3.1. Prelinguistic communication as joint action
Clark (1996) has pointed out that language is a type of joint action. We would argue that prelinguistic communication can be considered joint action as well, at least by 12–14 months of age. The best evidence comes from infants’ production and comprehension of the pointing gesture. Around 12 months, infants begin pointing “declaratively,” to share attention and interest about objects with a social partner. There are two recent findings that highlight the joint nature of this activity. First, in response to claims that 12-month-olds do not point to share attention and interest but rather for more egocentric reasons, simply to gain rewarding positive emotions to the self (Moore & D’Entremont, 2001), Liszkowski, Carpenter, Henning, Striano, and Tomasello (2004) manipulated an experimenter’s reaction to 12-month-olds' declarative points to interesting objects. The adult reacted either by engaging in joint attention with infants about the object, by just responding with positive emotion to infants (ignoring the object), by just looking at the object, or by ignoring the point. Infants’ patterns of repeated pointing within trials vs. continued pointing across trials indicated that they were only satisfied when the adult reacted in joint attention. Their point was an invitation to a joint action: to share attention to this object. A second study suggests that 12-month-olds can help their partner achieve this goal, once she has taken it up. If the adult is trying to share attention but misunderstands the referent of the infant’s point and “shares” to a different object nearby, infants help her by repeating their point (Liszkowski, Carpenter, & Tomasello, 2007).
Infants’ understanding of joint action is also reflected in their comprehension of others’ pointing gestures. For example, Liebal et al. (2009) had 14-month-olds participate in a “clean-up” game with an adult, E1. After throwing a series of objects into a basket together, E1 pointed to another, target object and simply said, “There!” Infants picked up that object and threw it into the basket as well, apparently seeing E1’s point as related to their joint activity. Infants’ responses in a control condition ruled out the possibility that infants were responding egocentrically, based on what they themselves were doing, instead of what they were doing jointly with E1. In this condition infants participated in the cleaning-up activity with E1, exactly as before, but then another adult, E2, pointed at the target object instead. Infants rarely cleaned up the target object in this condition.
Some of the earliest evidence of joint action may thus come from prelinguistic communication (see Tomasello, Carpenter, & Liszkowski, 2007, for a fuller discussion of shared intentionality in infants’ comprehension and production of pointing gestures). But, more prototypically, infants around this age are also beginning to engage in more instrumental collaborations involving problem-solving and social games with objects.
3.2. Instrumental collaboration
Warneken, Chen, and Tomasello (2006) presented 18- and 24-month-old children with four tasks in which collaboration was needed in order to achieve some goal. Two of the tasks were problem-solving tasks (involving retrieving an object from an apparatus) and two were social games (e.g., jointly bouncing a block on a small trampoline). In order to test how rich infants’ participation in the joint activities was, Warneken et al. had the adult partner suddenly stop playing his role in the middle of the activity, and they coded communicative attempts by children to reengage the adult in the activity. The reasoning was that if children had formed a joint goal with the adult, and understood the commitment this entailed, then they should try to persuade the adult to recommit to the joint goal when he stopped. Warneken et al. also coded how well children coordinated actions with the adult before and after the interruption periods, during the joint activities. They found that children at both ages succeeded in coordinating with the adult in at least some of the tasks, although 24-month-olds did this more skillfully than 18-month-olds. In some cases children were also able to reverse roles in the activity. Importantly, during the interruption periods, all children at both ages actively attempted to reengage the adult communicatively at least once, for example by pointing to the apparatus or pushing it toward the adult.
There are two important things to note here. First, in about half the trials, children’s predominate response was either to wait for the adult or to try to reengage him—instead of disengaging from the task or attempting to perform it individually. This is consistent with the idea that since the adult was committed to the shared goal, children had a right to expect that he would continue playing, and thus either waited for him to do so or else communicated with him in some manner (either to request his continued participation or to help him perform his role). Second, Warneken and colleagues also noted two other findings that speak to the motivation children had to cooperate: (a) children participated enthusiastically in the social games, in which there was no material reward; and (b) once they had successfully solved the problem-solving tasks (i.e., retrieved the object from the apparatus), almost all the children replaced the object at some point and repeated the task. The collaborative activity was thus an end in itself, not just a means to retrieving some material reward.
Warneken and Tomasello (2007) subsequently tested 14-month-old infants on two of the same tasks and found that they, too, showed some evidence both of coordination of actions and of reengagement attempts during the interruption periods (see also, e.g., Brownell & Carriger, 1990; Eckerman & Didow, 1989; Hay, 1979; and Ross & Lollis, 1987, for other studies of collaboration in infants).
