Neuron

Perspective

Centrality of Social Interactionin Human Brain Function

Riitta Hari,1,* Linda Henriksson,1 Sanna Malinen,1 and Lauri Parkkonen11Department of Neuroscience and Biomedical Engineering, Aalto University, PO Box 15100, FI-00076 AALTO, Espoo, Finland*Correspondence: riitta.hari@aalto.fihttp://dx.doi.org/10.1016/j.neuron.2015.09.022

People are embedded in social interaction that shapes their brains throughout lifetime. Instead of emergingfrom lower-level cognitive functions, social interaction could be the default mode via which humans commu-nicate with their environment. Should this hypothesis be true, it would have profound implications on howwethink about brain functions and howwe dissect and simulate them.We suggest that the research on the brainbasis of social cognition and interaction should move from passive spectator science to studies includingengaged participants and simultaneous recordings from the brains of the interacting persons.

People among PeopleThe importance of interacting with other people is evident for hu-

man cognition, development, and well-being but has only

recently started to gain attention in experimental neuroscience.

The reasons for the earlier ignorance are obvious: human-to-

human interactions are extremely complex, especially because

the interaction unfolds in time with an unpredictable trajectory,

and it is challenging to analyze and interpret brain-imaging

data collected within the diverse and ever-changing social set-

tings. It is also clear that ‘‘person stimuli,’’ including dynamic

faces and bodies, do not only comprise a large set of complex

sensory features, but our interpretations of them go far beyond

the immediate information given.

The importance of social interaction is evident in our everyday

life: we teach, learn, converse, treat, and deceive. We are

shaped by other people and crave for social contacts to the

extent that isolation is used as punishment and even as torture.

According to a recent meta-analysis, social isolation and loneli-

ness are risk factors for increased mortality (Holt-Lunstad et al.,

2015).

We often feel sad or happy with others. However, it is difficult

to define to which extent such concurrent emotions result from

direct contagion. In fact, one person’s emotions can elicit oppo-

site feelings in the other, for example, when an aggressive per-

son frightens a peaceful bystander on the street. On the other

hand, mother-infant dyads, which often show clear synchrony

of emotional states, may reflect emotional regulation rather

than sharing of the same emotional state.

During social interaction, people receive both conscious and

unconscious social cues from others’ expressions, gestures,

postures, actions, and intonation. Thus, they automatically align

at many levels, starting from bodily synchrony to similar orienta-

tions of interests and attention. Such an alignment facilitates pre-

diction and understanding of the others’ aims and future actions.

A good example of automatic alignment is the smooth turn-tak-

ing during conversation: over different languages and cultures,

the gaps between the turns are typically only up to a few hundred

milliseconds, and the speech turns may even overlap (Stivers

et al., 2009). So brief intervals cannot reflect just reactions to

the end of the previous speaker’s utterances; instead, the con-

versation participants have to be aligned to predict when the pre-

vious speaker is going to finish her turn of talk.

Although verbal communication is often emphasized in the

analysis of social interaction, a major part of human-to-human

interaction is nonverbal, including exchanges of glances, frowns,

and prosody. In contrast to this kind of clearly embodied interac-

tion, disembodied communication commonly takes place in the

modern society via various technical tools; still many people

consider it necessary to augment their written messages with

emotional icons stemming from embodied interaction.

Attending and Neglecting OthersEven when we cannot identify other people, we easily categorize

them on the basis of external factors, such as profession,

clothing, or the way they speak. During social encounters, these

features—in addition to, e.g., gender, seniority, professional de-

gree, or expertise—determine the interaction order, often in a

culture-dependent manner.

To some people, we do not pay any attention; they just co-

exist without havingmuch effect on us. Naturally, such a different

allocation of attention affects the obtained information and the

possibilities to understand the intentions and behavior of others.

Moreover, sensory defects and brain disorders put people to

different footings as the sensory data on which they base their

understanding of the world are distorted (Kennedy and Adolphs,

2012).

Sociologist Erving Goffman accurately described how

behavior in public places is governed by institutionalized norms

and rules (Goffman, 1966). If people need to cut in line at an

airport because they need to catch their plane, they may ask

for privilege and are usually granted for that. But to behave in

an acceptablemanner, they are expected to apologize and thank

for the favor. Many of these unwritten rules obviously help to

keep the society in a good order, and for the same reason people

are eager to punish misbehaving individuals, even those who

they do not know and will never meet again. On the other

hand, help is offered even to strangers from whom no corre-

sponding favor can be expected. This kind of cognitively

demanding ‘‘indirect reciprocity’’ may have been central in the

evolution of human societies, functioning via reputation building

Neuron 88, October 7, 2015 ª2015 Elsevier Inc. 181

Neuron

Perspective

of the interacting people who often are strangers to each other

(Nowak and Sigmund, 2005).

During a smooth social encounter involving two or more peo-

ple, the interaction is dynamic and bilateral: partners notice each

other and are mutually regulating and co-adapting their own

behavior. Leader–follower relationships easily emerge during

the interaction but with dynamical and unpredictable changes

of the lags between the participants. For example, when musi-

cians play in small ensembles, they show different levels of

mutual adjustments in tempo to synchronize tone onsets (Badino

et al., 2014; Wing et al., 2014); such corrections are based on

auditory feedback but also on monitoring the movements of

the other players.

Behavioral synchrony is the natural key in various group

performances in music and sports: without too much effort the

participants mutually adapt to the rhythm of others’ movements

(Coey et al., 2012; Richardson et al., 2007) and even to the

speech rhythm. In the latter case, in the absence of any

instructions to synchronize, the entrainment can be of similar

strength as in typical finger-tapping tasks where the partici-

pants are specifically instructed to synchronize their behavior,

implying that speech automatically and strongly entrains

the participants (Himberg et al., 2015), in agreement with the

astonishing easiness of conversation (Garrod and Pickering,

2004).

Self versus OthersAlthough healthy adults typically have no difficulties in discrimi-

nating themselves from others, the concept of self is multifac-

eted and hierarchical. Charles H. Cooley described already in

1902 humans as looking-glass selves who know themselves as

reflections from the other people (Cooley, 1998). Similarly, a

recent treatise on the concept of self made a distinction between

‘‘me’’ and ‘‘I.’’ Here, ‘‘me’’ refers to how I am seen, or I think that I

am seen by others, involving both affective and cognitive self-

related brain processing, whereas ‘‘I’’ is a coherent self that re-

sults from multisensory integration of an embodied actor with

continuous contact with the surrounding world (Christoff et al.,

2011).

Other people may sometimes know better than we ourselves

how we would react in a new situation (Gilbert et al., 2009).

Yet, we generally think that we know much more about others

than those others can know, by means of similar observation

of behavior, about us; this phenomenon is known as the illusion

of asymmetric insight (Pronin, 2008; Pronin et al., 2001). The self-

perception theory (Bem, 1967), although much criticized, even

states that people learn about themselves by observing their

own actions similarly as they observe others.

Already a young baby expects interaction from others, getting

nervous very soon if the mother ‘‘freezes’’ her face. Such ‘‘still-

face’’ experiments suggest that the baby has from early on

communicative intentions and that she and the caretaker form

an interactive system that is mutually regulated (Sravish et al.,

2013). Accordingly, children with Moebius syndrome, congenital

bilateral facial nerve paralysis, experience difficulties in social

interaction as the caregivers cannot respond properly to the

child’s emotions because the expressions of emotions are not

visible on such a face (Cole and Spalding, 2009). Although we

182 Neuron 88, October 7, 2015 ª2015 Elsevier Inc.

do not (fortunately) have data of people who would have devel-

oped without any human contact, impoverished interaction

and social deprivation, e.g., in orphanages, has been reported

to lead to behavioral disabilities (Rutter et al., 1999).

Private Brains, Shared WorldWehumans have private brains but share theworld,more sowith

people close to us. It is thus a nontrivial question how we, with

our own brains that differ extensively in their detail, can under-

stand each other. Such mutual understanding requires a be-

tween-participant similarity in variousmechanisms of perception

and action (Hari et al., 2013). It is already known that in some

brain areas activity can ‘‘tune in’’ in subjects receiving similar

sensory-affective stimulation, e.g., during listening to a spoken

narrative (Stephens et al., 2010) or while watching professionally

directed movies that effectively engage the viewers (Hasson

et al., 2004, 2008; Jaaskelainen et al., 2008; Malinen and Hari,

2011; Nummenmaa et al., 2012). Movie episodes eliciting

emotions—especially negative ones such as fear—are particu-

larly effective in increasing ‘‘collective ticking,’’ evident as

across-spectators synchronization of wide brain areas (Num-

menmaa et al., 2012, 2014b). Moreover, people consistently

report similar bodily feelings during viewing of emotionally laden

movie clips, which emphasizes the seamless integration of body

in emotions, as well as the synchronization of both brain and

bodily functions across people embedded in an absorbing social

situation (Nummenmaa et al., 2014a).

The ‘‘social brain hypothesis,’’ put forward by Robin Dunbar,

suggests that the computational demands of living in social

groups have driven the evolution of the large human brain. In

such groups, one constantly needs to monitor others and to

anticipate their future actions. It is thus the complexity of social

relationships and the group size rather than social learning or

general intelligence that are considered decisive for develop-

ment of the human brain (Dunbar and Shultz, 2007; Dunbar,

1998). Importantly, sociality is a group property rather than just

an individual personality trait.

Neural Substrates of Reading Other MindsRecent flourishing of social neuroscience has considerably

enlarged and deepened our understanding about social stimuli,

tasks, and contexts that affect human brain function. Especially

revolutionary has been the research on mirroring systems, start-

ing with the discovery of (motor) mirror neurons in the monkey

frontal cortex (di Pellegrino et al., 1992; Rizzolatti and Craighero,

2004). These studies have brought together various disciplines

focusing on human cognition.

