2 Bridging between actionrepresentation and infanttheory of mind

Andrew N Meltzoff

Normal adults share a network of ideas about human psychology thatare often described as "commonsense" or "folk" psychology. Although wedirectly observe other people's behavior, we interpret people to be morethan their biological movements. We think of them as having internal mentalstates just like us. We think that human beings want, think, and feel, and thatthese states lead to their actions. Our ideas about these mental states playa crucial role in our interactions with others and the regulation of our ownbehavior.

People with autism illustrate how important mental-state attribution isin everyday life. Individuals with autism do not conceptualize other peopleas psychological agents with a rich palette of mental states - there is animpairment in commonsense psychology. Autism has been likened to akind of "mind-blindness" with a devastating impact on social relations(Baron-Cohen, 1995).

Commonsense psychology may be common among typical adults, butit has a developmental history (Meltzoff, Kuhl, Movellan, & Senjowski,2009). Here I argue that if we want to discover the origins of commonsensepsychology, we must start by analyzing typical infant development and theelementary building blocks of social cognition. I marshal theoretical andempirical support for a developmental framework that rests on two comple­mentary processes, one going from the inside out ("like-me") and the otherfrom the outside in ("like-you"). According to this view, the bedrock on whichcommonsense psychology is constructed is the bidirectional apprehensionthat others are similar to the self. Infants are launched on their careerof interpersonal relations with the basic perception: "Here is something likeme" and "I can act like you." The aim of this chapter is to explore the earlymanifestations and developmental sequelae of these preverbal building blocksfor social cognition.

My thesis is that the equivalence between self and other at the level ofaction supports bidirectional learning effects. Going from the inside out,infants' understanding of others' acts is imbued with new meaning byperforming similar acts themselves. Going from the outside in, infantslearn about themselves and the consequences of their own potential action,

16 Meltzo.fJ

before and without producing it, simply by observing the behavior of others.The same underlying mechanism supports learning in both directions: Actionperception influences production and imitation ("like-you"); conversely,action production and self-experience influence the perception and inter­pretation of the acts of others, which I interpret as "like-me" effects. Thekey point is that human infants can run the system in both directions, andthis recognition of interpersonal equivalence is what sets the human childon course for developing their species-typical social cognition. Deficits inthese building blocks may be what we see in cases of autism spectrumdisorders.

A proposed mechanism for change in social cognition

This chapter describes how infants use the recognition of equivalence withothers at the level of action to bootstrap themselves into an increasinglymore adult-like view of other people as intentional agents. A sketch for theproposed developmental pathway follows.

Step 1: Innate equipment

Typical newborns detect and use equivalences between observed and exe­cuted acts. When newborns see adult biological motion, including handand face movements, these acts are mapped onto the infant's body move­ments. This mapping is manifest by newborn imitation. Self and other areintrinsically bound through an innate coding of human acts that is abstractenough to unite the perception and production of behavior. My own feltacts and the acts I watch you perform are registered by the same abstractcode. Meltzoff and Moore (1977, 1997) called this a supramodal representa­tion because it cuts across particular modalities. It is because of the infant'saction representation - the supramodal code - that the movements of peopleare special to young babies. The supramodal representation of human actionprovides the lingua franca, the abstract code, for connecting self and other inthe first instance.

Step 2: First-person experience

Through everyday experience infants map the relation between their ownbodily states and mental experiences. For example, there is an intimate rela­tion between striving to achieve a goal and the concomitant facial expressionof concentration and effortful bodily acts. Infants experience their ownunfulfilled desires and their own matching facial/postural/vocal reactions.They experience their own inner feelings and behavioral facial expressionsand construct a detailed bidirectional map linking internal experiences andbehavior (Meltzoff & Brooks, 2008).

Action representation and theory ofmind 17

Step 3: Attributions to others

When infants see others acting in a way that is similar to how they have actedin the past, they make an attribution. They ascribe the internal feelings thatregularly go with those behaviors, based on their self-experience. This givesinfants leverage for grasping other minds before language can be used.Infants integrate the ability to connect the acts of self and other (Step 1) withtheir own self-experience (Step 2) to yield a deeper understanding of whatlies behind the behavior of others. As children's own experiences expand,they have an enriched understanding of what another might be feeling,desiring, and perceiving when he or she acts in certain telltale ways.

This attributes a good deal of machinery to infants, but it is not Fodoriannativism (Fodor, 1983). Newborns do not innately possess the adult theory ofmind. Nor is the adult conception of others simply triggered by particularcues or due to the maturation of modules. Rather, infants' initial grasp thatthe actions of others are "like-me," coupled with their observations of others'acts and self-experiences, provides them with a developmental mechanismfor coming to understand the internal experiences of others (for furtherdiscussion, see Meltzoff & Brooks, 2008).

Clarifications and caveats

When doing developmental psychology, using English words to describeinfant mental states always presents a problem. The words refer to the adultmeanings, and infant concepts are, for the most part, not the same as adults'.The use of the English words "me" and "you" is not meant to imply thatyoung infants adopt the adult meanings. The "me" of the adult entails verbalself-reflection, which is absent in the infant; it may also involve my appear­ance when I look in the mirror and whether I think of myself as havingcertain attributes, such as being a generous person. When I say that the infantconstrues another as "like-me," this is a gloss for the infant recognizing anequivalence at the level of actions. What I mean is that the infant recognizes across-modal equivalence between acts that they observe in others and pro­duce themselves: "That seen act is like this felt act." The acts of others areseen as "like-me," but it is the isomorphism at the level of action representa­tion that I am pointing to, not a fully mature sense of "me" or "self," which isa developmental achievement (Meltzoff, 2007a; Meltzoff & Moore, 1995).

