A nexus model of the temporal–parietal junctionR. McKell Carter1,2 and Scott A. Huettel1,2,3

1 Center for Cognitive Neuroscience, Duke University, Durham, NC, USA2 Brain Imaging and Analysis Center, Duke University, Durham, NC, USA3 Department of Psychology and Neuroscience, Duke University, Durham, NC, USA

Review

The temporal–parietal junction (TPJ) has been proposedto support either specifically social functions or non-spe-cific processes of cognition such as memory and atten-tion. To account for diverse prior findings, we propose anexus model for TPJ function: overlap of basic processesproduces novel secondary functions at their convergence.We present meta-analytic evidence that is consistent withthe anatomical convergence of attention, memory, lan-guage, and social processing in the TPJ, leading to ahigher-order role in the creation of a social context forbehavior. The nexus model accounts for recent examplesof TPJ contributions specifically to decision making in asocial context and provides a potential reconciliation forcompeting claims about TPJ function.

IntroductionThe term ‘temporal-parietal junction’ (TPJ) refers to theregion of the cerebral cortex that lies along the boundary ofthe temporal and parietal lobes. This abstract label, whichin practice does not map precisely onto any specific anatom-ical features, has become increasingly common in the cogni-tive neuroscience literature. Neuroimaging studies havefound activation in the TPJ within diverse experimentalparadigms that manipulate memory [1–3], attention [4–6],language [7,8], and social cognition [9–12] (Figure 1). Per-haps more strikingly than for any other cerebral region, thefunctional properties of the TPJ have been subject to con-tinual debate. There exists no consensus on whether acommon function can account for these diverse experimen-tal results, whether there is spatial heterogeneity acrossthese functions, or even whether the TPJ should be consid-ered a unified brain region (Box 1).

The challenge of understanding TPJ function has becomemost evident in research on social cognition. One perspec-tive holds that the TPJ plays a critical role in processesrelated to social cognition, such as mentalizing [13], per-spective taking [14], and detection of social agents in theenvironment [15–17]. Supporting this perspective is awealth of data from functional neuroimaging studies, whichreveal almost ubiquitous activation in the TPJ during socialcognition tasks (Figure 1), as well as less-extensiveelectrophysiological [18,19] and lesion data [20,21]. Recentwork, for example, has linked TPJ structure and function to

1364-6613/$ – see front matter

� 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tics.2013.05.007

Corresponding author: Huettel, S.A. (scott.huettel@duke.edu).

328 Trends in Cognitive Sciences July 2013, Vol. 17, No. 7

pro-social tendencies, such as altruism [16,22]. An alterna-tive perspective contends that the TPJ plays no specific rolein social cognition. Neurologists have long recognized thatdamage to the right inferior parietal lobe can lead to thecondition of hemispatial neglect, which is characterized byfrank deficits in attentional control but not necessarily byspecific deficits in social cognition (reviewed in [23,24]).Accordingly, activation of the TPJ in social cognition taskscould reflect component processes (e.g., selective attention)that are critical for, but not unique to, social interactions[25,26].

In this review, we evaluate evidence for each of theseperspectives. Although each makes distinct predictions forthe cognitive function supported by the TPJ, if the region isto be considered a single functional unit, the perspectivescan be reconciled by relaxing the assumption that TPJ is anintegrated whole. Here, we propose a nexus model for TPJfunction. We contend that the functionally defined TPJreflects anatomically convergent streams that supportdistinct cognitive processes. The combination of thosestreams in the TPJ supports the construction of a socialcontext for behavior, which in turn alters modes of proces-sing elsewhere in the brain.

Social contexts shape behaviorResearch from economics, psychology, and other socialsciences indicates that, under some circumstances, theestablishment of a social context can change behavior. Ifa logical reasoning problem is reframed into a questionabout social norms, then it can be more easily and accu-rately solved [27]. Decisions in economic games changedepending on whether the opponent is a human or com-puter program, even if there is no difference in the objectiveplay of that opponent [28,29]. These and other examplescould be criticized, however, as being more engaging thantheir non-social counterparts [25]. Although such criti-cisms can be answered within individual studies by con-trolling for familiarity or task difficulty, it is difficult to ruleout all possible confounding factors.

To obtain conclusive evidence that social context repre-sents a qualitatively distinct state, researchers would needto demonstrate that the presence of a social context doesnot simply exaggerate or dampen behavior but ratherqualitatively changes it. Such evidence comes from studiesof the effects of monetary incentives on social behavior[30,31]. Beginning in the early 1970s, economists recog-nized that money sometimes served as a disincentive for

Mentalizing ToMInten�on Expression

EBA Body Biological Gaze

Social

AG SMG

AG SMG

AG SMG

AG SMG

AG SMG

AG SMG

AG SMG

AG SMG

AG SMG

TRENDS in Cognitive Sciences

Figure 1. Meta-analytic evidence of social function encoding in the greater temporal–parietal junction (TPJ). Reverse-inference maps from http://neurosynth.org, rendered

on the right lateral surface of a single brain, illustrate the location of activations that are likely associated with social cognition. Within the sampled studies (approximately

4000 included), activation in the colored region was significantly predictive of the displayed phrase. The reverse-inference map for the term ‘TPJ’ is displayed as a black

outline for reference and includes portions of the angular gyrus (AG, white broken line [85]), supramarginal gyrus (SMG, white broken line [85]), and superior temporal

gyrus and/or sulcus. Activation within the lower portion of the TPJ, along the superior temporal sulcus, is significantly associated with studies using the term ‘Social’.