Basically, two different mechanisms, mirroring and mentaliz-

ing or theory of mind (Frith, 2007; Frith and Frith, 2012), have

been proposed to support ‘‘reading’’ of other minds. Instead of

being competitive or mutually exclusive, as they are often

considered, these two mechanisms likely work together in a

complementary fashion but with different temporal scales, as

is typical for many dual cognitive processes that comprise im-

plicit and explicit components (Bohl and van den Bos, 2012).

Mirroring is fast, implicit, automatic, and intuitive, and it is

considered to contribute at subconscious level to the under-

standing of other person’s intentions or goals. According to

Neuron

Perspective

our magnetoencephalographic (MEG) recordings, the time lags

from human visual to motor cortex during action observations

are of the order of 250 ms (Nishitani and Hari, 2000; Nishitani

and Hari, 2002), in line with a fast mirroring route. In contrast,

mentalizing—referring to the ability to attribute mental states to

others and to believe that they have their own beliefs about the

world—is explicit, controlled, conscious, and reflective, and

therefore slower than mirroring. Although the exact time courses

of mentalizing still remain to be specified, a recent electroen-

cephalographic (EEG) study of the neural time courses of brain

processes supporting empathy and sympathy implied that acti-

vation of thementalizing network is lagging themirroring network

by 200–300 ms (Thirioux et al., 2014). Because the mirroring and

mentalizing brain circuitries overlap only in part (Spengler et al.,

2009), they can operate in parallel, and even independently

although they usually participate in the same tasks. One

assumption is that mirroring provides data for further processing

in the mentalization network.

The main circuitry of motor mirroring, also called the action-

perception network, comprises the premotor areas in the inferior

frontal cortex and a dense frontoparietal network combining vi-

sual, somatosensory, and motor processing; the system also

has close connections to limbic areas (Rizzolatti and Craighero,

2004).Whether the primarymotor cortex also belongs to the core

of the mirror-neuron system is still under debate (Hari et al.,

2014).

Functional brain imaging studies in humans have identified

cortical regions involved in different social tasks, such as the

engagement of the fusiform face area (FFA) in face perception

(Kanwisher et al., 1997) and the temporoparietal junction (TPJ)

in mentalizing tasks (Saxe and Kanwisher, 2003). In general, a

network of brain regions involved in, e.g., seeing a face engages,

in addition to FFA, also the occipital face area (Gauthier et al.,

2000) that is associated with face-feature processing (Henriks-

son et al., 2015), regions in the anterior temporal lobe that pro-

cess the identity of the face (Kriegeskorte et al., 2007), and

also the medial prefrontal cortex (MPFC) when the face belongs

to a familiar person (Wagner et al., 2012).

The network of brain regions supporting mentalizing (theory of

mind) includes TPJ, MPFC, and superior temporal sulcus (STS).

STS is especially sensitive to biological motion (Vaina et al.,

2001) and has also been associated with joint social attention

(Nummenmaa and Calder, 2009), where eye gaze, as an impor-

tant social cue, indicates the focus of one’s attention. Recently,

human STS was shown to contain both broadly tuned regions

and more narrowly tuned subregions, each responsive to a sub-

set of social information (Deen et al., 2015). Altogether, STS re-

sponds to a wide range of social stimuli, and it hence can be

considered as a ‘‘hub’’ for social perception (Lahnakoski et al.,

2012).

TPJ and MPFC seem to have complementary roles in inferring

others’ mental states: TPJ has been associated with inferring

temporary states of others (e.g., goals and intentions), whereas

MPFC has been associated with inferring more enduring states

(permanent characteristics, self-relevance) and self-knowledge

(for a review, see Van Overwalle, 2009). MPFC comprises func-

tional subregions involved in either more cognitive or more

emotional tasks (Amodio and Frith, 2006).

All these circuitries work in close connection with other brain

areas and networks that support social cognition (see e.g., Frith,

2007). For example, amygdala and temporal pole are considered

important for social scripts, emotions, and judgments. The

medial prefrontal cortex and midline structures in general are

related to self-referential processing (Northoff and Bermpohl,

2004). Moreover, several brain areas—including sensorimotor

cortices—support shared sensorimotor representations for self

and others. Important is also, e.g., striatum as a brain region

related to reward learning, as well as the prefrontal cortex that

controls meta-cognition and higher-order thinking (Frith and

Frith, 2012).

Several other networks have a role in social cognition. For

example, the salience network (with nodes in anterior cingulate

cortex and anterior insulae of both hemispheres) is assumed to

recruit other brain regions to process sensory stimuli, and the vi-

sual dorsal attention network (parietal cortex, frontal eye fields,

visual cortices) is central in all kinds of voluntary attention

(Barrett and Satpute, 2013). In addition, a separate network

has been suggested for empathy, including the anterior insula

and dorsal-anterior/anterior-midcingulate cortex (Bernhardt

and Singer, 2012).

Despite the improving understanding of brain areas involved in

social cognition, the details of the neural representations and

especially the neural dynamics of the different parts of the net-

works are still poorly understood and require further experi-

mental and theoretical work. Moreover, the details of analysis

may determine whether one sees robust wide-spread networks

or their task-related division to subnetworks (Pamilo et al., 2012).

One promising framework for modeling various levels of brain

functions, including mirroring and mentalizing, is predictive cod-

ing, a Bayesian approach to perception, action, and various

cognitive functions (de Bruin and Strijbos, 2015; Kilner et al.,

2007; Koster-Hale and Saxe, 2013). Here the basic assumption

is that since the brain operates in uncertain conditions, it likely

has to maintain probabilistic models of the surrounding world,

updating them on the basis of sensory information.

The predictive coding works at several nested levels so that

higher-level cortical areas generate predictions of forthcoming

events for lower-level areas, and the error signals—informing

about discrepancies between the expected and received sen-

sory feedback—are evaluated to adjust the model in the

higher-level areas. Although these mechanisms were originally

considered for single person’s behavior, even including theory-

of-mind attributions (Koster-Hale and Saxe, 2013), they may be

extended to cover others’ mental states by inferring the states

of mind in which the observers themselves are. A generalized

synchrony emerges when an observer is modeling the behavior

of another person who is modeling the observer (Friston and

Frith, 2015).

Spectator Science?Despite the enormous accumulation of important brain-imaging

data, it has to be noted that most approaches to investigate the

brain basis of social cognition and interaction have so far been

based on the assumption that humans (and their brains) are

reactive and that the baseline state of the reacting brain remains

about the same whatever stimuli are presented. In other words,

Neuron 88, October 7, 2015 ª2015 Elsevier Inc. 183

stimulus stimuluspassivespectator

engagedinteractor

Figure 1. Conceptual Differences between Spectator and InteractorBrain-Imaging SettingsLeft: the subject is a passive spectator and the stimulus–response effects areunidirectional. Right: in a real-life situation (in setups of two-person neuro-science), the subject to be studied is an engaged interactor, and the (closed-loop) stimuli change according to both subjects’ reactions.

Box 1. Current Status of the Field

d Regularities of the external world and people, with their

characteristic spatiotemporal appearance, as part of our

environment, inevitably leave their traces into our brains.d Nervous systems have developed for guiding movements

and for prediction of future events, with a special sensitivity

to social stimuli.d Bodies are strongly involved in cognition and emotions,

and both brain and bodily functions are synchronized

by external events, especially when the events are

emotionally laden.d Two-person neuroscience (2PN) is an approach to study

two interacting persons at the same time, with a focus on

the dyad rather than the individuals. Recent technical ad-

vances in 2PN brain imaging include setups that enable

simultaneous scanning of two subjects in the same fMRI

scanner, simultaneous imaging using two connected

MEG scanners, as well as simultaneous EEG and NIRS

recordings.d If the social interaction unfolds in tens of milliseconds, as

does, e.g., face-to-face interaction, 2PN brain imaging

setups are needed to capture the essential features of the

related brain functions, whereas brain mechanisms of

slowly paced (seconds or slower) interaction, such as writ-

ten communication, can be reduced to sequential brain im-

aging of the participants (Figure 2).

Neuron

Perspective

modulations of brain states triggered by the interacting partner’s

behavior are largely neglected.

This kind of research comprises most of the current brain im-

aging literature and can be considered to represent ‘‘spectator

science.’’ Yet, during naturalistic social interactions, the baseline

brain state of the examined person changes the more engaging

the stimuli are (Figure 1). In other words, people are participating

in the events of their world, and they do not only serve as passive

observers as is implicitly assumed in most of the current brain-

imaging experiments.

Interactive Approaches into Social NeuroscienceWhile discussing research of social interaction as part of social

neuroscience, Stanley and Adolphs (Stanley and Adolphs,

2013) called it ‘‘.an unusually rich and interesting topic,

exactly what social psychologists would wish to study and

many neurobiologists think is too fuzzy to study.’’ Conse-

quently, they list social interaction as a topic of future social

neuroscience and suggest that real social interactions should

be studied in well-controlled animal models. While fully

acknowledging the challenges and possible caveats involved

in uncontrolled, natural social experiments, we believe that so-

cial interaction has such a profound impact on our brain func-

tions that it can no longer be ignored in experimental human

neuroscience (see Box 1).

