To live with others is to make concessions. You may want to go to the movies tonight, but your partner may prefer the theatre: reaching a mutually desirable goal – that is, spending time together – requires adjusting your preferences to theirs. Many other social species also make such decisions, in particular monkeys that live in large groups.

Conceptually, these interactions are known as coordination games. In such scenarios, two players must coordinate their actions to attain a coveted reward, but they must also resolve a conflict about who gets the larger share. This makes the joint strategy non-trivial, and different pairs of players might resort to different strategies. In the laboratory, coordination games are often tested in settings which do not allow participants to monitor each other’s behaviors as they make these complex choices. In real life, however, individuals making a joint decision can often observe each other and receive immediate feedback.

In response, Moeller et al. developed a new way to test coordination games that allows more realistic social interactions. In their setup, two participants face each other and use a shared see-through touchscreen to perform a task. This new design was used to test how humans and macaque monkeys solved a simplified version of the ‘Bach or Stravinsky’ coordination game, which involves choosing between a red and blue target on the screen. Players in a pair had been trained to ‘prefer’ opposite colors. In this game, collaboration is beneficial (both individuals get a better prize if they choose the same color) but also creates unfairness (the reward is higher for the participant whose ‘favorite’ color is selected).

When paired up, both humans and monkeys learned to collaborate and to go for the same color (or, in some monkey pairs, the same side of the screen). However, only humans took turns selecting red or blue so that players could alternate getting the highest reward. Monkeys usually settled on one color throughout the game, unless they had learned the ‘turn-taking’ strategy from a human partner; in that case, the color chosen in each trial was typically determined by the monkey who was the faster to move.

These experiments show how monkeys and humans use visual information about their partner’s actions to coordinate their choices, paving the way for further decision-making studies that accurately reflect how interactions unfold in real life. Moeller et al. expect that this will help to understand how cooperation and competition emerge in these two species, including how direct face-to-face contact, or lack thereof in some aspects of our modern world, shapes our social behavior.

The majority of primate species live in complex social groups, in which interactions range from intense competition to cooperation (Byrne and Bates, 2010; Chang et al., 2013; Jaeggi et al., 2010; Joly et al., 2017; Platt et al., 2016). To adjust behavior optimally, individuals need to assess their own actions and goals, and those of other group members, while taking into account the history of interactions with and between these individuals (Frith and Frith, 2010; Platt et al., 2016; de Waal, 2007). Coordination is essential for maintaining cohesion between group members, avoiding conflicts, and achieving individual and joint action goals. Here, we focus on understanding how such coordination can be achieved and maintained in a dyadic ‘transparent’ setting in which agents can observe each other’s ongoing actions and social cues (Unakafov et al., 2020; Unakafov et al., 2019), like in many natural situations (Conty et al., 2012; Fruteau et al., 2013; Rizzolatti et al., 2001; Silk, 2007).

Game theory, developed to study strategic interactions in rational decision-makers (Von Neumann and Morgenstern, 1944), offers a powerful framework to investigate dyadic social interactions (Rilling and Sanfey, 2011; Sanfey, 2007; Tremblay et al., 2017) and the evolution of coordination (Brosnan, 2018; Santos and Rosati, 2015; Smith, 1979). In 2×2 games (Rapoport et al., 1976; Rapoport, 1973), each of the two agents chooses one of two actions and the outcome depends on the combination of their choices. Non-zero-sum 2×2 games that have both cooperative and competitive elements with aligned and opposing interests are particularly interesting for understanding many realistic scenarios. Here, the agents need to coordinate their choices to maximize individual and/or joint rewards (Rapoport, 1967). For example, in the Stag Hunt (or Assurance) game, agents choose between maximizing the individual as well as joint reward and minimizing the individual risk (Skyrms, 2003). Humans (Homo sapiens), chimpanzees (Pan troglodytes), rhesus macaques (Macaca mulatta), and capuchin monkeys (Cebus apella) coordinate at the above chance level on mutually beneficial high-reward/high-risk choices in the iterated Stag Hunt (Brosnan et al., 2012; Brosnan et al., 2011; Bullinger et al., 2011; Duguid et al., 2014). Similar to Stag Hunt, in Prisoner’s Dilemma (PD) two agents maximize their joint reward if they both choose to cooperate. Yet, an individual agent obtains the highest reward when defecting while the partner cooperates (and receives the lowest reward), creating a dilemma between the selfish temptation to defect and a riskier cooperation that requires trust. Humans often show a bias towards cooperative behavior (Camerer, 2003; Rilling et al., 2002; Wedekind and Milinski, 1996). Remarkably, it was demonstrated that macaques, while mostly defecting, chose mutual cooperation significantly more and mutual defection significantly less than expected by chance, especially after preceding cooperation, in an iterated PD game (Haroush and Williams, 2015). This finding suggests that macaques can at least partially overcome the selfish motives and reciprocate cooperation.

Another type of conflict that contrasts selfish and prosocial preferences is implemented in the Conflict game (also known as Hawk-Dove, or game of Chicken) (Brosnan et al., 2017; Smith, 1979). These anti-coordination games model a competition over a shared resource that can be monopolized by one agent, while an actual clash is highly detrimental to both. Humans, capuchins, and macaques often converge on anti-coordination (Brosnan et al., 2017), but importantly, only humans alternate between the two individually-optimal anti-coordinated choices (Brosnan et al., 2017; Grueneisen and Tomasello, 2017; Helbing et al., 2005). Such alternating behavior results in the maximal and equal payoff for both agents and is called ‘cooperative turn-taking’ (Bornstein et al., 1997; Colman and Browning, 2009; Noë, 2006).

In the above games, reward-optimizing rational agents should coordinate on one option (Stag Hunt and Prisoner’s Dilemma) or anti-coordinate on opposite options (Conflict game). A third, less studied type of game that emphasizes the coordination of either of two options is known as Battle of the Sexes, or Bach-or-Stravinsky game (BoS) (Kilgour and Fraser, 1988). Each agent has an individually preferred option, but coordinating on either one of these options adds the same bonus to both agents. This renders any coordinated choice better than no coordination for both agents, but one coordinated choice is better for the first agent and the other better for the second agent. The rational choice is to coordinate, but unlike the conceptually simpler optimal convergence on one option in Stag Hunt and PD, BoS includes an inherent conflict about who profits the most. This combination of cooperation and conflict offers an interesting opportunity for studying social interactions in nonhuman primates. As in the Conflict game, games of the BoS type revealed that humans often take turns, switching between the two coordinated options to ensure fairness (Ioannou and Romero, 2014; Sonsino and Sirota, 2003). Interestingly, in a rope-pulling task that requires cooperation but rewards only one agent at a time, trial-by-trial turn-taking frequently took place in 5-year-old children, but neither in 3-year-old children nor in chimpanzees (Melis et al., 2016). This result begs the question whether turn-taking requires special social or cognitive abilities that are unique to humans. We therefore compared the behavior of (adult) humans and rhesus macaques in the BoS economic game.

Traditionally, economic games are played either ‘simultaneously’ (neither agent knows the choice of the other before making its own decision) or sequentially in a predetermined order. Yet, real dyadic interactions often play out with the partner’s actions in direct sight (Dugatkin et al., 1992; van Doorn et al., 2014). Thus, the timing of one’s own and the other’s actions becomes part of the strategy space (McDonald et al., 2019). Such a ‘transparent’ continuous time setting can change choice strategies as compared to the classic simultaneous and sequential settings (Noë, 2006).

To account for changes in strategies during action visibility compared to ‘opaque’ (simultaneous) choices, we recently developed the concept of ‘transparent games’ that extends the classical evolutionary game-theoretic analysis to real-time interactions in which the visibility of partners’ actions depends on their relative action times (Unakafov et al., 2020; Unakafov et al., 2019). Here, we use the theoretical insights from that work and analyze short- and long-term dynamics of choices, mutual information, and action times in humans and rhesus monkeys playing a transparent version of an iterated movement-based BoS game. To this end, we designed a novel dyadic interaction platform where two human or monkey agents face each other and act on the same visual objects on a vertical touch-sensitive transparent display between them. This configuration provides action visibility, i.e. it allows agents to monitor and react to each other’s actions in real-time, emulating naturalistic interactions under well-controlled laboratory conditions.

To facilitate inter-species comparability, we paired human or macaque agents without explicit task instructions. We expected that the transparency and instantaneous coaction (van Doorn et al., 2014) would facilitate efficient coordination in both species. Based on their cognitive abilities (Proto et al., 2019), propensity for fairness (Peysakhovich et al., 2014), and perspective-taking (Devaine et al., 2017; Herrmann et al., 2007), we predicted humans to engage in pro-social turn-taking, where the reward maximization is alternated between partners. We predicted that macaques would also utilize the information about the partner’s actions, but would be less likely to grant an advantage to the other player. Instead, we expected that the monkeys would coordinate to increase individual (and joint) rewards, but without turn-taking to balance rewards between players.

Two human or macaque agents sat face-to-face and performed a task on a dyadic interaction platform using a shared transparent vertical workspace (Figure 1A). Agents could simultaneously see the stimuli and each other’s actions. In our variant of the Bach-or-Stravinsky (BoS) paradigm (Materials and methods), each agent could choose one of two simultaneously presented color targets (red and blue, Figure 1B). The left/right position of each color was randomized across trials (50% left, 50% right). The rewards followed the payoff matrix shown in Figure 1C and depended on the choices of both agents. Prior to the first dyadic session, each human or macaque agent in a pair was individually trained to associate one of the two color targets with a larger reward, so that the two agents preferred different colors. In the dyadic trials, selecting the same target resulted in an additional reward on top of the individually trained values - such that choosing the same target was always better than choosing different targets, for both agents individually, and as a group. But in such coordinated trials, the agent whose individually preferred (‘own’) color was selected receives a larger reward. Therefore, agents may attempt to behave selfishly/competitively in order to gain more by insisting on their own target color – which pays off as long as the partner accommodates. This paradigm probes the ability to coordinate in order to obtain higher rewards, and the capacity to deal with the unfairness introduced in each coordinated trial.

Figure 1

Download asset Open asset

Human pairs mainly converge on fair coordination

We recruited 38 naïve human subjects to assess performance in the transparent BoS paradigm (Supplementary file 1). Figure 1E and F illustrates an exemplary human pair that developed such alternating turn-taking behavior, showing the fraction of choosing own color (FCO) - the likelihood for each agent to select their own preferred color, and the fraction of choosing left side (FCL) - the likelihood for each agent to select the target on the left side of the display. These measures provide direct insight into each agent’s target and side biases. Here, the FCO curves show that this pair mostly jointly selected A’s or B’s color. After an initial period of long blocks of trials, the pair switched to shorter blocks. The FCL curves show that the agents did not have a side bias, as was also the case for the majority of pairs (Figure 2—figure supplement 1B).

