1. Prosocial behavior and moral motivation: A mystery hiding in plain sight

Recent decades have witnessed an explosion of fascinating scientific work concerning prosocial, cooperative, and altruistic behavior in both human and nonhuman organisms. Researchers have now documented an impressive range of instances in which other animals, especially nonhuman primates, exhibit sympathetic and empathetic concern for others, aggression in response to failures of reciprocation, inequity aversion, and other behaviors that would in humans be regarded as characteristic responses to the demands of morality.

Indeed, the pace of recent progress in this area has sometimes obscured the depth of the explanatory challenges that remain. Primatologist Frans de Waal, for example, though careful to concede that nonhuman primates fall short of “full-blown” human morality, nonetheless repeatedly insists that the existing evidence already demonstrates that the “foundations,” “pillars,” or “building blocks” of human moral psychology also exist in nonhuman primates and are explicable in fairly straightforward evolutionary terms (e.g., De Waal Reference De Waal1996; De Waal et al. Reference De Waal, Ober, Macedo, Ober and Macedo2006). But many moral philosophers have resisted this suggestion, arguing that these admittedly fascinating discoveries fail to establish that other organisms share the most central and distinctive features of human moral psychology. Perhaps most importantly, humans experience the demands of morality as somehow imposed on us externally: We do not simply enjoy or prefer to act in ways that satisfy the demands of morality; we see ourselves as obligated to do so regardless of our subjective preferences and desires, and we regard such demands as imposing unconditional obligations not only on ourselves, but also on any and all agents whatsoever, regardless of their preferences and desires. As philosopher Richard Joyce notes, even if we concede that chimpanzees have abundant “inhibitions, aversions, and inclinations” motivating prosocial behavior, and even that these are internalized into what De Waal calls “a sense of social regularity,” we must still ask,

Even if one chimpanzee who punishes another for violating her social expectations is motivated (as De Waal suggests) by concern for the welfare of the larger community, she need not experience this as a matter of respecting or enforcing any kind of objective or externally imposed moral order, rather than simply discouraging a fellow group member from repeating a violation of what her own subjective preferences (regarding community welfare or anything else) happen to be. De Waal's slide between these distinct possibilities is evident even in his explicit definition of nonhuman primates’ “sense of social regularity” as a set of “expectations about the way in which oneself (or others) ought to be treated and how resources ought to be divided” whose violation provokes protest or punishment (De Waal Reference De Waal1996, p. 95, my emphasis).

Of course, other organisms may well experience demands for prosocial emotions or behaviors as somehow externally imposed upon them, but moral philosophers are right to insist that simply demonstrating the presence of such emotions and behaviors fails to establish that they are motivated in other organisms by any sense of obligation and/or prohibition rather than inclination and aversion. Although De Waal sometimes claims (e.g., 1996, p. 210; see also Suchak et al. Reference Suchak, Eppley, Campbell, Feldman, Quarles and de Waal2016) that it is simply parsimonious to assume that similar behavior in similar circumstances arises from similar motivations in nonhuman and human organisms, we would surely want a convincing explanation of why humans (or any organisms) would externalize moral motivations, demands, and obligations in this way before deciding whether we think that any particular nonhuman organisms do so.

After all, from an evolutionary point of view, such externalization or objectification is deeply puzzling even in our own case: It seems that mere subjective preferences for social interactions with those who are kind, generous, fair, loyal, and so forth, and for avoiding those who are cruel, selfish, deceitful, exploitative, and so forth, would equally effectively guide us towards things that are good for us and away from things that are bad for us, just as it does in the case of our subjective preferences for keeping our heads dry, our tummies full, our orgasms frequent, and for the vast majority of other fitness-enhancing behaviors in which we routinely engage. The real challenge from an evolutionary point of view, then, is not to explain why we prefer interacting with kind and generous conspecifics to cruel and selfish ones, but instead why we treat moral demands and obligations as anything more than such preferences – as anything more than how we ourselves happen to want to behave and/or what we find appealing in others. Why, that is, do we experience our attitude towards Nazis or slavers any differently than an aversion to kale or a preference for chocolate ice cream over vanilla?

In what follows, I will first appeal to recent findings in cognitive science to more precisely characterize those features of our moral experience and cognition that require such further explanation, and I will argue that existing evolutionary proposals concerning human moral psychology simply do not help meet the distinctive explanatory demand that emerges. I will then go on to argue that externalizing moral demands in this way enabled us to safely discover and take advantage of novel opportunities for productive cooperative interaction by establishing a crucial connection between our own motivation to conform to any given distinctively moral norm of behavior and the extent to which we demand that others conform to that same norm if we are to view them as attractive or desirable partners in social interaction more generally. Externalization thus protected our expanding prosocial motivations and commitments from exploitation by ensuring correlated interaction between ourselves and those who share those same motivations and commitments, even as new norms governing such behavior were introduced and existing norms were extended in novel ways and applied to novel circumstances. I will then seek to show how this proposal both explains and unifies a number of important further findings from recent experimental cognitive science and how the need to recruit others and advertise ourselves as attractive partners in such forms of interaction structures much more of our resulting social behavior than we realize, including features of our social lives so familiar that we have difficulty recognizing that they require any explanation at all. But before asking whether this proposal offers a convincing explanation for the distinctive character of human moral motivation, let us first seek a more precise characterization of just what phenomena we might be seeking to explain by invoking it.

2. Acquiring the explanatory target

Recent work in cognitive science allows us to characterize our experience of moral demands, considerations, and judgments with unprecedented precision. We should start by noting that our distinctive experience of moral motivation and cognition is not simply a function of either its rule-governed or normative character, for human societies also characteristically recognize normative rules of the sort we regard as mere social conventions (such as norms of etiquette or fashion) to which the further distinctive characteristics of moral demands or norms are not attributed. Social psychologists have documented that, across an impressively heterogeneous variety of societies and cultures, children between the ages of 2.5 and 3 years begin to reliably and systematically distinguish norms and transgressions recognized as genuinely moral in character, such as pulling hair or stealing, from those regarded instead as merely conventional, such as talking out of turn or drinking soup from a bowl (Smetana Reference Smetana, Killen and Smetana2006; Turiel Reference Turiel1983; Turiel et al. Reference Turiel, Killen, Helwig, Kagan and Lamb1987). Like moral demands, social conventions or rules are understood normatively, but moral violations are nonetheless regarded as more serious and more deserving of punishment than violations of merely conventional norms, and their wrongness is typically explained by appeal to considerations of harm, fairness, justice, rights, or the welfare of victims, whereas comparable explanations for conventional norms appeal instead to considerations of social utility or acceptability. Moral norms are systematically viewed both as more “generalizable” (i.e., applicable to people in other places and historical periods, whether or not they care about them) than mere social conventions and as “authority-independent,” meaning that their force cannot be suspended by an appropriate individual or institution: for example, children report that it is not wrong to chew gum in class if a teacher has no rule against it, but that hitting another student is wrong no matter what the teacher says. Nucci (Reference Nucci and Harding1986) showed that this distinction is salient even to Amish and Mennonite teenagers with respect to God's authority: 100% of these subjects agreed that it would not be wrong to leave their heads uncovered or to ignore the prescribed day of worship if God had made no rule against this, but more than 80% insisted that stealing or hitting would still be wrong even if God had made no rule forbidding it. Some fascinating evidence suggests that psychopaths may constitute the exception that proves this rule (treating all social norms and rules as purely conventional in character; see Blair Reference Blair1995; Nichols Reference Nichols2004), but this claim remains controversial (cf. Aharoni, et al. Reference Aharoni, Sinott-Armstrong and Kiehl2012).