184 Neuron 88, October 7, 2015 ª2015 Elsevier Inc.

Consequently, we have proposed ‘‘two-person neuroscience’’

(2PN) as a suitable conceptual andmethodological framework to

study the physiological basis of human social interaction (Hari

and Kujala, 2009). 2PN refers to an approach to study two inter-

acting persons at the same time, and the related conceptual

framework, with a focus on dyads rather than individuals. Of spe-

cial interest is the emerging smooth and rapid social interaction

(see Item 5 below). Our definition of 2PN allows the subjects

under study to take the first-person, second-person, or third-

person perspective according to the task demands. Thus, the

2PN approach is a more general (and may bemore methodolog-

ically oriented) than second-person neuroscience (Schilbach,

2015; Schilbach et al., 2013). However, both concepts empha-

size the importance of active participation so that the persons

under study are not just observers of social situations. Both

these approaches will hopefully promote studies on real social

interaction in which the reactions of one subject are the stimuli

for the other subject, and vice versa.

Studies of the brain basis of social interaction would benefit

from, or even require, more naturalistic stimuli and setups than

are currently used in most imaging laboratories. Ultimately, one

would like to study people who are in real social interaction.

Next, we lay out five steps on this road from usingwell-controlled

artificial stimuli to truly interactive social experiments. In Figure 2,

Items 1–5 refer to the corresponding subsections below.

(1) Well-Controlled, Artificial Stimuli

Most of our knowledge about human brain function has been

achieved using simple andwell-controlled stimuli, such as check-

erboards with different check sizes, contrasts and luminance, or

Figure 2. Examples of Brain Imaging Setups(1)–(3) are single-person setups, (4) can be studied either in a single-personsetup (studying the members of the dyad sequentially) or in a 2PN setup,(5) requires a 2PN setup. (1) Subject receives simplified but well-controlledsensory stimuli. (2) Subject receives static pictures of social stimuli, e.g., faces.(3) Subject receives dynamical stimuli, e.g., movies or, for mirroring studies,live actions. (4) Two persons are interacting, e.g., in an economic game or bywriting. Here the time lags between the two subjects’ interactive actions orresponses are long. (5) Two persons are engaged in true, real-time socialinteraction, such as conversation. The interaction is dynamic and the re-sponses and alignment can overlap in time.

Neuron

Perspective

simple sounds with different frequency compositions, rise times

and presentation sequences, or somatosensory stimuli applied

as electric pulses to peripheral nerves or as touch on the skin. In

traditional experimental neuroscience, such accurately defined

stimuli and well-controlled experimental setups have provided

invaluable information about the transfer functions of all sensory

systems from peripheral sensory receptors to sensory cortices,

as well as—in tasks requiring motor actions—from the brain to

the periphery. The tradition has been to build on the responses

to elementary stimuli to address the functional principles underly-

ing the processing of more complex stimuli, which has been, and

still continues to be, an efficient approach in sensory neurosci-

ence (Rust and Movshon, 2005). Yet, when it comes to under-

standing how the dynamic social environment is processed in

the brain, this approach—even when complemented with effi-

cient computational approaches—may miss some essential in-

gredients of the complex interactive social processing.

(2) Snapshots of Complex, Naturalistic Stimuli

Presentation of more complex stimuli allows studying how hu-

mans react to, e.g., faces of other humans. Even though the brain

is extremely sensitive to faces, presenting a still image of a face

already introduces a challenge for modeling the brain activity

starting from, e.g., single-unit responses to low-level visual ele-

ments comprising spatial frequencies, orientations, contrasts,

and luminance. Comparing responses to different complex stim-

uli has revealed functionally specialized brain regions (Kanw-

isher, 2010), including the fusiform face area, which responds

strongly to faces as explained above. However, only a very

limited number of stimuli of particular perceptual significance

can be expected to show such functional specialization, and

most tasks engage distributed brain networks without any of

the nodes necessarily showing functional specialization for the

task or stimulus at hand.

(3) Dynamic Stimuli

The next step is to incorporate motion to the stimuli to make

them more natural. Motion, vividness, and action—either a live

person or a video—in the experimental setup can help unravel

how humans react to actions and intentions of other people.

Moving stimuli are highly engaging, both perceptually and at

neural level, as can be inferred from the robust brain activations

that they elicit. For example, several areas of the face-process-

ing circuitry are activated more effectively by video clips of natu-

rally moving faces than by still pictures of faces (Schultz and Pilz,

2009). Moreover, the human STS is known to be very sensitive to

biological motion (Blake and Shiffrar, 2007). Motor actions, either

live or on video, continue to serve as essential dynamic stimuli in

action observation studies designed to explore mirroring mech-

anisms of the human brain.

Movies, in general, provide an effective means to study both

low-level sensory processes and social cognition. Cinema is

an extremely rich visual (and often multimodal) stimulus but still

well controlled as it can be repeated in an identical form to the

same subject or to a group of subjects. It thereby serves as a

staging post on the route toward real naturalistic setups. A

well-directed movie stimulates and engages the human mind,

and it also includes the unique ability to manipulate space and

time, as well as to create an illusion of intimacy by showing

close-ups of faces. Although people just view the movie, without

interaction, they easily identify themselves with the protagonists

and capture emotions from film episodes in a strikingly similar

manner (Nummenmaa et al., 2012).

(4) Two Persons in Slowly Paced Social Interaction

When progressing beyond the single-person setups in studies

even in naturalistic environments, the next step is to search for

brain correlates of social interaction. Methodologically, and

probably also conceptually, the easiest way may be to start

from temporally clearly separate brain signals from persons

who are engaged in relatively slowly paced social interaction,

such as sending and receiving text messages. This type of

communication can be characterized as reactive rather than

interactive, meaning that the receiver reacts to the most recent

output of the partner instead of forming a dynamically adapting

dyad with her. Although we are dealing here with a real social

interaction, either embodied or disembodied, we do not neces-

sarily need simultaneous recordings from the two subjects. We

can, for example, record the brain signals separately from the

sender and receiver of writtenmessages, keeping the interacting

person outside the scanner, and then swapping the roles. Other

suitable experimental paradigms are, e.g., economical decision

games (Tomlin et al., 2006).

(5) Two Persons in Dynamic, Embodied Interaction

True social interaction occurs at a fast pace and the responses

can overlap in time. Examples include the very quick turn-taking

during conversation and the unconscious mutual adaptation

during a joint motor task, such as carrying a big heavy object.

Simultaneous brain imaging from both interacting subjects is

required to capture the full picture of the real-time (embodied)

interaction; such dynamic interaction cannot be reduced from

2PN settings to sequential one-person measurements.

Neuron 88, October 7, 2015 ª2015 Elsevier Inc. 185

Box 2. Future Directions

d Development of more real-life-like 2PN setups that employ

novel measurement technologies and analysis methods to

track at the same time both brain and bodily functions,

including movement synchrony.d Development of experiments and analysis methods to test

whether, and how, the interactive versus reactive modes of

brain function differ.d Characterization of the different timings of the mentalizing

and mirroring systems and their relationships to the dy-

namics of other brain networks subserving social cognition

and interaction.d Further topics to be studied in 2PN settings include the

following:

o Brain basis of the synchrony and turn-taking of motor

actions, vocal expressions, and gestures during nat-

ural interaction

o The effect of presence of another person on resting-

state and task-related activity, and the mutual adap-

tation of these effects in both persons

o Brain correlates of feeling of togetherness during

smooth fast-spaced interaction

o Emotional contagion (between the participants, and

from an external source but experienced together

with the other participant in a situation similar to

viewing a movie or listening to a political speech

together)

o Social error monitoring (entailing predictions and

their violations to inform of the neural systems sup-

porting social interaction)

Neuron

Perspective

The temporal structure of the interactive behavior defines

when a 2PN setup is the most efficient and informative, or the

only possibility. For example, investigating the neural basis of

natural face-to-face interaction likely calls for 2PN approaches,

whereas monitoring brain activity during an exchange of written

messages can well be studied sequentially (see Item 4 above).

Some estimates of the relevant timescales calling for 2PN setups

arise from the speed of articulation movements and other

dynamical facial changes, such as eye blinks (Mandel et al.,

2015), to which we react either consciously or unconsciously

during face-to-face communication. For example, since a

phoneme typically lasts about 100 ms, the sampling of the brain

signals should take place at least at 50-ms intervals, corre-

sponding to about 20 Hz. Thus, time-resolved brain-imaging

methods, such as MEG or EEG, are needed to track the under-

lying brain activity in such experiments.

Finally, it is important to note that these rapid-paced interac-

tions needing 2PN-imaging setups can never be repeated (in

contrast to many slow interactions in Item 4), because they are

specific to that particular time and interaction.

Simultaneous Neuroimaging of Two or More SubjectsNatural social interaction comprises unique spatiotemporal

events whose exact content and timing are usually unpredict-

able; the same interaction sequence cannot be repeated to sepa-

186 Neuron 88, October 7, 2015 ª2015 Elsevier Inc.

rately record the brain activity of the other interactor. Hence, if we

only have data from one interactor, it would be necessary to

quantify the interaction to identify the brain processes supporting

it. This quantification is challenging since itmay necessitate high-

level interpretations of the behavior; for example, the analysis

may involve a continuous assessment of the mentalizing per-

formed by the other participant. In contrast, when data are avail-

able from both brains, we can seek for dependencies between

the two sets of brain signals, without explicit reference to the

external events, to reveal brain processes that support the inter-

action. Thereafter, we can aim at building models that capture

and predict the coupling between the brain signals and behavior,

both within and between the interacting subjects (see Box 2 for

suggestions of topics to be resolved using 2PN setups).

However, care must be exercised in the design, analysis, and

interpretation of such experiments: trivial correlations between

the participants’ brains can emerge simply due to low-level sen-

sory input reaching both subjects. Therefore, one should include

appropriate control conditions (e.g., passive observation) and

employ suitable analytical approaches that help tease apart

the contributions due to mere shared sensory input.