These studies leave open several questions, however. First, although the findings of Warneken et al. (2006) are consistent with the idea that children were engaging in truly joint action and understood the mutual commitments this entailed, their understanding of commitments was not directly tested. And second, because all the activities in this study required two players, it is possible that children reengaged the adult simply as a means to achieving an individual goal, not a shared goal. That is, children could simply have wanted to achieve the effect (e.g., retrieving the object or seeing the block bounce) and needed the adult as a sort of “social tool” to make this happen. In the next series of studies we addressed both of these issues, to determine the extent to which young children really do participate in collaborative activities with shared goals and joint commitments.
Gräfenhain, Behne, Carpenter, and Tomasello (in press) tested slightly older children, 2- and 3-year-olds, on a set of games that could be played either jointly or individually. For example, in one game, each player could use a tool to press one of two levers to make two toy rabbits hop up inside a box (or else a single player could press the levers individually). After the experimenter and an assistant demonstrated the games to children both jointly and individually, children were given the chance to play too, in one of two ways. In the commitment condition, the experimenter invited children to play the game with her (and waited for them to agree to this), then played in a joint manner, attending to children and playing contingently with them. In the no-commitment condition, in contrast, children were told by the assistant that they could play the game; then as soon as they started playing the experimenter approached and played in parallel with them on the same apparatus. After this brief play period, in both conditions there was an interruption period, as in the Warneken et al. (2006) study. We expected that in the commitment condition—but not in the no commitment condition—children’s behavior during this interruption period would suggest that they knew they had a right to expect the experimenter’s continued participation.
That is exactly what we found with the 3-year-olds. In the commitment condition, most children either waited for the experimenter or attempted to reengage her into the game communicatively, whereas in the no commitment condition children mostly played the game alone. The fact that children waited for or attempted to reengage the adult in the commitment condition, even in these types of games, when they could just have easily played the games alone, suggests that they were not just using the adult as a social tool to achieve their own individual ends—that instead they saw the game as a joint, committed activity.
The 2-year-olds also often waited for or attempted to reengage the adult but, unexpectedly, they did this in both conditions. It is unclear what to conclude from these results. If children had no understanding of commitments, they might be expected not to wait for or reengage the adult in either condition. The finding that 2-year-olds did wait for or reengage the adult is consistent with the idea that they understood the adult’s commitment to the activity; the finding that they did this in both conditions suggests that they may have seen the parallel play in the no commitment condition as a joint activity too. It is possible that young children over-attribute shared goals and joint action: as long as their partner is sitting next to them and playing with the same object they are playing with, they see this as a joint action. On the other hand, it could be that 2-year-olds do not yet understand commitments and responded as they did in the no commitment condition simply because they preferred acting on the apparatus with another person, even if that person was only playing in parallel to them.
The 3-year-olds in the Gräfenhain et al. (in press) study did show the expected difference between conditions. However, because it was not clear on exactly what basis they did so—the commitment (based on the invitation and agreement) or the type of play (contingent vs. not)—Gräfenhain et al. (in press) conducted a second study in which we made the type of play constant across both conditions and only manipulated how the games were initiated. We also looked at a different aspect of commitments and obligations: leave-taking. Instead of measuring children’s reactions when the adult partner stopped playing for no apparent reason, in this study we measured how children themselves took leave of a committed joint activity when they were done participating.
In the commitment condition of this study, 3- and 4-year-olds were encouraged to invite the experimenter to play a game with them. Most children did this, and the experimenter accepted their invitation and began to play (basically in parallel to children, but with a verbal reminder halfway through that they were playing together). In the no-commitment condition, the experimenter simply announced that she would like to play and started playing in parallel to children. After 30 s of play, the response period began: across the room, an assistant began playing another, highly attractive game by herself. She gradually attempted to entice children to come play with her, first by simply playing loudly on her own, and eventually by asking if they wanted to play too. The main measure of interest was whether upon leaving the first game for the second one, children spontaneously would acknowledge their leaving to the experimenter, their partner in the first game, for example, by giving her the tool they had used in the game or telling her that they wanted to leave.
We found that both 3- and 4-year-olds acknowledged their leaving to the experimenter significantly more often in the commitment condition than in the no-commitment condition. Since in this study the only difference between conditions was the commitment, as embodied by the initial invitation to play and the acceptance of this invitation (the play itself was identical), this suggests that by 3 years of age children are both sensitive to whether they are in a committed joint activity and also are beginning to know what obligations such committed activities engender. The results of this last study in particular are not easily explainable in terms of children’s own individual goals in the activity. Children were not trying to get the adult to do anything; we were measuring how they ended an already successful joint action—whether they made the (technically unnecessary) effort to “close” their joint action with their partner or not.
By 12–18 months, infants are beginning to participate in a variety of joint actions which show many of the characteristics of adult joint action. They can coordinate actions and support their partners in joint activities, for example, by repairing miscommunications, waiting for the partner, and encouraging her to reengage. Whether children younger than 3 years of age understand joint commitments is unclear, but by 3 years, there is evidence that children are even beginning to feel some of the commitments and obligations inherent in joint action themselves, as when they excuse themselves when they wish to leave a joint action.