More recent work, however, has challenged whether moral norms must concern or be justified by appeal to considerations of harm, fairness, justice, rights, or welfare. Haidt et al. (Reference Haidt, Koller and Dias1993), for instance, showed that groups of low socioeconomic status (SES) subjects in both Brazil and the United States judged harmless transgressions such as privately washing the toilet bowl with the national flag and privately masturbating with a dead chicken to be both serious and distinctively moral in character. In a study of children in traditionally Arab villages in Israel, Nisan (Reference Nisan1987) found that a moralized response was provoked by norm violations of all kinds, including mixed-sex bathing and addressing a teacher by his first name. Nichols (Reference Nichols2004) showed that rules of etiquette prohibiting disgusting activities in particular were judged by American children to be serious, generalizable, and authority-independent, while American college students judged the same prohibitions to be serious and authority-independent, though not generalizable. Such findings illustrate that norms need not concern harm, justice, or rights in order to exhibit further characteristics associated with distinctively moral normativity. Moreover, Kelly et al. (Reference Kelly, Stich, Haley, Eng and Fessler2007, pp. 118–21) show in addition that scenarios that do involve harm, justice, and rights-based violations (beyond those likely familiar to young children from a school or playground) can be presented in ways that undermine subjects’ commitments to features like their generalizability or authority-independence.Footnote ¹ These results suggest that although moral norms may be frequently or even prototypically concerned with harm, fairness, justice, rights, or welfare, it would be a mistake to treat such concerns as a defining feature of moral norms themselves. Indeed, we might hope for a satisfying account of the reliable and robustly cross-cultural ontogenetic emergence of a distinction between merely conventional and genuinely moral norms to explain why the latter are frequently but not invariably concerned with harm, fairness, justice, rights, or welfare.

More recently, Nichols and Folds-Bennett (Reference Nichols and Folds-Bennett2003) have shown that in addition to distinguishing moral from merely conventional norms and violations, even very young children also distinguish moral properties from “response-dependent” properties (like “icky,” “yummy,” or “boring”) that are seen as constituted by our own reactions to features or aspects of the world. In these experiments, 4- to 6-year-old children did not treat the existence of such response-dependent properties as depending on the presence of actual responders, holding, for example, that grapes were yummy even before anyone was around to taste them. But faced with disagreement, children judged that grapes were yummy and cleaning house boring only “for some people,” whereas children facing precisely parallel disagreement did not judge that one monkey helping another who is hurt is good only “for some people,” but instead that it is good “for real.”

Perhaps most revealing of all, Goodwin and Darley (Reference Goodwin and Darley2008; Reference Goodwin and Darley2012) show that subjects reliably locate moral judgments at a particular point along a scale of increasing objectivity ranging from judgments of taste or preference (“Frank Sinatra was a better singer than is Michael Bolton”) to judgments of social convention (“Wearing pajamas and bath robe to a seminar meeting is wrong behavior”) to moral judgments (“Robbing a bank in order to pay for an expensive holiday is a morally bad action”) to judgments of straightforward empirical or scientific fact (“Boston [MA] is farther north than Los Angeles [CA]”). Specifying such objectivity as a question of (i) whether disagreements require that at least one party be mistaken or (ii) whether there can be a right answer regarding whether the statement in question is true, these authors found that:

These categorical differences in subjects’ willingness to tolerate the possibility of disagreement without error concerning various kinds of judgments provide perhaps the clearest way to characterize the sense in which we treat moral norms and judgments as systematically more objective than judgments of taste or preference or judgments of social convention. Note, however, that along with such categorical differences, Goodwin and Darley (Reference Goodwin and Darley2012) also found considerable variation in the degree of objectivity assigned to particular ethical statements by different experimental subjects, as well as in the relative degree of objectivity assigned to moral judgments with different sorts of content – the statement that it is wrong to rob a bank, for example, was reliably judged far more objective than the statement that anonymous giving is good. Such variation has been particularly emphasized by Wright et al. (Reference Wright, Grandjean and McWhite2013; Reference Wright, McWhite, Grandjean, Lombrozo, Nichols and Knobe2014) in work that confirms Goodwin and Darley's central findings.

Having specified with somewhat greater precision the features of distinctively moral motivation for which any satisfying explanation of moral externalization or objectification will have to account, I will now go on to argue that existing efforts to understand aspects or features of human moral psychology in evolutionary terms do not help satisfy these explanatory demands, whether or not they are explicitly addressed to them. That is, extant attempts to solve the puzzle explicitly are unpersuasive, while other influential views concerning the function(s) of human moral psychology (even if correct) do not actually help explain why our experience of moral motivation and moral norms systematically differs from that of mere preferences, desires, and merely conventional norms in the salient ways just described.

3. Illuminating failures

Following his own characterization of our puzzle (see previous sect. 1), Joyce (Reference Joyce2006) himself offers two independent lines of explanation for the distinctively externalized character of our moral experience. He begins with the familiar suggestion that moral considerations will motivate us more effectively if we take them to be perceptions of the world's own qualities rather than our subjective reactions to it. But this claim would seem to turn on a pair of confusions.

First, truly adaptive moral motivation would not be as strong as possible but would instead provide the right degree of motivation as balanced against a wide array of other evolutionarily important motivational impulses: Even if acting in morally praiseworthy ways turns out to be important from an evolutionary point of view, so are alleviating hunger, avoiding predation, having sex, and much else besides. Thus, simply increasing the motivational force of any and all moral considerations across the board is not advantageous unless we implausibly suppose that it always favors our evolutionary interests to behave in what we see as the most morally admirable way possible rather than balancing moral considerations against other evolutionarily significant motivations like hunger, fear, and sexual desire. Note that much the same problem afflicts Daniel Dennett's influential proposal (see Dennett Reference Dennett1995, Ch. 17) that moral considerations serve as “conversation stoppers” terminating what would otherwise be endless chains of internal deliberation about what to do. Even more clearly than Joyce's, Dennett's proposal suggests that it is somehow advantageous for moral considerations to trump all other motivational impulses, and this seems to describe neither a plausible strategy for evolutionary survival nor our actual internal deliberations: Rare indeed are those for whom the recognition that an act would less than fully satisfy the demands of morality precludes any further deliberative consideration of whether or not to perform it.

Even if this challenge can be overcome, however, another looms for both Joyce and Dennett, because it seems clear in any case that mere subjective states can be as strongly motivating as any others. After all, pain is a paradigmatically subjective response to the world, but there are few more powerful motivations for humans than avoiding serious pain, so it seems that even purely subjective states can generate motivation of arbitrarily high magnitude. Hence, even if creatures who externalize moral judgments are indeed more motivated by those judgments than they would be by subjective preferences or desires of equal strength (cf. Young & Durwin Reference Young and Durwin2013), the real question is how and why we came to externalize or objectify moral demands in this way in the first place, and the need for more effective or powerful motivation simply does not help explain that.

At the heart of this second difficulty lies the fact that sufficiently strong desires and subjective preferences (of just the sort we have for one ice cream flavor over another) seem capable of doing the proposed job of moral motivation just as well as moral considerations that we externalize or objectify. And once this general problem is recognized, we can see how it prevents a wide range of familiar proposals concerning the evolutionary function(s) of human moral psychology from helping us resolve our puzzle. Robert Frank (Reference Frank1988) has influentially argued, for example, that our moral emotions serve to commit us to courses of action that would otherwise not serve our immediate self-interests (ensuring that we will not need to actually carry out such courses of action very often so long as others are aware of these commitments). But this does not help address our puzzle, for there is no reason that the motivation for such commitments could not be (and be advertised as) sufficiently compelling preferences, desires, or affective states experienced as entirely subjective in character. Indeed, at least some paradigmatic instances of such commitment devices, such as jealousy, are ones that we do characteristically think of as subjective responses to the world (much like ice cream preferences) rather than responses to any externalized or objectified demand, as Frank perhaps acknowledges in not raising or addressing the question of why moral demands are characteristically externalized or objectified in the first place. Even if Frank's influential proposal is entirely correct, then, it does not help explain the distinctively externalized phenomenology of our moral experience.

