Chimpanzees spontaneously take turns in a shared serial ordering task

Social coordination can provide optimal solutions to many kinds of group dilemmas, and non-human subjects have been shown to perform single actions successively or simultaneously with partners to maximize food rewards in a variety of experimental settings. Less attention has been given to showing how animals are able to produce multiple (rather than single) intermixed and co-regulated actions, even though many species’ signal transmissions and social interactions rely on extended bouts of coordinated turn-taking. Here we report on coordination behaviour in three pairs of chimpanzees (mother/offspring dyads) during an experimentally induced turn-taking scenario. Participants were given a “shared” version of a computer-based serial ordering task that they had previously mastered individually. We found that minimal trial-and-error learning was necessary for the participants to solve the new social version of the task, and that information flow was more pronounced from mothers toward offspring than the reverse, mirroring characteristics of social learning in wild chimpanzees. Our experiment introduces a novel paradigm for studying behavioural coordination in non-humans, able to yield insights into the evolution of turn-taking which underlies a range of social interactions, including communication and language.


Figure S1. Comparison between social and automated error types.
Note: Errors on the task were separated into two categories: those which the target was located on the partner's side of the screen when the subject made a mistake (grey bars), and those in which the target was located on the subject's side of the screen when the subject touched incorrectly. Asterisks indicate significant differences (P < .05, Chi Square Test).

Supplementary Methods
Pairs of chimpanzees were called into the laboratory from their outdoor enclosure, and were led one by one into the two adjacent but separate experimental booths (see main text for further information on size of and touch-screen equipment installed in each booth). Once subjects were settled inside their respective booths, we activated the joint numerical sequencing task. Each trial commenced with the presentation of a small white circle in the outer lower corner of each half of the screen. Both subjects were required to touch this "start key", thus initiating the presentation of the numerical stimuli. Subjects were then shown numerals at random locations within a 4-by-8 matrix on the screen, and were required to respond by touching these numerals in ascending order. Upon being touched, a numeral disappeared from the screen. After the numerals were all touched in the correct order, a chime sounded and a food reward (small cube of apple) was delivered to both subjects simultaneously. If the numerals were touched in the incorrect order (i.e. a higher number touched before a lower one had been touched), a buzzer sounded, no food was delivered, and the trial was terminated.
Phase 1. In the first phase, only two numerals were presented to the subjects on each trial. Sessions consisted of 48 trials, and two sessions were carried out each day for each pair of participants. The 48 trials in a session were comprised of four trial types: the first two trial types consisted of both numerals being presented on the same half of the screen, and the second two trial types consisted of the two numerals being divided across the separate halves of the screen. The four trial types were distributed in a pseudorandom order, such that no one trial type was presented more than three times in a row. The first phase was divided into four stages. In the first stage, the numerical stimuli consisted of numbers 1 and 2. The second stage consisted of numbers 3 and 4, the third stage consisted of numbers 5 and 6, and the fourth stage consisted of numbers 7 and 8. Progress from stage to stage depended on individual subjects' ability to reach an 80% correct criterion level. The 80% criterion was said to have been achieved when subjects performed at or above that level of accuracy within a sliding window of 24 consecutive trials of each trial type. After both subjects in a given pair passed the criterion level for each of the trial types, the pair moved on to the next stage. As the minimum number of trials that were presented before a subject could reach the criterion level in each trial type was 24, if a subject gave correct responses to 20 or more of the initial 24 trial presentations (i.e., was at or above 80% correct), its ability to complete the task was said to be spontaneous rather than the result of trial-and-error learning.
Phase 2. In the second phase, three numerals were presented. As in the first phase, sessions consisted of 48 trials, and two sessions on average were given each day to each pair of participants. The 48 trials in a session were divided into eight trial types: the first two trial types consisted of all three numerals being presented to a single subject on one half of the screen, and the last six trial types consisted of the numerals being divided across both halves of the screen, in all possible configurations. The eight trial types were distributed pseudo-randomly such that no one trial type was presented more than two times in a row. The second phase was divided into three stages. In the first stage, the numerical stimuli consisted of the number set 1-2-3. The second stage consisted of the number set 4-5-6, and the third stage consisted of the number set 7-8-9. As in the first phase, we set the criterion level at 80% correct over a sliding window of 24 trials for each of the eight trial types. Progress from one stage to the next was made as soon as both subjects of a pair reached criterion for all of the trial types.
Phase 3. In the third phase, four numerals were presented. The number of possible permutations of four numerals spread across two halves of the screen was 16 -due to this large number of possible trial types, we increased the number of trials per session to 96. Two sessions were given to each pair of participants on each day. The third phase was divided into two stages. In the first stage, the numerical stimuli consisted of the number set 1-2-3-4, and in the second stage of the number set 5-6-7-8. As in the first two phases, we set the criterion level at 80% correct over a sliding window of 24 trials for each trial type. Progress from the first to second stage was made as soon as both subjects of a pair reached criteria on all trial types.
Phase 4. After the subjects had completed the initial three phases, the fourth phase consisted of presenting eight numerals at random locations on the screen, with the constraint that the two halves always displayed four numerals each, resulting in a set of 70 possible permutations (trial types). 32 blocks were given with each block containing one each of the 70 possible trial types presented in random order. Data from trials in which subjects failed to complete all eight touches to numerals in the correct order were discarded from subsequent response latency analyses.