Cortical thickness of the dorsolateral predicts strategic choices in economic games

Toshio Yamagishia,b,1, Haruto Takagishib, Alan de Souza Rodrigues Ferminb, Ryota Kanaic, Yang Lib, and Yoshie Matsumotob

aGraduate School of International Corporate Strategy, Hitotsubashi University, 2-1-2 Hitotsubashi, Chiyoda-ku, Tokyo 101-8439, Japan; bBrain Science Institute, Tamagawa University, 6-1-1 Tamagawagakuen, Machida, Tokyo 194-8610, Japan; and cDepartment of Neuroinformatics, Araya Brain Imaging, 3-16-16 Daizawa, Setagayaku, Tokyo 155-0032, Japan

Edited by Susan T. Fiske, Princeton University, Princeton, NJ, and approved March 30, 2016 (received for review December 5, 2015) Human prosociality has been traditionally explained in the social cooperated more (23). Additionally, increased activity in the sciences in terms of internalized social norms. Recent neuroscien- lateral prefrontal cortex was negatively related to fairness-seeking tific studies extended this traditional view of human prosociality behavior in an economic game (24). According to the heuristic by providing evidence that prosocial choices in economic games prosociality model (14–17), humans are predisposed to cooperate require cognitive control of the impulsive pursuit of self-interest. in social exchange situations. People fail to behave in a prosocial However, this view is challenged by an intuitive prosociality view manner in social exchanges when this predisposition is overridden emphasizing the spontaneous and heuristic basis of prosocial by strategic reasoning to secure their self-interest. By comparing choices in economic games. We assessed the brain structure of participants’ behaviors in two economic games with brain structural 411 players of an ultimatum game (UG) and a dictator game (DG) differences and strategic reasoning abilities, we provide evidence and measured the strategic reasoning ability of 386. According to that strategic reasoning controls, and thus reduces rather than the reflective norm-enforcement view of prosociality, only those promotes, game players’ prosocial behavior. capable of strategically controlling their selfish impulses give a fair The contrast between two simple, two-person economic games— share in the UG, but cognitive control capability should not affect namely, the dictator game (DG) and the UG—is often used to behavior in the DG. Conversely, we support the intuitive proso- support the reflective prosocial model by demonstrating how strategic ciality view by showing for the first time, to our knowledge, that reasoning affects game players’ decisions. In both games, one player COGNITIVE SCIENCES strategic reasoning and cortical thickness of the dorsolateral freely decides how much of a fixed reward to take and how much to PSYCHOLOGICAL AND prefrontal cortex were not related to giving in the UG but were leave for the other player. The difference between the two games is negatively related to giving in the DG. This implies that the that the other player in the UG (termed “responder”)hastheoption uncontrolled choice in the DG is prosocial rather than selfish, and to reject the decision made by the first player (termed “proposer”), those who have a thicker dorsolateral prefrontal cortex and are causing both to earn nothing. This option is not provided to the capable of strategic reasoning (goal-directed use of the theory of second player in the DG, who plays the role of a “recipient”.The mind) control this intuitive drive for prosociality as a means to “ ” maximize reward when there are no future implications of choices. recipient simply receives whatever the first player ( dictator ) gives. The level of giving by the proposer in the UG is usually higher than ultimatum game | dictator game | strategic reasoning | DLPFC | that by the dictator in the DG (25). This is attributed to the pro- ’ ’ prosocial behavior poser s strategic reasoning, which requires inference of the recipient s internal state and prediction of the resulting response (e.g., anger on the basis of unfair giving and subsequent rejection) (8, 9, 13). Given umans are a cooperative species, and the question of why that neuroimaging and neuroendocrinological studies showed that humans are so cooperative has been a subject of consider- H negative emotions are associated with rejection of unfair offers able interest in social and biological sciences (1–4). The tradi- (24, 26, 27), UG proposers may anticipate negative responses to tional answer in the social sciences highlights critical roles of unfair offers. UG proposers anticipate norm-enforcing responses social norms and cultural values internalized as personal values (rejection of the offer) to norm-violating behavior (taking most of the and social preferences (5, 6). Recent neuroscientific studies of brain structure and activity extended this traditional view of human prosociality by showing that players of economic games act Significance prosocially when they cognitively control selfish impulses (7–13). Experimental evidence shows that prosocial choices in economic Is human prosociality a consequence of cognitive control of selfish games positively relate to local gray matter volume and thickness impulses? Alternatively, is it a default option that most people use and the activation of brain areas that control selfish impulsive unless they are cognitively persuaded that a given situation does drives, such as the dorsolateral prefrontal cortex (DLPFC) and not require them to behave prosocially? Our results support the (TPJ) (7–9). Furthermore, impairment of latter argument. Participants with weaker cognitive control fairly cognitive control by disruption of DLPFC function prevents re- shared a reward with another participant even when there was no jection of unfair offers in the ultimatum game (UG), which some chance of punishing unfair behavior, whereas those more capable authors considered prosocial and fairness-seeking behavior (10– of cognitive control behaved selfishly in the same situation. These findings demonstrate that participants’ intuitive choices in eco- 13). Recently, this reflective view of human prosociality has been nomic games are prosocial. challenged by an alternative view emphasizing the intuitive na-