Moreover, it is clear that recognizing self-other equivalence at the level ofaction provides only a partial story about how children come to understandothers in a mature adult manner. The mental states most penetrable aredesires, visual perception, emotions, and simple intentions. For these, there isa relatively close coupling between the underlying mental states and theirexpression in bodily action, so the equivalence in self-other behavior is of atleast some use. When someone sees x or desires y, there are usually bodilymovements that correlate with such mental states (leaning towards, grasping,

18 Meltzoff

looking in the direction of, etc.). These bodily movements are the ones infantssee and the ones they feel themselves make. Other mental states are morequarantined from the action stream - for example, the "belief that x" (whichmay be false) can be held in the absence of any telltale action. Further devel­opment is needed for grasping the idea that another who is equivalent to mebelieves something that conflicts with my beliefs or perceptions (e.g., Astington& Gopnik, 1991; Flavell, 1999; Moll & Meltzoff, in press; Perner, 1991;Wellman, 1990). Thus, the proposal offered here is merely the initial footholdfor interpreting others as bearers of psychological properties commensuratewith one's own - the bare first step in developing a theory of mind.

Action representation and imitation

What is the empirical evidence that young infants use a supramodal codethat links perception and production? Perhaps the best evidence comesfrom infant imitation. Imitation demonstrates that, at some level of process­ing, infants use the observed behavior of others as a basis for forming acorresponding action plan. Through imitation, infants make manifest theconnection between self and other at the level of shared actions. The infantcan quite literally act "like-you."

When in development does this coding ofaction begin? Meltzoff and Moore(1977) reported that 12- to 21-day-old infants imitate facial expressions.Subsequent studies tested newborns in a hospital setting (Meltzoff & Moore,1983, 1989). The youngest infant was only 42 minutes old. The results showedthat these newborns imitate adult gestures. Imitation in early infancy has nowbeen replicated in more than 24 studies from around the world (for a review,see Meltzoff & Moore, 1997).

It is of interest that the neonatal imitative matching is quite specific;it is not a global response or a general arousal reaction. Infants responddifferentially to two types of lip movements (mouth opening versus lipprotrusion) and two types of protrusion actions (lip protrusion versus tongueprotrusion). Infants also differentiate two different types of tongue movementsfrom one another (a tongue thrust at midline versus one off-midline, from thecorner of the mouth), establishing that early behavioral matching is notmerely a global response (Meltzoff & Moore, 1994).

Infant imitation cannot be reduced to automatic resonance. The responsecan be displaced in time and space from the demonstration. In one study, apacifier was put in infants' mouths as they watched the display. After thepacifier was removed, the infants imitated the earlier displays (Meltzoff& Moore, 1977). In another study, an adult showed 6-week-old infants agesture, and then the infants were taken home for a 24-hour memory delay.The following day they were presented with the same adult sitting with aneutral facial expression. If the adult had shown mouth opening the daybefore, the infants initiated that gesture from memory; if the adult hadshown tongue protrusion, infants responded with that gesture (Meltzoff &

Action representation and theory ofmind 19

Moore, 1994). The imitative response is not rigidly preprogrammed in theform of a fixed-action pattern. Infants correct their imitative attempts so thatthey converge more and more closely on the model demonstrated (Meltzoff &Moore, 1997). This is genuine imitation, where the goal is to match the adult'starget.

Solving the correspondence problem

Meltzoff and Moore (1997) proposed a mechanism that underlies imitation.They dubbed it the AIM theory - an acronym for "active intermodal map­ping." On this view infants can, at some primitive level, recognize equivalencebetween the acts they see others do and the acts they do themselves. This isnot a complex mechanism that requires cognitive machinations by the infant.Rather, there appears to be a very primitive and foundational body schemethat allows infants to unify the seen acts of others and their own felt acts intoone common framework. The infants' own facial gestures are invisible tothem, but they are not unperceived. They are monitored by proprioception.Conversely, the adult's acts are not felt by proprioception, but they canbe seen. Infants can link perception and production through what AIMterms a common supramodal coding of human acts. This is why they cancorrect their imitative movements. And it is why they can imitate frommemory: Infants store a representation of the adult's act and it is the tar­get against which they compare their own acts. A more detailed descriptionof the metric of equivalence between self and other is provided elsewhere(Meltzoff & Moore, 1997).

The idea of a supramodal representation of action that we used to explainearly imitation fits well with modern neuroscience discoveries about sharedneural circuits for perception and action and so-called neural mirroringsystems (Hari & Kujala, 2009; Iacoboni et aI., 1999; Rizzolatti, Fogassi, &Gallese, 2001). An important task for the future is to analyze the commonal­ities and differences between the supramodal code, which is proposed at thepsychological/personal level, and shared neural circuits, which are proposedat the neural/sub-personal level. For example, the neural mirroring systemsare better suited to explain fast, automatic, non-effortful resonance than toexplain deferred imitation, response correction, and the imitation of novelacts. Analyses are beginning to emerge that compare these notions and othersfor uniting perception and action (e.g., Hommel, Miisseler, Aschersleben, &Prinz, 2001; Meltzoff & Decety, 2003; Meltzoff, 2007b; Rizzolatti, Fadiga,Fogassi, & Gallese, 2002; Sommerville & Woodward, 2005). Moreover,relevant infant social-cognitive neuroscience studies are being done usingEEG measures of mu-rhythm desynchronization in response to actionperception and production (e.g., Marshall, Young, & Meltzoff, 2010).

The unique contribution from the human developmental literature is thatnewborn imitation demonstrates that self-other connectedness is functionalat birth. At this moment. we lack the corresponding newborn studies with

20 Meltzojf

monkeys to establish whether mirror neurons are present in the newbornmonkey prior to postnatal experience and associative learning (for a discus­sion about the potential importance of prenatal experience, see Meltzoff &Moore, 1997). Importantly, I do not argue that the human imitative capacityhas reached adult-like levels at birth, and I have described several developmentsin imitation in later infancy and early childhood (e.g., Gleissner, Meltzoff, &Bekkering, 2000; Meltzoff, 2007b; Repacholi & Meltzoff, 2007; Repacholi,Meltzoff, & Olsen, 2008; Williamson, Jaswal, & Meltzoff, 2010; Williamson,Meltzoff, & Markman, 2008). However, the behavioral" discoveries doestablish that human infants are born learning from observing other people.