Within the TPJ, in the AG, reverse inferences are statistically valid for terms that reflect the process of thinking about another person’s mental processes [‘Intention’,

‘Mentalizing’, and ‘Theory of Mind’ (ToM)], as well as terms reflecting the interpretation of physical actions of social agents, such as eye gaze (‘Gaze’), biological motion

(‘Biological’), and facial expressions (‘Expression’). Immediately inferior to the TPJ, significant reverse inference clusters are present for the representation of social stimuli

such as bodies [‘Body’ and extrastriate body area (‘EBA’)]. The displayed statistical maps, downloaded during the fall of 2012, are whole-brain corrected for multiple

comparisons using a false discovery rate of P < 0.05 [51].

Box 1. TPJ: combining processes to create novel functions

A diverse set of cognitive functions has been linked to the TPJ:

attentional reorienting, language processing, numerical cognition,

episodic memory encoding, and various processes of social

cognition. No consensus yet exists as to whether these functions

are each supported by distinct subregions within the TPJ or whether

they all rely, at least in part, on some similar computational process.

A recent theoretical review by Cabeza and colleagues [67] advances

the perspective that all these diverse functions rely on bottom-up

attention, which is considered to involve capture of attention both

by external stimuli and by information entering working memory

from long-term memory. This view supports an overarching

function of ventral parietal cortex, which subsumes the TPJ, and

argues against the idea that the ventral parietal cortex is fractionated

into domain-specific subregions. The core idea of this perspective,

that convergent processes produce novel functions, can be general-

ized to other combinations of processes known to engage lateral

parietal cortex. By relaxing the constraint on spatial homogeneity

and incorporating the anatomical positions of contributing path-

ways (two potential changes alluded to by Cabeza and colleagues),

the TPJ could be envisioned to play an even broader integrative

role, one that goes beyond interactions between attention and

memory.

Review Trends in Cognitive Sciences July 2013, Vol. 17, No. 7

engaging in pro-social behavior; for example, paying peopleto donate blood led to lower rates of donation [32,33]. In alandmark study, Gneezy and Rustichini introduced fines todiscourage parents from being late when picking up theirchildren from daycare [34], but the fines in fact increasedthe number of late arrivals. Across this and many otherstudies [35,36], it has been shown that the introduction ofeconomic incentives can eliminate social motivations fordecision making (e.g., concern for others’ time), changingbehavior dramatically. This phenomenon has becomeknown as ‘reward undermining’ or ‘motivation crowding’[37] and has been demonstrated to coincide with changes invalue representations in the brain [38].

Reward undermining provides a prototypic example ofhow establishment of a social context, or elimination ofthat social context, determines the decisions that humansmake. However, it is hardly the only such example. Anartificial social context can be created within the laborato-ry through a minimal-group paradigm, in which uncon-nected individuals are separated into groups by somerandom or arbitrary assignment (e.g., distributing t-shirts;

329

Box 2. Meta-analyses: advantages and limitations

Neurosynth ([51]; http://neurosynth.org) and other meta-analytic

efforts such as BrainmapWeb ([86], http://brainmap.org), SumsDB

([87], http://sumsdb.wustl.edu/), and Talairach Daemon ([88];

http://talairach.org) seek to leverage the explosive growth in

neuroimaging over the past decade to produce a reliable

functional brain map. These efforts have become powerful tools,

allowing robust specification of regions of interest and quantita-

tive descriptions of functional specificity. They can even provide

information about the practice of research by mapping language

use and providing pointers to open research questions [89]. With

this promise, meta-analytic tools also bring new pitfalls. A meta-

analysis acts as a sort of popularity contest that can produce either

positive or negative consequences. For example, a finding that

was incorrect but influential might enter into meta-analyses and

thus persist long after it had been corrected by a single crucial

study. The fragmentation of cognitive neuroscience has enabled

some correction in this arena to take place by allowing multiple

fields to label the function of a single region. In addition, even for a

large database such as Neurosynth (approximately 4000 studies

by early 2013), the sample of studies may not be sufficient to make

strong conclusions about particular terms or concepts; neither

does it necessarily represent the literature as a whole. How to

optimize meta-analytic approaches to neuroscience has become

an active area of research, one that holds promise for undisturbed

and open evaluations of commonalities across an expansive

literature.

Review Trends in Cognitive Sciences July 2013, Vol. 17, No. 7

[39]). After such groups are established, individuals be-come more willing to allocate money to in-group individu-als [40] or to punish out-group individuals [41]. Theseeffects are exaggerated by the administration of the hor-mone oxytocin [42] and dampened when a personal rela-tionship is established with an individual from within oroutside the group [43]. The common thread across thesedisparate areas of research is the idea that the presence orabsence of a social context changes the mode of processing,in turn altering the decisions that humans make.

Social function in the TPJSelect lesion [20,21,44] and stimulation [45,46] studiesprovide evidence for social function in the TPJ (Figure 1,black outline), but more generally, findings are highlydependent on spatial location. As described above, the bestevidence for social function in the TPJ region comes fromneuroimaging studies [47,48]. To establish the replicabilityof these findings, data from a large literature base can bereadily combined using meta-analytic techniques thatevaluate how selectively activation in a particular brainregion is linked to specific cognitive processes; this practiceis labeled ‘reverse inference’ [49]. Reverse inference, espe-cially when based on the results of a single study, may have

Sta�c

Moving

Context

‘EBA’‘EBA’

‘Biological’‘Biological’

‘Inten�ons’‘Inten�ons’

TRENDS in Cognitive Sciences

Figure 2. Social context is constructed from information represented immediately inferior to the temporal–parietal junction (TPJ). Social context construction is proposed

to begin with the neural representation of static social images [extrastriate body area (‘EBA’)] in the extrastriate cortex. These static images could then be used to construct

representations of moving social entities [biological motion (‘Biological’)]. Representations of moving individuals could then be incorporated in a context for interpreting

intentions (‘Intention’). As in Figure 1 (main text), the black outline represents the extent of the reverse-inference map for the term ‘TPJ’. Reverse inference maps were

downloaded from http://neurosynth.org during the fall of 2012, are rendered on the right lateral surface of a single brain, and are whole-brain corrected for multiple

comparisons [51] using false discovery rate correction at P < 0.05.

