the absence of group differences is notmerely a lack of statistical power Taken together thesefindings suggest that normal face-related activation inventral occipito-temporal regions as measured withfMRI is not sufficient to ensure intact face processingand does not necessarily reflect normal face represen-tation within those regions

Can the BehaviorndashBOLD Dissociationbe Reconciled

The central result is that the BOLD activation in CP inventral occipito-temporal regions is largely not differen-tiable from that of the control subjects notwithstandingtheir marked behavioral impairment How might thisdissociation be explained

It has been previously suggested that the FFA ismainly involved in face detection that is discriminatingbetween faces and stimuli from other object categories(Tong et al 2000) an ability that is largely preserved inCP The apparently normal FFA activation in the CPsubjects might then reflect the intact involvement ofthe FFA in deriving a rather coarse and rudimentarystructural representation which suffices for some tasks(ie detection) but not for more taxing tasks such asidentification Evidence compatible with this possibleinterpretation suggests that the final stage of face iden-tification is not completed at the level of high-order

visual face-related regions such as the FFA rathersuccessful recognition requires activation in more ante-rior regions of the temporal lobe where a more abstractor semantic representation of faces is derived (HaxbyHoffman et al 2000 Leveroni et al 2000 SergentOhta amp Macdonald 1992) Further support for the roleof anterior temporal regions in mediating face identifi-cation and particularly for being a site of facial memory(long-term representation) comes from lesion studiesFor example Barton and Cherkasova (2003) foundthat patients with anterior temporal lesions were themost impaired in a task that required generating long-term representations of famous faces as in mentalimagery compared to patients with more posterioroccipito-temporal lesions Similarly patients with ante-rior temporal lobectomy (following intractable epilepsy)were found to be impaired in identifying familiar faces(Glosser Salvucci amp Chiaravalloti 2003) The source ofthe behavioral problem in CP might then arise in theabnormal propagation of activation from the intact FFAto more anterior temporal regions

However the idea that the FFA only mediates facedetection is not universally accepted and the cleandivision of labor between the FFA and more anteriorregions might not be totally correct Although notdenying the contribution of more anterior regions toface processing and particularly to face identificationrecent experimental evidence suggests that the FFA isinvolved not only in face detection but also in faceidentification Thus normal subjects exhibited differen-tial responses for face detection and identification in thefusiform gyrus with the two processes resulting in eitherdifferential fMRI amplitude (Grill-Spector et al 2004) ortemporal scales (Puce Allison et al 1999 and seeSugase Yamane Ueno amp Kawano 1999 for similarfindings from single-unit recordings in monkeys) Thesedifferential responses in the fusiform gyrus may poten-tially be the outcome of feedback from more anteriorregions (as above) andor from additional computationtaking place within the fusiform gyrus itself Which ofthese is perturbed in CP remains unclear and they arenot mutually exclusive either Propagation to or fromanterior regions may be affected in CP However recallthat CP subjects exhibit a clear deficit not only in faceidentification (naming of famous faces) but also indiscrimination of unfamiliar faces (Figure 1B) (see alsoBehrmann et al 2005) This deficit in the more per-ceptual (not just memorial) aspects of face processingimplicates posterior occipito-temporal regions ratherthan exclusively the more anterior areas (Barton PressKeenan amp OrsquoConnor 2002 Wada amp Yamamoto 2001)It remains a possibility therefore that the fusiformactivation itself may be abnormal in CP Because thecurrent study employed detection and discriminationstasks the abnormality may still be uncovered undermore challenging conditions such as face identificationor other fine-grained face tasks

Avidan et al 1161

We also cannot rule out the possibility that differ-ences in the activation patterns in the fusiform gyrusandor other parts of the face processing networkbetween CP and normal control subjects might exist inthe temporal domain Such differences could underliethe behavioral dysfunction of CP subjects but may notbe evident with the temporal resolution afforded byfMRI (in the order of seconds) Some evidence sup-porting this view comes from ERP studies showing thatthe relatively early N170 potential which shows faceselectivity in normal subjects failed to show such se-lectivity in CP subjects (Kress amp Daum 2003b BentinDeouell et al 1999)

Finally it might be that the small differences inBOLD activation in CP subjects such as the slight re-duction in face selectivity in the lateral occipital region(eg Figures 2A and 6B) might be lsquolsquoamplifiedrsquorsquo whentranslated into behavioral performance If so the lateraloccipital region would serve as a critical link in the faceprocessing network Further support for the role ofthe occipital region in face processing comes fromstudies showing that acquired prosopagnosic patientsoften have lesions that involve not only the fusiformgyrus but also the lateral occipital region (Barton et al2002 Wada amp Yamamoto 2001) Moreover brain dam-age to this region but not to the FFA in some individ-uals may be associated with severe prosopagnosia(Rossion Caldara et al 2003)

As is evident there are a host of potential expla-nations to account for the dissociation between the im-paired behavioral profile and the apparently normalBOLD activation pattern We are currently exploringthese issues using new imaging experiments and hopethat these will uncover in greater detail the mecha-nisms giving rise to CP Suffice it to say that no one ex-planation captures the BOLDndashbehavior dissociationWhat is crucial however is that a simple assignmentof face processing to the FFA is no longer defensibleand a deeper examination of the computational prop-erties of the FFA and associated regions is demanded bythese findings

Face-related Activation Outside theOccipito-temporal Cortex

In the current study we acquired whole-brain functionalscans and thus were able to explore differences be-tween the CP and control groups in the responsepatterns in all regions of the brain not only in theoccipito-temporal regions A particularly novel findingwas that robust bilateral face-related activation wasevident in prefrontal regions (precentral sulcus inferiorfrontal sulcus anterior lateral sulcus) in the CP groupIn contrast the control group only exhibited right-lateralized activation in the precentral sulcus (Figure 3)Studies of normal individuals performing workingmemory tasks have revealed face-related activation in

prefrontal regions (Druzgal amp DrsquoEsposito 2003 Salaet al 2003 Haxby Petit et al 2000) and some ERPstudies have detected some inferior prefrontal cortexsites that generate small but specific face-specific re-sponses (Allison Puce Spencer amp McCarthy 1999)

The poor performance exhibited by CP subjectscompared to their controls on the one-back memorytask specifically for face stimuli is consistent with theinterpretation of this prefrontal activity being related toworking memory However given that the conven-tional face and object mapping experiment was notspecifically designed to look at the different componentsthat usually comprise a memory study (encoding mem-ory delay response) further research that specificallyand parametrically manipulates the memory load offaces and other object stimuli is required in order todetermine the role of this activation in face processing inCP subjects

This study has exploited the unique possibility ofexamining the activation of the fusiform gyrus andother cortical regions in an unusual group of individ-uals with CP Delineating the neural mechanism thatgives rise to abnormal behavior and by logical infer-ence normal behavior is a critical goal of neurosci-ence As might be predicted the correspondencesbetween the brain and behavior are not transparentand in this study are dissociated with the behavioralimpairment being evident concomitant with normalFFA activation Taken together however this studyhas indicated patterns of similarities and patterns ofpotential difference in this population of CP individ-uals relative to their normal counterparts The challengethat lies before us is to use these patterns to explainthe face processing deficit in detail

METHODS

Subjects

Four healthy CP subjects (2 men) aged between 29 and60 years with no discernable cortical lesion or anyhistory of neurological disease participated in all ex-periments (for details see Table 1) Another CP sub-

Table 1 Biographical Information about CP Subjects

Congenital Prosopagnosia (CP) Subjects

Subjectsrsquo Initials Sex Age

TM M 27

KM F 60

MT M 41

BE F 29

KM and TM are a mother and a son

1162 Journal of Cognitive Neuroscience Volume 17 Number 7

ject (NI male aged 40) was tested behaviorally (seeBehrmann et al 2005) but because we could not ac-quire his imaging data he was excluded from thepresent article Seven of the 12 control subjects whoparticipated in the behavioral studies along with threenew participants completed the imaging experimentsThe control group included at least two age- and sex-matched controls for each CP individual All CP andcontrol subjects were right-handed and had normal orcorrected-to-normal vision All subjects consented to par-ticipate in the experiments and the protocol was ap-proved by the Institutional Review Boards of CarnegieMellon University and the University of Pittsburgh

Imaging Experiments

Visual Stimulation

Visual stimuli were generated using the E-prime IFISsoftware (Psychology Software Tools Pittsburgh PAUSA) and projected via LCD to a screen located in theback of the scanner bore behind the subjectrsquos head Sub-jects viewed the stimuli through a tilted mirror mountedabove their eyes on the head coil Before the scan sub-

jects were familiarized with the visual stimuli and tasksof each of the imaging experiments

MRI Setup

All subjects were scanned in a 3-T Siemens Allegrascanner equipped with a standard head coil Subjectsparticipated in one scanning session that lasted 15 to2 hr in which all the functional and anatomical scanswere acquired The order of the functional experimentswithin the scanning session was generally constant forall subjects BOLD contrast was acquired using gradient-echo echo-planar imaging sequence Specific parame-ters TR = 3000 msec TE = 35 msec flip angle = 908FOV = 210 210 mm2 matrix size 64 64 35 axialslices 3 mm thickness no gap High-resolution anatom-ical scans (T1-weighted 3-D MPRAGE) were also acquiredto allow accurate cortical segmentation reconstructionand volume-based statistical analysis Specific parame-ters TE = 349 flip angle = 88 FOV = 256 256 mm2matrix size = 256 256 slice thickness = 1 mm numberof slices = 160ndash192 orientation of slices was eitherhorizontal or sagittal For three of the CP subjects the

Table 2 Laterality in force-selective activation

1 Talairach Coordinates Fusiform Gyrus

Left Hemisphere Right Hemisphere

x y z x y z

CP subjects 40 plusmn 2 53 plusmn 3 19 plusmn 2 38 plusmn 5 47 plusmn 8 17 plusmn 3

Control subjects 37 plusmn 4 41 plusmn 5 18 plusmn 4 36 plusmn 2 40 plusmn 7 17 plusmn 3

2 Activation Foci

Localizer Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 4 ndash ndash

Control subjects 9 1 ndash 4 2 1

Rubin FacendashVase Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 1 3 ndash

Control subjects 9 ndash ndash 3 2 4

Part 1 Talairach coordinates of the face-related activation in the fusiform gyrus of CP and control subjects extracted from the conventionalface and object mapping experiment Part 2 Number of CP and control subjects that showed bilateral right-lateralized or left-lateralizedface-related activation in the fusiform gyrus and in the lateral occipital region in the conventional face and object mapping experiment andin the Rubin facendashvase experiment

Avidan et al 1163

3-D brain reconstruction was performed on imagespreviously acquired on a 15-T GE scanner

Experiments

Conventional Face and Object Mapping Experiment

Stimuli included line drawings of faces buildings com-mon objects and geometric patterns presented in ashort block design (for details see Levy et al 2001Figure 1A) Each epoch lasted 9 sec and contained ninestimuli from the same category there were eight differ-ent images and one immediate repetition of one of thestimuli Each stimulus was presented for 800 msecfollowed by a 200-msec interval of blank screen Therewere seven repetitions of each condition and their orderwas pseudorandomized across subjects Visual epochswere interleaved with 6-sec blank epochs The experi-ment lasted 450 sec and started and ended with extend-ed blank epochs of 27 and 9 sec respectively Subjectswere instructed to fixate on a red central fixation dot toperform a one-back memory task and to press a key ona response glove to indicate the repeated image Bothaccuracy and reaction time were recorded

The Motion Pictures Experiment

The experiment lasted 10 minutes and was composed of32 epochs (15 sec long) of movie clips each containingimages from one of four possible categories people invarious situations navigation of the camera through citybuildings or through open fields and miscellaneousimages of objects from different categories (machinesfalling water etc for details see Hasson Nir et al2004 Figure 4) The experiment started with a 51-secblank period followed by 9 sec of pattern stimuli andended with a 1-minute blank Subjects were simplyinstructed to watch the movie clips

The Adaptation Experiment

The experiment included line drawing images fromthree different categories faces buildings and cars(for details see Avidan et al 2002 Figure 5A) Stimuliwere presented in 12-sec epochs that contained either12 different images (lsquolsquodifferentrsquorsquo condition) from thesame category or 12 repetitions of the same identicalstimulus (lsquolsquoidenticalrsquorsquo condition) Within epochs eachstimulus was presented for 800 msec followed by200 msec of fixation only The experiment lasted480 sec and started and ended with a long blank pe-riod of 21 and 15 sec respectively The first visualblock always contained geometric patterns and was ex-cluded from the analysis Visual epochs were interleavedwith 6-sec blank epochs and the order of visual epochswas pseudorandomized across subjects Subjects wereinstructed to continuously fixate a central red dot and

to covertly categorize the images as follows man wom-an or a child for the face stimuli apartment townhouseor public facility for the buildings and private car truckor bus for the vehicles

The Rubin FacendashVase Experiment

The experiment included modified versions of the Rubinfacendashvase stimuli as well as line drawings of front facesand household objects (for details see Hasson Hendleret al 2001 Figure 6A) The Rubin facendashvase stimuli weremodified so that subjectsrsquo perception was biased towardsone perceptual state or the other This was done by firstuniformly coloring the vase area within each of thestimuli and placing stripes over the profile Stimuli werethen placed either over a striped background (biastowards perception of the vase due to closure of thevase stimulus) or over a uniform-color background (biastowards perception of the face due to closure of theface stimulus) Critically the Rubin face and vase sharethe same local contours but give rise to different globalperceptions In order to ensure that subjectsrsquo perceptionremained biased to one perceptual interpretation (ei-ther the profile or vase) during the experiment stimuliwere presented briefly (200 msec) and were immediate-ly followed by a masking grid that was presented for800 msec Stimuli were presented in 9-sec blocks thatwere interleaved with 6-sec blanks Each condition wasrepeated eight times and the order of presentation waspseudorandomized across subjects Subjects maintainedcentral fixation throughout the experiment and per-formed a one-back memory task however due to atechnical failure their responses were not recorded Theexperiment lasted 507 sec and started and ended withextended blank periods

Data Analysis

Data analysis was performed using the BrainVoyager48 software package (Brain Innovation MaastrichtNetherlands) and complementary in-house softwareDetailed description of the data analysis is provided inHasson Harel et al (2003) Briefly for each subject thecortical surface was reconstructed and unfolded intothe flattened format Functional data for each subjectfrom each experiment were analyzed separately Pre-processing of functional scans included 3-D motioncorrection and filtering of low frequencies up to fivecycles per experiment (slow drift) Statistical analysiswas based on the GLM For each experiment each ofthe conditions (except for blank) was defined as a sepa-rate predictor a boxcar shape was used to model eachpredictor and a 3-sec lag was assumed Percent signalchange for each subject in each experiment (except forthe motion pictures experiment) was calculated as thepercent activation from a blank baseline For the mo-

1164 Journal of Cognitive Neuroscience Volume 17 Number 7

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

Figure 4 Activation maps and activation profiles from the motion pictures experiment (A) Activation maps for faces (red) as well as buildings

and navigation (green) are shown for one representative CP subject and one control subject The data are shown on the inf lated and unfoldedmap representation of each individual using the same statistical threshold (GLM statistics p lt 0005) The white dotted lines on the unfolded maps

indicate the borders of the retinotopic areas that were mapped for each subject in a separate experiment Abbreviations LOS = lateral occipital

sulcus IOG = inferior occipital gyrus IPS = intraparietal sulcus PCS = postcentral sulcus Other abbreviations as in Figures 1 and 3 (B) Averaged

activation profiles of the CP (black) and control subjects (white) are shown for face-related voxels in the fusiform gyrus and lateral occipital regionand for building-related voxels in the collateral sulcus The signal is shown along the time-axis of the experiment and the vertical colored bars

indicate the different experimental conditions The correlation coefficient values between the entire activation profile of CP and control subjects

are shown for each ROI Error bars indicate SEM across subjects in each group

Avidan et al 1157

Figure 5 Experimental design and activation profiles of the adaptation experiment (A) Examples of stimuli and experimental design Face

building and car line drawings were presented in epochs containing either 12 different or 12 identical images (B) Averaged activation profiles inface-related voxels in the fusiform gyrus (top) and the lateral occipital region (bottom) for the CP and control subjects Error bars indicate SEM

across subjects in each group (C) Averaged activation profiles in the building-related voxels in the collateral sulcus

1158 Journal of Cognitive Neuroscience Volume 17 Number 7

dependently for each subject using the conventionalmapping experiment Figure 5B shows the activationlevels within the face-related foci in the fusiform gyrusand lateral occipital region averaged across all repeti-tions of each experimental condition and across thetwo hemispheres for the two groups Again despitethe fact that the face-related regions were selected usingan external localizer CP individuals exhibited clear faceselectivity in the lsquolsquodifferentrsquorsquo conditions in these ROIsNote the similarity between the groups in the magni-tude of the adaptation that is the signal decrease fol-lowing face repetition (dark gray compared with lightgray bars) suggesting that face representations in CPsubjects are indeed sensitive to face repetition Theface-related regions also exhibit a strong adaptationeffect for the nonface stimuli (buildings and cars) in-dicating that like normal subjects these regions in theCP subjects contain neurons that are highly selectivefor those stimuli (Avidan et al 2002) These observa-tions were confirmed in a repeated-measure ANOVAin which the only significant effect in both regions wasa main effect of stimulus type ( p lt 0001) There wasno main effect of group nor an interaction betweenthe factors

Figure 5C shows a similar analysis for the building-related focus in the anterior collateral sulcus This regionexhibited strong adaptation for the building stimuli aswell as for the other object categories in both groupsOnce again the ANOVA revealed only a main effect ofstimulus type ( p lt 0001) The strong building selectiv-ity in the collateral sulcus combined with the findingthat this region exhibits similar adaptation levels forall stimulus categories including faces further suggeststhat the entire occipito-temporal object representationnetwork in CP subjects is intact and mirrors that of thenormal control subjects

Rubin FacendashVase Experiment

One possible interpretation of the apparently normalface-selective BOLD response in the CP subjects foundin the previous experiments is that it reflects a responseto the presence of face parts (eyes mouth and nose) inthe absence of an intact representation of the face as awhole To examine whether the activation in CP ismerely a product of a part-representation we employeda paradigm used previously which utilizes the famousRubin facendashvase illusion to demonstrate evidence forglobal face representations in normal subjects (HassonHendler Ben Bashat amp Malach 2001) Critically in theRubin facendashvase illusion similar local elements are al-ways present However depending on figurendashgroundsegmentation and global integration the local elementscan induce two different visual percepts either of a faceor of a vase (Figure 6A) If face-selective activation in thefusiform gyrus and other face-related regions depends

only on processing of the local elements then activa-tion in these regions should be similar in both theRubin-face and Rubin-vase conditions (same local ele-ments are present) However if face representation inthe fusiform gyrus and other face-related regions isaffected by the global integration of the face elementsinto the percept of either a face or a vase then increasedactivation should be found for the Rubin-face perceptualstates compared to the Rubin-vase perceptual states In-deed in normal subjects selective activation within face-related regions has been previously demonstrated forthe Rubin-face compared to the Rubin-vase stimuli sug-gesting that normal face-related activation is modulatedby global grouping processes and is not only induced bythe representation of local face parts (Hasson Hendleret al 2001)

Here face-selective regions were independently lo-calized initially for each subject by contrasting theepochs of frontal faces with those of household ob-jects (Figure 6A) Data from one control subject wereexcluded from this experiment as they were too noisyto localize face-related activation The activation profilewas extracted for the Rubin-face and Rubin-vase stimulifrom both the fusiform gyrus and lateral occipital foci

Figure 6 The Rubin facendashvase experiment (A) Examples of the

stimuli including the Rubin facendashvase stimuli (left) and frontal faces

and household objects (localizer stimuli) used to independently

localize face-related voxels (in the experiment the uniform surfacesshown here in gray were colored in light pink) (B) Activation profiles

for the Rubin facendashvase conditions in face-related voxels in the fusiform

gyrus and lateral occipital region Error bars indicate SEM acrosssubjects in each group

Avidan et al 1159

for each subject (see Table 2 Part 2 for foci) The av-eraged percent signal change across all repetitions ofeach condition and across the left and right hemi-spheres of each subject are presented in Figure 6BAs in the control group the average fusiform gyrusactivation in the CP subjects exhibited a selective re-sponse for the Rubin-face compared to Rubin-vaseindicating that this region in CP subjects is indeedinvolved in computing global information about facesand is sensitive to the context not simply to the localcomponents of the display The similarity between theactivation profiles of the two groups in the fusiformgyrus was confirmed by a repeated-measure ANOVAwhich revealed only a main effect of stimulus type( p lt 01)