Likewise, advocates of so-called strong reciprocity (such as Boyd, Richerson, Bowles, Gintis, and Fehr) have sometimes suggested that the central function of our moral psychology is to motivate costly punishment (especially by third parties) of those who violate prosocial norms, whereas proponents of “indirect reciprocity” (such as Alexander, Nowak, and Sigmund) have instead argued that our moral psychology motivates us to keep track of others’ reputations (and manage our own) for engaging in appropriately altruistic, cooperative, and prosocial behavior. But once again, even if one or both of these influential suggestions are entirely correct, such punishment and/or reputation-tracking could be motivated equally effectively by the sorts of desires and subjective preferences that motivate the vast majority of our fitness-enhancing behaviors: We would simply need to be disposed to punish others when they violated those subjective preferences (concerning our interactions with them or even with third parties) and/or to keep track of (and manage) reputations for satisfying or violating them. Influential extant efforts to explain human moral psychology in evolutionary terms thus seem to fail quite spectacularly at explaining why moral demands can better do the job for which they are or were supposedly needed if they are externalized or objectified in the ways noted previously (sect. 2).Footnote ²

A more promising recent approach by DeScioli and Kurzban (Reference DeScioli and Kurzban2009; Reference DeScioli and Kurzban2013) suggests that the central problem to which our moral psychology offers a solution is the need for bystanders to a conflict to coordinate in choosing sides: These authors suggest that moral condemnation is a form of signaling that allows bystanders to avoid the costs of escalating conflicts by all choosing the same side, while making such side-taking responsive to actions themselves rather than the identity of actors avoids despotism by preventing the power of such allegiance from becoming permanently concentrated in the hands of particular individuals. This elaborate proposal at least promises to explain our characteristic generalization of moral judgments (i.e., why they must apply to everyone in the same way), but unfortunately it does so in a way that ignores the most notable characteristic of those judgments, namely, their pejorative character: We do not merely affiliate with those who share our own moral commitments against those who do not – in addition, we think that we are right to do so and would be wrong to do otherwise. But if mere act-guided coordination of side-taking in bystanders to a conflict is the point of our moral psychology, there is no need for violators to be wrong rather than merely different from us.

To see this point more clearly, note that advocates of strong reciprocity might also sensibly claim that the generalizability of moral judgments is explained by the need for punishment of norm violators no matter the identity of those violators, but we would not thereby explain why norm violators could not be (generalizably) punished simply for being different from us rather than for failing to respect externalized moral obligations imposed on us all. Indeed, DeScioli and Kurzban suggest that our “dynamic coordination” strategy uses a common cultural store of proscribed actions to allow bystanders to coordinate their affiliation in disputes without forming persistent allegiances that allow power to concentrate in the hands of individuals. But the pejorative character of moral judgment would seem to make it harder rather than easier for such coordination to remain dynamic: A person whose conduct is immoral has behaved wrongly rather than just differently than we ourselves would have and is therefore someone with whom we will be less inclined to interact or ally ourselves in the future (see sect. 5 below), no matter the rightness of her cause the next time around. Thus, the pejorative character of genuine moral condemnation actually leads it to fit rather less well than simple morally grounded local affiliation would with the function of dynamic coordination DeScioli and Kurzban propose for our moral psychology more generally.

These challenges might simply magnify our interest in Joyce's second proposal, which begins instead from the suggestion that our general tendency to externalize (or “project”) features of our experience is “the predictable result of natural selection's tight-fisted efficiency” (Joyce Reference Joyce2006, p. 128), as he seeks to illustrate explicitly with respect to sensory qualities such as color and pain. We do not project pain, he suggests, because the point of pain is to orient us towards a problem with the body rather than features of the world that cause pain in us. But we project the redness, warmth, and crackling sound of the fire into or onto the world as we experience it because there is simply nothing to be gained by representing these qualities instead as our subjective responses to the fire's sensation-inducing properties: “A perfectly adequate and simpler solution is if our experience presents itself as being of the world: of the fire being red, hot, and crackling” (Joyce Reference Joyce2006, p. 128). Similarly, Joyce suggests, we project moral qualities into the world, regarding our reaction to a particular behavior as appropriate and the behavior as deserving our praise or blame, for instance, because this is just phenomenologically simpler than representing such judgments to ourselves as subjective responses and because it motivates the relevant fitness-enhancing behaviors equally effectively. If so, the “projection” of moral demands and considerations into or onto the world itself might well be an ancestral condition from which no shift to an externalized or objectified moral psychology was ever required.

The fundamental problem with this proposed line of explanation is both subtle and illuminating, in a way that we can best begin to appreciate by recalling why the ubiquity of altruistic, prosocial, and cooperative behavior in nature once seemed to present a striking challenge to evolutionary theory more generally (one famously posed by Darwin [Reference Darwin1871] himself in Ch. 5 of The Descent of Man). It seems that creatures who systematically contribute to the fitness of others at some cost to their own would be reliably outcompeted or invaded by those who welcome the assistance or sacrifices of others but do not themselves engage in any such costly prosocial or altruistic behavior. The story of how this challenge was ultimately met by the discovery of evolutionary mechanisms like kin selection, reciprocal altruism, and group (or multilevel) selection is by now familiar, but far less widely appreciated is the subsequent recognition that, in a deep and important sense, all of these mechanisms turn out to be variations on a common theme – namely, correlated interaction. As Brian Skyrms (Reference Skyrms1996) and others have emphasized, what is most fundamentally required to protect cooperators from exploitation in a game like the Prisoner's Dilemma or establish cooperation in a game like the Stag Hunt is that there be some mechanism in place to ensure that cooperators or altruists are more likely to interact with one another than with members of the population at large, and the various distinct mechanisms that ensure the persistence of altruistic or exploitable cooperative behavior in nature are just different ways of achieving that. Moreover, further inquiry has revealed a fascinating variety of ways in which correlated interaction can come about in naturally occurring populations ranging from bacteria to chimpanzees: everything from simple population viscosity or structure to the most cognitively complex forms of signaling, individual recognition, recollection of past behavior, and conditional strategies of interaction.

But the need to actively maintain such correlated interaction among cooperators and/or altruists in order to protect their behavioral dispositions from exploitation renders Joyce's (Reference Joyce2006) parallel between the “projection” of moral demands or motivations and the color, heat, and crackling sound of the fire profoundly suspect. As the occasional struggles of those who are color-blind remind us, we have every evolutionary incentive to perceive colors and the like just as fellow group members do, but the same is just not true of prosocial, cooperative, and altruistic motivations, where the possibility of increasing our own payoffs by defecting and/or exploiting our interactive partners is omnipresent. To experience moral demands and considerations as aspects of the external world itself in the way we experience the redness, warmth, and crackling sound of the fire is effectively to presume that others’ experiences of those demands and considerations are identical to our own, thus putting ourselves in danger of exploitation by those who experience the moral demands of a given situation quite differently than we ourselves do (or who would quickly evolve to do so). We therefore cannot remain sensitive to the need to restrict our cooperative and/or altruistic interactions only to those whose perceptions of the moral demands and obligations of a given situation are sufficiently similar to our own while simultaneously “projecting” the source of moral motivation into or onto the world itself as we do the color, warmth, and sound of the fire. Nor, therefore, can we rest content with the suggestion that natural selection's “tight-fisted efficiency” made the “projection,” externalization, or objectification of moral motivation the evolutionary path of least resistance or a likely ancestral condition from which no shift to an externalized moral phenomenology would have been required.

In fact, I will ultimately suggest that externalizing moral motivation was itself favored in ancestral environments precisely because it allowed prosocial, altruistic, and/or cooperative agents to more quickly, efficiently, and effectively correlate their interactions and thereby take better advantage of the adaptive possibilities constituting our transition to a vastly more cooperative form of social life. To evaluate this suggestion, however, we must first consider some characteristic features of human cooperative interaction itself.