330

Review Trends in Cognitive Sciences July 2013, Vol. 17, No. 7

limited utility beyond guiding future hypothesis testing[50]. When drawn from many studies, however, reverseinferences can provide insight into functional dissociationsbetween brain regions.

We utilized the functional (f)MRI meta-analytic re-source Neurosynth (http://neurosynth.org, [51], approxi-mately 4000 total studies) to identify brain regionswhere activation is likely to have been produced by socialfunction (Figure 1). Although these results are correctedfor the whole-brain volume, we describe only those in thegreater TPJ region, focusing particularly on the right TPJ,where evidence for social function has been consideredparticularly strong, although similar arguments can bemade for the left TPJ. Below we present reverse-inferencestatistical maps for specific terms (displayed in singlequotes) searched for on the Neurosynth website.

Across many studies, there is considerable evidence thatactivation in the TPJ and surrounding regions is selectivelyrelated to the perception and evaluation of social stimuli(Figure 1). Two regions on the lateral surface of the occipitalcortex are selectively active for biological stimuli [52]: onerepresenting faces (the lateral occipital face area [53,54]),and the other representing other parts of the body (‘EBA’and ‘Body’, [55]). Regions overlapping with the inferiorportion of the TPJ are linked to more complex social func-tions, such as biological motion (‘Biological’, [56–58]), gaze

AG SMG

Figure 3. Attention and social cognition evoke overlapping but separable statistical m

activations produced by attention-reorienting (orange, ‘Corbetta’; one of the primary rese

tasks. This spatial bias in activation peaks is consistent with that found when activations

(inset) and with comparison of the two tasks in the same experiment [26,74]. The black o

term ‘TPJ’. The angular gyrus (AG, white broken line [85]) and supramarginal gyrus (SM

were downloaded from http://neurosynth.org during the fall of 2012, are rendered on th

comparisons [51] using a false discovery rate of P < 0.05. Reproduced, with permission

detection (‘Gaze’, reviewed in [59]), and identification offacial expressions (‘Expression’, [57,60,61]). In addition,activation in the angular gyrus (AG), in particular, predictscomplex social functions, such as mentalizing (‘Mentaliz-ing’), intention (‘Intention’), and theory of mind (‘Theory ofMind’). Although these reverse-inference maps provide anunbiased representation of the sampled literature, thereremains the possibility of publication bias; that is, thatstudies observing these relations would be more likely toenter the literature in the first place (Box 2). Even with thiscaveat, the evidence seems strong enough to conclude that, ifa study finds activation in the AG (especially the inferiorAG), the brain was likely engaged in some form of socialprocessing.

Collectively, these results point toward a topography ofsocial processing along the lateral surface of the posteriorcortex, beginning with face-sensitive regions that borderretinotopic visual areas and continuing to regions repre-senting social context in the AG (Figure 2). This configu-ration evokes one of the organizing principles of neuralsystems in the brain: that more abstracted stimulus repre-sentations are constructed from simpler inputs [53,62].Here, static social stimuli are first transformed into mov-ing agents and then placed in a context where their actionscan be interpreted. The potential existence of a socialprocessing stream was recognized in some of the earliest

Difference map:Reorien�ng versus theory of mind

TRENDS in Cognitive Sciences

aps within the temporal–parietal junction (TPJ). Spatial differentiation between

archers of attention-reorienting) and theory-of-mind [blue, ‘Theory of Mind’ (ToM)]

were compared for the two functions in a meta-analysis by Decety and Lamm [66]

utline represents the extent of the significant reverse-inference map cluster for the

G, white broken line [85]) are also displayed for reference. Reverse inference maps

e right lateral surface of a single brain, and are whole-brain corrected for multiple

, from [66] (inset).

331

Review Trends in Cognitive Sciences July 2013, Vol. 17, No. 7

fMRI studies [47], was a motivating factor behind thelabeling of the TPJ as playing a role in social processing[11], and has implications for clinical cognitive neurosci-ence, including the links between TPJ function and autism[63,64].

The idea that the TPJ integrates information to con-struct a social context is consistent with the more narrowinterpretations of social activation in the TPJ [65]. How-ever, it stands in opposition to alternative explanations ofsocial context activation in the TPJ, including argumentsfor effects specific to attention [25,66], memory [67], andlanguage [68,69]. Next, we consider the most prominentsuch argument, which reinterprets TPJ activation duringsocial interaction as a reorienting of attention to an exter-nal source.

Broader explanations for the role of the TPJ in socialcognitionA key challenge for theories positing a social function forthe TPJ has been the consistent link between lateraltemporal and parietal cortex and aspects of attention(reviewed in [70]). Patients with lesions near the TPJregion in the right hemisphere often display a strikinginability to attend to stimuli that are presented in the leftvisual field (i.e., neglect, [23,24]). Moreover, tasks thatinvolve the reorienting of attention to an external cue(e.g., in the Posner cuing task) reliably evoke activationin the TPJ (see the meta-analysis in [66] as well as theNeurosynth summary in Figure 3). Reorienting attentionto important unexpected targets (e.g., invalidly signaled

Social

Memory

A�en�on

Figure 4. The temporal–parietal junction (TPJ) is anatomically positioned at the nex

inference statistical maps highlight those regions of the brain where activations are likel

(blue), and ‘language’ (gold). Processing streams associated with these terms converge

neighboring processes. One such novel function that requires both the representation

theory of mind (reverse inference map, white broken line). Statistical maps were downl

lateral surface of a single brain, and are whole-brain corrected for multiple comparison

332

targets), but not to expected events (e.g., validly signaledtargets), has been interpreted as a mechanism for externalobjects to pull attention away from internally driven pro-cesses. With this role in reorienting in mind, it was pro-posed that previous evidence of social function in the TPJcould be reinterpreted as the orientation of attention to aparticular class of important external stimuli [25].

