Vrije Universiteit Brussel

Neural representations of others in the medial prefrontal cortex do not depend on ourknowledge about themHeleven, Elien; Van Overwalle, Frank

Published in:Social Neuroscience

DOI:10.1080/17470919.2018.1472139

Publication date:2019

Document Version:Proof

Link to publication

Citation for published version (APA):Heleven, E., & Van Overwalle, F. (2019). Neural representations of others in the medial prefrontal cortex do notdepend on our knowledge about them. Social Neuroscience, 14(3), 286-299.https://doi.org/10.1080/17470919.2018.1472139

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright ownersand it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portalTake down policyIf you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediatelyand investigate your claim.

Download date: 07. Jan. 2021

Full Terms & Conditions of access and use can be found athttps://www.tandfonline.com/action/journalInformation?journalCode=psns20

Social Neuroscience

ISSN: 1747-0919 (Print) 1747-0927 (Online) Journal homepage: https://www.tandfonline.com/loi/psns20

Neural representations of others in the medialprefrontal cortex do not depend on our knowledgeabout them

Elien Heleven & Frank Van Overwalle

To cite this article: Elien Heleven & Frank Van Overwalle (2019) Neural representations of othersin the medial prefrontal cortex do not depend on our knowledge about them, Social Neuroscience,14:3, 286-299, DOI: 10.1080/17470919.2018.1472139

To link to this article: https://doi.org/10.1080/17470919.2018.1472139

View supplementary material

Published online: 07 May 2018.

Submit your article to this journal

Article views: 63

View Crossmark data

Citing articles: 3 View citing articles

ARTICLE

Neural representations of others in the medial prefrontal cortex do not dependon our knowledge about themElien Heleven and Frank Van Overwalle

Pyschology, Vrije Universiteit Brussel, Brussel, Belgium

ABSTRACTEarlier neuroimaging studies on social inferences applying repetition suppression indicated thatpsychological entities of persons are represented in the ventral medial prefrontal cortex (mPFC).These representations were identified by suppression of activation in neural populations afterrepetition of the same persons, and are interpreted as abstract summary representations for therepeated social entity. However, an alternative explanation might be that suppression for personsdoes not reflect the representation of a person as such, but rather some degree of knowledgeabout this person. The current study contrasted these hypotheses by manipulating repetition notonly of a person, but also of the knowledge about that person. If a high level of knowledge playsa role in person representation, suppression effects in the mPFC would be larger for well-knownpersons as opposed to less-known persons (e.g., close friends versus acquaintances). Contrary tothis alternative hypothesis, but in line with the original interpretation, our results revealed onlysuppression for person repetition in the ventral mPFC irrespective of knowledge level, suggestingthat both well-known and lesser-known persons are represented in this area. Suppression areasfor well-known and lesser-known persons were, however, only partly overlapping and the area forless-known others extended towards the dorsal mPFC.

ARTICLE HISTORYReceived 20 December 2017Revised 4 April 2018Published online 5 May 2018

KEYWORDSFmri repetition suppression;person representation; closeothers; acquaintances

Introduction

Every day we engage in different social situations. Thesmoothness and efficiency with which we navigate inthese contexts largely depends on our ability to correctlyattribute relevant characteristics to different types ofpersons from our social network (e.g., well vs lessknown others) such as their goals, strategies, desires orpersonality traits (Amodio & Frith, 2006; Schurz, Radua,Aichhorn, Richlan, & Perner, 2014; Van Overwalle, 2009).This process is referred to as mentalizing or theory ofmind. Particularly the ventral medial prefrontal cortex(vmPFC), which is part of a system of brain regionsknown as the mentalizing network, is an importantbrain area involved in these attributions. This region isnot only recruited during the mere processing of socialinformation, but is also recognized as a central hub thatintegrates and abstracts these attributed characteristicsout of lower level (e.g., observed behavior or interocep-tive signals) and higher level information (e.g., subjectivevalue or self-relevance). This integration forms dynamicsummary representations of persons and their character-istics, that can be retrieved with relatively little effort,and can help us easily and quickly apprehend our envir-

onment in order to behave appropriately (Baetens, Ma,Steen, & Van Overwalle, 2014; Forbes & Grafman, 2010;Krueger, Barbey, & Grafman, 2009; Roy, Shohamy, &Wager, 2012). Representations are “memories that arelocalized in neural networks that encode informationand when activated, enable access to this stored infor-mation” (Wood & Grafman, 2003, p. 139), so that thisinformation does not need to be computed anew, but isimmediately available and can be rapidly accessed andspeed up processing. This may greatly facilitate fast andsmooth social interaction with other people.

To reveal these neural representations, researchersapplied functional magnetic resonance imaging (fMRI)repetition suppression. This technique is based on theassumption that repeated processing of the same stimu-lus or concept results in suppressed activation in theneural population processing and representing knowl-edge about this stimulus or concept. This view is con-sistent with predictive coding (Friston, 2005; Gotts,Chow, & Martin, 2012; Grill-Spector, Henson, & Martin,2006) and connectionist models of neural functioning(McClelland & Rumelhart, 1988) since suppression canbe seen as a decrease in prediction error to the same

CONTACT Elien Heleven Elien.Heleven@vub.be Department of Psychology, Vrije Universiteit Brussel, Pleinlaan 2, B - 1050 Brussel, BelgiumSupplemental data can be accessed here.

SOCIAL NEUROSCIENCE2019, VOL. 14, NO. 3, 286–299https://doi.org/10.1080/17470919.2018.1472139

© 2018 Vrije Universiteit Brussel

stimulus or concept in a neural representation. Contraryto fMRI activation, this technique thus enables us toidentify the population involved in the representationof a specific stimulus or summary construct (e.g., person)and so controls for variable aspects of processing thatelicits non-specific brain activation (de Graaf, Hsieh, &Sack, 2012) such as prerequisite processes (e.g., bodymovement) or consequential post processes (e.g., emo-tional expression). Moreover, the identification of theseneural populations can give us insight in the organiza-tion of the brain (i.e., which concepts have a commonneural basis or are represented close to each other).This suppression technique already demonstrated thepresence of representations for several social constructsin the vmPFC such as implied personality traits(Ma, Baetens, Vandekerckhove, Kestemont et al., 2014;Ma, Wang, Yang, Feng, & Van Overwalle, 2016; VanOverwalle, Ma, & Baetens, 2015), the self (Jenkins,Macrae, & Mitchell, 2008), close others (Heleven & VanOverwalle, 2016) and friends (Szpunar, St Jacques,Robbins, Wig, & Schacter, 2014) and less familiar otherpersons (Heleven, Boukhlal, & Van Overwalle, 2017).

Heleven and Van Overwalle (2016) applied therepetition suppression technique to reveal represen-tations of people in our close social network. Theyasked their participants to read two consecutivebehavioral sentences in which the repetition of aclose other was manipulated, and asked them toinfer the implied trait of the person. They foundsuppression for person repetitions in the vmPFC, sug-gesting specific representations for close others inthis area. This area overlapped partially with traitrepresentations which these authors identified bymanipulating the repetition of traits implied in thebehavior (see also Ma, Baetens, Vandekerckhove,Kestemont et al., 2014; Ma, Baetens,Vandekerckhove, Van der Cruyssen, & Van Overwalle,2014).