The results in the lateral occipital region were onceagain more variable across subjects in both groups (seeTable 2 Part 2) The ANOVA revealed a marginallysignificant main effect of group ( p lt 06) but not ofstimulus type and no significant interaction Note thatalthough not statistically significant the selectivity forthe Rubin-face compared to the Rubin-vase conditionin the lateral occipital focus was somewhat reduced inthe CP subjects compared to their controls This resultis analogous to the somewhat reduced face selectivityfound in the lateral occipital region in the conventionalmapping experiment (Figure 2A)

DISCUSSION

In a series of four functional imaging experimentswe have shown that individuals with CP exhibit nor-mal face- and object-related fMRI activation patternsin the ventral occipito-temporal cortex particularly inthe face-related fusiform gyrus (FFA) and largely nor-mal activations in the lateral occipital region despitetheir profound behavioral impairment Importantly thispattern was evident across different types of stimuli(eg line drawings and movie clips) and across differ-ent paradigms

In the first study in a conventional mapping experi-ment a normal pattern of BOLD activation was foundwhen subjects viewed line drawings of face and nonfacestimuli and performed a one-back memory task Thisparadigm has been used extensively in the literature andserves as a rigorous baseline to demarcate areas of ac-tivation in the ventral occipito-temporal cortex Onepossible interpretation of the normal activation is thatthe task was not sufficiently taxing for the CP individ-uals Behavioral data collected concurrently with theimage acquisition rules out this possibilitymdashalthoughthe task was easy for the controls this was not so forthe CP individuals who were clearly impaired on thistask Thus the normal face-related activation found inCP cannot merely be explained by the claim that the taskwas too simplistic for the CP subjects

A second possible interpretation and one that isdiametrically opposed to the first one offered above isthat the task was excessively difficult for the CP and itis this difficulty that drove the face-related activationparticularly in the fusiform gyrus of the CP subjects In-deed previous studies have shown that activation in theFFA increases as a function of working memory load(Druzgal amp DrsquoEsposito 2001) This explanation is nottenable either as the CP subjects continued to exhibitnormal face-related activation even when they were notrequired to perform any behavioral task during themotion pictures experiment Further research is clearlyneeded to unequivocally determine the contribution oftask difficulty to face-related activation in CP individ-uals We are now starting to parametrically explore theeffects of task difficulty in these CP individuals usingnew paradigms The critical point at present is thatnormal face-related activation is obtained at the verysame time that the behavioral impairment is manifest

The finding that face- and object-related activation inthe ventral visual cortex is apparently normal is replicat-ed in the motion picture experiment As mentionedabove this replication attests to the generalizability ofthe finding from the first conventional mapping exper-iment to more natural stimuli and viewing conditions

Importantly the normal signature of face-related ac-tivation in CP was also found when additional experi-mental manipulations were used CP subjects showednormal recovery from adaptation when different faceswere shown suggesting that neurons within these re-gions were sufficiently sensitive to differentiate betweenthe face exemplars used in this experiment Althoughone might think that normal adaptation is not thatsurprising given that the face stimuli used in this exper-iment were very different from each other it is impor-tant to bear in mind that CP subjects were actuallyimpaired in discriminating between such stimulimdashinthe conventional mapping experiment similar stimuliwere used and the CP subjects were impaired suggest-ing that the stimuli were confusing and not as percep-tually dissimilar for them as one might assume Of noteis that in addition to the adaptation for faces found inthe present experiment CP subjects also exhibitednormal adaptation levels in the fusiform gyrus and otherface-related regions for nonface stimuli suggestingthat similar to control subjects and replicating previousfindings these regions are involved in processing otherobject categories in addition to faces (Avidan et al 2002Haxby Gobbini et al 2001)

Although normal adaptation for faces is apparentdetermining its origin is less obvious Because the exactsame picture of an individual face was repeated duringthe adaptation condition we cannot determine at thisstage whether the adaptation found in the fusiformgyrus was due to the repetition of the exact same shape(ie based on perceptual geometry or structure) or therepetition of the same facial identity Differentiating

1160 Journal of Cognitive Neuroscience Volume 17 Number 7

between these alternatives and establishing whetherneural activity in the FFA is sufficient for normal rep-resentation of fine distinctions among individual faceswill require further experiments in which more subtlevariations will be made to face stimuli The use of anevent-related adaptation paradigm in such future ex-periments will also minimize possible attentional varia-tions that are inherent in the block-design adaptationparadigm employed here

Finally in the last experiment similar to control sub-jects once again CP subjects also exhibited a selectiveBOLD signal for Rubin-face compared to Rubin-vasestimuli in the fusiform gyrus Importantly these stimulicontain similar local elements but induce two differentvisual percepts either of a face or of a vase dependingon figurendashground segmentation and global integrationand shape processing Hence selectivity for the Rubin-face over the Rubin-vase stimuli implies that as in thecontrol group CP subjectsrsquo face-related activation ismodulated by global grouping processes and is notsimply induced by the representation of local face parts(Hasson Hendler et al 2001)

In sum the CP subjects exhibit normal face- andobject-related activation patterns in multiple regions ofthe ventral occipito-temporal cortex under various ex-perimental paradigms Of note however is that groupdifferences between CP subjects and their controls werefound in regions outside the occipito-temporal cortexindicating that the absence of group differences is notmerely a lack of statistical power Taken together thesefindings suggest that normal face-related activation inventral occipito-temporal regions as measured withfMRI is not sufficient to ensure intact face processingand does not necessarily reflect normal face represen-tation within those regions

Can the BehaviorndashBOLD Dissociationbe Reconciled

The central result is that the BOLD activation in CP inventral occipito-temporal regions is largely not differen-tiable from that of the control subjects notwithstandingtheir marked behavioral impairment How might thisdissociation be explained

It has been previously suggested that the FFA ismainly involved in face detection that is discriminatingbetween faces and stimuli from other object categories(Tong et al 2000) an ability that is largely preserved inCP The apparently normal FFA activation in the CPsubjects might then reflect the intact involvement ofthe FFA in deriving a rather coarse and rudimentarystructural representation which suffices for some tasks(ie detection) but not for more taxing tasks such asidentification Evidence compatible with this possibleinterpretation suggests that the final stage of face iden-tification is not completed at the level of high-order

visual face-related regions such as the FFA rathersuccessful recognition requires activation in more ante-rior regions of the temporal lobe where a more abstractor semantic representation of faces is derived (HaxbyHoffman et al 2000 Leveroni et al 2000 SergentOhta amp Macdonald 1992) Further support for the roleof anterior temporal regions in mediating face identifi-cation and particularly for being a site of facial memory(long-term representation) comes from lesion studiesFor example Barton and Cherkasova (2003) foundthat patients with anterior temporal lesions were themost impaired in a task that required generating long-term representations of famous faces as in mentalimagery compared to patients with more posterioroccipito-temporal lesions Similarly patients with ante-rior temporal lobectomy (following intractable epilepsy)were found to be impaired in identifying familiar faces(Glosser Salvucci amp Chiaravalloti 2003) The source ofthe behavioral problem in CP might then arise in theabnormal propagation of activation from the intact FFAto more anterior temporal regions

However the idea that the FFA only mediates facedetection is not universally accepted and the cleandivision of labor between the FFA and more anteriorregions might not be totally correct Although notdenying the contribution of more anterior regions toface processing and particularly to face identificationrecent experimental evidence suggests that the FFA isinvolved not only in face detection but also in faceidentification Thus normal subjects exhibited differen-tial responses for face detection and identification in thefusiform gyrus with the two processes resulting in eitherdifferential fMRI amplitude (Grill-Spector et al 2004) ortemporal scales (Puce Allison et al 1999 and seeSugase Yamane Ueno amp Kawano 1999 for similarfindings from single-unit recordings in monkeys) Thesedifferential responses in the fusiform gyrus may poten-tially be the outcome of feedback from more anteriorregions (as above) andor from additional computationtaking place within the fusiform gyrus itself Which ofthese is perturbed in CP remains unclear and they arenot mutually exclusive either Propagation to or fromanterior regions may be affected in CP However recallthat CP subjects exhibit a clear deficit not only in faceidentification (naming of famous faces) but also indiscrimination of unfamiliar faces (Figure 1B) (see alsoBehrmann et al 2005) This deficit in the more per-ceptual (not just memorial) aspects of face processingimplicates posterior occipito-temporal regions ratherthan exclusively the more anterior areas (Barton PressKeenan amp OrsquoConnor 2002 Wada amp Yamamoto 2001)It remains a possibility therefore that the fusiformactivation itself may be abnormal in CP Because thecurrent study employed detection and discriminationstasks the abnormality may still be uncovered undermore challenging conditions such as face identificationor other fine-grained face tasks

Avidan et al 1161

We also cannot rule out the possibility that differ-ences in the activation patterns in the fusiform gyrusandor other parts of the face processing networkbetween CP and normal control subjects might exist inthe temporal domain Such differences could underliethe behavioral dysfunction of CP subjects but may notbe evident with the temporal resolution afforded byfMRI (in the order of seconds) Some evidence sup-porting this view comes from ERP studies showing thatthe relatively early N170 potential which shows faceselectivity in normal subjects failed to show such se-lectivity in CP subjects (Kress amp Daum 2003b BentinDeouell et al 1999)

Finally it might be that the small differences inBOLD activation in CP subjects such as the slight re-duction in face selectivity in the lateral occipital region(eg Figures 2A and 6B) might be lsquolsquoamplifiedrsquorsquo whentranslated into behavioral performance If so the lateraloccipital region would serve as a critical link in the faceprocessing network Further support for the role ofthe occipital region in face processing comes fromstudies showing that acquired prosopagnosic patientsoften have lesions that involve not only the fusiformgyrus but also the lateral occipital region (Barton et al2002 Wada amp Yamamoto 2001) Moreover brain dam-age to this region but not to the FFA in some individ-uals may be associated with severe prosopagnosia(Rossion Caldara et al 2003)

As is evident there are a host of potential expla-nations to account for the dissociation between the im-paired behavioral profile and the apparently normalBOLD activation pattern We are currently exploringthese issues using new imaging experiments and hopethat these will uncover in greater detail the mecha-nisms giving rise to CP Suffice it to say that no one ex-planation captures the BOLDndashbehavior dissociationWhat is crucial however is that a simple assignmentof face processing to the FFA is no longer defensibleand a deeper examination of the computational prop-erties of the FFA and associated regions is demanded bythese findings

Face-related Activation Outside theOccipito-temporal Cortex

In the current study we acquired whole-brain functionalscans and thus were able to explore differences be-tween the CP and control groups in the responsepatterns in all regions of the brain not only in theoccipito-temporal regions A particularly novel findingwas that robust bilateral face-related activation wasevident in prefrontal regions (precentral sulcus inferiorfrontal sulcus anterior lateral sulcus) in the CP groupIn contrast the control group only exhibited right-lateralized activation in the precentral sulcus (Figure 3)Studies of normal individuals performing workingmemory tasks have revealed face-related activation in

prefrontal regions (Druzgal amp DrsquoEsposito 2003 Salaet al 2003 Haxby Petit et al 2000) and some ERPstudies have detected some inferior prefrontal cortexsites that generate small but specific face-specific re-sponses (Allison Puce Spencer amp McCarthy 1999)

The poor performance exhibited by CP subjectscompared to their controls on the one-back memorytask specifically for face stimuli is consistent with theinterpretation of this prefrontal activity being related toworking memory However given that the conven-tional face and object mapping experiment was notspecifically designed to look at the different componentsthat usually comprise a memory study (encoding mem-ory delay response) further research that specificallyand parametrically manipulates the memory load offaces and other object stimuli is required in order todetermine the role of this activation in face processing inCP subjects

This study has exploited the unique possibility ofexamining the activation of the fusiform gyrus andother cortical regions in an unusual group of individ-uals with CP Delineating the neural mechanism thatgives rise to abnormal behavior and by logical infer-ence normal behavior is a critical goal of neurosci-ence As might be predicted the correspondencesbetween the brain and behavior are not transparentand in this study are dissociated with the behavioralimpairment being evident concomitant with normalFFA activation Taken together however this studyhas indicated patterns of similarities and patterns ofpotential difference in this population of CP individ-uals relative to their normal counterparts The challengethat lies before us is to use these patterns to explainthe face processing deficit in detail

METHODS

Subjects

Four healthy CP subjects (2 men) aged between 29 and60 years with no discernable cortical lesion or anyhistory of neurological disease participated in all ex-periments (for details see Table 1) Another CP sub-

Table 1 Biographical Information about CP Subjects

Congenital Prosopagnosia (CP) Subjects

Subjectsrsquo Initials Sex Age

TM M 27

KM F 60

MT M 41

BE F 29

KM and TM are a mother and a son

1162 Journal of Cognitive Neuroscience Volume 17 Number 7

ject (NI male aged 40) was tested behaviorally (seeBehrmann et al 2005) but because we could not ac-quire his imaging data he was excluded from thepresent article Seven of the 12 control subjects whoparticipated in the behavioral studies along with threenew participants completed the imaging experimentsThe control group included at least two age- and sex-matched controls for each CP individual All CP andcontrol subjects were right-handed and had normal orcorrected-to-normal vision All subjects consented to par-ticipate in the experiments and the protocol was ap-proved by the Institutional Review Boards of CarnegieMellon University and the University of Pittsburgh

Imaging Experiments

Visual Stimulation

Visual stimuli were generated using the E-prime IFISsoftware (Psychology Software Tools Pittsburgh PAUSA) and projected via LCD to a screen located in theback of the scanner bore behind the subjectrsquos head Sub-jects viewed the stimuli through a tilted mirror mountedabove their eyes on the head coil Before the scan sub-

jects were familiarized with the visual stimuli and tasksof each of the imaging experiments

MRI Setup

All subjects were scanned in a 3-T Siemens Allegrascanner equipped with a standard head coil Subjectsparticipated in one scanning session that lasted 15 to2 hr in which all the functional and anatomical scanswere acquired The order of the functional experimentswithin the scanning session was generally constant forall subjects BOLD contrast was acquired using gradient-echo echo-planar imaging sequence Specific parame-ters TR = 3000 msec TE = 35 msec flip angle = 908FOV = 210 210 mm2 matrix size 64 64 35 axialslices 3 mm thickness no gap High-resolution anatom-ical scans (T1-weighted 3-D MPRAGE) were also acquiredto allow accurate cortical segmentation reconstructionand volume-based statistical analysis Specific parame-ters TE = 349 flip angle = 88 FOV = 256 256 mm2matrix size = 256 256 slice thickness = 1 mm numberof slices = 160ndash192 orientation of slices was eitherhorizontal or sagittal For three of the CP subjects the

Table 2 Laterality in force-selective activation

1 Talairach Coordinates Fusiform Gyrus

Left Hemisphere Right Hemisphere

x y z x y z

CP subjects 40 plusmn 2 53 plusmn 3 19 plusmn 2 38 plusmn 5 47 plusmn 8 17 plusmn 3

Control subjects 37 plusmn 4 41 plusmn 5 18 plusmn 4 36 plusmn 2 40 plusmn 7 17 plusmn 3

2 Activation Foci

Localizer Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 4 ndash ndash

Control subjects 9 1 ndash 4 2 1

Rubin FacendashVase Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 1 3 ndash

Control subjects 9 ndash ndash 3 2 4

Part 1 Talairach coordinates of the face-related activation in the fusiform gyrus of CP and control subjects extracted from the conventionalface and object mapping experiment Part 2 Number of CP and control subjects that showed bilateral right-lateralized or left-lateralizedface-related activation in the fusiform gyrus and in the lateral occipital region in the conventional face and object mapping experiment andin the Rubin facendashvase experiment

Avidan et al 1163

3-D brain reconstruction was performed on imagespreviously acquired on a 15-T GE scanner

Experiments

Conventional Face and Object Mapping Experiment

Stimuli included line drawings of faces buildings com-mon objects and geometric patterns presented in ashort block design (for details see Levy et al 2001Figure 1A) Each epoch lasted 9 sec and contained ninestimuli from the same category there were eight differ-ent images and one immediate repetition of one of thestimuli Each stimulus was presented for 800 msecfollowed by a 200-msec interval of blank screen Therewere seven repetitions of each condition and their orderwas pseudorandomized across subjects Visual epochswere interleaved with 6-sec blank epochs The experi-ment lasted 450 sec and started and ended with extend-ed blank epochs of 27 and 9 sec respectively Subjectswere instructed to fixate on a red central fixation dot toperform a one-back memory task and to press a key ona response glove to indicate the repeated image Bothaccuracy and reaction time were recorded

The Motion Pictures Experiment

The experiment lasted 10 minutes and was composed of32 epochs (15 sec long) of movie clips each containingimages from one of four possible categories people invarious situations navigation of the camera through citybuildings or through open fields and miscellaneousimages of objects from different categories (machinesfalling water etc for details see Hasson Nir et al2004 Figure 4) The experiment started with a 51-secblank period followed by 9 sec of pattern stimuli andended with a 1-minute blank Subjects were simplyinstructed to watch the movie clips

The Adaptation Experiment

The experiment included line drawing images fromthree different categories faces buildings and cars(for details see Avidan et al 2002 Figure 5A) Stimuliwere presented in 12-sec epochs that contained either12 different images (lsquolsquodifferentrsquorsquo condition) from thesame category or 12 repetitions of the same identicalstimulus (lsquolsquoidenticalrsquorsquo condition) Within epochs eachstimulus was presented for 800 msec followed by200 msec of fixation only The experiment lasted480 sec and started and ended with a long blank pe-riod of 21 and 15 sec respectively The first visualblock always contained geometric patterns and was ex-cluded from the analysis Visual epochs were interleavedwith 6-sec blank epochs and the order of visual epochswas pseudorandomized across subjects Subjects wereinstructed to continuously fixate a central red dot and

to covertly categorize the images as follows man wom-an or a child for the face stimuli apartment townhouseor public facility for the buildings and private car truckor bus for the vehicles

The Rubin FacendashVase Experiment

The experiment included modified versions of the Rubinfacendashvase stimuli as well as line drawings of front facesand household objects (for details see Hasson Hendleret al 2001 Figure 6A) The Rubin facendashvase stimuli weremodified so that subjectsrsquo perception was biased towardsone perceptual state or the other This was done by firstuniformly coloring the vase area within each of thestimuli and placing stripes over the profile Stimuli werethen placed either over a striped background (biastowards perception of the vase due to closure of thevase stimulus) or over a uniform-color background (biastowards perception of the face due to closure of theface stimulus) Critically the Rubin face and vase sharethe same local contours but give rise to different globalperceptions In order to ensure that subjectsrsquo perceptionremained biased to one perceptual interpretation (ei-ther the profile or vase) during the experiment stimuliwere presented briefly (200 msec) and were immediate-ly followed by a masking grid that was presented for800 msec Stimuli were presented in 9-sec blocks thatwere interleaved with 6-sec blanks Each condition wasrepeated eight times and the order of presentation waspseudorandomized across subjects Subjects maintainedcentral fixation throughout the experiment and per-formed a one-back memory task however due to atechnical failure their responses were not recorded Theexperiment lasted 507 sec and started and ended withextended blank periods

Data Analysis

Data analysis was performed using the BrainVoyager48 software package (Brain Innovation MaastrichtNetherlands) and complementary in-house softwareDetailed description of the data analysis is provided inHasson Harel et al (2003) Briefly for each subject thecortical surface was reconstructed and unfolded intothe flattened format Functional data for each subjectfrom each experiment were analyzed separately Pre-processing of functional scans included 3-D motioncorrection and filtering of low frequencies up to fivecycles per experiment (slow drift) Statistical analysiswas based on the GLM For each experiment each ofthe conditions (except for blank) was defined as a sepa-rate predictor a boxcar shape was used to model eachpredictor and a 3-sec lag was assumed Percent signalchange for each subject in each experiment (except forthe motion pictures experiment) was calculated as thepercent activation from a blank baseline For the mo-

1164 Journal of Cognitive Neuroscience Volume 17 Number 7

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

Figure 5 Experimental design and activation profiles of the adaptation experiment (A) Examples of stimuli and experimental design Face

building and car line drawings were presented in epochs containing either 12 different or 12 identical images (B) Averaged activation profiles inface-related voxels in the fusiform gyrus (top) and the lateral occipital region (bottom) for the CP and control subjects Error bars indicate SEM

across subjects in each group (C) Averaged activation profiles in the building-related voxels in the collateral sulcus