Externally directing attention may indeed be critical foradaptive social cognition. One hypothesized mechanism forrepresenting and responding to social stimuli is to simulateanother person’s perceptions, thoughts, and actions usingthe same systems that support those processes for self [71–73]. Doing so, in principle, introduces the problem ofdifferentiating one’s own perceptions and behaviors fromthose of an individual whom one is observing. If the role ofthe TPJ in social function were one of externally orientingattention, it could serve exactly this role by signaling thatsimulated social processes originate from an externalsource. Consistent with the idea that a similar region isimplicated in externally reorienting attention and in socialcognition, both meta-analyses [66] and direct comparisonsof attention reorienting and theory-of-mind tasks [26,74]find overlapping brain activation for both processingdomains in the TPJ.

Although both within the TPJ, social cognition andattentional processes do evince some spatial heterogeneity:activation peaks associated with theory of mind fall largelyin the inferior AG and peaks associated with reorientingactivation fall within the supramarginal gyrus (SMG)(Figure 3). This spatial bias is also present in the original

Language

Social

A�en

�on

Memory

TRENDS in Cognitive Sciences

us of the social, attention, memory, and language processing streams. Forward

y to be found for studies using the terms ‘social’ (green), ‘attention’ (red), ‘memory’

on the ‘TPJ’ (black line), where novel functions are produced from combinations of

of social stimuli and the context provided by attention, memory, and language is

oaded from http://neurosynth.org during the fall of 2012, are rendered on the right

s [51].

Review Trends in Cognitive Sciences July 2013, Vol. 17, No. 7

work proposing overlapping functions (e.g., contrasting theactivation for theory of mind with the reorienting of atten-tion, [66]) and in previous anatomical segmentations of theTPJ [75,76]. An alternative interpretation of this corre-spondence, therefore, is that social cognition not onlyshares some processing with attentional reorienting, butalso contains dissociable elements. In particular, buildinga social context depends on information that cannot readilybe extracted from ongoing perception but instead arisesfrom internal goals, memory, and semantic abstraction.Thus, social cognition may not be subsumed by attentionbut instead may integrate information derived from atten-tion with that derived from other processing streams, usingthat integrated information to construct a social context.

A nexus model for TPJ functionIncorporating social input into decision making requiresmore than the recognition of another person or agent. Agesture from that agent might be interpreted in differentways depending on other available information (e.g.,past experience or the description by the agent of their

Box 3. Construction of social context in decision making

A recent study investigated decision making in social and nonsocial

contexts, as fMRI participants played a simplified poker game against

human and computer opponents [82]. The game was a variant of von

Neumann’s one-card poker [90], in which a player was dealt a single

high or low card and must decide whether to bet or fold. The authors

used combinatoric multivariate pattern analysis (cMVPA) to predict

bluffing decisions (whether someone bets when dealt a low card) from

patterns of activation in anatomically defined regions. For each brain

region, the authors computed the degree to which its predictive power

was independent of that of other regions. Consistent with prior studies,

most regions typically associated with social cognition (e.g., frontal

pole or precuneus) carried independent information that predicted

participants’ eventual bluffing decisions against both human and

computer opponents (Figure IA). For most such regions, however, the

results were consistent with an overall increase in information when

playing against the human opponent. Information content in the TPJ,

by contrast, was selective for the social context (Figure IB). Moreover,

the selectivity of activation in the TPJ was eliminated in those

participants who judged the computer to be the more competent

opponent, consistent with research on dehumanization [91,92].

These results support the conception that social decision making

engages processes that would also be engaged in nonsocial contexts,

at least for regions such as the medial prefrontal cortex and

precuneus. However, they also argue that the TPJ contributes to

decisions only when those decisions are embedded in a social

context; that is, when interacting with another agent whose internal

states can be modeled and whose actions are relevant for one’s own

behavior.

Figure I. A distinct role for the temporal–parietal junction (TPJ) in social

cognition. Recent work has shown that the TPJ contains unique information

that predicted bluffing against a social but not a nonsocial opponent [82]. (A)

Multivariate pattern analysis was used to measure the independent information

contained within each region toward predicting poker bluffs against a human

opponent (Y-axis) and poker bluffs against a computer opponent (X-axis).

Orthogonal distance regression was used to identify a pattern of greater

information content in the social context, consistent with perspectives that

assume a single process that is more engaged when interacting with other

humans (broken line). (B) One region, the TPJ, stood out as an outlier: it was one

of the regions that most predicted decisions against the human opponent, but

the least predictive region against the computer opponent. This effect was

modulated by the participant’s opinion of their opponent. If the computer was

considered to be a better opponent than the human, the TPJ was no longer

predictive. Adapted from [82].

intentions). Ambiguity associated with the perceptual stim-ulus (so that perception can be mapped to the appropriateaction) can be resolved through the construction of a socialcontext.

To construct such a context, the brain must engageprocessing associated with nominally disparate domains:perception (especially of biological stimuli), attention,memory, and language and/or semantics. The functionalorganization of the posterior lateral surface of the cerebralcortex hints at the integration of these domains in the TPJ(Figure 4). As described previously (Figure 2), processingassociated with perception of social agents falls along atopography that begins with the identification of faces andbodies in the lateral occipital cortex and builds in abstrac-tion dorsally into the TPJ, where a context for the socialagent can be recalled and prioritized to support inferentialprocesses such as mentalizing.