ture of prosocial behavior, subsumed under intuitive prosociality Author contributions: T.Y., Y.L., and Y.M. designed research; H.T., Y.L., and Y.M. per- (14) or heuristic cooperation (15–17). Support for the intuitive formed research; T.Y., H.T., A.d.S.R.F., and R.K. analyzed data; and T.Y., Y.L., and Y.M. and automatic nature of prosocial behavior is provided by find- wrote the paper. ings that prosocial choices are promoted under time pressure The authors declare no conflict of interest. (15, 16, 18), under cognitive load (19–21), or after priming This article is a PNAS Direct Submission. by successful experiences of intuitive decision making (15, 22). 1To whom correspondence should be addressed. Email: [email protected]. Also, participants who expressed more positive emotional words This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. and less inhibitory words during and after an economic game 1073/pnas.1523940113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1523940113 PNAS Early Edition | 1of6 Downloaded by guest on October 1, 2021 thus make distinct predictions regarding the relationship between DLPFC thickness and behavior in the UG and DG. The reflective model predicts a positive relationship between DLPFC thickness and giving in the UG, whereas the intuitive model predicts a neg- ative relationship between DLPFC thickness and giving in the DG. These two alternative accounts of differences in giving in the UG and DG (8, 9) provide a way to test the intuitive selfishness assumption against the intuitive prosociality assumption. We first successfully replicated earlier findings that strategic behavior is more pronounced among those who had a thicker DLPFC than those who had a thinner DLPFC (8) in a study of 411 adult, nonstudent participants who played both the UG and DG and Fig. 1. Schematic representations of how strategic considerations generate from whom brain structural images were obtained. Then, we the difference between strategists (ST) and nonstrategists (Non-ST) in the UG found for the first time, to our knowledge, that local gray matter and DG. A shows the prediction that strategic considerations should improve thickness of the DLPFC negatively correlated with giving in the fair behavior in the UG. B shows the prediction that strategic considerations DG but was not correlated with giving in the UG (Fig. 2 C and should depress fair behavior in the DG. D). We further measured the strategic reasoning of 386 of these participants using a newly developed test of strategic reasoning, reward) and strategically adjust giving behavior to secure acceptance measured 411 participants’ Machiavellianism (34, 35) score, and by the responder. Thus, those capable of using strategic reasoning are found that those exhibiting better strategic reasoning behaved expected to make fair offers in the UG compared with those who more selfishly in the DG than those with poor strategic reasoning, struggle to control their selfish drive for immediate reward. but no relationship was found between task performance and In contrast, in the DG, which requires no strategic reasoning fairness in the UG. These striking findings provide strong evidence B to earn as much as possible, strategic control over selfish impulses is supporting the intuitive prosociality prediction depicted in Fig. 1 A expected not to influence the player’s choices. Spitzer et al. (9) but not the reflective prosociality prediction shown in Fig. 1 . confirmed this by showing a positive correlation between the dif- Results ference in giving in the UG and the DG (i.e., a measure of strategic Cortical Thickness and Game Behavior. The participants gave a reasoning) and activity of the right DLPFC and the lateral orbito- = frontal cortex. Given earlier findings implicating the DLPFC in mean proportion of the endowment of 0.410 (SE 0.007) in the – UG, which was significantly higher than the mean proportion in cognitive control of impulsive behavior (28 33),thisistakento = = t = P < support the reflective model of prosociality, in which prosocial be- the DG (M 0.324, SE 0.010), (410) 8.97, 0.0001. These havior requires cognitive control of the impulsive drive toward results agreed with previous findings (8, 9). We first sought to selfish behavior. Steinbeis et al. (8) provided further support via a replicate the earlier finding of the positive correlation between cortical thickness of the DLPFC and strategic behavior, shown as comparison of young children’s choices in the two economic games. difference in money provided to the partner in the UG and DG. Children took a large share in the DG while providing fairer Cortical thickness was estimated using the FreeSurfer package amounts to responders in the UG. The children’s more generous (see Methods), and the parcellated according to giving in the UG may be based upon strategic reasoning regarding the Destrieux Anatomical Atlas (36). Given that neuroimaging the possible consequences of not giving enough in the UG—that is, studies with humans associate the DLFPC broadly with the receiving no reward due to rejection by the other child—which plays middle and superior (37), we liberally matched the no role in the DG. Thus, the difference in giving between the UG DLPFC to regions 15 (middle frontal ), 16 (superior frontal and the DG is considered to reflect the use of strategic reasoning in gyrus), 52 (inferior frontal ), 53 (middle frontal sulcus), the UG. The strategic choices of more giving in the UG than in the ’ and 54 (superior frontal sulcus) of the Destrieux Atlas (see Fig. DGisrelatedtochildrens age, cortical thickness, and activity of 2A), with the aim of identifying component regions of the their left DLPFC. As children age and their DLPFC develops DLPFC related to strategic behavior. More specifically, we fo- further, they become able to control their selfish drive and adjust cused on the as the DLPFC because it con- their behavior to the anticipated negative consequences. – tains cytoarchitectonic regions ( 46 and 9) classically This interpretation of UG DG difference in prosocial giving as a identified as the DLPFC in macaques and humans (38). reflection of strategic reasoning (8, 9) assumes that the default Significant correlations were found between strategic behavior choice in the DG is impulsive and selfish. Younger children and and cortical thickness of the right (r = 0.117, P = 0.018) and left those with a thinner DLPFC are presumably less capable of stra- (r = 0.121, P = 0.015) DLPFCs, after adjusting for age, sex, and tegically adjusting their decisions to deal with anticipated responses intracranial volume (ICV) (Fig. 2B), replicating earlier findings and would impulsively pursue their own benefits in both the UG (8). A similar relationship was also found for the superior frontal and the DG. In contrast, older children and those with a thicker gyrus (Fig. S1). Relationships between giving in the two games DLPFC are more likely to have enhanced cognitive control, which and cortical thickness were also found with regard to the left and can be used to strategically adjust their choices, especially in the right superior frontal sulcus, although they were less clear for the UG but not in the DG. Therefore, a UG–DG positive reward latter (Fig. S2). Neither cortical thickness of the middle frontal transfer