4. Homo sapiens: Spontaneous, plastic, domain-general cooperators

Biologically altruistic, prosocial, and cooperative behavior is by no means limited to human beings, but in nonhuman organisms such behavior is characteristically restricted to a relatively narrow and well-defined set of recurring contexts (such as alarm-calling or guarding, cooperative breeding, collective defense, or food-sharing) in which the costs and benefits of quite specific forms of behavior in just these contexts have been sufficiently stable to selectively favor equally specific dispositions to engage in them. By contrast, human cooperative, altruistic, and prosocial behavior is not only domain-general, but also remarkably spontaneous and (like human behavior more generally) extremely facultative, plastic, and flexible: We frequently extend or adapt existing patterns of prosocial or altruistic behavior in new ways and into circumstances that are unfamiliar and/or infrequently occurring, and we routinely innovate entirely novel forms of cooperative interaction and problem-solving in both familiar and unfamiliar contexts. Indeed, among researchers who study behavior comparatively, humans are famous for the extraordinarily wide range of circumstances in which they actively and flexibly try to find ways to assist and cooperate with one another (e.g., Boyd & Richerson Reference Boyd and Richerson2005; Cheney Reference Cheney2011; Fehr & Fischbacher Reference Fehr and Fischbacher2003; Melis & Semmann Reference Melis and Semmann2010; Moll & Tomasello Reference Moll and Tomasello2007; Silk & House Reference Silk and House2011).

In addition, recent experimental studies (e.g., Warneken & Tomasello Reference Warneken and Tomasello2006; Reference Warneken and Tomasello2007) have shown that even prelinguistic or barely linguistic (14- to 18-month-old) children spontaneously seek out and take advantage of novel opportunities to cooperate with and assist even nearly complete strangers in achieving their goals to a much greater extent than chimpanzees do. Moreover, even very young children reliably take advantage of low-cost or cost-free opportunities to benefit strangers (Thompson et al. Reference Thompson, Barresi and Moore1997), and although chimpanzees are certainly capable of similarly low-cost or cost-free altruistic behavior (e.g., Horner et al. Reference Horner, Carter, Suchak and de Waal2011; Melis et al. Reference Melis, Warneken, Jensen, Schneider, Call and Tomasello2011b), they seem to take advantage of such opportunities far less readily or reliably, and in a far more restricted set of contexts, even for group members with whom they are already familiar (Jensen et al. Reference Jensen, Hare, Call and Tomasello2006; Silk & House Reference Silk and House2011; Silk et al. Reference Silk, Brosnan, Vonk, Henrich, Povinelli, Richardson, Lambeth, Mascaro and Schapiro2005). Similarly, using an experimental apparatus and procedure specifically designed to facilitate cross-species comparisons, Burkhardt et al. (Reference Burkhardt, Allon, Amici, Fichtel, Finkenwirth, Heschl, Huber, Isler, Kosonen, Martins, Meulman, Richiger, Rueth, Spilmann, Wiesendanger and van Schaik2014) found human children to be markedly more spontaneously prosocial than even our nearest primate relatives.

Indeed, Michael Tomasello has recently argued in considerable detail that chimpanzees lack at least some of the motivational and cognitive capacities that make humans’ extraordinarily robust cooperative dispositions possible (Tomasello Reference Tomasello2009; Reference Tomasello2016). More specifically, at about 9 months of age, human infants begin to engage in “joint attentional activities” in which they not only pursue a simple shared goal, but also monitor the attention of a partner to the activity and the goal, whereas chimpanzees seem neither to establish joint attention with others nor to participate in activities with genuinely shared goals (Tomasello Reference Tomasello2009, pp. 63–75): Unlike children, for example, chimpanzees do not seek to re-engage a partner who suddenly breaks off her participation in a coordinated problem-solving activity, even when they are highly motivated to complete the task. In further sharp contrast to even very young children, chimpanzees cannot reverse roles with a partner in a task they have learned to complete successfully (Fletcher et al. Reference Fletcher, Warneken and Tomasello2012), they prefer individual over collaborative problem-solving strategies even when the rewards are equal (Bullinger et al. Reference Bullinger, Melis and Tomasello2011), and they have great difficulty cooperating in circumstances in which obtained resources can easily be monopolized by one party, whereas even very young children (3 years old) spontaneously divide resources obtained collaboratively into equitable shares (Warneken et al. Reference Warneken, Lohse, Melis and Tomasello2011) and continue to cooperate to complete tasks and secure rewards for all participants in a task even if their own reward has already been received (Tomasello Reference Tomasello2009, pp. 65–66). Obtaining resources collaboratively makes children but not chimpanzees more likely to share those resources with those who collaborated to obtain them (Hamann et al. Reference Hamann, Warneken, Greenberg and Tomasello2011; Warneken et al. Reference Warneken, Lohse, Melis and Tomasello2011), and children but not chimps exclude free riders from the spoils of a collective enterprise (Melis et al. Reference Melis, Schneider and Tomasello2011a; Reference Melis, Altricher and Tomasello2013; though cf. Suchak et al. Reference Suchak, Eppley, Campbell, Feldman, Quarles and de Waal2016). Moreover, chimps show no interest in engaging in cooperative social games with no instrumental objective, whereas human children not only do so, but also will often turn a task with an instrumental objective into a cooperative game by replacing the reward in the experimental apparatus to restart the activity (Tomasello Reference Tomasello2009, pp. 63–65). Thus, although chimpanzees are undoubtedly impressive individual problem-solvers who can effectively learn to use a partner's behavior as an instrumental resource, they seem to lack many of the distinctive cognitive abilities and motivations that enable humans to work with one another to solve such problems together so frequently, flexibly, and spontaneously.Footnote ³

This remarkable affinity for spontaneous cooperation and prosociality appears even in infants of 6 months (Hamlin et al. Reference Hamlin, Wynn and Bloom2007) and does not appear to be a function of learning or enculturation. The age at which prelinguistic or barely linguistic children begin to spontaneously cooperate with or assist others and the range of situations in which they do so appear not to be sensitive to the culture in which they are raised: These results have now been replicated, for example, in children from “traditional, small-scale cultural settings in Peru and India” (Callaghan et al. Reference Callaghan, Moll, Rakoczy, Warneken, Liskowski, Behne and Tomasello2011, p. vii). And remarkably, not only does providing verbal encouragement by parents or the prospect of an external reward make no difference to whether or not children help, but also providing such a reward actually decreases the extent of helping on future occasions, a well-known “overjustification” effect thought to occur when an intrinsically rewarding activity or behavior has had its intrinsically rewarding character partially displaced by the substitution of such an external reward (Tomasello Reference Tomasello2009, pp. 7–10). Further evidence of this phylogenetic legacy of spontaneous and flexible human hypercooperation can be found in everything from the fact that humans alone seem to engage in direct, active teaching (Kline Reference Kline2015) to the fact that we alone among mammals have eyes with enlarged white sclera, the better to allow a potential cooperative partner to see where our attention is directed (Tomasello et al. Reference Tomasello, Hare, Lehman and Call2007).

Revealingly, influential earlier research suggesting that nonhuman primates had surprisingly limited theory of mind abilities (e.g., Povinelli Reference Povinelli2000, Ch. 2) was conducted using experimental paradigms in which subjects were expected to distinguish among human experimenters or conspecifics who could or could not help them find hidden food. When the same abilities were probed in experimental paradigms in which subjects competed for food, the performance of nonhuman primates improved dramatically. Most nonhuman primates apparently struggle to understand even the possibility that an interactive partner might be spontaneously offering them voluntary assistance in obtaining food. In one striking example, chimpanzees (unlike 2-year-old human children) could not use the communicative gestures (e.g., pointing) of a human or trained conspecific partner as cues to the location of hidden food, but could find the same food if the partner was instead reaching for it unsuccessfully (Hare & Tomasello Reference Hare and Tomasello2004). These experiments also reveal something about the humans who designed them, however, to whom it did not occur that such spontaneously prosocial or cooperative scenarios might be dramatically more difficult than competitive ones for nonhuman primates to recognize, interpret, or understand. Although chimps are certainly capable of impressive feats of altruism and cooperation (Melis et al. Reference Melis, Warneken, Jensen, Schneider, Call and Tomasello2011b; Suchak et al. Reference Suchak, Eppley, Campbell, Feldman, Quarles and de Waal2016), the sort of domain-general, spontaneous, frequent, and flexible cooperative and altruistic interaction so characteristic of human sociality does not seem to be a similarly ubiquitous feature of ordinary chimpanzee social life.