Human dyads showed mainly one of three different choice patterns. Figure 2A plots the larger of the two average fractions of choosing own values in a pair versus the smaller of the two, for all 19 pairs. The FCOs of 53% (10 of 19 pairs, teal underlay) had balanced values close to 0.5 for both agents, indicating alternating between colors, using either trial-by-trial or block-wise turn-taking. The remaining 47% (9 of 19 pairs) had FCO of at least one agent close to one or to zero (both cases indicate a fixed color selection). Five pairs coordinated on a fixed color favoring one agent (blue underlay). Of the four pairs that did not coordinate successfully, in two pairs both agents selected only their own preferred color (red underlay). In the remaining two pairs, one agent selected the non-preferred color while the other switched between the two colors or selected one side. Based on the debriefing responses (Appendix 1, Supplementary file 5), most subjects from the turn-taking dyads (17/20, 85%) realized the coordination benefit, as compared to only half of non-turn-takers (9/18, 50%; Fisher’s exact test p=0.035). In addition, 14 of 20 turn-takers (70%) understood the payoff structure, compared to only 1 of 18 non-turn-takers (6%; Fisher’s exact test p=0.00005).

Figure 2 with 3 supplements see all

Download asset Open asset

All but three pairs achieved an average joint reward above the chance level (2.5 units), and about half the pairs were close to the optimum for the dyad (3.5 units) (Figure 2C). Note that turn-taking is not maximizing average joint reward over coordination on a fixed color (in fact, out of the three pairs reaching the maximum average joint reward of 3.5, only one did so by perfect turn-taking, while the other two used static coordination); it does, however, equalize the rewards within a pair. Over all pairs, the average joint reward increased significantly in the course of the session (Figure 2—figure supplement 1A).

To characterize the dynamics of coordinated behavior, we derived a measure of the dynamic coordination reward (DCR). The DCR describes the average reward a pair earned above (or below) the reward expected for independent random choices, given the observed color choice frequencies per agent. Zero DCR indicates coordination by chance, values significantly different from zero indicate above chance levels of dynamic coordination. The DCR values (Figure 2D) and the fairly balanced average rewards (Figure 2—figure supplement 2B) suggest that 10 of 19 human pairs converged on a form of turn-taking, or reciprocal coordination. These are the same 10 pairs that showed roughly balanced FCO values around 0.5 (teal underlay; Figure 2A).

Comparison of behavior between species

To better understand what drives high dynamic coordination in humans (Figure 2D), we calculated the mutual information (MI) between the sequences of target color choice (MI target) and side choice (MI side) of both agents. These measures indicate how well one agent’s choice of color or side can be predicted from the other’s choice. Comparing the DCR and the MI measures revealed that high DCR values coincided with non-zero values of both MI side and MI color, indicating that dynamic coordination corresponds to both correlated side and correlated color choices between the agents (Figure 2—figure supplement 2D and E).

In contrast to humans, monkeys showed either high values of MI color or MI side indicating that the resulting coordination was not achieved by dynamic turn-taking (Figure 2D). Instead, the four pairs in the middle of Figure 2B (green underlay) that had roughly equal FCO and appeared to be similar to the human turn-takers achieved the coordination by consistently selecting the same side regardless of the target color, which was randomized by the task (Materials and methods).

To directly compare the strategies of human and macaque pairs, we plotted the mutual information values for the side and target color choices of each pair (Figure 3). The distance of each pair’s location to the origin denotes the strength of (anti)coordination. Locations close to the x-axis denote static side strategies; locations close to the y-axis denote static target color strategies. Ten of 19 human pairs were close to the main diagonal, corresponding to similar values of MI side and MI color. These locations denote trial-by-trial or block-wise turn-taking – a signature of dynamic coordination. In contrast to humans, nine macaque pairs showed simple static side-based (four pairs) or single target color-based (five pairs) coordination, with different levels of coordination strength, but did not employ dynamic coordination.

Figure 3

Download asset Open asset

We next asked whether a difference in action times between the two agents in a pair influenced the coordination by affecting the visibility of each other’s actions. Remarkably, all five macaque pairs coordinating on the fixed color converged on the faster monkey’s preferred color (Supplementary file 2). Likewise, for the static side coordination, all three (out of four) pairs where there was a significant difference in action times between the two agents converged on the faster monkey’s less effortful side closer to the acting arm (Supplementary file 2). In humans, such static Leader-Follower relationship also transpired in non-turn-taking dyads coordinating on the fixed color (Supplementary file 1), where the faster agent led to its own color and the slower agent followed (e.g., pairs 10 and 14), or the faster agent led to the partner’s color and the partner followed (e.g. pair 18). This implies that the static coordination in macaques and some human dyads was shaped by the faster agent’s preferences.

Macaques paired with a human confederate learn to follow

The absence of dynamic coordination in macaque pairs does not preclude the possibility that a slower agent could have monitored the faster partner trial-by-trial in order to coordinate. Alternatively, the faster agent might have reduced the possible outcomes for the slower agent to two options by consistently selecting a fixed side or color, and the slower agent then found the best of the remaining options without resorting to the trial-by-trial following. To test if macaques can actively monitor and adjust to dynamic choice behavior, we trained two monkeys in the dyadic condition with a human confederate who followed a strict pattern of alternating between the two colors in blocks of approximately 20 trials (‘confederate training’). If the rhesus monkeys were insensitive to partner’s diverse choices, the pattern of the confederate’s choices should not affect the monkey’s behavior.

Before confederate training, both monkeys adopted either a fixed side or fixed color strategy with the macaque partners (Supplementary file 2). When paired with the confederate, both monkeys started with a strong bias for selecting their own preferred color but over time changed to reliable coordination (Figure 4A, B, and Figure 4—figure supplement 1A, B for another monkey). Their action times (Figure 4 and Figure 4—figure supplement 1, right columns) were initially faster than those of the confederate, and did not vary strongly depending on the confederate’s actions. In later sessions, the animals reacted considerably slower in those blocks in which the confederate was selecting the animals’ non-preferred color, as if waiting for the confederate to commit to an action. We therefore labeled this as ‘following’ behavior.

Figure 4 with 2 supplements see all

Download asset Open asset

The aggregate measures of choice behavior and coordination development over multiple confederate-paired sessions showed similarities in monkeys’ learning (Figure 4—figure supplement 2). Both animals reached close to 0.8 DCR units, similar to the values reached by humans employing turn-taking (Figure 2C). These results indicate that macaques can take information from their partner into consideration when making decisions in the BoS paradigm. Action time dynamics suggested that the macaques started to actively monitor the confederate’s actions and only selected their non-preferred color if the partner started to reach there first.

Due to the repeating block strategy employed by the human confederate, it is not immediately clear whether the monkeys made use of the action visibility and based their decision on the action observed within a trial, or rather on the reward history. We therefore performed a control experiment in which we placed an opaque barrier on the confederate’s side of the display for roughly the middle third of a session. The barrier blocked the monkey’s view of the confederate hand, while keeping the face visible. In both monkeys, the ‘following’ behavior ceased and they reverted to selecting their own preferred colors irrespective of the confederate’s choices, who – by design – continued the block strategy (Figure 4C, Figure 4—figure supplement 1C). In line with this observation, there was a significant difference in the monkey’s action times when selecting the monkey’s preferred (blue) and the non-preferred (red) color in the transparent condition, but no difference in the opaque condition. In other words, monkeys failed to coordinate when they could not see the confederate’s actions. Note that reverting to mostly own choices in combination with a block-switching confederate resulted in a reward alternating between two and four units, clearly different from the constant reward of two units in solo trials. Hence, the experienced reward schedule and the visible presence of a confederate render the opaque condition distinct from the solo trials. These results imply that the immediate visual information about partner’s actions and not merely monitoring reward history drove and maintained the macaques’ ‘following’ behavior in transparent settings.

Confederate-trained macaque pair shows dynamic coordination

After two macaques had learned to monitor and follow a partner’s actions, we tested how these animals would perform when paired with each other again. At the end of the initial naïve sessions of this pair, and before the training with a human confederate, monkey F (agent A) mostly selected his own preferred color while monkey C (agent B) either selected monkey F’s preferred color or the left side. At the end of the confederate training, both F and C followed the confederate. The question was: would they now express turn-taking behavior or would they fall back to their initial static color or side coordination behavior?

The pair’s coordination behavior changed noticeably after training with a confederate. In the first session, after some initial back and forth, the pair converged on mostly selecting B’s color, with B acting faster (Figure 5A). The most interesting phase was during the second session, in which the animals repeatedly alternated between the two colors in long blocks (Figure 5B). In the third session, the pattern was opposite to the first session, with the pair converging mostly on A’s color who acted faster (Figure 5C). This time, B tried occasionally to switch to his preferred color, and A followed for three short periods. The mutual information measures as well as the DCR values indicate that after confederate training, the coordination strength increased and dynamic coordination resulted in turn-taking behavior in two sessions (Figure 5D, Figure 5—figure supplement 1).

Figure 5 with 1 supplement see all

Download asset Open asset

If action visibility is driving this form of dynamic coordination, then the question of who successfully ‘insists’ on own preference and who ‘accommodates’ the other’s preference should depend on the relative action times of the two agents. Indeed, in the session in which the animals exhibited apparent turn-taking behavior (Figure 5B), the faster animal (Leader) selected the preferred color, while the slower animal (Follower) had to accommodate. Thus, the dynamic coordination arose within a competitive interaction.

Comparison of dynamic coordination between species

At first glance, the turn-taking in the confederate-trained macaques resembled the long turn-taking observed in some human pairs (Figure 5D with Figure 3). Yet, the analysis of action time distributions suggests that macaques employed a competitive version of turn-taking in which the faster agent selects its own color (insists) while the slower agent follows (accommodates). Such behavior would be in line with theoretical predictions derived from our evolutionary simulation (Unakafov et al., 2020), showing that competitive turn-taking provides the most effective strategy for a BoS type game when players have a high probability to observe partner’s choices. To further test the hypothesis that this macaque pair developed a competitive Leader-Follower strategy, we performed a correlation analysis between the modeled visibility of the faster agent’s action by the slower agent and the observed likelihood of following to the faster agent’s color (the faster animal nearly always selected his own color). Figure 6A shows macaque agent A’s probability to select his non-preferred color, the ‘fraction of choosing other’s’ (the inverted fraction of choosing own) and the modeled probability for A to see B’s choice before A made own choice, and the same for macaque agent B (see Materials and methods for details on modeling). In both monkeys, there was a strong correlation between the model and the observation, ranging from 0.92 for smoothed data to 0.64 for unsmoothed data, all significant at p<0.001. This analysis further supports the notion that the confederate-trained macaque pair had developed a competitive turn-taking (alternating Leader-Follower) strategy. Strong positive correlations between action visibility and ‘following’ behavior were present in all six post-confederate-trained sessions, indicating bouts of dynamic coordination (Figure 6—figure supplement 1; Supplementary file 3).