difference is produced by strategists’ control over the sulcus (Fig. S3) nor the (Fig. S4) were selfish impulses in the UG, whereas those who fail to control such correlated with strategic behavior or the level of giving in the two impulses in the UG claim a considerable share in both games (Fig. games. We then tested the relationships between DLPFC thick- 1A). In contrast, the alternative, intuitive prosociality model as- ness and proportions of giving in the UG and DG separately. sumes that the uncontrolled choice is prosocial in both the UG and Neither the cortical thickness of the right nor the left DLPFC the DG, rather than selfish. Strategists control this impulse toward significantly correlated with the level of giving in the UG (Fig. 2 C prosociality in the DG where immediate pursuit of self-interest and D). In contrast, cortical thickness significantly negatively causes no strategic problem (Fig. 1B). Nonstrategists do not control correlated with the level of giving in the DG (Fig. 2 C and D). In a this impulse and provide a fair share in both games. Thus, a dif- repeated-measures analysis of variance including interaction be- ference due to strategic reasoning is predicted to exist in the DG tween the two games, controlling for age, sex, and ICV, the in- but not in the UG. The reflective and intuitive prosociality models teraction effect was significant—rDLPFC, F(1, 406) = 5.68, P = 0.018,

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1523940113 Yamagishi et al. Downloaded by guest on October 1, 2021 Fig. 2. Brain areas in the Destrieu Atlas (A), the relationship of cortical thickness of the DLPFC (middle frontal gyrus) and strategic choice (UG–DG) (B), the re- lationship of its right cortical thickness and giving in the UG and DG (C), and the relationship of its left cortical thickness and giving in the UG and DG (D). The horizontal axis represents the residual cortical thickness adjusted for participants’ age, sex, and ICV. The vertical axis represents the mean strategic choice of the players, giving in the UG or the DG within 0.1-mm intervals of residual cortical thickness. Each interval spans 0.1 mm on the horizontal axis segment. The size of each circle shows the number of players who fell within the interval. Error bars are SEs. n = 411. Correlations are after adjusting for age, sex, and ICV.

and lDLPFC, F(1, 406) = 5.99, P = 0.015. These findings are P < 0.0001, IQ was not significantly correlated with strategic be- consistent with the prediction based on the intuitive prosociality havior (UG–DG; r = 0.080, P = 0.105), UG proposal (r = –0.031, model illustrated in Fig. 1B and inconsistent with the reflexive P = 0.529), or DG giving (r = –0.096, P = 0.053). Regression COGNITIVE SCIENCES prosociality model prediction (Fig. 1A). analyses of strategic behavior indicated that the effect of strategists PSYCHOLOGICAL AND versus nonstrategists remained highly significant after controlling Strategic Reasoning and Game Behavior. We further tested the for IQ (β = 0.069), t(383) = 3.17, P = 0.002. predictions, illustrated in Fig. 1 A and B, specifying how be- haviors in the UG and the DG are related to strategic decision Machiavellian Personality and Game Behavior. We successfully making. Despite suggestions that differences in choice behavior replicated the earlier finding (9) of a positive correlation between in the UG may reflect the use of theory of mind to anticipate Machiavellian personality score and strategic behavior (r = 0.185, others’ reactions, no direct evidence supported this. We first P < 0.001). Although Machiavellian personality score was nega- tested participants’ strategic reasoning using the Machiavelli tively correlated with giving in both the DG (r = –0.255, P < 0.0001) (Mac) game, which requires forward planning to select the and UG (r = –0.121, P = 0.014), the interaction effect was signifi- ’ choice that maximizes own reward based on estimation of others cant, F(1, 409) = 14.55, P < 0.001. Given that the Machiavellian mental states and likely choices. Participants played the Mac personality scale is a mixture of strategic pursuit of self-interest and game eight times with two other players, each time with a dif- general distrust of others (39) and that the latter component was ferent set of parameters, and without feedback after each trial. irrelevant in our analysis of strategic reasoning, we used Yama- One strategic response in each trial would give the participant gishi’s general trust score (40) as a control variable. When general the highest monetary reward if he or she could successfully infer trust was controlled, Machiavellian personality score correlated with other two players’ choices and influence their decisions using DG giving (r = –0.171, P < 0.001) but not UG proposal (r = –0.062, monetary incentives (see Methods for further details). According P = 0.214). When general trust was added as an independent to a binomial distribution of random choices with a probability of variable, the interaction effect remained significant, F(1, 408) = 0.5 for each choice, the probability of making seven (P = 0.046) 7.65, P = 0.006. or eight (P = 0.004) strategic responses in the eight trials was less than 0.05. We thus classified those whose number of optimal Discussion strategic responses were seven or eight as strategists and those The reflective model of prosociality assumes that prosocial be- whose number of optimal strategic responses was six or less havior requires control of selfish impulses to meet demands of as nonstrategists. The proportion of strategists was 0.352 (see Fig. S5 for the distribution of the number of strategic responses). Strategists had a higher level of strategic behavior (i.e., UG–DG giving) compared with nonstrategists, t(239.7) = 3.36, P < 0.001 (see Fig. 3). Consistent with heuristic prosociality model predic- tions (Fig. 1B), there was no significant difference in the level of giving between strategists and nonstrategists in the UG, t(384) = 0.05, P = 0.959 (Fig. 3). Conversely, in the DG, strategists’ level of giving was significantly lower than nonstrategists, t(384) = 3.21, P = 0.001 (Fig. 3). The interaction effect of the UG versus DG and the dichotomized Mac game score was highly significant, F(1, 384) = 12.52, P < 0.001.Thesameconclusionsweredrawnwhenthe original Mac game scores instead of the dichotomized scores were r = – P = r = – P = used (UG: 0.009, 0.854; DG: 0.163, 0.001; in- Fig. 3. Means of strategic choice (difference in giving in the UG and DG), teraction: F(1, 384) = 11.43, P < 0.001). Although strategists’ mean proportion of giving in the UG, and proportion of giving in the DG. Error IQ was higher than nonstrategists’ (105.70 vs. 95.86), t(384) = 7.66, bars show 95% confidence interval. n = 386.

