Supernatural Punishing Agents:Cultural or Genetic Adaption?
A number of theorists (e.g., Harris & McNamara, 2008; Johnson & Bering, 2006) have proposed that religious beliefs, such as those associated with supernatural watchers, are genuine genetically evolved adaptations for enhancing human cooperation. That is, they suggested that there are modules for religious beliefs that originated in genetic mutations and have been favored by natural selection because of their cooperation-enhancing abilities. Johnson and Bering (2006), specifically, suggested that the belief in supernatural agents served the adaptive purpose of the wholesale suppression of selfish behavior.
Although we are in agreement with much of Johnson and Bering’s (2006) argument regarding the effects of moralizing supernatural agents on cooperation, we disagree with their suggestion that these beliefs emerged as genetic adaptations. The position that we endorse places many of the important details of religious beliefs in general, and the beliefs about the characteristics of supernatural agents more specifically, in greater debt to cultural evolution (see also Atran & Norenzayan, 2004; Henrich, 2007). We argue that the fear of punishing supernatural policing agents, instead of being a specific genetic adaptation, developed as evolutionary by-products honed over generations by cultural evolution. The evolved structure of the brain resulted in a mind that was very receptive to ideas about supernatural agents, a receptivity that was capitalized on by competing cultural variants of supernatural agents.
There are a number of factors that favor our approach over that of the “god beliefs as genetic adaptation.” First, theoretically, the reputational models of cooperation verbally described by these authors (Bering, 2006) are actually unlikely to favor or explain larger scale cooperation in purely genetic evolutionary models, although they can work well for cultural evolution. Second, it not clear how beliefs in supernatural agents could be encoded in DNA, and even if they can be, it’s not clear why natural selection would resort to programming supernatural beliefs into the human genome, as opposed to pursuing a variety of other, seemingly less costly routes to addressing the adaptive problem created by reputation management. Third, the genetic adaptation approach seems to flounder with the empirical evidence indicating that many small-scale societies lack moralizing high gods that act as omniscient supernatural punishers. We briefly discuss each of these issues next.
Evolutionary Modeling: The Selection Between Multiple Stable Strategies
Formal genetic evolutionary models based on purely within-group natural selection do not provide a solution to larger scale cooperative dilemmas (Henrich, 2006; Henrich & Henrich, 2007; Panchanathan & Boyd, 2004). These models— whether they involve costly punishment or reputation-based withdrawal of help— show that the same process can stabilize any costly behavior (including costly maladaptive behaviors that hurt the group and the individual), not merely cooperative behaviors. This means that these approaches suffer from an “equilibrium selection problem,” and we have no theoretical reason to expect within-group genetic selection to favor larger scale cooperation. Within-group transmission processes, therefore, cannot provide a complete solution to the dilemma of larger scale cooperation.
If we consider cultural evolution, however, and allow these alternative stable equilibria to compete in a process called cultural group selection, cultural evolution can favor norms and beliefs that lead to larger scale cooperation. This process, described next, is well modeled and does not suffer from the problems often associated with arguments for the genetic group selection of cooperation (Henrich & Henrich, 2007).
The previous description of reputation and cooperation may be surprising, as some psychologists have repeatedly claimed that “individual-level selection” based on reputation can favor larger scale cooperation (Bering, 2006). There are three issues that seem to need clarifying. First, we emphasize that we are referring to the analysis of mathematical models, not verbal models. Whenever theorists, deploying the mathematical tools that have long formed the bedrock of the study of evolutionary processes (Nowak, 2006), have sought to model reputationbased processes for solving larger scale cooperative dilemma, the previously mentioned issue of equilibrium selection emerges (Panchanathan & Boyd, 2004). That is, there is simply no mathematical model that supports the purely verbal models that some evolutionary psychologists have so frequently asserted: All such models generate multiple stable equilibria that include cooperative outcomes along with numerous noncooperative ones. Viewed as a genetic evolutionary process, these models require some mechanism, such as genetic group selection, to shift among these equilibria.
Second, part of this confusion may result from a failure to distinguish cooperation in dyads from larger scale cooperation in big groups. Reputation can favor cooperation in dyads (Leimar & Hammerstein, 2001; Panchanathan & Boyd, 2003), but this is not the kind of cooperation at issue. The models typically cited by psychologists, if any are cited at all, are limited to dyadic cooperation and do not extend to larger cooperative groups. Reputation-based reciprocity can provide a foundation for human concerns about reputation (Nowak & Sigmund, 2005), but either cultural evolution or culture–gene coevolution is needed to explain why reputation extends to cover all manners of social norms, including those that stabilize larger scale cooperation.
Third, the kind of cultural group selection we are discussing involves groups stable in equilibria, some of which are cooperative and some of which are not. This is not the kind of between-group influence on individual fitness that most non-specialists are accustomed to reading about, and it is not susceptible to the usual concerns that target the genetic group selection of altruism. In an ecology of different groups, defectors entering cooperative groups are suppressed by withingroup selective processes (via punishment or reputational damage). This is unlike the usual case of genetic group selection in which defectors reap a fitness bonanza when they enter cooperative groups (lots of people to free ride on). The effect of this suppression of free riding is to maximize the importance of the variation between groups and to magnify the importance of competition between groups (Henrich & Boyd, 2001).
Evolutionary Fit: Wholesale Versus Selective Suppression of Selfishness
Our second concern is the suggestion that the fear of supernatural policing agents was a genetic adaptation rests heavily on the assumption that such a belief could be genetically encoded, an assumption that can by no means be casually overlooked. Despite rampant speculation, there is no evidence to support the idea that modules evolve at the level of particular beliefs. Moreover, many have criticized the extension of biological evolutionary explanations to this level of specificity on theoretical and empirical grounds (e.g., Fodor, 1987; Panksepp & Panksepp, 2000).
Granting, however, that beliefs could develop as mutations and ignoring the empirical record of religion in small-scale societies, is it plausible that such a mutation would proliferate? According to Johnson and Bering (2006), the fear of supernatural watchers emerged in response to the ability that humans developed to communicate information about reputations. In this new environment where one’s slights and transgressions could be broadcast beyond the offended party, the selfish strategies of yore became much more costly. As a result, those possessing the mutation of a fear of omniscient watchers would have acted less selfishly in general, as they were in constant fear of being judged by the watcher and thereby would be less likely to attract the negative repercussions of being caught and exposed as a selfish operator. The wholesale suppression of selfish behavior, they concluded, would be ultimately adaptive at the individual level.