Figure 6 with 1 supplement see all

Download asset Open asset

In contrast, human participants did not show a strong trial-by-trial relationship between relative action time differences and choices (Supplementary file 4). Even in the pair with the strongest positive correlations (pair 12, Figure 6B), the ‘following’ behavior was weak and resulted from a gradual transition from B initially following A (to both colors) to A following B (predominately to B’s color) later in the session, rather than from alternating competitive turn-taking. The lack of such a correlation does not mean that action times and action visibility played no role in human choices. Indeed, in 47% (9 of 19) of human pairs, the absolute difference in action times between the agents significantly differed in coordinated trials compared to uncoordinated trials (Supplementary file 1). Furthermore, turn-taking dyads had a larger absolute action time difference (102.7 ± 38.7 ms, N 10) than non-turn-taking dyads (61.3 ± 21.6 ms, N 9; t-test correcting for unequal variance t(14.4) 2.918; p=0.011); this difference was mostly driven by the coordinated trials (105.0 ± 41.5 ms, N 10 versus 64.7±24.1 ms, N 7; t(14.7) 2.523; p=0.024), while a similar trend in trials without coordination did not reach significance (93.8 ± 29.2 ms, N 10 versus 69.5 ± 39.7 ms, N 9; t(14.6) 1.512; p=0.152). This implies that turn-takers tended to pay more attention to partner’s actions and wait more for each other during the coordination. Hence, similarly to macaques, seeing the faster agent’s choice can help the slower agent to coordinate. But unlike the confederate-trained macaque pair, dynamic human coordination was not driven by the faster agents striving to select only the own preferred color (except human pair 12 shown in Figure 6B): for instance, in the turn-taking pairs 3 and 17, the faster agent B led both to own and A’s preferred colors, while A followed (Supplementary file 1).

The action time analyses show that for the confederate-trained macaques, the visibility of the other’s action before making one’s own decision correlated with the probability to follow the choice of the other, with the faster monkey ‘selfishly’ selecting its own preference. Turn-taking in humans, however, did not rely on temporal competition. To highlight this difference further, we asked how the two species maintained the coordination. Macaques transitioned between periods of coordination using short periods of non-coordination, while turn-taking humans tended to switch seamlessly between the two coordination options without verbal communication (Figure 7—figure supplement 1A and B). Turn-taking humans showed a high and roughly equal amount of benevolent and selfish seamless switches by a faster player (Figure 7A, teal underlay). The four macaque pairs showing the highest number of seamless switches used a side-based strategy, which due to the color-to-side randomization trivially generated seamless color switches. Confederate-trained macaques and non-turn-taking humans showed only very few seamless switches.

Figure 7 with 2 supplements see all

Download asset Open asset

Alternative to seamless switches, agents can end the current coordination block by switching to non-coordination. These trials can be used to initiate or ‘signal’ a switch of coordination. When turn-taking humans showed uncoordinated trials they lay along the unity diagonal, indicating a benevolent intention to initiate the switch to the other’s preferred color in half of these trials (Figure 7B). All macaque pairs were below the unity diagonal, indicating a selfish preference for own color. In the confederate-trained macaques, the coordination epochs were separated by epochs of non-coordination that looked like each agent ‘challenged’ the other to accommodate. Indeed, during challenge initiations and resolutions, the faster macaques tended to act selfishly, while turn-taking humans showed balanced switching behavior (Figure 7—figure supplement 1C and D). The analysis of action times in confederate-trained macaques showed that the faster agent slowed down and the slower agent sped up when transiting to non-coordination, and conversely, the insisting agent sped up to initiate the coordination on the own color and the accommodating agent slowed down (Figure 7—figure supplement 2).

In sum, there were no indications of competitive turn-taking in humans, as the faster agents showed a balanced selection of both coordination options. Furthermore, in four out of the ten turn-taking pairs, one agent was significantly faster than the other in coordinated trials as compared to uncoordinated trials, regardless if it was coordination on its own or the other’s color (Supplementary file 1). This means that such an agent initiated switches to and from the own color; while being able to see the faster agent’s action likely helped the slower agent to maintain coordination and to accomplish seamless switches. In contrast, the confederate-trained macaques showed no seamless switches, but displayed temporal competition in which the faster agent led to his own color.

We studied macaque and human pairs in a coordination game, which offered higher rewards for selecting the same option but entailed an inherent conflict about which of the two coordinated options to select. Both species largely converged on coordinated behavior but in a markedly different fashion. Half of the human pairs achieved nearly optimal coordination and fair (cooperative) turn-taking that equalized the rewards of the two partners, without explicit communication. Macaques, instead, exploited simpler strategies that maximized reward without the need to track the actions of the partner in real-time. That is, they used static instead of dynamic coordination. After two macaques were trained to observe and attend to the partner’s choice with a human confederate, their behavior shifted to competitive turn-taking. In post-confederate-training sessions, the choice behavior was strongly correlated with trial-by-trial differences in the reach action times: the faster monkey chose his preferred option and the slower one followed. Our results show that both humans and macaques make use of action visibility by taking information about the other’s action into account before making their own decisions. There was, however, a fundamental difference: when coordinating dynamically, monkeys showed competitive turn-taking while in humans the faster agent often ‘offered’ switching to the partner’s preferred option, exhibiting a form of benevolent and fair turn-taking.

The datasets used in the current study and the links to the public GitHub code repositories are uploaded to a public Open Science Framework data repository (https://osf.io/f5u8z/).

1. 1. Moeller S
   2. Kagan I

   (2023) Open Science Framework

   Human and macaque pairs employ different coordination strategies in a transparent decision game.

   https://doi.org/10.17605/OSF.IO/F5U8Z

1. 1. Arifovic J
   2. Ledyard J

   (2018) Learning to alternate

   Experimental Economics 21:692–721.

   https://doi.org/10.1007/s10683-018-9568-1

   • Google Scholar
2. 1. Báez-Mendoza R
   2. van Coeverden CR
   3. Schultz W

   (2016) A neuronal reward inequity signal in primate striatum

   Journal of Neurophysiology 115:68–79.

   https://doi.org/10.1152/jn.00321.2015

   • PubMed
   • Google Scholar
3. 1. Ballesta S
   2. Duhamel JR

   (2015) Rudimentary empathy in macaques’ social decision-making

   PNAS 112:15516–15521.

   https://doi.org/10.1073/pnas.1504454112

   • PubMed
   • Google Scholar
4. 1. Berger M
   2. Calapai A
   3. Stephan V
   4. Niessing M
   5. Burchardt L
   6. Gail A
   7. Treue S

   (2018) Standardized automated training of rhesus monkeys for neuroscience research in their housing environment

   Journal of Neurophysiology 119:796–807.

   https://doi.org/10.1152/jn.00614.2017

   • PubMed
   • Google Scholar
5. 1. Bornstein G
   2. Budescu D
   3. Zamir S

   (1997) Cooperation in intergroup, N-person, and two-person games of chicken

   Journal of Conflict Resolution 41:384–406.

   https://doi.org/10.1177/0022002797041003003

   • Google Scholar
6. 1. Brocas I
   2. Carrillo JD
   3. Sachdeva A

   (2018) The path to equilibrium in sequential and simultaneous games: a mousetracking study

   Journal of Economic Theory 178:246–274.

   https://doi.org/10.1016/j.jet.2018.09.011

   • Google Scholar
7. 1. Brosnan SF
   2. Parrish A
   3. Beran MJ
   4. Flemming T
   5. Heimbauer L
   6. Talbot CF
   7. Lambeth SP
   8. Schapiro SJ
   9. Wilson BJ

   (2011) Responses to the assurance game in monkeys, apes, and humans using equivalent procedures

   PNAS 108:3442–3447.

   https://doi.org/10.1073/pnas.1016269108

   • Google Scholar
8. Conference

   1. Brosnan SF
   2. Wilson BJ
   3. Beran MJ

   (2012) Old world monkeys are more similar to humans than new world monkeys when playing a coordination game

   Proceedings. Biological Sciences. pp. 1522–1530.

   https://doi.org/10.1098/rspb.2011.1781

   • Google Scholar
9. 1. Brosnan SF
   2. Price SA
   3. Leverett K
   4. Prétôt L
   5. Beran M
   6. Wilson BJ

   (2017) Human and monkey responses in a symmetric game of conflict with asymmetric equilibria

   Journal of Economic Behavior & Organization 142:293–306.

   https://doi.org/10.1016/j.jebo.2017.07.037

   • Google Scholar
10. 1. Brosnan SF

    (2018) Insights into human cooperation from comparative economics

    Nature Human Behaviour 2:432–434.

    https://doi.org/10.1038/s41562-018-0383-7

    • PubMed
    • Google Scholar
11. 1. Bullinger AF
    2. Wyman E
    3. Melis AP
    4. Tomasello M

    (2011) Coordination of chimpanzees (Pan troglodytes) in a STAG Hunt game

    International Journal of Primatology 32:1296–1310.

    https://doi.org/10.1007/s10764-011-9546-3

    • Google Scholar
12. 1. Byrne RW
    2. Bates LA

    (2010) Primate social cognition: uniquely primate, uniquely social, or just unique?

    Neuron 65:815–830.

    https://doi.org/10.1016/j.neuron.2010.03.010

    • PubMed
    • Google Scholar
13. 1. Calapai A
    2. Berger M
    3. Niessing M
    4. Heisig K
    5. Brockhausen R
    6. Treue S
    7. Gail A

    (2017) A cage-based training, cognitive testing and enrichment system optimized for rhesus macaques in neuroscience research

    Behavior Research Methods 49:35–45.

    https://doi.org/10.3758/s13428-016-0707-3

    • PubMed
    • Google Scholar
14. Book

    1. Camerer CF

    (2003)

    Behavioral Game Theory: Experiments in Strategic Interaction

    New York: Princeton University Press.

    • Google Scholar
15. 1. Chang SWC
    2. Brent LJN
    3. Adams GK
    4. Klein JT
    5. Pearson JM
    6. Watson KK
    7. Platt ML

    (2013) Neuroethology of primate social behavior

    PNAS 2:10387–10394.

    https://doi.org/10.1073/pnas.1301213110

    • Google Scholar
16. 1. Clavien C
    2. Chapuisat M

    (2016) The evolution of utility functions and psychological altruism

    Studies in History and Philosophy of Biological and Biomedical Sciences 56:24–31.

    https://doi.org/10.1016/j.shpsc.2015.10.008

    • PubMed
    • Google Scholar
17. 1. Colman AM
    2. Browning L

    (2009)

    Evolution of cooperative turn-taking

    Evolutionary Ecology Research 11:949–963.