Yamagishi et al. PNAS Early Edition | 3of6 Downloaded by guest on October 1, 2021 social norms. Contrary to this, our findings support an alterna- clarify the evolutionary and neuropsychological foundations of tive scenario such that cognitive control operates in the opposite rejection of unfair UG offers, particularly in relation to internalized direction—that is, to control the automatic demands for prosocial social norms of nonaggression versus norms of punishing norm- behavior. The UG is often considered a game in which proposers violators. Another threat to the validity of this interpretation strategically adjust their choices to accommodate anticipated re- concerns the possible confound of general intelligence in perfor- sponses (8, 9). The player’s choices in the DG are considered a mance in the Mac game, as implied by the correlation between IQ straightforward reflection of social preferences, uncontaminated by scores and Mac game scores (r = 0.355, P < 0.0001). However, strategic considerations (8, 9, 41, 42). Our findings challenge this correlations between Mac game scores and strategic behavior or widely accepted understanding. Contrary to the previous un- DG giving remain highly significant even after adjusting for IQ, derstanding, the proposer’s choices in the UG are not related to indicating that the Mac game’s assessment regarding strategic the cortical thickness of the DLPFC or strategic reasoning choices in the DG is independent of general intelligence. ability. Furthermore, both the cortical thickness of the DLPFC Finally, earlier studies found a positive relationship between and strategic reasoning ability were negatively associated with young children’s performance on a cognitive, but not emotional, giving in the DG, where strategic reasoning has been understood theory of mind test and the levels of proposed UG giving (50, 51). to play no role (8, 9, 41, 42). Although these findings are counter- Although these findings are inconsistent with the current lack of intuitive because they oppose predictions based on the reflective effect of strategic reasoning in the UG, the inconsistency may be model of prosociality as depicted in Fig. 1A, they are consistent with due to age differences in the earlier study population (8, 50, 51) the intuitive prosociality predictions shown in Fig. 1B.Furthermore, versus the current study. Cognitive control of normative demands these findings are consistent with the findings of a transcranial di- in the DG may require the strategic use or goal-directed mobili- rect current stimulation (tDCS) study by Ruff et al. (13), who zation of theory of mind (52). Very young children, before ac- reported that anodal tDCS of the right LPFC increased strategic quiring cognitive theory of mind, do not make fair offers in either behavior (UG–DG, which the authors called “sanction-induced the UG or the DG (50, 51). As they grow and acquire cognitive transfer”) and decreased giving in the DG while not affecting giving theory of mind, they start adjusting their behavior to anticipated in the UG. Cathodal tDCS produced opposite effects, reducing responses, resulting in fair and acceptable offers in the UG. Up to strategic behavior and improving giving in the DG, again while not this stage of cognitive development, acquisition of cognitive theory affecting giving in the UG. of mind promotes fair offers in the UG and engenders the pattern Theoretical and methodological issues and limitations of the depicted in Fig. 1A. By the age of 5, children start internalizing current study need to be discussed here. The first concerns the others’ responses and give more in the DG when they are monitored relationship between cognitive control and cognitive assessment by the experimenter than when they are not (53). The internalization of the situation. We argued that participants with a thick DLPFC of social norms and anticipated responses by others further develops and an ability for strategic reasoning are better at discerning throughout elementary school, as shown by the finding that DG irrelevance of social norms on their own future welfare when giving increases first if the player’s choice is observed by other facing an experiment where anonymity with no future interac- players, followed by a gradual increase in the situations where the tions was ensured. Those who do not see the irrelevance of social player’s choice is anonymous (54). Generally, children’s giving in norms for their future welfare in one-shot laboratory experi- the DG increases as they age (54–56). Consequently, default ments will automatically follow the demands of social norms. We giving in the DG increases and the pattern shifts from that in Fig. used the term “cognitive control” to refer to the individual’s 1A to that in Fig. 1B. At the stage of full internalization of social inclination toward assessment of the consequences of their own norms, taking a large share in the DG requires goal-directed action for immediate and long-term future welfare. One model mobilization of theory of mind. Only those who understand the of the evolution of dual-process (intuitive versus deliberative) irrelevance of norm-abiding behavior in a one-shot, anonymous decision makers (43) indicates that the level of this inclination DG control the internalized demand for norm-abiding behavior depends on the nature of the social environment surrounding and freely take a majority share. Recent studies of theory of mind individuals, such as the relative frequency of repeated versus reveal that many adults who are capable of theory of mind do not one-shot interactions. It is also unclear whether cortical thick- use it routinely (57, 58). Given DLPFC involvement in DG decision ness of the DLPFC translates into cortical function and plays a making, involvement of executive control in adult use of theory of role in decision making in the UG and DG. A previous study (8) mind (58), and the established relationship between the DLPFC found a positive relationship between strategic behavior and and executive control (59), future studies of human prosociality cortical thickness and