1158 Journal of Cognitive Neuroscience Volume 17 Number 7

dependently for each subject using the conventionalmapping experiment Figure 5B shows the activationlevels within the face-related foci in the fusiform gyrusand lateral occipital region averaged across all repeti-tions of each experimental condition and across thetwo hemispheres for the two groups Again despitethe fact that the face-related regions were selected usingan external localizer CP individuals exhibited clear faceselectivity in the lsquolsquodifferentrsquorsquo conditions in these ROIsNote the similarity between the groups in the magni-tude of the adaptation that is the signal decrease fol-lowing face repetition (dark gray compared with lightgray bars) suggesting that face representations in CPsubjects are indeed sensitive to face repetition Theface-related regions also exhibit a strong adaptationeffect for the nonface stimuli (buildings and cars) in-dicating that like normal subjects these regions in theCP subjects contain neurons that are highly selectivefor those stimuli (Avidan et al 2002) These observa-tions were confirmed in a repeated-measure ANOVAin which the only significant effect in both regions wasa main effect of stimulus type ( p lt 0001) There wasno main effect of group nor an interaction betweenthe factors

Figure 5C shows a similar analysis for the building-related focus in the anterior collateral sulcus This regionexhibited strong adaptation for the building stimuli aswell as for the other object categories in both groupsOnce again the ANOVA revealed only a main effect ofstimulus type ( p lt 0001) The strong building selectiv-ity in the collateral sulcus combined with the findingthat this region exhibits similar adaptation levels forall stimulus categories including faces further suggeststhat the entire occipito-temporal object representationnetwork in CP subjects is intact and mirrors that of thenormal control subjects

Rubin FacendashVase Experiment

One possible interpretation of the apparently normalface-selective BOLD response in the CP subjects foundin the previous experiments is that it reflects a responseto the presence of face parts (eyes mouth and nose) inthe absence of an intact representation of the face as awhole To examine whether the activation in CP ismerely a product of a part-representation we employeda paradigm used previously which utilizes the famousRubin facendashvase illusion to demonstrate evidence forglobal face representations in normal subjects (HassonHendler Ben Bashat amp Malach 2001) Critically in theRubin facendashvase illusion similar local elements are al-ways present However depending on figurendashgroundsegmentation and global integration the local elementscan induce two different visual percepts either of a faceor of a vase (Figure 6A) If face-selective activation in thefusiform gyrus and other face-related regions depends

only on processing of the local elements then activa-tion in these regions should be similar in both theRubin-face and Rubin-vase conditions (same local ele-ments are present) However if face representation inthe fusiform gyrus and other face-related regions isaffected by the global integration of the face elementsinto the percept of either a face or a vase then increasedactivation should be found for the Rubin-face perceptualstates compared to the Rubin-vase perceptual states In-deed in normal subjects selective activation within face-related regions has been previously demonstrated forthe Rubin-face compared to the Rubin-vase stimuli sug-gesting that normal face-related activation is modulatedby global grouping processes and is not only induced bythe representation of local face parts (Hasson Hendleret al 2001)

Here face-selective regions were independently lo-calized initially for each subject by contrasting theepochs of frontal faces with those of household ob-jects (Figure 6A) Data from one control subject wereexcluded from this experiment as they were too noisyto localize face-related activation The activation profilewas extracted for the Rubin-face and Rubin-vase stimulifrom both the fusiform gyrus and lateral occipital foci

Figure 6 The Rubin facendashvase experiment (A) Examples of the

stimuli including the Rubin facendashvase stimuli (left) and frontal faces

and household objects (localizer stimuli) used to independently

localize face-related voxels (in the experiment the uniform surfacesshown here in gray were colored in light pink) (B) Activation profiles

for the Rubin facendashvase conditions in face-related voxels in the fusiform

gyrus and lateral occipital region Error bars indicate SEM acrosssubjects in each group

Avidan et al 1159

for each subject (see Table 2 Part 2 for foci) The av-eraged percent signal change across all repetitions ofeach condition and across the left and right hemi-spheres of each subject are presented in Figure 6BAs in the control group the average fusiform gyrusactivation in the CP subjects exhibited a selective re-sponse for the Rubin-face compared to Rubin-vaseindicating that this region in CP subjects is indeedinvolved in computing global information about facesand is sensitive to the context not simply to the localcomponents of the display The similarity between theactivation profiles of the two groups in the fusiformgyrus was confirmed by a repeated-measure ANOVAwhich revealed only a main effect of stimulus type( p lt 01)

The results in the lateral occipital region were onceagain more variable across subjects in both groups (seeTable 2 Part 2) The ANOVA revealed a marginallysignificant main effect of group ( p lt 06) but not ofstimulus type and no significant interaction Note thatalthough not statistically significant the selectivity forthe Rubin-face compared to the Rubin-vase conditionin the lateral occipital focus was somewhat reduced inthe CP subjects compared to their controls This resultis analogous to the somewhat reduced face selectivityfound in the lateral occipital region in the conventionalmapping experiment (Figure 2A)

DISCUSSION

In a series of four functional imaging experimentswe have shown that individuals with CP exhibit nor-mal face- and object-related fMRI activation patternsin the ventral occipito-temporal cortex particularly inthe face-related fusiform gyrus (FFA) and largely nor-mal activations in the lateral occipital region despitetheir profound behavioral impairment Importantly thispattern was evident across different types of stimuli(eg line drawings and movie clips) and across differ-ent paradigms

In the first study in a conventional mapping experi-ment a normal pattern of BOLD activation was foundwhen subjects viewed line drawings of face and nonfacestimuli and performed a one-back memory task Thisparadigm has been used extensively in the literature andserves as a rigorous baseline to demarcate areas of ac-tivation in the ventral occipito-temporal cortex Onepossible interpretation of the normal activation is thatthe task was not sufficiently taxing for the CP individ-uals Behavioral data collected concurrently with theimage acquisition rules out this possibilitymdashalthoughthe task was easy for the controls this was not so forthe CP individuals who were clearly impaired on thistask Thus the normal face-related activation found inCP cannot merely be explained by the claim that the taskwas too simplistic for the CP subjects

A second possible interpretation and one that isdiametrically opposed to the first one offered above isthat the task was excessively difficult for the CP and itis this difficulty that drove the face-related activationparticularly in the fusiform gyrus of the CP subjects In-deed previous studies have shown that activation in theFFA increases as a function of working memory load(Druzgal amp DrsquoEsposito 2001) This explanation is nottenable either as the CP subjects continued to exhibitnormal face-related activation even when they were notrequired to perform any behavioral task during themotion pictures experiment Further research is clearlyneeded to unequivocally determine the contribution oftask difficulty to face-related activation in CP individ-uals We are now starting to parametrically explore theeffects of task difficulty in these CP individuals usingnew paradigms The critical point at present is thatnormal face-related activation is obtained at the verysame time that the behavioral impairment is manifest

The finding that face- and object-related activation inthe ventral visual cortex is apparently normal is replicat-ed in the motion picture experiment As mentionedabove this replication attests to the generalizability ofthe finding from the first conventional mapping exper-iment to more natural stimuli and viewing conditions

Importantly the normal signature of face-related ac-tivation in CP was also found when additional experi-mental manipulations were used CP subjects showednormal recovery from adaptation when different faceswere shown suggesting that neurons within these re-gions were sufficiently sensitive to differentiate betweenthe face exemplars used in this experiment Althoughone might think that normal adaptation is not thatsurprising given that the face stimuli used in this exper-iment were very different from each other it is impor-tant to bear in mind that CP subjects were actuallyimpaired in discriminating between such stimulimdashinthe conventional mapping experiment similar stimuliwere used and the CP subjects were impaired suggest-ing that the stimuli were confusing and not as percep-tually dissimilar for them as one might assume Of noteis that in addition to the adaptation for faces found inthe present experiment CP subjects also exhibitednormal adaptation levels in the fusiform gyrus and otherface-related regions for nonface stimuli suggestingthat similar to control subjects and replicating previousfindings these regions are involved in processing otherobject categories in addition to faces (Avidan et al 2002Haxby Gobbini et al 2001)

Although normal adaptation for faces is apparentdetermining its origin is less obvious Because the exactsame picture of an individual face was repeated duringthe adaptation condition we cannot determine at thisstage whether the adaptation found in the fusiformgyrus was due to the repetition of the exact same shape(ie based on perceptual geometry or structure) or therepetition of the same facial identity Differentiating

1160 Journal of Cognitive Neuroscience Volume 17 Number 7

between these alternatives and establishing whetherneural activity in the FFA is sufficient for normal rep-resentation of fine distinctions among individual faceswill require further experiments in which more subtlevariations will be made to face stimuli The use of anevent-related adaptation paradigm in such future ex-periments will also minimize possible attentional varia-tions that are inherent in the block-design adaptationparadigm employed here

Finally in the last experiment similar to control sub-jects once again CP subjects also exhibited a selectiveBOLD signal for Rubin-face compared to Rubin-vasestimuli in the fusiform gyrus Importantly these stimulicontain similar local elements but induce two differentvisual percepts either of a face or of a vase dependingon figurendashground segmentation and global integrationand shape processing Hence selectivity for the Rubin-face over the Rubin-vase stimuli implies that as in thecontrol group CP subjectsrsquo face-related activation ismodulated by global grouping processes and is notsimply induced by the representation of local face parts(Hasson Hendler et al 2001)

In sum the CP subjects exhibit normal face- andobject-related activation patterns in multiple regions ofthe ventral occipito-temporal cortex under various ex-perimental paradigms Of note however is that groupdifferences between CP subjects and their controls werefound in regions outside the occipito-temporal cortexindicating that the absence of group differences is notmerely a lack of statistical power Taken together thesefindings suggest that normal face-related activation inventral occipito-temporal regions as measured withfMRI is not sufficient to ensure intact face processingand does not necessarily reflect normal face represen-tation within those regions

Can the BehaviorndashBOLD Dissociationbe Reconciled

The central result is that the BOLD activation in CP inventral occipito-temporal regions is largely not differen-tiable from that of the control subjects notwithstandingtheir marked behavioral impairment How might thisdissociation be explained

It has been previously suggested that the FFA ismainly involved in face detection that is discriminatingbetween faces and stimuli from other object categories(Tong et al 2000) an ability that is largely preserved inCP The apparently normal FFA activation in the CPsubjects might then reflect the intact involvement ofthe FFA in deriving a rather coarse and rudimentarystructural representation which suffices for some tasks(ie detection) but not for more taxing tasks such asidentification Evidence compatible with this possibleinterpretation suggests that the final stage of face iden-tification is not completed at the level of high-order

visual face-related regions such as the FFA rathersuccessful recognition requires activation in more ante-rior regions of the temporal lobe where a more abstractor semantic representation of faces is derived (HaxbyHoffman et al 2000 Leveroni et al 2000 SergentOhta amp Macdonald 1992) Further support for the roleof anterior temporal regions in mediating face identifi-cation and particularly for being a site of facial memory(long-term representation) comes from lesion studiesFor example Barton and Cherkasova (2003) foundthat patients with anterior temporal lesions were themost impaired in a task that required generating long-term representations of famous faces as in mentalimagery compared to patients with more posterioroccipito-temporal lesions Similarly patients with ante-rior temporal lobectomy (following intractable epilepsy)were found to be impaired in identifying familiar faces(Glosser Salvucci amp Chiaravalloti 2003) The source ofthe behavioral problem in CP might then arise in theabnormal propagation of activation from the intact FFAto more anterior temporal regions

However the idea that the FFA only mediates facedetection is not universally accepted and the cleandivision of labor between the FFA and more anteriorregions might not be totally correct Although notdenying the contribution of more anterior regions toface processing and particularly to face identificationrecent experimental evidence suggests that the FFA isinvolved not only in face detection but also in faceidentification Thus normal subjects exhibited differen-tial responses for face detection and identification in thefusiform gyrus with the two processes resulting in eitherdifferential fMRI amplitude (Grill-Spector et al 2004) ortemporal scales (Puce Allison et al 1999 and seeSugase Yamane Ueno amp Kawano 1999 for similarfindings from single-unit recordings in monkeys) Thesedifferential responses in the fusiform gyrus may poten-tially be the outcome of feedback from more anteriorregions (as above) andor from additional computationtaking place within the fusiform gyrus itself Which ofthese is perturbed in CP remains unclear and they arenot mutually exclusive either Propagation to or fromanterior regions may be affected in CP However recallthat CP subjects exhibit a clear deficit not only in faceidentification (naming of famous faces) but also indiscrimination of unfamiliar faces (Figure 1B) (see alsoBehrmann et al 2005) This deficit in the more per-ceptual (not just memorial) aspects of face processingimplicates posterior occipito-temporal regions ratherthan exclusively the more anterior areas (Barton PressKeenan amp OrsquoConnor 2002 Wada amp Yamamoto 2001)It remains a possibility therefore that the fusiformactivation itself may be abnormal in CP Because thecurrent study employed detection and discriminationstasks the abnormality may still be uncovered undermore challenging conditions such as face identificationor other fine-grained face tasks

Avidan et al 1161

We also cannot rule out the possibility that differ-ences in the activation patterns in the fusiform gyrusandor other parts of the face processing networkbetween CP and normal control subjects might exist inthe temporal domain Such differences could underliethe behavioral dysfunction of CP subjects but may notbe evident with the temporal resolution afforded byfMRI (in the order of seconds) Some evidence sup-porting this view comes from ERP studies showing thatthe relatively early N170 potential which shows faceselectivity in normal subjects failed to show such se-lectivity in CP subjects (Kress amp Daum 2003b BentinDeouell et al 1999)

Finally it might be that the small differences inBOLD activation in CP subjects such as the slight re-duction in face selectivity in the lateral occipital region(eg Figures 2A and 6B) might be lsquolsquoamplifiedrsquorsquo whentranslated into behavioral performance If so the lateraloccipital region would serve as a critical link in the faceprocessing network Further support for the role ofthe occipital region in face processing comes fromstudies showing that acquired prosopagnosic patientsoften have lesions that involve not only the fusiformgyrus but also the lateral occipital region (Barton et al2002 Wada amp Yamamoto 2001) Moreover brain dam-age to this region but not to the FFA in some individ-uals may be associated with severe prosopagnosia(Rossion Caldara et al 2003)

As is evident there are a host of potential expla-nations to account for the dissociation between the im-paired behavioral profile and the apparently normalBOLD activation pattern We are currently exploringthese issues using new imaging experiments and hopethat these will uncover in greater detail the mecha-nisms giving rise to CP Suffice it to say that no one ex-planation captures the BOLDndashbehavior dissociationWhat is crucial however is that a simple assignmentof face processing to the FFA is no longer defensibleand a deeper examination of the computational prop-erties of the FFA and associated regions is demanded bythese findings

Face-related Activation Outside theOccipito-temporal Cortex

In the current study we acquired whole-brain functionalscans and thus were able to explore differences be-tween the CP and control groups in the responsepatterns in all regions of the brain not only in theoccipito-temporal regions A particularly novel findingwas that robust bilateral face-related activation wasevident in prefrontal regions (precentral sulcus inferiorfrontal sulcus anterior lateral sulcus) in the CP groupIn contrast the control group only exhibited right-lateralized activation in the precentral sulcus (Figure 3)Studies of normal individuals performing workingmemory tasks have revealed face-related activation in

prefrontal regions (Druzgal amp DrsquoEsposito 2003 Salaet al 2003 Haxby Petit et al 2000) and some ERPstudies have detected some inferior prefrontal cortexsites that generate small but specific face-specific re-sponses (Allison Puce Spencer amp McCarthy 1999)

The poor performance exhibited by CP subjectscompared to their controls on the one-back memorytask specifically for face stimuli is consistent with theinterpretation of this prefrontal activity being related toworking memory However given that the conven-tional face and object mapping experiment was notspecifically designed to look at the different componentsthat usually comprise a memory study (encoding mem-ory delay response) further research that specificallyand parametrically manipulates the memory load offaces and other object stimuli is required in order todetermine the role of this activation in face processing inCP subjects

This study has exploited the unique possibility ofexamining the activation of the fusiform gyrus andother cortical regions in an unusual group of individ-uals with CP Delineating the neural mechanism thatgives rise to abnormal behavior and by logical infer-ence normal behavior is a critical goal of neurosci-ence As might be predicted the correspondencesbetween the brain and behavior are not transparentand in this study are dissociated with the behavioralimpairment being evident concomitant with normalFFA activation Taken together however this studyhas indicated patterns of similarities and patterns ofpotential difference in this population of CP individ-uals relative to their normal counterparts The challengethat lies before us is to use these patterns to explainthe face processing deficit in detail

METHODS

Subjects

Four healthy CP subjects (2 men) aged between 29 and60 years with no discernable cortical lesion or anyhistory of neurological disease participated in all ex-periments (for details see Table 1) Another CP sub-

Table 1 Biographical Information about CP Subjects

Congenital Prosopagnosia (CP) Subjects

Subjectsrsquo Initials Sex Age

TM M 27

KM F 60

MT M 41

BE F 29

KM and TM are a mother and a son

1162 Journal of Cognitive Neuroscience Volume 17 Number 7

ject (NI male aged 40) was tested behaviorally (seeBehrmann et al 2005) but because we could not ac-quire his imaging data he was excluded from thepresent article Seven of the 12 control subjects whoparticipated in the behavioral studies along with threenew participants completed the imaging experimentsThe control group included at least two age- and sex-matched controls for each CP individual All CP andcontrol subjects were right-handed and had normal orcorrected-to-normal vision All subjects consented to par-ticipate in the experiments and the protocol was ap-proved by the Institutional Review Boards of CarnegieMellon University and the University of Pittsburgh

Imaging Experiments

Visual Stimulation

Visual stimuli were generated using the E-prime IFISsoftware (Psychology Software Tools Pittsburgh PAUSA) and projected via LCD to a screen located in theback of the scanner bore behind the subjectrsquos head Sub-jects viewed the stimuli through a tilted mirror mountedabove their eyes on the head coil Before the scan sub-

jects were familiarized with the visual stimuli and tasksof each of the imaging experiments

MRI Setup

All subjects were scanned in a 3-T Siemens Allegrascanner equipped with a standard head coil Subjectsparticipated in one scanning session that lasted 15 to2 hr in which all the functional and anatomical scanswere acquired The order of the functional experimentswithin the scanning session was generally constant forall subjects BOLD contrast was acquired using gradient-echo echo-planar imaging sequence Specific parame-ters TR = 3000 msec TE = 35 msec flip angle = 908FOV = 210 210 mm2 matrix size 64 64 35 axialslices 3 mm thickness no gap High-resolution anatom-ical scans (T1-weighted 3-D MPRAGE) were also acquiredto allow accurate cortical segmentation reconstructionand volume-based statistical analysis Specific parame-ters TE = 349 flip angle = 88 FOV = 256 256 mm2matrix size = 256 256 slice thickness = 1 mm numberof slices = 160ndash192 orientation of slices was eitherhorizontal or sagittal For three of the CP subjects the

Table 2 Laterality in force-selective activation

1 Talairach Coordinates Fusiform Gyrus

Left Hemisphere Right Hemisphere

x y z x y z

CP subjects 40 plusmn 2 53 plusmn 3 19 plusmn 2 38 plusmn 5 47 plusmn 8 17 plusmn 3

Control subjects 37 plusmn 4 41 plusmn 5 18 plusmn 4 36 plusmn 2 40 plusmn 7 17 plusmn 3

2 Activation Foci

Localizer Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 4 ndash ndash

Control subjects 9 1 ndash 4 2 1

Rubin FacendashVase Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 1 3 ndash

Control subjects 9 ndash ndash 3 2 4

Part 1 Talairach coordinates of the face-related activation in the fusiform gyrus of CP and control subjects extracted from the conventionalface and object mapping experiment Part 2 Number of CP and control subjects that showed bilateral right-lateralized or left-lateralizedface-related activation in the fusiform gyrus and in the lateral occipital region in the conventional face and object mapping experiment andin the Rubin facendashvase experiment

Avidan et al 1163

3-D brain reconstruction was performed on imagespreviously acquired on a 15-T GE scanner