In the current study we raise the question: what dothese representations comprise of? Are they reallyabstract summaries of persons and their trait characteris-tics or is another aspect of these social stimuli responsiblefor the suppression effects? Potential alternative candi-dates for suppression effects are specific features insteadof integrated person characteristics. Prior work has shownthat such features may recruit activation in the vmPFC(although no repetition suppression has yet been demon-strated). Among others, candidates are closeness(Krienen, Tu, & Buckner, 2010), attributed personal value(D’Argembeau, 2013; Lin, Horner, & Burgess, 2016), andperson descriptiveness (Araujo, Kaplan, Damasio, &Damasio, 2014). According to Welborn and Lieberman(2015), all these different aspects can be interpreted as

an expression of how well we know a specific person,suggesting that knowledge is the key aspect that deter-mines vmPFC activation. In their meta-analyses, Murray,Schaer, and Debbané (2012) also found evidence forknowledge as an important factor for vmPFC involve-ment. They found a region in this area related to selfand close other processing, which could be differentiatedfrom a more dorsal region (dmPFC) related to the proces-sing of less-known others such as acquaintances or mediafigures. This ventral-dorsal differentiation in the mPFC isgenerally accepted in mentalizing research (see Bzdoket al., 2013; and meta-analysis by Van Overwalle, 2009)with the ventral part being responsible for self and closeother processing and the dorsal part for general mentaliz-ing processes that are applicable to anyone, also less orunknown others (Murray et al., 2012; Welborn &Lieberman, 2015). The role of knowledge in vmPFC activa-tion was also demonstrated by research that found largerinvolvement of this area when autobiographic informa-tion is processed about a domain for which we havemoreas opposed to less memories or experiences (Araujo et al.,2014; Lieberman, Jarcho, & Satpute, 2004). Chavez andHeatherton (2015) suggested that vmPFC involvement isdetermined by closeness and similarity, thus our knowl-edge level about the information. Taken together, thesestudies suggest a role for knowledge level in vmPFCactivation. The question is: does knowledge level alsoplay a role in the neural representation of persons?

A recent suppression study suggests that this isnot the case. Heleven et al., 2017) found suppressionfor repeated (compared to different) unknown personprocessing in the vmPFC. They concluded thatbesides previously identified person-specific represen-tations for close others in this area, there seem to bemore generic representations for unknown others.These generic representations are abstractions fromthe idiosyncrasies of many individual experiences thatretain an essential generalizability which facilitate theunderstanding of unfamiliar social encounters.However, as far as we are aware, there are no studiesdirectly comparing representations for persons withdifferent levels of knowledge. It is therefore stillunclear how these different types of representationsare organized.

In the current research we investigate whether knowl-edge plays a role in person representation in the vmPFC.We manipulated not only person repetition as in previousresearch (Heleven et al., 2017; Heleven & Van Overwalle,2016) but also the level of knowledge about these per-sons. We focused on persons from participants’ socialnetwork instead of unknown persons, since these personsare most informative for our daily functioning and theirneural representation is therefore probably very strong.

SOCIAL NEUROSCIENCE 287

Our participants selected from their social network well-known persons (e.g., friends or family members) as well asless-known persons (e.g., acquaintances). During scanning,they read two behavioral descriptions that implied thesame moral trait for the persons involved. Critically, twokey person aspects were orthogonally manipulated.Specifically, the persons in the two sentences were either(a) exactly the same versus different and (b) the knowl-edge level was either the same (i.e., well-known or less-known) versus different (a combination of a well-knownwith a less-known person). We always implied the sametrait in the two sentences in order to keep trait suppres-sion constant in each condition.

We pitted three hypotheses against each other. Inline with previous research, we expected as firsthypothesis to find suppression in the vmPFC whenthe same instead of no person is repeated.Alternatively, if the assumption is correct that highlevel of knowledge is critical in person representation,two hypotheses may follow: a strict and weak one. Asstrict (second) hypothesis, we expect larger suppres-sion effects for well-known than for less-known others,irrespective of whether the same or different person isinvolved. This hypothesis implies that simply proces-sing a lot of knowledge about various social targetsunderlies the representation of those targets. Thisseems reasonable because person knowledge (i.e.,traits and other characteristics) is shared among per-sons we know well, and therefore also shared (directlyor indirectly through associative links) among repre-sentations of these persons. As weaker (third) hypoth-esis we predict that person repetition leads tosuppression, and more so for well-known as opposedto less-known persons. It seems reasonable that exten-sive knowledge might constrain the ways we makejudgments about well-known targets in such a wayas to produce more reliable psychological and neuralrepresentations.

In addition, we measure participants’ social networksize in order to investigate the relation between neuralactivation during person and knowledge processing andthe ease by which somebody navigates in a social contextas expressed by the persons’ number of social contacts.Previous studies demonstrated a relation betweenreported social contacts and mPFC volume and activationduring person processing (Heleven & Van Overwalle,2016; Lewis, Rezaie, Brown, Roberts, & Dunbar, 2011;Powell, Lewis, Roberts, García-Fiñana, & Dunbar, 2012).These findings are in line with the social brain hypothesis(Dunbar, 1998) which states that a more variable andcomplex social environments should coincide withchanges in the neocortex.

Method

Participants

Participants were 28 right handed, native Dutch speak-ing individuals (10 men) with ages varying from 19 to30 years (M = 23.27). They all reported no abnormalneurological history and had normal or corrected-to-normal vision. Informed consent was obtained in amanner approved by the Free University Brussels (ofthe primary investigator, FVO) and the Medical EthicsCommittee at the Hospital of University of Ghent,where the study was conducted. Participants werepaid 20 euro in exchange for their participation.

Stimulus material

The stimulus material consisted of sentences describinga person performing a behavior. They were borrowedfrom prior studies (Heleven & Van Overwalle, 2016; Maet al., 2014, 2014; Van der Cruyssen, Heleven, Ma,Vandekerckhove, & Van Overwalle, 2015). The agentsin the behavioral descriptions were names (or nick-names) provided by the participants of five personsthey reported to know very well such as family mem-bers or friends (> 5 on a 7-point scale anchored at 1 = Islightly know this person, 4 = I know this person well,and 7 = I know this person extremely well) and fiveacquaintances they vaguely know such as a neighbor ora person from their school or work (< 3 on the samescale). We only selected persons about whom the par-ticipants reported to have a positive valence (> 4 on a7-point scale anchored at 1 = I don’t like this person atall, 4 = I like this person a little, and 7 = I extremely likethis person). Well-known and less-known persons werematched on valence, that is, for each well-known per-son a less-known with the same reported valence waschosen. Consequently, mean ratings for well-knownand less-known persons differed significantly on knowl-edge (M = 6.10 & 2.21 respectively, t(27) = 37.20,p < .05), but not on valence (M = 5.65 for both).