Goals and intentions of other people

In mature adult social cognition, other people do not solely act in ways thatare like the self - they share deeper similarities as well. Others not only act"like me," they also are understood to have mental states, including beliefs,emotions, and intentions that are like my own. When do preverbal infantsbegin to ascribe intentionality to human movement patterns? The behavioralre-enactment procedure was designed to provide a nonverbal technique forexploring intention reading in preverbal infants (Meltzoff, 1995). The pro­cedure capitalizes on children's natural tendency to re-enact or imitate butuses it in a more abstract way to investigate whether infants can read belowthe literal surface behavior to the goals or intentions of the actor.

The experimental procedure involves showing infants an unsuccessful act.For example, the adult accidentally under- or overshoots his target, or he triesto pull apart a dumbbell-shaped toy, but his hand slips off the ends and he isunsuccessful. Thus the goal-state is not achieved. To an adult, it is easy toread the actor's intentions even though he never fulfills them. The experi­mental question is whether children read through the literal body movementsto the underlying goal or intention of the act. The measure of how theyinterpreted the event is what they choose to re-enact, in particular whetherthey choose to ignore what the adult literally did and instead produce theintended act, despite the fact that it was never present to the senses. In a sense,the correct answer is to copy not the literal movement, but the intended actthat remains unfulfilled and invisible.

Meltzoff (1995) showed 18-month-old infants an unsuccessful act, a failedeffort. The study compared infants' tendency to perform the target act inseveral situations: (1) after they saw the full-target act demonstrated, (2) afterthey saw the unsuccessful attempt to perform the act, and (3) after it wasneither shown nor attempted. The results showed that 18-month-olds caninfer the unseen goals implied by unsuccessful attempts. Infants who saw theunsuccessful attempt and infants who saw the full-target act both producedtarget acts at a significantly higher rate than controls. Evidently, toddlers canunderstand our goals even if we fail to fulfill them.

In the adult framework, people act intentionally, but inanimate things do

Action representation and theory ofmind 21

not. To begin to examine this question of the ascription of intentionality,Meltzoff (1995) tested how 18-month-olds responded to a mechanical devicethat mimicked the same movements as the actor in the failed-attempt con­dition. An inanimate device was constructed that had poles for arms andmechanical pincers for hands. It did not look human, but it could movevery similarly to a human. For the test, the pincers "grasped" the dumbbellat the two ends just as the human hands did. One mechanical arm wasthen moved outwards, just as in the human case, and its pincer slipped offthe end of the dumbbell, just as the human hand did. The movement patternsof machine and man were closely matched from a purely spatiotemporaldescription of movements in space.

The experimental results showed that infants did not attribute a goal orintention to the movements of the inanimate device. Although they werenot frightened by the device and looked at it as long as at the human display,they did not see the sequence of actions as implying a goal. Infants were nomore - or less - likely to pull the toy apart after seeing the failed attempt ofthe inanimate device than in baseline conditions when they had seen nothing.

Taken together, this research shows that by 18 months of age children havealready adopted a fundamental aspect of a mature commonsense psychology- persons are understood within a framework involving goals and intentions.Just as importantly, the work shows that children did not slavishly imitate theunsuccessful motion by letting their fingers slip from the object but, instead,they completed the intended goal. Even though they had never seen the com­pleted act, the children inferred the goal of the act from the actor's try-and­try again behavior. This and other related work strongly suggests that infantscan interpret what the adult is trying to do. They re-enact the goal of the actand do not slavishly duplicate what was literally done (Brandone & Wellman,2009; Carpenter, Akhtar, & Tomasello, 1998; Gergely, Bekkering, & Kinily,2002; Hamlin, Hallinan, & Woodward, 2008; Williamson et aI., 2008).

Visual perceptions of other people

Another important aspect of social cognition is the realization that people aresources of information about external objects. For adults, particular bodymovements have special meanings - they are about something. If a personlooks up into the sky, bystanders follow his or her gaze. Adults realize thatpeople acquire information from afar, despite the spatial gap between viewerand object. Visual perception is a kind of psychological contact at a distance.

When do infants begin to ascribe visual perception to others? Is there astage when head turns are interpreted as purely physical motions, with nonotion that they are directed toward the external object, no notion of aperceiver? In fact, children with autism may regard adults' looking behaviorin this way. Children with autism have gaze-following deficits, and it can bespeculated that they process adults' looking behavior more as a physicalmovement in space than as a psychological act that connects the perceiver

22 Meltzoff

and the world (Hobson & Meyer, 2005; Mundy, 2009; Mundy & Newell,2007; Toth, Munson, Meltzoff, & Dawson, 2006).

There is a debate about the mechanisms underlying gaze following intypically developing infants and whether it shows that they have a primitivegrasp of seeing and visual contact in others (Flom, Lee, & Muir, 2007). Oneview is that young infants initially treat others' looking behaviors as meremovements. In the leanest version, young infants visually track the adult'shead movement in space and are pulled into the correct hemi-field where theycatch sight of the salient target object by happenstance (Butterworth & Jar­rett, 1991). Over time, infants then learn that the adult's head turn is a reliablecue, indicating where an object can be seen (Moore, 1999,2006). Conversely,others have offered a nativist view suggesting that infants have a built-inmodule that takes eye gaze as input and automatically makes attributionsabout seeing and visual experience in others (Baron-Cohen, 1995). A third,developmental view is that infants' understanding of others' vision emergesfrom more primitive beginnings. Me1tzoff and Brooks (2008) propose that amechanism of change is infants' experience with their own vision: Infantsdevelop an understanding of the vision of others in part through their ownacts of turning-in-order-to-see and opening/shutting their eyes to cut off andreinstate visual experience.