Experiments

Conventional Face and Object Mapping Experiment

Stimuli included line drawings of faces buildings com-mon objects and geometric patterns presented in ashort block design (for details see Levy et al 2001Figure 1A) Each epoch lasted 9 sec and contained ninestimuli from the same category there were eight differ-ent images and one immediate repetition of one of thestimuli Each stimulus was presented for 800 msecfollowed by a 200-msec interval of blank screen Therewere seven repetitions of each condition and their orderwas pseudorandomized across subjects Visual epochswere interleaved with 6-sec blank epochs The experi-ment lasted 450 sec and started and ended with extend-ed blank epochs of 27 and 9 sec respectively Subjectswere instructed to fixate on a red central fixation dot toperform a one-back memory task and to press a key ona response glove to indicate the repeated image Bothaccuracy and reaction time were recorded

The Motion Pictures Experiment

The experiment lasted 10 minutes and was composed of32 epochs (15 sec long) of movie clips each containingimages from one of four possible categories people invarious situations navigation of the camera through citybuildings or through open fields and miscellaneousimages of objects from different categories (machinesfalling water etc for details see Hasson Nir et al2004 Figure 4) The experiment started with a 51-secblank period followed by 9 sec of pattern stimuli andended with a 1-minute blank Subjects were simplyinstructed to watch the movie clips

The Adaptation Experiment

The experiment included line drawing images fromthree different categories faces buildings and cars(for details see Avidan et al 2002 Figure 5A) Stimuliwere presented in 12-sec epochs that contained either12 different images (lsquolsquodifferentrsquorsquo condition) from thesame category or 12 repetitions of the same identicalstimulus (lsquolsquoidenticalrsquorsquo condition) Within epochs eachstimulus was presented for 800 msec followed by200 msec of fixation only The experiment lasted480 sec and started and ended with a long blank pe-riod of 21 and 15 sec respectively The first visualblock always contained geometric patterns and was ex-cluded from the analysis Visual epochs were interleavedwith 6-sec blank epochs and the order of visual epochswas pseudorandomized across subjects Subjects wereinstructed to continuously fixate a central red dot and

to covertly categorize the images as follows man wom-an or a child for the face stimuli apartment townhouseor public facility for the buildings and private car truckor bus for the vehicles

The Rubin FacendashVase Experiment

The experiment included modified versions of the Rubinfacendashvase stimuli as well as line drawings of front facesand household objects (for details see Hasson Hendleret al 2001 Figure 6A) The Rubin facendashvase stimuli weremodified so that subjectsrsquo perception was biased towardsone perceptual state or the other This was done by firstuniformly coloring the vase area within each of thestimuli and placing stripes over the profile Stimuli werethen placed either over a striped background (biastowards perception of the vase due to closure of thevase stimulus) or over a uniform-color background (biastowards perception of the face due to closure of theface stimulus) Critically the Rubin face and vase sharethe same local contours but give rise to different globalperceptions In order to ensure that subjectsrsquo perceptionremained biased to one perceptual interpretation (ei-ther the profile or vase) during the experiment stimuliwere presented briefly (200 msec) and were immediate-ly followed by a masking grid that was presented for800 msec Stimuli were presented in 9-sec blocks thatwere interleaved with 6-sec blanks Each condition wasrepeated eight times and the order of presentation waspseudorandomized across subjects Subjects maintainedcentral fixation throughout the experiment and per-formed a one-back memory task however due to atechnical failure their responses were not recorded Theexperiment lasted 507 sec and started and ended withextended blank periods

Data Analysis

Data analysis was performed using the BrainVoyager48 software package (Brain Innovation MaastrichtNetherlands) and complementary in-house softwareDetailed description of the data analysis is provided inHasson Harel et al (2003) Briefly for each subject thecortical surface was reconstructed and unfolded intothe flattened format Functional data for each subjectfrom each experiment were analyzed separately Pre-processing of functional scans included 3-D motioncorrection and filtering of low frequencies up to fivecycles per experiment (slow drift) Statistical analysiswas based on the GLM For each experiment each ofthe conditions (except for blank) was defined as a sepa-rate predictor a boxcar shape was used to model eachpredictor and a 3-sec lag was assumed Percent signalchange for each subject in each experiment (except forthe motion pictures experiment) was calculated as thepercent activation from a blank baseline For the mo-

1164 Journal of Cognitive Neuroscience Volume 17 Number 7

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

dependently for each subject using the conventionalmapping experiment Figure 5B shows the activationlevels within the face-related foci in the fusiform gyrusand lateral occipital region averaged across all repeti-tions of each experimental condition and across thetwo hemispheres for the two groups Again despitethe fact that the face-related regions were selected usingan external localizer CP individuals exhibited clear faceselectivity in the lsquolsquodifferentrsquorsquo conditions in these ROIsNote the similarity between the groups in the magni-tude of the adaptation that is the signal decrease fol-lowing face repetition (dark gray compared with lightgray bars) suggesting that face representations in CPsubjects are indeed sensitive to face repetition Theface-related regions also exhibit a strong adaptationeffect for the nonface stimuli (buildings and cars) in-dicating that like normal subjects these regions in theCP subjects contain neurons that are highly selectivefor those stimuli (Avidan et al 2002) These observa-tions were confirmed in a repeated-measure ANOVAin which the only significant effect in both regions wasa main effect of stimulus type ( p lt 0001) There wasno main effect of group nor an interaction betweenthe factors

Figure 5C shows a similar analysis for the building-related focus in the anterior collateral sulcus This regionexhibited strong adaptation for the building stimuli aswell as for the other object categories in both groupsOnce again the ANOVA revealed only a main effect ofstimulus type ( p lt 0001) The strong building selectiv-ity in the collateral sulcus combined with the findingthat this region exhibits similar adaptation levels forall stimulus categories including faces further suggeststhat the entire occipito-temporal object representationnetwork in CP subjects is intact and mirrors that of thenormal control subjects

Rubin FacendashVase Experiment

One possible interpretation of the apparently normalface-selective BOLD response in the CP subjects foundin the previous experiments is that it reflects a responseto the presence of face parts (eyes mouth and nose) inthe absence of an intact representation of the face as awhole To examine whether the activation in CP ismerely a product of a part-representation we employeda paradigm used previously which utilizes the famousRubin facendashvase illusion to demonstrate evidence forglobal face representations in normal subjects (HassonHendler Ben Bashat amp Malach 2001) Critically in theRubin facendashvase illusion similar local elements are al-ways present However depending on figurendashgroundsegmentation and global integration the local elementscan induce two different visual percepts either of a faceor of a vase (Figure 6A) If face-selective activation in thefusiform gyrus and other face-related regions depends

only on processing of the local elements then activa-tion in these regions should be similar in both theRubin-face and Rubin-vase conditions (same local ele-ments are present) However if face representation inthe fusiform gyrus and other face-related regions isaffected by the global integration of the face elementsinto the percept of either a face or a vase then increasedactivation should be found for the Rubin-face perceptualstates compared to the Rubin-vase perceptual states In-deed in normal subjects selective activation within face-related regions has been previously demonstrated forthe Rubin-face compared to the Rubin-vase stimuli sug-gesting that normal face-related activation is modulatedby global grouping processes and is not only induced bythe representation of local face parts (Hasson Hendleret al 2001)

Here face-selective regions were independently lo-calized initially for each subject by contrasting theepochs of frontal faces with those of household ob-jects (Figure 6A) Data from one control subject wereexcluded from this experiment as they were too noisyto localize face-related activation The activation profilewas extracted for the Rubin-face and Rubin-vase stimulifrom both the fusiform gyrus and lateral occipital foci

Figure 6 The Rubin facendashvase experiment (A) Examples of the

stimuli including the Rubin facendashvase stimuli (left) and frontal faces

and household objects (localizer stimuli) used to independently

localize face-related voxels (in the experiment the uniform surfacesshown here in gray were colored in light pink) (B) Activation profiles

for the Rubin facendashvase conditions in face-related voxels in the fusiform

gyrus and lateral occipital region Error bars indicate SEM acrosssubjects in each group

Avidan et al 1159

for each subject (see Table 2 Part 2 for foci) The av-eraged percent signal change across all repetitions ofeach condition and across the left and right hemi-spheres of each subject are presented in Figure 6BAs in the control group the average fusiform gyrusactivation in the CP subjects exhibited a selective re-sponse for the Rubin-face compared to Rubin-vaseindicating that this region in CP subjects is indeedinvolved in computing global information about facesand is sensitive to the context not simply to the localcomponents of the display The similarity between theactivation profiles of the two groups in the fusiformgyrus was confirmed by a repeated-measure ANOVAwhich revealed only a main effect of stimulus type( p lt 01)

The results in the lateral occipital region were onceagain more variable across subjects in both groups (seeTable 2 Part 2) The ANOVA revealed a marginallysignificant main effect of group ( p lt 06) but not ofstimulus type and no significant interaction Note thatalthough not statistically significant the selectivity forthe Rubin-face compared to the Rubin-vase conditionin the lateral occipital focus was somewhat reduced inthe CP subjects compared to their controls This resultis analogous to the somewhat reduced face selectivityfound in the lateral occipital region in the conventionalmapping experiment (Figure 2A)

DISCUSSION

In a series of four functional imaging experimentswe have shown that individuals with CP exhibit nor-mal face- and object-related fMRI activation patternsin the ventral occipito-temporal cortex particularly inthe face-related fusiform gyrus (FFA) and largely nor-mal activations in the lateral occipital region despitetheir profound behavioral impairment Importantly thispattern was evident across different types of stimuli(eg line drawings and movie clips) and across differ-ent paradigms

In the first study in a conventional mapping experi-ment a normal pattern of BOLD activation was foundwhen subjects viewed line drawings of face and nonfacestimuli and performed a one-back memory task Thisparadigm has been used extensively in the literature andserves as a rigorous baseline to demarcate areas of ac-tivation in the ventral occipito-temporal cortex Onepossible interpretation of the normal activation is thatthe task was not sufficiently taxing for the CP individ-uals Behavioral data collected concurrently with theimage acquisition rules out this possibilitymdashalthoughthe task was easy for the controls this was not so forthe CP individuals who were clearly impaired on thistask Thus the normal face-related activation found inCP cannot merely be explained by the claim that the taskwas too simplistic for the CP subjects

A second possible interpretation and one that isdiametrically opposed to the first one offered above isthat the task was excessively difficult for the CP and itis this difficulty that drove the face-related activationparticularly in the fusiform gyrus of the CP subjects In-deed previous studies have shown that activation in theFFA increases as a function of working memory load(Druzgal amp DrsquoEsposito 2001) This explanation is nottenable either as the CP subjects continued to exhibitnormal face-related activation even when they were notrequired to perform any behavioral task during themotion pictures experiment Further research is clearlyneeded to unequivocally determine the contribution oftask difficulty to face-related activation in CP individ-uals We are now starting to parametrically explore theeffects of task difficulty in these CP individuals usingnew paradigms The critical point at present is thatnormal face-related activation is obtained at the verysame time that the behavioral impairment is manifest

The finding that face- and object-related activation inthe ventral visual cortex is apparently normal is replicat-ed in the motion picture experiment As mentionedabove this replication attests to the generalizability ofthe finding from the first conventional mapping exper-iment to more natural stimuli and viewing conditions

Importantly the normal signature of face-related ac-tivation in CP was also found when additional experi-mental manipulations were used CP subjects showednormal recovery from adaptation when different faceswere shown suggesting that neurons within these re-gions were sufficiently sensitive to differentiate betweenthe face exemplars used in this experiment Althoughone might think that normal adaptation is not thatsurprising given that the face stimuli used in this exper-iment were very different from each other it is impor-tant to bear in mind that CP subjects were actuallyimpaired in discriminating between such stimulimdashinthe conventional mapping experiment similar stimuliwere used and the CP subjects were impaired suggest-ing that the stimuli were confusing and not as percep-tually dissimilar for them as one might assume Of noteis that in addition to the adaptation for faces found inthe present experiment CP subjects also exhibitednormal adaptation levels in the fusiform gyrus and otherface-related regions for nonface stimuli suggestingthat similar to control subjects and replicating previousfindings these regions are involved in processing otherobject categories in addition to faces (Avidan et al 2002Haxby Gobbini et al 2001)

Although normal adaptation for faces is apparentdetermining its origin is less obvious Because the exactsame picture of an individual face was repeated duringthe adaptation condition we cannot determine at thisstage whether the adaptation found in the fusiformgyrus was due to the repetition of the exact same shape(ie based on perceptual geometry or structure) or therepetition of the same facial identity Differentiating

1160 Journal of Cognitive Neuroscience Volume 17 Number 7

between these alternatives and establishing whetherneural activity in the FFA is sufficient for normal rep-resentation of fine distinctions among individual faceswill require further experiments in which more subtlevariations will be made to face stimuli The use of anevent-related adaptation paradigm in such future ex-periments will also minimize possible attentional varia-tions that are inherent in the block-design adaptationparadigm employed here

Finally in the last experiment similar to control sub-jects once again CP subjects also exhibited a selectiveBOLD signal for Rubin-face compared to Rubin-vasestimuli in the fusiform gyrus Importantly these stimulicontain similar local elements but induce two differentvisual percepts either of a face or of a vase dependingon figurendashground segmentation and global integrationand shape processing Hence selectivity for the Rubin-face over the Rubin-vase stimuli implies that as in thecontrol group CP subjectsrsquo face-related activation ismodulated by global grouping processes and is notsimply induced by the representation of local face parts(Hasson Hendler et al 2001)

In sum the CP subjects exhibit normal face- andobject-related activation patterns in multiple regions ofthe ventral occipito-temporal cortex under various ex-perimental paradigms Of note however is that groupdifferences between CP subjects and their controls werefound in regions outside the occipito-temporal cortexindicating that the absence of group differences is notmerely a lack of statistical power Taken together thesefindings suggest that normal face-related activation inventral occipito-temporal regions as measured withfMRI is not sufficient to ensure intact face processingand does not necessarily reflect normal face represen-tation within those regions

Can the BehaviorndashBOLD Dissociationbe Reconciled

The central result is that the BOLD activation in CP inventral occipito-temporal regions is largely not differen-tiable from that of the control subjects notwithstandingtheir marked behavioral impairment How might thisdissociation be explained

It has been previously suggested that the FFA ismainly involved in face detection that is discriminatingbetween faces and stimuli from other object categories(Tong et al 2000) an ability that is largely preserved inCP The apparently normal FFA activation in the CPsubjects might then reflect the intact involvement ofthe FFA in deriving a rather coarse and rudimentarystructural representation which suffices for some tasks(ie detection) but not for more taxing tasks such asidentification Evidence compatible with this possibleinterpretation suggests that the final stage of face iden-tification is not completed at the level of high-order

visual face-related regions such as the FFA rathersuccessful recognition requires activation in more ante-rior regions of the temporal lobe where a more abstractor semantic representation of faces is derived (HaxbyHoffman et al 2000 Leveroni et al 2000 SergentOhta amp Macdonald 1992) Further support for the roleof anterior temporal regions in mediating face identifi-cation and particularly for being a site of facial memory(long-term representation) comes from lesion studiesFor example Barton and Cherkasova (2003) foundthat patients with anterior temporal lesions were themost impaired in a task that required generating long-term representations of famous faces as in mentalimagery compared to patients with more posterioroccipito-temporal lesions Similarly patients with ante-rior temporal lobectomy (following intractable epilepsy)were found to be impaired in identifying familiar faces(Glosser Salvucci amp Chiaravalloti 2003) The source ofthe behavioral problem in CP might then arise in theabnormal propagation of activation from the intact FFAto more anterior temporal regions

However the idea that the FFA only mediates facedetection is not universally accepted and the cleandivision of labor between the FFA and more anteriorregions might not be totally correct Although notdenying the contribution of more anterior regions toface processing and particularly to face identificationrecent experimental evidence suggests that the FFA isinvolved not only in face detection but also in faceidentification Thus normal subjects exhibited differen-tial responses for face detection and identification in thefusiform gyrus with the two processes resulting in eitherdifferential fMRI amplitude (Grill-Spector et al 2004) ortemporal scales (Puce Allison et al 1999 and seeSugase Yamane Ueno amp Kawano 1999 for similarfindings from single-unit recordings in monkeys) Thesedifferential responses in the fusiform gyrus may poten-tially be the outcome of feedback from more anteriorregions (as above) andor from additional computationtaking place within the fusiform gyrus itself Which ofthese is perturbed in CP remains unclear and they arenot mutually exclusive either Propagation to or fromanterior regions may be affected in CP However recallthat CP subjects exhibit a clear deficit not only in faceidentification (naming of famous faces) but also indiscrimination of unfamiliar faces (Figure 1B) (see alsoBehrmann et al 2005) This deficit in the more per-ceptual (not just memorial) aspects of face processingimplicates posterior occipito-temporal regions ratherthan exclusively the more anterior areas (Barton PressKeenan amp OrsquoConnor 2002 Wada amp Yamamoto 2001)It remains a possibility therefore that the fusiformactivation itself may be abnormal in CP Because thecurrent study employed detection and discriminationstasks the abnormality may still be uncovered undermore challenging conditions such as face identificationor other fine-grained face tasks

Avidan et al 1161

We also cannot rule out the possibility that differ-ences in the activation patterns in the fusiform gyrusandor other parts of the face processing networkbetween CP and normal control subjects might exist inthe temporal domain Such differences could underliethe behavioral dysfunction of CP subjects but may notbe evident with the temporal resolution afforded byfMRI (in the order of seconds) Some evidence sup-porting this view comes from ERP studies showing thatthe relatively early N170 potential which shows faceselectivity in normal subjects failed to show such se-lectivity in CP subjects (Kress amp Daum 2003b BentinDeouell et al 1999)

Finally it might be that the small differences inBOLD activation in CP subjects such as the slight re-duction in face selectivity in the lateral occipital region(eg Figures 2A and 6B) might be lsquolsquoamplifiedrsquorsquo whentranslated into behavioral performance If so the lateraloccipital region would serve as a critical link in the faceprocessing network Further support for the role ofthe occipital region in face processing comes fromstudies showing that acquired prosopagnosic patientsoften have lesions that involve not only the fusiformgyrus but also the lateral occipital region (Barton et al2002 Wada amp Yamamoto 2001) Moreover brain dam-age to this region but not to the FFA in some individ-uals may be associated with severe prosopagnosia(Rossion Caldara et al 2003)

As is evident there are a host of potential expla-nations to account for the dissociation between the im-paired behavioral profile and the apparently normalBOLD activation pattern We are currently exploringthese issues using new imaging experiments and hopethat these will uncover in greater detail the mecha-nisms giving rise to CP Suffice it to say that no one ex-planation captures the BOLDndashbehavior dissociationWhat is crucial however is that a simple assignmentof face processing to the FFA is no longer defensibleand a deeper examination of the computational prop-erties of the FFA and associated regions is demanded bythese findings

Face-related Activation Outside theOccipito-temporal Cortex

In the current study we acquired whole-brain functionalscans and thus were able to explore differences be-tween the CP and control groups in the responsepatterns in all regions of the brain not only in theoccipito-temporal regions A particularly novel findingwas that robust bilateral face-related activation wasevident in prefrontal regions (precentral sulcus inferiorfrontal sulcus anterior lateral sulcus) in the CP groupIn contrast the control group only exhibited right-lateralized activation in the precentral sulcus (Figure 3)Studies of normal individuals performing workingmemory tasks have revealed face-related activation in

prefrontal regions (Druzgal amp DrsquoEsposito 2003 Salaet al 2003 Haxby Petit et al 2000) and some ERPstudies have detected some inferior prefrontal cortexsites that generate small but specific face-specific re-sponses (Allison Puce Spencer amp McCarthy 1999)

The poor performance exhibited by CP subjectscompared to their controls on the one-back memorytask specifically for face stimuli is consistent with theinterpretation of this prefrontal activity being related toworking memory However given that the conven-tional face and object mapping experiment was notspecifically designed to look at the different componentsthat usually comprise a memory study (encoding mem-ory delay response) further research that specificallyand parametrically manipulates the memory load offaces and other object stimuli is required in order todetermine the role of this activation in face processing inCP subjects

This study has exploited the unique possibility ofexamining the activation of the fusiform gyrus andother cortical regions in an unusual group of individ-uals with CP Delineating the neural mechanism thatgives rise to abnormal behavior and by logical infer-ence normal behavior is a critical goal of neurosci-ence As might be predicted the correspondencesbetween the brain and behavior are not transparentand in this study are dissociated with the behavioralimpairment being evident concomitant with normalFFA activation Taken together however this studyhas indicated patterns of similarities and patterns ofpotential difference in this population of CP individ-uals relative to their normal counterparts The challengethat lies before us is to use these patterns to explainthe face processing deficit in detail

