Discovery of Hidden Variables for the Evolution of Ethical Religions

Discovery of Hidden Variables for the Evolution of Ethical Religions

RESEARCH ARTICLE

Multiple Hypotheses and Divergent Explanations: The Evolution of Moralizing High Gods

Douglas R. White*1,2, Tolga Oztan1, Giorgio Gosti1, Elliott Wagner1, and John Snarey3

1Institute for Mathematical Behavioral Sciences, UCI, 2Santa Fe Institute, 3Human Development and Ethics, Emory University

SFI working group on robustness of models for sociocultural evolution 10-25-2011

*To whom correspondence should be addressed. E-mail: SOM:Supplementary On Line Materials

Shortened version intended for PNAS

Notes: Periods after references such as (3..) are to facilitate searches in revising the bibliography.

The abstract will be integrated into the main text.

Abstract. As Darwin noted (1..), religious beliefs are evolved cultural characteristics (see SOM#0.5: What did Darwin say about Religion?). In the 1980s, evolutionary biologist Richard Alexander (3..) sparked new interest in the factors affecting the evolution of religious belief and in particular of ethically-centered religions. His work stimulated diverse theories about the adaptive benefits of belief in moralizing high gods focused on reproductive fitness, prosociality within and between groups, group-level adaptation, costly signaling and group commitment, supernatural punishment, and game-theoretic models of indirect reciprocity, altruism and the free-rider and punishment problems.

Comparative studies of religion often use Swanson’s (2..) moral gods variable (5..,6..,8.., 10..,12..,39..) as coded for the Standard Cross-Cultural Sample (SCCS, 25..,37..) to study ethical religious beliefs within an evolutionary framework. Few of these studies, however, have followed up on Alexander’s ideas of how ethical dilemmas (SOM#1: Are the “ethical principles” of moral gods universal?) faced by different societies have influenced systems of beliefs. Rich cross-cultural datasets are available to identify the ranges of societal, adaptive, and environmental variables that would help clarify the specifics of more general (“under fitted”) theories of evolution in religious beliefs. Yet, studies that have tested hypotheses and developed models with SCCS data have failed to simultaneously test multiple working hypotheses, a recurrent topic of the journal Science (18..,19..). This methodological shortcoming has led to conflicting hypotheses and misidentification of important factors in the origin and diffusion of beliefs about moralizing gods. We use multi-method and multi-hypothesis approaches to compare models and results with complementary types of analytic techniques. Our approaches lead to two new models that illustrate how complex problems of sociocultural evolution might be resolved and results of various studies could be synthesized. These models identify the factors that are likely to lead to grossly unequal disparities of wealth and which, as Alexander argues, may have encouraged the invention or diffusion of belief in moralizing high gods. In this study, we find that, among other variables, including scarcity of water, the critical predictor identified for pastoralists is the social inequality produced by cyclical variations in animal stock and for agricultural societies it is the inequality produced by cyclical variations in the ownership of land.

Introduction

Despite being a topic long neglected by researchers since Darwin (1..), who argued that religious belief was an evolved characteristic, explanations of the origins and diffusion of moralizing religions (2..) have recently received much attention from evolutionary biologists (3.., 4.., 5..), anthropologists (6.., 7..), scholars of religion (8..,9..), psychologists (10.., 11..), and a political economist (12..). This resurgence of interest in the evolution of religious beliefs began with Alexander (3..), who, while not familiar with the work of Swanson (2..), perceived many of the same issues in morality, ethics, and religion. Both suggested that the adoption of ethical principles into religion would often serve common issues of all members of a group, and/or the interests of the powerful, those inadvertently harmed by excessive inequalities of power, and instabilities in social relationships that are harmful to the group or society. Oddly, many of the problems that defined as theorists (See SOM#1: Alexander and Swanson) have been ignored in favor of investigating the role of indirect reciprocity in the evolution of prosociality. Many researchers have been inspired to hypothesize that the invention of moralizing “high gods” is an adaptation to promote cooperation within society. Other hypotheses are that religious beliefs serve as hard-to-fake costly signals of group commitment (13..,14..); that religion’s moral dictates are group-level adaptations that promote group success (2..,15..,16. SOM#2: Intergroup competition); that fear of supernatural punishment serves to increase contributions to public goods (12..); or that belief in moralizing “high gods” facilitates costly cooperation between strangers in societies too large for reputations to be easily tracked (10..). Few of the game-theoretic models thought to be relevant to these issues or to explaining the origin of moralizing high gods are supported by statistical evidence from cross-cultural studies. Even fewer evolutionary theorists incorporate full consideration of what Alexander and/or Swanson (1..,2..) saw as the relation between the ethical concepts of moral-god religions and concepts of justice or fairness.