Attention and memory processing streams converge inthe TPJ, as well. Studies using the term ‘attention’ arelikely to report activation in two distinct attentional path-ways, reminiscent of the ‘action’ and ‘perception’ pathways

(A)10

0

8

0

–2

0

–2

FP

pCG

PCuPCC

ACG

HumanComputer

0 8

10

Hum

an o

ppon

ent i

nfor

ma�

on

Computer opponent informa�on

TPJ

(B)

TPJ

Be�er opponentKey:Hu

man

opp

onen

t inf

orm

a�on

Computer opponent informa�on

TRENDS in Cognitive Sciences

333

Box 4. Outstanding questions

� What computations are supported by regions within the parietal

cortex where attention and memory processing streams overlap,

compared to the processing within each stream independently?

� Are there common topographic features that predict processing

stream convergence in other areas of the brain?

� What is the relation between number processing and attention in

the dorsal parietal cortex?

� How does information in the TPJ shape processing in regions

involved with value computations?

� Can meta-analytic descriptions of function (such as Neurosynth)

be used to improve the prediction of behavioral deficits from

lesion mapping data?

� What is the appropriate spatial resolution for description of

neuroimaging results, given evidence for overlapping functions?

� How might hemispheric asymmetries in TPJ function interact with

distinct types of social context?

Review Trends in Cognitive Sciences July 2013, Vol. 17, No. 7

proposed by Goodale and Milner [77]. Studies of ‘memory’reliably report activation in parallel neighboring dorsaland ventral clusters. These two regions have been pro-posed to follow a similar action–perception dichotomy; thatis, endogenous, indirect, or goal-directed recall is distinctfrom exogenous, direct, or stimulus-driven recall [2,67].The clusters of activation likely to be found in attentionand memory studies nearly surround the TPJ. The SMG,within the TPJ, is perfectly situated to allow the conver-gence of the dorsal and ventral attention and/or memorystreams. Consistent with a nexus model of TPJ function,crosstalk between the two streams in the SMG couldsupport a new, otherwise unavailable function: allowinga goal to be interrupted to reorient attention to a salientstimulus (as proposed for the TPJ by Corbetta and collea-gues [25]).

Finally, studies using the term ‘language’ evoke activa-tion along the superior temporal sulcus (Figure 4). Alongthe posterior portion of the superior temporal sulcus, socialand language processing streams converge, and we findactivations (Figure 2) commonly evoked by studies of eyegaze, facial expressions, and the perception of agency innonbiological objects. The social, attention, memory, andlanguage processing streams evince partially overlappingactivation within the TPJ, consistent with the idea thattheir output is integrated to create a social context. Inthese and other cases, the spatial juxtaposition of morebasic processes may allow neighboring groups of neurons tointeract in a way that produces more complex and other-wise unavailable functions.

We propose that TPJ activation signals not only thepresence of another social agent, but also the convergenceof multiple functions in the TPJ nexus to establish a socialcontext: that is, a decision situation that involves anotheragent whose goals or actions can influence one’s own be-havior. As examples, self-reported altruism has been linkedto changes in TPJ activation associated with observinganother agent [16], pro-social behavior in economic gameshas been linked to TPJ structure and function [22], andventromedial prefrontal cortex (vmPFC) value signals as-sociated with charitable giving have been shown to bemodulated by activation in TPJ [78]. Two recent studiesindicate that TPJ activation can be modulated by theperceived relevance of another’s actions for one’s own be-havior. Participants who engaged in strategic deceptionwithin a bargaining game showed activation in the TPJthat was greatest on high-value trials (i.e., those where thedeception led to large rewards); no such effect of value wasobserved in participants who engaged in systematic decep-tion across all trials [79]. In addition, it has been recentlyshown that the TPJ contains unique information predictiveof bluffing behavior in an interactive poker game, but onlywhen participants are competing against a human oppo-nent who is seen as a credible opponent (Box 3). Theimportance of context to social function has a parallel inexperiments using oxytocin, which amplifies social behaviorwhen administered to members of a group but has reducedeffects in a nonsocial or out-group context [42,80]. Collec-tively, this evidence indicates that the TPJ contributes todecision making specifically when there is a social contextrelevant for current behavior.

334

Concluding remarks and predictionsThe perspective proffered here points to a reconciliation ofthe ongoing debate on the contributions of the TPJ to socialcognition. So far, that debate has revolved around thegenerally accepted evidence that the TPJ is active duringtasks involving social cognition, with one perspective ar-guing for a specifically social function (e.g., mentalizing)and another for a general, nonsocial function (e.g., selectiveattention). Impeding resolution of the debate has beenimprecision in the functional characterization of theTPJ; as shown in Figure 3, meta-analytic evidence sup-ports the view that multiple functions produce spatiallydifferentiable (but overlapping) activations within theTPJ. Based on that evidence, and on the specific conditionsunder which TPJ activation is elicited in decision making,we argue that within the TPJ, the AG supports the con-struction of a social context for behavior based on integrat-ed information from converging processing streams.

Several predictions–some tested, some testable–followfrom this proposal (Box 4). First, changing from a nonsocialto a social context should increase activation within theTPJ and its functional connectivity with other task-rele-vant regions (e.g., [81]). Should the TPJ indeed be criticalfor establishment of that social context, then its activationshould be observed earlier than that of other regionsimplicated in social processing. In addition, when socialinformation becomes irrelevant for behavior, the TPJshould be disengaged even if a social agent is still present[82].