METHODS

Subjects

Four healthy CP subjects (2 men) aged between 29 and60 years with no discernable cortical lesion or anyhistory of neurological disease participated in all ex-periments (for details see Table 1) Another CP sub-

Table 1 Biographical Information about CP Subjects

Congenital Prosopagnosia (CP) Subjects

Subjectsrsquo Initials Sex Age

TM M 27

KM F 60

MT M 41

BE F 29

KM and TM are a mother and a son

1162 Journal of Cognitive Neuroscience Volume 17 Number 7

ject (NI male aged 40) was tested behaviorally (seeBehrmann et al 2005) but because we could not ac-quire his imaging data he was excluded from thepresent article Seven of the 12 control subjects whoparticipated in the behavioral studies along with threenew participants completed the imaging experimentsThe control group included at least two age- and sex-matched controls for each CP individual All CP andcontrol subjects were right-handed and had normal orcorrected-to-normal vision All subjects consented to par-ticipate in the experiments and the protocol was ap-proved by the Institutional Review Boards of CarnegieMellon University and the University of Pittsburgh

Imaging Experiments

Visual Stimulation

Visual stimuli were generated using the E-prime IFISsoftware (Psychology Software Tools Pittsburgh PAUSA) and projected via LCD to a screen located in theback of the scanner bore behind the subjectrsquos head Sub-jects viewed the stimuli through a tilted mirror mountedabove their eyes on the head coil Before the scan sub-

jects were familiarized with the visual stimuli and tasksof each of the imaging experiments

MRI Setup

All subjects were scanned in a 3-T Siemens Allegrascanner equipped with a standard head coil Subjectsparticipated in one scanning session that lasted 15 to2 hr in which all the functional and anatomical scanswere acquired The order of the functional experimentswithin the scanning session was generally constant forall subjects BOLD contrast was acquired using gradient-echo echo-planar imaging sequence Specific parame-ters TR = 3000 msec TE = 35 msec flip angle = 908FOV = 210 210 mm2 matrix size 64 64 35 axialslices 3 mm thickness no gap High-resolution anatom-ical scans (T1-weighted 3-D MPRAGE) were also acquiredto allow accurate cortical segmentation reconstructionand volume-based statistical analysis Specific parame-ters TE = 349 flip angle = 88 FOV = 256 256 mm2matrix size = 256 256 slice thickness = 1 mm numberof slices = 160ndash192 orientation of slices was eitherhorizontal or sagittal For three of the CP subjects the

Table 2 Laterality in force-selective activation

1 Talairach Coordinates Fusiform Gyrus

Left Hemisphere Right Hemisphere

x y z x y z

CP subjects 40 plusmn 2 53 plusmn 3 19 plusmn 2 38 plusmn 5 47 plusmn 8 17 plusmn 3

Control subjects 37 plusmn 4 41 plusmn 5 18 plusmn 4 36 plusmn 2 40 plusmn 7 17 plusmn 3

2 Activation Foci

Localizer Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 4 ndash ndash

Control subjects 9 1 ndash 4 2 1

Rubin FacendashVase Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 1 3 ndash

Control subjects 9 ndash ndash 3 2 4

Part 1 Talairach coordinates of the face-related activation in the fusiform gyrus of CP and control subjects extracted from the conventionalface and object mapping experiment Part 2 Number of CP and control subjects that showed bilateral right-lateralized or left-lateralizedface-related activation in the fusiform gyrus and in the lateral occipital region in the conventional face and object mapping experiment andin the Rubin facendashvase experiment

Avidan et al 1163

3-D brain reconstruction was performed on imagespreviously acquired on a 15-T GE scanner

Experiments

Conventional Face and Object Mapping Experiment

Stimuli included line drawings of faces buildings com-mon objects and geometric patterns presented in ashort block design (for details see Levy et al 2001Figure 1A) Each epoch lasted 9 sec and contained ninestimuli from the same category there were eight differ-ent images and one immediate repetition of one of thestimuli Each stimulus was presented for 800 msecfollowed by a 200-msec interval of blank screen Therewere seven repetitions of each condition and their orderwas pseudorandomized across subjects Visual epochswere interleaved with 6-sec blank epochs The experi-ment lasted 450 sec and started and ended with extend-ed blank epochs of 27 and 9 sec respectively Subjectswere instructed to fixate on a red central fixation dot toperform a one-back memory task and to press a key ona response glove to indicate the repeated image Bothaccuracy and reaction time were recorded

The Motion Pictures Experiment

The experiment lasted 10 minutes and was composed of32 epochs (15 sec long) of movie clips each containingimages from one of four possible categories people invarious situations navigation of the camera through citybuildings or through open fields and miscellaneousimages of objects from different categories (machinesfalling water etc for details see Hasson Nir et al2004 Figure 4) The experiment started with a 51-secblank period followed by 9 sec of pattern stimuli andended with a 1-minute blank Subjects were simplyinstructed to watch the movie clips

The Adaptation Experiment

The experiment included line drawing images fromthree different categories faces buildings and cars(for details see Avidan et al 2002 Figure 5A) Stimuliwere presented in 12-sec epochs that contained either12 different images (lsquolsquodifferentrsquorsquo condition) from thesame category or 12 repetitions of the same identicalstimulus (lsquolsquoidenticalrsquorsquo condition) Within epochs eachstimulus was presented for 800 msec followed by200 msec of fixation only The experiment lasted480 sec and started and ended with a long blank pe-riod of 21 and 15 sec respectively The first visualblock always contained geometric patterns and was ex-cluded from the analysis Visual epochs were interleavedwith 6-sec blank epochs and the order of visual epochswas pseudorandomized across subjects Subjects wereinstructed to continuously fixate a central red dot and

to covertly categorize the images as follows man wom-an or a child for the face stimuli apartment townhouseor public facility for the buildings and private car truckor bus for the vehicles

The Rubin FacendashVase Experiment

The experiment included modified versions of the Rubinfacendashvase stimuli as well as line drawings of front facesand household objects (for details see Hasson Hendleret al 2001 Figure 6A) The Rubin facendashvase stimuli weremodified so that subjectsrsquo perception was biased towardsone perceptual state or the other This was done by firstuniformly coloring the vase area within each of thestimuli and placing stripes over the profile Stimuli werethen placed either over a striped background (biastowards perception of the vase due to closure of thevase stimulus) or over a uniform-color background (biastowards perception of the face due to closure of theface stimulus) Critically the Rubin face and vase sharethe same local contours but give rise to different globalperceptions In order to ensure that subjectsrsquo perceptionremained biased to one perceptual interpretation (ei-ther the profile or vase) during the experiment stimuliwere presented briefly (200 msec) and were immediate-ly followed by a masking grid that was presented for800 msec Stimuli were presented in 9-sec blocks thatwere interleaved with 6-sec blanks Each condition wasrepeated eight times and the order of presentation waspseudorandomized across subjects Subjects maintainedcentral fixation throughout the experiment and per-formed a one-back memory task however due to atechnical failure their responses were not recorded Theexperiment lasted 507 sec and started and ended withextended blank periods

Data Analysis

Data analysis was performed using the BrainVoyager48 software package (Brain Innovation MaastrichtNetherlands) and complementary in-house softwareDetailed description of the data analysis is provided inHasson Harel et al (2003) Briefly for each subject thecortical surface was reconstructed and unfolded intothe flattened format Functional data for each subjectfrom each experiment were analyzed separately Pre-processing of functional scans included 3-D motioncorrection and filtering of low frequencies up to fivecycles per experiment (slow drift) Statistical analysiswas based on the GLM For each experiment each ofthe conditions (except for blank) was defined as a sepa-rate predictor a boxcar shape was used to model eachpredictor and a 3-sec lag was assumed Percent signalchange for each subject in each experiment (except forthe motion pictures experiment) was calculated as thepercent activation from a blank baseline For the mo-

1164 Journal of Cognitive Neuroscience Volume 17 Number 7

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

for each subject (see Table 2 Part 2 for foci) The av-eraged percent signal change across all repetitions ofeach condition and across the left and right hemi-spheres of each subject are presented in Figure 6BAs in the control group the average fusiform gyrusactivation in the CP subjects exhibited a selective re-sponse for the Rubin-face compared to Rubin-vaseindicating that this region in CP subjects is indeedinvolved in computing global information about facesand is sensitive to the context not simply to the localcomponents of the display The similarity between theactivation profiles of the two groups in the fusiformgyrus was confirmed by a repeated-measure ANOVAwhich revealed only a main effect of stimulus type( p lt 01)

The results in the lateral occipital region were onceagain more variable across subjects in both groups (seeTable 2 Part 2) The ANOVA revealed a marginallysignificant main effect of group ( p lt 06) but not ofstimulus type and no significant interaction Note thatalthough not statistically significant the selectivity forthe Rubin-face compared to the Rubin-vase conditionin the lateral occipital focus was somewhat reduced inthe CP subjects compared to their controls This resultis analogous to the somewhat reduced face selectivityfound in the lateral occipital region in the conventionalmapping experiment (Figure 2A)

DISCUSSION

In a series of four functional imaging experimentswe have shown that individuals with CP exhibit nor-mal face- and object-related fMRI activation patternsin the ventral occipito-temporal cortex particularly inthe face-related fusiform gyrus (FFA) and largely nor-mal activations in the lateral occipital region despitetheir profound behavioral impairment Importantly thispattern was evident across different types of stimuli(eg line drawings and movie clips) and across differ-ent paradigms

In the first study in a conventional mapping experi-ment a normal pattern of BOLD activation was foundwhen subjects viewed line drawings of face and nonfacestimuli and performed a one-back memory task Thisparadigm has been used extensively in the literature andserves as a rigorous baseline to demarcate areas of ac-tivation in the ventral occipito-temporal cortex Onepossible interpretation of the normal activation is thatthe task was not sufficiently taxing for the CP individ-uals Behavioral data collected concurrently with theimage acquisition rules out this possibilitymdashalthoughthe task was easy for the controls this was not so forthe CP individuals who were clearly impaired on thistask Thus the normal face-related activation found inCP cannot merely be explained by the claim that the taskwas too simplistic for the CP subjects

A second possible interpretation and one that isdiametrically opposed to the first one offered above isthat the task was excessively difficult for the CP and itis this difficulty that drove the face-related activationparticularly in the fusiform gyrus of the CP subjects In-deed previous studies have shown that activation in theFFA increases as a function of working memory load(Druzgal amp DrsquoEsposito 2001) This explanation is nottenable either as the CP subjects continued to exhibitnormal face-related activation even when they were notrequired to perform any behavioral task during themotion pictures experiment Further research is clearlyneeded to unequivocally determine the contribution oftask difficulty to face-related activation in CP individ-uals We are now starting to parametrically explore theeffects of task difficulty in these CP individuals usingnew paradigms The critical point at present is thatnormal face-related activation is obtained at the verysame time that the behavioral impairment is manifest

The finding that face- and object-related activation inthe ventral visual cortex is apparently normal is replicat-ed in the motion picture experiment As mentionedabove this replication attests to the generalizability ofthe finding from the first conventional mapping exper-iment to more natural stimuli and viewing conditions

Importantly the normal signature of face-related ac-tivation in CP was also found when additional experi-mental manipulations were used CP subjects showednormal recovery from adaptation when different faceswere shown suggesting that neurons within these re-gions were sufficiently sensitive to differentiate betweenthe face exemplars used in this experiment Althoughone might think that normal adaptation is not thatsurprising given that the face stimuli used in this exper-iment were very different from each other it is impor-tant to bear in mind that CP subjects were actuallyimpaired in discriminating between such stimulimdashinthe conventional mapping experiment similar stimuliwere used and the CP subjects were impaired suggest-ing that the stimuli were confusing and not as percep-tually dissimilar for them as one might assume Of noteis that in addition to the adaptation for faces found inthe present experiment CP subjects also exhibitednormal adaptation levels in the fusiform gyrus and otherface-related regions for nonface stimuli suggestingthat similar to control subjects and replicating previousfindings these regions are involved in processing otherobject categories in addition to faces (Avidan et al 2002Haxby Gobbini et al 2001)

Although normal adaptation for faces is apparentdetermining its origin is less obvious Because the exactsame picture of an individual face was repeated duringthe adaptation condition we cannot determine at thisstage whether the adaptation found in the fusiformgyrus was due to the repetition of the exact same shape(ie based on perceptual geometry or structure) or therepetition of the same facial identity Differentiating

1160 Journal of Cognitive Neuroscience Volume 17 Number 7

between these alternatives and establishing whetherneural activity in the FFA is sufficient for normal rep-resentation of fine distinctions among individual faceswill require further experiments in which more subtlevariations will be made to face stimuli The use of anevent-related adaptation paradigm in such future ex-periments will also minimize possible attentional varia-tions that are inherent in the block-design adaptationparadigm employed here

Finally in the last experiment similar to control sub-jects once again CP subjects also exhibited a selectiveBOLD signal for Rubin-face compared to Rubin-vasestimuli in the fusiform gyrus Importantly these stimulicontain similar local elements but induce two differentvisual percepts either of a face or of a vase dependingon figurendashground segmentation and global integrationand shape processing Hence selectivity for the Rubin-face over the Rubin-vase stimuli implies that as in thecontrol group CP subjectsrsquo face-related activation ismodulated by global grouping processes and is notsimply induced by the representation of local face parts(Hasson Hendler et al 2001)

In sum the CP subjects exhibit normal face- andobject-related activation patterns in multiple regions ofthe ventral occipito-temporal cortex under various ex-perimental paradigms Of note however is that groupdifferences between CP subjects and their controls werefound in regions outside the occipito-temporal cortexindicating that the absence of group differences is notmerely a lack of statistical power Taken together thesefindings suggest that normal face-related activation inventral occipito-temporal regions as measured withfMRI is not sufficient to ensure intact face processingand does not necessarily reflect normal face represen-tation within those regions

Can the BehaviorndashBOLD Dissociationbe Reconciled

The central result is that the BOLD activation in CP inventral occipito-temporal regions is largely not differen-tiable from that of the control subjects notwithstandingtheir marked behavioral impairment How might thisdissociation be explained

It has been previously suggested that the FFA ismainly involved in face detection that is discriminatingbetween faces and stimuli from other object categories(Tong et al 2000) an ability that is largely preserved inCP The apparently normal FFA activation in the CPsubjects might then reflect the intact involvement ofthe FFA in deriving a rather coarse and rudimentarystructural representation which suffices for some tasks(ie detection) but not for more taxing tasks such asidentification Evidence compatible with this possibleinterpretation suggests that the final stage of face iden-tification is not completed at the level of high-order

visual face-related regions such as the FFA rathersuccessful recognition requires activation in more ante-rior regions of the temporal lobe where a more abstractor semantic representation of faces is derived (HaxbyHoffman et al 2000 Leveroni et al 2000 SergentOhta amp Macdonald 1992) Further support for the roleof anterior temporal regions in mediating face identifi-cation and particularly for being a site of facial memory(long-term representation) comes from lesion studiesFor example Barton and Cherkasova (2003) foundthat patients with anterior temporal lesions were themost impaired in a task that required generating long-term representations of famous faces as in mentalimagery compared to patients with more posterioroccipito-temporal lesions Similarly patients with ante-rior temporal lobectomy (following intractable epilepsy)were found to be impaired in identifying familiar faces(Glosser Salvucci amp Chiaravalloti 2003) The source ofthe behavioral problem in CP might then arise in theabnormal propagation of activation from the intact FFAto more anterior temporal regions

However the idea that the FFA only mediates facedetection is not universally accepted and the cleandivision of labor between the FFA and more anteriorregions might not be totally correct Although notdenying the contribution of more anterior regions toface processing and particularly to face identificationrecent experimental evidence suggests that the FFA isinvolved not only in face detection but also in faceidentification Thus normal subjects exhibited differen-tial responses for face detection and identification in thefusiform gyrus with the two processes resulting in eitherdifferential fMRI amplitude (Grill-Spector et al 2004) ortemporal scales (Puce Allison et al 1999 and seeSugase Yamane Ueno amp Kawano 1999 for similarfindings from single-unit recordings in monkeys) Thesedifferential responses in the fusiform gyrus may poten-tially be the outcome of feedback from more anteriorregions (as above) andor from additional computationtaking place within the fusiform gyrus itself Which ofthese is perturbed in CP remains unclear and they arenot mutually exclusive either Propagation to or fromanterior regions may be affected in CP However recallthat CP subjects exhibit a clear deficit not only in faceidentification (naming of famous faces) but also indiscrimination of unfamiliar faces (Figure 1B) (see alsoBehrmann et al 2005) This deficit in the more per-ceptual (not just memorial) aspects of face processingimplicates posterior occipito-temporal regions ratherthan exclusively the more anterior areas (Barton PressKeenan amp OrsquoConnor 2002 Wada amp Yamamoto 2001)It remains a possibility therefore that the fusiformactivation itself may be abnormal in CP Because thecurrent study employed detection and discriminationstasks the abnormality may still be uncovered undermore challenging conditions such as face identificationor other fine-grained face tasks

Avidan et al 1161

We also cannot rule out the possibility that differ-ences in the activation patterns in the fusiform gyrusandor other parts of the face processing networkbetween CP and normal control subjects might exist inthe temporal domain Such differences could underliethe behavioral dysfunction of CP subjects but may notbe evident with the temporal resolution afforded byfMRI (in the order of seconds) Some evidence sup-porting this view comes from ERP studies showing thatthe relatively early N170 potential which shows faceselectivity in normal subjects failed to show such se-lectivity in CP subjects (Kress amp Daum 2003b BentinDeouell et al 1999)

Finally it might be that the small differences inBOLD activation in CP subjects such as the slight re-duction in face selectivity in the lateral occipital region(eg Figures 2A and 6B) might be lsquolsquoamplifiedrsquorsquo whentranslated into behavioral performance If so the lateraloccipital region would serve as a critical link in the faceprocessing network Further support for the role ofthe occipital region in face processing comes fromstudies showing that acquired prosopagnosic patientsoften have lesions that involve not only the fusiformgyrus but also the lateral occipital region (Barton et al2002 Wada amp Yamamoto 2001) Moreover brain dam-age to this region but not to the FFA in some individ-uals may be associated with severe prosopagnosia(Rossion Caldara et al 2003)

As is evident there are a host of potential expla-nations to account for the dissociation between the im-paired behavioral profile and the apparently normalBOLD activation pattern We are currently exploringthese issues using new imaging experiments and hopethat these will uncover in greater detail the mecha-nisms giving rise to CP Suffice it to say that no one ex-planation captures the BOLDndashbehavior dissociationWhat is crucial however is that a simple assignmentof face processing to the FFA is no longer defensibleand a deeper examination of the computational prop-erties of the FFA and associated regions is demanded bythese findings

Face-related Activation Outside theOccipito-temporal Cortex

In the current study we acquired whole-brain functionalscans and thus were able to explore differences be-tween the CP and control groups in the responsepatterns in all regions of the brain not only in theoccipito-temporal regions A particularly novel findingwas that robust bilateral face-related activation wasevident in prefrontal regions (precentral sulcus inferiorfrontal sulcus anterior lateral sulcus) in the CP groupIn contrast the control group only exhibited right-lateralized activation in the precentral sulcus (Figure 3)Studies of normal individuals performing workingmemory tasks have revealed face-related activation in

prefrontal regions (Druzgal amp DrsquoEsposito 2003 Salaet al 2003 Haxby Petit et al 2000) and some ERPstudies have detected some inferior prefrontal cortexsites that generate small but specific face-specific re-sponses (Allison Puce Spencer amp McCarthy 1999)

The poor performance exhibited by CP subjectscompared to their controls on the one-back memorytask specifically for face stimuli is consistent with theinterpretation of this prefrontal activity being related toworking memory However given that the conven-tional face and object mapping experiment was notspecifically designed to look at the different componentsthat usually comprise a memory study (encoding mem-ory delay response) further research that specificallyand parametrically manipulates the memory load offaces and other object stimuli is required in order todetermine the role of this activation in face processing inCP subjects