Procedure

Participants were instructed to read the behavioral sen-tences and to visualize the person performing thebehavior and to infer the implied trait (nice, friendly,trustworthy, honest, generous, and their opposites).Each trial started by the word ”Trait” in the middle ofthe screen during 2 seconds to remind participants ofthe task. After this, two sentences (prime and target)which implied the same trait, were presented. The per-sons in the sentences varied, depending on the

288 E. HELEVEN AND F. VAN OVERWALLE

condition. Note that throughout the experiment thetrait implications were always identical, so that ourmanipulation reveals repetition effects of persons only.

We created six conditions (see Table 1). In theRepeated person/High knowledge condition, theagent in two consecutive sentences was the same well-known person. In the Repeated person/Low knowledgecondition the agent was the same less-known person. Inthe Different person/High knowledge condition theagents were two different well-known persons. In theDifferent person/Low knowledge condition two differentless-known persons were the agents. In the Differentperson/Low-High condition the agent in the prime sen-tence was a less-known person and the one in the targetsentence a well-known person. This was reversed in theIn the Different person/High-Low knowledge condition.The latter two conditions of mixed knowledge wereadded to control for effects of differential activation inprime or target sentences due level of knowledge alone,rather than repetition of level of knowledge. We added asingleton condition in which only a target sentence waspresented, to avoid that participants would ignore theprime sentence.

There were 20 trials in each condition. We presentedone of six versions of the material to each participant,counterbalanced between conditions and participants.All trials were presented in a random order across con-ditions. Sentences were presented in the middle of thescreen during 5.5 seconds. Each sentence was precededby a jittered interstimulus interval, varying from 2.5 to4.5 seconds randomly drawn from a uniform distribu-tion, during which participants viewed passively a fixa-tion crosshair. In 60% of the trials, to keep participantsfocused on the task, they were asked to rate howapplicable the implied trait or its opposite was to theperson in the second sentence, using one of fourresponse buttons: 1 = never 2 = sometimes, 3 = often,and 4 = always. In the remaining 40% of the trails, noquestion was asked in order to keep the total length ofthe experiment under one hour.

After participants left the scanner, we measured theirsocial network size, using an adapted form of the SocialNetwork Questionnaire designed by Lewis et al. (2011).Participants were requested to list the initials of every-one with whom they had social contact (a) during thelast 7 days and (b) during the last month (i.e., approx.30 days) and to distinguish people from their supportclique (i.e., people that would give emotional or finan-cial support in difficult times) from others. Contact wasdefined as some form of interaction, including face-to-face, phone call, email or text-messaging, or a letter.People that were contacted for professional reasons(e.g., doctor, lawyer, hairdresser, priest, plumber,employer or supervisor etc.) were excluded, unless thisinteraction was mainly considered of a social nature atthe time. Participants could look up a list of names intheir phone/address book, and indicated the gender ofeach contact.

Imaging procedure

Images were collected with a Siemens Magnetom TrioTIM scanner system (Siemens Medical Systems,Erlangen, Germany) using a 32-channel radiofrequencyhead coil. Stimuli were projected onto a screen at theend of the magnet bore that participants viewed byway of a mirror mounted on the head coil. Stimuluspresentation was controlled by E-Prime 2.0 (www.pstnet.com/eprime; Psychology Software Tools) runningunder Windows XP. Participants were placed head firstand supine in the scanner bore and were instructed notto move their heads to avoid motion artifacts. Foamcushions were placed within the head coil to minimizehead movements. First, a high-resolution anatomicalimages were acquired using a T1-weighted 3DMPRAGE sequence [TR = 2530 ms, TE = 2.58 ms,TI = 1100 ms, acquisition matrix = 256 × 256 × 176,sagittal FOV = 220 mm, flip angle = 7, voxelsize = 0.9 × 0.86 × 0.86 mm3 (resized to1 × 1 × 1 mm)]. Second, a fieldmap was calculated to

Table 1. Schematic presentation of the design.Condition Sentences

Person Knowledge Prime Sentence Target Sentence

Repeated Repeated High [Friend] shows the instructor her fault [Friend] tells her the truthRepeated Low [Neighbor] shows the instructor her fault [Neighbor] tells her the truth

Different Repeated High [Mother] shows the instructor her fault [Friend] tells her the truthRepeated Low [Colleague] shows the instructor her fault [Neighbor] tells her the truth

Different Low – High [Neighbor] shows the instructor her fault [Friend] tells her the truthHigh – Low [Friend] tells her friend the truth [Neighbor] tells her the truth

Singleton – – [Friend/Neighbor] tells her the truth

The persons between strait brackets are people from participants’ social network, matched on valence. The same personality trait is implied in twoconsecutive sentences.

SOCIAL NEUROSCIENCE 289

correct for inhomogeneities in the magnetic field(Cusack & Papadakis, 2002). Next, whole brain func-tional images were collected in a single run using aT2*-weighted gradient echo sequence, sensitive toBOLD contrast (TR = 2000 ms, TE = 35 ms, imagematrix = 64 × 64, FOV = 224 mm, flip angle = 80º,slice thickness = 3.0 mm, distance factor = 17%, voxelsize = 3.5 × 3.5 × 4.0 mm [resized to 2 x 2 × 2 mm], 30axial slices).

Image processing

SPM12 (Wellcome Department of Cognitive Neurology,London, UK) was used to process and analyze the fMRIdata. To remove sources of noise and artifact, data werepreprocessed. Inhomogeneities in the magnetic fieldwere corrected using the fieldmap (Cusack & Papadakis,2002). Functional data were corrected for differences inacquisition time between slices for each whole-brainvolume, realigned to correct for head movement, andco-registered with each participant’s anatomical data.Then, the functional data were transformed into a stan-dard anatomical space (2 mm isotropic voxels) based onthe ICBM152 brain template (Montreal NeurologicalInstitute), which approximates Talairach and Tournouxatlas space. Normalized data were then spatiallysmoothed (6 mm full-width at half-maximum, FWHM)using a Gaussian Kernel. Finally, using the ArtifactDetection Tool software package (ART; http://web.mit.edu/swg/art/art.pdf; http://www.nitrc.org/projects/artifact_detect), the preprocessed data were examined forexcessive motion artifacts and for correlations betweenmotion and experimental design, and between globalmean signal and experimental design. Outliers were iden-tified in the temporal differences series by assessingbetween-scan differences (Z-threshold: 3.0 mm, scan toscan movement threshold: 0.5 mm; rotation threshold:0.02 radians). These outliers were omitted from the ana-lysis by including a single regressor for each outlier. Nocorrelations between motion and experimental design orglobal signal and experimental design were identified. Sixdirections of motion parameters from the realignmentstep as well as outlier time points (defined by ART) wereincluded as nuisance regressors. A default high-pass filterwas used of 128s and serial correlations were accountedfor by the default auto-regressive AR(1) model.

Statistical analysis

The general linear model of SPM12 (WellcomeDepartment of Cognitive Neurology, London, UK) wasused to conduct the analyses of the fMRI data. At thefirst (single participant) level, the event-related design

was modeled with two regressors for each conditiontime-locked at the presentation of the two sentences(and one regressor for the singleton condition), andconvolved with a canonical hemodynamic responsefunction with event duration set to 0 for all conditions.We did not model the response of the participants as aseparate regressor.

