REPORTS nenta t for helnine from parents However n if they had received ne pne n any ctive the facts hat human parents urage thei Out of reach Physical obstacle Wron result Wrona m eans by helpin with ials,the mean percen ehavior per total number of trias wa p A number of theorists have claimed tha .30. 0.00 cially pon-kin)in the Flap task with only trial per individua l,we com ted Fisher's exact test (=4.P not found in othe cies (26-28).This er's exact tests (N 24),P 1.0,0.48, ieanur sho of testing and had and with mor life by humans.Each especially salient they still did that the co on an object,the goal is un- a daily basis ame on ativenes SK s S5 to S8) A出 ast to Re instar help.but the the oher'do tal,control)were as follows:Alex,5.0; These experimental results demonstrate in. P.Nunn.Int L Behar Dev 9.265 3 Annet,(each als)in Eds (.Ne neg the current study unlike thos in21,22),w helped.the but a human. Press,New Yo with cach of the extre 24.2. m40 sh 10.0.g anything.As wi the hum n in nts,the did and mo A handing humans ther children). each objects for hu is.in those involving physical tacles,wrong s clearly not kin).Of special note,they helped L Cogn.Dev in pre al tasks of thesc zees bened in on This vard Oniv. types- -designed to make the human's problem due to a greater propensity to help in children,or 1302 3 MARCH 2006 VOL 311 SCIENCE www.sciencemag.org 54 months of age at the time of testing and had been raised their whole life by humans. Each chimpanzee performed both conditions of each task in two different sessions on consecutive days. They were tested by a highly familiar human caretaker with whom they spent time on a daily basis. The chimpanzees helped in some of the tasks (movies S5 to S8). All three chimpanzees helped reliably in the five tasks involving reaching: Across all such trials, the chimpanzees could retrieve objects for the human from 0 to 13 times in both the experimental and control conditions. The scores of the three individuals (experimen￾tal, control) were as follows: Alex, 5, 0; Alexandra, 10, 3; Annet, 9, 0 (each pair is sig￾nificantly different from a chance distribution: Fisher_s exact test, P 0 0.039; P 0 0.017, P 0 0.0005, respectively). Because it was more difficult to control the behavior of the chim￾panzees than that of the children, the human had to call each one by name to pay attention more often and sooner in the process. None￾theless, when the chimpanzees helped, they did so relatively quickly (average latency 0 12.9 s of reaching for the object), with each of the three individuals helping the human from 4 to 7 times across all tasks before she verbalized anything. As with the human infants, they did so without receiving any benefit (reward or praise) for helping (although they retained the object in their possession for some seconds before handing it over more often than did the children). However, the chimpanzees did not help the human reliably in the other types of tasks—that is, in those involving physical obstacles, wrong results, or wrong means. In a follow-up study, we gave them two additional tasks of these types—designed to make the human_s problem especially salient and with more time for a response—and they still did not help in these tasks (14). Presumably, when someone is reaching with an outstretched arm toward an object, the goal is in principle easier to un￾derstand and the kind of intervention follows straightforwardly. This could explain why out￾of-reach tasks (in contrast to the other scenar￾ios) elicited more helping by children and the only instances of helping by chimpanzees. Children and chimpanzees are both willing to help, but they appear to differ in their ability to interpret the other_s need for help in different situations. These experimental results demonstrate in￾strumental helping (toward goals) in a nonhuman primate. It is possible that helping behaviors are more likely when they involve objects that are not food, and that this explains why we obtained positive results when others, using different tasks involving food, have found negative results. It should also be noted that the chimpanzees of the current study, unlike those in (21, 22), were helping not a conspecific but a human. This might be important because chimpanzees are extremely competitive with one another (24, 25), but when they grow up interacting with humans, they seem to develop some more cooperative skills and motivations as well. Although our chimpanzees had been rewarded in the past for handing humans objects already in their posses￾sion upon