functional activity of the DLPFC. It is thus should address the role of strategic (or goal-oriented) use of theory plausible to suggest, based on current findings, that strategists of mind in controlling adherence to default prosocial choices. not only have thicker DLPFCs but also recruit their deliberative Deliberate pursuit of self-interest without respect to social norms functions more heavily than nonstrategists. The tDCS study by yields better outcomes for the self than automatically observing so- Ruff et al. (13) is of particular importance here because it shows cial norms, insofar as the player can accurately assess the anonymity that an induced increase in LPFC excitability is causally related to and one-shot nature of the experiment. Assessment of the nature of reduced DG but not UG giving, suggesting that DLPFC activity is real-world social interactions is more difficult than in experiments, causally related to norm-violating behavior in the DG. A finding and making errors in assessment is always possible. Erroneously that disrupting the DLPFC made UG responders more accepting assessing that selfish behavior will not be noticed in some situations, of unfair offers (10) seems to contradict the current findings if but when such behavior is actually detected, it can lead to devas- rejection of unfair UG offers is considered a norm-enforcing tating consequences—punishment or ostracism from the community. behavior. However, the latter assumption has been increasingly Individuals who are poor at assessing the nature of social situations criticized (40, 44–49), and the critics provide evidence that re- may give up chances to selfishly maximize profit to reduce the jection of unfair offers as emotion-based aggression is against the probability of committing serious social errors (16, 17). From this dominant social norm, at least in highly industrialized societies (40, logic of error management (17), suppressing internalized demands 44, 45, 48, 49), and is instead a long-term strategy to protect oneself of social norms to pursue immediate self-interest in appropriate from possible future exploitation by others (47, 48). From this situations, such as the DG, can be adaptive for capable strategists alternative view, not rejecting and accepting the disadvantageous who can discern salient situational aspects. However, this strategy can offer is an internalized and intuitive norm-abiding behavior that be maladaptive for those more prone to make mistakes in assessing must be controlled to execute strategies to protect the player’s such situations. Those individuals avoid making errors with poten- reputation and standing (47–49). More studies are needed to tially serious consequences by always adhering to internalized

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1523940113 Yamagishi et al. Downloaded by guest on October 1, 2021 demands of social norms. This can be a more adaptive strategy. The the responses of all relevant players. We developed a task called the Mac adaptive advantages of intuitive prosociality and deliberative decision game to measure the player’s use of strategic choices in an interdependent making discussed above also flow from a recently proposed formal situation. In this task, we defined a strategic choice as one that maximizes ’ model of the evolution of dual-process decision makers who co- one s own reward based on the expected choices of two other players, based operate by default but spend cognitive cost to assess the benefit of on the assumption that other people prefer more money than less. To suc- cessfully make the strategic choice requires an understanding of the nature noncooperation (43). This model asserts that everyone spends a of interdependence between players—that is, that what a player gets de- cognitive cost to assess the benefit of one type of decision against pends on other players’ choices. Strategic choices in this sense require the the other when the strategic nature of the situation is obvious, as in unsolicited use of theory of mind. Studies of theory of mind with adult the UG. In other situations, as in the DG, whether or not to spend participants (57, 58) indicate that some adults do not in an unsolicited cognitive cost to discern the true one-shot nature of the game de- manner use a theory of mind that most children over the age of 6 can use, pends on the individual’s inclination toward intuitive prosociality, even when it must be used to successfully perform a task. Among adults, being which reflects the nature of the social environment surrounding the able to understand the actions of others in terms of their mental states does player in real life. not necessarily mean that this ability is used voluntarily in everyday decision- We would also like to emphasize that our finding that the cortical making. In the Mac game, correct prediction of the other players’ choices is thickness of the DLPFC and strategic reasoning ability reduces giving required to earn the most money. Another feature of the Mac game is that ’ in the DG but does not affect giving in the UG makes us reconsider the players must understand that they can influence other players choices. For the way we interpret the findings of the studies using these games and example, player A has a choice between a1 and a2.IfAchoosesa1,Bearns$10 (not A’s money). If A chooses a , B earns $1. It is thus in B’sbestinterestto the way we address the evolutionary puzzle of human prosociality. 