At the second (group) level, in all whole-brain ana-lyses clusters were defined at threshold p < .001, uncor-rected with a minimum cluster extent of 10 voxels, andwe restricted the analysis further to clusters with aFamily Wise Error (FWE) corrected threshold at clusterlevel with p < .05. First, in order to investigate differencesbetween well- and less-known person processing (with-out repetition), we conducted a whole-brain randomeffects analysis contrasting the prime sentence of condi-tions with well-known and less-known agents. Second,to investigate repetition suppression effects in each con-dition in particular, we contrasted prime and target sen-tences in all conditions. Third, to further investigate ourspecific hypotheses on the repetition of person versusknowledge level, in line with earlier repetition researchwe defined repetition suppression main effects with theprime > target contrast involving suppression givenrepetition of the same person (hypothesis 1) and givenrepetition of high knowledge (hypothesis 2). A first set ofcontrasts tested only the relevant prime > target con-trasts between repetitions of a person or high knowl-edge, and ignored non-repetitions. We termed these“main contrasts given person/high knowledge repeti-tion” (see Table 2 for the contrast weights). A secondset of contrasts provided a stronger test of these repeti-tion hypotheses, in which the remaining non-repetitionconditions were restricted to show no changes betweenprimes and targets, by clamping the primes and targetsto a weight of 1. We termed these “main contrasts givenperson/high knowledge repetition controlling for non-repetition” (see also Table 2 for the contrast weights).Note that this actually also controls for trait repetition(which occurred in all conditions as the implied traitswere always similar). Finally, to test hypothesis 3, wedefined an interaction to compare person suppressionfor high as opposed to low knowledge level (see Table 2for the contrast weights). In addition, the reverse, target> prime enhancement contrasts were defined for allcontrasts above to explore whether the brain areas iden-tified are not involved in the complementary process ofrepetition enhancement which refers to an increase ofactivation from prime to target (see Supplementary tablefor specific contrast weights).

Prior research (Heleven et al., 2017; Heleven &Van Overwalle, 2016; Ma, Baetens, Vandekerckhove,Kestemont et al., 2014) documented that a whole brain

290 E. HELEVEN AND F. VAN OVERWALLE

Table 2. Repetition Suppression (Prime > Target contrast) effects from the whole brain analysis.Anatomical Label X y z Voxels Max t

Simple Prime > Target ContrastsRepeated person/High knowledge

[1–1 0 0 0 0 0 0]vmPFC 6 50 −2 396 4.15 ***vmPFC 6 38 −2 4.14vmPFC −8 48 −4 3.94

Repeated person/Low knowledge[0 0 1–1 0 0 0 0]

vmPFC 4 54 4 165 3.81 *vmPFC 6 42 10 3.55dmPFC 4 46 22 3.52R Superior Frontal Gyrus 22 26 60 192 4.75 *R Superior Frontal Gyrus 32 22 58 4.28R Superior Frontal Gyrus 22 40 52 3.86

Different person/High knowledge[0 0 0 0 1–1 0 0]

R PFC 26 38 2 270 5.04 **R Inferior Frontal Gyrus 22 32 −2 4.37vmPFC 14 34 −6 4.08

Different person/Low knowledge[0 0 0 0 0 0 1–1]

vmPFC 4 52 0 687 4.84 ***vmPFC −10 44 −2 4.40vmPFC −10 32 0 3.81

Different person/Low-High knowledgeNo suprathreshold clusters

Different person/High-Low knowledgevmPFC 6 36 0 3144 5.56 ***vmPFC 4 56 6 5.35vmPFC −2 42 −6 5.12R Middle Temporal Gyrus 66 −18 −12 315 5.73 **R Middle Temporal Gyrus 66 −36 0 4.61R Inferior Temporal Gyrus 62 −20 −22 4.27Precuneus 4 −40 50 342 4.75 **Paracentral Lobule −6 −40 52 4.03Cingulate Gyrus −4 −34 44 3.92Cerebellum (Pyramis) −38 −72 −44 303 4.29 **Cerebellum (Tuber) −44 −66 −38 4.27Cerebellum (Declive) −16 −84 −30 4.08

Main Suppression Prime > Target Contrasts given Person repetition (Hypothesis 1)Main suppression contrast given person repetition:

[1–1 1–1 0 0 0 0]vmPFC 4 52 2 678 4.98 ***vmPFC 8 40 −2 4.70vmPFC −10 48 −2 3.38R Inferior Frontal Gyrus 26 36 −4 195 5.71 *R Superior Frontal Gyrus 34 22 58 231 5.07 *R Middle Frontal Gyrus 28 36 52 4.67R Superior Frontal Gyrus 22 28 60 4.08Precuneus 0 −38 48 214 4.94 *

Main suppression contrast given person repetition controlling for non-repetition:[1–3 1–3 1 1 1 1]

R Inferior Frontal Gyrus 24 34 −4 849 5.50 ***vmPFC 8 40 −2 5.08vmPFC 6 48 −2 4.64R Middle Frontal Gyrus 32 32 52 277 5.03 **R Superior Frontal Gyrus 34 22 58 4.54R Superior Frontal Gyrus 24 28 58 3.72L Precuneus 0 −38 48 445 5.53 ***L Cingulate gyrus −12 −38 34 3.95

(Continued )

SOCIAL NEUROSCIENCE 291

analysis may be sensitive to many activation differencesthat are not due to suppression alone (e.g., due todifferences in the activation of unrelated primes andtargets), and may so reveal “false” suppression effects.To identify only brain areas showing the predicted repe-tition suppression pattern and exclude “false” suppres-sion effects, we conducted signal change analyses likethis earlier work (see also Jenkins et al., 2008).Specifically, we defined regions of interest (ROI) as asphere of 3 mm around all coordinates of the mainand interaction contrasts revealed in testing our

hypotheses. We then extracted the percentage signalchange in these ROIs for each participant using theMarsBar toolbox (http://marsbar.sourceforge.net) andcalculated repetition indexes for each condition, definedas the percentage signal change of target minus primesentences for each condition. We looked for differencesbetween the indexes using a repeated measures ANOVAwith Person (repeated vs. different) and Knowledge(High vs. Low) as within-participant factors with a thresh-old of p < 0.05. Significant differences between condi-tions were identified using t-tests with a threshold of

Table 2. (Continued).