    • Google Scholar
18. 1. Conty L
    2. Dezecache G
    3. Hugueville L
    4. Grèzes J

    (2012) Early binding of gaze, gesture, and emotion: neural time course and correlates

    The Journal of Neuroscience 32:4531–4539.

    https://doi.org/10.1523/JNEUROSCI.5636-11.2012

    • PubMed
    • Google Scholar
19. 1. Dal Monte O
    2. Chu CCJ
    3. Fagan NA
    4. Chang SWC

    (2020) Specialized medial prefrontal-amygdala coordination in other-regarding decision preference

    Nature Neuroscience 23:565–574.

    https://doi.org/10.1038/s41593-020-0593-y

    • PubMed
    • Google Scholar
20. 1. de Waal FBM

    (1989) Food sharing and reciprocal obligations among chimpanzees

    Journal of Human Evolution 18:433–459.

    https://doi.org/10.1016/0047-2484(89)90074-2

    • Google Scholar
21. 1. Devaine M
    2. San-Galli A
    3. Trapanese C
    4. Bardino G
    5. Hano C
    6. Saint Jalme M
    7. Bouret S
    8. Masi S
    9. Daunizeau J

    (2017) Reading wild minds: a computational assay of theory of mind sophistication across seven primate species

    PLOS Computational Biology 13:e1005833.

    https://doi.org/10.1371/journal.pcbi.1005833

    • PubMed
    • Google Scholar
22. Book

    1. de Waal F

    (2007) Chimpanzee Politics: Power and Sex Among Apes

    JHU Press.

    https://doi.org/10.56021/9780801886560

    • Google Scholar
23. 1. Dugatkin LA
    2. Mesterton-Gibbonsand M
    3. Houston AI

    (1992) Beyond the prisoner’s dilemma: toward models to discriminate among mechanisms of cooperation in nature

    Trends in Ecology & Evolution 7:202–205.

    https://doi.org/10.1016/0169-5347(92)90074-L

    • PubMed
    • Google Scholar
24. 1. Duguid S
    2. Wyman E
    3. Bullinger AF
    4. Herfurth-Majstorovic K
    5. Tomasello M

    (2014) Coordination strategies of chimpanzees and human children in a stag hunt game

    Proceedings. Biological Sciences 281:20141973.

    https://doi.org/10.1098/rspb.2014.1973

    • PubMed
    • Google Scholar
25. 1. Falcone R
    2. Cirillo R
    3. Ferraina S
    4. Genovesio A

    (2017) Neural activity in macaque medial frontal cortex represents others’ choices

    Scientific Reports 7:12663.

    https://doi.org/10.1038/s41598-017-12822-5

    • PubMed
    • Google Scholar
26. 1. Ferrari-Toniolo S
    2. Visco-Comandini F
    3. Battaglia-Mayer A

    (2019) Two brains in action: joint-action coding in the primate frontal cortex

    The Journal of Neuroscience 39:3514–3528.

    https://doi.org/10.1523/JNEUROSCI.1512-18.2019

    • PubMed
    • Google Scholar
27. 1. Formaux A
    2. Paleressompoulle D
    3. Fagot J
    4. Claidière N

    (2022) The experimental emergence of convention in a non-human primate

    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 377:20200310.

    https://doi.org/10.1098/rstb.2020.0310

    • PubMed
    • Google Scholar
28. 1. Friedman D
    2. Oprea R

    (2012) A continuous dilemma

    American Economic Review 102:337–363.

    https://doi.org/10.1257/aer.102.1.337

    • Google Scholar
29. 1. Frith U
    2. Frith C

    (2010) The social brain: allowing humans to boldly go where no other species has been

    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 365:165–176.

    https://doi.org/10.1098/rstb.2009.0160

    • PubMed
    • Google Scholar
30. 1. Fruteau C
    2. van Damme E
    3. Noë R

    (2013) Vervet monkeys solve a multiplayer “ forbidden circle game ” by queuing to learn restraint

    Current Biology 23:665–670.

    https://doi.org/10.1016/j.cub.2013.02.039

    • PubMed
    • Google Scholar
31. 1. Fujii N
    2. Hihara S
    3. Iriki A

    (2007) Dynamic social adaptation of motion-related neurons in primate parietal cortex

    PLOS ONE 2:e397.

    https://doi.org/10.1371/journal.pone.0000397

    • PubMed
    • Google Scholar
32. 1. Goffe AS
    2. Fischer J

    (2016) Meat sharing between male and female guinea baboons (papio papio)

    Primate Biology 3:1–8.

    https://doi.org/10.5194/pb-3-1-2016

    • Google Scholar
33. 1. Grassberger P

    (1988) Finite sample corrections to entropy and dimension estimates

    Physics Letters A 128:369–373.

    https://doi.org/10.1016/0375-9601(88)90193-4

    • Google Scholar
34. 1. Grueneisen S
    2. Tomasello M

    (2017) Children coordinate in a recurrent social dilemma by taking turns and along dominance asymmetries

    Developmental Psychology 53:265–273.

    https://doi.org/10.1037/dev0000236

    • PubMed
    • Google Scholar
35. 1. Haroush K
    2. Williams ZM

    (2015) Neuronal prediction of opponent’s behavior during cooperative social interchange in primates

    Cell 160:1233–1245.

    https://doi.org/10.1016/j.cell.2015.01.045

    • PubMed
    • Google Scholar
36. 1. Hawkins RXD
    2. Goldstone RL

    (2016) The formation of social conventions in real-time environments

    PLOS ONE 11:e0151670.

    https://doi.org/10.1371/journal.pone.0151670

    • PubMed
    • Google Scholar
37. 1. Helbing D
    2. Schönhof M
    3. Stark H
    4. Hołyst JA

    (2005) How individuals learn to take turns: emergence of alternating cooperation in a congestion game and the prisoner’s dilemma

    Advances in Complex Systems 08:87–116.

    https://doi.org/10.1142/S0219525905000361

    • Google Scholar
38. 1. Herrmann E
    2. Call J
    3. Hernàndez-Lloreda MV
    4. Hare B
    5. Tomasello M

    (2007) Humans have evolved specialized skills of social cognition: the cultural intelligence hypothesis

    Science 317:1360–1366.

    https://doi.org/10.1126/science.1146282

    • PubMed
    • Google Scholar
39. 1. Ioannou CA
    2. Romero J

    (2014) A generalized approach to belief learning in repeated games

    Games and Economic Behavior 87:178–203.

    https://doi.org/10.1016/j.geb.2014.05.007

    • Google Scholar
40. 1. Isoda M
    2. Noritake A
    3. Ninomiya T

    (2018) Development of social systems neuroscience using macaques

    Proceedings of the Japan Academy. Series B, Physical and Biological Sciences 94:305–323.

    https://doi.org/10.2183/pjab.94.020

    • PubMed
    • Google Scholar
41. 1. Jaeggi AV
    2. Burkart JM
    3. Van Schaik CP

    (2010) On the psychology of cooperation in humans and other primates: combining the natural history and experimental evidence of prosociality

    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 365:2723–2735.

    https://doi.org/10.1098/rstb.2010.0118

    • PubMed
    • Google Scholar
42. 1. Jaeggi AV
    2. Gurven M

    (2013) Natural cooperators: food sharing in humans and other primates

    Evolutionary Anthropology 22:186–195.

    https://doi.org/10.1002/evan.21364

    • PubMed
    • Google Scholar
43. 1. Jahng J
    2. Kralik JD
    3. Hwang DU
    4. Jeong J

    (2017) Neural dynamics of two players when using nonverbal cues to gauge intentions to cooperate during the prisoner’s dilemma game

    NeuroImage 157:263–274.

    https://doi.org/10.1016/j.neuroimage.2017.06.024

    • PubMed
    • Google Scholar
44. 1. Joly M
    2. Micheletta J
    3. De Marco A
    4. Langermans JA
    5. Sterck EHM
    6. Waller BM

    (2017) Comparing physical and social cognitive skills in macaque species with different degrees of social tolerance

    Proceedings. Biological Sciences 284:20162738.

    https://doi.org/10.1098/rspb.2016.2738

    • PubMed
    • Google Scholar
45. 1. Kilgour DM
    2. Fraser NM

    (1988) A taxonomy of all ordinal 2 x 2 games

    Theory and Decision 24:99–117.

    https://doi.org/10.1007/BF00132457

    • Google Scholar
46. 1. Martin CF
    2. Biro D
    3. Matsuzawa T

    (2017) Chimpanzees spontaneously take turns in a shared serial ordering task

    Scientific Reports 7:14307.

    https://doi.org/10.1038/s41598-017-14393-x

    • PubMed
    • Google Scholar
47. 1. Massen JJM
    2. van den Berg LM
    3. Spruijt BM
    4. Sterck EHM

    (2010) Generous leaders and selfish underdogs: pro-sociality in despotic macaques

    PLOS ONE 5:e9734.

    https://doi.org/10.1371/journal.pone.0009734

    • PubMed
    • Google Scholar
48. 1. McDonald KR
    2. Broderick WF
    3. Huettel SA
    4. Pearson JM

    (2019) Bayesian nonparametric models characterize instantaneous strategies in a competitive dynamic game

    Nature Communications 10:1808.

    https://doi.org/10.1038/s41467-019-09789-4

    • PubMed
    • Google Scholar
49. 1. Melis AP
    2. Grocke P
    3. Kalbitz J
    4. Tomasello M

    (2016) One for you, one for me: humans’ unique turn-taking skills

    Psychological Science 27:987–996.

    https://doi.org/10.1177/0956797616644070

    • PubMed
    • Google Scholar
50. 1. Noë R

    (2006) Cooperation experiments: coordination through communication versus acting apart together

    Animal Behaviour 71:1–18.

    https://doi.org/10.1016/j.anbehav.2005.03.037

    • Google Scholar
51. 1. Nougaret S
    2. Ferrucci L
    3. Genovesio A

    (2019) Role of the social actor during social interaction and learning in human-monkey paradigms

    Neuroscience and Biobehavioral Reviews 102:242–250.

    https://doi.org/10.1016/j.neubiorev.2019.05.004

    • PubMed
    • Google Scholar
52. 1. Ong WS
    2. Madlon-Kay S
    3. Platt ML

    (2021) Neuronal correlates of strategic cooperation in monkeys

    Nature Neuroscience 24:116–128.

    https://doi.org/10.1038/s41593-020-00746-9

    • PubMed
    • Google Scholar
53. Book

    1. Osborne MJ
    2. Rubinstein A

    (1994)

    A Course in Game Theory

    Cambridge, Mass: The MIT Press.

    • Google Scholar
54. 1. Parrish A
    2. Brosnan S
    3. Beran M

    (2015) Capuchin monkeys alternate play and reward in a dual computerized task

    Animal Behavior and Cognition 2:334–347.

    https://doi.org/10.12966/abc.11.03.2015

    • Google Scholar
55. 1. Pethel S
    2. Hahs D

    (2014) Exact test of independence using mutual information

    Entropy 16:2839–2849.

    https://doi.org/10.3390/e16052839

    • Google Scholar
56. 1. Peysakhovich A
    2. Nowak MA
    3. Rand DG

    (2014) Humans display a “ cooperative phenotype ” that is domain general and temporally stable

    Nature Communications 5:4939.

    https://doi.org/10.1038/ncomms5939

    • PubMed
    • Google Scholar
57. 1. Platt ML
    2. Seyfarth RM
    3. Cheney DL

    (2016) Adaptations for social cognition in the primate brain

    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 371:20150096.

    https://doi.org/10.1098/rstb.2015.0096

    • PubMed
    • Google Scholar
58. 1. Poncela-Casasnovas J
    2. Gutiérrez-Roig M
    3. Gracia-Lázaro C
    4. Vicens J
    5. Gómez-Gardeñes J
    6. Perelló J
    7. Moreno Y
    8. Duch J
    9. Sánchez A

    (2016) Humans display a reduced set of consistent behavioral phenotypes in dyadic games

    Science Advances 2:e1600451.

    https://doi.org/10.1126/sciadv.1600451

    • PubMed
    • Google Scholar
59. 1. Proto E
    2. Rustichini A
    3. Sofianos A

    (2019) Intelligence, personality, and gains from cooperation in repeated interactions

    Journal of Political Economy 127:1351–1390.

    https://doi.org/10.1086/701355

    • Google Scholar
60. 1. Rapoport A

    (1967) Exploiter, leader, hero, and martyr: the four archetypes of the 2 times 2 game

    Behavioral Science 12:81–84.

    https://doi.org/10.1002/bs.3830120202

    • PubMed
    • Google Scholar
61. Book

    1. Rapoport A.

    (1973)

    Two-Person Game Theory

    University of Michigan Press.