This study has exploited the unique possibility ofexamining the activation of the fusiform gyrus andother cortical regions in an unusual group of individ-uals with CP Delineating the neural mechanism thatgives rise to abnormal behavior and by logical infer-ence normal behavior is a critical goal of neurosci-ence As might be predicted the correspondencesbetween the brain and behavior are not transparentand in this study are dissociated with the behavioralimpairment being evident concomitant with normalFFA activation Taken together however this studyhas indicated patterns of similarities and patterns ofpotential difference in this population of CP individ-uals relative to their normal counterparts The challengethat lies before us is to use these patterns to explainthe face processing deficit in detail

METHODS

Subjects

Four healthy CP subjects (2 men) aged between 29 and60 years with no discernable cortical lesion or anyhistory of neurological disease participated in all ex-periments (for details see Table 1) Another CP sub-

Table 1 Biographical Information about CP Subjects

Congenital Prosopagnosia (CP) Subjects

Subjectsrsquo Initials Sex Age

TM M 27

KM F 60

MT M 41

BE F 29

KM and TM are a mother and a son

1162 Journal of Cognitive Neuroscience Volume 17 Number 7

ject (NI male aged 40) was tested behaviorally (seeBehrmann et al 2005) but because we could not ac-quire his imaging data he was excluded from thepresent article Seven of the 12 control subjects whoparticipated in the behavioral studies along with threenew participants completed the imaging experimentsThe control group included at least two age- and sex-matched controls for each CP individual All CP andcontrol subjects were right-handed and had normal orcorrected-to-normal vision All subjects consented to par-ticipate in the experiments and the protocol was ap-proved by the Institutional Review Boards of CarnegieMellon University and the University of Pittsburgh

Imaging Experiments

Visual Stimulation

Visual stimuli were generated using the E-prime IFISsoftware (Psychology Software Tools Pittsburgh PAUSA) and projected via LCD to a screen located in theback of the scanner bore behind the subjectrsquos head Sub-jects viewed the stimuli through a tilted mirror mountedabove their eyes on the head coil Before the scan sub-

jects were familiarized with the visual stimuli and tasksof each of the imaging experiments

MRI Setup

All subjects were scanned in a 3-T Siemens Allegrascanner equipped with a standard head coil Subjectsparticipated in one scanning session that lasted 15 to2 hr in which all the functional and anatomical scanswere acquired The order of the functional experimentswithin the scanning session was generally constant forall subjects BOLD contrast was acquired using gradient-echo echo-planar imaging sequence Specific parame-ters TR = 3000 msec TE = 35 msec flip angle = 908FOV = 210 210 mm2 matrix size 64 64 35 axialslices 3 mm thickness no gap High-resolution anatom-ical scans (T1-weighted 3-D MPRAGE) were also acquiredto allow accurate cortical segmentation reconstructionand volume-based statistical analysis Specific parame-ters TE = 349 flip angle = 88 FOV = 256 256 mm2matrix size = 256 256 slice thickness = 1 mm numberof slices = 160ndash192 orientation of slices was eitherhorizontal or sagittal For three of the CP subjects the

Table 2 Laterality in force-selective activation

1 Talairach Coordinates Fusiform Gyrus

Left Hemisphere Right Hemisphere

x y z x y z

CP subjects 40 plusmn 2 53 plusmn 3 19 plusmn 2 38 plusmn 5 47 plusmn 8 17 plusmn 3

Control subjects 37 plusmn 4 41 plusmn 5 18 plusmn 4 36 plusmn 2 40 plusmn 7 17 plusmn 3

2 Activation Foci

Localizer Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 4 ndash ndash

Control subjects 9 1 ndash 4 2 1

Rubin FacendashVase Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 1 3 ndash

Control subjects 9 ndash ndash 3 2 4

Part 1 Talairach coordinates of the face-related activation in the fusiform gyrus of CP and control subjects extracted from the conventionalface and object mapping experiment Part 2 Number of CP and control subjects that showed bilateral right-lateralized or left-lateralizedface-related activation in the fusiform gyrus and in the lateral occipital region in the conventional face and object mapping experiment andin the Rubin facendashvase experiment

Avidan et al 1163

3-D brain reconstruction was performed on imagespreviously acquired on a 15-T GE scanner

Experiments

Conventional Face and Object Mapping Experiment

Stimuli included line drawings of faces buildings com-mon objects and geometric patterns presented in ashort block design (for details see Levy et al 2001Figure 1A) Each epoch lasted 9 sec and contained ninestimuli from the same category there were eight differ-ent images and one immediate repetition of one of thestimuli Each stimulus was presented for 800 msecfollowed by a 200-msec interval of blank screen Therewere seven repetitions of each condition and their orderwas pseudorandomized across subjects Visual epochswere interleaved with 6-sec blank epochs The experi-ment lasted 450 sec and started and ended with extend-ed blank epochs of 27 and 9 sec respectively Subjectswere instructed to fixate on a red central fixation dot toperform a one-back memory task and to press a key ona response glove to indicate the repeated image Bothaccuracy and reaction time were recorded

The Motion Pictures Experiment

The experiment lasted 10 minutes and was composed of32 epochs (15 sec long) of movie clips each containingimages from one of four possible categories people invarious situations navigation of the camera through citybuildings or through open fields and miscellaneousimages of objects from different categories (machinesfalling water etc for details see Hasson Nir et al2004 Figure 4) The experiment started with a 51-secblank period followed by 9 sec of pattern stimuli andended with a 1-minute blank Subjects were simplyinstructed to watch the movie clips

The Adaptation Experiment

The experiment included line drawing images fromthree different categories faces buildings and cars(for details see Avidan et al 2002 Figure 5A) Stimuliwere presented in 12-sec epochs that contained either12 different images (lsquolsquodifferentrsquorsquo condition) from thesame category or 12 repetitions of the same identicalstimulus (lsquolsquoidenticalrsquorsquo condition) Within epochs eachstimulus was presented for 800 msec followed by200 msec of fixation only The experiment lasted480 sec and started and ended with a long blank pe-riod of 21 and 15 sec respectively The first visualblock always contained geometric patterns and was ex-cluded from the analysis Visual epochs were interleavedwith 6-sec blank epochs and the order of visual epochswas pseudorandomized across subjects Subjects wereinstructed to continuously fixate a central red dot and

to covertly categorize the images as follows man wom-an or a child for the face stimuli apartment townhouseor public facility for the buildings and private car truckor bus for the vehicles

The Rubin FacendashVase Experiment

The experiment included modified versions of the Rubinfacendashvase stimuli as well as line drawings of front facesand household objects (for details see Hasson Hendleret al 2001 Figure 6A) The Rubin facendashvase stimuli weremodified so that subjectsrsquo perception was biased towardsone perceptual state or the other This was done by firstuniformly coloring the vase area within each of thestimuli and placing stripes over the profile Stimuli werethen placed either over a striped background (biastowards perception of the vase due to closure of thevase stimulus) or over a uniform-color background (biastowards perception of the face due to closure of theface stimulus) Critically the Rubin face and vase sharethe same local contours but give rise to different globalperceptions In order to ensure that subjectsrsquo perceptionremained biased to one perceptual interpretation (ei-ther the profile or vase) during the experiment stimuliwere presented briefly (200 msec) and were immediate-ly followed by a masking grid that was presented for800 msec Stimuli were presented in 9-sec blocks thatwere interleaved with 6-sec blanks Each condition wasrepeated eight times and the order of presentation waspseudorandomized across subjects Subjects maintainedcentral fixation throughout the experiment and per-formed a one-back memory task however due to atechnical failure their responses were not recorded Theexperiment lasted 507 sec and started and ended withextended blank periods

Data Analysis

Data analysis was performed using the BrainVoyager48 software package (Brain Innovation MaastrichtNetherlands) and complementary in-house softwareDetailed description of the data analysis is provided inHasson Harel et al (2003) Briefly for each subject thecortical surface was reconstructed and unfolded intothe flattened format Functional data for each subjectfrom each experiment were analyzed separately Pre-processing of functional scans included 3-D motioncorrection and filtering of low frequencies up to fivecycles per experiment (slow drift) Statistical analysiswas based on the GLM For each experiment each ofthe conditions (except for blank) was defined as a sepa-rate predictor a boxcar shape was used to model eachpredictor and a 3-sec lag was assumed Percent signalchange for each subject in each experiment (except forthe motion pictures experiment) was calculated as thepercent activation from a blank baseline For the mo-

1164 Journal of Cognitive Neuroscience Volume 17 Number 7

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

between these alternatives and establishing whetherneural activity in the FFA is sufficient for normal rep-resentation of fine distinctions among individual faceswill require further experiments in which more subtlevariations will be made to face stimuli The use of anevent-related adaptation paradigm in such future ex-periments will also minimize possible attentional varia-tions that are inherent in the block-design adaptationparadigm employed here

Finally in the last experiment similar to control sub-jects once again CP subjects also exhibited a selectiveBOLD signal for Rubin-face compared to Rubin-vasestimuli in the fusiform gyrus Importantly these stimulicontain similar local elements but induce two differentvisual percepts either of a face or of a vase dependingon figurendashground segmentation and global integrationand shape processing Hence selectivity for the Rubin-face over the Rubin-vase stimuli implies that as in thecontrol group CP subjectsrsquo face-related activation ismodulated by global grouping processes and is notsimply induced by the representation of local face parts(Hasson Hendler et al 2001)

In sum the CP subjects exhibit normal face- andobject-related activation patterns in multiple regions ofthe ventral occipito-temporal cortex under various ex-perimental paradigms Of note however is that groupdifferences between CP subjects and their controls werefound in regions outside the occipito-temporal cortexindicating that the absence of group differences is notmerely a lack of statistical power Taken together thesefindings suggest that normal face-related activation inventral occipito-temporal regions as measured withfMRI is not sufficient to ensure intact face processingand does not necessarily reflect normal face represen-tation within those regions

Can the BehaviorndashBOLD Dissociationbe Reconciled

The central result is that the BOLD activation in CP inventral occipito-temporal regions is largely not differen-tiable from that of the control subjects notwithstandingtheir marked behavioral impairment How might thisdissociation be explained

It has been previously suggested that the FFA ismainly involved in face detection that is discriminatingbetween faces and stimuli from other object categories(Tong et al 2000) an ability that is largely preserved inCP The apparently normal FFA activation in the CPsubjects might then reflect the intact involvement ofthe FFA in deriving a rather coarse and rudimentarystructural representation which suffices for some tasks(ie detection) but not for more taxing tasks such asidentification Evidence compatible with this possibleinterpretation suggests that the final stage of face iden-tification is not completed at the level of high-order

visual face-related regions such as the FFA rathersuccessful recognition requires activation in more ante-rior regions of the temporal lobe where a more abstractor semantic representation of faces is derived (HaxbyHoffman et al 2000 Leveroni et al 2000 SergentOhta amp Macdonald 1992) Further support for the roleof anterior temporal regions in mediating face identifi-cation and particularly for being a site of facial memory(long-term representation) comes from lesion studiesFor example Barton and Cherkasova (2003) foundthat patients with anterior temporal lesions were themost impaired in a task that required generating long-term representations of famous faces as in mentalimagery compared to patients with more posterioroccipito-temporal lesions Similarly patients with ante-rior temporal lobectomy (following intractable epilepsy)were found to be impaired in identifying familiar faces(Glosser Salvucci amp Chiaravalloti 2003) The source ofthe behavioral problem in CP might then arise in theabnormal propagation of activation from the intact FFAto more anterior temporal regions

However the idea that the FFA only mediates facedetection is not universally accepted and the cleandivision of labor between the FFA and more anteriorregions might not be totally correct Although notdenying the contribution of more anterior regions toface processing and particularly to face identificationrecent experimental evidence suggests that the FFA isinvolved not only in face detection but also in faceidentification Thus normal subjects exhibited differen-tial responses for face detection and identification in thefusiform gyrus with the two processes resulting in eitherdifferential fMRI amplitude (Grill-Spector et al 2004) ortemporal scales (Puce Allison et al 1999 and seeSugase Yamane Ueno amp Kawano 1999 for similarfindings from single-unit recordings in monkeys) Thesedifferential responses in the fusiform gyrus may poten-tially be the outcome of feedback from more anteriorregions (as above) andor from additional computationtaking place within the fusiform gyrus itself Which ofthese is perturbed in CP remains unclear and they arenot mutually exclusive either Propagation to or fromanterior regions may be affected in CP However recallthat CP subjects exhibit a clear deficit not only in faceidentification (naming of famous faces) but also indiscrimination of unfamiliar faces (Figure 1B) (see alsoBehrmann et al 2005) This deficit in the more per-ceptual (not just memorial) aspects of face processingimplicates posterior occipito-temporal regions ratherthan exclusively the more anterior areas (Barton PressKeenan amp OrsquoConnor 2002 Wada amp Yamamoto 2001)It remains a possibility therefore that the fusiformactivation itself may be abnormal in CP Because thecurrent study employed detection and discriminationstasks the abnormality may still be uncovered undermore challenging conditions such as face identificationor other fine-grained face tasks

Avidan et al 1161

We also cannot rule out the possibility that differ-ences in the activation patterns in the fusiform gyrusandor other parts of the face processing networkbetween CP and normal control subjects might exist inthe temporal domain Such differences could underliethe behavioral dysfunction of CP subjects but may notbe evident with the temporal resolution afforded byfMRI (in the order of seconds) Some evidence sup-porting this view comes from ERP studies showing thatthe relatively early N170 potential which shows faceselectivity in normal subjects failed to show such se-lectivity in CP subjects (Kress amp Daum 2003b BentinDeouell et al 1999)

Finally it might be that the small differences inBOLD activation in CP subjects such as the slight re-duction in face selectivity in the lateral occipital region(eg Figures 2A and 6B) might be lsquolsquoamplifiedrsquorsquo whentranslated into behavioral performance If so the lateraloccipital region would serve as a critical link in the faceprocessing network Further support for the role ofthe occipital region in face processing comes fromstudies showing that acquired prosopagnosic patientsoften have lesions that involve not only the fusiformgyrus but also the lateral occipital region (Barton et al2002 Wada amp Yamamoto 2001) Moreover brain dam-age to this region but not to the FFA in some individ-uals may be associated with severe prosopagnosia(Rossion Caldara et al 2003)

As is evident there are a host of potential expla-nations to account for the dissociation between the im-paired behavioral profile and the apparently normalBOLD activation pattern We are currently exploringthese issues using new imaging experiments and hopethat these will uncover in greater detail the mecha-nisms giving rise to CP Suffice it to say that no one ex-planation captures the BOLDndashbehavior dissociationWhat is crucial however is that a simple assignmentof face processing to the FFA is no longer defensibleand a deeper examination of the computational prop-erties of the FFA and associated regions is demanded bythese findings

Face-related Activation Outside theOccipito-temporal Cortex

In the current study we acquired whole-brain functionalscans and thus were able to explore differences be-tween the CP and control groups in the responsepatterns in all regions of the brain not only in theoccipito-temporal regions A particularly novel findingwas that robust bilateral face-related activation wasevident in prefrontal regions (precentral sulcus inferiorfrontal sulcus anterior lateral sulcus) in the CP groupIn contrast the control group only exhibited right-lateralized activation in the precentral sulcus (Figure 3)Studies of normal individuals performing workingmemory tasks have revealed face-related activation in

prefrontal regions (Druzgal amp DrsquoEsposito 2003 Salaet al 2003 Haxby Petit et al 2000) and some ERPstudies have detected some inferior prefrontal cortexsites that generate small but specific face-specific re-sponses (Allison Puce Spencer amp McCarthy 1999)

The poor performance exhibited by CP subjectscompared to their controls on the one-back memorytask specifically for face stimuli is consistent with theinterpretation of this prefrontal activity being related toworking memory However given that the conven-tional face and object mapping experiment was notspecifically designed to look at the different componentsthat usually comprise a memory study (encoding mem-ory delay response) further research that specificallyand parametrically manipulates the memory load offaces and other object stimuli is required in order todetermine the role of this activation in face processing inCP subjects

This study has exploited the unique possibility ofexamining the activation of the fusiform gyrus andother cortical regions in an unusual group of individ-uals with CP Delineating the neural mechanism thatgives rise to abnormal behavior and by logical infer-ence normal behavior is a critical goal of neurosci-ence As might be predicted the correspondencesbetween the brain and behavior are not transparentand in this study are dissociated with the behavioralimpairment being evident concomitant with normalFFA activation Taken together however this studyhas indicated patterns of similarities and patterns ofpotential difference in this population of CP individ-uals relative to their normal counterparts The challengethat lies before us is to use these patterns to explainthe face processing deficit in detail

METHODS

Subjects

Four healthy CP subjects (2 men) aged between 29 and60 years with no discernable cortical lesion or anyhistory of neurological disease participated in all ex-periments (for details see Table 1) Another CP sub-

Table 1 Biographical Information about CP Subjects

Congenital Prosopagnosia (CP) Subjects

Subjectsrsquo Initials Sex Age

TM M 27

KM F 60

MT M 41

BE F 29

KM and TM are a mother and a son

1162 Journal of Cognitive Neuroscience Volume 17 Number 7

ject (NI male aged 40) was tested behaviorally (seeBehrmann et al 2005) but because we could not ac-quire his imaging data he was excluded from thepresent article Seven of the 12 control subjects whoparticipated in the behavioral studies along with threenew participants completed the imaging experimentsThe control group included at least two age- and sex-matched controls for each CP individual All CP andcontrol subjects were right-handed and had normal orcorrected-to-normal vision All subjects consented to par-ticipate in the experiments and the protocol was ap-proved by the Institutional Review Boards of CarnegieMellon University and the University of Pittsburgh

Imaging Experiments

Visual Stimulation

Visual stimuli were generated using the E-prime IFISsoftware (Psychology Software Tools Pittsburgh PAUSA) and projected via LCD to a screen located in theback of the scanner bore behind the subjectrsquos head Sub-jects viewed the stimuli through a tilted mirror mountedabove their eyes on the head coil Before the scan sub-

jects were familiarized with the visual stimuli and tasksof each of the imaging experiments

MRI Setup

All subjects were scanned in a 3-T Siemens Allegrascanner equipped with a standard head coil Subjectsparticipated in one scanning session that lasted 15 to2 hr in which all the functional and anatomical scanswere acquired The order of the functional experimentswithin the scanning session was generally constant forall subjects BOLD contrast was acquired using gradient-echo echo-planar imaging sequence Specific parame-ters TR = 3000 msec TE = 35 msec flip angle = 908FOV = 210 210 mm2 matrix size 64 64 35 axialslices 3 mm thickness no gap High-resolution anatom-ical scans (T1-weighted 3-D MPRAGE) were also acquiredto allow accurate cortical segmentation reconstructionand volume-based statistical analysis Specific parame-ters TE = 349 flip angle = 88 FOV = 256 256 mm2matrix size = 256 256 slice thickness = 1 mm numberof slices = 160ndash192 orientation of slices was eitherhorizontal or sagittal For three of the CP subjects the

Table 2 Laterality in force-selective activation

1 Talairach Coordinates Fusiform Gyrus

Left Hemisphere Right Hemisphere

x y z x y z

CP subjects 40 plusmn 2 53 plusmn 3 19 plusmn 2 38 plusmn 5 47 plusmn 8 17 plusmn 3

Control subjects 37 plusmn 4 41 plusmn 5 18 plusmn 4 36 plusmn 2 40 plusmn 7 17 plusmn 3

2 Activation Foci

Localizer Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 4 ndash ndash

Control subjects 9 1 ndash 4 2 1

Rubin FacendashVase Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 1 3 ndash

Control subjects 9 ndash ndash 3 2 4

Part 1 Talairach coordinates of the face-related activation in the fusiform gyrus of CP and control subjects extracted from the conventionalface and object mapping experiment Part 2 Number of CP and control subjects that showed bilateral right-lateralized or left-lateralizedface-related activation in the fusiform gyrus and in the lateral occipital region in the conventional face and object mapping experiment andin the Rubin facendashvase experiment

Avidan et al 1163

3-D brain reconstruction was performed on imagespreviously acquired on a 15-T GE scanner

Experiments

Conventional Face and Object Mapping Experiment

Stimuli included line drawings of faces buildings com-mon objects and geometric patterns presented in ashort block design (for details see Levy et al 2001Figure 1A) Each epoch lasted 9 sec and contained ninestimuli from the same category there were eight differ-ent images and one immediate repetition of one of thestimuli Each stimulus was presented for 800 msecfollowed by a 200-msec interval of blank screen Therewere seven repetitions of each condition and their orderwas pseudorandomized across subjects Visual epochswere interleaved with 6-sec blank epochs The experi-ment lasted 450 sec and started and ended with extend-ed blank epochs of 27 and 9 sec respectively Subjectswere instructed to fixate on a red central fixation dot toperform a one-back memory task and to press a key ona response glove to indicate the repeated image Bothaccuracy and reaction time were recorded