Simultaneous brain imaging of multiple subjects is commonly

referred to as hyperscanning or dual scanning. The technical

feasibility of hyperscanning has already been demonstrated

using a variety of different brain imaging methods (see e.g.,

Babiloni and Astolfi, 2014). In the following, we concisely review

some hyperscanning studies.

fMRI-to-fMRI: From Connected Scanners to Dual-Coil

Experiments

Montague et al. (Montague et al., 2002) were the first to record

brain signals from two persons at the same time: the subjects

were located in different 1.5 T fMRI scanners situated over

1,000 miles apart and connected via the Internet. During the

scanning, the subjects played a simple interactive game where

the receiver subject needed to guess whether the color (two

options) mediated by the sender was the same as what the

sender had seen on her screen. The receiver won the game if

she guessed right, otherwise the sender won. Data were

analyzed using both temporal regressors and independent

component analysis (ICA) by concatenating the data from two

brains into a single ‘‘hyperbrain.’’ Results showed between-sub-

ject similarity in the supplementary motor area, and the signals

from both brains had the same task-related frequency of about

0.04 Hz, with slightly different phases. This study demonstrated

the feasibility of dual-fMRI scanning and introduced the idea that

social interaction could be best understood by simultaneously

scanning the brains of both interacting persons.

Subsequent two-person fMRI experiments have revealed

changes in brain activity related to, e.g., reciprocity (King-Casas

et al., 2005), agency (Tomlin et al., 2006), and social comparison

on reward processing (Fliessbach et al., 2007). Recently, two-

person fMRI was applied to study information flow between

two brains during a joint attention paradigm applied with an im-

mersive audiovisual interface between the two scanners (Bilek

et al., 2015). Noteworthy is that simultaneous measurements

from both brains are not always necessary in communicative

tasks (Figure 2, Item 4), but the brain activity of the two players

can be recorded sequentially.

Neuron

Perspective

In most published two-person fMRI studies, twoMRI scanners

have been connected via the Internet. The latest setups enable

scanning two people simultaneously even in the same MRI de-

vice. Currently, we are aware of three technical realizations for

two-person recordings in one fMRI scanner with dual-head bird-

cage coils. A setup with subjects laying side-by-side in the scan-

ner became operational first at Princeton University, USA (Lee,

2015; Lee et al., 2010, 2012). The first two-person fMRI setup

in our laboratory at Aalto University, Finland was a pair of

helmet-like surface coils with which the subjects were imaged

in a face-to-face position while they were lying on their stomach,

with upper bodies tilted slightly upward so that the first subject

entered the scanner feet first and the other in the normal manner

head first (V. Renvall and S. Malinen, 2012, OHBM, conference);

for a video of the setup, see https://vimeo.com/98542820.

Because of neck strain in the applied position, our laboratory’s

latest setup also allows the two people to be lying on their sides

in face-to-face position, thus being able to, e.g., touch each

other’s faces (V. Renvall, J. Kauramaki, S. Malinen, R. Hari,

and L. Nummenmaa, 2015, Soc. Neurosci., conference). Prelim-

inary studies with these pioneering two-person fMRI setups have

confirmed that it is technically possible to record fMRI signals

simultaneously from two subjects within the same scanner.

Thebenefitsofdual-coil setups for two-person fMRIareobvious

for real 2PN recordings because the dyad can interactwithout any

delay, the presence of the other is very strong, the signal quality in

the same scanner is the same for both persons, and there are no

delays in data acquisition from the two brains. However, in the

side-by-side implementations (Leeet al., 2010; V.Renvall, J.Kaur-

amaki, S. Malinen, R. Hari, and L. Nummenmaa, 2015, Soc. Neu-

rosci., conference), the distance between the bodies of the two

participants may be too intimate for natural interaction, at least

for strangers. Without vision correction, the focusing distance to

the other person’s face may also be too short, and in our current

setup (V. Renvall, J. Kauramaki, S. Malinen, R. Hari, and L. Num-

menmaa, 2015, Soc. Neurosci., conference) the coils are so

snug that one cannot wear headphones. It is thus obvious that

further technical development is eagerly awaited for in this area.

fNIRS-to-fNIRS

Functional near-infrared spectroscopy (fNIRS) has also been

applied to study brain activity simultaneously from multiple sub-

jects. Themain benefit of fNIRS comparedwith fMRI is the porta-

bility of the equipment, enabling more naturalistic experimental

setups. In the first two-person fNIRS experiment, the subjects

performed a cooperative button-press task while sitting face-

to-face across a table (Funane et al., 2011). Subsequently, fNIRS

hyperscanning has been applied, for example, to study the

uniqueness of face-to-face communication (Jiang et al., 2012)

and the emergence of a leader during communication (Jiang

et al., 2015). The recent introduction of a wearable multi-channel

fNIRS system (Piper et al., 2014) opens a possibility of simulta-

neous brain imaging from freely moving interacting subjects.

However, fNIRS is limited by the poor penetration of light through

the scalp, skull, and brain tissue, so that one can only assess the

superficial brain structures. Compared with other brain imaging

techniques, fNIRS has a relatively low spatial resolution and its

temporal resolution is limited by the inherent sluggishness of

the underlying hemodynamic phenomena, similarly as in fMRI.

EEG-to-EEG

The first dual-EEG data were recorded already 50 years ago in

a bizarre study designed to explore extrasensory perception

(Duane and Behrendt, 1965), but more serious EEG hyperscan-

ning has gained popularity during the last decade (Astolfi et al.,

2010; Babiloni and Astolfi, 2014; Babiloni et al., 2007a, 2007b).

The benefits of EEG include high temporal resolution, relatively

low cost, and high portability, enabling naturalistic experimental

setupsandsimultaneousmeasurements frommore than twosub-

jects. Consequently, EEG hyperscanning has been applied to

study the brain basis of people’s tendency to mutually adapt to

each other’s rhythm during motor tasks (see, e.g., Dumas et al.,

2010; Konvalinka et al., 2014) and speech (Kawasaki et al., 2013).

Music and especially playing in musical ensembles provides

an interesting setting to study social interaction as the synchrony

between players can be monitored behaviorally at many levels

(D’Ausilio et al., 2015). The coordination of actions during

music performance has also been studied using simultaneous

EEG recordings, suggesting, for example, brain signatures for

emotional empathy and musical roles during the performance

(see, for example, Babiloni et al., 2012; Lindenberger et al.,

2009; Muller et al., 2013).

During EEG-to-EEG recordings, the participants can move

quite freely (although movement artifacts easily arise). The ca-

veats of EEG include poor discrimination of different signal sour-

ces, even those of the most prominent brain rhythms. Many EEG

hyperscanning studies have reported brain-to-brain synchrony

at single frequencies, such as around the 20-Hz beta oscilla-

tions, raising questions about the origin of the modulation taken

that brain rhythms have clear individual signatures in frequencies

and modulations. Moreover, most brain rhythms have several

sources, the dominance of which changes with a short (a few

hundredmillisecond) timescale (Salmelin and Hari, 1994). A safer

approach would be to look at the modulations of the envelopes

of the rhythms but here the problem is their slowness so that

although the electrophysiological signals as such have milli-

second temporal resolution, the rhythms wax and wane with a

time constant slower than a few hundred milliseconds (Ramku-

mar et al., 2010).

MEG-to-MEG

We realized the first simultaneous MEG-to-MEG recordings be-

tween two MEG labs 5 km apart (Baess et al., 2012). In these

initial recordings, the subjects interacted through a short-latency

audio connection. More recently, we augmented the two-MEG

setup to include also a video connection between the partici-

pants and implemented the audiovisual link via the Internet,

allowing MEG hyperscanning of participants at arbitrarily large

geographical distances. The setup delivers video with an end-

to-end delay of about 130 ms, which does not hamper smooth

and natural interaction (Zhdanov et al., 2015). In Japan, a

setup with two MEG scanners in the same room was recently

established and its feasibility was demonstrated by studying

simultaneously a mother and her child who were lying on their

backs in scanners located side-by-side, seeing each other’s

facial expressions in real time via a mirror system (Hirata et al.,

2014).

Although connecting two MEG devices is more complex and

expensive than using a dual-EEG system, the unique benefits

Neuron 88, October 7, 2015 ª2015 Elsevier Inc. 187

Neuron

Perspective

ofMEGmake such attemptsworthwhile.WhileMEGhas a similar

high temporal resolution as EEG, it offers significantly better

spatial resolution since the neuromagnetic fields, unlike the

EEG signals, are not smeared by the combination of the poorly

conducting skull and the well-conducting cerebrospinal fluid

and scalp (Hamalainen et al., 1993; Hari and Parkkonen, 2015;

Hari and Salmelin, 2012). The better spatial resolution of MEG

not only enables more precise localization of the neural activity

but, perhaps more importantly here, better separability of simul-

taneously active brain regions, especially those generating brain

rhythms that are often recorded in studies of social interaction (for

a recent reviewof theMEGmethod, seeHari andSalmelin, 2012).

Compared with EEG, however, MEG ismost sensitive to activ-

ity in the fissural cortex and typically does not pick up signals

from deep brain regions. In addition, asynchronous neural activ-

ity is poorly represented in both MEG and EEG, which makes

MEG/EEG less sensitive than fMRI to activity, which is not accu-

rately time locked to external events.

The Challenge of Interpreting 2PN DataNow that simultaneous brain imaging of multiple subjects has

become technically feasible, we should turn our attention to

the analysis of the recordings. In general, a major concern in

the interpretation of hyperscanning data is the inability to disen-

tangle the correlations evoked by social interaction from other

possible common sources between the subjects. A synchronous

change in the data from two subjects does not necessarily imply

coupling related to the social interaction but can reflect, for

example, a difference in the experimental conditions affecting

both subjects (Burgess, 2013). In addition, synchronous

changes in physiological (cardiac and respiratory) signals during

behavioral synchrony (Muller and Lindenberger, 2011) could also

manifest as spurious correlations between the two subjects’

neuroimaging data. As discussed in the preceding sections,

identifying the true brain signatures of social interaction from

two-person data remains a grand challenge for future 2PN

research.