5. Moral externalization: A cooperation-building machine

For creatures whose behavior is as generally facultative and plastic as our own, the benefits of such spontaneous, domain-general, and flexible cooperative tendencies seem evident: They enable us to more quickly, easily, and (therefore) frequently identify and take advantage of novel cooperative opportunities in both familiar circumstances as well as those that arise infrequently or are even completely unprecedented. There are many unfamiliar as well as familiar things a group of hominins working together can do far more effectively, quickly, efficiently, safely, or otherwise better than if each worked alone, and many others that hominins working separately could not reliably accomplish at all. And such dispositions towards spontaneous and flexible hypercooperation would have become increasingly useful and important as they enabled ancestral hominins (unlike other primates) to radiate into a wide variety of novel ecological habitats uninhabitable by their phylogenetic ancestors and as humans evolved to become obligate cooperators in virtually every environment they occupy.

But of course, such hypercooperative dispositions also expose humans to enormously elevated risks of exploitation (cf. Enquist & Leimar Reference Enquist and Leimar1993). Accordingly, even very young children do not cooperate indiscriminately. For one thing, cooperative and altruistic dispositions appear to be dramatically influenced by indicators of shared membership in a culturally defined in-group, and sensitivity to such indicators is found even among infants (Hamlin et al. Reference Hamlin, Mahajan, Liberman and Wynn2013; Kinzler et al. Reference Kinzler, Dupoux and Spelke2007; Mahajan & Wynn Reference Mahajan and Wynn2012). Even within a well-defined in-group, however, spontaneous hypercooperators remain persistently at risk of exploitation by those who might happily enjoy the fruits of their cooperative and/or altruistic inclinations but not themselves share rewards, forego opportunities for exploitation, maintain costly alliances, or pay the other costs involved in the nearly limitless forms of cooperation possible for human beings in both familiar and unfamiliar or infrequently occurring contexts. A disposition to cooperate with any willing partner in each new and unfamiliar context of interaction is effectively a ticket to sustained, serial exploitation.

The transition to a hypercooperative form of social life therefore required not only that early hominins be motivated to initiate and sustain cooperative and collaborative efforts far more readily than even their nearest primate relatives in response to both familiar and novel environmental challenges, but also some mechanism for ensuring that those efforts be selectively directed (even within a well-defined in-group) towards those who share those same motivations. Faced with a rapidly proliferating array of cooperative and collaborative opportunities, increasingly frequent and consequential decisions concerning the desirability of partners would need to be sensitive not only to characteristics of those partners likely to influence chances of success (like competence, knowledge, or physical prowess), but also to the critical importance of avoiding exploitation. And recent evidence suggests that our moral psychology plays a central and distinctive role in mediating such preferences.

Perhaps most importantly, Skitka et al. (Reference Skitka, Bauman and Sargis2005) showed that the desirability of partners in social interaction generally is mediated by our awareness of the extent of our distinctively moral disagreement with them. These experimenters found that subjects preferred greater social (i.e., closeness in relationship) and even physical distance from those with whom they had identified moral disagreements over and above the social and physical distances they preferred from those with whom they differed in equally strongly held non-moral attitudes (also controlling for importance, certainty, and centrality of these attitudes). A stronger moral conviction on the issue in question was associated with greater intolerance of attitudinally dissimilar others in both intimate (e.g., friend) and distant (e.g., owner of a store one frequents) relationships, and these effects were robust when experimenters controlled for age, gender, individual differences in political orientation, and even tendency to see issues overall in a moral light. Measures of attitude strength alone, however, had an inconsistent relationship with preferred social distance, tending to vary across both relationship types and the specific issue, but more often than not were simply unassociated with preferred social distance from those who held conflicting attitudes. It therefore seems to be whether a particular attitude or conviction is seen as moral in character that most reliably determines whether it inclines us to include or exclude those who do or do not share it from social interaction with us.

Skitka et al. (Reference Skitka, Bauman and Sargis2005) also found lower levels of goodwill and cooperativeness in groups that were heterogeneous with respect to a moral as opposed to an equally strongly held non-moral attitude and a greater inability to generate procedural solutions for resolving disagreements. Importantly, however, groups seeking to resolve non-moral disagreements were constructed by the experimenters so as to be in fact heterogeneous with respect to a moral attitude as well (regarding capital punishment or abortion), but were charged only with finding procedures for resolving their non-moral disagreement instead. Thus, it is not the mere existence but the awareness of distinctively moral disagreement that depresses goodwill and cooperativeness in the members of such groups substantially below that of groups with equally strong non-moral disagreements.

Once again, however, moral convictions would not need to be externalized in order to mediate our preferred social distance or enthusiasm for social interaction with others: It would be simpler and more continuous with the rest of our motivational psychology for us to regard those who do not share a particular subset of our own attitudes as merely different from ourselves in ways that we dislike or find unappealing, rather than as violating demands that we regard all agents as unconditionally responsible for satisfying. Hence, even if we allow that a special conceptual category of motivations and attitudes emerged to mediate our enthusiasm for potential partners in various forms of social interaction, we must still ask how and why such motivating attitudes became externalized into distinctively moral obligations.

The key to resolving this further puzzle is to recognize that although we might more simply and easily use subjective preferences or desires to motivate either our own conformity to a given set of norms or standards of behavior, or our social exclusion of those who do not similarly conform, experiencing moral motivation as externally imposed on both ourselves and others simultaneously is nonetheless an extremely effective and efficient way to ensure that these otherwise distinct motivations arise together and remain systematically connected to one another even as particular norms themselves come to be modified, applied, and extended in new ways and into new circumstances, and even as entirely new norms come to be adopted (or come to be moralized) in the first place. That is, experiencing moral demands and obligations as externally imposed simultaneously on both ourselves and others ensures that if I myself come to be motivated to conform to a particular norm or standard of behavior that I experience as distinctively moral in character, I automatically demand that others conform to it as well, judging them to be less attractive potential partners in social interaction generally if they do not. If instead my motivation to engage in some particular exploitable prosocial behavior or conform to a given norm were experienced as a purely subjective preference or desire, I would not thereby automatically require that desirable social partners must also engage in that behavior or conform to that norm, leaving me open to exploitation by those who are not similarly motivated. (Similarly, if my motivation to exclude those who do not adhere to or accept a given norm were a merely subjective preference or desire, it would not automatically ensure that I myself am also motivated to adopt or adhere to the norm, ultimately leading to my own exclusion from valuable opportunities for cooperative social interaction by those who externalize it.) Thus, I regard my affinity for ice cream or for chocolate over vanilla as a subjective preference or desire that I neither expect nor require you to share. But I experience my own opposition to Nazis or slavery (or selfish hunting partners, or cowardly defenders of the tribe, or untrustworthy coalition partners) as a response to an externally imposed demand by which I insist that you be similarly motivated.Footnote ⁴ And if I become convinced that you are not motivated by the same distinctively externalized moral demands that I myself am, I will find you a correspondingly less desirable candidate partner for social interaction generally.Footnote ⁵

The creation of a novel conceptual category of norms or standards of behavior to which I hold both others and myself responsible simultaneously thus established a mechanism for safely extending prosocial, altruistic, and cooperative behavior in new ways and into new contexts. That is, it ensured that each such extension was automatically protected from exploitation from its inception rather than exposing those motivated by it to exploitation unless and until they evolved a distinct and independent subjective desire or preference to exclude those who do not share that disposition from social interaction with them. From an evolutionary point of view, our characteristic externalization of moral motivation thus represents a mechanism for establishing and maintaining correlated interaction under plasticity – that is, for establishing and maintaining correlated interaction between those prepared to accept any given norm, even as the content of the particular norms we embrace remains free to change over time as we apply them in novel ways, extend them into novel or unfamiliar circumstances, and innovate entirely new norms governing our social interactions. As noted previously, such correlated interaction is the crucial condition which must be satisfied in order for cooperative, altruistic, and otherwise exploitable prosocial dispositions to emerge and/or remain evolutionary stable.