Anatomical Label X y z Voxels Max t

Main suppression contrast given High Knowledge repetition (Hypothesis 2)Main suppression contrast given high knowledge repetition:

[1–1 0 0 1–1 0 0]R Inferior Frontal Gyrus 24 36 −2 871 6.43 ***vmPFC 6 48 −2 4.45vmPFC −6 28 8 4.26Precuneus 0 −38 48 295 4.59 **Cingulate Gyrus −10 −38 40 4.00Cingulate Gyrus 8 −30 42 3.33

Main suppression contrast given high knowledge repetition controlling for non-repetition:[1–3 1 1 1–3 1 1]

R Inferior Frontal Gyrus 24 34 −4 261 5.60 *R Anterior Cingulate Gyrus 24 38 8 4.29– 30 34 12 4.19Cingulate Gyrus 6 −36 44 161 3.65 *Cingulate Gyrus −8 −36 42 3.55

Interaction contrast of Person suppression and High > Low Knowledge (Hypothesis 3)

[1–3 1 1 0 0 0 0]No suprathreshold clusters

Interaction of person suppression and high > low knowledge controlling for non-repetition:[1–7 1 1 1 1 1 1]

No suprathreshold clusters

Additional Suppression Contrasts given Low KnowledgeMain suppression contrast given low knowledge repetition:

[0 0 1–1 0 0 1–1]vmPFC 4 52 2 1482 5.63 ***vmPFC 6 44 −4 4.83dmPFC 4 60 34 4.61Superior Frontal Gyrus 22 26 60 410 4.72 ***Medial Frontal Gyrus 14 36 42 3.93Superior Frontal Gyrus 32 24 56 3.93Posterior Cingulate 4 −38 22 414 4.43 ***Paracentral Lobule 0 −36 46 4.27Cingulate Gyrus 2 −22 36 3.69Inferior Parietal Lobule 44 −60 56 341 5.28 **Inferior Parietal Lobule 54 −56 46 4.19Inferior Parietal Lobule 44 −72 42 3.98

Conjunction of main suppression contrast given high and low knowledge repetition:[1–1 0 0 1–1 0 0][0 0 1–1 0 0 1–1]

vmPFC 6 48 −2 255 4.45 **vmPFC 6 40 −4 4.24

Notes: Coordinates refer to the MNI (Montreal Neurological Institute) stereotaxic space. Whole-brain analysis thresholded at p < 0.001, uncorrected with avoxel extent of ≥ 10. Reported coordinates are from clusters with p < 0.05, FWE cluster corrected. The contrast weights between parentheses refer to thePrime and Target in the first four conditions respectively: (1) Repeated person/High knowledge, (2) Repeated person/Low knowledge, (3) Different person/High knowledge and (4) Different person/Low knowledge. The contrast weights of the other conditions were set to zero (not shown). vmPFC = ventralmedial prefrontal cortex, dmPFC = dorsal medial prefrontal cortex, PFC = prefrontal cortex, L = left, R = right.

* p < 0.05, **p < 0.01, ***p < 0.001 (FWE cluster corrected)

292 E. HELEVEN AND F. VAN OVERWALLE

p < 0.05, one-sided since we were testing for increasedsuppression only. Note that this analysis has been typi-cally used for fMRI repetition designs in order to omitfalse positives due to differences (e.g., between primes)other than the predicted suppression prime > targetpattern in the hypothesized conditions (Heleven et al.,2017; Ma, Baetens, Vandekerckhove, Van der Cruyssenet al., 2014; Ma et al., 2016).

To explore the relationship between brain activationand social network size, we calculated the correlation ofpercent signal change and reported number of contactsduring each month for each of the first four conditionsseparately.

Multivoxel pattern analysis (MVPA)

In order to further investigate potential differencesbetween high versus low knowledge (without repe-tition), we applied a MVPA which allows to investi-gate differences between neural representations byanalyzing the differential pattern of activation in acluster rather than the mean level of activationabove threshold in a cluster as in the classical gen-eral linear model (GLM). We used PRoNTo 2.0(Schrouff et al., 2013; www.mlnl.cs.ucl.ac.uk/pronto)and compared high versus low knowledge of allprimes using a Support Vector Machine (SVM;Burges, 1998). We used prime rather than targetsentences, because the first sentences are unconta-minated by possible suppression effects, and thusreflect pure encoding (i.e., an MVPA on target sen-tences would simply reflect suppression effect). Thisanalysis was applied at the group level as well as ata within-individual level of each participant. For thegroup level, we analyzed the beta coefficients of thewhole brain of each participant per condition com-puted by the 1st level GLM in SPM. The multi-voxelclassification estimations were cross-validated usinga leave-one-subject-out scheme. We performed asimilar analysis at the individual level in order tosee whether a person-specific approach would bet-ter differentiate high vs low knowledge processing.Analysis at this level was performed on participants’betas per trial (not per condition as in the groupanalysis) computed by another 1st level GLM using aLeast Squares All method (Abdulrahman & Henson,2016). Validation used a leave-one-block-out cross-validation scheme. The significance of the perfor-mance of all classifications is determined by a non-parametric permutation test.

Results

Whole brain analysis

Before starting the analysis of interest, we began with awhole-brain random effects analysis of the activation ofthe primes to explore whether differences in the pro-cessing of primes might affect our repetition results.The well-known > less-known prime contrast and itsreverse did not show significant differences betweenprimes. This implies that the analyses below are notcontaminated by differences in processing of theprimes.

To investigate repetition suppression effects, we firstcontrasted prime and target sentences in all conditionsseparately. These simple prime > target repetition con-trasts revealed a significant degree of repetition sup-pression in the vmPFC for all conditions except for theDifferent person/Low-High knowledge condition(Table 2). This general suppression effect is unsurprisingsince the same trait implication is repeated in allconditions.

Second and more critically, to compare our personversus knowledge hypotheses, we conducted severalprime > target contrast analyses across several con-ditions, termed “main effects”. To reveal person sup-pression, we conducted a main effect of personrepetition (across the conditions of high and lowknowledge; hypothesis 1). This revealed a significantmain suppression effect of the repeated person in thevmPFC, but also in a few additional areas involvingthe right frontal cortex and precuneus (Table 2).When additionally controlling for no change in thenon-repetition (different) person conditions by clamp-ing the weights in these latter conditions to “1”,similar results were found, including the vmPFC. Amain prime > target contrast for repeated highknowledge (across similar or different persons;hypothesis 2), revealed suppression in the vmPFCand other areas. Importantly, when additionally con-trolling for no change in the low knowledge condi-tions (weights clamped to “1”), the analysis did notshow any suppression in the vmPFC (Table 2). Thissuggests that high knowledge alone is not predictiveof repetition effects in the vmPFC, contrary to ourknowledge hypothesis. To test our weaker knowledgeprediction (i.e., whether person repetition was stron-ger given a high versus low knowledge; hypothesis3), we conducted an interaction of the suppression(prime > target) contrast comparing repeated personprocessing for high > low knowledge. This interactiondid not reveal any significant clusters (Table 2), even

SOCIAL NEUROSCIENCE 293

when further controlling for non-repetition. Together,these results suggest that persons are represented inthe vmPFC irrespective of the level of knowledge wehave about them. Other contrasts to further explorethese suppression effects comparing high versus lowknowledge in person repetition revealed similarresults.