One key data point for adjudicating this debate is to determine whetherinfants are, as the lean view suggests, simply processing the salient physi­cal movements in space caused by the head. Brooks and Meltzoff (2002)developed a protocol that zeroed in on the importance of eyes in infant gazefollowing. In this procedure, an adult turned to look at one of two targets.The manipulation was that the adult turned to the target with eyes open forone group and with eyes closed for the other group. If infants rely simplyon head motions, they should turn in both cases. If infants appreciate thatthe eyes are relevant for connecting a perceiver and an object, they shoulddifferentiate between the two conditions and turn to look at the target in onecase but not in the other. The reason such a manipulation is crucial is thatwe do, in fact, see with our eyes and not with our head. It is an importantstep forward in social cognition for infants to put special emphasis on eyes(something children with autism may not do; see Mundy, 2009). It is, after all,the eyes that are the window to the soul - the head is not such a portal.

Brooks and Meltzoff (2002) used the Gaze Following: Eyes Open/ClosedTest to assess 12-, 14-, and 18-month-old infants. Each infant at each age wasrandomly assigned to a condition in which the adult turned to the target witheither open or closed eyes. Infants at all three ages followed the adult signifi­cantly more often when the adult turned with open than with closed eyes.At the ages tested, our current findings disprove the leanest interpretation,because head movement was controlled and infants were more likely to lookat the correct target when the social partner could see it.

Eye closure is only one way to block a person's line of sight. Another wayis to use an inanimate object as a barrier. For an adult, an opaque physical

Action representation and theory ofmind 23

barrier has the same function as closed eyes - both prevent visual access.This is not the case for 12-month-olds. We found that infants understandthat vision is cut off by the biological motion of eye closure in advance ofunderstanding that an inanimate barrier does so.

In the study of inanimate occluders, the person turned toward a targetwearing either a headband or a blindfold (Brooks & Meltzoff, 2002). In bothinstances, the same cloth covered part of the experimenter's face, but in onesituation the adult could see and in the other she could not. We found that 14­and 18-month-old infants looked at the adult's target significantly more oftenin the headband than in the blindfold condition. In contrast, the 12-month­aids infants did not distinguish between the two conditions. They systematic­ally looked at the indicated target, whether the adult turned wearing theblindfold or the headband. This is not just a matter of blindfolds causing ageneral suppression of activity. Rather, 12-month-old infants make the mis­take of following the "gaze" of the adult wearing the blindfold. Evidently,they recognize that the human act of eye closure blocks contact with externalobjects, but they do not yet understand the same about inanimate occluders.

Self-experience changes infants' understanding of others:A training study

Infants understand eye closure in advance of view-blockage by inanimatebarriers, but the adult's line of sight is blocked in both cases. Why is therea developmental difference? Eye closure is a biological motion with whichinfants have extensive first-person experience: They can control their ownvision by closing their eyes when they do not want to look at something.The experience of turning visual access to the world off and on by eyeclosing/opening might serve as a framework for understanding such behaviorin others. We wanted to know whether infants who are given systematicexperience that blindfolds block their own view might make different attribu­tions to others. To test this, we designed a novel training study that providedspecific blindfold experience to infants (Meltzoff & Brooks, 2008).

In one study, 12-month-old infants were randomly assigned to one of threegroups. Infants in the treatment group were provided with various interestingobjects to play with on the table; when they looked down to visually inspectan object, the experimenter held an opaque blindfold between the objectand the child's eyes (Figure 2.1). Thus, the infants experienced that theirown view was blocked when the opaque blindfold was held in front of theireyes and was restored again when the blindfold was removed. This experiencehad nothing to do with the experimenter's viewpoint; it was a first-personexperience. Two control groups were used. One involved a specially con­structed windowed cloth, which was made from the same material as theblindfold but had a window cut out of the center. Infants in the windowgroup received the same protocol as just described, thus controlling forthe experience of an adult inserting a cloth between them and the objects;

24 MeltzoJf

A.

B.

Figure 2.1 A 12-month-old boy in Meltzoff and Brooks' (2008) training study. Infantswho were randomly assigned to the experimental group were given self­experience with a blindfold. Infants first looked at an interesting object(A). The blindfold then blocked their view of the toy (B). This was repeatedover a 7-min training session. The results show that infants subsequentlyused this self-experience to make an inference about the effect of theblindfold on other people's visual perception. This is an example of a"like-me" attribution: "The blindfold blocks my own perception and there­fore must have the same effect on another person's visual perception."For a more elaborated theoretical discussion, consideration of alternatives,and description of control groups see Meltzoff and Brooks (2008).

however, they could peer through the windowed cloth. Infants in the baselinecontrol group were simply familiarized with the opaque cloth while it waslying flat on the table, so they could see and touch it, but they did not receive atutorial on its view-blocking properties.

At the end of training all three groups were given a standard gaze-followingtest: Infants were presented with the blindfolded adult who turned toward thedistal objects. The results showed that infants who had received first-persontraining on the opaque blindfold now interpreted the adult's blindfoldedturning correctly. They did not turn when the adult wore the blindfold.

Action representation and theory ofmind 25

Importantly, infants who had the windowed-cloth experience and the baselineinfants mistakenly followed the blindfolded adult's "gaze" to the distal objectwhen she wore the blindfold (replicating Brooks & Meltzoff, 2002).

In the natural course of development, infants change their understandingof visual perception. By 18 months of age, infants do not act as though adultscan see through opaque barriers (Brooks & Meltzoff, 2002; Butler, Caron, &Brooks, 2000; Dunphy-Lelii & Wellman, 2004). Meltzoff and Brooks (2008,Expt. 2) capitalized on this by providing 18-month-olds with a completelynovel self-experience - one they would not have encountered outside thelaboratory. We constructed a trick blindfold that looked opaque from theoutside but was made of special material that could be seen through whenheld close to the eyes. Infants were randomly assigned to one of three groups:(a) experience with the trick blindfold, (b) experience with the opaque blind­fold, and (c) baseline experience in which they simply played with the trickblindfold as an object while it lay flat on the table. As in the previous study,for infants in the first two groups the blindfold was interposed between theireyes and the toys during the training period. The opaque blindfold blockedtheir view, and the trick blindfold provided infants with the experience thatthe - apparently opaque - blindfold could in fact be seen through.