In fact, among studies that have used cross-cultural data to test hypotheses about the relevant correlates of moral gods and the effects of multiple variables in regression models, little agreement exists about which hypotheses are confirmed by the data. The failure to consider alternative hypotheses has led many researchers to neglect the nuances of earlier theories or misconstrue the variables of earlier studies. A common mistake in such cases has been the failure to follow the lead of Alexander (3..), who cautioned researchers to consider effects of different periods in evolutionary time, as for example, the independent invention of moral gods in earlier periods versus increasing cultural diffusion in later periods. Table 1 shows seven sets of variables considered in five prior cross-cultural studies (2.., 5.., 6..,8.., 12..) and two new variables in models of our own that test and synthesize these results while using improved regression techniques and controls for the nonindependence of observation. The table shows the factors that each study found to be associated with a society’s belief in moralizing gods. Salient in the discrepancies among the prior studies listed in the Table is the tendency of conclusions to diverge according to whether the authors used simple correlations, multiple variables, or controls for spatial and linguistic (common origin) autocorrelation. One study (5..), without autocorrelation controls, for instance, surmised that high rates of external warfare and plentiful subsistence resources might affect multiple levels of political hierarchy that in turn might predict moralizing gods (R2=.29, p = 0.0001, inflated by autocorrelation).

View this table in …

Table 1: Sets of variables are considered in prior cross-cultural studies and two of our own models that synthesize prior results and add refinements to prior variables. The major conflicting hypotheses are those of Alexander, Snarey, and Brown and Eff versus Swanson and Roes and Raymond.

Types of Topic Vars. / Controls / Positive predictors of moralizing gods
Spatial / Language / Community Structure / Habitat / Conflict / Pastoralist
Topics Þ ßStudies / Diff-usion / Common origin / Political hierarchy (Superjh) & Prosociality / Community size (Commsize) or Fixity / Food & water resources / External or Internal War / Pastoral society features
Roes & Raymond (5..)-2003 / Not con-trolled / Not controlled / Levels of political hierarchy
Superjh / Rich resources cor.with political hierarchy / External war cor .w/ political hierarchy / Animal husbandry not tested
Johnson (12..)-2005 / Not con-trolled / Not controlled / **Prosocial Political-Economic Institutions / Community size Commsize / Not considered / No external war Internal war (?) / Not considered
Our OLS Model 1* (17..)-2011 / Not con-trolled / Not controlled / Levels of political hierarchy (Superjh) given other variables / Fixed comm- unities & wages (episodic inequality) FxCmtyWages Def / Scarce water: Lo_rain_Dry Def / No Effects / Pronounced episodic inequality in herd sizes AnimXbwealth Def
Brown & Eff 2SLS*
(6..)-2010 / Yes / No effect / 1st principal component of Superjh & Commsize / Moderate community size (Non-monotonic & curvilinear) / Food scarcity
Scarc / No external war / Animal husb. problems of theft & moral regulation
Our primary 2SLS-IS* Model 2 (17..)-2011 / Yes / No effect / Levels of political hierarchy (Superjh) given other variables / Fixed comm- unities & wages (labor inequality) FxCmtyWages Def / Scarce water Lo_rain_Dry Def / No Effects / Episodes of pronounced inequality in herd sizes AnimXbwealth Def
Inter-societal & ecological interactions / Population, community size & structure increase in last three centuries, relevant for all models / Ecological spread of deserts *** / Not applicable / Pastoral societies spread with deserts
Theories supported by our models / Swanson (2..) partial support; Models 1,2 and Brown & Eff / Alexander (3..);
Models 1,2; elsewhere uncertain / Snarey (8..) Modified in Models 1,2 / None / Brown & Eff (6..) as Modified in Models 1,2
Theories disconfirmed by our models / Roes & Raymond (5..) / Johnson (12..) / Roes & Raymond (5..) / Roes & Raymd. Brown Eff (6..) / Brown & Eff (6..)
Column 1 / Column 2 / Column 3 / Column 4 / Column 5 / Column 6 / Column 7 / Column 8
Def - Variable defined in the text. *Each imputes missing data / *** add quote to strengthen argument
**Johnson: Money, Credit; Levels of political hierarchy, Police, Taxes, Formal sanctions; No variables are found to be significant for SCCS measures of prosocial individual behaviors: Compliance of individuals with community norms, Loyalty to the local community (sccs$v778), Loyalty to the wider society (sccs$v779), Compliance of Individuals with Community Norms (sccs$v775), Formal Sanctions and Enforcement for Community Decisions (sccs$v775), and Sharing of food (sccs$v1718).