To explore further the impact of knowledge on per-son representations, we analyzed suppression in each(High and Low) knowledge condition, irrespective ofpersons. These contrasts revealed suppression in thevmPFC for well-known persons (Repeated person/Highknowledge), and additionally suppression in the dmPFCfor less-known persons only (Repeated person/Lowknowledge; Figure 1). This finding is in line with theventral-dorsal distinction within the mPFC for self ver-sus other processing (Murray et al., 2012; Welborn &Lieberman, 2015). Note that suppression in the vmPFCfor less-known others corroborates the finding abovethat High knowledge is not specifically predictive ofperson repetition effects, and does not impact onknowledge representations.

For enhancement, the reverse, target > prime contrastor their interaction contrasts revealed repetition enhance-ment in many regions (see supplementary Table). Notethat this analysis for repeated high knowledge (acrosssimilar or different persons) did not reveal enhancementin the vmPFC, except when we additionally controlled fornon-repetitions (MNI: 2 36 4). However, this contrast didnot show any suppression in the vmPFC (Table 2).

Percent signal change analysis

In order to identify only the expected suppression patternin the significant contrasts of the whole brain analysisabove and to exclude false suppression effects due to

other reasons (e.g., between unrelated primes and tar-gets), we computed the percent signal change in ROIscentered on all the peaks of the repetition main effectsand interactions reported in Table 2. We then calculated asuppression index by subtracting the percent signalchanges in the prime sentence from those in the targetsentence.

Figure 2 shows the most pronounced person sup-pression pattern in the vmPFC (hypothesis 1). Arepeated measures ANOVA with Person (repeatedvs. different) and Knowledge (high vs. low) as factorsrevealed a main effect for person repetition in thevmPFC, F (1, 26) = 6.60, p < .05 partial η2 = 0.196,but none for knowledge, F (1, 26) = 0.01, ns. Simplet-tests confirmed that the suppression indices of theconditions involving person repetition did not differfrom each other, while they differed significantlyfrom the other conditions (p < 0.05, one-sided),except that one difference was only marginally sig-nificant (p = 0.08; Different person/High knowledgeversus the Different person/Low knowledge).Additional analyses on the High-Low and Low-Highconditions further confirmed that the suppressionindices of the conditions involving person repetitiondiffered (marginally) significantly from these condi-tions (p < 0.08 and p < 0.05), except for theRepeated person/Low knowledge condition and theDifferent person/High-Low (p = 0.13). Together, thepercent signal change analysis confirms our wholebrain analysis, and also identifies person suppressionin the vmPFC irrespective of the level of knowledge.

We also searched for a pattern in line with ourmain knowledge repetition prediction (hypothesis 2)and in line with our weaker knowledge repetitionprediction (hypothesis 3). Such patterns were notfound in the ROIs based on the peaks in Table 2.

Figure 1. Identified clusters from the whole-brain analyses of the repetition suppression main effects given High and Lowknowledge (with respective main contrasts weights [1–1 0 0 1–1 0 0] and [0 0 1–1 0 0 1–1] as defined in Table 2), and theirconjunction (thresholded at p < .001, uncorrected). The clusters reflecting person representation are located more dorsal for less-known persons and more ventral for well-known persons.

294 E. HELEVEN AND F. VAN OVERWALLE

Multivoxel pattern analysis

The results of the multi-voxel pattern analysis (Schrouffet al., 2013) revealed that the accuracy of classificationof High versus Low Knowledge in the prime sentences(across all respective conditions) at the group level wasbelow or at random level (i.e., 33 and 51% respectively)and at the individual level with a classification abovechance for only one participant (i.e., 3.57% of all parti-cipants). Classification accuracy below chance level of50% indicates that the obtained classification is gener-ally opposite to the predicted classification (e.g., whenmore items end up in another category than the pre-dicted category). This demonstrates that the presentmulti-voxel analysis was not able to reliably classifypersons with high versus low knowledge based oninformation in the prime sentences (i.e., before repeti-tion), not even when using a within-individual analysis.Note that applying the classifier after repetition wouldbe superfluous, as repetition effects were demonstratedalready using the typical univariate SPM analysis.

Correlation with social network size

The mean number of all reported social contacts was27.82 (SD = 10.39) for the last week and 51.39(SD = 18.30) for the last month. Reported contacts

with members of the support clique specifically were7.75 (SD = 2.78) for the last week and 12.36 (SD = 5.12)for the last month. We investigated the relationshipbetween the number of reported social contacts duringthe last month and the percent signal change of targetsentences for the ROI showing the strongest suppres-sion (Figure 2; MNI −10 48–2). There was a significantpositive correlation between the number of social con-tacts with support clique members during the lastmonth and activation during target processing for theDifferent person/High knowledge condition (r = 0.39;p < 0.05; Figure 3). There was no significant correlationwith percent signal change for any other condition. Toexplore potential alternative explanations, we com-puted the same correlations for the other ROIs showingno person suppression, but none was significant.

Discussion

The current study investigated whether the knowledgewe have about persons determines their neural repre-sentation, as identified by the fMRI repetition suppres-sion method. We pitted against each other threeplausible hypotheses: (1) a suppression effect due toperson repetition (reflecting only a representation ofpersons), (2) a suppression effect due to high knowl-edge in general (reflecting only a representation of the

Figure 2. Percentage signal change in the ventral medial prefrontal cortex (vmPFC) on MNI coordinate −10 48–2. The left side ofthe graph shows prime-target pairs for each condition, the right side the suppression indexes. The inset above the indices showsthe cluster and peak (blue crosshairs) revealed in the whole-brain analysis thresholded at p < .001, uncorrected, from which thesignal change was extracted. Repetition suppression in the first two (Repeated person) conditions differs from next two (Differentperson) conditions, * p < .05 (F (1, 26) = 6.60, p < .05 partial η2 = 0.196). This pattern is indicative of a neural representation ofpersons (cf. repeated person suppression) irrespective of knowledge level.

SOCIAL NEUROSCIENCE 295

high knowledge about persons) or (3) an interaction(the “weaker” knowledge hypothesis reflecting strongerrepresentations for well-known persons than less-known persons). In line with the person representationhypothesis, we found suppression in the vmPFC whenthe same person from participants’ social networkswas processed twice, replicating the finding ofHeleven and Van Overwalle (2016). More importantly,this suppression effect was the same for well-knownand less-known persons. In contrast to the knowledgerepresentation hypothesis, the mere repetition of well-known information without specific person repetitiondid not elicit suppression effects. Moreover, the“weaker” knowledge hypothesis predicting that highlyknowledgeable persons are better represented thanless knowledgeable persons, was not supported either.These results demonstrate that person representationsin the vmPFC are not determined by the level of knowl-edge we have about these people, and that theobserved suppression effects in the current and pre-vious studies are not just a result of the repeatedintegration of the same level of knowledge about per-sons, but about the persons themselves. In line with theidea of the vmPFC as a central hub region (Baetenset al., 2014; Forbes & Grafman, 2010; Krueger et al.,2009; Roy et al., 2012), we interpret these suppressioneffects as reflecting abstract summary representationsof persons irrespective of our knowledge level aboutthem.