After training, infants in all three groups saw the adult wear the blindfoldin our standard test. As expected, infants in the baseline group and theopaque-blindfold groups refrained from following the adult's head turnswhen the adult wore the blindfold. The new finding is that infants who hadfirst-person experience with the trick see-through blindfold followed theadult's head turns significantly more often than did infants in the othertwo groups.

This underscores the power of infant self-experience. Infants were given aparticular novel experience under experimental control. They immediatelyused the novel self-experience to change their construal of the behavior ofothers. They assumed that the other could see through the blindfold, eventhough the adult's eyes were covered and it looked, from the outside, as if shecould not see.

This is the first study showing that infants use first-person experience abouta psychological state such as seeing to make interpretations about anotherperson. I think these training effects are a case of like-me projection withimplications for how infants' self-experience transforms their understanding ofthe minds of others who act like them, as is elaborated on later in this chapter.

Developmental theory

The puzzle ofsocial cognition

The fundamental puzzle of social cognition stems from the fact that personsare more than physical objects. Enumerating a person's height, weight, andeye color does not exhaust our description of that person. We have skipped

26 MeltzofJ

over their psychological makeup. If a self-mobile, human-looking body wasdevoid of psychological characteristics, it would not be a person at all but arobot or, to use the philosopher's favorite, a zombie. A fundamental issueis how we come to know others as persons like ourselves. Each of us hasthe phenomenological experience that we are not alone in the world, not theunique bearer of psychological properties. We know that we perceive, feel,and intend, and we believe that others have psychological states just like ours.

Philosophers seek to justify the inference that the dynamic sacks of skin wesee are animated by psychological states. They contemplate whether this is afiction and assemble criteria for knowing whether it is or is not (e.g., Ryle,1949). Developmental psychologists ask different questions. We inquire howsuch a view takes hold regardless of whether it is logically justified. Is itinnately specified? Does it differ in children with autism?

Fodor (1987) thinks that infants innately assign adult commonsense psy­chology to people. The opposing school is that newborns lack any inklingthat other humans have psychological properties. It is claimed, for example,that the child is born a solipsist (Piaget, 1954) or is in a state of so-callednormal autism (Mahler, Pine, & Bergman, 1975), treating people the same asthings. It is a long way - I would sayan impossible path - to get from there tocommonsense psychology.

Starting-state nativism: A third way

Modern developmental scientists, including myself, have been trying todevelop a third way. It grants far more to the newborn than the secondview, while stopping short of the first. In my view, infant imitation andthe neural representations that underlie it provide an innate foundation forbuilding the adult understanding of people, but infants do not possess theadult framework. Infants imitate at birth, but they do not infer intentions orfully understand visual perception as a mental state in others; there areimportant developments, as shown in Brooks and Meltzoff (2005), Meltzoff(1999), and Meltzoffand Moore (1995). This is hardly grounds for Fodoriannativism. It is equally true that young infants outstrip Piagetian theory.What we need is a new theory of social development that includes a rich initialunderstanding and a mechanism of change that can transform this into theadult state based on structured interpersonal experience. We have called thisstarting-state nativism (Gopnik & Meltzoff, 1997; Meltzoff & Gopnik, 1993)to indicate a highly structured innate groundwork and then transformationaldevelopment based on social interaction and experience (see earlier sectionheading: A Proposed Mechanismfor Change in Social Cognition).

Development

Infants' action representation and imitation demonstrate that they map otherpeople's behavior onto their own bodies. Because human acts are seen in

Action representation and theory ofmind 27

others and performed by the self, the infant can grasp the social connection.This construal of certain movements in the environment as me-relevant thenhas cascading developmental effects. First, the world of material objectscan be divided into those entities that perform these acts (people) and thosethat do not (things). Second, the lingua franca of human acts providesaccess and interpretations of others' behaviors that are not afforded by thespatio-temporal movements of things.

It has long been thought that the commonality between self and other isintegral to our understanding ofother minds (e.g., Smith, 1759/1976). The waythat the classic philosophers went wrong is that they thought the self-otherequivalence was late to develop - emerging from language or complex cogni­tive analyses. The last quarter-century of research forces a revision in this view.The recognition of self-other equivalences is the starting point for socialcognition - a precondition for infant social understanding, not the outcome ofit. Contemporary philosophers of the mind are influenced by the new findingsabout infant imitation, joint attention, and intention-reading from preverbalinfants (Campbell, 1994; Gallagher, 2001; Goldman, 2005; Gordon, 2005).

Interdisciplinary implications: Autism, social robots, neuroscience, andhuman empathy

The idea that action representation and self-other equivalence is a basicbuilding block for developing social cognition is proving to be useful instudies of autism, robotics, and social neuroscience. For example, deficienciesin the "like-me"I"like-you" comparisons may help illuminate the puzzlingpattern of impairments exhibited by children with autism. They have specificdeficits in imitation, gaze following, and emotion understanding (e.g.,Dapretto et a!., 2006; Dawson, Meltzoff, Osterling, & Rinaldi, 1998; Hobson& Meyer, 2005; Meyer & Hobson, 2004; Mundy, 2009; Nadel, 2006; Rogers,2006), all of which are underwritten by the equivalence between self andother. Thus, what theorists have laboriously put together in our discussionof supramodal action representation and self-other isomorphisms is tornasunder in autism. Exactly what piece of the "like-me" and "like-you" build­ing blocks is impaired and how that impacts and alters development is agrand challenge remaining to be solved.

There is also a connection between modern infant studies and emergingwork in adult social neuroscience. Based in part on the infant data, wedesigned brain-imaging studies on imitation in adults. In one study, we foundthat the left inferior parietal lobe (IPL) is activated when adults imitate theacts of others ("like-you") and the right IPL is activated when they recognizethat someone is imitating them ("like-me"), providing distinguishable neuralmechanisms for the hypothesized cognitive differentiation (Decety, Chami­nade, Grezes, & Meltzoff, 2002). In another study, we found that observingbody actions from a first- versus a third-person perspective (me versus not­me) leads to different neural processing (Jackson, Meltzoff, & Decety, 2006).