2 induce A to choose a1 instead of a2. B has the opportunity to indicate to A that Methods he or she will pay A $1 if A chooses a1 and nothing if A chooses a2.PayingA$1 for choosing a will induce A to choose that option, and consequently, B will – 1 Sample. Five hundred and sixty-four nonstudent residents (ages 19 59 as of earn $10. In this simplified example, paying $1 to A so that he or she chooses January 2012) living in Machida, a suburb of Tokyo, and its surroundings par- a1 is a strategic choice, which requires that B understands he or she can in- ticipated in the initial wave of a longitudinal study consisting of eight waves to fluence A’s choice such that A will choose a1. The Mac game involves two steps date and continuing about 3.5 y from its inauguration in 2013. Of these 564, of this type of reasoning and is thus more complex than this simple example. It 411 (197 female) participated in both a DG experiment and a UG experiment is explained fully in SI Methods: The Mac Game and Dataset S1 for the actual and submitted to brain structural scans. The Mac game was conducted with 386 instructions used in the study. participants (185 female), and the Machiavellianism Scale was administered to IQ. We administered an IQ test (Kyoto University NX15 (60) in wave 1, which all 411 participants who participated in the two games and received brain scans. took about 1.5 h, including instructions. All experimental protocols were approved by the ethics committee at the General trust. Yamagishi’s general trust scale (40) was administered three times COGNITIVE SCIENCES Brain Science Institute, Tamagawa University, where the study was conducted, PSYCHOLOGICAL AND in waves 1, 3, and 6, and the mean of the three measures of general trust after according to the requirements of the Declaration of Helsinki; the methods were standardizing each was used as the overall measure of general trust. carried out in accordance with approved guidelines. One informed consent form was signed by each participant in the first wave of the study to confirm MRI Data Acquisition. MRI images were recorded on a 3 Tesla Siemens Trio A their overall agreement to participate, and another was signed separately to Tim MRI scanner. High-resolution anatomical images were acquired using a provide permission for the brain scan. Data from this project have been used elsewhere, but the comparison of proposals in the UG and giving in the DG has T1-weighted 3D magnetization prepared rapid acquisition gradient echo × not been reported in earlier studies. sequence (repetition time, 2,000 ms; echo time, 1.98 ms; field of view, 256 256 mm; number of slices, 192; voxel size, 1 × 1 × 1 mm; average, 3 times). Games and Measures. DG. All participants first played a one-shot DG (DG1) as dictators. Each participant MRI Data Analysis. Gray matter thickness of the regions labeled as the middle was given an endowment of JPY 1,000 and decided how much of the endowment frontal gyrus, middle frontal sulcus, superior frontal sulcus, inferior frontal sulcus, to give to their partner (the recipient). Participants actually received the money they and was extracted as the volume of the DLPFC. Gray allocatedtothemselvesaswellasthemoney allocated by a randomly matched matter thickness values were estimated for the five regions on both hemispheres dictator. Following DG1, participants played similar games six times as a dictator, using the FreeSurfer package (version 5.1.0 for Linux CentOS 4; surfer.nmr.mgh. with a different recipient each time. The size of the endowment varied each time, harvard.edu). Three T1-weighted MRI images were registered and averaged for ranging from JPY 300 to JPY 1,300 (300, 400, 600, 700, 1,200, and 1,300). Participants each participant. The mean images were submitted to a fully automated pro- were told that they would receive payment once as a dictator and once as a re- cedure that reconstructed 3D models of the pial surface and the boundary cipient. We used the mean proportion of endowment given to the recipient in the between the gray and white matter. The initial part of the reconstruction seven games as participant’s giving in the DG. procedure included registration to a common stereotactic space, image in- UG. Each participant played the game once as a proposer and once as a tensity correction for magnetic field inhomogeneity, and skull stripping. The responder and was actually paid the sum of the money received in two of the boundary between the gray and white matter for each hemisphere was seg- games. All participants played first as proposers, without knowing that they mented, tessellated, and corrected for topological errors. The resulting surface would later play the game in the other role. As proposers, participants de- models of the boundary were aligned to a surface template by matching the cided how much of an endowment of JPY 1,500 to provide to a randomly gyral and sulcal patterns to the template. We computed the cortical thickness of matched responder, in increments of JPY 100. After all participants made the regions of interest using cortical parcellation based on the Destrieux Atlas their decisions as proposers, they were told that they would play the game (36), which divides each cortical hemisphere into 74 regions. Gray matter again in the other role and decide whether to accept or reject each of 16 thickness was calculated as the closest distance between the gray/white matter possible proposals (JPY 0 to JPY 1500) made by a newly matched proposer. boundary and the pial surface. We used the Destrieux Atlas to match the DLPFC When all participants had played the game in both roles, random pairs were with the regions specified in the atlas. Specifically, the DLPFC was matched with formed twice, once where the participant was the proposer and once as the regions 15 (middle frontal gyrus), 16 (superior frontal gyrus), 52 (inferior frontal responder. Each participant was paid according to the choice made in each sulcus), 53 (middle frontal sulcus), and 54 (superior frontal sulcus) in the Destrieux pairing. We used the proportion of JPY 1,500 the participant gave the re- Atlas (see Fig. 2A). All relevant data used for analysis are included in Dataset S2. sponder as proposed giving in the UG. Mac game. A strategic choice in an interdependent situation is goal-oriented ACKNOWLEDGMENTS. The studies reported in this paper were supported by and is realized by reading the intentions of other individuals. Niccolo Grants-in-aid 23223003 and 15H05730. We thank Profs. Minoru Kimura and Machiavelli is the most well-known advocate of strategic choices, or choices Masamichi Sakagami (Tamagawa University Brain Science Institute) for their that maximize one’s own benefits (i.e., the goal) by correctly determining continuous support.