Regarding the location of the neural representations,although suppression for well-known and less-known

persons overlapped in the vmPFC, we found that repe-tition for well-known persons was confined within thevmPFC, while the area for less-known others extendedtowards the dmPFC. This is in line with the generalaccepted ventral-dorsal differentiation in the medialprefrontal cortex, with the ventral part considered asbeing responsible for person-specific processing of theself and close others, and the dorsal part for less knownor unknown others (Murray et al., 2012; Welborn &Lieberman, 2015). Note that these suppression clustersoverlapped and hence merely show gradients betweenfamiliar and unfamiliar others rather than clear-cut dis-tinctions. Moreover, these clusters were unconfoundedfor non-repetitions, and should thus be considered withcaution. Moreover, we were unable to further differenti-ate the neural population involved in high versus lowknowledge processing with an additional MVPA. Thisindicates that not only the mean activation at clusterlevel in a univariate analysis, but also the analysis of thedistributed pattern of activation at the level of singlevoxels, was unable to reliably differentiate low fromhigh knowledge.

Perhaps the present null results for knowledgerepresentation are due to the fact that well-knownand less-known people are still quite familiar andpart of participants’ social network (i.e., family, friends,acquaintances and colleagues). However, it is unlikelythat more reliable knowledge representations will beuncovered when comparing well-known others withtotally unknown others, since a recent study byHeleven et al. (2017) revealed person representations

Figure 3. Positive correlation between number of support clique members seen last month and activation in the vmPFC for theDifferent person/High knowledge condition. This pattern indicates that processing information about a new well-known other astarget recruits more vmPFC activation when the number of contacts with support clique members increases. Note that thiscorrelation is absent for less-known others, as well as for repeated others. * p < .05 (two-sided).

296 E. HELEVEN AND F. VAN OVERWALLE

for unknown others in the vmPFC. Perhaps futuresuppression studies should investigate whether wecan differentiate persons based on other characteris-tics than knowledge, in order to understand the neuralorganization of person representations.

The absence of a knowledge effect in the mereactivation of the primes is in line with earlier studiesthat did not reveal differential activation in the mPFCwhen processing close and distant others (Araujo,Kaplan, & Damasio, 2013), or close and public others(Murray et al., 2012). It however contradicts other find-ings such as those of Welborn and Lieberman (2015)who revealed elevated activation for subjectively well-known as opposed to less-known persons. Future stu-dies should investigate which factors determine theseconflicting results. Note that the lack of differences inprime activation by itself does not invalidate our failureto find high knowledge repetition suppression. On thecontrary, differences in target activation (in comparisonwith primes) are of most importance to reveal repeti-tion suppression, and differences in prime activationmay only tend to mask such target differences.

A potential limitation of the current study is thatwe presented people to whom the participantsattributed a positive valence. Therefore, it is possiblethat the observed suppression effect for personrepetition resulted from valence repetition. Sincewell and less known persons were matched onvalence, the equivalent suppression effect forrepeated persons on both levels of knowledge canalso be explained in terms of valence. We cannotentirely dismiss this interpretation in the presentstudy, but note that an earlier study by Ma,Baetens, Vandekerckhove, Van der Cruyssen et al.(2014) contrasted valence and trait repetition, andfound no ground for a valence explanation. Futureresearch can investigate more directly the possiblerole in vmPFC representations for valence or otherperson aspects for which we did not control in thecurrent study such as personal relevance ordescriptiveness.

Another potential alternative explanation for thepresent results might be that repetition effects resultedfrom the anticipation of a trait question that mightpotentially follow right after the target sentence.However this is unlikely, since no behavioral differencesin reaction time between conditions were observed.

Another possible limitation is that we cannot controlhow well participants imagined the persons performingthe described behavior. Future studies can take this intoaccount by letting participant rate every description onhow easy they were to visualize. We would not recom-mend to show specific images of the persons since this

could lead to activation in other brain areas such as thevisual (e.g., fusiform face area) or mirror system (VanOverwalle & Baetens, 2009).

Correlation with social network size

We found a positive correlation between the number ofreported social contacts with support clique membersand vmPFC activation when processing informationabout a second, but different well-known person. Thissuggests that a larger support clique corresponds withmore in-depth processing of novel well-known personinformation. Conversely, individuals with a smaller sup-port clique probably have a smaller number of well-known other representations, so that they retrieve theright person more easily, resulting in lesser activation.This positive correlation stands in contrast with theresult of Heleven and Van Overwalle (2016) whofound that a higher number of reported social contactswas related to less (rather than more) vmPFC activationwhen processing information about a second but dif-ferent close other. One possible explanation for thisdiscrepancy is that Heleven and Van Overwalle (2016)asked about the number of social contacts in general,and did not focus on support clique members alone aswas the case in this research. Moreover, Heleven andVan Overwalle (2016) reported that the identified cor-relation was heavily dependent on its exact location.Future research should try to clarify these conflictingresults.

Conclusion

In this study we examined the role of knowledge inperson representations in the vmPFC. We did not finddifferential suppression effects for well-known asopposed to less-known persons. Suppression regionsfor both levels of knowledge were partly overlappingwithin the vmPFC, with the representation of well-known others being confined to the vmPFC, and therepresentation of less-known others extending towardsthe dmPFC. However, an additional MVPA did not con-firm this ventral-dorsal differentiation. We conclude thatneural representations for persons from our social net-work (ranging from family to acquaintances) do notdepend much on the level of knowledge we haveabout them.

Disclosure statement

No potential conflict of interest was reported by the authors.

SOCIAL NEUROSCIENCE 297

ORCID

Frank Van Overwalle http://orcid.org/0000-0002-2538-9847

References

Abdulrahman, H., & Henson, R. N. (2016). Effect of trial-to-trialvariability on optimal event-related fMRI design:Implications for Beta-series correlation and multi-voxel pat-tern analysis. NeuroImage, 125, 756–766.

Amodio, D. M., & Frith, C. D. (2006). Meeting of minds: Themedial frontal cortex and social cognition. Nature Reviews.Neuroscience, 7(4), 268–277.

Araujo, H. F., Kaplan, J., & Damasio, A. (2013). Cortical midlinestructures and autobiographical-self processes: An activa-tion-likelihood estimation meta-analysis. Frontiers in HumanNeuroscience, 7(September), 1–10.

Araujo, H. F., Kaplan, J., Damasio, H., & Damasio, A. (2014).Involvement of cortical midline structures in the processingof autobiographical information. PeerJ, 2, e481.

Baetens, K., Ma, N., Steen, J., & Van Overwalle, F. (2014).Involvement of the mentalizing network in social andnon-social high construal. Social Cognitive and AffectiveNeuroscience, 9(6), 817–824.

Burges, C. J. C. (1998). A tutorial on support vector machines forpattern recognition. Boston: Kluwer Academic Publishers.

Bzdok, D., Langner, R., Schilbach, L., Engemann, D. A., Laird, A.R., Fox, P. T., & Eickhoff, S. B. (2013). Segregation of thehuman medial prefrontal cortex in social cognition.Frontiers in Human Neuroscience, 7(May), 1–17.

Chavez, R. S., & Heatherton, T. F. (2015). Representational simi-larity of social and valence information in the medial pre-frontal cortex. Journal of Cognitive Neuroscience, 27(1), 73–82.