28 MeltzofJ

The framework discussed here has also been useful in thinking abouthuman empathy. Studies of human empathy show that adults' neural respon­ses to injury vary as a function of the like-me-ness of that entity. One recentstudy from our laboratory manipulated the injured entity along a continuumrepresenting self, to another person, to a life-like but inanimate mannequin(Jackson, Brunet, Meltzoff, & Decety, 2006). In other work, we measuredbrain and behavioral responses to seeing another person have their handinjected with a needle. There were two different instructional conditions, butthe visible stimulus was the same in both cases: The participants in thefMRI study were instructed either that the hand was anesthetized before theneedle injection or that it was not. Results showed an automatic bottom-upresponse in the observer's pain matrix in both cases, but a top-down regula­tion of the reaction in the case of the anesthetized hand. We interpret this asshowing that observers modulate their empathic pain responses depending onthe target's situation even if it differs from one's own (Lamm, Nusbaum,Meltzoff, & Decety, 2007). In our most recent work, we tested empathy forpain in cases that varied whether the person was "like-me" or not. We did thisby having participants in an fMRI study observe patients who had a specialneurological condition that made the touch of a Q-tip painful for them,although the patient experienced no pain when injected with a needle. In linewith theory, we found that empathy in the situation that was painful onlyfor the patient but not the self (Q-tip) recruited specific areas involved inself-other distinction (dorso-medial prefrontal cortex) and cognitive control(right inferior frontal cortex) (Decety & Meltzoff, in press; Lamm, Meltzoff,& Decety, 2010).

Finally, the developmental work is influencing the field of robotics andmachine learning. Researchers are beginning to design algorithms that enablerobots to learn from observing and imitating the behavior of others. Insteadof programming a fixed set of skills, the robot can be given imitative skills sothat it can learn flexibly from watching humans. The long-term goal is toconstruct socially intelligent robots that imitate the way a baby does - map­ping equivalent body parts and actions between self and other (e.g., Demiris& Hayes, 2002; Demiris & Meltzoff, 2008; Nehaniv & Dautenhahn, 2007;Rao, Shon, & Meltzoff, 2007; Shon, Storz, Meltzoff, & Rao, 2007).

Conclusions

Some of the most interesting advances in the next decade will come fromcombining psychological studies of children and social neuroscience. This willallow us to investigate the neural mechanisms underlying human imitation,empathy, gaze following, and intersubjectivity. The goal will be to crack oneof the most urgent and ancient cries for human meaning: Do others feel whatI am feeling? Is there anybody out there like me?

Fodor is correct that solipsism and blank-slate empiricism are too impover­ished a starting state. He is correct that there is no known learning mechanism

Action representation and theory o/mind 29

that can generate the observed richness of adult social cognition from suchimpoverished beginnings. It is not the poverty of the stimulus but the povertyof the proposed starting state that makes these theories a dead end. However,this does not mean that adult theory of mind is present at birth or maturesindependently of social experience. The evidence from modern developmentalscience suggests a third way: Nature designed a baby with an imitative brain.Culture immerses the child in social play with psychological agents whomthe baby can imitate and whose contingent responding and communicativeactions are recognized to be like the child's own. The adult understanding ofmind and empathy for others is the developmental outcome.

We have begun to specify the initial state of social cognition; now we needmore work on the mechanisms of change. The training study by Meltzoffand Brooks (2008) shows, in an experimental fashion, that we can inducedevelopment of social understanding by delivering the right experience atthe right time. The over-arching goal is a new science of learning (Meltzoff etaI., 2009) that puts special emphasis on humans' remarkable gift for sociallearning - learning about me by watching you and interpreting your behaviorvia my own self-experience. Without this bidirectional bridge, our species­specific commonsense psychology would not be so common at all.

Acknowledgments

Support for the preparation of this chapter was provided by the NationalScience Foundation, LIFE Science of Learning Center (SBE-0354453) andthe National Institutes for Health (HD-225 14). The views expressed in thischapter are the author's and do not necessarily represent those of NSF orNIH. I thank Calle Fisher and Craig Harris for their assistance.

References

Astington, 1. w., & Gopnik, A. (1991). Theoretical explanations of children's under­standing of the mind. British Journal ofDevelopmental Psychology, 9, 7-31.

Baron-Cohen, S. (1995). Mindblindness: An essay on autism and theory of mind.Cambridge, MA: MIT Press.

Brandone, A. c., & Wellman, H. M. (2009). You can't always get what you want:Infants understand failed goal-directed actions. Psychological Science, 20, 85-91.

Brooks, R., & Meltzoff, A. N. (2002). The importance of eyes: How infants interpretadult looking behavior. Developmental Psychology, 38, 958-966.

Brooks, R., & Meltzoff, A. N. (2005). The development of gaze following and itsrelation to language. Developmental Science, 8, 535-543.

Butler, S. c., Caron, A. 1., & Brooks, R. (2000). Infant understanding of the referentialnature of looking. Journal of Cognition and Development, 1,359-377.

Butterworth, G., & Jarrett, N. (1991). What minds have in common is space:Spatial mechanisms serving joint visual attention in infancy. British Journal ofDevelopmental Psychology, 9,55-72.

Campbell,1. (1994). Past, space. and self. Cambridge, MA: MIT Press.

30 MeltzofJ

Carpenter, M., Akhtar, N., & Tomasello, M. (1998). 14 through 18 month oldinfants differentially imitate intentional and accidental actions. Infant Behavior andDevelopment, 21,315-330.

Dapretto, M., Davies, M. S., Pfeifer, 1. H., Scott, A. A., Sigman, M., Bookheimer,S. Y, et al. (2006). Understanding emotions in others: Mirror neuron dysfunctionin children with autism spectrum disorders. Nature Neuroscience, 9, 28-30.

Dawson, G., Meltzoff, A. N., Osterling, 1., & Rinaldi, 1. (1998). Neuropsychologicalcorrelates of early symptoms of autism. Child Development, 69,1276-1285.

Decety, 1., Chaminade, T., Grezes, 1., & Meltzoff, A. N. (2002). Rapid communication:A PET exploration of the neural mechanisms involved in reciprocal imitation.Neurolmage, 15, 265-272.