1. Bowles S, Gintis H (2011) A Cooperative Species (Princeton Univ Press, Princeton). 4. Tomasello M, Melis AP, Tennie C, Wyman E, Hermann E (2012) Two key steps in the 2. Boyd R, Richerson PJ (2009) Culture and the evolution of human cooperation. Philos evolution of human cooperation: The interdependence hypothesis. Curr Anthropol Trans R Soc Lond B Biol Sci 364(1533):3281–3288. 53(6):673–692. 3. Nowak M (2011) SuperCooperators: Altruism, Evolution, and Why We Need Each 5. Mead G (1934) Mind, Self, and Society (University of Chicago Press, Chicago). Other to Succeed (Free Press, New York). 6. Parsons T (1937) The Structure of Social Action (McGraw Hill, New York).

Yamagishi et al. PNAS Early Edition | 5of6 Downloaded by guest on October 1, 2021 7. Morishima Y, Schunk D, Bruhin A, Ruff CC, Fehr E (2012) Linking brain structure and 33. Miller EK, Cohen JD (2001) An integrative theory of prefrontal cortex function. Annu activation in temporoparietal junction to explain the neurobiology of human altru- Rev Neurosci 24:167–202. ism. Neuron 75(1):73–79. 34. Nakamura T, et al. (2012) Development and validation of a Japanese version of the 8. Steinbeis N, Bernhardt BC, Singer T (2012) Impulse control and underlying functions Machiavellianism Scale. Jpn J Pers 20(3):233–235. of the left DLPFC mediate age-related and age-independent individual differences in 35. Christie R, Geis FL (1970) Studies in Machiavellianism (Academic, New York). strategic social behavior. Neuron 73(5):1040–1051. 36. Destrieux C, Fischl B, Dale A, Halgren E (2010) Automatic parcellation of human 9. Spitzer M, Fischbacher U, Herrnberger B, Grön G, Fehr E (2007) The neural signature cortical gyri and sulci using standard anatomical nomenclature. Neuroimage 53(1): of social norm compliance. Neuron 56(1):185–196. 1–15. 10. Knoch D, Pascual-Leone A, Meyer K, Treyer V, Fehr E (2006) Diminishing reciprocal 37. Barbey AK, Colom R, Grafman J (2013) Dorsolateral prefrontal contributions to hu- fairness by disrupting the right prefrontal cortex. Science 314(5800):829–832. man intelligence. Neuropsychologia 51(7):1361–1369. 11. Knoch D, et al. (2008) Studying the neurobiology of social interaction with trans- 38. Passingham RE, Wise SP (2012) The Neurobiology of the Prefrontal Cortex: Anatomy, cranial direct current stimulation–The example of punishing unfairness. Cereb Cortex Evolution, and the Origin of Thought (Oxford Univ Press, Oxford). 18(9):1987–1990. 39. Dahling JJ, Whitaker BG, Levy PE (2009) The development and validation of a new 12. Baumgartner T, Knoch D, Hotz P, Eisenegger C, Fehr E (2011) Dorsolateral and ven- Machiavellianism scale. J Manage 35(2):219–257. tromedial prefrontal cortex orchestrate normative choice. Nat Neurosci 14(11): 40. Yamagishi T, et al. (2013) Is behavioral pro-sociality game-specific? Pro-social pref- – 1468 1474. erence and expectations of pro-sociality. Organ Behav Hum Dec 120(2):260–271. 13. Ruff CC, Ugazio G, Fehr E (2013) Changing social norm compliance with noninvasive 41. Andreoni J, Miller J (2002) Giving according to GARP: An experimental test of the – brain stimulation. Science 342(6157):482 484. consistency of preferences for altruism. Econometrica 70(2):737–753. – 14. Zaki J, Mitchell JP (2013) Intuitive prosociality. Curr Dir Psychol Sci 22(2):466 470. 42. Fehr E, Schmidt KM (1999) A theory of fairness, competition and cooperation. QJ 15. Rand DG, Greene JD, Nowak MA (2012) Spontaneous giving and calculated greed. Econ 114(3):817–868. – Nature 489(7416):427 430. 43. Bear A, Rand DG (2016) Intuition, deliberation, and the evolution of cooperation. 16. Rand DG, et al. (2014) Social heuristics shape intuitive cooperation. Nat Commun 5:3677. Proc Natl Acad Sci USA 113(4):936–941. 17. Yamagishi T, Terai S, Kiyonari T, Mifune N, Kanazawa S (2007) The social exchange 44. Eriksson K, Andersson PA, Strimling P (2015) Moderators of the disapproval of peer pun- heuristic: Managing errors in social exchange. Rationality Soc 19(3):259–291. ishment. Group Process Intergroup Relat; epub ahead of print 10.1177/1368430215583519. 18. Cappelletti D, Goth W, Ploner M (2011) Being of two minds: Ultimatum offers under 45. Espín AM, Brañas-Garza P, Herrmann B, Gamella JF (2012) Patient and impatient cognitive constraints. J Econ Psychol 32(6):940–950. punishers of free-riders. Proc Biol Sci 279(1749):4923–4928. 