Social interaction involves a plethora of brain processes that

work at multiple temporal scales, related to monitoring of past

behavior, reacting to current sensory input, and predicting the

actions of the partner. Intersubject synchrony and alignment

can take place at all these levels. The analysis methods of the

corresponding brain signals should thus be able to tackle all

this complexity.

Repeated joint actions, such as synchronized movements,

likely appear as temporal correlations in brain signals. In

contrast, e.g., conversation should also involve reciprocal or

‘‘antagonistic’’ brain activations due to the dynamic asymmetry

between the speaker and the listener. Moreover, the mere pres-

ence of another person likely influences brain responses to stim-

uli or tasks that are not even related to the other. Disentangling

these different sources of modulations of brain activity forms a

challenge for both data analysis and experimental design.

Since brain measurements have traditionally been confined to

single subjects, appropriate and established analysis methods

for hyperscanning data are not readily available. Currently known

analytic approaches to 2PN data could be broadly categorized

as follows.

188 Neuron 88, October 7, 2015 ª2015 Elsevier Inc.

(1) Hyperconnectivity

Functional connectivity refers to detecting temporal correlations

in signals from different brain regions and considering those to

reflect couplings of, or common drive to, these regions. In 2PN

studies, this approach can be extended to the signals measured

simultaneously from the brains of both participants. The hyper-

connectivity (Astolfi et al., 2011) analysis aims to detect how

the brain of each participant in a dyad influences the brain of

the other and how joint behavior may mediate interbrain func-

tional coupling.

Functional hyperconnectivity between two brains can be as-

sessed between (1) homologous areas in brain 1 and brain 2,

(2) between one area in brain 1 and all areas in brain 2, and (3)

between networks of areas in brain 1 and brain 2. The methodo-

logical and computational challenges as well as difficulties in

visualizing and interpreting the results naturally increase when

moving from (1) to (3).

Hyperconnectivity analysis has been performed on EEG data,

e.g., by assessing phase synchronization across the signals from

subject pairs who were playing guitar together (Lindenberger

et al., 2009), by quantifying Granger causality when subjects

were playing a card game (Astolfi et al., 2010), and by computing

phase-locking values while subjects were imitating each other’s

hand movements (Dumas et al., 2012).

Despite the success in applying connectivity metrics originally

developed for within-brain analysis, real social interaction is pre-

sumably associated with more complex dependencies between

the brain signals. For example, each subject may be constantly

switching between active, reactive, and anticipatory modes in

the course of the interaction, which affects the relative timing

of the recorded signals. During interaction and joint tasks,

leader–follower relationships may spontaneously emerge, but

with varying time lags between the partners, modulating the

timing of the task-related brain signals. Similarly, during a con-

versation, brains of the speaker and listener, in addition to

showing activations of overlapping brain areas (Stephens

et al., 2010), likely exhibit antagonistic behavior. Thus, an ideal

method for hyperconnectivity analysis should be able to esti-

mate not only brain-to-brain couplings but also allow mode-

dependent time lags and directions of these couplings. Along

these lines, we have proposed non-linear canonical correlation

analysis, with a dynamic delay between the signal sets, for the

analysis of dual-MEG signals (Campi et al., 2013). Very recently

Bilek and colleagues (Bilek et al., 2015) assessed hyperconnec-

tivity in a dual-fMRI experiment on joint attention by computing

cross-correlations of those independent components (deter-

mined jointly from all participants’ data) that showed more activ-

ity during interaction versus non-interaction; they also estimated

the delay between the two signal sets.

(2) Correlations with External Measures

Behavioral signals associated, e.g., with movements or speech

could help disentangle 2PN data. For example, computing corti-

covocal coherence between the fundamental frequency of the

acoustic signal from the speaker and the brain activity of the

listener (Bourguignon et al., 2013) could indicate which part of

the brain-to-brain coupling is due to the mere shared sensory

stimulus. In general, bodies are closely involved in cognition

and emotions (Nummenmaa et al., 2014a) and peripheral

Neuron

Perspective

measures, such as galvanic skin response, heart rate, and

breathing rhythm, could index the dynamics and strength of

mutual coupling and thus alleviate the problems in isolated anal-

ysis of 2PN imaging data.

(3) Joint Statistics and Multivariate Pattern Analysis

Statistical methods addressing signals from both brains simul-

taneously, e.g., independent or principal component analysis

(ICA/PCA) performed at once on functional data from both

brains, could be used to search for functional networks from

the two instead of one brain. Such analysis is expected to

pool, e.g., the network involved in speech comprehension in

one brain with the speech production network in the other brain.

However, in real conversation, the participants do not merely

respond to the partner’s message they just heard but they are

aligned toward the partner and largely anticipate her/his next

lines. The brain processes supporting such predictions may

not follow the time course of turn taking and thus may not be

considered as components reflecting speaking and listening

even though these processes can be crucial for the social

aspect of the interaction. Therefore, approaches that aim at

forming predictions about the future (instead of present) signals

in brain 2 using data from brain 1 may turn successful in 2PN

analysis.

If the 2PN data are assumed to contain cross-brain correla-

tions during discrete temporal windows only, group factor anal-

ysis (probabilistic extension of canonical correlation analysis)

has been shown to provide a meaningful separation of the 2PN

data in indicating which of the persons is speaking at a given

time (Remes et al., 2013).

Multivariate pattern analysis (MVPA) methods, also known as

‘‘decoding’’ of brain signals, can also be applied to 2PN data.

Konvalinka and colleagues (Konvalinka et al., 2014) employed

such approaches to dual-EEG recordings to predict whether

the subjects were following mutual synchrony in finger tapping

or following a computer metronome. Selecting appropriate

features (evoked responses, amplitude envelopes of on-going

oscillations, etc.) from the 2PN analysis is crucial for success

in these approaches. As with hyperconnectivity measures of

brain-to-brain coupling (Burgess, 2013), a caveat with MVPA is

that successful ‘‘decoding’’ of brain responses may rely on brain

signals that are synchronous but not because of the interaction.

Hence, future studies should aim at building and estimating

computational encoding models (Naselaris et al., 2011) that

describe the coupling between the subjects to really ‘‘decode’’

the underlying brain signals.

Social Interaction: The Brain’s Default Mode?Social interaction is among the most complex functions humans

(and their brains) perform. Yet, the interaction typically appears

surprisingly easy. For example, during conversation, turns of

speaking are usually taken effortlessly, smoothly, and in a

temporally accurate manner without conscious effort. According

to Garrod and Pickering (2004), humans are ‘‘designed’’ for dia-

logs rather than monologues. These authors consider conversa-

tion so easy because unconscious interactive processing aligns

the linguistic representations of the interlocutors so that the

cognitive load is alleviated as it can be divided on an ‘‘implicit

common ground.’’

One behavioral feature that might facilitate social perception is

the innate tendency to see agency andmental states in inanimate

objects, even inmoving geometrical shapes (Heider and Simmel,

1944). Assuming the world to be animated and full of agents is

especially common in childhood; in adulthood, active inhibition

of the attribution of agency seems to be required for rational infer-

ences about events in the world (Lindeman et al., 2013).

Would it then be possible that humans have an innate ten-

dency to interact and synchronize with others? The origin of

such a tendency would be easy to understand in all mammals

who are entrained with the mother’s motor and vocalization

rhythm already in the womb. Moreover, because of their imma-

turity, the human newborns have to totally rely on their care-

givers. Being in synchrony seems to please the caregiver and

result in better care. Preference for synchronous movements

with others continues throughout the adult life.

The primacy of interaction is also supported by findings that

children learn best during interaction, and much less by

observing the behaviors of others (Moll and Meltzoff, 2011).

Moreover, it is inherent for small children to assume shared

perceptual experience even when there is none (Moll and Meltz-

off, 2011). These experimental findings agree with view that, ‘‘In-

fant human beings imitate other humans, not just to act like them,

but to enter into a communicative and cooperative relationship

with them by some transfer of the feeling of body action’’ (Tre-

varthen, 2011).

In an innovative behavioral study, Noy et al. (Noy et al., 2011)

asked dyads of subjects to play a one-dimensional mirror game

in which the players, placed on two sides of a narrow table,

moved their own handle along a track either in a leader–follower

fashion or without any designated leader being just instructed to

‘‘imitate each other, create interesting and synchronized move-

ments and enjoy playing together.’’ Analysis of the velocity pat-

terns showed closer similarity between the two players in the

latter condition, indicating that the players had entered co-lead-

ership states during which the synchrony of behavior was much

better than in the leader–follower conditions. One sees here a

close resemblance to various joint tasks, with continuous

smooth adaptation to other persons’ actions without any of the

participants working consciously as leaders or followers. In gen-

eral, interpersonal synchrony is considered to promote social

connectedness, and in improvisation theater such connected-

ness is specifically trained.

Two more recent studies focused on the feelings of ‘‘together-

ness’’ during improvisedmotion, a phenomenon familiar tomany

dancers, musicians, and actors acting in synchrony (Hart et al.,

2014; Noy et al., 2015). Hart et al. (Hart et al., 2014) noticed

that although individuals have their characteristic signatures of

velocity patterns in the mirror game (that they use while acting

as leaders), during the togetherness epochs these movement

patterns were different; they were not of either of the partici-

pants, nor were they just average or intermediate patterns but

distinctly different from the individual patterns. It is thus likely

that during joint actions the participants construct their complex

movements from simpler and smoother elements that are easier

for the other to follow.