This proposal is strongly supported by Goodwin and Darley's (Reference Goodwin and Darley2012) further finding of a direct relationship between the degree or extent to which we externalize or objectify a given norm that we ourselves accept and the extent to which we are prepared to exclude those who do not accept it from social interaction with us. These authors found that even when they controlled for the strength of agreement with a given moral belief, the more objective subjects held the belief in question to be the less comfortable they were with social engagement with (e.g., being a roommate of) someone who disagreed and the more immoral they regarded such a person as being (see also Wright et al. Reference Wright, McWhite, Grandjean, Lombrozo, Nichols and Knobe2014). These results nicely complement Skitka et al.'s finding (see above) that within the category of externalized norms (but not more generally) we find a direct relationship between the strength of our own commitment to a given norm and the extent to which we exclude those who do not follow it from social interaction with us. That is, our enthusiasm for social interaction with those who violate a given moral norm decreases not only in proportion to the strength with which we ourselves are committed to and motivated by that norm (Skitka et al. Reference Skitka, Bauman and Sargis2005), but also in proportion to the extent to which we ourselves have externalized or objectified it into a distinctively moral obligation (Goodwin & Darley Reference Goodwin and Darley2012).

Note that this proposal obviates the need for any costly punishment of norm-violators in order to get the benefits of externalization off the ground: Each individual simply protects herself from exploitation by excluding those who engage in either transgressions of commission (e.g., theft) or omission (e.g., failing to share) from her own circle of voluntary social interaction. As the pool of available partners comes to consist of an ever-higher proportion of exploitative or otherwise undesirable partners rejected by other community members, the selective pressure to identify and avoid undesirable partners becomes even stronger, and such discrimination in the desirability of partners will serve to carve out networks of cooperative interaction within a single community. Baumard et al. (Reference Baumard, André and Sperber2013) describe ethnographic evidence suggesting the ubiquity of such partner choice among contemporary hunter-gatherer groups, noting that “punishment as normally understood … is uncommon in societies of foragers … in these societies, most disputes are resolved by self-segregation” (p. 66; see also Guala Reference Guala2012). More costly forms of punishment can await the explicit recognition and/or agreement that persistently exploitative or norm-violating partners represent a public nuisance or a threat to the public welfare that can be addressed by coordinated action at the community level with relatively low costs or risks to each individual member.

Moreover, this account explains many of the distinctive central features of our moral psychology noted previously (sect. 2), such as the moral/conventional distinction and the distinction we draw between moral properties and response-dependent properties (like being icky or boring) seen as constituted by our heterogeneous subjective reactions to objects or events in the world. Perhaps most importantly, it offers a natural explanation of the systematic differences in objectivity that Goodwin and Darley (Reference Goodwin and Darley2008) found between judgments of taste/preference, convention, morality, and empirical fact. While judgments of taste or preference carry no demand for or expectation of intersubjective agreement, conventional norms are imposed (explicitly or implicitly) on the members of a group by that group itself (or a recognized authority within it) and thus apply to its members regardless of the subjective preferences and desires of any particular individual but not those of the group as a whole. But we experience prototypical moral norms, demands, and obligations as imposed unconditionally, irrespective of not only our own preferences and desires, but also those of any and all agents whatsoever, including those of any social group or arrangement to which we belong: We apply them not only to ourselves and our fellow community members, but also to absolutely any candidate social partner with whom we might interact, no authority is capable of suspending the demands they impose, and we regard violations of them as more serious and more deserving of punishment than violations of the merely conventional rules that we collectively self-impose. There is correspondingly less room for intersubjective disagreement without error concerning them, indeed nearly as little as in the case of empirical or scientific fact.

The characteristic plasticity of the process by which particular norms become moralized also helps to explain why there is considerable variation in the degree of objectivity (see Goodwin & Darley Reference Goodwin and Darley2012; Wright et al. Reference Wright, Grandjean and McWhite2013; Reference Wright, McWhite, Grandjean, Lombrozo, Nichols and Knobe2014) ascribed to different norms by different individuals: It seems natural to see this as a consequence of differences in the extent to which particular norms have become moralized by different individuals or groups in the process of acquiring them. Whether and to what extent particular norms become moralized in an individual will presumably be quite strongly influenced by the cultural environment in which she is raised, but a wide range of further factors might also influence whether or to what extent any particular norm becomes moralized (such as the presence of a strong affective response; see Nichols Reference Nichols2004). Such factors might also help explain the various curiously partial or incomplete forms of moralization discussed by Haidt et al. (Reference Haidt, Koller and Dias1993) and Nichols (Reference Nichols2004), while the comparatively high importance of avoiding exploitation might help explain Goodwin and Darley's (Reference Goodwin and Darley2012) finding that subjects typically judge the wrongness of immoral acts as significantly more objective than the goodness of positive moral acts even when the strength of their convictions in the wrongness or rightness of each act is precisely the same.

This remarkable plasticity in the acquisition, externalization, and application of distinctively moral norms does not, of course, show that there are no constraints on the content of the norms that can become (or cease to be) moralized, and it is certainly noteworthy that norms regarding harm and fairness appear to be either universal or found across a much wider range of human societies than others (see Nichols Reference Nichols2004). Indeed, we are now finally well-positioned to understand why human moral norms are so often concerned with harm, justice, fairness, and protecting the rights of victims, even though they need not be so concerned in order to trigger some or all aspects of the moralized response: Although it may be possible for nearly any norm or behavior to become moralized, norms protecting spontaneous prosociality and cooperation from exploitation are those which must be preserved to maintain the most important evolutionary benefits of moralization itself, and norms concerning harm, fairness, justice, and the rights of victims seem like natural candidates for this role.

This suggestion leaves open a wide variety of mechanisms, however, by which such cooperation-enhancing norms could become and remain ubiquitous among human cultures. This might be a straightforward consequence of the fact that existing sets of norms have come from pre-existing sets of norms by descent with modification, or it might be a product of convergent cultural evolution favoring groups better able to foster fitness-enhancing cooperation among their members. It may be that we are biologically predisposed to regard norms with particular kinds of content as moral in character, or their moralization might be the product of the sort of strongly biased learning mechanism that threatens to make the very distinction between learning and predisposition seem unhelpful or inapposite. But we need not here prejudge the answer to the daunting question of which of these mechanisms or what combination of them is most likely to have produced the distinctive patterns we find in the moralization of particular kinds of norms among human cultures and societies, nor is any single explanation likely to be correct in all cases. The ubiquity of norms concerning harm, fairness, justice, and the rights of victims can be explained by recognizing the relative importance of such norms in fostering fitness-enhancing cooperative, altruistic, and prosocial interaction among humans, even if we remain agnostic about the specific evolutionary mechanism(s) at work.

The account also leaves open a wide range of important further empirical questions. For example, it may be that the further distinction between conventional and moral norms emerged phylogenetically following the constitution of a more general category of normative beliefs, or it may be that some norms originally seen as merely conventional became externalized or objectified in the ways we have noted to create the distinctive conceptual category of moral norms. It leaves open a further and distinct question concerning how moral norms acquire their characteristic status in the course of individual ontogeny. It could be that at a particular point in human development a subset of existing norms become externalized or objectified and thereby acquire a distinctively moral character, but there is at least some evidence that children instead start out “reifying” (externalizing or objectifying) all social norms and rules and simply learn over time to treat some (but only some) particular norms and rules as merely conventional instead (see Gabennesch Reference Gabennesch1990). Once again, the fundamental attractions of the idea that moral externalization establishes or preserves correlated interaction between those prepared to adopt any particular norm (or to extend an existing norm in a particular way) while leaving the content of those norms free to evolve on a cultural (rather than phylogenetic) timescale do not depend on the answers to these admittedly important further questions.