Cusack, R., & Papadakis, N. (2002). New robust 3-D phaseunwrapping algorithms: Application to magnetic field map-ping and undistorting echoplanar images. NeuroImage, 16(3),754–764.

D’Argembeau, A. (2013). On the role of the ventromedialprefrontal cortex in self-processing: The valuation hypoth-esis. Frontiers in Human Neuroscience, 7(July), 372.

de Graaf, T. A., Hsieh, P.-J., & Sack, A. T. (2012). The “correlates”in neural correlates of consciousness. Neuroscience andBiobehavioral Reviews, 36(1), 191–197.

Dunbar, R. I. M. (1998). The social brain hypothesis. EvolutionaryAnthropology: Issues, News, and Reviews, 6(5), 178–190.

Forbes, C. E., & Grafman, J. (2010). The role of the humanprefrontal cortex in social cognition and moral judgment.Annual Review of Neuroscience, 33, 299–324.

Friston, K. (2005). A theory of cortical responses. PhilosophicalTransactions of the Royal Society of London. Series B,Biological Sciences, 360(1456), 815–836.

Gotts, S. J., Chow, C. C., & Martin, A. (2012). Repetition primingand repetition suppression: Multiple mechanisms in needof testing. Cognitive Neuroscience, 3(3–4), 250–259.

Grill-Spector, K., Henson, R., & Martin, A. (2006). Repetition andthe brain: Neural models of stimulus-specific effects. Trendsin Cognitive Sciences, 10(1), 14–23.

Heleven, E., Boukhlal, S., & Van Overwalle, F. (2017). A stran-ger in my brain: Neural representation for unfamiliar per-sons using fMRI repetition suppression. SocialNeuroscience, 1–11. doi:10.1080/17470919.2017.1358663

Heleven, E., & Van Overwalle, F. (2016). The person within:Memory codes for persons and traits using fMRI repetitionsuppression. Social Cognitive and Affective Neuroscience, 11(1), 159–171.

Jenkins, A. C., Macrae, C. N., & Mitchell, J. P. (2008). Repetitionsuppression of ventromedial prefrontal activity duringjudgments of self and others. Proceedings of the NationalAcademy of Sciences of the United States of America, 105(11),4507–4512.

Krienen, F. M., Tu, P.-C., & Buckner, R. L. (2010). Clan men-tality: Evidence that the medial prefrontal cortexresponds to close others. The Journal of Neuroscience,30(41), 13906–13915.

Krueger, F., Barbey, A. K., & Grafman, J. (2009). The medialprefrontal cortex mediates social event knowledge. Trendsin Cognitive Sciences, 13(3), 103–109.

Lewis, P. A., Rezaie, R., Brown, R., Roberts, N., & Dunbar, R. I. M.(2011). Ventromedial prefrontal volume predicts under-standing of others and social network size. NeuroImage,57(4), 1624–1629.

Lieberman, M. D., Jarcho, J. M., & Satpute, A. B. (2004).Evidence-based and intuition-based self-knowledge: AnFMRI study. Journal of Personality and Social Psychology,87(4), 421–435.

Lin, W.-J., Horner, A. J., & Burgess, N. (2016). Ventromedialprefrontal cortex, adding value to autobiographical mem-ories. Scientific Reports, 6(28630), 1–10.

Ma, N., Baetens, K., Vandekerckhove, M., Kestemont, J.,Fias, W., & Van Overwalle, F. (2014). Traits are repre-sented in the medial prefrontal cortex: An fMRI adapta-tion study. Social Cognitive and Affective Neuroscience, 9(8), 1185–1192.

Ma, N., Baetens, K., Vandekerckhove, M., Van der Cruyssen, L.,& Van Overwalle, F. (2014). Dissociation of a trait and avalence representation in the mPFC. Social Cognitive andAffective Neuroscience, 9(10), 1506–1514.

Ma, N.,Wang, S., Yang, Q., Feng, T., & VanOverwalle, F. (2016). Theneural representation of competence traits : An fMRI study.Nature Publishing Group, (June), 1–9. doi:10.1038/srep39609

McClelland, J. L., & Rumelhart, D. E. (1988). Explorations inparallel distributed processing: A handbook of models, pro-grams and exercises. Cambridge, MA: Bradford.

Murray, R. J., Schaer, M., & Debbané, M. (2012). Degrees ofseparation : A quantitative neuroimaging meta-analysisinvestigating self-specificity and shared neural activationbetween self- and other-reflection. Neuroscience andBiobehavioral Reviews, 36(3), 1043–1059.

Powell, J. L., Lewis, P. A., Roberts, N., García-Fiñana, M., &Dunbar, R. I. M. (2012). Orbital prefrontal cortex volumepredicts social network size: An imaging study of individualdifferences in humans. Proceedings. Biological Sciences/TheRoyal Society, 279(1736), 2157–2162.

Roy, M., Shohamy, D., & Wager, T. D. (2012). Ventromedialprefrontal-subcortical systems and the generation ofaffective meaning. Trends in Cognitive Sciences, 16(3),147–156.

Schrouff, J., Rosa, M. J., Rondina, J. M., Marquand, A. F., Chu, C.,Ashburner, J., . . . Mourão-Miranda, J. (2013). PRoNTo:Pattern recognition for neuroimaging toolbox.Neuroinformatics, 11(3), 319–337.

Schurz, M., Radua, J., Aichhorn, M., Richlan, F., & Perner, J.(2014). Fractionating theory of mind: A meta-analysis of

298 E. HELEVEN AND F. VAN OVERWALLE

functional brain imaging studies. Neuroscience andBiobehavioral Reviews, 42, 9–34.

Szpunar, K. K., St Jacques, P. L., Robbins, C. A., Wig, G. S., &Schacter, D. L. (2014). Repetition-related reductions in neuralactivity reveal component processes of mental simulation.Social Cognitive and Affective Neuroscience, 9(5), 712–722.

Van der Cruyssen, L., Heleven, E., Ma, N., Vandekerckhove, M.,& Van Overwalle, F. (2015). Distinct neural correlates ofsocial categories and personality traits. NeuroImage.doi:10.1016/j.neuroimage.2014.09.022

Van Overwalle, F. (2009). Social cognition and the brain: Ameta-analysis. Human Brain Mapping, 30(3), 829–858.

Van Overwalle, F., & Baetens, K. (2009). Understanding others’actions and goals by mirror and mentalizing systems: Ameta-analysis. NeuroImage, 48(3), 564–584.

Van Overwalle, F., Ma, N., & Baetens, K. (2015). Nice or nerdy? Theneural representation of social and competence traits. SocialNeuroscience, 1–12. doi:10.1080/17470919.2015.1120239

Welborn, B. L., & Lieberman, M. D. (2015). Person-specifictheory of mind in medial pFC. Journal of CognitiveNeuroscience, 27(1), 1–12.

Wood, J. N., & Grafman, J. (2003). Human prefrontal cortex:Processing and representational perspectives. NatureReviews. Neuroscience, 4(2), 139–147.

SOCIAL NEUROSCIENCE 299