Furthermore, Noy et al. (Noy et al., 2015), using the same

game, noticed that during the moments of togetherness, or

Neuron 88, October 7, 2015 ª2015 Elsevier Inc. 189

Neuron

Perspective

‘‘being in the zone’’—that formed about 15% of the whole play-

ing period and were assessed by subjective ratings and by kine-

matics of the players—were characterized by increased heart

rates regardless of motion intensity. Such a pattern was seen

for subjectively defined periods of togetherness but clearly less

so for the kinematically defined epochs of togetherness. The au-

thors suggested a connection of the participants’ heart rates to

enhanced engagement and enjoyment during the epochs of

togetherness.

In good agreement with these experimental findings is the

‘‘Interactive Brain Hypothesis’’ (IBH) of Di Paolo and De

Jaegher (Di Paolo and De Jaegher, 2012). IBH assumes that

interactive experience and interactive skills play an enabling

role for the development and function of social brain functions

in an analogous manner as, e.g., electricity has an enabling

role in boiling water in a kettle. If IBH would turn out to be

true, we would need to revise many current ideas about the

brain basis of human social interaction. This is because social

interaction is at present considered to emerge from simple

perceptual and cognitive mechanisms that should be explored

using well-controlled stimuli before moving towards more

naturalistic experiments, in a way resembling the evolution of

brain-imaging setups in Figure 2. Stated boldly, the assump-

tions of IBH would form the big picture of interactive behavior,

decorated by all the details of perception and action, and not

vice versa.

Social interaction as the default mode of human brain would

also have profound implications for the large-scale research ini-

tiatives aiming to simulate functions of the human brain. If the

goal is to achieve human-like behavior, it is not enough to build

on the bottom-up stimulus-driven effects, but the centrality,

eventually primacy, of social interaction should be incorporated

to the models.

The principles of human brain function are to some extent

simulated and tested in social robotics that aims at building ro-

bots with human-like social skills (for a review, see Scassellati

et al., 2012), such as abilities to recognize faces, follow eye

gaze, and react to gestures and other social signals. However,

social robots are, at least so far, built to mimic and recognize hu-

man behavior, not to simulate human brain functions.

For smooth human-robot interaction, the robot does not

necessarily need to appear human-like (humanoid), and too real-

istic external features can even make the humans to feel the

robot ‘‘uncanny’’ (Mori, 1982). Desirable properties of future so-

cial robots could include the adaptive dynamics of smooth social

interaction. Even a very simple robot can be highly engaging and

evoke, especially in children, an urge to attribute mental states to

that nonliving apparatus. Preliminary studies suggest that social

robots could be used in education to provide peer support for

children, e.g., in language learning (Kanda et al., 2004), or in

behavioral therapy to encourage autistic children to develop

and employ social skills (Scassellati et al., 2012).

In our mind, the future of social neuroscience should aim at

both conceptual andmethodological advances in understanding

the brain basis of social interaction, targeting questions such as

how to transfer the obtained information about the centrality of

social interaction into simulating of human brain function, how

to build socially smart interactive robots, and how to better un-

190 Neuron 88, October 7, 2015 ª2015 Elsevier Inc.

derstand and eventually treat social dysfunction in various

neurological and psychiatric disorders.

ACKNOWLEDGMENTS

This work was financially supported by the European Union Seventh Frame-work Programme (FP7/2007–2013) under grant agreement no. 604102(Human Brain Project), and by the Academy of Finland (grant 278957 toL.H.). The experimental work was supported by the European Research Coun-cil (Advanced Grant 232946 Brain2Brain to R.H., 2009–2014), the Academy ofFinland (grants 131483 and 263800 to R.H.), and the aivoAALTO project of theAalto University. We thank Lotta Hirvenkari for help with Figure 2. Conflict ofinterests: L.P. is a part-time employee of Elekta Oy.

REFERENCES

Amodio, D.M., and Frith, C.D. (2006). Meeting of minds: the medial frontal cor-tex and social cognition. Nat. Rev. Neurosci. 7, 268–277.

Astolfi, L., Toppi, J., De Vico Fallani, F., Vecchiato, G., Salinari, S., Mattia, D.,Cincotti, F., and Babiloni, F. (2010). Neuroelectrical hyperscanning measuressimultaneous brain activity in humans. Brain Topogr. 23, 243–256.

Astolfi, L., Toppi, J., De Vico Fallani, F., Vecchiato, G., Cincotti, F., Wilke, C.T.,Yuan, H., Mattia, D., Salinari, S., He, B., and Babiloni, F. (2011). Imaging thesocial brain by simultaneous hyperscanning during subject interaction. IEEEIntell. Syst. 26, 38–45.

Babiloni, F., and Astolfi, L. (2014). Social neuroscience and hyperscanningtechniques: past, present and future. Neurosci. Biobehav. Rev. 44, 76–93.

Babiloni, F., Astolfi, L., Cincotti, F., Mattia, D., Tocci, A., Tarantino, A., Mar-ciani, M., Salinari, S., Gao, S., Colosimo, A., and De Vico Fallani, F. (2007a).Cortical activity and connectivity of human brain during the prisoner’sdilemma: an EEG hyperscanning study. Conf. Proc. IEEE Eng. Med. Biol.Soc. 2007, 4953–4956.

Babiloni, F., Cincotti, F., Mattia, D., De Vico Fallani, F., Tocci, A., Bianchi, L.,Salinari, S., Marciani, M., Colosimo, A., and Astolfi, L. (2007b). High resolutionEEG hyperscanning during a card game. Conf. Proc. IEEE Eng.Med. Biol. Soc.2007, 4957–4960.

Babiloni, C., Buffo, P., Vecchio, F., Marzano, N., Del Percio, C., Spada, D.,Rossi, S., Bruni, I., Rossini, P.M., and Perani, D. (2012). Brains ‘‘in concert’’:frontal oscillatory alpha rhythms and empathy in professional musicians. Neu-roimage 60, 105–116.

Badino, L., D’Ausilio, A., Glowinski, D., Camurri, A., and Fadiga, L. (2014).Sensorimotor communication in professional quartets. Neuropsychologia55, 98–104.

Baess, P., Zhdanov, A., Mandel, A., Parkkonen, L., Hirvenkari, L., Makela, J.P.,Jousmaki, V., and Hari, R. (2012). MEG dual scanning: a procedure tostudy real-time auditory interaction between two persons. Front. Hum.Neurosci. 6, 83.

Barrett, L.F., and Satpute, A.B. (2013). Large-scale brain networks in affectiveand social neuroscience: towards an integrative functional architecture of thebrain. Curr. Opin. Neurobiol. 23, 361–372.

Bem, D.J. (1967). Self-perception: An alternative interpretation of cognitivedissonance phenomena. Psychol. Rev. 74, 183–200.

Bernhardt, B.C., and Singer, T. (2012). The neural basis of empathy. Annu. Rev.Neurosci. 35, 1–23.

Bilek, E., Ruf, M., Schafer, A., Akdeniz, C., Calhoun, V.D., Schmahl, C., Dema-nuele, C., Tost, H., Kirsch, P., and Meyer-Lindenberg, A. (2015). Informationflow between interacting human brains: Identification, validation, and relation-ship to social expertise. Proc. Natl. Acad. Sci. USA 112, 5207–5212.

Blake, R., and Shiffrar, M. (2007). Perception of human motion. Annu. Rev.Psychol. 58, 47–73.

Bohl, V., and van den Bos, W. (2012). Toward an integrative account of socialcognition: marrying theory of mind and interactionism to study the interplay ofType 1 and Type 2 processes. Front. Hum. Neurosci. 6, 274.

Neuron

Perspective

Bourguignon, M., De Tiege, X., Op de Beeck, M., Ligot, N., Paquier, P., VanBogaert, P., Goldman, S., Hari, R., and Jousmaki, V. (2013). The pace of pro-sodic phrasing couples the listener’s cortex to the reader’s voice. Hum. BrainMapp. 34, 314–326.

Burgess, A.P. (2013). On the interpretation of synchronization in EEG hyper-scanning studies: a cautionary note. Front. Hum. Neurosci. 7, 881.

Campi, C., Parkkonen, L., Hari, R., and Hyvarinen, A. (2013). Non-linear canon-ical correlation for joint analysis of MEG signals from two subjects. Front.Neurosci. 7, 107.

Christoff, K., Cosmelli, D., Legrand, D., and Thompson, E. (2011). Specifyingthe self for cognitive neuroscience. Trends Cogn. Sci. 15, 104–112.

Coey, C.A., Varlet, M., and Richardson, M.J. (2012). Coordination dynamics ina socially situated nervous system. Front. Hum. Neurosci. 6, 164.

Cole, J., and Spalding, H. (2009). The Invisible Smile: Living without FacialExpression (Oxford University Press).

Cooley, C.H. (1998). On Self and Social Organization (The University of Chi-cago Press).

D’Ausilio, A., Novembre, G., Fadiga, L., and Keller, P.E. (2015).What canmusictell us about social interaction? Trends Cogn. Sci. 19, 111–114.

de Bruin, L., and Strijbos, D. (2015). Direct social perception, mindreading andBayesian predictive coding. Conscious. Cogn. 36, 565–570.

Deen, B., Koldewyn, K., Kanwisher, N., and Saxe, R. (2015). Functional orga-nization of social perception and cognition in the superior temporal sulcus.Cereb. Cortex. Published online June 5, 2015. http://dx.doi.org/10.1093/cercor/bhv1111.