Second, there should be spatial biases of processingwithin the TPJ (and neighboring regions) that mirrorthe anatomical organization of the contributing processingstreams. That is, the topography within the TPJ should bederived from the functional properties of surroundingregions. For example, recent quantitative analyses havelinked the incidence of spatial neglect specifically to lesionsof the anterior TPJ, specifically the posterior superiortemporal gyrus (pSTG), and not to more posterior regionsof parietal cortex [44]. That result is consistent with aconvergence of attentional and semantic streams in thepSTG. Convergence of other pairs of processing streamswill contribute to different higher-order processes (e.g.,social and memory processing to contribute to autobio-graphical memory).

Review Trends in Cognitive Sciences July 2013, Vol. 17, No. 7

Third, the neural machinery used to construct a socialcontext can be co-opted for use in nonsocial contexts. Undersome conditions, stimuli that are not intrinsically socialmay be best processed through the introduction of socialcontext, even if one would not exist, otherwise. This phe-nomenon is well documented within the domain of socialcognition. As examples, the TPJ shows robust activation tosimple visual displays where the movements of shapes fit asocial narrative [83] or where the change in the displayimplies the action of another agent [16]. However, similareffects should be possible within decision making, particu-larly when individuals can improve decisions by simulat-ing others’ perspectives [84].

As a concluding note, we emphasize that heterogeneitywithin the TPJ is still consistent with a role for that regionin social decision making. Under the perspective advocat-ed, the contributions of the TPJ to social cognition arisebecause it is a heterogeneous region: it sits at the nexus ofseveral distinct processing streams, each of which carriesinformation relevant to social cognition. When the infor-mation in those streams becomes relevant for a particulardecision, the TPJ helps establish a social context for deci-sion making, shaping processing elsewhere in the brain.

AcknowledgmentsThe authors would like to acknowledge constructive conversations withRoberto Cabeza and manuscript feedback from Edward McLaurin,Lawrence Ngo, and Amy Winecoff. R.M.C. was supported by NIMHR01-86712. S.A.H. was supported by an Incubator Award from the DukeInstitute for Brain Sciences.

References1 Wagner, A.D. et al. (2005) Parietal lobe contributions to episodic

memory retrieval. Trends Cogn. Sci. 9, 445–4532 Cabeza, R. et al. (2008) The parietal cortex and episodic memory: an

attentional account. Nat. Rev. Neurosci. 9, 613–6253 Vilberg, K.L. and Rugg, M.D. (2008) Memory retrieval and the parietal

cortex: a review of evidence from a dual-process perspective.Neuropsychologia 46, 1787–1799

4 Arrington, C.M. et al. (2000) Neural mechanisms of visual attention:object-based selection of a region in space. J. Cogn. Neurosci. 12 (Suppl.2), 106–117

5 Corbetta, M. and Shulman, G.L. (2002) Control of goal-directed andstimulus-driven attention in the brain. Nat. Rev. Neurosci. 3, 201–215

6 Serences, J.T. et al. (2005) Coordination of voluntary and stimulus-driven attentional control in human cortex. Psychol. Sci. 16, 114–122

7 Kobayashi, C. et al. (2007) Cultural and linguistic effects on neuralbases of ‘Theory of Mind’ in American and Japanese children. BrainRes. 1164, 95–107

8 Binder, J.R. et al. (2009) Where is the semantic system?. A criticalreview and meta-analysis of 120 functional neuroimaging studies.Cereb. Cortex 19, 2767–2796

9 Goel, V. et al. (1995) Modeling other minds. Neuroreport 6, 1741–174610 Happe, F. et al. (1996) ‘Theory of mind’ in the brain. Evidence from a

PET scan study of Asperger syndrome. Neuroreport 8, 197–20111 Gallagher, H.L. et al. (2000) Reading the mind in cartoons and stories:

an fMRI study of ‘theory of mind’ in verbal and nonverbal tasks.Neuropsychologia 38, 11–21

12 Saxe, R. and Kanwisher, N. (2003) People thinking about thinkingpeople: the role of the temporo-parietal junction in ‘theory of mind’.Neuroimage 19, 1835–1842

13 Van Overwalle, F. (2009) Social cognition and the brain: a meta-analysis. Hum. Brain Mapp. 30, 829–858

14 Aichhorn, M. et al. (2006) Do visual perspective tasks need theory ofmind? Neuroimage 30, 1059–1068

15 Schultz, J. et al. (2005) Activation in posterior superior temporal sulcusparallels parameter inducing the percept of animacy. Neuron 45, 625–635

16 Tankersley, D. et al. (2007) Altruism is associated with an increasedneural response to agency. Nat. Neurosci. 10, 150–151

17 Mar, R.A. et al. (2007) Detecting agency from the biological motion ofveridical vs animated agents. Soc. Cogn. Affect. Neurosci. 2, 199–205

18 Sabbagh, M.A. and Taylor, M. (2000) Neural correlates of theory-of-mindreasoning: an event-related potential study. Psychol. Sci. 11, 46–50

19 Liu, D. et al. (2004) Decoupling beliefs from reality in the brain: an ERPstudy of theory of mind. Neuroreport 15, 991–995

20 Samson, D. et al. (2004) Left temporoparietal junction is necessary forrepresenting someone else’s belief. Nat. Neurosci. 7, 499–500

21 Apperly, I.A. et al. (2007) Testing the domain-specificity of a theory ofmind deficit in brain-injured patients: evidence for consistentperformance on non-verbal, ‘reality-unknown’ false belief and falsephotograph tasks. Cognition 103, 300–321

22 Morishima, Y. et al. (2012) Linking brain structure and activation intemporoparietal junction to explain the neurobiology of humanaltruism. Neuron 75, 73–79