    • Google Scholar
62. Book

    1. Rapoport A
    2. Guyer MJ
    3. Gordon DG.

    (1976)

    The 2x2 Game

    University of Michigan Press.

    • Google Scholar
63. 1. Rapoport A

    (1997) Order of play in strategically equivalent games in extensive form

    International Journal of Game Theory 26:113–136.

    https://doi.org/10.1007/BF01262516

    • Google Scholar
64. 1. Rilling JK
    2. Gutman DA
    3. Zeh TR
    4. Pagnoni G
    5. Berns GS
    6. Kilts CD

    (2002) A neural basis for social cooperation

    Neuron 35:395–405.

    https://doi.org/10.1016/s0896-6273(02)00755-9

    • PubMed
    • Google Scholar
65. 1. Rilling JK
    2. Sanfey AG

    (2011) The neuroscience of social decision-making

    Annual Review of Psychology 62:23–48.

    https://doi.org/10.1146/annurev.psych.121208.131647

    • PubMed
    • Google Scholar
66. 1. Rizzolatti G
    2. Fogassi L
    3. Gallese V

    (2001) Neurophysiological mechanisms underlying the understanding and imitation of action

    Nature Reviews. Neuroscience 2:661–670.

    https://doi.org/10.1038/35090060

    • PubMed
    • Google Scholar
67. 1. Roelfsema PR
    2. Treue S

    (2014) Basic neuroscience research with nonhuman primates: A small but indispensable component of biomedical research

    Neuron 82:1200–1204.

    https://doi.org/10.1016/j.neuron.2014.06.003

    • PubMed
    • Google Scholar
68. 1. Rollwage M
    2. Pannach F
    3. Stinson C
    4. Toelch U
    5. Kagan I
    6. Pooresmaeili A

    (2020) Judgments of effort exerted by others are influenced by received rewards

    Scientific Reports 10:1868.

    https://doi.org/10.1038/s41598-020-58686-0

    • PubMed
    • Google Scholar
69. 1. Sanfey AG

    (2007) Social decision-making: insights from game theory and neuroscience

    Science 318:598–602.

    https://doi.org/10.1126/science.1142996

    • PubMed
    • Google Scholar
70. 1. Santos LR
    2. Rosati AG

    (2015) The evolutionary roots of human decision making

    Annual Review of Psychology 66:321–347.

    https://doi.org/10.1146/annurev-psych-010814-015310

    • PubMed
    • Google Scholar
71. 1. Schultz W

    (2015) Neuronal reward and decision signals: from theories to data

    Physiological Reviews 95:853–951.

    https://doi.org/10.1152/physrev.00023.2014

    • PubMed
    • Google Scholar
72. 1. Silk JB

    (2007) The strategic dynamics of cooperation in primate groups

    Advances in the Study of Behavior 37:1–41.

    https://doi.org/10.1016/S0065-3454(07)37001-0

    • Google Scholar
73. Book

    1. Skyrms B

    (2003) The Stag Hunt and the Evolution of Social Structure

    Cambridge University Press.

    https://doi.org/10.1017/CBO9781139165228

    • Google Scholar
74. 1. Smith JM

    (1979) Game theory and the evolution of behaviour

    Proceedings of the Royal Society of London. Series B, Biological Sciences 205:475–488.

    https://doi.org/10.1098/rspb.1979.0080

    • PubMed
    • Google Scholar
75. 1. Sonsino D
    2. Sirota J

    (2003) Strategic pattern recognition—experimental evidence

    Games and Economic Behavior 44:390–411.

    https://doi.org/10.1016/S0899-8256(03)00040-X

    • Google Scholar
76. 1. Steuer R
    2. Kurths J
    3. Daub CO
    4. Weise J
    5. Selbig J

    (2002) The mutual information: detecting and evaluating dependencies between variables

    Bioinformatics 18 Suppl 2:S231–S240.

    https://doi.org/10.1093/bioinformatics/18.suppl_2.s231

    • PubMed
    • Google Scholar
77. 1. Stevens JR
    2. Hauser MD

    (2004) Why be NICE? psychological constraints on the evolution of cooperation

    Trends in Cognitive Sciences 8:60–65.

    https://doi.org/10.1016/j.tics.2003.12.003

    • PubMed
    • Google Scholar
78. 1. Tang H
    2. Mai X
    3. Wang S
    4. Zhu C
    5. Krueger F
    6. Liu C

    (2016) Interpersonal brain synchronization in the right temporo-parietal junction during face-to-face economic exchange

    Social Cognitive and Affective Neuroscience 11:23–32.

    https://doi.org/10.1093/scan/nsv092

    • PubMed
    • Google Scholar
79. 1. Tremblay S
    2. Sharika KM
    3. Platt ML

    (2017) Social decision-making and the brain: A comparative perspective

    Trends in Cognitive Sciences 21:265–276.

    https://doi.org/10.1016/j.tics.2017.01.007

    • PubMed
    • Google Scholar
80. 1. Unakafov AM
    2. Schultze T
    3. Kagan I
    4. Moeller S
    5. Gail A
    6. Treue S
    7. Eule S
    8. Wolf F

    (2019) Evolutionary successful strategies in a transparent iterated prisoner’s dilemma

    Evolutionary Biology 11454:204–219.

    https://doi.org/10.1007/978-3-030-16692-2

    • Google Scholar
81. 1. Unakafov AM
    2. Schultze T
    3. Gail A
    4. Moeller S
    5. Kagan I
    6. Eule S
    7. Wolf F

    (2020) Emergence and suppression of cooperation by action visibility in transparent games

    PLOS Computational Biology 16:e1007588.

    https://doi.org/10.1371/journal.pcbi.1007588

    • Google Scholar
82. 1. van Doorn GS
    2. Riebli T
    3. Taborsky M

    (2014) Coaction versus reciprocity in continuous-time models of cooperation

    Journal of Theoretical Biology 356:1–10.

    https://doi.org/10.1016/j.jtbi.2014.03.019

    • Google Scholar
83. 1. Vaziri-Pashkam M
    2. Cormiea S
    3. Nakayama K

    (2017) Predicting actions from subtle preparatory movements

    Cognition 168:65–75.

    https://doi.org/10.1016/j.cognition.2017.06.014

    • PubMed
    • Google Scholar
84. Book

    1. Von Neumann J
    2. Morgenstern O

    (1944)

    Theory of Games and Economic Behavior

    Princeton University Press.

    • Google Scholar
85. 1. Wedekind C
    2. Milinski M

    (1996) Human cooperation in the simultaneous and the alternating prisoner’s dilemma: pavlov versus generous tit-for-tat

    PNAS 93:2686–2689.

    https://doi.org/10.1073/pnas.93.7.2686

    • PubMed
    • Google Scholar
86. 1. Yoo SBM
    2. Hayden BY
    3. Pearson JM

    (2021) Continuous decisions

    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 376:20190664.

    https://doi.org/10.1098/rstb.2019.0664

    • PubMed
    • Google Scholar
87. 1. Yoshida K
    2. Saito N
    3. Iriki A
    4. Isoda M

    (2011) Representation of others’ action by neurons in monkey medial frontal cortex

    Current Biology 21:249–253.

    https://doi.org/10.1016/j.cub.2011.01.004

    • PubMed
    • Google Scholar
88. 1. Zhao K
    2. Smillie LD

    (2015) The role of interpersonal traits in social decision making: exploring sources of behavioral heterogeneity in economic games

    Personality and Social Psychology Review 19:277–302.

    https://doi.org/10.1177/1088868314553709

    • PubMed
    • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "Human and macaque pairs employ different coordination strategies in a transparent decision game" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Joshua Gold as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Steve W C Chang (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission. Full comments by each of the reviewers are appended to this summary for your reference.

Essential revisions:

1. Some human pairs (about half) did not employ a turn-taking strategy. The authors should discuss potential reasons for these differences, for instance, whether some subjects were not as socially engaged or were less flexible when choosing different options. It would be most helpful if the authors are able to perform additional data analyses to characterize the differences between the two groups of humans (social attentional differences, reaction time differences, etc.) to examine why the turn-taking and non-turn-taking pairs differed.

2. The authors should further consider the degree to which maximizing one's own rewards could drive different strategies in the monkeys. Although cooperating maximized the group's rewards, it wasn't clear whether one monkey could do even better by behaving competitively.

3. It was felt that the manuscript would benefit from including additional information on the monkeys' interactions with the human confederate. Specifically, how did their behavior differ with humans compared to other monkeys, how long do effects of training last, and would the behavior of a trained monkey be expected to translate to an untrained monkey?

4. A notable limitation of the study's ability to compare across species is that the rewards given to the monkeys and humans differed qualitatively (primary vs. secondary rewards). Although this is touched on in the discussion, it was felt that further consideration is warranted. For instance, to what extent might motivation to obtain each outcome drive the tendency to be 'selfish' versus cooperative? Is there any evidence that monkeys would adopt different strategies if they were more sated?

5. Can the authors more precisely consider the role of action visibility? Specifically, whether visibility influences social cognition or aides sensorimotor coordination.

6. Finally, can the authors provide further discussion of the effects of learning and socialization on this task across species?

Reviewer #1 (Recommendations for the authors):

1. It would be helpful to know more about how the goal of maximizing reward might impact the behaviors described here. To start, how does turn-taking compare with the simpler strategies used by the monkeys in terms of reward received? Figure 2C seems to show that the average reward earned by two subjects was actually quite similar between monkeys and humans. The highest-earning human pairs (presumably those who were turn-taking) might have earned slightly more than the monkeys, but there doesn't appear to be a big advantage to turn-taking, in terms of simple reward maximization. Is this true, and if so, is it possible to interpret their behavior as arising from other advantages (e.g., placing value on a partner's reward specifically vs. perceived equity)? What about rewards received by each individual and not the average across pairs? If an agent only values rewards received by themself, is there a more or less optimal approach to this task?

2. Related to the above, it is notable that rewards given to the monkeys and humans differed qualitatively, as did their motivation to pursue them. The humans were receiving small amounts of money in a fairly abstract arrangement in which money was obtained at the end of the session, whereas the monkeys were thirsty and received fluid rewards continuously depending on trial outcomes. To what extent is the motivation to be 'selfish' versus cooperative driven by the motivation to obtain each outcome? For instance, if monkeys were not water deprived, would they be more likely to cooperate rather than compete? Is there any evidence that motivation to get the juice reward changed toward the end of the session?

3. Are there action time plots similar to those shown for the monkeys for the human pairs? Although Figure 6B is compelling, if I understand it correctly, it just shows that the humans' choices can't be predicted based on who responded first. But it's hard to understand, then, how this type of coordination would emerge if one participant didn't watch and follow the other. The results in 6B could occur if the person who responds faster also drives the shifts to the other's target from time to time, as suggested in Figure 7. Therefore, is it possible that there is a leader and follower in the human pairs (perhaps these roles switch off), but the leader just doesn't choose their own color reliably?