Finally, this proposal explains an intriguing pattern of variation in our characteristic generalization of moral norms identified by Sarkissian et al. (Reference Sarkissian, Park, Tien, Wright and Knobe2011). These authors found that subjects are much more likely to insist that at least one party to a moral disagreement must be in error if that disagreement arises among the subject's peers than if it arises instead between a peer and a member of an isolated Amazonian tribe or an extraterrestrial civilization. Although the authors interpret these results as showing that “the folk” become increasingly relativist as they are forced to confront alternative moral frameworks or perspectives, the account presented here would suggest instead that the demand that others must share our own moral convictions becomes progressively more relaxed as salient contextual cues render the realistic possibility of significant interaction with the agent who makes a contrary moral judgment seem ever more remote. This is perfectly intelligible, of course, if human moral psychology evolved to mediate the evaluation and selection of partners in social interaction: An act can be wrong for agents in all times, places, and cultures even if the demand for intersubjective agreement among judges of that act is dramatically weakened when salient further considerations substantially reduce or preclude the possibility of the dissenting judge serving as a partner in significant forms of social interaction in any case. This analysis suggests that if salient cues instead emphasize the realistic prospect of ongoing social interactions with those who dissent from the subject's own moral judgments, we should see a corresponding increase in the extent to which those subjects insist that at least one party to the disagreement must be in error.

6. Becoming human

I have suggested that the emergence of a distinctive conceptual category of motivations or attitudes which others must share (on pain of being judged less desirable partners in social interaction generally) served to protect the extension of increasingly spontaneous cooperation and other forms of adaptively advantageous prosociality from exploitation and thereby facilitated a transition in the hominin line to a vastly more cooperative form of social life. But the degree of protection against exploitation afforded by such moral externalization in the course of extending prosociality and cooperation is strictly limited by the accuracy and detail of the information agents are able to acquire about one another's distinctively moral commitments. Perhaps some of this information might be gleaned merely by observing one another's behavior, but recognizing how important the extent and detail of such information are to the degree of protection against exploitation that it provides invites us to notice the otherwise quite remarkable proportion of our ordinary conversational interaction devoted to soliciting and providing just this information to one another.

As the anthropologist Robin Dunbar has documented, an astounding 60–70% or more of our ordinary conversation is taken up with discussion of “who-was-doing-what-with-whom and personal social experiences” (Dunbar Reference Dunbar1996, p. 174; see also Emler Reference Emler1990; Reference Emler, Goodman and Ben-Ze'ev1994; Reference Emler, Robinson and Giles2001) – an activity colloquially called “gossip” that is extremely widespread in traditional hunter-gatherer societies (Haviland Reference Haviland1977) just as it is in our own. What these researchers have not sufficiently emphasized, however, is that such gossip does not consist simply of reports of our own and others’ behavior: We are also extraordinarily interested in expressing our own and hearing others’ attitudes towards, rationales for, and justifications of that behavior. I suggest that we judge the desirability of potential partners in cooperative endeavors and other social entanglements neither exclusively nor even primarily by gathering intelligence on cheats, liars, free-riders, and other undesirables from our interlocutors (cf. Dunbar Reference Dunbar1996; Reference Dunbar2004; Enquist & Leimar Reference Enquist and Leimar1993), but instead by evaluating the extent to which those very interlocutors share our own attitudes and reactions towards these and other specific examples of behavior, including observed, reported, rumored, and even merely imagined or counterfactual instances.

After all, we eagerly engage in such practices of moral evaluation, instruction, and advertisement with our peers (cf. Sarkissian et al. Reference Sarkissian, Park, Tien, Wright and Knobe2011; and see sect. 5 above) regarding conduct in rumors, stories, films, television programs, legends, fables, and fairy-tales regarding characters who are fictional, long dead, or otherwise present no possibility themselves of serving as actual prospective partners in social interaction. Of course, we engage in such discussion because we find it “interesting,” but the extraordinary interest we take in these particular subjects of conversation can no more be simply taken for granted than facts about what we find painful, pleasurable, or delicious. It is obviously useful for creatures who live in stable social groups and depend upon one another to take an interest in who does what with whom, but our otherwise remarkable interest in sharing, hearing, validating, ratifying, and calibrating our opinions about such doings with one another arises only because we thereby learn or establish as much or more about the moral convictions of the prospective cooperative partners who relate and respond to such gossip as we do about those whom they discuss, and because this affords us the opportunity to similarly advertise our own moral convictions, negotiate points of substantive agreement and disagreement, and try to establish the widest possible range of genuinely shared moral commitments with our interlocutors. In fact, such moral advertisement, evaluation, navigation, and negotiation seem to make up a substantial proportion of ordinary conversational interaction among humans.

Indeed, I suggest that such practices of moral advertisement and evaluation have played a crucial role in facilitating a range of further distinctive features of human cognition and communication. Because we so often spontaneously extend cooperative or exploitable interaction in novel ways and into novel or infrequently occurring contexts, we do well to learn as much as we can about the dispositions and judgments of potential cooperative partners (as well as to advertise our own) across as wide a range of circumstances as possible, including those that are at present merely imagined, hypothetical, or counterfactual. And of course, the ability to advertise, assess, and negotiate agreements concerning such dispositions or judgments effectively would be dramatically expanded not only by the sophistication of our linguistic abilities, but also by other cognitive abilities either unique to humans or present in them to an extraordinary degree, such as theory of mind abilities, perspective-taking, counterfactual reasoning or simulation, chronesthesia (mental time travel), and abstract categorical thought.

I do not mean to suggest that the demands of cooperation and moral advertisement were ultimately the key driver of higher cognitive abilities in humans. Indeed, Sterelny (Reference Sterelny2012) has argued convincingly that the extremely rapid cognitive evolution that produced behaviorally modern humans at least 40,000–50,000 years ago was most likely the product of one or more feedback loops, in which the acquisition or sophistication of some cognitive abilities made the acquisition or sophistication of others even more valuable in an ongoing cycle. I do want to suggest, however, that a feedback loop of precisely this sort connects the various sophisticated linguistic and cognitive capacities described in the previous paragraph, with improvements in each rendering further improvements in the others even more useful specifically for the purposes of moral advertisement, evaluation, and the navigation or negotiation of shared moral judgments concerning particular instances of behavior (and their possible variations), character traits, and (much more rarely) general and abstract moral commitments.Footnote ⁶ Increasing the sophistication of any or all of the distinctive cognitive and linguistic abilities noted above would certainly have allowed us to increase the range, complexity, and precision of the information we provide one another in the course of such moral advertisements, evaluations, and negotiation. Therefore, among other important advantages, such cognitive and linguistic sophistications would have allowed us to better correlate our interactions with desirable cooperative partners, reduce the risk of exploitation or failed cooperation in new and unfamiliar contexts, and thereby dramatically expand the range of circumstances into which our cooperative dispositions could be safely and effectively extended.Footnote ⁷

Of course, agents have incentives to represent themselves as more desirable interactive partners than they really are, and in fact people do (intentionally or not) overestimate how ethically they themselves will behave in a given set of circumstances while predicting others’ behavior more accurately (Epley & Dunning Reference Epley and Dunning2000). But such misrepresentations must typically remain fairly modest in character because they are by no means cost-free. Other members of the community have a wide variety of more-or-less continuous opportunities to compare the actual behavior of each community member with moralized attitudes she has previously expressed regarding behavior in circumstances that were at the time merely reported, rumored, imagined, or counterfactual, and damaging the credibility of one's moral advertisements is another way to become marked as an undesirable cooperative partner by others in the community. This also helps to explain why the extension of moral concern and spontaneous prosociality is particularly sensitive to indicators of in-group membership: Especially in ancestral environments, it would generally be all and only the members of an in-group with whom such moral advertisements and evaluations could be regularly exchanged.