request, they had not been encour￾aged to retrieve, nor rewarded for retrieving, out-of-reach objects for humans. The human infants helped much more, and they did so for an adult they had just met (who was clearly not kin). Of special note, they helped in four different kinds of situations, whereas the chimpanzees helped in only one. This could be due to a greater propensity to help in children, or to children_s more sophisticated cognitive skills in discerning the goal of the other in a variety of different situations. Infants 18 months of age are too young to have received much verbal en￾couragement for helping from parents. However, even if they had received some prior encourage￾ment, many of the current tasks would have been unfamiliar for them, and the recipient of the help was an unfamiliar adult as well. In any case, viewed from a larger evolutionary perspective, the facts that human parents encourage their children to help others and that children comply by helping (even before they are linguistic) are noteworthy as the teaching and learning of prosocial norms. A number of theorists have claimed that human beings cooperate with one another and help one another (especially non-kin) in ways not found in other animal species (26–28). This is almost certainly so, and the current results demonstrate that even very young children have a natural tendency to help other persons solve their problems, even when the other is a stranger and they receive no benefit at all. However, our nearest primate relatives show some skills and motivations in this direction as well, and this suggests that the common an￾cestor to chimpanzees and humans already pos￾sessed some tendency to help before humans began down their unique path of hypercooper￾ativeness (25, 29). References and Notes 1. D. Bischof-Ko¨hler, in Infant Development: Perspectives from German-Speaking Countries, M. E. Lamb, H. Keller, Eds. (Erlbaum, Hillsdale, NJ, 1991), pp. 245–273. 2. D. Bischof-Ko¨hler, Z. Psychol. Z. Angew. Psychol. 202, 349 (1994). 3. D. Bischof-Ko¨hler, Psychol. Erzieh. Unterr. 47, 142 (2000). 4. J. Dunn, P. Munn, Int. J. Behav. Dev. 9, 265 (1986). 5. N. Eisenberg, R. A. Fabes, in Handbook of Child Psychology: Vol. 3. Social, Emotional, and Personality Development, W. Damon, N. Eisenberg, Eds. (Wiley, New York, ed. 5, 1998), pp. 701–778. 6. J. E. Grusec, M. Davidov, L. Lundell, in Blackwell Handbook of Childhood Social Development, P. K. Smith, C. H. Hart, Eds. (Blackwell, Malden, MA, 2002), pp. 457–474. 7. M. L. Hoffman, Empathy and Moral Development: Implications for Caring and Justice (Cambridge Univ. Press, New York, 2000). 8. C. Zahn-Waxler, M. Radke-Yarrow, R. A. King, Child Dev. 50, 319 (1979). 9. C. Zahn-Waxler, M. Radke-Yarrow, E. Wagner, M. Chapman, Dev. Psychol. 28, 126 (1992). 10. D. B. Johnson, Merrill Palmer Q. 28, 379 (1982). 11. Infants and young children readily participate in typical household chores such as cleaning up (12) and also sometimes provide information for others (4, 13), but in the studies done to date, there have been no control or baseline conditions to determine whether the children are actually helping others with their goals or just engaging in the activity for its own sake, independently of the other actually needing help. 12. H. L. Rheingold, Child Dev. 53, 114 (1982). 13. U. Liszkowski, M. Carpenter, T. Striano, M. Tomasello, J. Cogn. Dev., in press. 14. See supporting material on Science Online. 15. S. M. O’Connell, Primates 36, 397 (1995). 16. F. de Waal, Good Natured (Harvard Univ. Press, Cambridge, MA, 1996). Fig. 1. Mean percentage of target behaviors as a function of task and condition. In tasks with multiple trials, the mean percentage of trials with target behavior per total number of trials was computed for each individual. Independent-sample t tests (df 0 22) revealed significant differences between conditions for the tasks Paperball (t 0 4.30, P G 0.001), Marker (t 0 2.70, P G 0.05), Clothespin (t 0 4.38, P G 0.001), Books (t 0 2.33, P G 0.05), and Cabinet (t 0 3.08, P G 0.01). For the Flap task with only one trial per individual, we computed Fisher’s exact test (N 0 24, P G 0.05). In these six tasks, children performed the target behavior significantly more often in the experimental than in the control condition. No difference between conditions was found for the tasks Clips (t 0 1.04, P 0 0.31), Cap, Chair, and Tool, Fisher’s exact tests (N 0 24), P 0 1.0, 0.48, and 0.22, respectively. Error bars represent SE; *P G 0.05. REPORTS 1302 3 MARCH 2006 VOL 311 SCIENCE www.sciencemag.org