23 Corbetta, M. and Shulman, G.L. (2011) Spatial neglect and attentionnetworks. Annu. Rev. Neurosci. 34, 569–599

24 Karnath, H.-O. and Rorden, C. (2012) The anatomy of spatial neglect.Neuropsychologia 50, 1010–1017

25 Corbetta, M. et al. (2008) The reorienting system of the human brain:from environment to theory of mind. Neuron 58, 306–324

26 Mitchell, J.P. (2008) Activity in right temporo-parietal junction is notselective for theory-of-mind. Cereb. Cortex 18, 262–271

27 Cosmides, L. (1996) Are humans good intuitive statisticians after all?Rethinking some conclusions from the literature on judgment underuncertainty. Cognition 58, 1–73

28 Sanfey, A.G. et al. (2003) The neural basis of economic decision-makingin the Ultimatum Game. Science 300, 1755–1758

29 Haley, K.J. and Fessler, D.M. (2005) Nobody’s watching? Evol. Hum.Behav. 26, 245–256

30 Lepper, M.R. et al. (1973) Undermining children’s intrinsic interestwith extrinsic reward: a test of the ‘overjustification’ hypothesis. J.Pers. Soc. Psychol. 28, 129–137

31 Deci, E.L. (1971) Effects of externally mediated rewards on intrinsicmotivation. J. Pers. Soc. Psychol. 18, 105–115

32 Titmuss, R.M. (1970) The Gift Relationship: From Human Blood toSocial Policy. George Allen and Unwin

33 Mellstrom, C. and Johannesson, M. (2008) Crowding out in blooddonation: was Titmuss right? J. Eur. Econ. Assoc. 6, 845–863

34 Gneezy, U. and Rustichini, A. (2000) A fine is a price. J. Legal Stud. 29,1–17

35 Frey, B.S. (1997) The cost of price incentives: an empirical analysis ofmotivation crowding-out. Am. Econ. Rev. 87, 746–755

36 Deci, E.L. et al. (1999) A meta-analytic review of experimentsexamining the effects of extrinsic rewards on intrinsic motivation.Psychol. Bull. 125, 627–668 discussion 692–700

37 Frey, B.S. and Jegen, R. (2001) Motivation crowding theory. J. Econ.Surv 15, 589–611

38 Murayama, K. et al. (2010) Neural basis of the undermining effect ofmonetary reward on intrinsic motivation. Proc. Natl. Acad. Sci. U.S.A.107, 20911–20916

39 Dunham, Y. et al. (2011) Consequences of ‘minimal’ group affiliations inchildren. Child Dev. 82, 793–811

40 Chen, Y. and Li, S.X. (2009) Group identity and social preferences. Am.Econ. Rev. 99, 431–457

41 Bernhard, H. et al. (2006) Parochial altruism in humans. Nature 442,912–915

42 De Dreu, C.K.W. et al. (2010) The neuropeptide oxytocin regulatesparochial altruism in intergroup conflict among humans. Science 328,1408–1411

43 Tanis, M. and Postmes, T. (2005) A social identity approach to trust:interpersonal perception, group membership and trusting behaviour.Eur. J. Soc. Psychol. 35, 413–424

44 Smith, D.V. et al. (2013) Decoding the anatomical network of spatialattention. Proc. Natl. Acad. Sci. U.S.A. http://dx.doi.org/10.1073/pnas.1210126110

45 Young, L. et al. (2010) Disruption of the right temporoparietal junctionwith transcranial magnetic stimulation reduces the role of beliefs inmoral judgments. Proc. Natl. Acad. Sci. U.S.A. 107, 6753–6758

46 Santiesteban, I. et al. (2012) Enhancing social ability by stimulatingright temporoparietal junction. Curr. Biol. 22, 2274–2277

335

Review Trends in Cognitive Sciences July 2013, Vol. 17, No. 7

47 Gallagher, H.L. and Frith, C.D. (2003) Functional imaging of ‘theory ofmind’. Trends Cogn. Sci. 7, 77–83

48 Carrington, S.J. and Bailey, A.J. (2009) Are there theory of mindregions in the brain?. A review of the neuroimaging literature.Hum. Brain Mapp. 30, 2313–2335

49 Poldrack, R.A. (2006) Can cognitive processes be inferred fromneuroimaging data? Trends Cogn. Sci. 10, 59–63

50 Poldrack, R.A. (2011) Inferring mental states from neuroimaging data:from reverse inference to large-scale decoding. Neuron 72, 692–697

51 Yarkoni, T. et al. (2011) Large-scale automated synthesis of humanfunctional neuroimaging data. Nat. Methods 8, 665–670

52 Pinsk, M.A. et al. (2009) Neural representations of faces and body partsin macaque and human cortex: a comparative FMRI study. J.Neurophysiol. 101, 2581–2600

53 Hasson, U. et al. (2003) Large-scale mirror-symmetry organization ofhuman occipito-temporal object areas. Neuron 37, 1027–1041

54 Tsao, D.Y. et al. (2008) Comparing face patch systems in macaques andhumans. Proc. Natl. Acad. Sci. U.S.A. 105, 19514–19519

55 Downing, P.E. et al. (2001) A cortical area selective for visualprocessing of the human body. Science 293, 2470–2473

56 Howard, R. et al. (1996) A direct demonstration of functionalspecialization within motion-related visual and auditory cortex ofthe human brain. Curr. Biol. 6, 1015–1019

57 Puce, A. et al. (1998) Temporal cortex activation in humans viewing eyeand mouth movements. J. Neurosci. 18, 2188–2199