Decety, 1., & Meltzoff, A. N. (iri press). Empathy, imitation, and the social brain. In A.Copland & P. Goldie (Eds.), Empathy: Philosophical and psychological perspectives.Oxford: Oxford University Press.

Demiris, Y, & Hayes, 0. (2002). Imitation as a dual-route process featuring pre­dictive and learning components: A biologically plausible computationai model.In K. Dautenhahn & C. L. Nehaniv (Eds.), Imitation in animals and artifacts(pp. 327-361). Cambridge, MA: MIT Press.

Demiris, Y, & Meltzoff, A. N. (2008). The robot in the crib: A developmentalanalysis of imitation skills in infants and robots. Infant and Child Development,17,43-53.

Dunphy-Lelii, S., & Wellman, H. M. (2004). Infants' understanding of occlusionof others' line-of-sight: Implications for an emerging theory of mind. EuropeanJournal ofDevelopmental Psychology, 1, 49-66.

Flavell, 1. H. (1999). Cognitive development: Children's knowledge about the mind.Annual Review of Psychology, 50,21--45.

Flom, R., Lee, K., & Muir, D. (Eds.). (2007). Gaze following: Its development andsignificance. Mahwah, NJ: Lawrence Erlbaum Associates, Inc.

Fodor, 1. A. (1983). The modularity ofmind: An essay onfaculty psychology. Cambridge,MA: MIT Press.

Fodor, 1. A. (1987). Psychosemantics: The problem of meaning in the philosophy ofmind. Cambridge, MA: MIT Press.

Gallagher, S. (2001). The practice of mind: Theory, simulation or primary interaction?Journal of Consciousness Studies, 8, 83-108.

Gergely, G., Bekkering, H., & Kiraly, 1. (2002). Rational imitation in preverbalinfants. Nature, 415, 755.

Gleissner, B., Meltzoff, A. N., & Bekkering, H. (2000). Children's coding of humanaction: Cognitive factors influencing imitation in 3-year-olds. DevelopmentalScience, 3, 405--414.

Goldman, A. 1. (2005). Imitation, mind reading, and simulation. In S. Hurley & N.Chater (Eds.), Perspectives on imitation: From neuroscience to social science: Vol. 2.Imitation, human development, and culture (pp. 79-93). Cambridge, MA: MIT Press.

Gopnik, A., & Meltzoff, A. N. (1997). Words, thoughts, and theories. Cambridge,MA: MIT Press.

Gordon, R. M. (2005). Intentional agents like myself. In S. Hurley & N. Chater (Eds.),Perspectives on imitation: From neuroscience to social science: Vol. 2. Imitation,human development, and culture (pp. 95-106). Cambridge, MA: MIT Press.

Hamlin, 1. K., Hallinan, E. Y., & Woodward, A. L. (2008). Do as I do: 7-month-oldinfants selectively reproduce others' goals. Developmental Science, 11,487--494.

Action representation and theory ofmind 31

Hari, R., & Kujala, M. V. (2009). Brain basis of human social interaction: Fromconcepts to brain imaging. Physiological Review, 89, 453-479.

Hobson, R. P., & Meyer, 1. A. (2005). Foundations for self and other: A study inautism. Developmental Science, 8, 481-491.

Hommel, B., Miisseler, 1., Aschersleben, G., & Prinz, W. (200 I). The theory of eventcoding (TEC): A framework for perception and action planning. Behavioral andBrain Sciences, 24, 849-878.

Iacoboni, M., Woods, R. P., Brass, M., Bekkering, H., Mazziotta, 1. C, & Rizzolatti,G. (1999). Cortical mechanisms of human imitation. Science, 286,2526-2528.

Jackson, P. L., Brunet, E., Meltzoff, A. N., & Decety, 1. (2006). Empathy examinedthrough the neural mechanisms involved in imagining how I feel versus how youfeel pain. Neuropsychologia, 44, 752-761.

Jackson, P. L., Meltzoff, A. N., & Decety, 1. (2006). Neural circuits involved inimitation and perspective-taking. Neurolmage, 31,429-439.

Lamm, C, Meltzoff, A. N., & Decety, 1. (2010). How do we empathize with someonewho is not like us? A functional magnetic resonance imaging study. Journal ofCognitive Neuroscience, 22,362-376.

Lamm, C, Nusbaum, H., Meltzoff, A. N., & Decety, 1. (2007). What are you feeling?Using functional magnetic resonance imaging to assess the modulation of sensoryand affective responses during empathy for pain. PLoS ONE, 2(12), e1292.

Mahler, M. S., Pine, E, & Bergman, A. (1975). The psychological birth of the humaninfant. New York: Basic Books.

Marshall, P. 1., Young, T., & Meltzoff, A. N. (20 I0). Neural correlates ofaction obser­vation and execution in 14-month-old infants: An event-related EEG desynchroniza­tion study. Poster presented at the International Congress on Infant Studies,Baltimore, MD, USA.

Meltzoff, A. N. (1995). Understanding the intentions of others: Re-enactment ofintended acts by 18-month-old children. Developmental Psychology, 31, 838-850.

Meltzoff, A. N. (1999). Origins of theory of mind, cognition and communication.Journal of Communication Disorders, 32, 251-269.

Meltzoff, A. N. (2007a). "Like me": A foundation for social cognition. DevelopmentalScience, 10,126-134.

Meltzoff, A. N. (2007b). The "like me" framework for recognizing and becoming anintentional agent. Acta Psychologica, 124, 26-43.

Meltzoff, A. N., & Brooks, R. (2008). Self-experience as a mechanism for learningabout others: A training study in social cognition. Developmental Psychology, 44,1257-1265.

Meltzoff, A. N., & Decety, 1. (2003). What imitation tells us about social cognition:A rapprochement between developmental psychology and cognitive neuroscience.Philosophical Transactions of the Royal Society of London. Series B, BiologicalSciences, 358, 491-500.

