Role Reversal Imitation and Language in Typically Developing Infants and Children With Autism Malinda Carpenter, Michael Tomasello, and Tricia Striano Max Planck Institute for Evolutionary Anthropology Leipzig, Germany Three types of role reversal imitation were investigated in typically developing 12- and 18-month-old infants and in children with autism and other developmental de- lays. Many typically developing infants at both ages engaged in each of the 2 types of dyadic, body-oriented role reversal imitation: self���self reversals, in which the adult acted on herself and the child then acted on himself, and other���other reversals, in which the adult acted on the child and the child then acted back on the adult. However, 12-month-olds had more difficulty than 18-month-olds with triadic, object-mediated role reversals involving interactions around objects. There was little evidence of any type of role reversal imitation in children with autism. Positive rela- tions were found between role reversal imitation and various measures of language development for 18-month-olds and children with autism. Children learn many of their most important skills by observing the behavior of other people. However, recent theoretical analyses and empirical research have re- vealed a wide variety of different social learning processes that might be involved, depending on how the process of observation interacts with children���s broader skills of cognition and social cognition���especially their ability to read the inten- tions of others. Recent reviews have been provided by Want and Harris (2002), Call and Carpenter (2002), and Tomasello and Carpenter (2005). Briefly stated, children are flexible imitators: Instead of copying surface behaviors literally, they do what others are trying to do, not what they actually do. For example, between 12 and 15 months, infants begin to reproduce the action they see the adult intending (trying) to perform, not the bodily movements she is actually performing INFANCY, 8(3), 253���278 Copyright �� 2005, Lawrence Erlbaum Associates, Inc. Correspondence should be addressed to Malinda Carpenter, Max Planck Institute for Evolutionary Anthropology, Deutscher Platz 6, D-04103 Leipzig, Germany. E-mail: carpenter@eva.mpg.de

(Bellagamba & Tomasello, 1999 Johnson, Booth, & O���Hearn, 2001 Meltzoff, 1995). They reproduce the actions adults produce intentionally, not the ones adults produce by accident (Carpenter, Akhtar, & Tomasello, 1998). They do not repro- duce an adult���s unusual action when they see a reason for the adult���s behavior that does not currently apply to them, whereas they do reproduce it if they see no exten- uating circumstances that justify the adult���s odd behavior (Gergely, Bekkering, & Kir��ly, 2002). They produce different behaviors in attempting to imitate an adult depending on what they perceive as the adult���s goal, for example, touching an ob- ject versus moving his or her arm in a certain way (Bekkering, Wohlschl��ger, & Gattis, 2000 Carpenter, Call, & Tomasello, 2005 see also Carpenter, Call, & Tomasello, 2002, for a related finding). These studies all involve others��� intentions toward objects. Yet many of the most interesting and important skills children develop involve a different type of intentions: intentions toward each other. For example, learning language and par- ticipating in other collaborative activities requires an understanding of others��� in- tentions toward the self, and an ability to coordinate intentions with others (Tomasello, Carpenter, Call, Behne, & Moll, 2005). There is much evidence that 18-month-old and older children can learn to use pieces of language by discerning what adults are intending to do with them in a wide variety of communicative cir- cumstances, including cases in which the adult has failed attempts and accidents, so the imitative learning of the novel word is not so straightforward (Tomasello & Barton, 1994 Tomasello, Strosberg, & Akhtar, 1996 see Tomasello, 2001, for a review). In this study we asked if younger children also use the same flexibility in imitation involving intentions toward the self. Specifically, we explored what Tomasello (1999) called role reversal imitation that is, when the child learns to perform an action toward an adult in the same way that the adult performed it toward him or her. For example, an adult demonstrator might slap his own knee, to which the child might respond by slapping her own knee (self���self role reversal). Alternatively, an adult might pat the child on her head, and the child might reciprocate by patting the adult on his head (other���other role reversal). Finally, a more triadic, cooperative type of role reversal is one that is mediated by objects, as when on one occasion the adult holds out a basket so the child can place toys into it, and on a later occasion the child then holds out the bas- ket to the adult so he can put toys into it, thus reversing roles in this activity. In all of these cases, the child discerns that the adult intends for her to do to or for him what he has done to or for her. What distinguishes this type of imitation from other forms of imitation is that it is performed on an object that is different from the one the demonstrator performed it on���but not just any object. In straightforward imitation the adult might place a pencil into a cup and the child might follow suit by doing the exact same thing with the exact same objects���or perhaps do something similar with similar yet different objects. However, in role reversal imitation the child does to other people what 254 CARPENTER, TOMASELLO, STRIANO