Reviewer #2 (Recommendations for the authors):

This manuscript is nicely organized and very thorough. However, it could be improved even further if some points can be discussed in more detail or with additional data analyses.

1. Related to Figure 2D: the authors mentioned that some human pairs (about half) were not employing the turn-taking strategy. From this figure, these pairs were actually not a small number (9 out of 19). The authors should discuss what might be the reason behind this. Is it possible that these subjects were not very socially engaged (i.e., not paying attention to others) or were less flexible when choosing different options (even in non-social situations)? It will be better if the authors could perform additional data analyses to characterize the differences between the two groups of humans (social attentional differences, reaction time differences, etc) to examine why they were different from the turn-taking pairs and more similar to the monkey pairs.

2. Related to Figure 5: It is still a little puzzling that monkeys' behaviors can be improved after being trained with humans. What was the reason for this change? Did the human subjects provide extra social cues to the monkeys in order to encourage them to pay more attention, or did the monkeys learn turn-taking simply from the diverse choice patterns of the human subjects? The authors need to have further discussion about this point.

3. It is not very clear whether the action visibility is influencing social cognition or simply aiding sensorimotor coordination. The authors should discuss this further in Discussion.

Essential revisions:

1. Some human pairs (about half) did not employ a turn-taking strategy. The authors should discuss potential reasons for these differences, for instance, whether some subjects were not as socially engaged or were less flexible when choosing different options. It would be most helpful if the authors are able to perform additional data analyses to characterize the differences between the two groups of humans (social attentional differences, reaction time differences, etc.) to examine why the turn-taking and non-turn-taking pairs differed.

This point is related to the questions raised by Reviewer 1, Recommendations to the authors Comment 3 [R1, C3] and Reviewer 2, Recommendations to the authors Comment 1 [R2, C1].

We thank the Reviewers for bringing up this question. First, we performed additional analysis of action times as well as absolute action time differences between the two agents – i.e. mean(|ATA – ATB|) – to compare the two groups: turn-takers and non turn-takers. There was no significant difference in mean action times (t-test correcting for unequal variance: turn-takers mean (M, in ms) 577.3 SD 80.7 N 20; non turn-takers M 536.1 SD 62.9 N 18; t(35.3) 1.766; p 0.086). Therefore, both groups appear to have a similar engagement in the task. However, when comparing means of the absolute trial-by-trial action time differences between the two agents, turn-taking dyads had a larger difference than non turn-taking dyads. We added this information to Results (lines 416-422) and the revised Table S1 (Supplementary file 1) (columns 11-17):

“Furthermore, turn-taking dyads had a larger absolute action time difference (102.7 ± 38.7 ms, N 10) than non turn-taking dyads (61.3 ± 21.6 ms, N 9; t-test correcting for unequal variance t(14.4) 2.918; p = 0.011); this difference was mostly driven by the coordinated trials (105.0 ± 41.5, N 10 vs. 64.7 ± 24.1, N 7; t(14.7) 2.523; p = 0.024), while a similar trend in trials without coordination did not reach significance (93.8 ± 29.2, N 10 vs. 69.5 ± 39.7, N 9; t(14.6) 1.512; p = 0.152). This implies that turn-takers tended to pay more attention to partner’s actions and wait more for each other during the coordination.”

We emphasize however that the leader-follower dynamics also transpired in non-turn-taking dyads and in macaques statically coordinating on a fixed color or side. We now address this in lines 254-263:

“We next asked whether a difference in action times between the two agents in a pair influenced the coordination by affecting visibility of each other’s actions. Remarkably, all 5 macaque pairs coordinating on the fixed color converged on the faster monkey’s preferred color (Table S2, Supplementary file 2). Likewise, for the static side coordination, all 3 (out of 4) pairs where there was a significant difference in action times between the two agents converged on the faster monkey’s less effortful side closer to the acting arm (Table S2, Supplementary file 2). In humans, such static Leader-Follower relationship also transpired in non-turn-taking dyads coordinating on the fixed color (Table S1, Supplementary file 1), where the faster agent led to own color and the slower agent followed (e.g. pairs 10 and 14), or the faster agent led to the partner’s color and the partner followed (e.g. pair 18). This implies that the static coordination in macaques and some human dyads was shaped by the faster agent’s preferences.”

If the mutual action visibility serves as means of coordination, one may expect that the action time differences should be larger when coordination took place. Therefore, we also directly compared absolute AT differences between the two agents in coordinated vs non-coordinated trials (revised Table S1, Supplementary file 1, columns 18-26). Six of 10 turn-taker dyads showed a significant difference between coordinated and uncoordinated trials, as well as 3 of 9 non turn-taker dyads (the difference in ratios was not significant, Fisher’s exact test, p 0.3698). In 4 turn-taking dyads but also in 2 non turn-taking dyads the difference was larger for coordinated trials, further suggesting leader-follower dynamics that was not limited to turn-takers.

Since our human subjects, like macaques, learned the payoff contingencies while playing the game, we asked several debriefing questions after the experiment (see Supplemental Information). We extracted from these free form reports whether subjects explicitly understood that coordination on “same” targets was beneficial and whether they also realized that the two coordination options were more favorable for one or another player. We now discuss these additional insights in lines 145-148:

“Based on the debriefing responses (Supplementary Information), most subjects from the turn-taking dyads (17/20, 85%) realized the coordination benefit, as compared to only half of non turn-takers (9/18, 50%; Fisher’s exact test p=0.035). In addition, 14 of 20 turn-takers (70%) understood the payoff structure, compared to only 1 of 18 non turn-takers (6%; Fisher’s exact test p=0.00005).”

and lines 501-508:

“What sets the turn-takers apart from the dyads that coordinated on a fixed color, or did not coordinate well? Based on their debriefing responses, turn-takers displayed a much better understanding of the implicit reward rules such as the value of coordination and unfairness. The non turn-taker group was more diverse: some subjects did not understand the game or only understood some aspects; other subjects had a better understanding but were selecting selfishly or could only follow the partner. Furthermore, the turn-taking dyads utilized action visibility driven by action time differences more extensively, although some non-turn-taking dyads overall also exploited action visibility to coordinate on a fixed color.”

We also note that 4 subjects from non-turn-taking pairs (and no subjects from turn-taking pairs) showed <25% own target color selection in the end of solo training (revised Table S1, red font, now mentioned in the legend), in line with the Reviewer’s suggestion of less flexibility in non-social situations.

Finally, the Reviewer 1 is absolutely correct that in many human dyads the faster leaders also drove the shifts to the other's target, as highlighted in Figures 7 and S8, and lines 422-426:

“Hence, similarly to macaques, seeing the faster agent’s choice can help the slower agent to coordinate. But unlike the confederate-trained macaque pair, dynamic human coordination was not driven by the faster agents striving to select only the own preferred color (except human pair 12 shown in Figure 6B): for instance, in the turn-taking pairs 3 and 17, the faster agent B led both to own and A’s preferred colors, while A followed (Table S1, Supplementary file 1).”

To sum up, as a group human turn-takers explored more, better understood the game and used action visibility more to coordinate, but the leaders did not exclusively strive to select their preferred option – either due to understanding of fairness or because they did not realize the payoff asymmetry. The non turn-taker group was diverse, where some subjects did not understand the game or only understood some aspects, whereas other subjects had a better understanding but could only follow the partner. We also note that the proportion of turn-takers in our study, in absence of explicit payoff instructions, is similar to what has been reported in the literature in similar tasks with more explicit instructions, suggesting the facilitation by direct action visibility (lines 497-500):

“The observed turn-taking frequency was close to other variants of BoS games with explicit instructions of the payoff matrix, where approximately 50-60% of human subjects developed turn-taking (Arifovic and Ledyard, 2018; Ioannou and Romero, 2014; Melis et al., 2016; Sonsino and Sirota, 2003).”

While the number of human dyads in our study is not sufficient to make a systematic link between in-game behavior to other measures such as personality traits, this should be addressed in future work in a larger sample (lines 692-696):

“Finally, it would be very interesting to relate diverse in-game behavioral strategies observed in the human pairs to individual differences in cognitive and socio-economic personality traits (Rollwage et al., 2020; Zhao and Smillie, 2015), and link them to stable behavioral phenotypes (Peysakhovich et al., 2014; Poncela-Casasnovas et al., 2016). This however will require a much larger sample and personality-oriented study design.”

2. The authors should further consider the degree to which maximizing one's own rewards could drive different strategies in the monkeys. Although cooperating maximized the group's rewards, it wasn't clear whether one monkey could do even better by behaving competitively.

This point is related to the question raised by Reviewer 1, Recommendations to the authors Comment 1 [R1, C1].

It is an important point to clarify. To recap, the BoS paradigm is constructed such that coordination (selecting the same target) is invariably more rewarded than not coordinating (selecting different targets). This is related to the fact that converging on either of the two targets results in a game theoretic Nash equilibrium – a strategy in which no individual player can improve its payoff by changing course unilaterally. (For completeness, BoS has a third mixed Nash equilibrium in which players select the two choices with specific probabilities, but it is inefficient since the average reward per player is below the reward of even less rewarded player in the two pure – unfair – Nash equilibria). If an agent is dissatisfied with the current mode of coordination (e.g. always following partner’s color and getting only 3 units, or engaging in side-based on dynamic turn-taking and getting only 3.5 units, on average), he/she can try to force/nudge the partner to coordinate on his/her own preferred target by insisting on it, but for this to be a successful reward maximizing strategy the partner needs to follow suit – otherwise the agent initiating the change risks a reward loss if the partner does not follow. In other words (lines 118-124):

“In the dyadic trials, selecting the same target resulted in an additional reward on top of the individually trained values – such that choosing the same target was always better than choosing different targets, for both agents individually, and as a group. But in such coordinated trials, the agent whose individually preferred (“own”) color was selected receives a larger reward. Therefore, agents may attempt to behave selfishly/competitively in order to gain more by insisting on their own target color – which pays off as long as the partner accommodates.”

We indeed observed such behavior in both humans and macaques.

Concerning human dyads, as can be seen in the revised Figure 2C (where we now indicate turn-taking dyads), all 10 turn-taking dyads earned >3 units average joint reward, as compared to only 4 of the 9 non turn-taker dyads (this difference was significant; Fisher’s exact test p < 0.011). But as should be clearer now from the revised figure and lines 150-153:

“Note that turn-taking is not maximizing average joint reward over coordination on a fixed color (in fact, out of the three pairs reaching maximum average joint reward 3.5, only one did so by perfect turn-taking, while the other two used static coordination); it does however equalize the rewards within a pair.”

The turn-taking does optimize “fairness” (with both agents ideally getting 3.5 on average, rather than one getting 4 and other 3). For coordinating non turn-takers, the individual reward reaches maximum (4) for the selfish agent and 3 for the accommodating agent, lower than 3.5, but still higher than 2 for no coordination (e.g. dyads 10 and 18). The revised Figure 2—figure supplement 2B shows the individual rewards and now also indicates the turn-taking pairs.