Di Paolo, E., and De Jaegher, H. (2012). The interactive brain hypothesis. Front.Hum. Neurosci. 6, 163.

di Pellegrino, G., Fadiga, L., Fogassi, L., Gallese, V., and Rizzolatti, G. (1992).Understanding motor events: a neurophysiological study. Exp. Brain Res. 91,176–180.

Duane, T.D., and Behrendt, T. (1965). Extrasensory electroencephalographicinduction between identical twins. Science 150, 367.

Dumas, G., Nadel, J., Soussignan, R., Martinerie, J., and Garnero, L. (2010).Inter-brain synchronization during social interaction. PLoS ONE 5, e12166.

Dumas, G., Chavez, M., Nadel, J., and Martinerie, J. (2012). Anatomical con-nectivity influences both intra- and inter-brain synchronizations. PLoS ONE7, e36414.

Dunbar, R.I.M. (1998). The social brain hypothesis. Evol. Anthropol. 6,178–190.

Dunbar, R.I., and Shultz, S. (2007). Understanding primate brain evolution.Philos. Trans. R. Soc. Lond. B Biol. Sci. 362, 649–658.

Fliessbach, K.,Weber, B., Trautner, P., Dohmen, T., Sunde, U., Elger, C.E., andFalk, A. (2007). Social comparison affects reward-related brain activity in thehuman ventral striatum. Science 318, 1305–1308.

Friston, K., and Frith, C. (2015). A duet for one. Consc. Cogn. 36, 390–405.

Frith, C.D. (2007). The social brain? Philos. Trans. R. Soc. Lond. B Biol. Sci.362, 671–678.

Frith, C.D., and Frith, U. (2012). Mechanisms of social cognition. Annu. Rev.Psychol. 63, 287–313.

Funane, T., Kiguchi, M., Atsumori, H., Sato, H., Kubota, K., and Koizumi, H.(2011). Synchronous activity of two people’s prefrontal cortices during acooperative task measured by simultaneous near-infrared spectroscopy.J. Biomed. Opt. 16, 077011.

Garrod, S., and Pickering, M.J. (2004). Why is conversation so easy? TrendsCogn. Sci. 8, 8–11.

Gauthier, I., Tarr, M.J., Moylan, J., Skudlarski, P., Gore, J.C., and Anderson,A.W. (2000). The fusiform ‘‘face area’’ is part of a network that processes facesat the individual level. J. Cogn. Neurosci. 12, 495–504.

Gilbert, D.T., Killingsworth, M.A., Eyre, R.N., and Wilson, T.D. (2009). The sur-prising power of neighborly advice. Science 323, 1617–1619.

Goffman, E. (1966). Behavior in Public Places: Notes on the Social Organiza-tion of Gatherings (Free Press).

Hamalainen, M., Hari, R., Ilmoniemi, R., Knuutila, J., and Lounasmaa, O.V.(1993). Magnetoencephalography – theory, instrumentation, and applicationsto noninvasive studies of the working human brain. Rev. Mod. Phys. 65,413–497.

Hari, R., and Kujala, M.V. (2009). Brain basis of human social interaction: fromconcepts to brain imaging. Physiol. Rev. 89, 453–479.

Hari, R., and Parkkonen, L. (2015). The brain timewise: how timing shapes andsupports brain function. Philos. Trans. R. Soc. Lond. B Biol. Sci. 370,20140170.

Hari, R., and Salmelin, R. (2012). Magnetoencephalography: From SQUIDs toneuroscience. Neuroimage 20th anniversary special edition. Neuroimage 61,386–396.

Hari, R., Himberg, T., Nummenmaa, L., Hamalainen, M., and Parkkonen, L.(2013). Synchrony of brains and bodies during implicit interpersonal interac-tion. Trends Cogn. Sci. 17, 105–106.

Hari, R., Bourguignon, M., Piitulainen, H., Smeds, E., De Tiege, X., and Jous-maki, V. (2014). Human primary motor cortex is both activated and stabilizedduring observation of other person’s phasic motor actions. Philos. Trans. R.Soc. Lond. B Biol. Sci. 369, 20130171.

Hart, Y., Noy, L., Feniger-Schaal, R., Mayo, A.E., and Alon, U. (2014). Individ-uality and togetherness in joint improvised motion. PLoS ONE 9, e87213.

Hasson, U., Nir, Y., Levy, I., Fuhrmann, G., and Malach, R. (2004). Intersubjectsynchronization of cortical activity during natural vision. Science 303, 1634–1640.

Hasson, U., Landesman, O., Knappmeyer, B., Vallines, I., Rubin, N., andHeeger, D. (2008). Neurocinematics: The neuroscience of film. Projections 2,1–26.

Heider, F., and Simmel, M. (1944). An experimental study of apparentbehavior. Am. J. Psychol. 57, 243–259.

Henriksson, L., Mur, M., and Kriegeskorte, N. (2015). Faciotopy-A face-featuremap with face-like topology in the human occipital face area. Cortex. Pub-lished online July 16, 2015. http://dx.doi.org/10.1016/j.cortex.2015.1006.1030.

Himberg, T., Hirvenkari, L., Mandel, A., and Hari, R. (2015). Word-by-word entrainment of speech rhythm during joint story building. Front. Psychol.6, 797.

Hirata, M., Ikeda, T., Kikuchi, M., Kimura, T., Hiraishi, H., Yoshimura, Y., andAsada, M. (2014). Hyperscanning MEG for understanding mother-child cere-bral interactions. Front. Hum. Neurosci. 8, 118.

Holt-Lunstad, J., Smith, T.B., Baker, M., Harris, T., and Stephenson, D. (2015).Loneliness and social isolation as risk factors for mortality: a meta-analytic re-view. Perspect. Psychol. Sci. 10, 227–237.

Jaaskelainen, I.P., Koskentalo, K., Balk, M.H., Autti, T., Kauramaki, J., Pomren,C., and Sams, M. (2008). Inter-subject synchronization of prefrontal cortex he-modynamic activity during natural viewing. Open Neuroimaging J. 2, 14–19.

Jiang, J., Dai, B., Peng, D., Zhu, C., Liu, L., and Lu, C. (2012). Neural synchro-nization during face-to-face communication. J. Neurosci. 32, 16064–16069.

Jiang, J., Chen, C., Dai, B., Shi, G., Ding, G., Liu, L., and Lu, C. (2015). Leaderemergence through interpersonal neural synchronization. Proc. Natl. Acad.Sci. USA 112, 4274–4279.

Kanda, T., Hirano, T., Eaton, D., and Ishiguro, H. (2004). Interactive robots associal partners and peer tutors for children: a field trial. Hum. Comput. Interact.19, 61–84.

Kanwisher, N. (2010). Functional specificity in the human brain: a window intothe functional architecture of the mind. Proc. Natl. Acad. Sci. USA 107, 11163–11170.

Neuron 88, October 7, 2015 ª2015 Elsevier Inc. 191

Neuron

Perspective

Kanwisher, N., McDermott, J., and Chun, M.M. (1997). The fusiform face area:a module in human extrastriate cortex specialized for face perception.J. Neurosci. 17, 4302–4311.

Kawasaki, M., Yamada, Y., Ushiku, Y., Miyauchi, E., and Yamaguchi, Y. (2013).Inter-brain synchronization during coordination of speech rhythm in human-to-human social interaction. Sci. Rep. 3, 1692.

Kennedy, D.P., and Adolphs, R. (2012). The social brain in psychiatric andneurological disorders. Trends Cogn. Sci. 16, 559–572.

Kilner, J.M., Friston, K.J., and Frith, C.D. (2007). Predictive coding: an accountof the mirror neuron system. Cogn. Process. 8, 159–166.

King-Casas, B., Tomlin, D., Anen, C., Camerer, C.F., Quartz, S.R., andMontague, P.R. (2005). Getting to know you: reputation and trust in a two-per-son economic exchange. Science 308, 78–83.

Konvalinka, I., Bauer, M., Stahlhut, C., Hansen, L.K., Roepstorff, A., and Frith,C.D. (2014). Frontal alpha oscillations distinguish leaders from followers: multi-variate decoding of mutually interacting brains. Neuroimage 94, 79–88.

Koster-Hale, J., and Saxe, R. (2013). Theory of mind: a neural prediction prob-lem. Neuron 79, 836–848.

Kriegeskorte, N., Formisano, E., Sorger, B., and Goebel, R. (2007). Individualfaces elicit distinct response patterns in human anterior temporal cortex.Proc. Natl. Acad. Sci. USA 104, 20600–20605.

Lahnakoski, J.M., Glerean, E., Salmi, J., Jaaskelainen, I.P., Sams, M., Hari, R.,and Nummenmaa, L. (2012). Naturalistic FMRI mapping reveals superior tem-poral sulcus as the hub for the distributed brain network for social perception.Front. Hum. Neurosci. 6, 233.

Lee, R.F. (2015). Dual logic and cerebral coordinates for reciprocal interactionin eye contact. PLoS ONE 10, e0121791.

Lee, R.F., Dai, W., and Dix, W. (2010). A decoupled circular-polarized volumehead coil pair for studying two interacting human brains with MRI. Conf. Proc.IEEE Eng. Med. Biol. Soc. 2010, 6645–6648.

Lee, R.F., Dai, W., and Jones, J. (2012). Decoupled circular-polarized dual-head volume coil pair for studying two interacting human brains with dyadicfMRI. Magn. Reson. Med. 68, 1087–1096.

Lindeman, M., Svedholm, A.M., Riekki, T., Raij, T., and Hari, R. (2013). Is it justa brick wall or a sign from the universe? An fMRI study of supernatural be-lievers and skeptics. Soc. Cogn. Affect. Neurosci. 8, 943–949.