Meltzoff, A. N., & Gopnik, A. (1993). The role of imitation in understanding personsand developing a theory of mind. In S. Baron-Cohen, H. Tager-Flusberg, & D. 1.Cohen (Eds.), Understanding other minds: Perspectives from autism (pp. 335-366).New York: Oxford University Press.

Meltzoff, A. N., Kuhl, P. K., Movellan, 1., & Sejnowski, T. 1. (2009). Foundations for anew science of learning. Science, 325, 284-288.

Meltzoff, A. N., & Moore, M. K. (1977). Imitation of facial and manual gestures byhuman neonates. Science, 198, 75-78.

32 Meltzof[

Meltzoff, A. N., & Moore, M. K. (1983). Newborn infants imitate adult facial gestures.Child Development, 54, 702-709.

Meltzoff, A. N., & Moore, M. K. (1989). Imitation in newborn infants: Exploringthe range of gestures imitated and the underlying mechanisms. DevelopmentalPsychology, 25,954--962.

Meltzoff, A. N., & Moore, M. K. (1994). Imitation, memory, and the representationof persons. Infant Behavior & Development, 17, 83-99.

Meltzoff, A. N., & Moore, M. K. (1995). Infants' understanding of people and things:From body imitation to folk psychology. In 1. L. Bermudez, A. Marcel, & N. Eilan(Eds.), The body and the seif(pp. 43-69). Cambridge, MA: MIT Press.

Meltzoff, A. N., & Moore, M. K. (1997). Explaining facial imitation: A theoreticalmodel. Early Development and Parenting, 6, 179-192.

Meyer,1. A., & Hobson, R. P (2004). Orientation in relation to self and other: The caseof autism. Interaction Studies, 5, 221-244.

Moll, H., & Meltzoff, A. N. (in press). How does it look? Level 2 perspective-takingat 36 months of age. Child Development.

Moore, C. (1999). Gaze following and the control of attention. In P Rochat (Ed.),Early social cognition: Understanding others in the first months oflife (pp. 241-256).Mahwah, NJ: Lawrence Erlbaum Associates, Inc.

Moore, C. (2006). The development of commonsense psychology. Mahwah, NJ:Lawrence Erlbaum Associates, Inc.

Mundy, P (2009). Lessons learned from autism: An information processing modelof joint attention and social cognition. In D. Cicchetti & M. R. Gunnar (Eds.),Minnesota symposia on child psychology: Meeting the challenge of translationalresearch in child psychology (Vol. 35, pp. 59-113). Hoboken, NJ: Wiley.

Mundy, P, & Newell, L. (2007). Attention, joint attention and social cognition.Current Directions in Psychological Science, 16, 269-274.

Nadel, 1. (2006). Does imitation matter to children with autism? In S. 1. Rogers &1. H. 0. Williams (Eds.), Imitation and the social mind (pp. 118-137). New York:Guilford Press.

Nehaniv, C. L., & Dautenhahn, K. (Eds.). (2007). Imitation and social learning inrobots, humans, and animals: Behavioural, social and communicative dimensions.Cambridge: Cambridge University Press.

Perner,1. (1991). Understanding the representational mind. Cambridge, MA: MIT Press.Piaget, 1. (1954). The construction of reality in the child, trans. M. Cook. New York:

Basic Books.Rao, R. P N., Shon, A. P, & Meltzoff, A. N. (2007). A Bayesian model of imitation in

infants and robots. In C. L. Nehaniv & K. Dautenhahn (Eds.), Imitation and sociallearning in robots, humans, and animals: Behavioural, social and communicativedimensions (pp. 217-247). Cambridge, UK: Cambridge University Press.

Repacholi, B. M., & Meltzoff, A. N. (2007). Emotional eavesdropping: Infants select­ively respond to indirect emotional signals. Child Development, 78, 503-521.

Repacholi, B. M., Meltzoff, A. N., & Olsen, B. (2008). Infants' understanding of thelink between visual perception and emotion: "If she can't see me doing it, shewon't get angry." Developmental Psychology, 44,561-574.

Rizzolatti, G., Fadiga, L., Fogassi, L., & Gallese, V. (2002). From mirror neurons toimitation, facts, and speculations. In A. N. Meltzoff& W. Prinz (Eds.), The imitativemind: Development, evolution, and brain bases (pp. 247-266). Cambridge, UK:Cambridge University Press.

Action representation and theory ofmind 33

Rizzolatti, G., Fogassi, L., & Gallese, V. (2001). Neurophysiological mechanismsunderlying the understanding and imitation of action. Nature Reviews Neuro­science, 2,661-670.

Rogers, S. 1. (2006). Studies of imitation in early infancy: Findings and theories.In S. 1. Rogers & 1. H. G. Williams (Eds.), Imitation and the social mind: Autism andtypical development (pp. 3-26). New York: Guilford Press.

Ryle, G. (1949). The concept ofmind. London: Hutchinson.Shon, A. P, Storz, 1. 1., Meltzoff, A. N., & Rao, R. P N. (2007). A cognitive model of

imitative development in humans and machines. International Journal ofHumanoidRobotics, 4, 387-406.

Smith, A. (1976). The theory ofmoral sentiments. New York: Oxford University Press.(Original work published in 1759.)

Sommerville, 1. A., & Woodward, A. L. (2005). Pulling out the intentional structure ofaction: The relation between action processing and action production in infancy.Cognition, 95, 1-30.

Toth, K., Munson, 1., Meltzoff, A. N., & Dawson, G. (2006). Early predictors ofcommunication development in young children with autism spectrum disorder:Joint attention, imitation, and toy play. Journal of Autism and DevelopmentalDisorders, 36, 993-1005.

Wellman, H. M. (1990). The child's theory ofmind. Cambridge, MA: MIT Press.Williamson, B. A., Jaswal, V. K., & Meltzoff, A. N. (2010). Learning the rules:

Observation and imitation of a sorting strategy by 36-month-old children.Developmental Psychology, 46, 57-65.

Williamson, R. A., Meltzoff, A. N., & Markman, E. M. (2008). Prior experiences andperceived efficacy influence 3-year-olds' imitation. Developmental Psychology,44,275-285.