For macaques, the most parsimonious explanation is that they indeed learned to maximize their own reward, either by finding one of the pure Nash equilibria (unfair to one of the two agents, but still better that no coordination), or by converging on the same side (which incidentally ensured fair distribution – we do not think that monkeys were motivated by fairness, see lines 669-672):

“Secondly, due to the task target color/side randomization, selecting targets on the same side of the display resulted in low effort, efficient and fair coordination. We cannot say if fairness was a contributing factor in convergence of the four macaque pairs to this strategy, although we deem it highly unlikely.”

As can be seen in revised Figure 2—figure supplement 2G which now indicates side strategy pairs, the individual reward is equalized in those macaque pairs but differs by ~0.7-1 units in fixed color-coordinating pairs.

To sum up, “well-coordinating” human dyads (all turn-takers and few non turn-takers) and all macaque dyads maximized their individual reward, achieving a high average reward above 3 units. The slight difference between well coordinating humans and macaques was not significant (Wilcoxon rank sum test: human dyads median 3.44 N 14; macaque dyads median 3.29 N 9, rank sum 192.5, p=0.297; we added this information to Figure 2 legend) and is likely due to more coordination lapses in macaques. For human and macaque dyads who coordinated statically on a fixed color, the difference between higher and lower individual rewards in a dyad was 0.7-1 units, but the accommodating agents still received more than they would for not coordinating and only selecting their own color.

3. It was felt that the manuscript would benefit from including additional information on the monkeys' interactions with the human confederate. Specifically, how did their behavior differ with humans compared to other monkeys, how long do effects of training last, and would the behavior of a trained monkey be expected to translate to an untrained monkey?

This point is related to the question raised by Reviewer 2, Recommendations to the authors Comment 1 [R2, C1]. We answer each of these important aspects below.

How did their behavior differ with humans compared to other monkeys?

The Reviewer is asking which aspects of macaque experience with the human confederate could have led to a turn-taking behavior in monkeys. As we now clarify in lines 280-283 and new Table S2:

“Before confederate training, both monkeys adopted either fixed side or fixed color strategy with the macaque partners (Table S2, Supplementary file 2). When paired with the confederate, both monkeys started with a strong bias for selecting their own preferred color but over time changed to reliable coordination (cf. Figure 4A with 4B, and Figure 4—figure supplement 1A with Figure 4—figure supplement 1B for another monkey).”

The human confederates did not intentionally provide additional social cues. The monkeys learned – in case of monkey F, fairly slowly, in the span of 15 sessions (Figure 4—figure supplement 2A-F) – to follow the diverse choices of consistent blockwise human turn-taker in order to maximize their reward. This learning was facilitated by the direct visibility of human actions.

While our new analysis shows that 8 of 9 monkey pairs, including pairs with monkeys F and C, converged on the faster agent’s preferred choice, we speculate that this type of “static” following was indeed different from trial-by-trial monitoring and following the dynamic human actions. We clarify these points in lines 265-279:

“The absence of dynamic coordination in macaque pairs does not preclude the possibility that a slower agent could have monitored the faster partner trial-by-trial in order to coordinate. Alternatively, the faster agent might have reduced the possible outcomes for the slower agent to two options by consistently selecting a fixed side or color, and the slower agent then found the best of the remaining options without resorting to the trial-by-trial following. To test if macaques can actively monitor and adjust to dynamic choice behavior, we trained two monkeys in the dyadic condition with a human confederate who followed a strict pattern of alternating between the two colors in blocks of approximately 20 trials (“confederate training”). If the rhesus monkeys were insensitive to partner’s diverse choices, the pattern of confederate’s choices should not affect monkey’s behavior.”

in lines 512-518:

“Although we found that all macaque pairs that exhibited a significant difference in action times between the agents (8 of 9 pairs) converged on the faster agent’s preferred choice, it seems that initially macaques did not rely on trial-by-trial action visibility. This is suggested by the fact that it required relatively long training before macaques started to follow a human confederate dynamically. Instead, we speculate that slower macaques adapted to their partner’s static choice by maximizing their reward over the remaining options.”

and lines 614-618:

“Very different dynamics transpired in additional experiments with macaques. Pairing them with a human confederate who consistently alternated between colors in short blocks but did not intentionally provide any additional social cues demonstrated that macaques are capable to closely follow partner’s actions (Nougaret et al., 2019), and adopt the imposed “turn-taking” to maximize their reward.”

How long do effects of training last?

As shown in Author response image 1, the confederate-following behavior, once established, was persistent across multi-week/multi-month breaks in confederate training. We did not explore this question specifically in our study and hence only can present data from the circumstantial breaks that happened in the course of continued training in the context of another study.

Author response image 1

Download asset Open asset

Would the behavior of a trained monkey be expected to translate to an untrained monkey?

We thank the Reviewer for this insightful question. It seems that for turn-taking behavior to occur a single confederate-trained monkey is not sufficient, since turn-taking is a joint strategy that requires active participation of both agents. We now mention this in the Discussion (lines 634-639) and in the legend of the new Table S2:

“Notably, we considered the interactions in the pairs where one monkey underwent the confederate training, and other did not (Table S2, Supplementary file 2). In all 4 instances, static strategies prevailed, suggesting that turn-taking as a joint strategy behavior requires active participation of both agents. We note, however, that in all 3 cases of fixed color strategy, the confederate-trained monkeys accommodated the selfish choice of their partners, perhaps because they had learned to monitor and follow partner’s actions.”

4. A notable limitation of the study's ability to compare across species is that the rewards given to the monkeys and humans differed qualitatively (primary vs. secondary rewards). Although this is touched on in the discussion, it was felt that further consideration is warranted. For instance, to what extent might motivation to obtain each outcome drive the tendency to be 'selfish' versus cooperative? Is there any evidence that monkeys would adopt different strategies if they were more sated?

This point is related to the question raised by Reviewer 1, Recommendations to the authors Comment 2 [R1, C2].

It is indeed an important point; we agree that the difference between primary vs secondary rewards might be crucial and it is essential to consider potential changes in motivation during macaque sessions. We considered the two aspects of this question: (i) coordination vs no coordination and (ii) selfish vs benevolent coordination.

Concerning the first aspect, in the BoS game the coordination is invariably better than no coordination (both for an individual and a dyad, please see also the point 2 above). Therefore, “selfish” tendency to select only own preferred color would result in less reward, unless a partner always accommodates. A “rational” less sated monkey should select the coordination on either color to gain more reward, so from this perspective one might expect that there would be more coordination at the start of the sessions compared to the end. We however observed the effect in the opposite direction: when the difference was significant, the coordination (as assessed by the average joint reward) was stronger in the last 200 trials of a session compared to the first 100. In fact, we see a similar effect in the human dyads, suggesting the development of coordination irrespective of the reward type (see Figure 2—figure supplement 1A,C,D). We now illustrate how the individual average reward, average action time, and the control measure “fraction of choosing left” (pertinent for the side coordination strategy) change between the start and the end of a session (Author response image 2 and 3). In line with the average joint reward, the individual rewards increased significantly in early macaque and human sessions (Author response image 2A,C), and across all sessions in confederate-monkey pairs, significantly for monkey F and marginally for monkey C (Author response image 3A,B).

Author response image 2

Download asset Open asset

Author response image 3

Download asset Open asset

Of course, we cannot rule out that in a less sated monkey the urge of (reflexively) selecting “own” pre-trained color might override the prospect of the coordination benefit, resulting in less coordination at the start of a session. But we cannot disentangle this effect from a need to (re)establish the coordination regime. We emphasize however that both prior to and during the confederate training, there was no significant difference in the coordination strength or individual rewards between the start and the end, in the late sessions – on which we base our strategy descriptions (while this difference was present in early sessions when the dyadic learning just started). In other words, the selection behavior was fairly stable in late macaque-macaque and last confederate-macaque sessions. Likewise, there was no significant increase in rewards in the macaque FC pair after the confederate training.

No change in the fraction of choosing left confirms that there was no shift in side-based strategies in the course of sessions. Finally, we checked how the action times changed in the course of each session. There was an increase from early to late trials (except in confederate-trained FC pair and in humans). This increase can be due to decreased motivation because of satiation and/or fatigue. But to re-iterate, these effects did not result in a shift of strategy.

Concerning the second aspect, selfish vs benevolent coordination, we now analyzed if the proportion of selfish vs benevolent coordinated trials might change in the course of each session (Author response image 4). In the macaque dyadic late sessions, the behavior was very stable, without any change. Likewise, in the last 4 confederate-training sessions of each monkey (where both monkeys successfully followed the confederate), there was no significant change in the course of any session. Only in the dynamically-coordinating macaque FC pair after the confederate training, in several sessions one of the two agents became more selfishly insistent on his preferred color towards the end of the session – while the other accommodated. We now added a note of this effect in the legend to Figure 5—figure supplement 1:

Author response image 4

Download asset Open asset

“To assess if the balance between selfish vs benevolent coordination changed in the course of each session, we analyzed the proportion of coordinating on F’s preferred color among all coordinated trials (note that this is complementary to the proportion of coordinating on C’s color) in first 100 vs last 200 trials (Fisher’s exact test with Bonferroni correction for the number of tested sessions). In several sessions (FC-1, FC-2, FC-3, FC-6) one of the two agents became more selfishly insistent on his preferred color towards the end of the session – while the other accommodated. Interestingly, the roles reversed in different sessions: in FC-1 C is the leader, while in FC-3 and FC-6 F is the leader. We conclude that during the dynamic coordination, there is a tendency for this pair to gradually adopt more consistent leader-follower roles, potentially because one monkey accommodates and follows more easily in a more sated state.”

Furthermore, we added the number of solo trials monkeys performed after each dyadic session, to show that the end of dyadic sessions did not correspond to fully waning motivation (lines 864-866):

“After the dyadic task we offered monkeys the opportunity to obtain additional liquid in the solo version of the task for as long as they continued to work (mean 169 trials, median 73, range 0 to 1163), essentially making participation in the dyadic trials optional.”

To sum up, while it is indisputable that the nature of the reward and the underlying motivations differed between the two species, there is very little indication that variations in satiation level in the course of each session played a substantial role in shaping macaque strategies – further highlighted in lines 682-691:

“Inherently, the nature of reward differed between the species: humans received monetary compensation while macaques worked for a liquid reward. Note, however, that in each session macaques were offered an opportunity to work and obtain more liquid in a solo regime after the dyadic experiment (Materials and methods), so the participation in a dyadic interaction was not forced, and the end of the dyadic sessions was not associated with a fully waning motivation. We also did not observe that variations in satiation level in the course of each session played a substantial role in shaping macaque strategies. Nevertheless, it is important to acknowledge differences between primary and secondary rewards, and potential divergence in the underlying motivations and learning, beyond the species differences in socialization and cognitive capacities considered above.”

5. Can the authors more precisely consider the role of action visibility? Specifically, whether visibility influences social cognition or aides sensorimotor coordination.

This point is related to the question raised by Reviewer 2, Recommendations to the authors Comment 3.