The Motion Pictures Experiment

The experiment lasted 10 minutes and was composed of32 epochs (15 sec long) of movie clips each containingimages from one of four possible categories people invarious situations navigation of the camera through citybuildings or through open fields and miscellaneousimages of objects from different categories (machinesfalling water etc for details see Hasson Nir et al2004 Figure 4) The experiment started with a 51-secblank period followed by 9 sec of pattern stimuli andended with a 1-minute blank Subjects were simplyinstructed to watch the movie clips

The Adaptation Experiment

The experiment included line drawing images fromthree different categories faces buildings and cars(for details see Avidan et al 2002 Figure 5A) Stimuliwere presented in 12-sec epochs that contained either12 different images (lsquolsquodifferentrsquorsquo condition) from thesame category or 12 repetitions of the same identicalstimulus (lsquolsquoidenticalrsquorsquo condition) Within epochs eachstimulus was presented for 800 msec followed by200 msec of fixation only The experiment lasted480 sec and started and ended with a long blank pe-riod of 21 and 15 sec respectively The first visualblock always contained geometric patterns and was ex-cluded from the analysis Visual epochs were interleavedwith 6-sec blank epochs and the order of visual epochswas pseudorandomized across subjects Subjects wereinstructed to continuously fixate a central red dot and

to covertly categorize the images as follows man wom-an or a child for the face stimuli apartment townhouseor public facility for the buildings and private car truckor bus for the vehicles

The Rubin FacendashVase Experiment

The experiment included modified versions of the Rubinfacendashvase stimuli as well as line drawings of front facesand household objects (for details see Hasson Hendleret al 2001 Figure 6A) The Rubin facendashvase stimuli weremodified so that subjectsrsquo perception was biased towardsone perceptual state or the other This was done by firstuniformly coloring the vase area within each of thestimuli and placing stripes over the profile Stimuli werethen placed either over a striped background (biastowards perception of the vase due to closure of thevase stimulus) or over a uniform-color background (biastowards perception of the face due to closure of theface stimulus) Critically the Rubin face and vase sharethe same local contours but give rise to different globalperceptions In order to ensure that subjectsrsquo perceptionremained biased to one perceptual interpretation (ei-ther the profile or vase) during the experiment stimuliwere presented briefly (200 msec) and were immediate-ly followed by a masking grid that was presented for800 msec Stimuli were presented in 9-sec blocks thatwere interleaved with 6-sec blanks Each condition wasrepeated eight times and the order of presentation waspseudorandomized across subjects Subjects maintainedcentral fixation throughout the experiment and per-formed a one-back memory task however due to atechnical failure their responses were not recorded Theexperiment lasted 507 sec and started and ended withextended blank periods

Data Analysis

Data analysis was performed using the BrainVoyager48 software package (Brain Innovation MaastrichtNetherlands) and complementary in-house softwareDetailed description of the data analysis is provided inHasson Harel et al (2003) Briefly for each subject thecortical surface was reconstructed and unfolded intothe flattened format Functional data for each subjectfrom each experiment were analyzed separately Pre-processing of functional scans included 3-D motioncorrection and filtering of low frequencies up to fivecycles per experiment (slow drift) Statistical analysiswas based on the GLM For each experiment each ofthe conditions (except for blank) was defined as a sepa-rate predictor a boxcar shape was used to model eachpredictor and a 3-sec lag was assumed Percent signalchange for each subject in each experiment (except forthe motion pictures experiment) was calculated as thepercent activation from a blank baseline For the mo-

1164 Journal of Cognitive Neuroscience Volume 17 Number 7

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

We also cannot rule out the possibility that differ-ences in the activation patterns in the fusiform gyrusandor other parts of the face processing networkbetween CP and normal control subjects might exist inthe temporal domain Such differences could underliethe behavioral dysfunction of CP subjects but may notbe evident with the temporal resolution afforded byfMRI (in the order of seconds) Some evidence sup-porting this view comes from ERP studies showing thatthe relatively early N170 potential which shows faceselectivity in normal subjects failed to show such se-lectivity in CP subjects (Kress amp Daum 2003b BentinDeouell et al 1999)

Finally it might be that the small differences inBOLD activation in CP subjects such as the slight re-duction in face selectivity in the lateral occipital region(eg Figures 2A and 6B) might be lsquolsquoamplifiedrsquorsquo whentranslated into behavioral performance If so the lateraloccipital region would serve as a critical link in the faceprocessing network Further support for the role ofthe occipital region in face processing comes fromstudies showing that acquired prosopagnosic patientsoften have lesions that involve not only the fusiformgyrus but also the lateral occipital region (Barton et al2002 Wada amp Yamamoto 2001) Moreover brain dam-age to this region but not to the FFA in some individ-uals may be associated with severe prosopagnosia(Rossion Caldara et al 2003)

As is evident there are a host of potential expla-nations to account for the dissociation between the im-paired behavioral profile and the apparently normalBOLD activation pattern We are currently exploringthese issues using new imaging experiments and hopethat these will uncover in greater detail the mecha-nisms giving rise to CP Suffice it to say that no one ex-planation captures the BOLDndashbehavior dissociationWhat is crucial however is that a simple assignmentof face processing to the FFA is no longer defensibleand a deeper examination of the computational prop-erties of the FFA and associated regions is demanded bythese findings

Face-related Activation Outside theOccipito-temporal Cortex

In the current study we acquired whole-brain functionalscans and thus were able to explore differences be-tween the CP and control groups in the responsepatterns in all regions of the brain not only in theoccipito-temporal regions A particularly novel findingwas that robust bilateral face-related activation wasevident in prefrontal regions (precentral sulcus inferiorfrontal sulcus anterior lateral sulcus) in the CP groupIn contrast the control group only exhibited right-lateralized activation in the precentral sulcus (Figure 3)Studies of normal individuals performing workingmemory tasks have revealed face-related activation in

prefrontal regions (Druzgal amp DrsquoEsposito 2003 Salaet al 2003 Haxby Petit et al 2000) and some ERPstudies have detected some inferior prefrontal cortexsites that generate small but specific face-specific re-sponses (Allison Puce Spencer amp McCarthy 1999)

The poor performance exhibited by CP subjectscompared to their controls on the one-back memorytask specifically for face stimuli is consistent with theinterpretation of this prefrontal activity being related toworking memory However given that the conven-tional face and object mapping experiment was notspecifically designed to look at the different componentsthat usually comprise a memory study (encoding mem-ory delay response) further research that specificallyand parametrically manipulates the memory load offaces and other object stimuli is required in order todetermine the role of this activation in face processing inCP subjects

This study has exploited the unique possibility ofexamining the activation of the fusiform gyrus andother cortical regions in an unusual group of individ-uals with CP Delineating the neural mechanism thatgives rise to abnormal behavior and by logical infer-ence normal behavior is a critical goal of neurosci-ence As might be predicted the correspondencesbetween the brain and behavior are not transparentand in this study are dissociated with the behavioralimpairment being evident concomitant with normalFFA activation Taken together however this studyhas indicated patterns of similarities and patterns ofpotential difference in this population of CP individ-uals relative to their normal counterparts The challengethat lies before us is to use these patterns to explainthe face processing deficit in detail

METHODS

Subjects

Four healthy CP subjects (2 men) aged between 29 and60 years with no discernable cortical lesion or anyhistory of neurological disease participated in all ex-periments (for details see Table 1) Another CP sub-

Table 1 Biographical Information about CP Subjects

Congenital Prosopagnosia (CP) Subjects

Subjectsrsquo Initials Sex Age

TM M 27

KM F 60

MT M 41

BE F 29

KM and TM are a mother and a son

1162 Journal of Cognitive Neuroscience Volume 17 Number 7

ject (NI male aged 40) was tested behaviorally (seeBehrmann et al 2005) but because we could not ac-quire his imaging data he was excluded from thepresent article Seven of the 12 control subjects whoparticipated in the behavioral studies along with threenew participants completed the imaging experimentsThe control group included at least two age- and sex-matched controls for each CP individual All CP andcontrol subjects were right-handed and had normal orcorrected-to-normal vision All subjects consented to par-ticipate in the experiments and the protocol was ap-proved by the Institutional Review Boards of CarnegieMellon University and the University of Pittsburgh

Imaging Experiments

Visual Stimulation

Visual stimuli were generated using the E-prime IFISsoftware (Psychology Software Tools Pittsburgh PAUSA) and projected via LCD to a screen located in theback of the scanner bore behind the subjectrsquos head Sub-jects viewed the stimuli through a tilted mirror mountedabove their eyes on the head coil Before the scan sub-

jects were familiarized with the visual stimuli and tasksof each of the imaging experiments

MRI Setup

All subjects were scanned in a 3-T Siemens Allegrascanner equipped with a standard head coil Subjectsparticipated in one scanning session that lasted 15 to2 hr in which all the functional and anatomical scanswere acquired The order of the functional experimentswithin the scanning session was generally constant forall subjects BOLD contrast was acquired using gradient-echo echo-planar imaging sequence Specific parame-ters TR = 3000 msec TE = 35 msec flip angle = 908FOV = 210 210 mm2 matrix size 64 64 35 axialslices 3 mm thickness no gap High-resolution anatom-ical scans (T1-weighted 3-D MPRAGE) were also acquiredto allow accurate cortical segmentation reconstructionand volume-based statistical analysis Specific parame-ters TE = 349 flip angle = 88 FOV = 256 256 mm2matrix size = 256 256 slice thickness = 1 mm numberof slices = 160ndash192 orientation of slices was eitherhorizontal or sagittal For three of the CP subjects the

Table 2 Laterality in force-selective activation

1 Talairach Coordinates Fusiform Gyrus

Left Hemisphere Right Hemisphere

x y z x y z

CP subjects 40 plusmn 2 53 plusmn 3 19 plusmn 2 38 plusmn 5 47 plusmn 8 17 plusmn 3

Control subjects 37 plusmn 4 41 plusmn 5 18 plusmn 4 36 plusmn 2 40 plusmn 7 17 plusmn 3

2 Activation Foci

Localizer Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 4 ndash ndash

Control subjects 9 1 ndash 4 2 1

Rubin FacendashVase Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 1 3 ndash

Control subjects 9 ndash ndash 3 2 4

Part 1 Talairach coordinates of the face-related activation in the fusiform gyrus of CP and control subjects extracted from the conventionalface and object mapping experiment Part 2 Number of CP and control subjects that showed bilateral right-lateralized or left-lateralizedface-related activation in the fusiform gyrus and in the lateral occipital region in the conventional face and object mapping experiment andin the Rubin facendashvase experiment

Avidan et al 1163

3-D brain reconstruction was performed on imagespreviously acquired on a 15-T GE scanner

Experiments

Conventional Face and Object Mapping Experiment

Stimuli included line drawings of faces buildings com-mon objects and geometric patterns presented in ashort block design (for details see Levy et al 2001Figure 1A) Each epoch lasted 9 sec and contained ninestimuli from the same category there were eight differ-ent images and one immediate repetition of one of thestimuli Each stimulus was presented for 800 msecfollowed by a 200-msec interval of blank screen Therewere seven repetitions of each condition and their orderwas pseudorandomized across subjects Visual epochswere interleaved with 6-sec blank epochs The experi-ment lasted 450 sec and started and ended with extend-ed blank epochs of 27 and 9 sec respectively Subjectswere instructed to fixate on a red central fixation dot toperform a one-back memory task and to press a key ona response glove to indicate the repeated image Bothaccuracy and reaction time were recorded

The Motion Pictures Experiment

The experiment lasted 10 minutes and was composed of32 epochs (15 sec long) of movie clips each containingimages from one of four possible categories people invarious situations navigation of the camera through citybuildings or through open fields and miscellaneousimages of objects from different categories (machinesfalling water etc for details see Hasson Nir et al2004 Figure 4) The experiment started with a 51-secblank period followed by 9 sec of pattern stimuli andended with a 1-minute blank Subjects were simplyinstructed to watch the movie clips

The Adaptation Experiment

The experiment included line drawing images fromthree different categories faces buildings and cars(for details see Avidan et al 2002 Figure 5A) Stimuliwere presented in 12-sec epochs that contained either12 different images (lsquolsquodifferentrsquorsquo condition) from thesame category or 12 repetitions of the same identicalstimulus (lsquolsquoidenticalrsquorsquo condition) Within epochs eachstimulus was presented for 800 msec followed by200 msec of fixation only The experiment lasted480 sec and started and ended with a long blank pe-riod of 21 and 15 sec respectively The first visualblock always contained geometric patterns and was ex-cluded from the analysis Visual epochs were interleavedwith 6-sec blank epochs and the order of visual epochswas pseudorandomized across subjects Subjects wereinstructed to continuously fixate a central red dot and

to covertly categorize the images as follows man wom-an or a child for the face stimuli apartment townhouseor public facility for the buildings and private car truckor bus for the vehicles

The Rubin FacendashVase Experiment

The experiment included modified versions of the Rubinfacendashvase stimuli as well as line drawings of front facesand household objects (for details see Hasson Hendleret al 2001 Figure 6A) The Rubin facendashvase stimuli weremodified so that subjectsrsquo perception was biased towardsone perceptual state or the other This was done by firstuniformly coloring the vase area within each of thestimuli and placing stripes over the profile Stimuli werethen placed either over a striped background (biastowards perception of the vase due to closure of thevase stimulus) or over a uniform-color background (biastowards perception of the face due to closure of theface stimulus) Critically the Rubin face and vase sharethe same local contours but give rise to different globalperceptions In order to ensure that subjectsrsquo perceptionremained biased to one perceptual interpretation (ei-ther the profile or vase) during the experiment stimuliwere presented briefly (200 msec) and were immediate-ly followed by a masking grid that was presented for800 msec Stimuli were presented in 9-sec blocks thatwere interleaved with 6-sec blanks Each condition wasrepeated eight times and the order of presentation waspseudorandomized across subjects Subjects maintainedcentral fixation throughout the experiment and per-formed a one-back memory task however due to atechnical failure their responses were not recorded Theexperiment lasted 507 sec and started and ended withextended blank periods

Data Analysis

Data analysis was performed using the BrainVoyager48 software package (Brain Innovation MaastrichtNetherlands) and complementary in-house softwareDetailed description of the data analysis is provided inHasson Harel et al (2003) Briefly for each subject thecortical surface was reconstructed and unfolded intothe flattened format Functional data for each subjectfrom each experiment were analyzed separately Pre-processing of functional scans included 3-D motioncorrection and filtering of low frequencies up to fivecycles per experiment (slow drift) Statistical analysiswas based on the GLM For each experiment each ofthe conditions (except for blank) was defined as a sepa-rate predictor a boxcar shape was used to model eachpredictor and a 3-sec lag was assumed Percent signalchange for each subject in each experiment (except forthe motion pictures experiment) was calculated as thepercent activation from a blank baseline For the mo-

1164 Journal of Cognitive Neuroscience Volume 17 Number 7

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

ject (NI male aged 40) was tested behaviorally (seeBehrmann et al 2005) but because we could not ac-quire his imaging data he was excluded from thepresent article Seven of the 12 control subjects whoparticipated in the behavioral studies along with threenew participants completed the imaging experimentsThe control group included at least two age- and sex-matched controls for each CP individual All CP andcontrol subjects were right-handed and had normal orcorrected-to-normal vision All subjects consented to par-ticipate in the experiments and the protocol was ap-proved by the Institutional Review Boards of CarnegieMellon University and the University of Pittsburgh

Imaging Experiments

Visual Stimulation

Visual stimuli were generated using the E-prime IFISsoftware (Psychology Software Tools Pittsburgh PAUSA) and projected via LCD to a screen located in theback of the scanner bore behind the subjectrsquos head Sub-jects viewed the stimuli through a tilted mirror mountedabove their eyes on the head coil Before the scan sub-

jects were familiarized with the visual stimuli and tasksof each of the imaging experiments

MRI Setup

All subjects were scanned in a 3-T Siemens Allegrascanner equipped with a standard head coil Subjectsparticipated in one scanning session that lasted 15 to2 hr in which all the functional and anatomical scanswere acquired The order of the functional experimentswithin the scanning session was generally constant forall subjects BOLD contrast was acquired using gradient-echo echo-planar imaging sequence Specific parame-ters TR = 3000 msec TE = 35 msec flip angle = 908FOV = 210 210 mm2 matrix size 64 64 35 axialslices 3 mm thickness no gap High-resolution anatom-ical scans (T1-weighted 3-D MPRAGE) were also acquiredto allow accurate cortical segmentation reconstructionand volume-based statistical analysis Specific parame-ters TE = 349 flip angle = 88 FOV = 256 256 mm2matrix size = 256 256 slice thickness = 1 mm numberof slices = 160ndash192 orientation of slices was eitherhorizontal or sagittal For three of the CP subjects the

Table 2 Laterality in force-selective activation

1 Talairach Coordinates Fusiform Gyrus

Left Hemisphere Right Hemisphere

x y z x y z

CP subjects 40 plusmn 2 53 plusmn 3 19 plusmn 2 38 plusmn 5 47 plusmn 8 17 plusmn 3

Control subjects 37 plusmn 4 41 plusmn 5 18 plusmn 4 36 plusmn 2 40 plusmn 7 17 plusmn 3

2 Activation Foci

Localizer Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 4 ndash ndash

Control subjects 9 1 ndash 4 2 1

Rubin FacendashVase Experiment

Fusiform Gyrus Lateral Occipital Region

Bilateral Right Lat Left Lat Bilateral Right Lat Left Lat

CP subjects 4 ndash ndash 1 3 ndash

Control subjects 9 ndash ndash 3 2 4

Part 1 Talairach coordinates of the face-related activation in the fusiform gyrus of CP and control subjects extracted from the conventionalface and object mapping experiment Part 2 Number of CP and control subjects that showed bilateral right-lateralized or left-lateralizedface-related activation in the fusiform gyrus and in the lateral occipital region in the conventional face and object mapping experiment andin the Rubin facendashvase experiment

Avidan et al 1163

3-D brain reconstruction was performed on imagespreviously acquired on a 15-T GE scanner

Experiments

Conventional Face and Object Mapping Experiment

Stimuli included line drawings of faces buildings com-mon objects and geometric patterns presented in ashort block design (for details see Levy et al 2001Figure 1A) Each epoch lasted 9 sec and contained ninestimuli from the same category there were eight differ-ent images and one immediate repetition of one of thestimuli Each stimulus was presented for 800 msecfollowed by a 200-msec interval of blank screen Therewere seven repetitions of each condition and their orderwas pseudorandomized across subjects Visual epochswere interleaved with 6-sec blank epochs The experi-ment lasted 450 sec and started and ended with extend-ed blank epochs of 27 and 9 sec respectively Subjectswere instructed to fixate on a red central fixation dot toperform a one-back memory task and to press a key ona response glove to indicate the repeated image Bothaccuracy and reaction time were recorded

The Motion Pictures Experiment

The experiment lasted 10 minutes and was composed of32 epochs (15 sec long) of movie clips each containingimages from one of four possible categories people invarious situations navigation of the camera through citybuildings or through open fields and miscellaneousimages of objects from different categories (machinesfalling water etc for details see Hasson Nir et al2004 Figure 4) The experiment started with a 51-secblank period followed by 9 sec of pattern stimuli andended with a 1-minute blank Subjects were simplyinstructed to watch the movie clips

The Adaptation Experiment

The experiment included line drawing images fromthree different categories faces buildings and cars(for details see Avidan et al 2002 Figure 5A) Stimuliwere presented in 12-sec epochs that contained either12 different images (lsquolsquodifferentrsquorsquo condition) from thesame category or 12 repetitions of the same identicalstimulus (lsquolsquoidenticalrsquorsquo condition) Within epochs eachstimulus was presented for 800 msec followed by200 msec of fixation only The experiment lasted480 sec and started and ended with a long blank pe-riod of 21 and 15 sec respectively The first visualblock always contained geometric patterns and was ex-cluded from the analysis Visual epochs were interleavedwith 6-sec blank epochs and the order of visual epochswas pseudorandomized across subjects Subjects wereinstructed to continuously fixate a central red dot and

to covertly categorize the images as follows man wom-an or a child for the face stimuli apartment townhouseor public facility for the buildings and private car truckor bus for the vehicles