Table 1: Sets of variables are considered in prior cross-cultural studies and two of our own models that synthesize prior results and add refinements to prior variables. The major conflicting hypotheses are those of Alexander, Snarey, and Brown and Eff versus Swanson and Roes and Raymond. The last two rows list studies that are at least partially supported or are disconfirmed with respect to certain variables. These include disconfirmations such as Roes & Raymond’s assertion that plentiful resources and external war have indirect effects on the likelihood of occurrence of moral gods and Brown & Eff’s conclusions that pastoralism and lack of external war are predictors of moral gods. The full set of variables (e.g.., Moralizing gods, Superjh) are defined in the SOM text (Description of Variables. SCCS society numbers for the 43 societies with moral god beliefs are 21 24 25 26 29 34* 36 37 38 39 40 41 42 43 44 46 47 48 50 51 52 54 55 56 57 58 59 64 65 70 82*/ 120* /140 /151 154 155 156 158 160 162 / 172 186 (*indicates those coded for Snarey’s HiGod4 variables that Murdock codes as missing data). Numbers are keyed to Figure 1 and those underlined refer to local rather than world religions.

Figure 1: Societies of the Standard Cross-Cultural Sample ordered from 1-186 (southern Africa to South America) by most similar neighbors for use in calculating diffusion effects. The smaller figure, 1a, shows the distribution of high gods (1=none, 2=present but inactive, 3=present but not concerned with human morality, 4= concerned with morality).

Fig. 1: Hi gods 1=none (black) 2=punish (gray) 3=indifferent (organe) 4=moralizing (red); green numbers show SCCS societies along the maximum diffusion path from 1 to 186

High gods on the maximum diffusion path of Murdock and White (1969)

A weakness of all but two of these studies (6..,17..) is that nonindependence among societies surveyed is not taken into account, resulting in inflated results for significance tests. A key study with controls for spatial diffusion within the same SCCS database (6..) found that the presence of moralizing high gods is not predicted by political hierarchy but predicted instead by low levels of external warfare, food scarcity, moderate community sizes, and dependence on animal husbandry. The latter (6..) is said to result because animal husbandry invites easy theft, for which belief in a moral god might provide protection.

Table 2 shows an evaluation if conflicting hypotheses and guided by the conception of “multiple working hypotheses” (18. 19) to reconcile contrasting results and reflect on the value of considering multiple working hypotheses and multiple methods. Chamberlain (18,19) warned against the premature adoption of favored hypotheses, and this recommendation is particularly relevant when seeking to understand and interpret discrepant findings. The key is not necessarily to select one or another method or model over others but, in the case of explaining the origins and spread of religious beliefs, to recognize that while controlling for autocorrelation may give us models for early origins of moralizing “high gods” prior to the extensive diffusion of world religions, standard regression methods such as OLS (ordinary least-squares) may provide information about factors that affect the wide-reaching diffusion of religious beliefs and institutions. Therefore, the contrasting results of a correlational or OLS model (5) and a 2SLS (two-stage least squares) model controlling for autocorrelation (6..) can be viewed as providing complementary information about the initial invention of moral gods and the eventual diffusion of world religions. We will show that in two improved models, one in which spatial and language autocorrelation is controlled for and one in which it is not, three types of societies are identified that form beliefs in moralizing high gods: those in dry zones; those with permanent settlements, rights over land, and wage labor; and those pastoral societies in which wealth is characterized by slow-reproducing animals used for trade and travel such as camels and horses. The latter two types of societies are subject to episodic cycles of resource abundance followed by periods of excessive inequality (20, 21). Consequently, these results suggest that the invention or adoption of belief in moralizing “high gods” may be the result not only of resource-poor environments (8..) or one in which property was prone to theft (6..) but of a culture’s need to cope with periods of heighted inequality that are amplified in periods of scarce resources relative to overabundance of population. Excessive inequality is generated more readily, for example, when property values relative to wages are amplified by scarcity and overpopulation or when pastoralist wealth in animal stock is unequally accumulated through exchanges that also increase the strength of some kinship groups at the expense of others.