they have previously done to her (or else does to herself what others have done to themselves). There is thus a reciprocal substitution between demonstrator and learner. It is important that the child could theoretically mistake the demonstrator���s intentions and so she might follow his patting of his own head by patting his head as well, or when the demonstrator pats the child���s head she could follow suit by patting her own head. The most well-known example of this ���error��� occurs in lan- guage acquisition when children sometimes make I���you substitutions, saying such things as ���Pick you up��� when they mean ���Pick me up��� that is, on hearing the adult say such things as ���Do you want me to pick you up?��� they fail to understand that I and you are referenced to the current speaker and listener in a way that most other words are not (Dale & Crain-Thoreson, 1993). There are no systematic investigations of children���s ability to engage in role re- versal imitation. However, in preferential looking studies, typically developing in- fants are able to recognize changes in others��� roles by around 9 months of age (Rochat, Striano, & Morgan, 2004). In their natural social interactions, infants be- gin to take the adult���s role in peek-a-boo and other such games on a regular basis at around 14 months (Ratner & Bruner, 1978), with dramatic increases in the fre- quency of this behavior between 9 and 18 months (Ross & Lollis, 1987). It is no ac- cident that this is the age at which children are first beginning to learn linguistic symbols, as the same type of reciprocal imitation applies to the learning of a piece of language. Thus, to learn to use a symbol like an adult, children must learn to use it toward the adult in the same way the adult used it toward them (Tomasello, 1999). They must be able to engage in other���other role reversals mediated by a conventional artifact. There is a special population of children who might have difficulties with role reversals: children with autism. We are aware of one previous study of something like self���self role reversals in children with autism. Peeters, Grobben, Hendrickx, Van den Eede, and Verlinden (2003) asked typically developing 3- to 8-year-old children, 5- to 9-year-old children with autism, and adults to play a series of games in which participants could respond either with a same or a different location re- sponse. For example, a red cloth and a blue cloth were lying on the floor. The par- ticipant chose a red or blue card and the experimenter took the other card. The ex- perimenter crawled under the cloth of the same color as her card and then asked the participant to ���do the same.��� In most of the games, all but the adult participants chose the opposite object as the adult, that is, they did not choose the same location response. However, in these games, children could have produced this response with a simple understanding of ���this object is hers and this object is mine,��� without engaging in self���self role reversal. In the single game that might have required self���self role reversal���when the experimenter showed the participant a photo- graph of herself and a photograph of the participant, tapped the photograph of her- self, and asked the participant to ���do the same������children with autism chose ran- domly (although note that all but the youngest group of typically developing ROLE REVERSAL IMITATION 255

children also chose randomly, and the youngest children chose the experimenter���s photograph). There are also suggestions in the literature that children with autism might have difficulties with self���self role reversals in other contexts. These children show pro- nounced deficits in gesture imitation, which requires self���self role reversals (and they might show somewhat less impairment on tasks involving imitating others��� actions on objects, which do not require role reversal because the child acts on the same object as the demonstrator see, e.g., Rogers, Cook, & Meryl, 2003, for a re- view). When they do copy gestures, children with autism often show an interesting pattern of errors that further indicates difficulties with role reversals. Ohta (1987), Whiten and Brown (1998), Hobson and Lee (1999), and Smith and Bryson (1998) all reported that children with autism