The Rubin FacendashVase Experiment

The experiment included modified versions of the Rubinfacendashvase stimuli as well as line drawings of front facesand household objects (for details see Hasson Hendleret al 2001 Figure 6A) The Rubin facendashvase stimuli weremodified so that subjectsrsquo perception was biased towardsone perceptual state or the other This was done by firstuniformly coloring the vase area within each of thestimuli and placing stripes over the profile Stimuli werethen placed either over a striped background (biastowards perception of the vase due to closure of thevase stimulus) or over a uniform-color background (biastowards perception of the face due to closure of theface stimulus) Critically the Rubin face and vase sharethe same local contours but give rise to different globalperceptions In order to ensure that subjectsrsquo perceptionremained biased to one perceptual interpretation (ei-ther the profile or vase) during the experiment stimuliwere presented briefly (200 msec) and were immediate-ly followed by a masking grid that was presented for800 msec Stimuli were presented in 9-sec blocks thatwere interleaved with 6-sec blanks Each condition wasrepeated eight times and the order of presentation waspseudorandomized across subjects Subjects maintainedcentral fixation throughout the experiment and per-formed a one-back memory task however due to atechnical failure their responses were not recorded Theexperiment lasted 507 sec and started and ended withextended blank periods

Data Analysis

Data analysis was performed using the BrainVoyager48 software package (Brain Innovation MaastrichtNetherlands) and complementary in-house softwareDetailed description of the data analysis is provided inHasson Harel et al (2003) Briefly for each subject thecortical surface was reconstructed and unfolded intothe flattened format Functional data for each subjectfrom each experiment were analyzed separately Pre-processing of functional scans included 3-D motioncorrection and filtering of low frequencies up to fivecycles per experiment (slow drift) Statistical analysiswas based on the GLM For each experiment each ofthe conditions (except for blank) was defined as a sepa-rate predictor a boxcar shape was used to model eachpredictor and a 3-sec lag was assumed Percent signalchange for each subject in each experiment (except forthe motion pictures experiment) was calculated as thepercent activation from a blank baseline For the mo-

1164 Journal of Cognitive Neuroscience Volume 17 Number 7

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

3-D brain reconstruction was performed on imagespreviously acquired on a 15-T GE scanner

Experiments

Conventional Face and Object Mapping Experiment

Stimuli included line drawings of faces buildings com-mon objects and geometric patterns presented in ashort block design (for details see Levy et al 2001Figure 1A) Each epoch lasted 9 sec and contained ninestimuli from the same category there were eight differ-ent images and one immediate repetition of one of thestimuli Each stimulus was presented for 800 msecfollowed by a 200-msec interval of blank screen Therewere seven repetitions of each condition and their orderwas pseudorandomized across subjects Visual epochswere interleaved with 6-sec blank epochs The experi-ment lasted 450 sec and started and ended with extend-ed blank epochs of 27 and 9 sec respectively Subjectswere instructed to fixate on a red central fixation dot toperform a one-back memory task and to press a key ona response glove to indicate the repeated image Bothaccuracy and reaction time were recorded

The Motion Pictures Experiment

The experiment lasted 10 minutes and was composed of32 epochs (15 sec long) of movie clips each containingimages from one of four possible categories people invarious situations navigation of the camera through citybuildings or through open fields and miscellaneousimages of objects from different categories (machinesfalling water etc for details see Hasson Nir et al2004 Figure 4) The experiment started with a 51-secblank period followed by 9 sec of pattern stimuli andended with a 1-minute blank Subjects were simplyinstructed to watch the movie clips

The Adaptation Experiment

The experiment included line drawing images fromthree different categories faces buildings and cars(for details see Avidan et al 2002 Figure 5A) Stimuliwere presented in 12-sec epochs that contained either12 different images (lsquolsquodifferentrsquorsquo condition) from thesame category or 12 repetitions of the same identicalstimulus (lsquolsquoidenticalrsquorsquo condition) Within epochs eachstimulus was presented for 800 msec followed by200 msec of fixation only The experiment lasted480 sec and started and ended with a long blank pe-riod of 21 and 15 sec respectively The first visualblock always contained geometric patterns and was ex-cluded from the analysis Visual epochs were interleavedwith 6-sec blank epochs and the order of visual epochswas pseudorandomized across subjects Subjects wereinstructed to continuously fixate a central red dot and

to covertly categorize the images as follows man wom-an or a child for the face stimuli apartment townhouseor public facility for the buildings and private car truckor bus for the vehicles

The Rubin FacendashVase Experiment

The experiment included modified versions of the Rubinfacendashvase stimuli as well as line drawings of front facesand household objects (for details see Hasson Hendleret al 2001 Figure 6A) The Rubin facendashvase stimuli weremodified so that subjectsrsquo perception was biased towardsone perceptual state or the other This was done by firstuniformly coloring the vase area within each of thestimuli and placing stripes over the profile Stimuli werethen placed either over a striped background (biastowards perception of the vase due to closure of thevase stimulus) or over a uniform-color background (biastowards perception of the face due to closure of theface stimulus) Critically the Rubin face and vase sharethe same local contours but give rise to different globalperceptions In order to ensure that subjectsrsquo perceptionremained biased to one perceptual interpretation (ei-ther the profile or vase) during the experiment stimuliwere presented briefly (200 msec) and were immediate-ly followed by a masking grid that was presented for800 msec Stimuli were presented in 9-sec blocks thatwere interleaved with 6-sec blanks Each condition wasrepeated eight times and the order of presentation waspseudorandomized across subjects Subjects maintainedcentral fixation throughout the experiment and per-formed a one-back memory task however due to atechnical failure their responses were not recorded Theexperiment lasted 507 sec and started and ended withextended blank periods

Data Analysis

Data analysis was performed using the BrainVoyager48 software package (Brain Innovation MaastrichtNetherlands) and complementary in-house softwareDetailed description of the data analysis is provided inHasson Harel et al (2003) Briefly for each subject thecortical surface was reconstructed and unfolded intothe flattened format Functional data for each subjectfrom each experiment were analyzed separately Pre-processing of functional scans included 3-D motioncorrection and filtering of low frequencies up to fivecycles per experiment (slow drift) Statistical analysiswas based on the GLM For each experiment each ofthe conditions (except for blank) was defined as a sepa-rate predictor a boxcar shape was used to model eachpredictor and a 3-sec lag was assumed Percent signalchange for each subject in each experiment (except forthe motion pictures experiment) was calculated as thepercent activation from a blank baseline For the mo-

1164 Journal of Cognitive Neuroscience Volume 17 Number 7

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

tion picture experiment the raw activation level ineach time point for each subject was z-normalizedand then smoothed with a moving average of threetime points These values were then averaged acrossall participants of each group (Figure 4B) To obtainthe multisubject maps time series of images of brainvolumes for each subject were converted into Talairachspace and z-normalized The multisubject maps wereobtained using a random-effects procedure Signifi-cance levels of activation maps were calculated takinginto account the minimum cluster size and the prob-ability threshold of a false detection of any givencluster This calculation was accomplished by a MonteCarlo simulation (lsquolsquoAlphaSimrsquorsquo software by B DouglasWard which is part of the AFNI analysis package CoxR W 1996) Specifically the probability of a false-positive detection per image was determined fromthe frequency count of cluster sizes within the entirecortical surface (not including white matter and sub-nuclei) using the combination of individual voxel prob-ability thresholding and a minimum cluster size of sixcontiguous functional voxels

lsquolsquoInternal Localizerrsquorsquo Test

To obtain an unbiased statistical test within a scan weused one set of epochs to define anatomical ROIs in theconventional face and object mapping experimentwhereas another set was used to estimate the percentsignal change within each region (for details see LernerHendler amp Malach 2002)

Definition of ROIs for the Motion Pictures and theAdaptation Experiments

ROIs were independently defined using the convention-al face and object mapping experiment Face-relatedregions (fusiform gyrus lateral occipital region) wereselected using a face contrast (faces vs buildings andobjects) and the collateral was defined using a buildingcontrast (building vs faces and objects) We then ex-tracted the unbiased activation profile from each ROIfor each subject

Statistical Comparison between CP Subjectsand Control Group

For the localizer adaptation and Rubin facendashvase ex-periments the statistical analysis was done by perform-ing a repeated-measure ANOVA with group as a factor(CPcontrols) the different experimental conditions asthe repeated measures and the percent signal changeaveraged across the left and right hemisphere as thedependent variable For the motion picture experimentthe correlation analyses were performed between theCP group and the control group

Acknowledgments

We thank Dr Kwan-Jin Jung Scott Kurdilla and Brent Fryefrom the Brain Imaging Research Center at the McGowan Cen-ter in Pittsburgh for their great help in acquiring the imagingdata and Grace Lee Leonard and Erin Reeve for their helpin collecting and analyzing the behavioral data We would alsolike to thank Michal Harel for her help with the 3-D brainreconstruction and Ifat Levy for fruitful discussions andcomments

Reprint requests should be sent to Marlene Behrmann Depart-ment of Psychology Baker Hall 331H Carnegie Mellon Univer-sity 5000 Forbes Avenue Pittsburgh PA 15213-3890 USA orvia e-mail behrmanncnbccmuedu

REFERENCES

Aguirre G K Singh R amp DrsquoEsposito M (1999) Stimulusinversion and the responses of face and object-sensitivecortical areas NeuroReport 10 189ndash194

Allison T Puce A Spencer D D amp McCarthy G(1999) Electrophysiological studies of human faceperception I Potentials generated in occipitotemporalcortex by face and non-face stimuli Cerebral Cortex 9415ndash430

Avidan G Hasson U Hendler T Zohary U amp Malach R(2002) Analysis of the neuronal selectivity underlying lowfMRI signals Current Biology 12 964ndash972

Barton J J amp Cherkasova M (2003) Face imagery and itsrelation to perception and covert recognition inprosopagnosia Neurology 61 220ndash225

Barton J J Press D Z Keenan J P amp OrsquoConnor M(2002) Lesions of the fusiform face area impair perceptionof facial configuration in prosopagnosia Neurology 5871ndash78

Behrmann M amp Avidan G (2005) Congenital prosopagnosiaface blind from birth Trends in Cognitive Sciences 9180ndash187

Behrmann M Avidan G Marotta J J amp Kimchi R (2005)Detailed exploration of face-related processing incongenital prosopagnosia 1 Behavioral findings Journalof Cognitive Neuroscience 17 1130ndash1149

Bentin S Allison T Puce A Perez E amp McCarthy G(1996) Electrophysiological studies of face perceptionin humans Journal of Cognitive Neuroscience 8551ndash565

Bentin S Deouell L Y amp Soroker N (1999) Selectivevisual streaming in face recognition Evidence fromdevelopmental prosopagnosia NeuroReport 10823ndash827

Bruyer R Laterre C Seron X Feyereisen P Strypstein EPierrard E amp Rectem D (1983) A case of prosopagnosiawith some preserved covert remembrance of familiar facesBrain and Cognition 2 257ndash284

Clark V P Keil K Maisog J M Courtney S UngerleiderL G amp Haxby J V (1996) Functional magnetic resonanceimaging of human visual cortex during face matchingA comparison with positron emission tomographyNeuroimage 4 1ndash15

Cox R W (1996) AFNI Software for analysis andvisualization of functional magnetic resonanceneuroimages Computers and Biomedical Research29 162ndash173

Damasio A R Tranel D amp Damasio H (1990) Face agnosiaand the neural substrates of memory Annual Review ofNeuroscience 13 89ndash109

Avidan et al 1165

De Haan E H amp Campbell R (1991) A fifteen yearfollow-up of a case of developmental prosopagnosiaCortex 27 489ndash509

De Renzi E (1997) Prosopagnosia In T E Feinberg ampM Farah (Eds) Behavioral neurology andneuropsychology (pp 245ndash256) New York McGraw-Hill

DrsquoEsposito M Deouell L Y amp Gazzaley A (2003)Alterations in the BOLD fMRI signal with ageing and diseaseA challenge for neuroimaging Nature ReviewsNeuroscience 4 863ndash872

Druzgal T J amp DrsquoEsposito M (2001) Activity in fusiformface area modulated as a function of working memoryload Brain Research Cognitive Brain Research 10355ndash364

Druzgal T J amp DrsquoEsposito M (2003) Dissectingcontributions of prefrontal cortex and fusiform face area toface working memory Journal of Cognitive Neuroscience15 771ndash784

Duchaine B C (2000) Developmental prosopagnosiawith normal configural processing NeuroReport 1179ndash83

Duchaine B amp Nakayuma K (2005) Dissociation of faceand object recognition in developmental prosopagnosiaJournal of Cognitive Neuroscience 17 249ndash261

Epstein R amp Kanwisher N (1998) A corticalrepresentation of the local visual environment Nature392 598ndash601

Farah M (2004) Visual agnosia (2nd ed) CambridgeMIT Press

Gauthier I Tarr M J Moylan J Skudlarski PGore J C amp Anderson A W (2000) The fusiformlsquolsquoface arearsquorsquo is part of a network that processes facesat the individual level Journal of Cognitive Neuroscience12 495ndash504

Glosser G Salvucci A E amp Chiaravalloti N D (2003)Naming and recognizing famous faces in temporal lobeepilepsy Neurology 61 81ndash86

Grill-Spector K Knouf N amp Kanwisher N (2004) Thefusiform face area subserves face perception not genericwithin-category identification Nature Neuroscience 7555ndash562

Grill-Spector K amp Malach R (2001) fMR-adaptationA tool for studying the functional properties of humancortical neurons Acta Psychologica (Amsterdam) 107293ndash321

Gross C G Rocha-Miranda C E amp Bender D B (1972)Visual properties of neurons in inferotemporal cortex ofthe Macaque Journal of Neurophysiology 35 96ndash111

Grueter M Grueter T Bell V Horst J Laskowski WSperling K Halligan P W Ellis H D amp Kennerknecht I(in press) Hereditary prosopagnosia The first case seriesCortex

Hadjikhani N amp De Gelder B (2002) Neural basis ofprosopagnosia An fMRI study Human Brain Mapping 16176ndash182

Halgren E Dale A M Sereno M I Tootell R B HMarinkovic K amp Rosen B R (1999) Location of humanface-selective cortex with respect to retinotopic areasHuman Brain Mapping 7 29ndash37

Hasson U Avidan G Deouell L Y Bentin S amp Malach R(2003) Face-selective activation in a congenitalprosopagnosic subject Journal of Cognitive Neuroscience15 419ndash431

Hasson U Harel M Levy I amp Malach R (2003) Large-scalemirror-symmetry organization of human occipito-temporalobject areas Neuron 37 1027ndash1041

Hasson U Hendler T Ben Bashat D amp Malach R (2001)Vase or face A neural correlate of shape-selective grouping

processes in the human brain Journal of CognitiveNeuroscience 13 744ndash753

Hasson U Nir Y Levy I Fuhrmann G amp Malach R (2004)Inter-subject synchronization of cortical activity duringnatural vision Science 303 1634ndash1640

Haxby J V Gobbini M I Furey M L Ishai ASchouten J L amp Pietrini P (2001) Distributed andoverlapping representations of faces and objects in ventraltemporal cortex Science 293 2425ndash2430

Haxby J V Hoffman E A amp Gobbini M I (2000) Thedistributed human neural system for face perception Trendsin Cognitive Sciences 4 223ndash233

Haxby J V Petit L Ungerleider L G amp Courtney S M(2000) Distinguishing the functional roles of multipleregions in distributed neural systems for visual workingmemory Neuroimage 11 380ndash391

Haxby J V Ungerleider L G Clark V P Schouten J LHoffman E A amp Martin A (1999) The effect of faceinversion on activity in human neural systems for face andobject perception Neuron 22 189ndash199

Henson R Shallice T amp Dolan R (2000) Neuroimagingevidence for dissociable forms of repetition primingScience 287 1269ndash1272

Hoffman E A amp Haxby J V (2000) Distinct representationsof eye gaze and identity in the distributed human neuralsystem for face perception Nature Neuroscience 380ndash84

Kanwisher N (2000) Domain specificity in face perceptionNature Neuroscience 3 759ndash763

Kanwisher N McDermott J amp Chun M M (1997) Thefusiform face area A module in human extrastriate cortexspecialized for face perception Journal of Neuroscience 174302ndash4311

Kanwisher N Tong F amp Nakayama K (1998) The effect offace inversion on the human fusiform face area Cognition68 1ndash11

Kress T amp Daum I (2003a) Developmental prosopagnosia Areview Behavioural Neurology 14 109ndash121

Kress T amp Daum I (2003b) Event-related potentialsref lect impaired face recognition in patients withcongenital prosopagnosia Neuroscience Letters 352133ndash136

Lerner Y Hendler T amp Malach R (2002) Object-completioneffects in the human lateral occipital complex CerebralCortex 12 163ndash177

Leveroni C L Seidenberg M Mayer A R Mead L ABinder J R amp Rao S M (2000) Neural systemsunderlying the recognition of familiar and newlylearned faces Journal of Neuroscience 20 878ndash886

Levy I Hasson U Avidan G Hendler T amp Malach R(2001) Centerndashperiphery organization of human objectareas Nature Neuroscience 4 533ndash539

McCarthy G Puce A Belger A amp Allison T (1999)Electrophysiological studies of human face perceptionII Response properties of face-specific potentials generatedin occipitotemporal cortex Cerebral Cortex 9 431ndash444

McCarthy G Puce A Gore J C amp Allison T (1997) Facespecific processing in the human fusiform gyrus Journal ofCognitive Neuroscience 9 605ndash610

McConachie H R (1976) Developmental prosopagnosia Asingle case report Cortex 12 76ndash82

Mundel T Milton J G Dimitrov A Wilson H WPelizzari C Uftring S Torres I Erickson R KSpire J P amp Towle V L (2003) Transient inability todistinguish between faces Electrophysiologic studiesJournal of Clinical Neurophysiology 20 102ndash110

Nakamura K Kawashima R Sato N Nakamura ASugiura M Kato T Hatano K Ito K Fukuda H

1166 Journal of Cognitive Neuroscience Volume 17 Number 7

Schormann T amp Zilles K (2000) Functional delineationof the human occipito-temporal areas related to face andscene processing A PET study Brain 123 1903ndash1912

Puce A Allison T Bentin S Gore J C amp McCarthy G(1998) Temporal cortex activation in humans viewing eyeand mouth movements Journal of Neuroscience 182188ndash2199

Puce A Allison T amp McCarthy G (1999)Electrophysiological studies of human face perception IIIEffects of top-down processing on face-specific potentialsCerebral Cortex 9 445ndash458

Puce A amp Perrett D (2003) Electrophysiology and brainimaging of biological motion Philosophical Transactionsof the Royal Society of London Series B BiologicalSciences 358 435ndash445

Rossion B Caldara R Seghier M Schuller A M LazeyrasF amp Mayer E (2003) A network of occipito-temporalface-sensitive areas besides the right middle fusiform gyrusis necessary for normal face processing Brain 1262381ndash2395

Rossion B Schiltz C Robaye L Pirenne D ampCrommelinck M (2001) How does the brain discriminatefamiliar and unfamiliar faces A PET study of face categoricalperception Journal of Cognitive Neuroscience 131019ndash1034

Sala J B Rama P amp Courtney S M (2003) Functionaltopography of a distributed neural system for spatial andnonspatial information maintenance in working memoryNeuropsychologia 41 341ndash356

Sergent J Ohta S amp Macdonald B (1992) Functionalneuroanatomy of face and object processing A positronemission tomography study Brain 115 15ndash36

Sergent J amp Signoret J L (1992) Functional andanatomical decomposition of face processing Evidencefrom prosopagnosia and PET study of normal subjectsPhilosophical Transactions of the Royal Society of LondonSeries B Biological Sciences 335 55ndash61

Sugase Y Yamane S Ueno S amp Kawano K (1999)Global and fine information coded by single neuronsin the temporal visual cortex Nature 400 869ndash873

Talairach J amp Tournoux P (1988) Co-planar stereotaxicatlas of the human brain New York Thieme MedicalPublishers

Tong F Nakayama K Moscovitch M Weinrib O ampKanwisher N (2000) Response properties of the humanfusiform face area Cognitive Neuropsychology 17 257ndash279

Wada Y amp Yamamoto T (2001) Selective impairment offacial recognition due to a haematoma restricted to the rightfusiform and lateral occipital region Journal of NeurologyNeurosurgery and Psychiatry 71 254ndash257

Avidan et al 1167