58 Grossman, E. et al. (2000) Brain areas involved in perception ofbiological motion. J. Cogn. Neurosci. 12, 711–720

59 Langton, S. et al. (2000) Do the eyes have it? Cues to the direction ofsocial attention. Trends Cogn. Sci. 4, 50–59

60 Adolphs, R. et al. (1996) Cortical systems for the recognition of emotionin facial expressions. J. Neurosci. 16, 7678–7687

61 Kesler-West, M.L. et al. (2001) Neural substrates of facial emotionprocessing using fMRI. Cogn. Brain Res. 11, 213–226

62 Felleman, D.J. and Van Essen, D.C. (1991) Distributed hierarchicalprocessing in the primate cerebral cortex. Cereb. Cortex 1, 1–47

63 Kaiser, M.D. et al. (2010) Neural signatures of autism. Proc. Natl. Acad.Sci. U.S.A. 107, 21223–21228

64 Lombardo, M.V. et al. (2011) Specialization of right temporo-parietaljunction for mentalizing and its relation to social impairments inautism. Neuroimage 56, 1832–1838

65 Saxe, R. (2010) The right temporo-parietal junction: a specific brainregion for thinking about thoughts. In Handbook of Theory of Mind(Leslie, A. and German, T., eds), pp. 1–35, Psychology Press

66 Decety, J. and Lamm, C. (2007) The role of the right temporoparietaljunction in social interaction: how low-level computational processescontribute to meta-cognition. Neuroscientist 13, 580

67 Cabeza, R. et al. (2012) Cognitive contributions of the ventral parietalcortex: an integrative theoretical account. Trends Cogn. Sci. 16,338–352

68 Ferstl, E.C. et al. (2008) The extended language network: a meta-analysis of neuroimaging studies on text comprehension. Hum. BrainMapp. 29, 581–593

69 Perner, J. and Aichhorn, M. (2008) Theory of mind, language and thetemporoparietal junction mystery. Trends Cogn. Sci. 12, 123–126

70 Behrmann, M. et al. (2004) Parietal cortex and attention. Curr. Opin.Neurobiol. 14, 212–217

336

71 Gallese, V. (2001) The shared manifold hypothesis. J. Conscious. Stud.8, 33–50

72 Mitchell, J.P. et al. (2005) The link between social cognition and self-referential thought in the medial prefrontal cortex. J. Cogn. Neurosci.17, 1306–1315

73 Oberman, L.M. and Ramachandran, V.S. (2007) The simulating socialmind: the role of the mirror neuron system and simulation in the socialand communicative deficits of autism spectrum disorders. Psychol.Bull. 133, 310–327

74 Scholz, J. et al. (2009) Distinct regions of right temporo-parietaljunction are selective for theory of mind and exogenous attention.PLoS ONE 4, e4869

75 Uddin, L.Q. et al. (2010) Dissociable connectivity within humanangular gyrus and intraparietal sulcus: evidence from functionaland structural connectivity. Cereb. Cortex 20, 2636–2646

76 Mars, R.B. et al. (2012) Connectivity-based subdivisions of thehuman right ‘temporoparietal junction area’: evidence for differentareas participating in different cortical networks. Cereb. Cortex 22,1894–1903

77 Goodale, M.A. and Milner, A.D. (1992) Separate visual pathways forperception and action. Trends Neurosci. 15, 20–25

78 Waytz, A. et al. (2012) Response of dorsomedial prefrontal cortexpredicts altruistic behavior. J. Neurosci. 32, 7646–7650

79 Bhatt, M.A. et al. (2010) Neural signatures of strategic types in a two-person bargaining game. Proc. Natl. Acad. Sci. U.S.A. 107, 19720–19725

80 Bartz, J.A. et al. (2011) Social effects of oxytocin in humans: context andperson matter. Trends Cogn. Sci. 15, 301–309

81 Smith, D.V. et al. (2010) Distinct value signals in anterior and posteriorventromedial prefrontal cortex. J. Neurosci. 30, 2490–2495

82 Carter, R.M. et al. (2012) A distinct role of the temporal-parietaljunction in predicting socially guided decisions. Science 337, 109–111

83 Castelli, F. et al. (2000) Movement and mind: a functional imagingstudy of perception and interpretation of complex intentionalmovement patterns. Neuroimage 325, 314–325

84 Bateman, A. and Fonagy, P. (2004) Psychotherapy for BorderlinePersonality Disorder. Oxford University Press

85 Desikan, R.S. et al. (2006) An automated labeling system forsubdividing the human cerebral cortex on MRI scans into gyralbased regions of interest. Neuroimage 31, 968–980

86 Laird, A.R. et al. (2005) BrainMap: the social evolution of a humanbrain mapping database. Neuroinformatics 3, 65–78

87 Van Essen, D.C. and Dierker, D.L. (2007) Surface-based andprobabilistic atlases of primate cerebral cortex. Neuron 56, 209–225

88 Lancaster, J.L. et al. (2000) Automated Talairach atlas labels forfunctional brain mapping. Hum. Brain Mapp. 10, 120–131

89 Yarkoni, T. et al. (2010) Cognitive neuroscience 2.0: building acumulative science of human brain function. Trends Cogn. Sci. 14,489–496

90 Rapoport, A. et al. (1997) Randomization and adaptive learning in asimplified poker game. Organ. Behav. Hum. Decis. Process. 69, 31–49

91 Fiske, S.T. et al. (2002) A model of (often mixed) stereotype content:competence and warmth respectively follow from perceived status andcompetition. J. Pers. Soc. Psychol. 82, 878–902

92 Harris, L.T. and Fiske, S.T. (2006) Dehumanizing the lowest of thelow: neuroimaging responses to extreme out-groups. Psychol. Sci. 17,847–853