This view of moral advertisement also explains our otherwise somewhat puzzling attitude towards hypocrisy. If I declare that it was wrong for you to abandon the camp instead of defend it or to keep all of the best nuts for yourself and then I proceed to do just the same in your position, my hypocrisy is itself a moral wrong over and above my cowardice or my greed. But of course, these behaviors may indeed deserve my condemnation even if I myself am unable to resist engaging in them! This puzzlement dissolves, however, when we note that public criticism of others’ moral failings expresses the recognition of an unconditional, generalizable, authority-independent externalized demand to which the author of such criticism is (therefore) automatically subject as well and thus serves as at least a prima facie commitment that she herself would not respond to the same circumstances in the same fashion. In this way, our advancing linguistic and cognitive abilities would have enabled simple moralizing gossip among humans to convey genuine (albeit implicit) commitments regarding our own future behavior, and undertaking such public commitments would have become yet another way to increase one's attractiveness as a potential partner in exploitable social interactions.Footnote ⁸ Indeed, such moral criticism seems a natural and plausible precursor to fully explicit practices of promise-making and commitment more generally. This also illuminates the distinctive role played in our social lives by practices of apology and repair, by which we seek to reassure others that we do indeed recognize a norm or obligation that we take ourselves to have violated in some particular case and that we can still be trusted to see it as placing demands or constraints on us in the future (despite recent contrary evidence). This is part of why such apologies are not cost free and why we are often reluctant to make them, as well as why they strike us as empty or insincere unless accompanied by modified behavior.

I have suggested not only that plasticity in the content of the norms or behaviors we moralize is what allows humans to so spontaneously, frequently, and flexibly take advantage of novel opportunities for cooperative and altruistic interactions, but also that the price of protecting that plasticity from exploitation is the continuous monitoring of others’ and advertisement of our own existing moral commitments in the course of ordinary social interaction. If this analysis is correct in even its fundamentals, it seems possible but unlikely that nonhuman organisms externalize distinctively moral motivation in the same ways that humans do. Although in principle nonhuman primates might better correlate their cooperative and other prosocial behavior if their own motivations for those behaviors arose only in conjunction with comparable demands on the behavior of othersFootnote ⁹ (and vice versa), it would nonetheless seem that the most salient advantages of such externalization arise for humans precisely because we so frequently and reliably face the need to correlate such interactions when (1) existing norms are applied in novel ways to familiar circumstances, (2) existing norms are adapted or extended into unfamiliar and/or infrequently occurring contexts in particular ways, and (3) entirely new norms come to be adopted (or come to be moralized) in the first place. After all, if exploitable prosocial behavior arose only in consistent ways across a narrow and/or fixed range of recurring contexts, the extent of an agent's subjective motivation to pursue a given course of action in those contexts and the extent to which she demands that desirable social partners do so as well could independently converge on stable values and then remain there indefinitely. But externalization ensures that this crucial relationship is immediately established and maintained even as we extend our exploitable prosocial behavior in novel ways, into novel contexts, or by means of novel norms of behavior. Thus, it is the combination of the extraordinary plasticity, flexibility, and facultative character of our behavior with the extraordinary spontaneity and frequency of our prosociality that renders moral externalization so enormously advantageous for human beings. More generally, the selective advantage of externalizing a distinctive category of motivations so as to protect them from exploitation increases directly with the spontaneity, frequency, and plasticity with which a creature engages in cooperative, altruistic, and other exploitable forms of prosociality, and present evidence (tentatively) suggests that humans’ enormously spontaneous, frequent, and plastic prosociality is simply not a pervasive feature of chimpanzee social life. Moreover, the advantages of such externalization seem to depend on the extent, precision, and accuracy of the information we can exchange with one another concerning our distinctively moral commitments, and here again it seems that humans are far better equipped to exchange such information than are chimps or other nonhuman primates. Indeed, perhaps this makes it less surprising that nonhuman primates seem to lack some of the fundamental cognitive capacities (e.g., joint attention) that facilitate such spontaneous, flexible, and domain-general cooperative interaction in humans (see sect. 4).

Of course, the very same plasticity in the content of our moralized norms that makes externalization so important for human cooperation allowed (and continues to allow) norms of behavior having nothing whatsoever to do with protecting cooperation, altruism, or prosociality from exploitation to become moralized: attitudes concerning homosexuality, say, or gambling, or masturbation, or female genital mutilation, or food taboos, or whatever else a particular person or social group might ultimately seek to convince us (for whatever reasons) are unconditionally prescribed or proscribed in the distinctively externalized manner of moral demands and obligations. And of course, human cultures famously demonstrate an incredible range of variation in the particular norms that have come to be moralized by their members. That is, although human beings reliably begin moralizing some norms or others at a particular point in their cognitive development, which particular norms become moralized is highly variable and depends sensitively on a complex surrounding social environment.

On the other hand, this sensitive dependence of the content of any particular person's norms on her social and cultural environment also ensures considerable homogeneity in the norms embraced by the members of any given cultural or social group at any given time. It might be tempting to view this group-level variation through the lens of cultural group selection, supposing (as Darwin famously did) that differences in the moral norms to which the members of distinct groups are committed generate selective advantages and disadvantages for each group in competition with others. But many, if not most, of the differences in the content of the particular moral convictions embraced by different social groups would seem to have little or no intrinsic adaptive significance. Indeed, the truly extraordinary range of variation in the content of the moral convictions embraced by even closely related social groups (e.g., Henrich et al. Reference Henrich, Heine and Norenzayan2010, p. 61) strongly suggests that the content of a group's moral convictions ultimately came to play a further role simply as an effective and reliable marker of in-group membership itself, just as it has in the case of linguistic dialects (cf. Boyd & Richerson Reference Boyd and Richerson2005, Ch. 6; Dunbar Reference Dunbar1996, p. 168f); styles of dress; rituals; tattooing, scarification, other bodily markings or adornment; and virtually every other sociocultural practice with easily recognizable variants whose differences have little or no intrinsic adaptive significance. Cultural evolution routinely takes advantage of non-adaptive or very weakly adaptive variation in such practices to serve as in-group markers helping to identify the members of one's own communities (at a variety of scales or sizes) and distinguish them from members of other such communities. Thus, once the general conceptual category of moralized norms was in place, selective pressures would have favored variation for its own sake in the particular norms moralized by different groups, just as they have in the case of many other similarly plastic social and cultural practices, generating in turn the extraordinary range of variation in the moral norms we find among even closely related human cultural groups. Following the inception of large-scale agriculture, it seems that this same plasticity was easily co-opted to establish, extend, and reinforce power asymmetries by means of new norms (perhaps especially new role-dependent and/or asymmetric norms) as human groups became larger, more complex, and more hierarchical.

On this view, our moral psychology represents something of a kludge, whose rough edges, scars, and imperfect fit with the rest of our motivational psychology is revealed in our endless philosophical puzzlement (at least since Plato) concerning how anything could have the distinctive combination of characteristics that we seem to unreflectively attribute to moral obligations, facts, and properties. The phenomenology of moral demands combines elements from each side of the fundamental division between the subjective and the objective in ways found nowhere else in nature: Perhaps most saliently, we experience such demands as imposed on us without regard for our preferences in something like the way objective empirical facts are, but as nonetheless intimately connected to our motivational states in ways that such empirical facts never are. Indeed, moral judgment and motivation are so deeply entangled that philosophers have sometimes argued that simply recognizing the existence of a moral demand or obligation automatically entails being motivated by it, unlike the existence of any ordinary matter of empirical fact. It is by now a familiar observation that evolution proceeds by making incremental modifications to whatever materials are already close at hand. And it seems eminently plausible to suppose that among creatures who go in for cognitively complex forms of representation at all, the most fundamental division embodied in their experience will be that between representations of how things stand in the world itself (e.g., “the cat is on the mat”) – from which others cannot dissent without somebody being wrong about something – and our subjective reactions to those states of the world, like pain or the desire for ice cream, that are intrinsically motivating but carry no such demand for intersubjective agreement. This fundamental division was surely the background phenomenological and conceptual framework into which moral norms, demands, considerations, and commitments had to be shoehorned by the conservative, tinkering process of evolution; and this in turn explains why we find their curiously hybrid character so endlessly puzzling. This ongoing puzzlement illustrates how and why our moral cognition is perhaps an imperfect adaptation, but of course, it need not be perfect to confer a selective advantage upon those who possess it. An old Darwinian joke tells of two hikers being chased by a bear, one of whom stops to put on snowshoes. “Are you crazy?” says the other, “You can't outrun a bear in snowshoes.” To which the first hiker replies: “I don't have to outrun a bear – I only have to outrun you.”