Children, chimpanzees, and bonobos adjust the visibility of their actions for cooperators and competitors

Chimpanzees and bonobos are highly capable of tracking other’s mental states. It has been proposed, however, that in contrast to humans, chimpanzees are only able to do this in competitive interactions but this has rarely been directly tested. Here, pairs of chimpanzees or bonobos (Study 1) and 4-year-old children (Study 2) were presented with two almost identical tasks differing only regarding the social context. In the cooperation condition, players’ interests were matched: they had to make corresponding choices to be mutually rewarded. To facilitate coordination, subjects should thus make their actions visible to their partner whose view was partially occluded. In the competition condition, players’ interests were directly opposed: the partner tried to match the subject’s choice but subjects were only rewarded if they chose differently, so that they benefited from hiding their actions. The apes successfully adapted their decisions to the social context and their performance was markedly better in the cooperation condition. Children also distinguished between the two contexts, but somewhat surprisingly, performed better in the competitive condition. These findings demonstrate experimentally that chimpanzees and bonobos can take into account what others can see in cooperative interactions. Their social-cognitive skills are thus more flexible than previously assumed.

Only two of the four towers were baited (randomly selected on each trial) and the door 860 between the two rooms was open so that subjects had access to the apparatus from both sides. 861 The ropes on the proposer side were moved into reach first, and once subjects had made a 862 choice they moved around and made a choice on the responder side. In the cooperation 863 condition subjects could then collect the rewards on both sides of the apparatus. In the 864 competition they could only collect the rewards on the responder's side. To enter the next 865 phase, subjects had to successfully access the food by choosing corresponding towers in at 866 least 10 out of 12 trials in two consecutive sessions which took them 3.6. sessions on average 867 (range: 2-7). One chimpanzee did not reach criterion within 8 sessions and were excluded from 868 the study. This phase emphasized the interdependence between proposer and responder (i.e. 869 that choices had to be made on both sides). They also learned that choosing the same tower 870 would only lead to rewards being released on both sides in the cooperation condition. 871 872

Barrier experience 873
A barrier spanning the whole apparatus prevented the responder from seeing the 874 proposer's choice (the proposer's visual access remained unobstructed). Through slits in the 875 barrier the responder could still pull the ropes and collapse any of the lower platforms. At the 876 start of each test session subjects received two trials in the responder role followed by two 877 trials in the proposer role. The partner was the same conspecific stooge as in the following test 878 trials. Trials were rigged so that subjects were successful once in each role in both conditions. 879 Hence, in both conditions subjects were rewarded equally often in the presence of a barrier 880 before entering the test. These experience trials familiarized subjects with the barrier and 881 showed them how the barrier obstructed the responder's view of the apparatus. They also 882 highlighted that all four options could still be chosen in the presence of the barrier. Main experiment 901 902 Apparatus familiarization 903 An experimenter (E1) explained to children (C) that the aim of the game was to win as 904 many marbles as possible. The partner was a same-sex puppet played by a second experimenter 905 (E2; the E2 had to be replaced halfway through data collection. Each E2 tested the same 906 number of children of each sex per condition). All towers were baited with red marbles. C were 907 in the proposer and the puppet in the responder role. Depending on the condition, E1 908 emphasized the cooperative or competitive nature of the game ("You can win the marbles 909 together." vs. "Max/Lola can try to steel the marbles."). On the first trial the puppet chose a 910 different tower than C. E1 then opened the sliding doors while saying "round is over." The 911 puppet exclaimed "Oh no, I made a mistake, this time, we won't get the marbles/I won't get 912 the marbles" (cooperation/competition condition, respectively). E1 reset the apparatus and the 913 trial was repeated twice except that now the puppet chose the same tower as C. In all 914 subsequent trials the puppet expressed joy ("Yes, this time we/I win the marbles") or 915 disappointment ("On no, this time we/I won't win any marbles") depending on the outcome of 916 the round. In the cooperation condition the child and the puppet collected their marbles 917 together in a joint container whereas they received separate containers in the competition 918 condition. While C and the puppet made their choices E1 always turned away and pretended to 919 be busy with something else. 920

921
Apparatus understanding 922 C assumed the responder and the puppet the proposer role. If C chose the correct 923 platform (i.e. of the same tower as the puppet) in three out of four trials they entered the next 924 training phase. If not, they were excluded from the experiment. Only one child failed to pass 927

Open-door 928
Two of the four towers were baited (randomly selected on each trial) and C had to 929 operate the apparatus alone by collapsing corresponding platforms on both sides. To enter the 930 final training phase, they had to successfully retrieve the rewards on three of four trials. All 931 children passed without mistakes. 932 933 Barrier experience trials 934 The setup was the same as for the apes. C first received two trials as responder and then 935 two trials as proposer. These experience trials familiarized C with the barrier while 936 highlighting that all four options could still be chosen in the presence of the barrier (pilot data 937 indicated that children often assumed towers covered by the barrier were no longer part of the 938 game). To further emphasize this, E1 pointed to each platform on both sides at the beginning of 939 each training section while saying "You can pull here, here, here, and here." 940 941 942 Follow-up preference tests 943 944 Apparatus understanding 945 All towers were baited and C collapsed one of the platforms and collected the rewards 946 on three trials while the puppet sat passively in the responder position. To ensure that C 947 experienced a similar number of trials as in Study 2, C and the puppet then switched positions 948 and the puppet collect the marbles on three trials. 949 950

Barrier experience 951
A barrier spanning the whole apparatus blocked C's view of the towers (C was in the 952 responder position). All towers were baited and the puppet collected the rewards on two trials. 953 The puppet and C then swapped positions and C collected the rewards on two trials.

Statistical analyses 955 956 a) General information 957
For the main analyses we used Generalized Linear Mixed Models (GLMM) as this allowed us to simultaneously test for multiple predictors (and 958 thus avoid multiple testing issues) while controlling for the influence of several potentially confounding variables. Moreover, since the 959 chimpanzees came from two different social groups this analytic approach allowed us to include the social group as a random effect in order to 960 avoid potential problems related to pseudoreplication. Finally, in Study 1b the GLMM enabled us to test for an interaction between the condition 961 and the experimental phase which would not have been possible with standard non-parametric tests (e.g. Wilcoxon Signed Rank Test). The first 962 GLMM in each study served to detect any condition differences to provide a first indicator for whether they successfully adjusted their choices 963 to the social context. This analysis was followed up by a second GLMM looking at whether subjects made more correct choices in one of the 964 conditions. All analyses were determined prior to data inspection.