We thank the Reviewer for highlighting this aspect. The answer is both. Action visibility is certainly influencing the sensorimotor coordination, as demonstrated by the differences in action times between opaque and transparent conditions (Figures 4C and S4C), convergence to a faster agent’s preference during static coordination (Table S2, Supplementary file 2), a tight relationship between action visibility and choice behavior (Figures 6 and S7), and inter-agent action time differences in coordinated vs non-coordinated trials (Table S1, Supplementary file 1, see also point 1). Furthermore, anecdotally human confederates reported that macaques paid close attention to their actions and often did not even start the trial unless they saw a human doing so. At the same time, we firmly believe (but can only speculate, based on our own impressions and the informal feedback from human subjects) that also the social cognition “per se” was influenced by the direct face-to-face context and by touching the “same” targets. We expanded discussion on this in lines 582-589:

“In our study, action visibility facilitated sensorimotor coordination, as demonstrated by action time differences in transparent vs. opaque conditions, convergence to a faster agent’s preference during static coordination, a tight relationship between action visibility and choice behavior in turn-taking macaques, and inter-agent action time differences in coordinated vs. non-coordinated trials. Seeing actual movements rather than relying on abstract representations of others such as cursor motions might be especially crucial in nonhuman experiments. But beyond immediate sensorimotor aspects, acting face-to-face on the same targets in close proximity likely influenced social cognition on a more general level.”

6. Finally, can the authors provide further discussion of the effects of learning and socialization on this task across species?

This point is related to the question raised by Reviewer 3. We readily acknowledge the differences between primary vs secondary rewards (please see also point 4 above) and different underlying motivations and expectations of the two species. We see within-session learning effects in both species, but only in monkeys we could study learning across multiple sessions, with the same and different partners. We extended the acknowledgment of these differences in lines 688-696:

“Nevertheless, it is important to acknowledge differences between primary and secondary rewards, and potential divergence in the underlying motivations and learning, beyond the species differences in socialization and cognitive capacities considered above.

Finally, it would be very interesting to relate diverse in-game behavioral strategies observed in the human pairs to individual differences in cognitive and socio-economic personality traits (Rollwage et al., 2020; Zhao and Smillie, 2015), and link them to stable behavioral phenotypes (Peysakhovich et al., 2014; Poncela-Casasnovas et al., 2016). This however will require a much larger sample and personality-oriented study design.”

Reviewer #1 (Recommendations for the authors):

1. It would be helpful to know more about how the goal of maximizing reward might impact the behaviors described here. To start, how does turn-taking compare with the simpler strategies used by the monkeys in terms of reward received? Figure 2C seems to show that the average reward earned by two subjects was actually quite similar between monkeys and humans. The highest-earning human pairs (presumably those who were turn-taking) might have earned slightly more than the monkeys, but there doesn't appear to be a big advantage to turn-taking, in terms of simple reward maximization. Is this true, and if so, is it possible to interpret their behavior as arising from other advantages (e.g., placing value on a partner's reward specifically vs. perceived equity)? What about rewards received by each individual and not the average across pairs? If an agent only values rewards received by themself, is there a more or less optimal approach to this task?

Please see the response to the point 2 of essential revisions.

In brief, the revised Figure 2C and associated analysis indeed show that the average joint reward was similar between monkeys and humans, and the advantage of the turn-taking (or a side strategy resulting in “automatic turn-taking” adopted by 4 monkey pairs) is not in maximizing the joint reward per pair, but in simultaneously maximizing both individual rewards. We surmise that human turn-takers – but not side-coordinating macaques – were indeed at least in part motivated by equity considerations.

The individual rewards are shown in Figure 2—figure supplement 2B,G. Indeed, if the agent only values own reward, the optimal approach is to force coordination on his/her preferred color, as long as the partner plays along. Clear examples of such behavior are human pairs 10, 18, 16 and macaque pairs CL, TF and TC (individual rewards ~4 and ~3). From a purely economic value perspective of a selfish agent, the turn-taking only makes sense to ensure coordination in a long term (to avoid a scenario where previously accommodating partner is fed up and starts refusing to coordinate at all on the selfish agent’s color).

2. Related to the above, it is notable that rewards given to the monkeys and humans differed qualitatively, as did their motivation to pursue them. The humans were receiving small amounts of money in a fairly abstract arrangement in which money was obtained at the end of the session, whereas the monkeys were thirsty and received fluid rewards continuously depending on trial outcomes. To what extent is the motivation to be 'selfish' versus cooperative driven by the motivation to obtain each outcome? For instance, if monkeys were not water deprived, would they be more likely to cooperate rather than compete? Is there any evidence that motivation to get the juice reward changed toward the end of the session?

Please see the response to the point 4 of essential revisions.

3. Are there action time plots similar to those shown for the monkeys for the human pairs? Although Figure 6B is compelling, if I understand it correctly, it just shows that the humans' choices can't be predicted based on who responded first. But it's hard to understand, then, how this type of coordination would emerge if one participant didn't watch and follow the other. The results in 6B could occur if the person who responds faster also drives the shifts to the other's target from time to time, as suggested in Figure 7. Therefore, is it possible that there is a leader and follower in the human pairs (perhaps these roles switch off), but the leader just doesn't choose their own color reliably?

Please see the response to the point 1 of essential revisions.

Since the point 1 of essential revisions also deals with a different emphasis on comparing turn-takers vs non turn-takers, raised by the Reviewer 2, we briefly summarize the specific answer here. The Reviewer is absolutely correct – indeed, there were temporal leaders and followers in the human pairs, and indeed, the faster players equally often selected their own color and the color of the partner, to equalize the rewards, as suggested by Figure 7 and Figure 7—figure supplement 1. We now emphasize this point more clearly in lines 422-426:

“Hence, similarly to macaques, seeing the faster agent’s choice can help the slower agent to coordinate. But unlike the confederate-trained macaque pair, dynamic human coordination was not driven by the faster agents striving to select only the own preferred color (except human pair 12 shown in Figure 6B): for instance, in the turn-taking pairs 3 and 17, the faster agent B led both to own and A’s preferred colors, while A followed (Table S1, Supplementary file 1).”

Figure 6—figure supplement 1 and Table S4 (Supplementary file 4) show correlation coefficients between the visibility of other’s actions and selecting other’s color for all human dyads. The pair 12 shown in Figure 6B shows the highest correlation among the human dyads. In this pair, firstly player A led (equally often to own and to B’s color) and B followed, and then B led (more to own color) and A followed. Among all human pairs, only in this pair there was a reliable difference between AT A – AT B for coordination on A’s color vs coordination on B’s color, but even here the separation between the AT difference histograms was not as clear as in the macaque pair FC (Author response image 5C).

Author response image 5

Download asset Open asset

Reviewer #2 (Recommendations for the authors):

This manuscript is nicely organized and very thorough. However, it could be improved even further if some points can be discussed in more detail or with additional data analyses.

1. Related to Figure 2D: the authors mentioned that some human pairs (about half) were not employing the turn-taking strategy. From this figure, these pairs were actually not a small number (9 out of 19). The authors should discuss what might be the reason behind this. Is it possible that these subjects were not very socially engaged (i.e., not paying attention to others) or were less flexible when choosing different options (even in non-social situations)? It will be better if the authors could perform additional data analyses to characterize the differences between the two groups of humans (social attentional differences, reaction time differences, etc) to examine why they were different from the turn-taking pairs and more similar to the monkey pairs.

Please see the response to the point 1 of essential revisions.

2. Related to Figure 5: It is still a little puzzling that monkeys' behaviors can be improved after being trained with humans. What was the reason for this change? Did the human subjects provide extra social cues to the monkeys in order to encourage them to pay more attention, or did the monkeys learn turn-taking simply from the diverse choice patterns of the human subjects? The authors need to have further discussion about this point.

Please see the response to the point 3 of essential revisions.

3. It is not very clear whether the action visibility is influencing social cognition or simply aiding sensorimotor coordination. The authors should discuss this further in Discussion.

Please see the response to the point 5 of essential revisions.

Author details

1. Sebastian Moeller

   1. Cognitive Neuroscience Laboratory, German Primate Center – Leibniz Institute for Primate Research, Göttingen, Germany
   2. Leibniz ScienceCampus Primate Cognition, Göttingen, Germany

   Contribution

   Conceptualization, Resources, Data curation, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-0381-6449

2. Anton M Unakafov

   1. Cognitive Neuroscience Laboratory, German Primate Center – Leibniz Institute for Primate Research, Göttingen, Germany
   2. Leibniz ScienceCampus Primate Cognition, Göttingen, Germany
   3. Georg-Elias-Müller-Institute of Psychology, University of Gottingen, Göttingen, Germany
   4. Max Planck Institute for Dynamics and Self-Organization, Göttingen, Germany
   5. Campus Institute for Dynamics of Biological Networks, Gottingen, Germany

   Contribution

   Conceptualization, Resources, Software, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft

   Competing interests

   No competing interests declared

3. Julia Fischer

   1. Leibniz ScienceCampus Primate Cognition, Göttingen, Germany
   2. Cognitive Ethology Laboratory, German Primate Center – Leibniz Institute for Primate Research, Göttingen, Germany
   3. Department of Primate Cognition, Johann-Friedrich-Blumenbach Institute for Zoology and Anthropology, University of Gottingen, Göttingen, Germany

   Contribution

   Funding acquisition, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-5807-0074

4. Alexander Gail

   1. Cognitive Neuroscience Laboratory, German Primate Center – Leibniz Institute for Primate Research, Göttingen, Germany
   2. Leibniz ScienceCampus Primate Cognition, Göttingen, Germany
   3. Georg-Elias-Müller-Institute of Psychology, University of Gottingen, Göttingen, Germany
   4. Bernstein Center for Computational Neuroscience, Göttingen, Germany

   Contribution

   Conceptualization, Supervision, Funding acquisition, Project administration, Writing – review and editing

   Competing interests

   No competing interests declared

   Additional information

   Co-senior authors

   "This ORCID iD identifies the author of this article:" 0000-0002-1165-4646

5. Stefan Treue

   1. Cognitive Neuroscience Laboratory, German Primate Center – Leibniz Institute for Primate Research, Göttingen, Germany
   2. Leibniz ScienceCampus Primate Cognition, Göttingen, Germany
   3. Georg-Elias-Müller-Institute of Psychology, University of Gottingen, Göttingen, Germany
   4. Bernstein Center for Computational Neuroscience, Göttingen, Germany

   Contribution

   Conceptualization, Supervision, Funding acquisition, Project administration, Writing – review and editing

   Competing interests

   No competing interests declared

   Additional information

   Co-senior authors

6. Igor Kagan

   1. Cognitive Neuroscience Laboratory, German Primate Center – Leibniz Institute for Primate Research, Göttingen, Germany
   2. Leibniz ScienceCampus Primate Cognition, Göttingen, Germany

   Contribution

   Conceptualization, Data curation, Supervision, Funding acquisition, Visualization, Writing – original draft, Project administration, Writing – review and editing

   For correspondence

   ikagan@dpz.eu

   Competing interests

   No competing interests declared

   Additional information

   Co-senior authors

   "This ORCID iD identifies the author of this article:" 0000-0002-1814-4200

• 731

  Page views

• 110

  Downloads

• 2

  Citations

Article citation count generated by polling the highest count across the following sources: PubMed Central, Crossref, Scopus.