19. Cornelissen G, Dewitte S, Warlop L (2011) Are social value orientations expressed 46. Strimling P, Eriksson K (2014) Regulating the regulation: Norms about punishment. automatically? Decision making in the dictator game. Pers Soc Psychol Bull 37(8): Reward and Punishment in Social Dilemmas, eds van Lange PAM, Rockenbach B, 1080–1090. Yamagishi T (Oxford Univ Press, Oxford), pp 52–69. 20. Schulz JF, Fischbacher U, Thöni C, Utikal V (2014) Affect and fairness: Dictator games 47. Burnham TC (2007) High-testosterone men reject low ultimatum game offers. Proc under cognitive load. J Econ Psychol 41:77–87. Biol Sci 274(1623):2327–2330. 21. Roch SG, Lane JAS, Samuelson CD, Allison ST, Dent JL (2000) Cognitive load and the 48. Yamagishi T, et al. (2009) The private rejection of unfair offers and emotional com- equality heuristic: A two-stage model of resource overconsumption in small groups. mitment. Proc Natl Acad Sci USA 106(28):11520–11523. Organ Behav Hum Decis Process 83(2):185–212. 49. Yamagishi T, et al. (2012) Rejection of unfair offers in the ultimatum game is no 22. Lotz S (2015) Spontaneous giving under structural inequality: Intuition promotes evidence of strong reciprocity. Proc Natl Acad Sci USA 109(50):20364–20368. cooperation in asymmetric social dilemmas. PLoS One 10(7):e0131562. 50. Takagishi H, Kameshima S, Schug J, Koizumi M, Yamagishi T (2010) Theory of mind 23. Rand DG, Kraft-Todd G, Gruber J (2015) The collective benefits of feeling good and – letting go: Positive emotion and (dis)inhibition interact to predict cooperative be- enhances preference for fairness. J Exp Child Psychol 105(1-2):130 137. havior. PLoS One 10(1):e0117426. 51. Takagishi H, et al. (2014) The role of cognitive and emotional perspective taking in 24. Tabibnia G, Satpute AB, Lieberman MD (2008) The sunny side of fairness: Preference economic decision making in the ultimatum game. PLoS One 9(9):e108462. ’ for fairness activates reward circuitry (and disregarding unfairness activates self- 52. Sher I, Koenig M, Rustichini A (2014) Children s strategic theory of mind. Proc Natl – control circuitry). Psychol Sci 19(4):339–347. Acad Sci USA 111(37):13307 13312. 25. Camerer C (2003) Behavioral Game Theory (Princeton Univ Press, Princeton). 53. Fujii T, Takagishi H, Koizumi M, Okada H (2015) The effect of direct and indirect 26. Sanfey AG, Rilling JK, Aronson JA, Nystrom LE, Cohen JD (2003) The neural basis of monitoring on generosity among preschoolers. Sci Rep 5:9025. economic decision-making in the Ultimatum Game. Science 300(5626):1755–1758. 54. Takagishi H, et al. (2015) The development of the effect of peer monitoring on 27. Takagishi H, Fujii T, Kameshima S, Koizumi M, Takahashi T (2009) Salivary alpha- generosity differs among elementary school-age boys and girls. Front Psychol 6:895. ’ amylase levels and rejection of unfair offers in the ultimatum game. Neuroendocrinol 55. Benenson JF, Pascoe J, Radmore N (2007) Children s altruistic behavior in the dictator Lett 30(5):643–646. game. Evol Hum Behav 28(3):168–175. 28. Dalwani M, et al. (2011) Reduced cortical gray matter volume in male adolescents 56. Fehr E, Bernhard H, Rockenbach B (2008) Egalitarianism in young children. Nature with substance and conduct problems. Drug Alcohol Depend 118(2-3):295–305. 454(7208):1079–1083. 29. Mostofsky SH, Cooper KL, Kates WR, Denckla MB, Kaufmann WE (2002) Smaller 57. Keysar B, Lin S, Barr DJ (2003) Limits on theory of mind use in adults. Cognition 89(1): prefrontal and premotor volumes in boys with attention-deficit/hyperactivity disor- 25–41. der. Biol Psychiatry 52(8):785–794. 58. Apperly IA, et al. (2010) Why are there limits on theory of mind use? Evidence from 30. Weygandt M, et al. (2013) The role of neural impulse control mechanisms for dietary adults’ ability to follow instructions from an ignorant speaker. Q J Exp Psychol (Hove) success in obesity. Neuroimage 83:669–678. 63(6):1201–1217. 31. Figner B, et al. (2010) Lateral prefrontal cortex and self-control in intertemporal 59. Miller BL, Cummings JL (2007) The Human Frontal Lobes: Functions and Disorders choice. Nat Neurosci 13(5):538–539. (Guilford, New York). 32. Hare TA, Camerer CF, Rangel A (2009) Self-control in decision-making involves 60. Osaka R, Umemoto T (1984) Revised Kyoto University NX-15: The Second Version of modulation of the vmPFC valuation system. Science 324(5927):646–648. an Intelligence Test (Taisei Shuppan, Tokyo). Japanese.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1523940113 Yamagishi et al. Downloaded by guest on October 1, 2021