sometimes reproduced gestures exactly as they saw them, without switching perspectives. For example, when the modeled action was ���waving with the open palm facing the subject,��� some children with au- tism waved with their own palm facing themselves, as opposed to with their palm facing the experimenter. We are unaware of any tests of other���other role reversal imitation in children with autism, but some characteristics of these children���s language use would sug- gest that they might have difficulty with this type of role reversal as well. The ten- dency of children with autism to engage in echolalia (mimicry of others��� speech) and to make personal pronoun (e.g., Jordan, 1989) and question���statement (Tager-Flusberg, 1993) reversal errors are well-known, and they might be a reflec- tion of these children���s ability to copy surface behavior but not to engage in imita- tive learning that takes into account the demonstrator���s intentions toward them (Carpenter & Tomasello, 2000). In this study we report initial explorations of all three types of role reversal imita- tion: the dyadic self���self and other���other role reversals, and the triadic, ob- ject-mediated role reversals. We do this for typically developing children of 12 and 18 months of age, as well as for a small group of children with autism and, as a com- parison,childrenwithotherdevelopmentaldisabilities.Wealsoinvestigatedinterre- lationsbetweenchildren���srolereversalimitationandtheirvocabularydevelopment. We predicted that children with autism (and perhaps 12-month-olds) would have difficulty with role reversal imitation, especially with other���other role reversals and triadic,object-mediatedrolereversals.Wealsopredictedpositiverelationsbetween role reversal imitation and language for all children, in particular between other���otherrolereversalimitationandcomprehensionandproductionofpronouns. STUDY 1 In this study we investigated dyadic self���self and other���other role reversal imita- tion in typically developing 12- and 18-month-old children. Each type of task had 256 CARPENTER, TOMASELLO, STRIANO

the same structure: An adult performed a simple action on either herself or the child, then gave the child a chance to perform the action on the person of her choice. Also included was a control task in which the adult performed the action on a third person, to investigate infants��� baseline rates of reproducing the actions on themselves or the first adult. In half of each type of task, the adult performed the action with an object and in half she simply performed an action directly on some body part (note that the object in this case was mainly a motivational tool and was not intended to make the interaction triadic). We chose simple, familiar actions be- cause our focus was on role reversals and not imitative learning of novel actions. Method Participants. Participants were 48 children of various ethnicities from a large metropolitan area in the United States. Half of the children were 12 months old (M = 11.7 months), and the other half were 18 months old (M = 17.7 months). There were 12 boys and 12 girls at each age. Three additional children were ex- cluded from the study: 1 due to parental interference, 1 due to lack of participation, and 1 because of equipment failure. Children were recruited from a psychology de- partment database comprised of parents and children who volunteered in various ways. Materials. In half of the imitation tasks, the experimenter acted on either self, child, or the other adult directly (e.g., tickling), and in the other half she used an ob- ject (e.g., rolling a car along the child���s leg). Materials for the object tasks included a red foam ball, an orange plastic ring, a small yellow car that looked like a dog, and a foam block with pictures of animals on each side. Parents of infants at both ages also completed (at the end of the session) the MacArthur Child Development Inventory of Words and Gestures (long form Fenson et al., 1993), measuring comprehension and production of 396 words. This checklist includes a section of 11 pronouns (e.g., I, mine, you, this). Procedure. Testing took place in a small university playroom. The child, a parent, and two female experimenters (E1 and E2) were in the testing room during the experiment. The entire session was taped by a video camera from the corner of the room. Following a warm-up ���normal��� imitation task (dropping a figurine down a chute), the parent was asked to sit in a chair or on the floor in the corner of the room opposite the video camera. The child could choose whether to sit on the parent���s lap or stand or sit in front of the parent���s knees. E1 sat cross-legged on the floor to the immediate right of the child and E2 sat in a similar fashion to the immediate left of the child. The Es were both close to the child (approximately 5���13 cm away) and were equidistant from the child during all the role reversal tests. ROLE REVERSAL IMITATION 257