Lindenberger, U., Li, S.C., Gruber, W., andMuller, V. (2009). Brains swinging inconcert: cortical phase synchronization while playing guitar. BMC Neurosci.10, 22.

Malinen, S., and Hari, R. (2011). Data-based functional template for sorting in-dependent components of fMRI activity. Neurosci. Res. 71, 369–376.

Mandel, A., Helokunnas, S., Pihko, E., and Hari, R. (2015). Brain responds toanother person’s eye blinks in a natural setting-themore empathetic the viewerthe stronger the responses. Eur. J. Neurosci. Published July 1, 2015. http://dx.doi.org/10.1111/ejn.13011.

Moll, H., andMeltzoff, A.N. (2011). Joint attention as the fundamental perspec-tive in understanding perspectives. In Joint Attention: New Developmentsin Philosophy, Psychology, and Neuroscience, S. Seeman, ed. (MIT),pp. 393–413.

Montague, P.R., Berns, G.S., Cohen, J.D., McClure, S.M., Pagnoni, G., Dha-mala, M., Wiest, M.C., Karpov, I., King, R.D., Apple, N., and Fisher, R.E.(2002). Hyperscanning: simultaneous fMRI during linked social interactions.Neuroimage 16, 1159–1164.

Mori, M. (1982). The Buddha in the Robot (Tuttle Publishing).

Muller, V., and Lindenberger, U. (2011). Cardiac and respiratory patterns syn-chronize between persons during choir singing. PLoS ONE 6, e24893.

Muller, V., Sanger, J., and Lindenberger, U. (2013). Intra- and inter-brain syn-chronization during musical improvisation on the guitar. PLoS ONE 8, e73852.

Naselaris, T., Kay, K.N., Nishimoto, S., and Gallant, J.L. (2011). Encoding anddecoding in fMRI. Neuroimage 56, 400–410.

192 Neuron 88, October 7, 2015 ª2015 Elsevier Inc.

Nishitani, N., and Hari, R. (2000). Temporal dynamics of cortical representationfor action. Proc. Natl. Acad. Sci. USA 97, 913–918.

Nishitani, N., and Hari, R. (2002). Viewing lip forms: cortical dynamics. Neuron36, 1211–1220.

Northoff, G., and Bermpohl, F. (2004). Cortical midline structures and the self.Trends Cogn. Sci. 8, 102–107.

Nowak, M.A., and Sigmund, K. (2005). Evolution of indirect reciprocity. Nature437, 1291–1298.

Noy, L., Dekel, E., and Alon, U. (2011). The mirror game as a paradigm forstudying the dynamics of two people improvising motion together. Proc.Natl. Acad. Sci. USA 108, 20947–20952.

Noy, L., Levit-Binun, N., andGolland, Y. (2015). Being in the zone: physiologicalmarkers of togetherness in joint improvisation. Front. Hum. Neurosci. 9, 187.

Nummenmaa, L., and Calder, A.J. (2009). Neural mechanisms of social atten-tion. Trends Cogn. Sci. 13, 135–143.

Nummenmaa, L., Glerean, E., Viinikainen, M., Jaaskelainen, I.P., Hari, R., andSams, M. (2012). Emotions promote social interaction by synchronizing brainactivity across individuals. Proc. Natl. Acad. Sci. USA 109, 9599–9604.

Nummenmaa, L., Glerean, E., Hari, R., and Hietanen, J.K. (2014a). Bodilymapsof emotions. Proc. Natl. Acad. Sci. USA 111, 646–651.

Nummenmaa, L., Smirnov, D., Lahnakoski, J.M., Glerean, E., Jaaskelainen,I.P., Sams, M., and Hari, R. (2014b). Mental action simulation synchronizesaction-observation circuits across individuals. J. Neurosci. 34, 748–757.

Pamilo, S., Malinen, S., Hlushchuk, Y., Seppa, M., Tikka, P., and Hari, R.(2012). Functional subdivision of group-ICA results of fMRI data collected dur-ing cinema viewing. PLoS ONE 7, e42000.

Piper, S.K., Krueger, A., Koch, S.P., Mehnert, J., Habermehl, C., Steinbrink, J.,Obrig, H., and Schmitz, C.H. (2014). A wearable multi-channel fNIRS systemfor brain imaging in freely moving subjects. Neuroimage 85, 64–71.

Pronin, E. (2008). How we see ourselves and how we see others. Science 320,1177–1180.

Pronin, E., Kruger, J., Savitsky, K., andRoss, L. (2001). You don’t knowme, butI know you: the illusion of asymmetric insight. J. Pers. Soc. Psychol. 81,639–656.

Ramkumar, P., Parkkonen, L., and Hari, R. (2010). Oscillatory response func-tion: towards a parametric model of rhythmic brain activity. Hum. Brain Mapp.31, 820–834.

Remes, S., Klami, A., and Kaski, S. (2013). Characterizing unknown events inMEGdatawithgroup factoranalysis. InProceedingsof the3rdWorkshoponMa-chine Learning and Interpretation in Neuroimaging (MLINI) (Lake Tahoe, USA),https://sites.google.com/site/mlininips2013/proceedings-of-mlini-2012-1.

Richardson, M.J., Marsh, K.L., Isenhower, R.W., Goodman, J.R., and Schmidt,R.C. (2007). Rocking together: dynamics of intentional and unintentional inter-personal coordination. Hum. Mov. Sci. 26, 867–891.

Rizzolatti, G., and Craighero, L. (2004). The mirror-neuron system. Annu. Rev.Neurosci. 27, 169–192.

Rust, N.C., and Movshon, J.A. (2005). In praise of artifice. Nat. Neurosci. 8,1647–1650.

Rutter, M., Andersen-Wood, L., Beckett, C., Bredenkamp, D., Castle, J., Groo-thues, C., Kreppner, J., Keaveney, L., Lord, C., and O’Connor, T.G.; Englishand Romanian Adoptees (ERA) Study Team (1999). Quasi-autistic patternsfollowing severe early global privation. J. Child Psychol. Psychiatry 40,537–549.

Salmelin, R., and Hari, R. (1994). Characterization of spontaneous MEGrhythms in healthy adults. Electroencephalogr. Clin. Neurophysiol. 91,237–248.

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

Neuron

Perspective

Scassellati, B., Admoni, H., and Matari�c, M. (2012). Robots for use in autismresearch. Annu. Rev. Biomed. Eng. 14, 275–294.

Schilbach, L. (2015). The neural correlates of social cognition and social inter-action. In Brain Mapping: An Encyclopedic Reference, A.W.W. Toga, ed.(Elsevier), pp. 159–164.

Schilbach, L., Timmermans, B., Reddy, V., Costall, A., Bente, G., Schlicht, T.,and Vogeley, K. (2013). Toward a second-person neuroscience. Behav. BrainSci. 36, 393–414.

Schultz, J., and Pilz, K.S. (2009). Natural facial motion enhances cortical re-sponses to faces. Exp. Brain Res. 194, 465–475.

Spengler, S., von Cramon, D.Y., and Brass, M. (2009). Control of shared rep-resentations relies on key processes involved in mental state attribution. Hum.Brain Mapp. 30, 3704–3718.

Sravish, A.V., Tronick, E., Hollenstein, T., and Beeghly, M. (2013). Dyadic flex-ibility during the face-to-face still-face paradigm: a dynamic systems analysisof its temporal organization. Infant Behav. Dev. 36, 432–437.

Stanley, D.A., and Adolphs, R. (2013). Toward a neural basis for socialbehavior. Neuron 80, 816–826.

Stephens, G.J., Silbert, L.J., and Hasson, U. (2010). Speaker-listener neuralcoupling underlies successful communication. Proc. Natl. Acad. Sci. USA107, 14425–14430.

Stivers, T., Enfield, N.J., Brown, P., Englert, C., Hayashi, M., Heinemann, T.,Hoymann, G., Rossano, F., de Ruiter, J.P., Yoon, K.E., and Levinson, S.C.(2009). Universals and cultural variation in turn-taking in conversation. Proc.Natl. Acad. Sci. USA 106, 10587–10592.

Thirioux, B., Mercier, M.R., Blanke, O., and Berthoz, A. (2014). The cognitiveand neural time course of empathy and sympathy: an electrical neuroimagingstudy on self-other interaction. Neuroscience 267, 286–306.

Tomlin, D., Kayali, M.A., King-Casas, B., Anen, C., Camerer, C.F., Quartz, S.R.,and Montague, P.R. (2006). Agent-specific responses in the cingulate cortexduring economic exchanges. Science 312, 1047–1050.

Trevarthen, C. (2011). What is It like to be a person who knows nothing?Defining the active intersubjective mind of a newborn human being. InfantChild Dev. 20, 119–135.

Vaina, L.M., Solomon, J., Chowdhury, S., Sinha, P., and Belliveau, J.W. (2001).Functional neuroanatomy of biological motion perception in humans. Proc.Natl. Acad. Sci. USA 98, 11656–11661.

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

Wagner, D.D., Haxby, J.V., and Heatherton, T.F. (2012). The representation ofself and person knowledge in the medial prefrontal cortex. Wiley Interdiscip.Rev. Cogn. Sci. 3, 451–470.

Wing, A.M., Endo, S., Bradbury, A., and Vorberg, D. (2014). Optimal feed-back correction in string quartet synchronization. J. R. Soc. Interface 11,20131125.

Zhdanov, A., Nurminen, J., Baess, P., Hirvenkari, L., Jousmaki, V., Makela,J.P., Mandel, A., Meronen, L., Hari, R., and Parkkonen, L. (2015). AnInternet-based real-time audiovisual link for dual MEG recordings. PLoSONE 10, e0128485.

Neuron 88, October 7, 2015 ª2015 Elsevier Inc. 193